Nizami Akturk 1993 PHD Thesis

Imperial College of Science, Technology and Medicine
University of London
DEPARTMENT OF MECHANICAL ENGINEERING

TRIBOLOGY
SECTIO
DYNAMICS OF A RIGID SHAFT

SUPPORTED BY
ANGULAR CONTACT BALL BEARINGS
by
Nizami AKTORK
B.Sc. (Eng.), M.Sc., D.I.C.
A Thesis submitted for the Degree of

DOCTOR OF PHILOSOPHY
of the
UNIVERSITY OF LONDON
April 1993
London
Abstract
ABSTRACT
In this thesis, the dynamics of a rigid shaft supported by angular contact

ball bearings is studied. Particular attention is paid to ball bearings
which have wide industrial applications. Despite their frequent usage,
little is known about the vibration characteristics of ball bearings.
Indeed it is generally not widely appreciated that even a defect free ball
bearing will cause vibrations. With the added effects of defects, the
vibrations and noise produced can be difficult to analyse.
The method used to investigate the vibration characteristics of angular

contact ball bearings was to consider the shaft bearing assembly as a
spring-mass system, where the shaft acts as a mass and the raceway and
balls act as massless nonlinear contact springs. The system therefore
undergoes nonlinear vibrations under dynamic conditions. In this thesis
the vibrations of such a system were assumed to occur in up to five
degrees of freedom. A computer program was written to model the radial
and axial vibrations. Results are obtained in both the time and
frequency domains.
It is shown that the vibration characteristics of the shaft-bearing

assembly change when the bearings operate in different regions of their
nonlinear load deflection characteristics or when the bearing design or
operating conditions are altered. The untoward effects of the natural
frequency and the ball passage frequency, which is one of the main
problems of shaft-bearing systems, are demonstrated for a perfect
bearing where when there is coincidence of natural frequency and ball
passage frequency, or their sub or superharmonics, resonances occur.
2
Abstract
In addition, the system behaviour was studied when defects such as; out
of roundness, waviness, off-sized balls, misalignment and localised
defects on the running surfaces, were present. The vibrations associated
with these defects are demonstrated with resonances occurring at the
natural frequency and certain of the shaft speeds, cage and ball speeds,
ball passage frequency, their sub and super harmonics and their
combinations.
Finally, the bearing outer races were supported on elastomer buttons,

cartridges or 0-rings. Since the dynamic properties of elastomers are
shape dependent, the available data was not usable in this thesis.
Therefore geometry dependent theoretical models were developed in
order to estimate the stiffness and damping coefficients of elastomers
from their material properties. One formula for cylindrical buttons, two
formulae for ring cartridges and three formulae for 0-rings were
developed and shown to be consistent with available experimental data.
In the modelling of elastomers, a mechanical Voigt model was employed
and the obtained spring dash-pot system was introduced to the shaft-ball
bearing assembly. It was demonstrated that a marked reduction in
vibration amplitude could be achieved by applying elastomeric dampers
to shaft-ball bearing systems.
3
3 dediade this work to the
people of Wiirkiye whose
money has financed my
studies abroad, in order that
our country be benefited by
the perfection of
technological culvancement
to vohich 3 hope my
research has contributed.
(Niany students for over a

century have stiulied abroad
but made limited
contribution to the industry
of our nation. 3 earnestly
hope that the students of my
generation will be better
equipped in knowledge to
enable them to enrich
their nation.
Acknowledgements
ACKNOWLEDGEMENTS
First and foremost I am thankful to The Creator of the

Heavens and the Earth, Who vibrates everything in order for
them to survive, for giving me the health, courage and ability
to complete this research.
I would like to express my deepest gratitude to my supervisor

Dr. Ramsey Gohar for his continual support, guidance and
help during the course of this research.
Thanks are also due to my colleagues and friends Dr. M.

Uneeb, Dr. T. Mahmood, Z. Nobari, Dr. C. Romo, Dr.
Abdun-Noor Yeo, Dr. A. Selcuk, F. Cali§kan, H. Yazgan, S.
Uzuno§lu, Dr. H. Celik, R. Sezer, M. Bayat, C. Kocyi§it,
M.E. Tuncaz, Z. Erman and members of Tribology Laboratory
who were always there in time of need.
I wish to thank Dr. R. Dwyer-Joyce for proof reading some

chapters of this thesis.
The help provided by the Mechanical Engineering Library staff

and 1.C. Computer Centre staff is greatly acknowledged.
1 am profoundly grateful to my family members and relatives, in

particular to my parents, Fazli and Behice Aktiirk for their
constant support and encouragement.
Finally, I would like to express my sincere thanks to the

Ministry of Education of Torkiye for the financial support
which has enabled me to do this research.
S
Contents
CONTENTS
Abstract 2
Dedication 4
Acknowledgements 5
Contents 6
Notation 11
CHAPTER 1General Introduction

1.1 Introduction 18
1.2 Angular Contact Ball Bearings as a Source of Vibrations 18
1.3 Damping of Vibrations 21
1.4 Objectives of the Thesis 22
1.5 Layout of the Thesis 22
1.6 Definition of Some Factors Influencing Ball Bearing Vibrations 23
CHAPTER 2 Previous Work

2.1 Introduction 30
2.2 Early Work on Ball Bearing Vibration 32
2.3 A Brief Look at Research Post 1960 34
2.4 Experimental Work 35
2.5 Elastic Contact Behaviour of Ball Bearings 41
2.6 Important Frequencies Associated with Ball Bearings 44
2.6.1 Natural Frequency 44
2.6.2 The Ball Passage Frequency 46
2.6.3 Flexural Vibrations of the Outer Ring due to Ball Loads 54
2.6.4 Natural Frequencies of Outer Ring 55
2.6.5 Natural Frequencies of Ball Motion 55
2.7 Vibration Monitoring and Frequencies due to
the Imperfections of Ball Bearings 57
2.7.1 Imperfections of Ball Bearings 59
2.7.1.1 Waviness 60
2.7.1.2 Oft-sized Balls 64
6
Contents
2.7.1.3 Misalignment 66
2.7.2 Defects 70
2.8 Shaft-Ball Bearings with External Dampers 74
2.9 Elastomers as External Dampers 77
2.10 Background of the Project 81
CHAPTER 3 Mechanics of Ball Bearings

3.1 Introduction 84
3.2 Elastic Contact Modelling 85
3.2.1 Contact Mechanics 85
3.2.2 Calculation of the Stiffness Factor 89
3.3 Bearing Geometry 90
3.3.1 Calculation of Preloaded Contact Angle 90
3.3.2 Calculation of Deflection 92
3.3.3 Calculation of Total Force on a Bearing 100
3.4 Dynamic Modelling of a Vibrating Shaft Supported by Ball Bearings 102
3.4.1 Assumptions 103
3.4.2 Equations of Motion 104
3.4.3 Dimensionless Groups 108
3.4.4 Computational Solution of the Equations of Motion 111
3.4.4.1 Initial Conditions 111
3.4.4.2 Computer Flow Chart and the Solution Procedure 112
CHAPTER 4 Vibration Frequencies of a Shaft-Bearing System

4.1 Introduction 114
4.2 The Inherent System Frequencies 115
4.2.1 System Natural Frequency 115
4.2.2 Cage Speed 117
4.2.3 Ball Rotational Speed 120
4.2.4 Relative Speeds 121
4.2.5 Ball Passage Frequency 121
4.3 Manufacturing Malfunctions 123
4.3.1 Waviness 124
4.3.1.1 Inner Ring Waviness 126
4.3.1.2 Outer Ring Waviness 127
4.3.1.3 Ball Waviness 128
4.3.2 Off-sized Balls 129
4.3.3 Misalignment 130
7
Contents
4.3.3.1 Deflection due to Misalignment 131

4.3.3.2 Contact Angle Variation due to Misalignment 132
4.3.4 Out of Balance of Mass Centre 134
4.3.5 Defects on the Rolling Surfaces 135
4.3.5.1 Defect on the Outer Race Surface 136
4.3.5.2 Defect on the Inner Race Surface 137
4.3.5.3 Defect on the Ball Surface 137
CHAPTER 5 Elastomers as External Dampers

5.2 Properties of Elastomeric Materials 140
5.2.1 Large Deformations 140
5.2.2 Dynamic Properties of Elastomers 141
5.2.3 Geometry Effect 141
5.2.4 Frequency Effect 142
5.3 Formulation of Elastomers 144
5.3.1 Rectangular Cross-Section Buttons 146
5.3.2 Cylindrical Buttons 147
5.3.3 Ring Cartridges 149
5.3.4 0-Rings 150
5.4 Experimental Results Used in This Thesis 153
5.4.1 Material Selection 156
5.4.2 Cylindrical Button Cartridge Test Samples 156
5.4.3 Ring Cartridge Test Samples 157
5.4.4 0-ring Test Sample 158
5.5 Representation of the Experimental Data 159
5.5.1 Generalised Viscoelastic Models 160
5.5.2 Generalised Force-Displacement Relationships 161
5.5.3 Direct Fit of Data to Mathematical Expressions 161
5.6 Brief Description of the Subroutine 163
5.7 Modelling of The System 166
5.8 Excitation Frequency 168
CHAPTER 6 Results and Discussion: Ball Bearings

6.2 Shaft-Ball Bearing Specifications 171
6.3 System Natural Frequency 174
6.4 Ball Passage Frequency 179
8
Contents
6.4.1 Effect of Varying the Number of Balls 186

6.6.2 Effect of Varying the Preload 189
6.5 Waviness 192
6.5.1 Outer ring Waviness 192
6.5.2 Inner ring Waviness 201
6.5.3 Ball Waviness 205
6.6 Off-sized Balls 208
6.7 Misalignment 217
6.7.1 Both Bearings are Misaligned 217
6.7.1.1 Outer ring Misalignment 217
6.7.1.2 Inner ring Misalignment 218
6.7.1 LHS Bearings is Misaligned 218
6.8 Defects on the Rolling Surfaces 219
6.8.1 Outer race Defect 220
6.8.2 Inner race Defect 221
6.8.3 Defect on the Ball Surface 223
6.9 Vibrations in 5 Degrees of Freedom 224
CHAPTER 7 Results and Discussion: Elastomers as External Dampers

7.2 Dynamic Properties of Elastomers 229
7.2.2 Ring Cartridges 233
7.2.3 0-rings 234
7.3 Simulation of Elastomer External Dampers 236
with Mechanical Voigt Model
7.3.2 0-rings 239
7.4 Elastomer Dampers in Assembly 241
CHAPTER 8 Overall Conclusions and Suggestion for Future Work

8.2 Conclusions for Ball Bearings 250
8.2.1 System Natural Frequency 250
8.2.2 The Ball Passage Frequency 251
8.2.3 Outer ring Waviness 251
8.2.4 Inner ring Waviness 252
8.2.5 Ball Waviness 252
9
Contents
8.2.6 Off-sized Balls 253

8.2.7 Misalignment 253
8.2.8 Defective Running Surfaces 254
8.3 Conclusions for Elastomers as External Dampers 255
8.4 Recommendation for Further Research 257
REFERENCES 259
APPENDIX 1 Equivalent Voigt Model 272
APPENDIX 2 Phase-Plane Representation 273
APPENDIX 3 Large Deformations of Elastomers 277
APPENDIX 4 Other Methods of Obtaining the Stiffness Coefficients of 279

Ring Cartridges
APPENDIX 5 Previously Derived Prediction Methods for Elastomers 284
APPENDIX 6 Voigt Model 289
APPENDIX 7 Ball Passage Frequency 291
APPENDIX 8 Fast Fourier Transform 294
APPENDIX 9 Runge-Kutta Method 296
APPENDIX 10 The Effect of Preload and Number of Balls on 297

The Natural Frequency
10
Notation
NOTATION
symbol Description Units
a the distance between the misalignment centre and bearing centre m

Or ball waviness amplitude m
or cross-sectional area of button (Chapter 5) m2
or width of the rectangular cross-sectional elastomer button (Chapter 5) m
Or distance between the external forces and the LHS bearing (see Fig.(3.16))m
al position of the LHS bearing from the C.G. (see Fig.(3.16)) m
A amplitude of the lobes m
or cross-sectional area of button (Chapter 5) m2
Or a constant (Chapter 5)
or distance between raceway groove centres m
b breadth of the rectangular cross-sectional elastomer button (Chapter 5) m
bi . position of the RHS bearing from the C.G. (see Fig.(3.16)) m
B total curvature, A / db
or a constant (Chapter 5)
c viscous damping factor N s/m
ce equivalent viscous damping factor N slm
d diameter m
or diameter of elastomer buttons (Chapter 5) m
db ball diameter m
di inner raceway diameter m
do, pitch diameter m
do outer raceway diameter m
D diameter of the buttons m
or shaft diameter at groove base m
E Young's modulus NI/n/2
Or error function (Chapter 5)
Eeff effective Young's modulus (Chapter 5) N/m2
11
Notation
E. static Young's modulus (Chapter 5) N/m2

E* complex Young's modulus of elasticity N/m2
E' real part of the complex Young's modulus of elasticity N/m2
E" imaginary part of the complex Young's modulus of elasticity N/rn2
f frequency Hz
fbp ball passage frequency Hz
fc cage rotation frequency Hz
fn natural frequency Hz
fw wave passage frequency Hz
nf n th super harmonic of one frequency Hz
F( p) curvature difference 1/m
F external forces N
Or external tensile force (Chapter 5) N
g acceleration due to gravity m/s2
G shear modulus of elasticity N/m2
G* complex shear modulus of elasticity N/m2
G real part of the complex shear modulus of elasticity Nim2
G"
G imaginary part of the complex shear modulus of elasticity N/rn2
h the height of the buttons (Chapter 5) m
or time increment s
H compliance rn/N
H* complex compliance rn/N
H 1 real part of the complex compliance m/N
H2 imaginary part of the complex compliance m/N
I second moment of area m4
k material constant, proportional to hardness

K contact stiffness factor N/m3/2
Or Runge-Kutta interval constant
K* complex stiffness N/m

K 1 real part of the complex stiffness N/m
K2 imaginary part of the complex stiffness N/m
K e equivalent stiffness N/m
K, inner raceway/ball contact stiffness factor N/m312
K0 outer raceway/ball contact stiffness factor N/m3r2
/ length of the specimen (Chapter 5) m
12
Notation
L a characteristic length m
or mean circumferential length of a cartridge ring (Chapter 5) m
m number of balls
or number of Voigt elements (Chapter 5)
mor mass of outer ring kg
M mass of the shaft kg
or moment Nm
n mode order
or shaft speed cp.m.
n i inner race (i.e., shaft) speed Ep.m.
no outer race speed Ep.m.
N number of data points

or number of wave lobes (Chapter 5)
or number of buttons (Chapter 5)
P empirical constant for a particular geometry
q instantaneous centre of movement m
3 radius m
r •
gi
inner groove radius m
r outer groove radius m
go
r• inner raceway radius m
I
r pitch radius m
m
r0 outer raceway radius m
Po the amplitude of forcing function N
PR preload N
R ball centre focus radius under zero load m
R, inner radius of the ring cartridge or 0-ring (Chapter 5) m
R 0 outer radius of the ring cartridge or 0-ring (Chapter 5) m
R mean radius of ring cartridge or 0-ring (Chapter 5 —see Fig.(5.8)) m
S entropy
S #. shape factor equivalent to loaded area divided by force free area
I time s
or thickness of the ring cartridge (Chapter 5) m
A t time step s
T temperature K
3 velocity m/s
13
Notation
or volume (Chapter 5) m3
Vi inner raceway velocity nVs
V° outer raceway velocity m/s
V c cage velocity m/s
w work done by the representative chain (Chapter 5) Nm
w 1 weighting constant
wz weighting constant
W total contact force between a ball and its race N
or the total work done (Chapter 5) Nm
W off total contact force due to off sized ball N
x, y, z local coordinates attached to the shaft (see Fig.(3.17))
or local coordinates attached to the reference ball (see Fig.(3.15))
X, Y, Z fixed coordinates (see Fig.(3.17))
X 1 , Yi , Zi local fixed coordinates at the C.G. (see Fig.(3.17))
X '1, Y11,
Z'i intermediate coordinates following Euler's angles (see Fig.(3.17))
a contact angle rad.
ai. Instantaneous contact angle rad.
a0 unloaded contact angle rad.
a preloaded contact angle rad.
P
P shape factor
(5 contact deflection m
or phase angle as in Equ.(5.5) rad.
Oc deflection due to misalignment m
3i total instantaneous contact deflection for the i th baU m

AS diameter difference of the off sized ball m
& amplitude of the vibration (Chapter 5) m
6* dimensionless contact deflection
lc elliptical eccentricity parameter
A wave length m
or extension ratio (Chapter 5) rad.
V Poisson's ratio
0 the angle between the fixed and moving
reference axes (see Fig.(3.15)) rad.
Or phase difference
14
Notation
e convergence criterion m
or strain
• rock angle of the shaft about y fads rad.
3 angle between neighbouring two balls (see Fig.(3.15)) rad.
or number of chains (Chapter 5)
g integration angle rad.
C damping ratio
a stress N1/12
0 angular position of balls rad.

0' arbitrary angular position of balls rad.
P curvature 1/m
or misalignment angle (Chapter 5) rad.
Zp curvature sum 1 /m
to angular velocity of the shaft radis
ball angular velocity in the same l pane asthe
te cage speedd rad./s
ball angular velocity about its centre rad./s
wb
a) c cage angular velocity rad./s
angular velocity about the instantaneous axis of zero velocity rad./s
WCL
a). Iinner race angular velocity radfs
to cr outer race angular velocity rad./s
a) s ball spinning velocity rad./s
12 The stiffness divided by the damping factor (Klc ) (Chapter 5) 1/s
17 angular speed vector of the shaft
V rock angle of the shaft about x axis rad.
X0 the amplitude of vibration in x direction m
Yo the amplitude of vibration in y direction m
n the amplitude of the waves at the contact angle m
FO initial wave amplitude m

3P maximum amplitude of the wave m
X complete elliptic integral of the first kind
3 complete elliptic integral of the second kind
radius of the loci groove centres of curvature for inner ring m
910 radius of the loci groove centres of curvature for outer ring m
/ a characteristic length m
15
Notation
subscripts
Symbol Description
0 initial conditions
1 relating to the real part of a complex variable
2 relating to the imaginary part of a complex variable
a relating to axial
b relating to ball
bp ball passage frequency
C relating to cage
e equivalent
i ith ball
or i th Voigt element (Chapter 5 and 7)
or iteration index
or relating to the initial conditions
j relating to the Voigt element being calculated

or iteration index
k relating to discrete time series
n relating to the natural frequency
3 relating to discrete time series
or radial
wp relating to the wave passage frequency
x, y, z coordinate axes
a) frequency dependence
Superscripts
5ymbol Describtion
relating to the first derivative with respect to time

relating to the second derivative with respect to time
Indicating a normalised variable
Or relating to a complex number
16
Abbreviations
BPF Ball Passage Frequency
C.G. Centre of Gravity
deg. degrees
Equ. equation
Equs. equations
Fig. figure
Figs. figures
FFT Fast Fourier Transform
LHS Left Hand Side
rad. radians
IMS root mean squares
rpm revolutions per minute
RHS Right Hand Side
wpc waves per circumference
WPF Wave Passage Frequency
17
CHAPTER 1
GENERAL INTRODUCTION
1.1 Introduction
A general introduction to the shaft-angular contact ball bearings assembly as a part of

vibratory system and elastomers as external dampers to such a system will be given in the
first part of this chapter. Later the scope and outline of the thesis will also be given and at
the end of this chapter some of the important terminology used in this thesis will be
defined.
1.2 Angular Contact Ball Bearings as a Source of Vibrations
In modern machine applications, ball bearings have become indispensable since,

whenever a shaft rotates, It must be supported in such a way that it will be maintained in
the correct position and be free to rotate in the machine's framework thus ensuring the
correct functioning of the machine.
In the transformation of rotating motion from one machine part to another, the main
objectives are to minimise radial run-out and frictional losses since friction consumes
energy as well as causing wear which brings about changes in dimensions until a machine
becomes ineffective. The machine part employed to achieve minimum frictional losses
and hence a free motion between moving and stationary parts of the machinery is called a
bearing. The simplest bearing is made up of a shaft, called a journal, rotating in a mating
hole which serves as the bearing. Even after making the running surfaces of the shaft
and bearings as smooth as possible and using lubricant between them in order to achieve
low friction torque, the losses due to friction are considerable. In engineering, rolling
bearings, as an alternative to journal bearings, are therefore employed in order to
minimise frictional losses, since their rolling resistance is much less than the journals'
sliding resistance.
18
Chapter 1 Introduction
Amongst rolling element bearings, ball bearings are frequently used for applications,
where load is not too high, on account of their exceptionally low friction and their ability to
take a preload.
Outer race
Cage
Inner race
Fig (1.1) Assembled Angular Contact Ball Bearing
A ball bearing is composed of an inner race, an outer race, a set of balls and a cage or
separator as shown in Fig.(1.1). Inner and outer races are basically two hardened steel
rings with annular grooves which serve as a guide to the precessing balls. The inner race
is generally force-fitted onto the shaft and the outer race is similarly fitted in a housing.
Inner or outer races can sometimes be dispensed with, and the balls mounted in such a
manner that they run on a hardened shaft itself or a supporting surface. This is needed in
some applications because of the space limitations. The cage or separator prevents balls
from contacting each other while they are precessing round the inner race and spaces
them uniformly. The balls are spherical to a high accuracy and are the most important parts
of a bearing since they transmit the forces from the moving parts of the machinery to the
stationary parts.
Angular contact ball bearings are a special type for they are able to accommodate both
axial and radial loads. The shafts under combined loading are generally supported by
angular contact ball bearings with a built-in preload. Depending on the arrangement
made or the type of angular contact bearing chosen, they can accommodate axial loads
acting in one or both directions. Another advantage of angular contact ball bearings is
their being producible with tight tolerances. In this case they are called precision angular
contact ball bearingswhich are also used in the simulation model in this thesis.
19
Ball bearings may appear simple and unsophisticated to the casual onlooker, yet their use
In applications involving high speeds, high temperature, heavy, or unusual loading,
requires a precise definition of their specification and performance. Kinematics, vibration
analysis, theory of elasticity, hydrodynamics and elastrohydrodynamic lubrication, wear,
friction and heat transfer are some of the topics which are employed in an analysis of
bearing performance. Some of these topics are widely employed in the investigations of
ball bearings while others remained relatively less used. In particular, kinematics and
vibrations of ball bearings can be classified under the latter heading. Hence in this thesis
vibrations associated with angular contact ball bearings will be investigated.
Angular contact ball bearings generally support a shaft rotating at a certain speed.
Naturally they are involved in the dynamics of this shaft and the load it carries. For quite a
long time researchers did not realise that ball bearings were one of the governing
parameters for the dynamic characteristics of the system. Balls were assumed to be rigid
or were taken as equivalent linear springs based on the complete bearing. One of the
main reasons for not realising the importance of ball bearings as a vibratory system was
that their measurements and simulations were not precise enough. As more precise
applications were carried out, due attention was given to ball bearings in shaft dynamics.
It is generally not widely known that a rolling element bearing may be perfect (i.e. defect
free) and yet still produce vibrations. With defects in addition, the vibrations and noise
produced can be quite difficult to analyse. As a first step in identifying the noise problem
caused by a ball bearing or in other words Its vibration characteristics, a mechanical
vibratory model of a shaft-bearing assembly should be considered as a spring-mass
system, where the shaft acts as a mass and the raceways and balls act as nonlinear
contact springs. Therefore, the system undergoes nonlinear vibrations under dynamic
conditions.
Previous investigations also showed that when a shaft supported by ball bearings
rotates, it undergoes vibrations. These vibrations can be quite severe if the wrong
operational speeds or bearings are chosen. Researchers try to eliminate these vibrations
for various reasons. Some may require a quiet environment, others very smooth running.
Whatever the reason, they endeavour to overcome these vibrations by careful design.
After each new design they would like to check whether their new design works. Without
a simulation model, they have to build an experimental rig and keep testing with frequent
modifications. This trial and error approach is very costly. The best thing is to have a
simulation model, like the one in this thesis, in order to check whether the new idea will
wort( at the design stage and then to verify the results with experiments afterwards. This
saves money and time.
1.3 Damping of Vibrations

Y
If the vibrations cannot be eliminated or lowered to an acceptable level by bearing desi gn
only, external dampers might have to be employed.
External dampers have become increasingly important in the control of shaft vibrations of
rotating equipment. Previous researchers tried to eliminate or reduce the untoward
effects of the vibrations by employing different dampers such as using elastrohydro-
dynamic lubrications or supporting the bearings on flexible mountings or by using oil
squeeze bearings round the outer race.
Another type of external damping is by employing viscoelastic damping which is

exhibited strongly in polymeric and glassy materials. In this thesis, elastomers are
employed. Elastomer is a material that consists of long-chain organic molecules linked at
various points along their length by strong chemical be nds giving it elastic properties
akin to those of rubber.
Although elastomers have been used in different applications, mainly in vibration and
shock isolation, for a long time, their dynamic properties were unknown for many years.
Today, elastomers are used in designs involving increasingly complex applications. Thus
accurate knowledge and a clear definition of elastomer dynamic properties (stiffness and
damping) is becoming important.
However, there are three problems in employing elastomers. The first one is limited data
since the dynamic properties are shape dependent. The second is the highly nonlinear
damping exhibited in them and the fact that their dynamic characteristics are obtained
only for a periodic excitation, whereas in the present study the initial transient vibrations
contribute to the vibration behaviour of the system and are not necessarily periodic. The
third is their relatively low rigidity that reduces one of the prime advantages of the ball
bearing which is its relatively high stiffness compared with, say, gas bearings.
In order to investigate these difficulties, a pilot study was carried out during the course of
this work and some interesting results were obtained. It is shown that the dynamic
properties of elastomers can be derived from their material properties and, using this data,
elastomer dampers can be successfully employed for attenuating untoward vibrations by
careful design.
In summary it may be said that ball bearings are one of the essential parts of rotating
machinery and their characteristics as well as imperfections in them, can cause severe
vibration problems. Hence It is important to understand vibration response and to identify
21
undesired effects in order to eliminate them through design considerations or to avoid

them by careful operational planning (i.e. shaft speed, preload, external forces etc.).
Therefore, the object of this investigation is to study the vibration characteristics of
angular contact ball bearings and to see whether elastomer dampers are able to reduce
vibration levels.
J.4 Objectives of the Thesis
The main objectives of this thesis are to investigate:
1.4.1 The variation of the natural frequency due to different positions of the ball set in a
bearing.
1.4.2 The effect of the ball passage frequency (it will be defined later in this chapter)
and the effects of varying the number of balls and the preload on it.
1.4.3 The effect of system anomalies such as waviness, off-sized ball, misalignment and
out of balance of shaft centre.
1.4.4 The presence of localised point defects on the running surfaces.
1.4.5 Obtaining the dynamic properties (stiffness and damping coefficients) of

elastomers from their material properties (Young's modulus or shear modulus).
1.4.6 Whether elastomers can successfully be employed as external dampers in shaft-

ball bearing assemblies.
1.5 Layout of the Thesis
This thesis is in eight chapters.
In Chapter 1, there is a general introduction, the outline of the project is defined and
the terms used in the thesis are clarified.
Chapter 2 gives the survey of previous work done. Since the subject is broad, this
review is done under different subheadings in order to show how the various topics
interact with each other.
Chapter 3 deals with angular contact ball bearings that are used to support a rotating
shaft. It gives a general definition of the dynamic problems associated with angular
contact ball bearings. The theories used are discussed and the equations developed.
22
The resulting equations of motion are developed and their solution method is
Investigated.
Solutions of the equations of motion result In peak amplitudes at certain frequencies.

These frequencies are investigated in Chapter 4. Some of the manufacturing
anomalies and defects that can lead to severe vibrations are also considered and their
effects are investigated. The effects of some external forces as well as the change in
some of the system parameters on the dynamic behaviour of the system is also
investigated in the fourth chapter.
In Chapter 5 elastomers as external dampers to angular contact ball bearing-shaft

systems are studied. Because of limited data available, an investigation is carried out to
obtain the dynamic behaviour of elastomers from their material properties and geometric
shapes. Different types of elastomer dampers, namely buttons, cartridges and 0-rings
are introduced to the system as external dampers in order to eliminate or reduce
untoward effects of natural frequencies of the system. In this a mechanical Voigt model is
employed.
In Chapter 6 the simulation models described in the third and fourth chapters are used
to obtain results which are then compared. First, a two degree of freedom system is
investigated since its simplicity allows some very important system parameters to reveal
themselves. The degrees of freedom are then increased, first to three and then to five in
order to obtain close simulation of real systems. The results obtained are compared with
other researchers' experimental and theoretical data and their differences are discussed.
In Chapter 7 the equations developed for elastomers are investigated and compared
with experimental data presented by other researchers. Later elastomers are introduced
to the simulation model andttfect of employing elastomer as external dampers to the
shaft-ball bearing system is studied.
Chapter 8 is the conclusion of the thesis. In this chapter the results obtained during
the course of this research are summarised. The chapter condudes with suggestions for
further work.
1,6 Definitions of Some Factors Influencing Ball Bearing

Vibrations
Today, as new branches of science come into existence, each branch specialises in a
particular topic, hence many new terms appear known only to specialists. Furthermore,
23
the same word is sometimes used with different meanings in other fields or by other
researchers in the same field. In some cases different words are employed to describe
the very same thing or occurrence. Therefore it has become necessary to clarify the
meanings of the words used in reports and theses. Thus some of the important key
words used in this thesis are defined here and have been kept unchanged throughout
the thesis unless otherwise stated.
1.6.1 Natural Freauency of the Shaft and Bearings Assembly
If a shaft on ball bearings is displaced from its position of equilibrium and released,
vibrations take place. These vibrations, which are maintained by the elastic force in the
bearings alone, are called the free or natural vibrations of the shaft and its bearings. For
this type of vibration there are certain frequencies where the amplitude of the shaft
displacement becomes large (see Fig.(1.2)). If the shaft is assumed to be elastic (i.e., not
rigid) there are a series of natural frequencies referred to as modal frequencies of the
shaft and bearings. If any forcing functions coincide with one of the natural modes, the
effect of resonance can be observed clearly.
0.03
First Natural Frequency
Second Natural Frequency
0.01 .
0.00 .

10 100 1000
Frequency
Fig.(1.2) Natural frequencies of a system for two degrees of freedom
lithe shaft is assumed to be rigid, there is only one natural frequency for each degree of
freedom it possesses.
24
1.6.2 Ball Passage Frequency
This is one of the important frequencies associated with ball bearings. Although it
appears as ball pass frequency or variable compliance frequency in different references,
it is referred to as the ball passage frequency (B P F for short) throughout this study The
vibrations associated with the BPF is called variable compliance vibrations or ball passage
vibrations. While the shaft is rotating, applied loads are supported by a few balls restricted
to a narrow load region defined by the radial position of the inner ring with respect to the
outer ring and depending upon the elastic deflections at the ball to raceways contacts
and the preload (see Fig.(1.3)). As the position of the balls change with respect to the
applied load vector the load distribution on the balls changes, producing a relative radial
movement between the inner and outer rings which repeats itself at a certain frequency
that is called the ball passage frequency. In mathematical terms the BPF can be
described as the cage speed times the number of balls. As a result of this, the system
can go into resonance at a fraction of a natural frequency.
This untoward behaviour is inherent in the design of rolling element bearings and cannot
be avoided even for a geometrically perfect bearing. More detailed information can be
found in Chapter 4.
Fig.(1.3) Deformation of balls when they rotate round the inner ring
1.6.3 Roundness. Roughness and Waviness
A machine part is said to be round in a specific cross section if all points on the periphery
are equidistant from the centre. lithe cross section is not round, it Is said to be out of
' 25
round. Out of roundness usually exists in the form of an irregular profile. The
imperfections found on any surface generally take the form of peaks and valleys of
varying height and width. In the simple case of out of roundness there are a certain
number of lobes per circumference. It has been a common practise to describe different
wavelength bands in different terms, referring to wavelengths of order of micrometers as
roughness and the longer wavelengths as waviness (see Fig.(1.4)). In this thesis, if the
wavelength of these lobes is of the order of the Hertzian contact width or less, in the
description of the surface feature, the term roughness is employed, whereas for surfaces
with longer wavelengths the term waviness is applicable. For longer wavelengths, rolling
motion is continuous with the ball rolling surface over contours of the race. More
information can be found in Chapter 4.
----"----
Fig.(1.4) Roundness, Waviness and Roughness
1.6.4 Contact Angle
For angular contact ball bearings, the most important parameter is the contact angle which
is defined as the angle between a plane perpendicular to the bearing axis and an axis
passing through the points of contact of the balls with the raceways (see Fig.(1.5)). The
contact angle influences the axial and radial specifications for a bearing. Low contact
angles are generally used for light axial loads and/or high speed applications, whereas
higher contact angles are chosen for low speed applications and when high axial loads
and/or axial rigidity are the main requirements.
26
Contact angle
II.
I" Ar
Points of
contact
The plane
perpendicular
to bearing axis
Bearing axis
Fig.(1.5) Contact angle
1.6.5 Preload
Preload is the application of a permanent axial load on a bearing. It is achieved by

applying internal loads to a pair of bearings, for instance by means of springs, or by
preloading a pair of bearings against each other through a clamping nut. Preload should
not be higher than is necessary for the application but should be sufficient to avoid the
preload being completely relieved from any bearing by the action of external loads. The
preload serves to guarantee a certain minimum load to seat all the balls and ensure firm
rolling contact, because otherwise balls can skid and roll and produce a cage instability.
High vibration levels can be produced by bearings which are too lightly loaded.
1.6.6 Bearing Internal Clearance
Bearing internal clearance is defined as the total distance through which one bearing ring
can be moved relative to the other in the radial direction (radial internal clearance -
Fig.(1.6.a)) or in the axial direction (axial internal clearance -Fig.(1.6.b)).
It is necessary to distinguish between the internal clearance of a bearing before mounting

and the internal clearance in a mounted bearing which has reached Its operating
temperature (operational clearance). The initial clearance (before mounting) is greater
than the operational clearance.
27
The radial clearance of a bearing is of considerable importance if satisfactory operation is

to be obtained.
Racial internal Axial internal

clearance clearance
,
b
a
Fig.(1.6) Internal clearance
Depending on the application, it is necessary for there to be a positive or a negative

operational clearance in a bearing arrangement. In the majority of applications, the
operational clearance should be positive, i.e. when in operation, the bearing should have
a residual clearance, however slight. There are some cases where normally a negative
clearance, i.e. a preload is desirable in order to enhance the stiffness of the bearing
arrangement or to increase running accuracy.
1.6.7 Misalignment
Ball bearing races are said to be misaligned when one bearing ring is not colinear or is
angularly displaced relative to the other, about an axis at right angles to the bearing
running axis (see Fig.(1.7)). This may be found descnbed as tilted, canted, cocked, out
of square or malaligned In other researchers work but the term misalignmentis employed
In this thesis. Misalignment in ball bearings is a commonly occurring problem which has a
predominant effect on basic running errors and can lead to lubrication problems, cage
failure, premature raceway damage and/or vibration at the running speed.
28
I
• •
..
a b
Fig.(1. Misalignment of ball bearings
Closure
In this chapter a general introduction to the subject was given, the thesis was outlined
and some of the keywords used were defined. Taking this general introduction as a
basis, in the next chapter, the previous research done on the subject will be investigated.
29
CHAPTER 2
PREVIOUS WORK
.1 Introduction
The earliest form of ball bearing was found amongst the remains of ships dating from the
time of Caligula (AD 42-54) [Dowson, 1979]. However, the idea of ball bearings for
supporting a weight started to appear in written records only as early as 1490. The
following text is an extract from two large unpublished manuscripts by Leonardo da Vinci
(AD 1452-1520) discovered in the National Library in Madrid in 1967 [The Unknown
Leonardo, 1974]:
"3 affirm that if a weight with a flat

surface moves on a similar surface, the
movement will be facilitate6 by interpos-
ing between them bans or rotten!. ... an6
3 6o not ate any 6ifference between balls
or rotten?. Gave the fact that balls have
universal motion whereas rollers can move
only in one 6irection. Zut if the balls or
rolleis touch each other in their motion,
then movement will be more 6ifficult than
if there were no contact between them,
because when they touch, the friction
CatI6et, a contrariwise motion an6 for this
reason the movements contrabict each
other. lowever, if the balls or rolleis are
pitche6 apart, they will touch only at one
point between the !oat) an6 its resistance
an6 consequently it will be easy to pro-
&ice this movement."
30
Chapter 2 Previous Work
From the text It is obvious that Leonardo understood the importance of the cage or
separator and advised its use for reducing friction (see Fig.(2.1)). The first patent to cover
a ball bearing, taken out by Philip Vaughan with the British Patent No 2006, goes back to
1794. Although these very early records may suggest that roller bearings were common
applications, they were not commonly adopted in machines until about 1900 [Roman and
Gohar, 1975].
4
(Tit:77n.
rz—ce
Fig.(2.1) Robbin type separator envisaged by Leonardo da Vinci

[Roman and Gohar, 1975]
From Leonardo to the 18th Century, many types and forms of ball bearings are described
in contemporary technical literature and presumably some of these designs have been
manufactured with varying degrees of success. However, the application of ball bearings
was largely empirical and specific designs of ball bearings did not emerge until the
eighteenth century By the 1880s because of the common usage and production of
bicycles, ball bearings were subject to intense development. Scientific work on the
development of rolling bearings also shows an increase in the latter part of the nineteenth
century Mass production of rolling bearings began In the early part of the twentieth
century [Tindale, 1988]. A complete presentation on the early history of ball and rolling
bearings can be found in a book by Dowson [1979].
After the Second World War the production of ball bearings as well as other type of
bearings gained importance and scientific progress occurred at an exponential rate. With
the advent in metallurgical processes and the increasing accuracy of machine tools, ball
bearings have evolved to the form we know today.
After 1950s the vibration of shafts supported by ball bearings also came to the attention
of researchers. An considerable increase in the theoretical research on the vibration of
ball bearings was made possible by the advancement in computer technology and
numerous research papers on the topic are available now. Although many books are
31
available on Rolling Bearings, only one book, entitled "Vibration of Bearings" [Ragulskis
et al., 1989], investigates solely the vibration problems associated with ball bearings. But
the importance of the topic is now realised and the vibration problems of rolling bearings
are addressed in recent editions of standard texts [Harris, 1991].
As mentioned earlier much research has already been done on the vibration of ball
bearings, such as; vibration monitoring, defects in bearings, natural frequencies of
bearings, external damping of vibration due to ball bearings, etc. The same researchers
have also worked on different aspects of ball bearing vibrations. Thus it will be more
convenient to review the previous research works under subtitles which will now be
reviewed individually.
2.2 Early Work on Ball Bearing Vibration

The first scientific papers on the subject were mainly concerned with producing perfect
bearings and therefore most of them described experiments with more attention paid to
stress levels, mechanical performance, wear etc. than to vibration problems, despite the
fact that many research study were describing a rotating shaft supported by ball bearings
(see, for example, Stribeck [1900], Hess [1908], SKF 11957]). A good example of the
first extensive scientific papers on ball bearings is a report from the Central Laboratory for
Scientific Technical Investigation written by Stribeck in 1900 under the title of "Ball
Bearings for Various Loads". In this report there is reference to earlier wort from 1897
and 1898. A summary of the development of roller bearing technology from 1907 to
1957 is given in a complimentary and extended article by SKF [1957] for the 50th year of
establishment of SKF. However, in this latter article, there is no reference to the problems
caused by the vibrations due to rolling bearings.
After solving major problems of production, researchers turned their attention to other
factors. As a result of this, the noise problem caused by ball bearings became important.
In any environment noise, even at a low level, from for example office apparatus, vacuum
cleaners, fans, shafts rotating in cars or ship etc., is irritating. Hence the vibration
characteristics of ball bearings attracted interest. Since the calculations involving vibrating
ball bearings are very complicated, experiments were carried out, but no analytical
approach was attempted for a long time. In the 1950s researchers tried to overcome the
noise and wear problems with lubrication. This was successful to some extent [Hall,
1957]. Although some research workers approached the theoretical investigation of ball
bearing dynamics pessimistically, even as recently as the 1970s [SKF, 1961; Wallin,
1966; Yhland and Johansson, 1970], some brilliant studies on the topic can be found as
early as the 1950s [Perret, 1950a; 1950b; Meldau, 1951; 1952]. Perret and Meldau
theoretically proved that even a perfect bearing may undergo vibrations due to ball
32
passage. But since they did not consider the shaft mass, their solutions were
incomplete.
Meanwhile some researchers started comparing antifriction bearings, particularly ball

bearings, with plain bearings for grinding spindles and concluded that although the plain
bearings are superior to ball bearings for lower speeds, the latter would give better
performance at higher speeds [Doi, 1958]. But this choice was not free of problems. A
shaft would vibrate less when mounted on good quality properly designed plain bearings
than on ball bearings [SKF, 1961]. Ball bearings certainly aggravated vibrations and make
the design of spindles more difficult because the dynamic characteristics of the system
were not well understood. Since there was not enough information on the dynamic
behaviour of grinding spindles supported by ball bearings, the manufacturers tried to
overcome the vibration problem by designing ball bearings with sufficient rigidity,
assuming that in this case even a small damping factor would keep the amplitude of
vibration well within permissible limits for most grinding conditions [Weimar, 1958]. They
were not completely successful, since the problem was not as simple as they thought and
vibration could not be avoided.
The importance of rolling bearings was recognised, but no attempt was made to Include
them in dynamic modelling of such problems as machine tool chatter, which were
established by Tobias and Fishwid( [1956]. The results were later experimentally verified
by Tobias [1959] for drilling and face milling operations. These results were improved by
Tobias et al. [1962] and published as book titled of "Machine Tool Vibrations" [Tobias,
1965]. However, Tobias et al. concentrated their research on the dynamic stability of the
machines subjected to external forces due to friction between tool and work-piece,
cutting force as a function of cutting velocity, rake angle variation, chip thickness variation
without giving the full credit to the effect of bearing characteristics on the performance of
machining spindles. Meanwhile, in the early 1960s, the vibration characteristics of ball
bearings were analytically studied under various simplifying assumptions by different
researchers [lgarashi, 1959; Tamura and Taniguchi, 1960]. Even with simplifications it
was difficult to analyse theoretically the vibratory behaviour caused by ball bearings
because of the complexity of the problem.
One of the most influential and extensive studies on vibration characteristics of rolling
bearings was done in 1963 by Gustafsson et al. The purpose of this study was to
Investigate the vibration and noise producing characteristics of large roller bearings for
propulsion machinery of ships, as well as smaller rolling bearings in order to reduce
vibration and noise at all audio and subsonic frequencies. Gustafsson et al. investigated
vibration of geometrically perfect bearings as well as those with some imperfections.
33
They showed that a shaft-ball bearing system may undergo vibrations even when they are
geometrically perfect. They pointed out two causes for this; the first is the baN passage
frequency (BPF) effect which was first realised and theoretically proven by Perret and
Meldau in 1950, and the second is bending of the outer ring due to variable ball bads.
Geometrical imperfections Investigated in this report includes consideration of waviness
and off-sized balls.
In the late 1960s some analytical studies of ball bearings appeared, but since there was
no help from powerful computers which were then inaccessible, even relatively simple
rolling bearing arrangements were difficult to analyse mathematically [Yhland and
Johansson, 1970]. Some proposals made in these studies were either too simple to be
generally applicable or too complicated to apply to the practical estimation of the spring
and damping properties of ball bearings. As an excellent example of this era, the studies
of Tamura and Shimizu can be given. In their first paper [1966] they investigated a bearing
with two balls, In the second paper [Tamura and Shimizu, 1967] they increased the
number of balls to three or four and in the third paper [Tamura and Shimizu, 1968] they
extended the study to a large number of balls. Although the authors employed many
equations and calculations, the vibration information was relatively limited.
From the late 1960$ onwards, with the common use of powerful computers, much
research was carried out on the vibration characteristics of ball bearings.
2.3 A Brief LookatResearch post 1960
The research on ball bearing vibrations covers a large spectrum. Thus different
researchers concentrated on various aspects of vibrations attributed to ball bearings. In
particular, there are some groups that have produced major contributions. Some of the
major research after 1960 is as follows:
2.3.1 Gustafsson and TaIlian et al. [1963; 1965] carried out an investigation of the
vibration and noise producing characteristics of ball bearings. They concentrated on
some very important vibration characteristics of ball bearings such as: the BPF and
vibration of the outer ring due to ball loads. The effects of waviness and off-sized ballson
the vibration of ball bearings were also investigated.
2.3.2 Tamura and Shimizu [1967; 1968; 1969] studied the static stiffness of ball
bearings. This research is important because It contnbuted towards investigations of the
Instabilities of shaft-bearing systems.
2.3.3 Gupta et al. [1977; 1979a; 1979b; 1979c; 1979d] modelled the rolling bearing in
34
Chapter2 Previous Work
cylindrical coordinates In six degrees of freedom. Since they did not consider the mass of
the shaft, but the rolling elements' mass only they concentrated their investigation on the
vibrations due to the balls and ball-cage Interactions.
2.3.4 Yamamoto et al. [1974; 1975; 1977; 1981a; 1981b; 1984; 1985] worked on
instabilities caused by ball bearings supporting an unsymmetric shaft. They also studied
sub and super harmonic vibrations due to ball bearings.
2.3.5 Gad et al. [1983; 1984; 1985a; 1985b] employed a computer simulation model of
ball bearing systems, including the shaft mass, in order to investigate the dynamic
properties of ball bearings and pointed out instabilities around the natural frequencies.
2.3.6 Igarashi et al. [1982; 1983; 1985] investigated the vibration due to point defects
in ball bearings.
2.3.7 Wardle et al. [1983; 1988a; 1988b] investigated the effect of distributed defects
in a ball bearing. Particular attention was paid to waviness.
2.3.8 Gohar et al. investigated different aspects of ball bearing vibrations. Rahnejat and
Gohar [1985] studied the effect of elastrohydrodynamic damping behaviour of the oil film
between the rolling elements and races. Matsubara and Gohar [1988] considered a
flexible shaft supported by ball bearings. Aini and Gohar [1990] modelled a grinding
spindle supported by ball bearings. Franco and Gohar [1992] investigated multiple
defects on a bearing and AktOrk and Gohar [1992] the ball passage vibrations.
2.4 Experimental Work
Although research on ball bearing vibrations has a strong background, with many
research papers available, there is no commonly accepted and justifiable standard
experimental method for vibration tests. Each manufacturer or researcher has
established his own rig and utilised his own method to extract information from ball
bearing vibrations. Every researcher has his different reasons for the experiments and
the rigs were designed to fit them. Therefore an acceptable method and test for one
researcher may be useless for another. This increased the number of methods available.
However, there is very little justification for the methods employed [Hemmings and Smith,
1976]. The standard test methods give little information that can be used in the
manufacture of ball bearings or in their usage by the customer, but the results available
are a useful general indicator that manufacturing quality is being maintained [Hemmings
and Smith, 1976] or it can be used to check the re-usability of the dismounted ball
bearings [Wallin, 1966].
35
Ever since ball bearings were first manufactured, checks for vibrations in them have been
carried out by spinning the bearings by hand. Since in normal bearing schemes the inner
ring usually rotates, the checking procedure has been improved by mounting the bearing
on a mandrel. The outer ring is then held and the inner ring rotated by means of the
mandrel. Spinning bearings by hand is quite acceptable for a quick assessment of
serviceability and this method is, to some extent, still in use today. However, the absolute
vibration level of the bearing cannot be determined by this method [SKF, 1961]. Testing
the bearing alone will only give information about manufacturing quality or re-usability of
the bearing; when it is assembled the bearing will have all together different vibration
characteristics. Therefore tests should be performed on bearings that are functioning in
the assembly [Watford and Stone, 1980]. When the bearings are in assembly, for
experimental investigations, another difficulty arises; the interaction between
predominant factors such as production accuracy, misaligned races, eccentric races,
elliptic rings, unequal ball sizes, radial clearance and geometrical errors of the running
surfaces, which are all of the same order of microns [Watford and Stone, 1980], cause
complicated vibrations. Therefore, it is difficult experimentally to separate, measure and
grasp the predominant factors clearly, and control the correspondence between the
effect of each factor and the resulting vibration [Gad et al., 1984]. Bearing characteristics,
therefore, have to be deduced indirectly from the behaviour of the system as a whole.
Furthermore, the interfaces between the shaft and the inner race, and the housing and
the outer race, can have a significant effect and must, therefore, be controlled carefully,
and the bearing should be tested under its proper operating conditions of rotation and
preload [Watford and Stone, 1980; Gad et al., 1984].
Electrodynamic
pickup
Fig. (2.2) Schematic presentation of test rig [Harris, 1991)
When the bearing arrangement is running under normal operating conditions it vibrates
and produces noise. Noise is a good indicator of vibrations. The major noise generation
is due to vibration from the bearing ring travelling through the structure and resonating
panels or sections of cas'ing to produce noise. Hemmings and Smith [1976] argue that
36
noise produced by ball bearings Is not useful In predicting bearing vibrations. This is
because in the majority of installations the direct noise from a bearing is of low power and
is generated inside a sealed Item of machinery. Under these conditions negligible noise
energy escapes through the casing or through seals and this source of noise may be
ignored. lgarashi and Yabe [1983] argues that when sophisticated instruments are
employed and experiments are carried out delicately, clear information can be obtained
from the sound generated by ball bearings. However, the best way to get information
from the shaft bearing assembly is by direct measurement of shaft vibration although this
Is not easily made either, when the bearings are assembled. As vibration forces are
transmitted through bearings, measurements are best made on bearing housings as near
to the shaft as possible [Dowson, 19701
Fig.(2.3) Vibration Test Schematic [Harris, 1931)
In spite of all these difficulties some figs and experimental measurements are described in
the references available. However, they are generally designed for a bearing in isolation
37
and cannot represent the bearing in normal operating conditions. Some experimental
results from the bearing In the assembly are given In the Section 2.7, health monitoring of
bearings, of this thesis. The measurement method is the same in principle for all the tests
available and as follows: the bearing to be measured is mounted on the mandrel against a
shoulder. The shaft speed is thus imparted to the inner ring while the outer ring is held
stationary. The radial and axial movements are registered by velocity pick-ups. This kind
of device was used throughout the SKF organisation [SKF, 1961] and established a
vibration standard in the early 19605 (see Fig.(2.2)). Very similar set-ups were also
employed by Gustafsson at al. [1963], Cana and Hobbs [1972], Hemmings and Smith
[1976], Braun and Danter [1979], lgarashi and Hamada [1982] and Karakurt [1989].
The schematic representation of a commonly used vibration tester is shown in Fig.(2.3).

The inner ring raceway of the bearing mounts on a precision arbour fastened to the
spindle, which rotates at 1800 rpm. A specified thrust load is applied to the side face of
the non-rotating outer ring. The top of the velocity transducer is lightly spring-loaded on
the outer diameter of the ring. The ring contacts the side face of the outer ring. The tool
and load combinations are sufficiently compliant to allow radial motion of the outer ring to
occur as balls roll over wavy surfaces or defects in races. The voltage signal from the
transducer is input to the analysing instrument, which amplifies the signal, bandpass
filters it, and displays the rms velocity values in each band. The three frequency bands
are 50-300 Hz., 300-1800 Hz. and 1800-10000 Hz. Larger bearings are tested at slower
rotational speed (700 rpm) with corresponding lower filter bands: 20-120 Hz., 120-700
Hz. and 700-4000 Hz. [Harris, 1991]. This basic testing technique has been successfully
used by bearing manufacturers and customers for many years.
Watford and Stone [1980] developed a rig suitable for measuring bearing characteristics
under controlled conditions of rotation, preload, and shaft and housing fit. In this rig they
measured the relative deflection of the shaft centre with respect to housing, since the
housing was no longer attached to a rigid abutment. They claim this gives better results
for ball bearing vibration tests.
ts
II
Is
ii
:-
/..••••el 12.
Fig.(2.4) The recurrence frequency of the pulses from noise and vibration
(Igarashi and Yabe, 1983)
38
Igarashi and Yabe [1983] carried out their experiments in a soundproof chamber. They
measured noise produced by ball bearings and employed the same sort of set used by
lgarashi and Hamada [1982] in order to compare the results. Their experimental rig shows
that the noise from a vibrating machine can be very good source of information if handled
correctly and vibrations caused by bearings can be isolated from the other vibrations.
They showed that the recurrence frequency of the pulses was the same for both sound
and vibration produced by ball bearings (see Fig.(2.4)).
Aini [1990] designed an experimental rig to investigate the vibration characteristics of a

grinding spindle. A shaft was supported by angular contact ball bearings and a force
could be exerted on the shaft-bearing assembly to simulate the generated forces in the
grinding processes. Since this was a simulation of a spindle under working conditions,
the vibration spectrum should be very similar to those obtained from a grinding
_
machine.
5511
IS 17
11K (assc) TUE (mg)
Nal MAIM EPOMITATIfil Of 1111141 MI NW& 11101KMWMTIO Of IMAM

SIM MAK WILLI AT flee SIC MI IMMO Al AM
I I IIIII II 1111 Ill'''' ill''''
11111t1 1 111 11 III II I I II

I IS Al is
IS Al Tou (We)
T311 basil MT 11161151 IIPIES(VIAT/IO Of 111111M
SIC 011t NIKO AT Irala
WO) NIIITAT •PILSErIATTIII or issur
1111 lave sIMMID AT WOW
Fig.(2.5) The influence of the sampling frequency on the discrete representation of a

given analogue waveform (Xistris et al., 1980)
Setting up the rig and obtaining the data is only one part of the problem in experimental
research. The other is the acquisition of data and its analysis. Xistris et al. [1980] showed
that if the data acquisition Is not done correctly the results obtained could be misleading.
Fig.(2.5) shows the influence of the sampling frequency on the discrete representation
of a given analogue waveform. The digitilized signal decipted in Fig.(2.5) were obtained
by sampling a 10 V peak 1000 Hz sine wave at different frequencies. This shows what a
wrongly selected sampling frequency may cause, for example. Filtering and windowing
the data may also produce misleading results [Xistris et al., 1980; Mitchell, 1985].
39
0.25 —
fS Spindle Speed . 1000 rpm
Axial preload ., 100 N
Lubricant flow rate-40 dpm
0.2 —
a 0.15—
-1,
f bl
0.05 —
a
II
0 .0
1 4
1
,. I A..........., -.... A.
I r
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 (KHz)
Fig.(2.6) An experimental frequency spectrum Mini, 1990)
Even the data acquisition and transferring it into the frequency domain are performed
correctly, the spectrum obtained may be very complicated to acquire any information from,
since the effects of geometrical imperfections and natural frequencies are all present in
the spectrum. Owing to the large number of factors, the experimental spectra (the FFT
diagrams) have a large number of peaks, some of which correspond to different
modulation effects. The successful isolation and attribution of certain effects to specific
causes is difficult to achieve [Aini, 1990]. Fig.(2.6) shows an experimental vibration
spectrum. Since waviness, misalignment, off-sized ball effect etc. produces peaks at side
band frequencies (cant%) and sub and super harmonics to the fundamental frequencies
present in the spectrum [Wardle, 1988a; 1988b; Franco et al., 1992], almost any
prediction can be fit to this experimental data with a little justification. Filtering the signal
into specific frequency bands was suggested in order to avoid this problem [Braun and
Danter, 1979; Li and lnasaki, 1988 and Karakurt, 1989]. This requires a good knowledge
of the vibration of ball bearings since some important frequencies may be left out
otherwise.
40
2.5 Elastic Contact Behaviour of Ball Bearings

The classical solution for the local stress and deformation of two elastic bodies apparently
contacting at a single point was established by Hertz [1896] and today elastic contact
deformations are called Hertzian in recognition of his achievement [Harris, 1991]. Hertz
[1896] showed that the load-deflection relationship of the contacts between the balls and
raceways can be represented in the form;
3
F=K3 2 (2.1)
The characteristic spring action of the shaft and ball bearing oscillation system mainly
comes from the Hertzian contact of balls and raceways, and it is characterised by a variable
stiffness and hence nonlinear vibrations [Shimizu and Tamura, 1966]. •
Gupta [1975] modelled ball motion around the inner ring in six degrees of freedom. He
assumed that the balls posses mass and the normal =tact load comes from the Hertzian
contact deflection. The interactions between the cage and the balls also modelled in
terms of hydrodynamic and metallic contact interactions [Gupta, 1977; 1979a; 1979b;
1979c; 1979d].
Gad et al. [1983] followed a similar approach to Gupta in modelling the elastic reaction
forces due to contact between balls and raceways; they, however, did not consider the
mass of the ball and hydrodynamic interactions. The ball bearing modelled was ideal
which means the inner and outer races were assumed to be perfectly circular, all balls
were perfectly spherical and of equal diameter. Since the friction forces between balls
and raceways was not considered the balls were investigated in five degrees of freedom.
However, most of the researchers did not follow this approach. Tamura and Shimizu
[1966; 1967; 1968], Rahnejat and Gohar [1985], Aini and Gohar [1990] did not consider
the ball in a six degrees of freedom system, rather they calculated the total deflection in
the direction of ball to inner raceway and ball to outer raceway contacts. Having calculated
the total force on one ball, this force was then resolved to its three components in x, y
and z directions. Hence the total force and moment on a bearing were calculated by
summing the forces and moments in three dimensions, resulting in five degrees of
freedom system in general.
While the first approach may be useful in the calculation of "skidding and *skewing"
Instabilities where skidding normally represents large sliding between the ball and the
race, while skewing denotes rotation of the ball about its transverse axes, and
elastrohydrodynamic contact lubrication effects, the latter will give a good estimation of
41
lower band frequency of the vibration spectrum. The reason for this is in the first
approach since the ball moves in five degrees of freedom, It may produce instabilities for
certain loads and clearances [Gupta, 1979a; 1979c; Gad et al., 1983].
Although it was known that the load-deflection characteristic for ball to race contacts was
nonlinear, it was assumed for a long time that a linear line can represent the actual curve in
dynamical problems to avoid the difficulties caused by the non-linearity. Today, with the
help of powerful computers, in dynamic as well as static problems, the relationship is
modelled as a non-linear Hertzian contact. However, in contemporary studies of ball
bearings, three different approaches for modelling the nonlinear ball to race contact
relationship can be found. In the first, the nonlinear relationship is assumed to be linear
over a narrow range of deflection, resulting in a piece-wise characteristic curve
[Timoshenko et al., 1974; Matsubara, 1983]. Matsubara et al. [1985] employed this
model to investigate the vibrations of an elastic shaft supported by a pair of radial ball
bearings as seen in Fig.(2.7). The shaft/bearing model was formulated as an eigenvalue
problem and the first (fundamental) and higher natural frequencies of the system were
obtained under the application of various cyclic forces and shaft speeds. The model was
convenient since the system was continuous, they needed to use a superposition
technique which holds only for linear systems.
The second method uses the Taylor series in order to obtain an approximate curve fit to
the Hertzian contact. This was first employed by Wardle [1988a; 1988b]. Wardle [1988a]
assumed that the shaft and angular contact ball bearings oscillate about an equilibrium
position. If the equilibrium position after preloading is given by the deflection 60 at each
contact point, the Hertzian contact force can be linearised as:
I 38
F=K3g(1+ --)
2 (5,, (2.2)
The study [Wardle, 1988a] points out that for this linearisation, the ratio of ($5 /80 ) has to
be small in order to obtain accurate solutions. But if the same expansion is employed
taking another term into account, the numerical error between the real solution and the
approximation will be less than 2% in the case when the ratio of (3 /So ) is less than 1.
Taking this study as a basis Franco [1991] showed that if the higher terms of the Taylors
expansion are taken into account, the Hertzian contact can be represented very closely.
After the first six terms the effect of additional terms will not change the result very much
whereas the process will increase the CPU time. Therefore, there is no need for use of
more than six terms for practical purposes. In a paper by Franco et al [1992] this method
was employed to investigate the the effect of multiple defects of ball bearings on system
vibration characteristics.
42
W kb"
2000 Piecewise-linear
approximation line
1800
1600
1400
1200
g
1000
3 800[
600
400
200 i
I ; I 1
1 I I I I I i I I
4 8 10 12 14 16 18 20 22 24 26 28
112
Deflection 6 (pm)
Fig.(2.7) A piece-wise presentation of the Hertzian contact relation Afatsubara, 198£1)
Gustafsson and Tallian et al. [1963] employed a similar technique to the Taylor's series in
order to linearise the Hertzian contact deflection curve as:
F = —8)+F
(2.3)
where
is average deformation occurring under a load F
— and,
K L is linearised coefficient of contact deformation, defined as:
dF) 3 5,—R( 1)
— .—„ u 2
( d8 o.g 2 (2.4)
Equs.(2.2-2.3) were assumed to be valid in the neighbourhood of the point 8 5.

Equ.(2.3) gives a satisfactory approximation of Equ.(2.1) as long as I 8-8I is small
compared to the average deformation 8 [Gustafsson and Tallian et al., 1963]. The
second term on the RHS of Equ.(2.3) was introduced to compensate the minus values of
(8 - 8').
The third method employs the Hertzian contact relation. This is the most commonly used
approach and will also be employed in this thesis (see Chapter 3).
43
2.6 Important Frequencies Associated with Ball Bearings

In the literature available there are certain identified frequencies of a shaft supported by
ball bearings that should be avoided. Some of these frequencies have been thoroughly
Investigated while others remain largely unresearched.
Frequencies of shaft-ball bearing systems may be divided into two main categories: the
frequencies of a geometrically perfect (i.e., defect free) system and the frequencies due
to defects. In this section only the frequencies of the defect free system, or in other
words the inherentfrequencies of the system, will be investigated while the latter will be
investigated in the next section.
2.6.1 Natural Frequency

The natural frequency is one of the most important frequencies of the system. However,
it has been investigated by only few researchers for a shaft-ball bearing system. This is
perhaps because it was taken for granted. Shimizu and Tamura [1966; 1967; 1968] made
an attempt to calculate the static stiffness of ball bearings and showed that the difference
in the overall stiffness comes from different positions of the ball set in the bearing with the
effect being cyclic. Later, experimental research by Yamamoto and lshida [1974] showed
that the position of the ball set causes a change in the natural frequency of the system.
Fig.(2.8) shows the experimental result of the change in the natural frequency when the
position of the ball set is changed. Since the mass centre of the shaft they studied did
not coincide with the geometrical centre of the shaft, a non-uniformity of the stiffness was
present. Therefore the researchers obtained the natural frequency of the system in the
direction of the maximum, po i (the natural frequency in the x direction) and minimum
stiffness, p021 (the natural frequency in the y direction) respectively by hitting the shaft in
x ory direction and recording the vibrations. When the natural frequencies were plotted
against different shaft positions as shown in Fig.(2.8), a cyclic change in the natural
frequencies was observed. However, these results seem to be caused by wrongly
selected experimental observation intervals (see Chapter 6 for a further discussion on
this). From these results Yamamoto and Ishida [1974] concluded that when the shaft is
rotated, the natural frequency will fluctuate periodically with shaft speed and its
superharmonics and hence the system will be excited at the shaft rotating speed and its
superharmonics.
Gupta [1974] has reported that ball bearings exhibit a hardening type spring and the
change in natural frequency is therefore the result of change in the stiffness. Gad et al.
[1984; 1985] observed jump behaviour around the natural frequencies and reported that
this is due to the hardening type spring characteristic of ball bearings.
44
Fig.(2.8) Natural Frequency vs. Different Positions of Balls

(Yamamoto and lshida, 1974)
Yamamoto and Ishida [1974] and Yamamoto et al. [1977] reported that outer ring
misalignment causes change in the spring characteristic of ball bearings and hence affect
the natural frequency of the system. These effects resulted in jump behaviour over a
large frequency range in the experimental researches by Yamamoto and Ishida [1974]
and Yamamoto et al. [1977].
15.
120 Hz. +
...........""°.. Stiffneu
Z 10 10
b
;
a
a
05 %
4.. 4....
+
Phase
+.
+ ,
o o
500 1000 1500 2000
Speed (revs/min)
Rg.(2.9) Stiffness and Phase Characteristics of Bearings as a Function of Speed

(Watford and Stone, 19801
However these results of changes in the natural frequency were obtained for a non-
rotating shaft. Later Watford and Stone [1980] and Yamamoto et al. [1981] did a series of
experiments in order to obtain the stiffnesses and natural frequencies of ball bearings
under oscillating conditions. Watford and Stone [1980] concludes that the stiffness and
the damping of the system varies with the rotational speed as seen in Fig.(2.9). Stiffness
45
Increases linearly as predicted, implying a hardening type stiffness. Yamamoto at al.

[1981] confirms this and shows that the natural frequency increases parabolically as seen
In Fig.(2.8), resulting in a system with a hardening type spring characteristic. Watford and
Stone [1980] reported a big difference between the measured stiffnesses and the
Hertzian contact spring. This was considered as a result of the interfaces between the
races and housing, respectively. Although it is not very clear how they calculated the
equivalent stiffness for the system, this big difference (7x10 8 N/m calculated, lx2 108
N/m from the practical tests) may come from modelling of the bearings as linear springs.
These experimental studies will be discussed in Chapter 6 in more detail and the
comparison with the results of computer programme will be given.
AktOrk et al. [1992] showed that increasing preload would force the system to oscillate
with a higher ball stiffness bearing and therefore the natural frequency of the system
would be increased (see Fig.(2.10). This effect will also be discussed further in Chapter 6.
Prelow)
a Amplitude of Vibrations about Steady-state Position, lx I

b. Preloaded Contact Angle, ap,
C. Natural Frequency, con,
d. Axial Deflection, to
Fig.(2.10) Some Factors Changing with Preload (AktOrk eta!., 1992)
2.6.2 Ball Passage Freauencv
The ball passage frequency (BPF) was first recognised by Perret-Meldau [1950] as a
static running accuracy problem. Perret and Meldau suggested that an increase in the
number of the balls in a bearing, reduces its untoward effect. However, Its nature and
importance was not properly understood until research by Gustafsson at al. [1963], who
called this vibration "variable compliance". Due to the difference in the overall stiffness,
46
for the different positions of the ball set, a force imposed on the inner ring and it excites
the system at the BPF as will be explained in more detail later in Chapter 4. This is an
Inherent frequency of the system, i.e., it is present even when the bearing is
geometrically defect free. The experimental studies of Gustafsson et al. [1963] showed
that clearance is an important parameter for the BPF. The greater the clearance, the more
effective the ball passage vibrations become. This was later confirmed by the
experimental research of Wardle and Poon [1983]. Wardle and Poon argue that the ball
passage vibration levels depend on the number of balls and radial clearance. For radially
loaded or misaligned bearings, running clearance determines the extent of ball passage
vibration level [Wardle and Poon, 1983]. In general the ball passage vibration increases
with the internal clearance as seen in Fig.(2.11).
Gustafsson et al. [1963] also observed that higher harmonics, which are multiples of the
BPF, are also present in the vibration spectrum (see Fig.(2.12)). The results shown in
Fig.(2.12) are approximate, higher harmonics, especially was used only for comparison of
orders of magnitude. The radial loads were varied in the range from 100 N to 10000 N
and the amplitudes of the ball passage vibrations in the figure represent the maximum
amplitudes of each harmonic which occur under different radial loads. Although the
physical explanation was not given, the researchers [Gustafsson et al., 1963] pointed out
that the amplitudes at the harmonics of the BPF depend on the radial load, radial
clearance, the rotational speed and the order of the harmonic. The same conclusion was
later reached theoretically by Meyer et al. [1980] for a perfect ball bearing, with linear
modelling of the spring characteristics of balls. Meyer et al. [1980] showed that for a
perfect system, ball passage vibration and its superharmonics will appear in the spectrum
(see Fig.(2.13) and Appendix 7).
O..", NW 63,3
Meal Lona SIX 4
...........-----
............5.............ir ,S X 25 Misr my
Om ove• yono
Fig. (2.11) The vibration due to the BPF vs. internal clearance
(Wardle and Poon, 1983)
47
OM IM t reterrqs1 it 11 MC PM
LOAD
INCE • ZO - /COO LIS
CONFORM II TOE 011ECTIN OF TOE 1.04
CC/MEW FT.IIFFAIOJUX TO TNT 011ECTION OF TIC LOAD
10,000'
ROO
Firvissev •s
Fig.(2.12) Maximum Values of Velocity Amplitude of the Harmonics of the BPF

pustalsson and Tallian eta!., 19631
A resonance was reported by different researchers when the BPF coincides with the
natural frequency [Gustafsson et al., 1963; Meyer et al., 1980; Wardle and Poon, 1983;
Rahnejat, 1984; Gad et al., 1984; 1985; Rahnejat and Gohar, 1985; Wardle, 1988;
AktOrk, 1988; Aini, 1990; Aini et al., 1990; AktOrk et al., 1992]. This means the system
may resonate at a fraction of the natural frequency since the BPF is defined as the cage
speed which is about the hat( of the shaft speed, times the number of the balls in a
bearing. Rahnejat and Gohar [1985] showed that even in the presence of an
elastrohydrodynamic lubricating film between the balls and the inner and outer races, a
peak at the BPF appears in the spectrum. Fig.(2.14) shows the steady-state frequency
48
spectrum of the oscillations in the vertical direction with a wavy inner race. The
subharmonics of the BPF also appear in the frequency spectrum. However Rahnejat
[1984] could not observe the BPF effect for a perfect ball bearing arrangement, probably
because of the preload or heavy mass present or insensitivity of his computer program.
IC Pn
;LI. TONES OF AMPLITUDE: vr
a; , • .. nX
. I r, ANGULAR
FREQUENCY
8
X.i 2X.• nX •si
Fig.(2.13) Frequency Spectrum for Bearing without defects (Meyer et al., 1980)
I Shaft rotation frequencyfi

2 '6 A
3 Ball-pass frequencyfb
4 Base natural frequency/.
5 1.-In
6 J. — fe
7 Wave-pass frequency/.
V
.0
41
M. i
E
<
I-0
Frequency
kHz
Fig.(2.14) Steady-state Frequency Spectrum of x oscillations with wavy inner race

Flahnejat and Gohar, 1985]
The study by Tamura and Gad et al. [1983] is a kind of exhibition of the Meldau-Perret
problem. They showed that due to the different positions of the balls in a bearing there
will be an oscillating force in the system at the frequency of the ball passage. This study
was extended and republished by Tamura and Gad et al. [1985] with more attention paid
to the BPF. Fig.(2.15) shows the presence of the BPF for different values of
dimensionless characteristic parameter r, which is a combination of radial clearance,
number of balls, the contact deflection proportionality and the radial load. Although the
49

▪

shape of the inner ring motion changed, the oscillations show the effect of the BPF. This
is further discussed in a paper by Tamura et al. [19861
:I
••

"" !Iwo sill
11.NA

1.101 1(90
8 • •'s 1 9 1:1 • • 55 1.6 1.1 -ea 89111
r.-I.175 r.I.ie
• 11
9
.0 1
1 111 • too
MO
.113
1 n
• • 1145 1 • 1.5 • • • I•
w- I .110 r-2.35
• S SI • SI
NM)
ell LI
...“
1.11 • 1.1
Se9.11111 Ispw9.11,1
" 1.1
11.5 1.0 1.1 OA 0.1 la 1.1 S.
1..4.025 r.2.se
•,,w1.81
a13
91 19 55 .4 81 • • • 01 • 9.1 IS 1.1
r.-0.1150 1.10
•pr53 I
I- Zed
r.-0.120
Fig. (2.15) Radial Motion of Inner ring (Tamura and Gad et at, 19857
Gad et al. 11984131 investigated the BPF and its sub and super harmonics for perfect ball
bearings. Fig.(2.16) shows a selection from this paper showing the effect of the BPF and
its sub and super harmonics in the vertical and horizontal directions. In the figures, the
50
system exhibits the 3rd order superharmonics of the BPF. The locus of the shaft centre
for these oscillations was also presented. Gad et al. 119&4a] also point out beat like and
chaotic vibrations of the shaft-ball bearing systems. As discussed earlier, It should be
taken with precaution and further investigated since these may be due to the modelling
of forces and moments for a single ball.
III r10-2
0.23'
II -0.33
-2.00
a
• .2.26
▪ -2.52
-0.24 -0.04 0.00 0.24 .0.33 -0.13 0.03 0.22

aagal alma. a (a.) 410 -1 04414. s (m4) 410-1
Fig. (2.16) Steady-state vibrations and paths of inner ring centre due to the BPF effect
pad et al. 198414
Gad et al. [1984b] later showed that resonance occurs when the BPF coincides with the
frequencies of system anomalies such as inner ring eccentricity and the misalignment of
inner ring relative to the shaft axis. In this study, researchers pointed out that for certain
speeds the BPF can exhibit its sub and super harmonic vibrations for a given shaft-ball
bearing system.
Aini [1990] argues that the ball passage vibrations appear as a result of contact spring
nonlinearity and become effective when there are system irregularities or when a forcing
frequency excites the system. But the analytical linear model of Meyer et al. [1980]
suggests the ball passage vibrations occur even for perfect bearings. The reason for this
is when a radial clearance is present the balls of the bottom half of bearing will always be
carrying the load and therefore the BPF effect will be present in the vibration spectrum.
However, for the nonlinear model, even in the case of interference, the BPF will cause
resonances due to the nonlinearity of the bearings. Aini [1991] argues that for perfect
bearings the BPF does not necessarily appear in the spectrum since the levels of
51

vibration due to the BPF depend generally on the number of balls supporting the applied
load, the greater the number of balls, the less the vibration amplitude. However, AktOrk et
al. [1992] points out that the effect of the BPF is always present in the spectrum, however
small it is. AktOrk et al. 0992] investigated the effect of number of balls and the preload
on the vibrations due to ball passage. Fig.(2.18) shows the effect of changing the
number of balls. When the number of balls Is increased the BPF will coincide with the
natural frequency at lower frequencies as seen in Fig.(2.18).
10
e 1 '
t .„. P k •-e_ -0 - -. I • e
• _. nom 0 .m. •Pat -
d.
P. 141•20 k , Pr=-0 N
•„
I 0 No. of Balls
: •
.
. (•—• ) • 5
.' '•0 0- —*) B
• ( • ••• ) 12 -
... ..
.001 • . . -. _. ..„ ,
0 2500 5000 7500 10000 12500 15000 17500
Speed (Rpm)
Fig. (2.18) Effect of varying the number of balls MktOric et al., 1992]
In spite of its importance, the effect of the BPF is generally not clear. Some researchers
observed the BPF in their experimental researches but could not identify it as such.
Three examples are given below:
GlOckner [BBJ 225] shows the effect of the BPF, even identifies its speed but did not
recognise that it is one of the inherent frequencies of the system. Fig.(2.19) shows the
the force acting on the shalt and the measured signal as a function of spindle load. The
in Fig.(2.19) coincides to the BPF of the ball bearings but the author calls it "the over-
rolling frequency" and fi coincides with the cage speed. As the figure suggest, although
it is not mentioned in the paper, that the bearings had 10 rolling elements.
52
F[N]
..............----......%................/
tts1_
Fig. (2.19) Measuring signal as function of spindle load (GlOckner , B13.) 225]
Fig.(2.20) shows the vibrations resulting from a once per cage revolution averaged during
a bearing test. The bearing had 8 balls. When the figure is investigated, 8 oscillations per
cage rotation are observed, which means vibrations at 80... This is the BPF. However,
Hemmings et al. claim that this is due to surface waviness. This seems more unlikely (see
the discussions on the BPF and waviness in this chapter). Perhaps the authors were
unaware that even geometrically perfect bearings can undergo vibrations.
—I
200 pmis
1 Revolution
Fig.(2.20) Bearing test results (Hemmings et al., 1976)
Karakurt [1989] identified the BPF which is marked with NORDF —Outer Race Defect
Frequency in Fig.(2.21a) very clearly with a defect free bearing in his experimental
investigation but attributed it to an outer race defect (see Fig.(2.21)). He admits that this
is not a defect frequency as the bearings can be assumed defect free (defects present in
the bearings are negligible when they are compared with the large local defects
Introduced by the researcher). It is know that the BPF and the outer race defect
frequency have the same frequency. Hence the author wrongly identified the BPF as the
outer race defect frequency.
53

- C ct r ;xi
24. 000 -
vs 18. 000 —
12. poo - A
0. 000
0. 0 100. 0 MO. 0 120C1. 0 1600. 0 2000.0
r equ4ru..-y (1.t.3
Zooivid Sp ect ruoi

SO. 523
a 24.4/9
I„ 4. .0
— 0
CO
01
•.-n
MI •
4. b. b.
13 12. 205 c.. 0 r...0
o = o b.
0 0 CQ C.
0 eV =en .7 0
Ill
• z- 6.105
0.000
0.0 SW. 0 1200. 0 IOW. 0 2000.0

equen,..-y (HZ]
Fig.(2.21) Zoomed spectrum of New and Damaged Bearings (Karakurt, 19891
2.6.3 Flexural Vibrations of the Outer Ring due to Ball Loads
This vibration was first studied by Gustafsson and Tallian et al. [1963]. They reported that
flexural vbrations of the outer ring were induced by ball loads in an axial as well as a radially
loaded ball bearing. It has the following characteristics [Gustafsson & Tallian et al., 1963]:
1. The fundamental frequency of vibrations is the frequency of balls passing a given

point on the stationary outer ring of the bearing, i.e., the BPF,fbp.
2. Higher harmonics, which are multiples of the fundamental frequency, are also present.
3. In a thin ring (such as the outer ring of most bearings) of mean radius R, and second
moment of area of the ring cross-section I, loaded by an axial load Fa the r.m.s. velocity
amplitude of the i th harmonic of the flexural vibrations is given by [Gustafsson and Tallian
et al., 1963]
54
Fa R3 i m co(1-10-cos(a))
0 2
71-i= x242- E4(im)--11 sin(a)
(2.5)
As is suggested by Equ.(2.5) the velocity amplitude is directly proportional to the applied
load and shaft speed and decreases inversely with the cube of the number of balls, m
and with the cube of the order i of the harmonic. The fundamental frequency is therefore
predominant. The velocity amplitude of the second harmonic is approximately 12% and
the third approximately 3.5% of that of the fundamental [Gustafsson and Tallian et al.,
1963]. Higher harmonics have successively lower amplitudes. Although the research
suggests that this vibration is relatively small for normal applied loads, for loads exceeding
these values, the effect of the flexural vibration due to ball loads may become an
important contributor to the vibration level below 800 Hz.
2.6.4 Natural Frequencies of' the Outer Ring
The natural frequency of rigid body motion of the outer ring was studied by Gustafsson
and Tallian et al. [1963] for a free outer ring, supported only by the balls and also for the
outer ring supported by the balls and mounted in a housing which is elastic both radially
and angularly. In this study they assumed that the outer ring deforms elastically only
where it contacts with the balls. Gustafsson and Tallian et al. [1963] calculated three
natural frequencies of rigid body motion for radial, axial and angular mode of vibrations
respectively. They also studied the natural frequencies of the elastic vibrations of the
outer ring. In this case the outer ring deforms elastically, e.g., by bending. This natural
frequency was computed under the assumption that the inner ring was rigid and joined to
the outer ring by a set of equally spaced, identical, elastic balls. Gustafsson and Tallian et
al. [1963] showed that a ring with multiple elastic supports (such as a bearing outer ring
supported by balls) has a sequence of natural frequencies of flexural vibrations which are
not integral multiples of a fundamental frequency and are dependent on ring mass and
the elastic properties of the ring and supports.
Gustafsson and Tallian et al. [1963] suggested to use a bearing with a thicker outer ring to
reduce the vibration level due to the rigid body motion of outer ring.
2.6.5 Natural Frequencies of Ball Motion

Gupta et al. [1977] modelled the motion of the ball in a cylindrical coordinate system. The
simulation model showed two natural frequencies, one being a lower frequency and the
other higher. Gupta et al. claimed that these were two unknown natural frequencies of
the ball bearing systems and called them elastic contact frequency and bearing
kinematic frequency (see Fig.(2.22)).
55
OMR
IOU
COVTICT
:AN
MOTU,
.10
mi. o music i
Fig. (2.22) Variations of computed elastic contact frequency and bearing kinematic
frequency as a function of ball-race contact loads (Gupta et al., 19771.
The elastic contact frequency is obtained from the linearised Hertzian contact deflection
relation as follows [Gupta et al., 1977]:
1
3Q )2
w = (____
1 2 MI, (5
(2.6)
where Mb is the ball mass and Q is the load. The bearing kinematic frequency is
obtained from the classical expression of the type CO - (g 11 )1/2 as [Gupta et al., 1977]:
i
1 Qy
wk = 111b1)
-- (2.7)
where 1 is the length of the oscillating pendulum and effective 1 is very difficult to
determine. Gupta et al. [1977] gave restricted experimental evidences for their theory.
However, specially the bearing kinematic frequency seems to be arguable since any
frequency can be matched playing with 1. In the same papers discussion [Gupta et al.,
1977] Frarey argues that the predictions does not give reliable results since the
experimental investigations showed that the predicted signals were not shifted for
different load conditions although a shift was suggested by Equ.(2.7). Fig.(2.23a) is
56

submitted by Gupta et al. as the proof of the mentioned frequencies (co e-53.23 KHz and
4=8.04 KHz) and Fig.(2.23b) for the same bearings but for different load conditions
shows no shift in the predicted vibrations hence they are fixed resonances and not the
elastic contact frequency and the bearing kinematic frequency signals as suggested.
1 MP 10110•111. ma la -.A-- NM ore 11•171

C low ••••••••11 Ma ea SP IP Uwe Ilamal
I ...MEM llin Ina.. SIM • MI ______________________ 6... la ••••

Ira. mar... I - • I I I ......sLAW11--
masa III •N 4 WM OM • I vara .." •.•
tom anum. Om.= ao aware ibs a a
•
111•••••••••• 11
It
ORM
MMEM MOM
I 11..--•-•4
UMW,
MM.
1111•11•1•11•••NO.1111
111•••111•••••111111
111•••••••••1111111 flOw.
J&A
11•••••••••111111
111111••1111•••1111111 elnWflaimmI aan
WAIIIIMIEMIPMEIMP.11.111111111 Peftlionw
•••• tames
•111, 1111•61•911•19
NM gm N a sown INInrd
1111/8 ai • NS 11111111 WWI
c112.44,
••••••n• kaal. ""w IMO •-t Ufa
I truwayar •••10,rn 1 • 4 •n•• - a.a.././1-

N of's • so . PI•11 r

••••••111 .M) OM Dow 11•II ••n•••
Wei MI - tla Ilat
.7%../..#"."...
t ..: 1 ' ! :
... A 4 kanh*rtzoliiii a W.
"re.t,iie't
. . .
•
11111111111111.11
111111111111111111111
Ila WEI.n••••n•n
Fig. (2.23) Radial acceleration spectrum (Frarey's discussion [Gupta et al., 1977]
2.7 Vibration Monitoring and Frequencies due to the

Imperfections of' Ball Bearings
As bearing applications become increasingly complicated because of the requirements

for continuous operation, automation, etc., failures in rotating machinery are expensive.
The breakdown of this kind of machinery always causes additional expense, not only
57
because repairs are costly, but also because of the loss of operating time and production.
Since bearings are a particularly critical component of machinery and their serviceability is
usually essential for the operation of any machine into which they are fitted, it is important
that bearing damage should be detected early enough to replace the bearing during a
scheduled servicing period, thus reducing unscheduled servicing of the machine. This is
particularly true where safety is a factor, as in aircraft engines, and certain process
Industries [Yhland and Johansson, 1970]. To illustrate the cost which can be involved it
has been reported that a simple bearing failure in a fully integrated steel mill can lead to a
total shut-down which at full output rate may cost £150-300 per minute [Braithwaite,
1969]. A similar bearing failure In a generator set could involve a loss of E1-20 per minute
until the set was again operational [Scott, 1972]. In order to avoid failures and losses,
"Machine Health Monitoring" can be the solution. When early failure of any part of
bearings is diagnosed and located at an early stage the remedial action will be relatively
inexpensive and shut down time for the replacement of the failed bearing can be planned
In advance [ Wallis, 1966; Karakurt, 1989]. The monitoring of rolling bearing installations
can serve two important purposes: firstly they provide continuous 'health checks' on
bearings without the need for dismantling the machinery; and secondly bearing damage
can be detected at such an early stage that failure, and the possibly serious
consequences of it, is avoided.
Health monitoring cannot be isolated from other areas of research as it has often
concentrated on understanding vibration characteristics of ball bearings with the main
objective of detecting problems of ball bearings after they are assembled. To obtain
information while bearings are mounted is difficult. Perhaps the only way to get
information from such bearings is to investigate the vibration and noise produced by
them. The condition of ball bearings can be judged by the nature of their vibration
pattern. If a change in the vibration pattern can be monitored and analysed, the health of
the ball bearings can be determined.
In previous years some work was done on early detection of rolling bearing failure
[Dowson, 1970; Hemmings and Smith, 1979; Braun and Danter, 1979]. In the early days
of vibration monitoring, a simple but very effective method was employed. A wooden
shaft acted as a good listening post because it transmits noise which relates to the
condition of the bearing; it also cuts out much of the noise from other machine
components which cause confusion and doubt when using a more sensitive
stethoscope. This method is still used today in practical applications for simple cases.
But for more complicated cases, the vibrations and sound produced by the machine
should be recorded and investigated carefully [SKF, 1961; Wallin, 1966].
58
However, to obtain the vibration due to ball bearings alone is another difficult subject. As
discussed earlier under Section 2.4 the noise produced by ball bearings may not be
useful information to predict the bearing vibrations since In the majority of installations the
direct noise from a bearing is of low power and is generated inside a sealed component.
Under these conditions negligible noise energy escapes through the casing or through
seals and this source of noise may be ignored. There are also other sources of conflicting
signals and obstacles to their collection [Hampson, 19841 Therefore the best way of
getting information from the shaft-bearing assembly is direct measurement of shaft
vibration but this is not easily made when bearings are assembled. As vibration forces are
transmitted through bearings, measurements are generally made on bearing housings as
near to the shaft as possible [Dowson, 1970]. However, obtaining the vibration signature
is not enough in health monitoring since this signal is useful only when it is properly
understood. To be able to investigate the vibration spectrum obtained by monitoring ball
bearings, their vibration characteristics must be known. In practice the acceptable
vibration levels for particular bearings and machines are largely a matter of experience
[Dowson, 1970]. Therefore, when bearings are installed they should be monitored and
the healthy bearing signals should be recorded. A trend of increasing vibration will
indicate that the trouble is becoming worse. Different type of defects will cause different
vibration patterns. Experience of various causes of bearing damage can be gained by
highly qualified people over a long period. An additional difficulty is that the information
obtained from the investigation of one design of bearing arrangement cannot necessarily
be applied directly to another [Yhland and Johansson, 1970]. Once considerable
experimental experience of a certain arrangement has been gained, it is usually known in
which section of the spectrum the changes tend to lie and this reduces the work of
analysis. On the other hand, it is not generally possible to predict a determined absolute
level of the vibrations, because even modern mass produced assemblies with close
tolerances on their components show great differences in their vibration spectra. Hence
the importance of a clear understanding of vibrations associated with ball bearings is
obvious.
Many researchers investigated the heath monitoring of ball bearings for different types of
defects with varying degrees of success [Wallin, 1968; Hemmings and Smith, 1976;
Taylor, 1980; Igarashi and Hamada, 1982; lgarashi and Yabe, 1983; Hampson, 1984;
McFadden and Smith, 1984; lgarashi and Kato, 1985; Kanai et al., 1987; Karakurt, 1989;
Khan, 1991]. Since there are many studies on the subject the papers will be reviewed
under subtitles according to the defect type they consider.
2.7.1 Imperfections of Ball Bearings
It is generally accepted that It is impossible to produce a perfect surface or contour even

with the best machine tools, and this applies also in ball bearing manufacture. The minute
59
surface irregularities of the bearing components and their effects on one another, result
in vibrations of the bearing. Parts of a machine which resonate with a frequency
produced by a bearing are a common source of high sound-level noise [Cena and Hobbs,
1972]. As a result of experiments a specific relationship has been found to exist between
the internal dimensions of bearings and the vibration factors for the different
components, the rolling elements being more important than the outer ring and inner ring
[SKF, 1961; Cena and Hobbs, 1972].
Waviness, off-sized balls, misalignment can be cited for the manufacturing and
assembling imperfections. These are discussed below.
2.7.1.1 Waviness
Anomalies may occur during the machining processes by the manufacturers which affect
the form and the finish of the surfaces produced [Wardle et al., 1988; Aini, 1990].
Comparison of the surface features compared with the deformed contact (Hertzian
contact) dimensions of the rolling element raceway, in terms of feature wavelength, leads
to the terms "Roughness" and "Waviness" [Wardle and Poon, 1983] (see Chapter 1 for
the definition of roughness and waviness).
The significance of the level of vibrations due to roughness is evident when asperities
break through the lubricant film and contact the opposing surface. It consists of a random
sequence of small impulses which can excite all natural modes of the bearing and the
supporting structure. This phenomena was first recognised by Sayles and Poon [1981].
Vibration caused by surface roughness is not considered in this thesis.
The importance of running surface waviness from the vibration point of view was known
for a long time. However, not until the 1960s was it studied. This was mainly because to
determine the surface waviness was very difficult. In the 1960s, with the availability of
vibration testing machines which could also measure the surface waviness (see the paper
by SKF [1961]), the research on the contribution of waviness to the shaft-ball bearing
vibrations increased. Gustafsson and Tallian et al. [1963] studied the effect of waviness
and came to two important conclusions: Firstly, low order outer ring waviness affects the
amplitudes of the vibrations at the BPF and secondly inner ring waviness of the order k
waves per circumference produces flexural vibrations of the outer ring (their pick-up was
on the outer ring) with a predominant peak of k times the rotational frequency. The
amplitudes at other frequencies were small compared to the peak at times the rotational
frequency and the amplitudes of vibrations were significant only for low orders of
waviness. An experimentally obtained spectrum of a bearing with inner ring 2 and 3
waves per circumference is shown in Fig.(2.24).
60

z notati•sal Froqueney
lie= Ball Pass Irrequ•ney
—Ii lonl000lowIerri 1 raw owl 1111.0411
U0
1.
m ?
•
.41
19 fit
10 ..
021 0.4 0.3 02 01 — CT + 01

WINCES WIDT14 /ACTOR

02

03

0.4

•

1111•Mii=1 11411.111 ••••n14.11 OW. U ellMat NEW IMRE 111..• MIL 100.1•Mia SOWS NI ...b.

50 100 150 200
Fig. (2.24) Narrow band spectrum of a ball bearing with inner ring 2 wpc and 3 wpc
waviness at a rotational speed of 1800 rpm (Gustafsson and Tallian et al., 1963)
Gustafsson and Tallian et al. [1963] further concluded that the following formulae could
be employed to find the approximate ranges of orders of waviness:
(2.8)
(2.9)
(2.10)
where kw ki and k b are the orders of inner ring, outer ring and ball waviness
respectively, generating vibrations of frequency off Hz. h is the rotational frequency of
the cage with respect to outer ring, ft the frequency of the cage with respect to the inner
ring andft, the polar rotation frequency of the ball, in Hz.
Yhland [1967] examined the correspondence between waviness and the resulting
vibration spectrum. For a bearing with m rolling elements, if p and q are integers equal to
or greater than 1 and 0, respectively, then for vibrations in the radial direction measured at
61

a point on the outer diameter of the outer ring, the vibration circular frequencies as
functions of inner ring, outer ring, and roller waviness are given in Table (2.1) [Yhland,
1967; Harris, 1991].
Waviness Vibrations
Component of orders (caused by waviness)
Inner Ring k=qmtp qm(01 - Wc)±PC0i
Outer Ring k=qm.±-p qm4
Roller k (even) k alb ±p Oic
Table (2.1) Vibration frequency vs. waviness [after Harris, 1991]
The bearing outer ring moves as a rigid body when p= 1. For p> 1, vibrations are of the
flexural type with p equal to the number of lobes per circumference of the outer ring
deflection curve. For a waviness spectrum obtained at an inner ring speed of 900 rpm, for
a bearing with an accentuated inner ring waviness, Yhland [1967] obtained the vibration
spectrum at 1800 rpm shown by Fig.(2.25). The Ta/yrond trace of the inner ring also
shown in Fig.(2.25), the tested tapered roller bearing contained very smooth rollers and
outer ring.
2 3 4 5 6 7 8 910 15 20
Waves per circumference
Peaks caused by wavi ness of order:

2 8 16 24 32 40
I /
s 4 1000 •
I
a /
1. 100
> a
Vibration frequency
10 1 1,1 ./ iLA (cycles/sewnd)

100 150 200 300 400 500 1000 1500
Speed: 900-1800 rev/min.
Fig.(2.25) Waviness and vibration spectra from inner ring with accentuated waviness
(Thland, 196
62
Cena and Hobbs [1972] concluded that ball waviness was of greater significance than
raceway waviness on the vibration of ball bearing systems. Meyer et al. [1980]
theoretically studied the outer ring waviness and concluded that the vibration spectrum of
a ball bearing with a wavy surface would have peaks at N co, coc , the BPF and its
superharmonics and peaks around the wave passage frequency and its harmonics at
±N co, where N is the number of waves (see Fig.(2.26)).
P Po
AMPLITUDE: wg-q- AMPLITUDE:
2 M atm watm
6(w)
0 11k 1 pt
mw
OR
A . 61
(Mato:a
OR )
. 21C . a, I
(2Ksatmar
0R )
Ia•as
ANGULAR
FREQUENCY
Fig.(2.26) Frequency Spectrum for Waviness (Meyer et al., 1980)
Thomas [1982] pointed out that waviness contributes to low frequency noise, so at
relatively low speeds it causes few problems with precision bearings. It only becomes a
serious problem when components possess relatively large amplitude features or high
speeds. However, Gupta [1988] argues that waviness does not have a significant effect
on overall bearing dynamic performance and for practical applications it can be ignored.
However, in contrast, Wardle and Poon [1983] found that the waviness causes most of
the severe vibrations and noise problem in bearings. They reported that waviness
produces vibrations at frequencies up to approximately 300 times the rotational speed
but is predominant at frequencies below about 60 times rotational speeds. Wardle and
Poon [1983] also pointed out the relation between the number of balls and the waves for
severe vibrations to occur. When the number of balls and waves are equal there would be
severe vibrations since in this case there is symmetry of loading and all balls vibrate in
phase. Later Wardle theoretically [1988a] and experimentally [1988b] showed that outer
race waviness produces vibrations at the harmonics of outer race ball passage frequency,
m cec as also calculated by Gustafsson and Tallian et al. [1963]. Rahnejat [1984] and
Rahnejat and Gohar [1985] realised that inner ring waviness is somewhat more
complicated than what Gustafsson and Tallian et al. [1963] and Meyer et al. [1980] have
predicted. Wardle [1988a] showed that in the case of inner ring waviness the axial
vibrations take place at frequencies harmonic with the ball to inner ring passage rate
m(o) — co), whereas radial vibrations occur at frequencies m(coc — cod ±coi . Wardle
[1988a] argues that only specific orders of waviness generate vibrations. Axial vibration is
63
produced when the number of waves per circumference is an Integer multiple of the
number of balls, whereas radial vibration is produced by waviness of the order of
N=i m .± 1. Some of Wardle's arguments were later proved by AktOrk [1988], Franco
[1990] and Franco et al. [1992].
Wardle [1988a] showed that ball waviness produced vibrations in the axial direction at
frequencies given by 2 i 4, while radial vibration occurs at frequencies given by
2 i Cob t 4. He points out that only even orders of ball waviness produce vibration; this
was also pointed out by Yhland [1967]. This is understandable since otherwise, ball
waviness would cancel itself out. In practice this is not the case of course. This problem
arose from the assumption that the ball was assumed to remain rotating about an axis
normal to the plane containing the centres of the inner and outer race contacts.
2..7.1.2_121L:Sizet_Balls
The existence of off-sized balls in a bearing introduces further untoward vibrations to the
shaft-ball bearing system. It is impossible to produce a set of identical balls even with the
best machine tools, there is always some difference between ball diameters (less than the
machining tolerances). _It is worth pointing out here that the balls for precision ball
bearings are selected to be as near as possible the same within tight tolerances and
surface finish. The cheaper the bearing, the more these tolerances are relaxed. These
tolerances are not usually significant dimensionally as far as circular or elliptical contact
footprints are concerned. However they may play an important role in the vibrations of ball
bearings. The off-sized ball effect has been studied by different researchers. Tamura
[1968], for example, experimentally showed that the ball bearing axial stiffness varies with
the cage position, since the number of loaded balls at any instant is changed by the balls'
diameter differences. The appearance of rotor axial resonance was also reported by
Tamura [1968] as a multiple of i (i .1, 2, 3,..) times the angular velocity of the cage, when
it was operating close to the natural frequency of the system. He further concluded that,
in the case of a perfect bearing with no variation of diameter among balls, the resonance
frequency did not appear.
Fig. (2.27) Frequency Spectrum for Unequal Ball Forces (Meyer et al., 1980)
64
Barash [1969] studied the effect of an off-sized ball on the ball speed variation. He
reported that if one ball is bigger than the rest of the balls in a bearing, It will lag
continuously since It will have a smaller contact angle and lower speed. This causes
peculiar vibrations.
Meyer et al. [1980] analytically investigated the off-sized ball problem and come to the
same conclusion as Tamura [1968]: k causes vibrations with the multiples of cage speed.
It will interact with the BPF and cause an increase in the vibration amplitude as seen in
Fig.(2.27). Yamamoto et al. [1981] also experimentally observed a peak at the cage
rotational speed and another one at twice the cage speed.
0.8
0.6 g
cp- 0.4 a
z
fa.
< 0.2-
hi
1
0.0
0 20 0 400 600 800 1000
Frequency, f [Hz]
Fig. (2.28) Calculated frequency spectrum for vertical vibrations of a bearing with defects
[Franco et aL, 1992]
AktOric [1988] using angular contact ball bearings and Franco [1990] using radial ball
bearings, studied the problem of an off-sized ball and both reported vibrations at cage
speed and influence of an off-sized ball on the other frequencies. Although AktOrk
[1988] observed vibrations at three times the cage speed for a ball bearing with 15 balls, 3
of which were 1 gm oversized and symmetrically distributed, he considered 11 simply as
the superharmonic of the cage speed. Franco [1990] and Franco et al. [1992] observed
vibrations due to off-sized balls at the cage speed due to the random distribution of off-
sized balls within the bearing. They showed that when there is one oversized ball in a
bearing the most dominant vibrations occur at the cage speed which is indicated by (a) in
Fig.(2.28).
65
Gupta [1988] showed that when the off-sized ball size increases, some performance
deterioration was observed. He reported that the shape of the cage whirl orbit changes
from circular to somewhat polygonal with increasing variation in ball size. This may be the
result of ball speed variation as also reported by Barash [1968].
2.7.1.3 Misalignment
Even bearings are themselves completely satisfactory, through inaccurate machining of

shafts or bearing housing, or by adopting poor assembly techniques, the bearing can be
deformed or the rings can be misaligned. The vibration caused by these factors can have
exactly the same character as those arising from poor quality bearings [Johansson,
BI3J200]. Bearing distortions may take place when force fitting them into' their housings.
This takes place when the outer race of the bearing is skewed at an angle to the housing
seating [Aini, 1990].
In earlier studies, misalignment was considered only because it imposed additional

stresses on the ball bearings. The stress distributions in a misaligned bearing were
obtained by different researchers [Harris, 1991]. When the research on bearing running
problems and early failure started, misalignment began to be identified as a major source
of bearing problems. For example a study by Barash [1968] into ball speed variation and
its effect on cage design identified bearing misalignment as the most common cause of
ball speed variation and cage failure.
An investigation into misalignment problems in ball bearings by Ellis [1970] showed that
misalignment of ball bearings can produce adverse conditions and shorten bearing life.
The cause of ball speed variations due to misalignment was found to be the contact angle
variations [Ellis, 1970]. When there is contact angle variation, each ball tries to rotate
about the bearing axis (cage rotation) at a different speed (see Fig.(2.29)). This is not a
satisfactory situation, some balls are compelled to slow down and others to speed up by
being forced to slide against the bearing rings. Each ball cycles through the total contact
angle variation for each relative revolution between the misaligned bearing ring and cage.
This effect was reported to be avoided by employing a plastic cage [Stolz, BBJ 228]. A
flexible cage will, as a result of deformation, avoid the kinematically produced stress and
permit the natural differences in ball velocity due to the variation in contact angles [Stolz,
BBJ 228]. Smooth running of bearings will be impaired by misalignment causing
additional vibrations and noise. The worst condition are likely to arise under combined
loads [Barash, 1968; Ellis, 1970]. The severity of the condition increases with the degree
of misalignment [Ellis, 1970]. When bearings failed due to misalignment are investigated,
66
a common running path can identify the cause of failure. Ellis [1970] gives an example of
a common running path on radial and angular contact ball bearings under axial load (see
Fig.(2.30)).
Contact angle
21° 25° 27 25° 21°
Fast balls
/
(C7
Ir Aer \
ring rotation
Resultant
reaction of ball
pocket loads
resisted by cage
centring surface

16° 11° 9° 11° 16°
Contact angle
Fig.(2.29) Possible cage loading resulting from contact angle variations

(frictional forces excluded) [Ellis, 1970]
Further theoretical research into the forces and moments produced by misalignment was
carried out by Andreason [1970]. Later BrAndlein [1971] investigated the effect of
misalignment on the life of ball bearings and calculated permissible misalignment for
different ball bearings.
Research into vibration problems caused by misalignment did not start until the 1980s
and today only a few papers are available on the subject. Meyer et al. [1980] theoretically
investigated the vibrations due to misalignment. Fig.(2.31) shows the predicted
frequency spectrum for a misaligned outer race [Meyer et al., 1980]. In the investigation,
the researchers assumed that the bearing has a misaligned rotating outer race and a
stationary inner race. Fig.(2.31) shows that the amplitudes at the BPF has not changed
but the amplitudes of the sidebands will be lower than the ball passage harmonic
components by an approximate factor of P12 where P is the degree of ball force variation
due to the defect [Meyer et al., 1980].
67

Case Wultgn- Rotating Itueniag path petters as revealed

anent ring by unwrapping Karla( slag
Inner Inner ring
11 Zero Outer ring I
Outer
Inner Inner ring —11i111.1111111.M.1.111...
9 Misaligned Outer ring —jimmit.r._

miter .n•n•=i-
Outer For angular contact bearings the ball path will generally be higher
up the side of the ring groove than for radial ball bearings. For
light loads the pattern may become discontinuous at positions A
beer ring A----7-A

Inner
10 Misaligned Outer ring

Inner
Outer For angular contact bearings the ball path will generally be higher
up the side of the ring groove than for radial ball bearings. For
light loads the pattern may become discontinuous at positions A
Fig.(2.30) Running path patterns on radial and angular contact ball bearings under axial
load [Ellis, 1970]
Experimental research into the cause and cure of bearing noise showed that
misalignment has a predominant effect on basic running errors (i.e., deviations in the path
of the ball) and hence vibration at the running speed [Wardle and Poon, 1983; Ellis,
1970]. Fig.(2.32) shows an example of the effect of a combination of misalignment and
thrust load on ball load distribution in a single row radial ball bearing. Wardle and Poon
[1983] point out that misalignment produces regions of high ball loads as seen in
Fig.(2.32) which leads to a secondary source of the BPF effect. Although it is always
small compared to the basic running errors, it occurs at high frequencies, i.e., at the
bearing's ball passage rate and its harmonics, where it may excite low order shaft
resonance.
RP
o LE P
AMPLITUDE. AMPLITUDE: —2 0
riTer — 11-7F-0
6 (te) ss sti sti
1 0 R • se 4 2E•eo
..LL.L... ANGULAR
nx•et
14 FREQUENCY
Kw — Y/ 1/4.1.61
OR 0R OR
Fig.(2.31) Frequency spectrum for misalignment [Meyer et aL, 1980]
68
Model end *ad Ihift

between Rey Centres
Me T. Plashbeeent
Bon Lee,
Confect Ante
Fig. (2.32) Example of ball load distnbution under thrust load and misalignment
(Wardle et at, 1983]

110 SOO ON INS WM MO
blew Speed • OW
Fig.(2.33) Resonance curve of axial vibration in the case of inner ring misalignment
(Gad et at, 1984b]
69

Gad et al. [1984b] introduced inner ring misalignment into their model. Fig.(2.33) shows
the effect of misalignment for different damping constants. The shaft-bearing system
with misalignment shows periodic motions. It can be seen from Fig.(2.33) that the axial
vibrations due to bearing inner ring misalignment is a forced vibration and exhibits
resonance behaviour. Fig.(2.34) shows transient and steady-state vibrations of the shaft-
ball bearing system with inner ring misalignment of 0.025 rad. at a speed of 850 r.p.m. and
damping constant of 100 Ns/m.
0.6
se
0.2
; -0.2
-0.6
110 -2
1 -1.6
-2.0
-2.2
a10-2
-2.0
-4.0
0.00 0.20 0.40 0.60 0.00 1.00 1.20 1.40 1.60 1.00 2.00
Tian t (sec.)
Fig.(2.34) Transient and steady-state motions in the case of inner ring misalignment
[Gad et aL, 198441
The harmonic axial and radial vibrations due to inner ring misalignment synchronise with
shaft revolution. This is expected since the inner ring is rotating at the speed of shaft.
2.7.2 Defects
Defects in bearings can cause severe vibrations. By monitoring the vibration

characteristics of a bearing, the surface roughness on the race or balls of ball bearings can
be estimated [Kanai et al., 1987]. The sources of these defects can be manufacturing
and assembling anomalies which was discussed in the previous section or external
70
effects. Some of the external sources of defects can be cited as wear of the contact
surfaces, a defect on the surface of mating ball and cage surfaces, debris and
contaminations in the contact etc. Defects can be local, i.e. point defect or discrete.
Discrete defect refers to a wider range of faults, e.g. scores, indentations, corrosion pits,
brinelling and contamination [Wardle and Poon, 1983]. Characteristically, discrete
defects produce impulsive vibrations and noise. This can be considered as roughness as
discussed earlier and will not be investigated In this thesis. In this section point defects
will be discussed. Point defects may occur due to fatigue or chattering marks.
Fatigue begins as a minute crack at a certain depth below the surface and this leads to
gradual flaking which sooner or later necessitates replacement of the bearing [Wallin,
1966]. If vibration monitoring is done correctly this crack can be identified long before it
comes to the surface since structural changes associated with fatigue phenomena and
cracks in the bearing components can affect the vibration spectrum even though they do
not affect the bearing geometry [Yhland and Johansson, 1970]. This is because of the
changes in the contact spring and damping characteristics of the bearing.
There are in fact at least three different but associated effects which occur when the
rolling elements roll over a damaged raceway surface, or when damage on one of the
rolling elements comes into contact with the raceway [Yhland and Johansson, 1970].
1. Characteristic changes in the vibration spectra at points on the boundaries of the

bearing arrangements.
2. Impacts which through wave propagation phenomena in bearing components and the
surrounding structure, take the form of transient oscillations at the boundaries of the
bearing arrangements.
3. More or less pronounced vibration peaks.
Fig (2.35) Vibration sequence in a damaged rolling bearing

(Yhland and Johansson, 19701
71
When a rolling element rolls over a damaged raceway surface, or when the damage on a
rolling element contact the raceway, the edges of the microscopic geometric changes in
the bearing caused by the damage give rise to sudden and large variations in the contact
pressure. The resulting sequence of dynamic behaviour can most closely be described
as an impact, or a series of impacts following each other closely (see Fig.(2.35)).
Yhland and Johansson [1970] investigated this dynamic behaviour and concluded that if
the damaged bearing component is a ring, the damage is rolled over at approximately
regular intervals. The frequency at which this occurs can easily be determined on the
basis of the geometry of the bearing and Its speed of rotation, and this is also true If the
damage is on a roller of a roller bearing. In the case of a damaged ball in a ball bearing,
there will be variable intervals during which the damaged point will periodically make
contact with the raceways and variable intervals during which there will be no such
contact. This due to the complex pattern of movements of the balls in a rotating ball
bearing.
If there is damage at more than one point in a bearing, as may occur if secondary damage
occurs fairly soon after the primary damage, several oscillation sequences will be
superimposed on each other. This more or less linear superposition gives a
correspondingly more complicated transducer signal, and this must be considered when
the signal is being studied.
Braun and Danter [1979] investigated a bearing with developing localised defects. The
vibration signature from each defect appeared with period T, for a constant rotational
speed, as in Fig.(2.36). Taylor [1980] argues that a machine with a defective bearing will
produce vibrations at least with five different period T1 , i.e., with frequency. They are
shaft speed, cage speed, ball passage speeds of the outer and inner race, and ball spin
frequency. Ball passage frequencies are generated as the balls pass over a defect on the
raceways.
k mipr-:-= fr-
Fig.(2.36) Typical bearing signatureiBraun and Danter, 1979]
72

Envel Si al
• 387.563 :
2
1. 310.051 -
: 232.538:
• 155. 025 -
>
... 77.513 -
-i 0.000 I
0.09000 0.01562 0.03125 0.04687
tie! Es]
Spectrum of Envelop.?
14.122
11.23? -
8.473
_
5.649 B
_ V.'
2.824
0.000 644i16"itill
0.0 400.0 800.0 1200.0 l600. 0 201XL 0

frequency (Hz)
Fig.(2.3 Envelope signal and related M1S spectrum of damaged bearing

(larakurt, 1989)
Igarashi et al [1982; 1983; 1985] carried out exhaustive experimental studies of

indentations caused by the debris in the lubricant on the raceway surfaces and the balls
with a view to establish a procedure for diagnosing rolling bearing defects from their
vibration and sound. lgarashi and Hamada [1982] used an anderometer, which is an
instrument that is designed to make a measurement of radial vibration at one point on the
outer ring of a rolling bearing while its inner ring is rotating, whereas Igarashi and Yabe
[1983] used a sound measuring instrument to investigate the information obtained from
the bearing with a single defect on the race surface of the inner or outer ring or on the ball
surface. It was concluded that the main peaks in the bearing sound frequency spectrum
coincide with those in the vibration acceleration frequency spectrum. McFadden and
Smith [1984; 1985] studied single and multiple point defects in a rolling element bearing.
They showed that in the case of inner case defect, for example, the vibration spectrum
will accommodate peaks at the shaft speed, cage speed, their harmonics and their
combinations (see Fig.(2.38)). Later Igarashi and Kato [1985] and Karakutt [1989] made a
similar study on a bearing with multiple defects. Fig.(2.37) shows the envelope signal and
related rms spectrum of a bearing with a defected inner race. The results of Karakurt
[1989] are very similar to those of McFadden and Smith [1984; 1985]. All studies show
that useful information about bearings can be obtained from their vibration spectrum.
73
4-4
4.4
21, 4.24
4-24
4/.
Fig.(2.38) Envelope spectrum for a bearing with early damage

(McFadden and Smith, 1985]
Zllfi..Ehattlltall!dIILE1‘Iern:LaL12aOjleIa
-
There are three commonly applied damping systems in the problem of the shaft-ball
bearing arrangements. These are arbitrary damping for mathematical convenience, oil film
damping and viscoelastic damping.
Generally when there is need for external damping in mathematical investigations, an

arbitrary (convenient) damping are introduced to the system. This is particularly useful
when the effects of the natural frequency Is to be avoided in the problem. Gad et al.
[1983; 1984a; 1984b], Gupta et al. [1975; 1979a;1979b; 1979c; 1979d], Rahnejat and
Gohar [1985], Aini et al. [1990], Franco et al. [1992], AktOrk et al. [1992] employed
arbitrary damping in their theoretical investigations.
Rahnejat [1984] and Rahnejat and Gohar [1985] investigated the effect of the
elastrohydrodynamic oil films between the ball to race ways contacts on the vibration
characteristics of the shaft-bearing systems. They concluded that the effect of internal oil
film damping is slight and can be omitted from a vibration point of view.
Fig.(2.39) shows the cyclic oil film behaviour in the contact of rolling members. It is
composed of two unequal portions, of similar shape, with a common point A. The larger
segment represents the film behaviour when the ball is in the least loaded region of the
bearings, where there is a significant hydrodynamic contribution. The smaller segment,
to the left of A, corresponds to the more highly loaded region which does not necessarily
apply always to the top or the bottom region of the shaft contact, as its centre itself is
74
oscillating. Rahnejat and Gohar [1985] noted that as the loops on the RHS of the point
A enclose larger areas than those on the LHS, there is more energy dissipation
associated with them, implying that squeeze film damping action is more significant in the
region of thicker films. Since damping in the oil film is caused by squeeze velocity,
relatively larger damping is obtained with thicker oil films. However, thinner oil films means
less speed and hence low damping. Therefore in Fig.(2.39) h the loaded zone there is
less damping due to the thin oil film. This is in agreement with the findings of Dareing and
Johnson [1975] who argue that lubrication film damping is greatest under conditions of
separation and this indicates that film damping is generated primarily by the squeeze film
mechanism. However, although Dareing and Johnson [1975] argue that in any case oil
film damping will increase the overall damping at least 30% and it is more effective under
conditions of severe vibration, Rahnejat and Gohar [1985] found it Insignificant in
vibration problems. This is perhaps because of the heavy load they employed in their
model (500 kg of shaft mass) since fluid film damping may be more important and effective
under light loading conditions where Hertz compression is small [Dareing and Johnson,
1975; Mehdigholi et al., 1990].
0.15
0•10
0•05
rnm/s
- 0.05
- 0.10-
- 0.15 I I 1
I-0 1 . 03 1.10 1 • 13 1-20
hat
mm
Fig.(2.39) Steady-state knit cycle (Rahnejat and Gohar, 19851
75
A dynamic analysis of the transient ball motion in an angular contact ball bearing,
operating under elastrohydrodynamic traction was formulated by Gupta [1975]. He
assumed a cage-free bearing for his analysis, thus eliminating the cage influence on the
stabilising/destabilising of the balls' motion. Gupta [1975] concluded that, his analysis
presented a design tool, capable of simulation the dynamics of bearings subjected to
skidding, due to the acceleration of races in lubricated contacts.
Later Gupta [1979c] and Aini [1990] reconfirmed that internal lubricant film damping in ball
bearings was insignificant from the vibration point of view but reduces the noise caused
by high frequency vibrations as suggested by Sayles and Poon [1981] and Wardle and
Poon [1983].
Elastomers are employed as external dampers, particularly in the form of 0-rings.

Gustafsson and Tallian et al. [1963] suggested use of elastomer external dampers for
shaft-ball bearing systems. However, elastomers have not been widely employed until
more recently in practical applications since many engineers do not regard a moulded
rubber component to be of sufficiently consistent dimensional accuracy for use in
precision engineering and the information available on elastomers did not contain the
required design data in terms of dynamic properties of elastomers and neither is this
information easy to obtain experimentally, particularly in the case of 0-rings [Powell and
Tempest, 1968]. Elastomers have been employed in gas bearings running at high
speeds with successful results [Powell and Tempest, 1968; Kazimierski and Jarzecki,
1979]. Hedman [1987] employed elastomer buttons as external dampers. His simulation
also shows some reduction in the vibration amplitudes. Nevertheless, since he
employed a relatively less sophisticated model, his results are not very reliable. There
may be found other kinds of external dampers employed in the shaft-ball bearing
systems. For example Gunter [1970] investigated the influence of damped externally
and linear flexibly mounted rolling element bearings on dynamic shaft unbalance
response. He showed that the untoward effects of the shaft unbalance can be
dramatically reduced by proper design of bearing support characteristics.
Olsson [1986] analysed the effect of introducing an external lubricant squeeze film
around the outer race as a damping element in an angular contact ball bearing assembly.
Although the simulation did not suggest really successful application of this approach,
the results showed some reduction in the vibration amplitudes at the natural frequencies.
In this thesis, different types of elastomer theoretical models are studied as alternative
external damping elements for the shaft-ball bearing systems. Therefore there is need
for further investigation into the literature on their dynamic and mechanical properties, in
particular their stiffness and damping characteristics. This will be done in the next section.
76
2.9 Elastomers as External Dampers

Dampers are important in the control of shaft vibration of rotating equipment which must
operate through one or more critical speeds. Although elastomers have been used In
different applications, mainly in vibration and shock Isolation for a long time, their dynamic
properties had remained unresearched for many years. As today elastomers are used in
the design of increasingly complex arrangements, accurate knowledge and clear
definitions of elastomer dynamic properties (stiffness and damping) are becoming more
important. For example modem turbomachines are running at very high speeds (e.g. 120
000 rpm) and their size Is reducing. These applications often lead to operation above
one or more flexible rotor critical speeds [Taylor and Fehr, 1982]. For these kinds of
applications it is very important to know the elastomer dynamic properties so that system
critical speeds can be avoided or their corresponding amplitudes reduced. Therefore,
there appears to be a need for a systematic study of the nonlinear dynamic behaviour of
elastomers, and subsequent development and/or evaluation of an appropriate material
model [Generiwala and Rotz, 1987].
Almost four decades ago the first papers on the dynamic applications of elastomers
appeared. Some of the literature available deals only with experiments which show how to
derive the dynamic stiffness and damping. A paper by Smalley et al [1975] investigates a
number of methods and apparatuses which have been used to measure material dynamic
properties for engineering application. More recently Kluesener [1986] tested two types
of elastomers using five different methods and reported some significant differences
between them. The literature review, made two decades ago contains many experimental
papers on the subject but no attempt has been made to investigate the problem
theoretically, since the subject is complicated and needs a computer. Snowdon [1963]
tried to handle the problem analytically without any computation in 1963. He derived
some equations that predict the response to vibration of rubberlike materials and
structures but the results were not very accurate. In the last decade theoretical
investigations on the subject have increased and some relevant literature is available.
Sommer and Meyer [1973] gave the brief summary of the changes in the spring stiffness
and damping coefficients produced by variations in rubbers, fillers, oils, level of cross-
linking and processing factors such as mixing, curing, and storage. Apart from these, the
dynamic properties of elastomers are strongly dependent upon many factors; the most
common of which are: frequency of the force exciting it, temperature, amplitude of
vibrations and geometry of the dampers.
77
In the 19705 some important research was carried out on the calculation and prediction of
the dynamic properties of elastomers by the National Aeronautics and Space
Administration (NASA). Five contractor reports were produced during this research
programme [Chiang at al., 1972; Gupta et al. 1974; Smalley and Tessarzk, 1975; Darlow
and Smalley, 1977; Smalley et al., 1977). In the first report [Chiang et al., 19723 the basic
methods are studied and a rig for the experiments was built as seen in Fig.(2.40)
Fig. (2.40) Schematic of base excitation resonant mass test rig (Chiang eta!., 19721
Using this test rig Gupta et al [19743 showed that the variation of the complex dynamic
stiffness with the frequency can be expressed in the form of a power function of the form
shown below:
(2.11)
where A and B are constants. Gupta et al. [19743 also showed that this data can
successfully be presented with Voigt or Maxwell mechanical models (see Appendix 6).
Fig.(2.41) shows the experimental data and the mechanical model curve fit obtained for it.
It is seen that most of the stiffness is determined by the static spring, and that the
stiffness and damping behaviour of the model follows reasonably closely the mean level
of the measured stiffness and damping values.
78

1 1
11-1-0111T111111C111 SPC1=1/ 11C/1- II . 1:1111 4-- 30 .
0
MAT ellaPil- so 1 .1,1abodlaso VWX6 7 MEMEL
LOPUIINS • Cs .Irvoma 1 011/1.) a Oh ..sonei.
PrIA4C It is .111
1 ....Were I.- 1.'16E•1
I
Pl tCl. M/R31 • 1 . NW 1 g
I; 1. el IC + 1 '4 . WIC *i
FIC1414 TEMP a II .111 C • C. 1. 11111 ( • 1 1.16111E+1
I SW
,
I Wi*
cr
14 Ns (441) it.P421) 6/11b
ORM 3C 1/ I/11 •
1.15 C-11 1 . 1.1111 1- Ise%
Weil/04T S 1. IN +1 •.1111111 C44
H1
H2
FREQUEN-CY <HZ>
Fig.(2.41) Typical correlation of experimental data with a double Voigt model

(Gupta eta!., 1974)
Smalley and Tessarzik [1975] investigated the effect of temperature, dissipation level and
geometry using the test rig established by Chiang et al. [1972] (see Fig.(2.40)). Smalley
and Tessarzik [1975] argue that if real and imaginary parts of complex stiffness, K (co) and
K2 (co) respectively, are calculated from the following equations Equ.(2.12) and
Equ.(2.13), an accurate curve fitting can be achieved. Note that Equ.(2.12) and
Equ.(2.13) include dimension factors and hence are prediction formulae where G is the
,
real part and G is the imaginary part of the complex shear modulus G- . is the real part
re
and /3 is the imaginary part of the shape factor.
702 N
Ki (co)= 3G'(w)-
4h 11+ 13'(w)sl (2.12)
K2(co)= 3G"(w)V1-[1+ fil0.0s2]

(2.13)
79
The experiment done by Smalley&Tessarzik[1975] and Darlow and Smalley [1977]

. II I
showed that if G ,G , p and p" in Equs.(2.12 and 2.13) are expressed at constant
temperature of 32 degrees Celsius, as below, the error for dynamic stiffness would be
less than 50% and with damping the error would be less than 20%.
G'(w)= 3.686 x10 6 Cele" N i M2 (2.14)
G1(0)= 8.333 x10 6 Co -CU 277 N / m2 (2.15)
Inco)= 12.33 co-t2" (2.16)
fr(o))= L726 CO "2" (2.17)
Darlow and Smalley [1977] established another rig for rotational loading in order to
compare their results with the previous findings and validate the previous results with the
closer simulation of the real life working conditions of elastomer dampers since most of
the elastomer damper work in the designs where the loads are rotational (see Chapter 5
for these experimental arrangements). Their findings confirmed the previous results.
Smalley et al. [1975] tested elastomer 0-rings of three different materials for dynamic
properties with the rig established by Chiang et al. [1972] with small alteration. They
showed that the dynamic properties of elastomer 0-rings 4-15 times the corresponding
static properties. A similar study done by Green and Etsion [1986] also confirmed the
findings of this report.
Taking these reports as a basis, Tem et al. [1979a, 1979b] designed elastomer dampers
for a high-speed flexible shaft. They showed that when the dynamic properties are
accordingly arranged, smooth running can be obtained with elastomer external dampers.
These two papers have a very similar arrangement to the assumed grinding spindle model
studied in this thesis. Therefore theoretical results obtained may be compared more
readily with the experimental data obtained from these tests.
In these studies the dynamic properties of elastomers were obtained in the form of the
real and imaginary parts of the complex stiffness. The real part represents the stiffness
and the imaginary part represents the damping factor for the system. All reports and
papers report that elastomers can be used as external dampers and they will be effective
in damping out the untoward effects of natural frequencies and reducing forced vibration
amplitudes if their dynamic properties are known and their application is made correctly
80
Different ways of representation of obtained data were also studied by several

researchers. Gupta et al. [1974], Dahl and Rice [1984], Fersht et al. [1986] employed
mechanical models whereas Smalley and Tessarzik [1975], Dariow and Smalley [1977],
Smalley et al. [1977] applied direct fits to the experimental data. Because the mechanical
model is more convenient for vibration studies of the shaft-ball bearing system, a modified
Voigt model is used for the presentation of the data In this thesis.
2.10 Background of the Project

Dr. R. Gohar started investigating the elastrohydrodynamic oil film behaviour under point
contact of a vibrating bearing In the early 1980s. Rahnejat [1984] investigated the
influence of vibrations on the oil film In the concentrated contacts under the supervision
of Dr. Gohar. Rahnejat modelled gears and rolling element bearings as bodies in
concentrated contact both with and without a lubricant film. During this work Rahnejat
also mathematically modelled a horizontal milling machine supported by a pair of deep
groove ball bearings. He studied the BPF and showed that it appears as a result of
contact spring nonlinearity of the ball bearing. However, he could only observe it when
the imperfections are present in the system. This may be because of his high preload
and heavy mass, existence of which will considerably reduce the effect of the BPF
[AktOrk et al., 1992]. Rahnejat [1984] also showed that damping of the oil film between
the ball to raceway contacts comes mainly from the squeeze effect as discussed earlier
(Section 2.9) and is negligible. Later, a dynamic simulation of a pair of deep groove ball
bearings supporting a shaft, Incorporating non-linear elastic behaviour of ball to race
contacts, contact rolling/squeezing film lubrication and inclusion of surface defects of
rolling members as well as eccentric rotation of inner raceway of radial ball bearings was
presented by Rahnejat and Gohar [1985]. Major frequencies relating to applied load-
stiffness characteristics of the bearings as well as the BPF, cage set rotational speed,
wave passage frequency and some modulation effects between these were observed.
The model was also applied to lubricated contact vibration problems in gears under
hydrodynamic and elastrohydrodynamic conditions [Rahnejat, 1984].
Meanwhile Matsubara [1982] studied the vibrations of a heavy elastic shaft supported by
a pair of radial contact ball bearings under the supervision of Dr. Gohar. Matsubara
modelled ball bearings with linear rotating springs as linearised load deflection curves
over a narrow region employing a piece-wise spring technique. Later, the results of this
study were published in a paper by Matsubara et al. [1988].
The work presented by Rahnejat [1984] was extended by Eisenbeis [1985] under the
supervision of Dr. Gohar. Eisenbeis substituted deep groove bearings support by a pair
of angular contact ball bearings. He presented the effect of off-centre rotation, off-sized
81
balls and waviness features on the vibration characteristics of the system as well as the
effect of changing the number of balls and preload. However, he could not observe the
effect of the BPF for a geometrically perfect bearing. This study was further extended by
AktOrk [1988] under the supervision of Dr. Gohar. AktOrk showed some Instabilities in the
computer simulation of the previous researchers. This led to a new simulation programme
which is used in this thesis. Some of the findings of this research were presented in a
paper by AidUric at al. [1992].
Further work in gear simulation involved the contact of wavy gears which was undertaken
by Mehdigholi [1987] under the supervision of Dr. Gohar. He simulated the disc machine
geometry reported by Dareing and Johnson [1975], where one disc was crowned and
the other was corrugated to present a wavy surface. The result of the theoretical
simulation and the experimental results confirmed closely to each other, except that the
bandwidth of frequencies in the vicinity of the resonance is larger with the theoretical
results, because of the failure in the oil film model under contact separation conditions.
The result of this research was published in a paper by Mehdigholi et al. [1990].
Franco [1990] also studied radial contact ball bearings under the supervision of Dr. Gohar.
He linearised the load deflection curve with Taylor's expansion and studied multiple
defects of ball bearings. The anomalies he included in his simulation were off-sized balls,
waviness, out-of-balance shaft centre. The results of this work was also presented in a
paper by Franco et al. [1992].
Aini [1990] studied a perfect precision grinding spindle supported by a pair of angular
contact ball bearings theoretically and experimentally. He also included a8i1 film into his
calculations and concluded that the presence of the oil film along the line of contacts do
not significantly after the position of the major modes of the system. This is also in
agreement with the results of Rahnejat [1984], Rahnejat and Gohar [1985], Mehdigholi
[1987] and Mehdigholi et al. [1990].
Gohar et al. also employed different external damping systems to the shaft ball bearing
systems. Olsson [1986] employed external oil squeeze film dampers and later Hedman
[1987] employed elastomers as external dampers. Although his results showed some
reductions in the vibration amplitudes, overall results were not satisfactory due to the
difficulty in modelling of elastomers.
Later Larsson [1989] also employed elastomers as external dampers to aerostatic

bearings and compared them with piezo-electric and oil film dampers. Since both
Hedman [1987] and Larsson [1989] suffered from the lack of much solid information on
the dynamic properties of elastomers, therefore, Grassano [1991] studied dynamic
82
characteristics of elastomer dampers for cylindrical buttons, ring cartridges and 0-rings
under the supervision of Dr. Gohar. He showed that close predictions for different
geometries of elastomers were possible.
Closure
The dynamics of the shaft-ball bearing system was reviewed in this chapter. Due to the
broadness of the subject and a plethora research papers available, only certain topics
were reviewed in depth. Although there are many research papers quoted, there are
some topics that have been only partially investigated. As was given in Section 2.10, the
research done by Gohar et al. will be consolidated and the dynamic characteristics of
shaft-ball bearing systems will be investigated in this thesis. This will also include the
study of elastomers as external dampers to shaft-ball bearing systems. Therefore the
following chapters 3, 4 and 5 will be outlining the theory involved in this fascinating
subject.
83
CHAPTER 3
MECHANICS OF BALL BEARINGS
3.1 Introduction
In the second chapter the previous on ball bearings research was explained. In this and
the next two chapters the theories for the forthcoming results and discussion will be
established. In this chapter the main concern are the vibrations of ball bearings.
However, vibration theory alone cannot handle the problem of the shaft-bearing system
since the elasticity of bearings has to be defined by contact stress relations and
geometrical considerations, hence the subject is very broad. Therefore there is a need to
be involved in contact mechanics, kinematics, dynamics, numerical analysis and
vibrations. Furthermore, the theory of vibrations should be placed in its appropriate
position amongst others for correct handling of the problem.
The study of vibration is concerned with the oscillatory motion of mechanical systems and
the dynamic conditions related thereto. This motion may be in a regular form or it may be
irregular i.e., of a rather random nature. A vibratory system, in general, includes an
element for storing potential energy, an element for storing kinetic energy, and an
element by which energy is gradually lost.
Whilst for small amplitude linear vibrations, superposition techniques may be used with
the mathematical solutions well developed, for nonlinear vibrations, the superposition
principle is not valid, and suitable analytical techniques are less well known. Linear
analysis is not always sufficient to describe the behaviour of a shaft-ball bearing assembly,
since it is nonlinear and causing totally unexpected behaviour that is not predicted or
even hinted at by linear vibration theory. Hence some knowledge of non-linear vibration
theory is also desirable.
84
Chapter 3 Mechanics of Ball Bearings
With nonlinear behaviour, the most difficult problem is usually to identify the source of
nonlinearity, which is generally caused either by nonlinear springs or by nonlinear
damping elements or both. In our case both are present. However, deflection dependent
stiffness is often the more dominant cause of the nonlinearity.
The nonlinear damping element present in ball bearings can be caused by friction
between the rolling/sliding elements, squeeze film lubrication or by an external damper.
The first two are found to be insignificant [Ragulskis, 1975; Rahnejat, 1984; Rahnejat and
Gohar, 1985] and the latter will be investigated in Chapter 5.
In order to predict the system vibrations correctly, it is necessary to analyse the dynamic
behaviour of the shaft bearing systems using a mathematical model. The elastic
properties of the shaft and the housing can be described reliably by established
mathematical models. The bearings are however modelled less reliably. This is generally
because the vibrations associated with ball bearings are very complex. Even with
simplifications the theoretical model produced is difficult to analyse.
Two different approaches are employed in investigating the vibration characteristics of

the
shaft-bearing assemblies. The first is called superposition method in which the vibration
of shaft and bearings are investigated separately while considering their interaction later.
The second method is to investigate the system as a whole.
Ball bearings cannot be isolated from the shaft since Its movement will cause deflection of
balls resulting in forces that will contributed to its dynamic behaviour. Therefore the
system should be modelled as a whole and as a result the contribution of bearing
vibrations is included in overall vibrations interactively. In this method the shaft acts as a
mass and the bearing balls act as nonlinear springs.
After the forces due to the deflection of a ball is found and the total force on a bearing is
calculated, the equations of motion for shaft centre vibrations are developed using these
forces, as well as other forces which will be defined. Finally the equations of motion will
be defined and their numerical solution will be discussed.
3.2. Elastic Contact Modelling
3.2.1 Contact Mechanics
The solution to the elastic problem for the concentrated contact of two elements was
established by Hertz [1896]. Some general knowledge of the stress-strain relationship of
contact mechanics is needed to investigate the problem.
85
If two elastic bodies in concentrated contact are considered as shown in Figure (4.1), the
following definitions may be employed:
The curvature is generally defined as inversely proportional of the radius of curvature and
given as
(3.1)
where curvature is positive for surfaces whose centres of curvature are on opposite sides
of the tangent plane and is negative for surfaces whose centres of curvature are on the
same side with the tangent plane.
Fig. (3.1) Geometry of contacting bodies
The curvature sum is defined as (see Fig.(3.1)):
EP = P Al + PA2 + Pal + Pin (3.2)
86
where A and B refers to the body and index 1 and 2 refers to the planes. The curvature
difference is defined as (see Fig.(3.1)):
F (p) = (P A1 PA2) (Pei — Pe2)

(3.3)
The curvature difference for a Hertzian contact is found to be also a function of the
elliptical parameters as follows (for more information see [Harris, 1991]):
(0+1)3 — 2N
F (p) — ( 2
Pc —1)3
(3.4)
where K is the elliptical eccentricity parameter and tt and S are complete elliptic
integrals of the first and second kind respectively.
By assuming values of the elliptical eccentricity parameter, a table of K versus F(p) can
be obtained and then it can be shown that the local force and deflection relationship of
two bodies may be written as follows [Harris, 1991]:
o
= w i-Lil121-v221)3Ep
2EpL E1 E2 2
(3.5)
where (5* is dimensionless deflection and is expressed as:
1
r• 211 IV )3
= --
7r tc2 2 (3.6)
6** is given as a function of F(p) in a table and is plotted on three figures in [Harris, 1991].
In order to write the computer programme the following 4 th order polynomial was fitted
along the curve represented by Equ.(3.6) [AktOrk,1988]:
S e = —327.6145 + 1883.338(F(p)) —3798.1121 (F (p))2
+ 3269.6154 (F (p))3 —1026.96 (F (p))4

(3.7)
87
Chapter3 Mechanics of Ball Bearings

0.90 0.92 0.94 0.96 0.98 1.00
Curvature Difference (F( p))
Fig.(3.2) Dimensionless Contact Factors
For angular contact ball bearings, the curvature difference is generally between 0.9 and
0.99. Therefore in order to achieve maximum accuracy the curve was fitted over this
range of data as seen in Fig.(3.2) and each time the curvature difference is checked if It is
within the limits. If it is out of range then the other two curves fitted over the data between
0 and 0.9 and 0.99 and 1 are employed as it is appropriate.
The curve fitting over 0 to 0.9 is:
(5 * = 0.99878+0.097536(F (p))— 0.94627 (F (p))2
+1.6462 (F (p))3 — 1.2784 (F (p))4 (3.8)
The curve fitting over 0.99 to 1.0 is:
(5. = 397890.-1205000 (F (p)) + 1216500 (F (p))2 — 409370 (F (p))3 (3.9)
88
Chapter 3 Mechanics of BailBearings
3.2.2 Load-deflection Relationship and Calculation of the

Stiffness Coefficient
From Equ.(3.5) it can be seen that for a given ball-raceway point contact loading there is a
relation between the deflection and the load. This can be expressed as:
2
— (3.10)
Inverting Equ.(3.9) and expressing it in equation format yields to [Harris, 1991]:
3
W = K82 (3.11)
If Equ. (3.5) is written in the form of Equ.(3.11) for the same material properties of the two
contacting bodies:
(E
2"11 u2) 116;
=[ 3 (EPY
5.
) (3.12)
Hence from Equ.(3.12) and the similarity of Equ.(3.11) and Equ.(3.12), the stiffness
coefficient can be given as:
K=
(
2
E
215 L1- v ) ( 1
—T. ___
y 3
3 ( Ep)i 3. (3.13)
where v is Poisson's ratio
For the i th ball, there are two contact surfaces: outer raceway/ball and baMnner raceway.
Hence, the total deflection at the i th ball is summation of these two local relative
movement:
8 11 = 8i + 80 (3.14)
2
Considering 5 = (f)3 and W.= Wi + w. and using Equ.(3.9):

K)
89
1
K. =[ (
(KJ (IT)
(3.15)
This is the total proportionality of deflection and is used in this thesis. In the previous
researches [Eisenbeis, 1985; Akturt, 1988; Aini, 1990] the stiffness factor is assumed to
be constant as the change in it due to contact angle variation is relatively small. However,
Fig. (3.3) shows that for certain contact angles the change becomes quite significant. In
this research, therefore, a subroutine was written in order to determine the stiffness factor
for each new contact angle.
7
IIIMIrvAi
11/111111
0 6
1111111111111111111011111
11111111111111111/1111111111
2
03 11111111111111111111111111
111111111M1111111111111111
1111Prid111111111111111111111
0
P1,2111111111111 11111

15 30 45 60
Contact Angle (DEG.)
Fig.(3.3) The effect of changing contact angle on the stiffness factor
3.3 Bearing Geometry
3.3.1 Calculation of the Preload Contact Angle
When the bearing is axially preloaded, the preload will cause the same amount of
deflection at each ball because it is uniformly supported by all the balls. In this case,
90
considering the Hertzian contact relationship, from Fig.(3.4), the the relation between the
preload and the preloaded contact angle is given as:
„
Pita = m (8 0)isin (ap)
(3.16)
where 50 is the preloaded contact deflection and given as
450 = B d iccas.)__ 1)
6 COS (ap)
(3.17)
Initial Position
of the Inner ring
The position
of the Inner ring
Fig.(3.4) Geometry of Angular Contact Ball Bearing
Substituting Equ.(3.17) into Equ.(3.16):
3
2
PRa = mK.[B d b (?=)s-1119.1 — sin(a p)
COS(CC
(3.18)
Equ.(3.18) includes sine and cosine simultaneously. Therefore, the preloaded contact
angle ap is solved by trial and error.
91
Chapter3 Mechanics of BailBearings
The initial axial deflection zo due to the preload can then be calculated from (see
Fig (3.4))
sin(a 0) sin(a, — a ))
zn = —+ D a,. — =
sin(c cp) sm(a) hi cos(a•)
(3.19)
3.3.2 Calculation of Deflection
The objective is to calculate the net force on a bearings due to the displacement of shaft
centre since this force can then be used in the equations of motion in order to find the
movement of the shaft centre.
In order to calculate the total force, the deflection at the i th ball in Fig.(3.5) will be
calculated first and this will be used in the calculation of the total force.
As seen in Fig.(3.5) the ball is rotating between inner and outer rings. During this rotation
the ball is continuously in contact with different points in circular grooves in each race. In
the initial position, without any preload, the loci of raceway groove centres of curvature will
produce circles as shown in Fig.(3.6).
If a ball is compressed by a force, for example the weight of the shaft, since the centres of
the curvature of the raceway grooves are fixed with respect to the corresponding
raceway, the distance between the centres is increased by the amount of the normal
approach between the raceways. Fig.(3.6) shows a ball bearing prior to the application of
load and after preloading. It can be determined from Fig.(3.6) that the locus of the
centres of the inner ring raceway groove curvature radii is expressed by:
91i = cl'—" +(r — Ncos a

2 2 (3.20)
91.= B db cos a (3.21)
Or
9te = Am +Er + Ncosoc,

2 2 (3.22)
92
Fig.(3.5) Angular Contact Ball Beating
If the outer ring of the bearing Is considered fixed in space as load is applied to the
bearing, then the inner ring will be displaced. It can be shown that the distance between
the centres of curvature of the inner and outer ring raceway grooves at ith ball before and
after the preload is equal to the contact deflection and can be found from Fig.(3.7) as
(also see Fig.(3.8)):
tie . B db (cos a0 i)
cos ap
(3.23)
The preloaded contact angle is:
B db sin (a0) + zo
tan(a )—
' B db cos(a 0)
(3.24)
The actual deflections caused by vibration in the x, y and z directions will be dealt with
from this point onwards. Fig.(3.9) describes the situation after this three dimensional
movement of the inner ring centre has taken place. The deflections along the x and y
axes can always be combined as a single radial deflection.
93
Fig. (3.6) Loci of raceway groove curvature radii centres before loading
The radial deflection for the i th ball (see Fig.(3.9)):
Sr = xcos(0,)+ysin(0,) (3.25)
z iz-
Fig.(3.7) Loci after the displacement due to the preload
94
Fig.(3.8) View normal to a plane at the angle ei from the x axis (see Fig. (3.77)
Fig.(3.9) Loci groove centres of curvature after three dimensional displacement of the
inner ring centre (x, y and z)
From Fig.(3.10) the deflection:
1
x 2ji
8:=[(B db sin(a0)+ ;I + z) +t13 db cos(a 0) + , ) — B db
(3.26)
95
and the contact angle:
Bdb sin(a0)+ zo + z
tan(al —
B di, cos (a0)+ (3.27)
Fig.(3.10) View normal to the plane at the angle 9; from the x axis (see Fig.(3.9))
Deflection of the inner ring centre due to rocking motion of the shaft may also be
considered in terms of deflections in three axes. For the LHS bearing the radial and axial
deflections will be (see Fig.(3.10)):
3' = a, sin (0,)cos(0,) + cos(03 sin (Osin(01) (3.28)
z'= a1 (1— cos(0))cos(84)+ al (1— cos(Mcos(0 ,)sin(9 ,) (3.29)
Since the i th ball centre is at a distance, R, from the bearing rotation axis this will cause
additional axial and radial displacements. Assuming small and w:
8". —R(1— cos(Ocos(93— R(1— cos(")) sin (0,) (3.30)

z"= CRcos(0)— vr1Rsin(91) (3.31)
96

a sin wi
Fig. (3.11) Displacement of the bearing centre due to rocking motion
Fig.(3.12) Effect of rocking angles
97
me
91i
8r
8"
Fig.(3.13) The final displacement as seen from the plane at the angle
The resulting deflections for the i th ball of the left hand side bearing will be:
1
2
[Bdb cos(a 0 ) + xcos(60+ ysin(0i)
(40cos(0 1 )+ ajtif3cos(0) sin (0
—R(1— cos(Mcos(60 — R(1— cos(vi)) sin (0)1
2
= —B cl,
[Bdb sin (a o ) + zo — z +a,(1— cos(0))cos(0) 2
+ +a1 (1— cos( vfMcos(03sin (60 + (0 )Rcos(0 ,)

_ —fri)Rsin(B,)
(3.32)
98
The contact angle for the I th ball of the LHS bearing is:
[B db sin(ao) + zo — z + a1 (1— cos(03)cos(60 \

+a1(1—cos(v/iNos(03sin(03+(03Rcos(e1)
—WRsin(0,)
tanRai)vol=[ <
Bdb cos(a0)+ xcos(ej+ ysin(0a)
a1 (03cos(0,)+ a1 (yr1)cos(0,) sin (9,)
—R(1— cos(03)cos(Bj— R(1— cos( lifi)) sin(OiL
(3.33)
Similar expressions can be derived for the RHS bearing. The deflections and contact
angles for the RHS bearing are
2 2
B db cos(cco)+ xcos(0,)+ y sin(0
—4(0)cos(0+ /21 vf1 )cos(03 sin (60
+R(1— cos(0j)cos(0j+ R(1— cos( y,)) sin (9,)
[ 1
(3.),„= — B db
[B db sin(a0)+ zo + z + b,(1— cos(Mcos(0 4) 2

+ +Fig' (1 — COS( Vi)) COS(C)Sin (0) — (OR COS (ej
_ +(iiii )R Sin (93
(3.34)
Bdb sin(ao) + zo + z+ b1 (1— cos(06))cos(0,) ]

+61 (1— cos( Mcos(0 jsin(e ,) —(40R cos(9 j
+0 irjR sin(e i)
[
tan[(aL] —i
Bd b cos(a0)+ xcos(9 j+ y sin(0 3
—b,(0)c-os(C oi) + bi (vr i)cos(0 ) sin(0)
+R(1—cos(03)cos(60+ R(1— cos()) sin(9)
(3.35)
99
3.3.3. Calculation of Total Contact Force on a Bearing

The deflection for the i th ball was calculated in the previous section. The force due to
this deflection for a single ball can be found and then the total force in the bearing can be
calculated as this ball rotates round the inner ring. In the model, the contacts of balls to
the inner and outer races are represented by nonlinear contact springs, i.e. balls act as
massless springs. The elastic model of the bearing is represented by Fig.(3.14). Any
deflections in any direction will produce a compressive force in that direction. Therefore,
the first step in calculating the total force is to find the deflection of each ball due to the
movement of the shaft centre.
Fig. (3.14). Nonlinear Elastic Model of the Beating
If the force on the i th ball in Fig.(3.5) is calculated, the reference axes should be set and
the total deflection and hence forces with respect to these axes should be calculated.
The deflection for each ball can be calculated as described in the previous section (see
Section 3.3.2). Having calculated the deflection for the I th ball in its contact direction,
the force in the same direction (Fi ) can easily be found. This force can be split into two
components in the radial (Fri ) and axial (Fm ) directions for angular contact ball bearings.
Hence the force in the X direction is:
Fx . = Fncosei
(3.36)
where 8 . is the angle between the X axis and the axis of the i th ball and this angle is
combination of different angles as shown in Fig.(3.15).
100
Fig.(3.15) The Reference axes
It is difficult to find the angle Oi since the balls are also rotating round the inner ring as time
passes. Therefore, three sets of reference axes are described as shown in Fig.(3.15).
The first one (X, Yand Z ) is fixed in space and the X axis is vertically downwards in the
gravitational force direction. The second set of the axes (u, v and w ) can be defined
arbitrarily in the space. It is considered stationary with the cage turning relative to it. In our
case the u axis has an off-set angle of v anti-clock-wise difference to the X axis. But in
this thesis v is assumed to be equivalent to zero unless otherwise stated. The last set
(x, y and z ) are set on the shaft and rotating at the speed of one of the balls without any
disturbance from the rotations of the ball around its own axes. Under normal conditions all
balls are rotating in a cage at the cage speed around the inner ring and this set of axes is
also rotating at the cage speed. Hence this set of axes is apart from the second reference
axes at the angle of qt. The ball calculated (i th baU) is, now, found easily since it is fixed
with respect to the last set of axis. If the angle between two balls is defined as y, k will be:
27r
7 =-
m (3.37)
where m is the number of balls in the bearing.
101
As a result the angle Oi can be expressed as
Oi=6+coct+iy (3.38)
Substituting Equ.(3.38) into Equ.(3.36), the forces along the main fixed axes are:
= Kcos(t, + coc t + i y)
(3.39)
W4 =Wn sin(19 + + i y)
(3.40)
= (3.41)
The total force in any direction will be the summation of all forces in that direction:
Wx, =EW,cos(z, + c t + y)
6.1 (3.42)
Wy =-EWn sin(t.5 + + i y)
(3.43)
W EW
(3.44)
3.4 Dynamic Modelling of a Vibrating Shaft Supported by

Ball Bearings
Vibrations can be classified in several ways, for example free and forced vibration,
undamped and damped vibration, linear and nonlinear vibration etc. If all basic
components of a vibratory system behave linearly, the resulting vibration is known as
linear vibration". In linear systems cause and effect are related linearly whereas in a
nonlinear system this relationship is no longer proportional. Hence there is a number of
oscillatory behaviours that cannot be explained by linear theory [Thomson, 1988; Stoker,
1950].
The differential equation describing a nonlinear oscillatory system may have the general
form:
d 2x x
+f 0
dt 2 dt (3.45)
102
For this sort of equation the principal of superposition does not hold for their solution
[Thomson, 1988; Stoker, 1950].
The vibrations of a shaft-bearing system has a nonlinear nature due to the Hertzian
contact relation of ball to race elastic contacts and the nonlinearity of the elastomers
employed as external dampers.
3.4.1 Assumptions
The real shaft-ball bearing system described above is generally very complicated and
difficult to model. Thus, in order to study this real system a number of simplifying
assumptions have to be made. The assumptions made in this thesis are as follows:
3.4.1.1 A maximum of five degrees of freedom system Is assumed for the shaft with
radial vibrations in the x and y directions, axial vibrations in the z direction
and two rocking motions about the x axis and y axis (v/ and 0 respectively).
The sixth degree of freedom is omitted by assuming that the friction
between the rolling elements and races is negligible (i.e., torsional vibrations
—about the z axis— are omitted).
3.4.1.2 The rolling elements (i.e. balls) are assumed to be massless in order to
eliminate 6m degrees of freedom (where m is the number of balls in a
bearing).
3.4.1.3 The rings are flexurally rigid and undergo only local deformation due to the
contact stresses.
3.4.1.4 Deformations occur according to the Hertzian theory of elasticity (i.e. elasto-
static footprints and pressure distributions are perfectly elliptical).
3.4.1.5 Rolling elements are positioned equi-pitched around the inner ring and
there is no interaction between them.
3.4.1.6 The sources of damping contributions either from the elastrohydrodynamic

film at the contacts or any friction present, between the cage and the rolling
elements and in various races to shaft and housing joints are assumed to be
negligible since their contribution to vibration damping is small within the
frequency range of interest in this thesis.
103
3.4.1.8 The outer ring Is assumed to be firmly fitted to a rigid housing, thus
eliminating its flexural vibrations which generally cause high frequency
vibration, while the inner ling is firmly fitted to the shaft and can be assumed
to be a part of IL
3.4.1.9 The shaft is considered to be rigid in order to eliminate its flexural natural
modes and hence only the first natural mode (rigid body mode) of the shaft
appears in the vibration spectrum. Since the other modes are not in the
range of running speed, with this assumption, while the problem is relatively
easier to solve, only the small effects of the higher modes on the peaks in
the frequency range of interest are neglected.
3.4.1.10 The pitching and yawing angles 0 and vi are assumed to be small, as are x,
y and z.
3.4.1.11 For a three degrees of freedom system analysis, both bearings are
positioned symmetrically such that they move in the same direction
simultaneously. In other words the balls are assumed to be in phase (i.e.
there is no rocking motions about the x ory axis).
3.4.2 Equations of Motion
The equations of motion are based on the second law of dynanics which leads to the net
force or moment on a mass balanced by the acceleration of mass. In a shaft bearing
system there are three kinds of applied forces. The first is external force. The second
force is due to deflected balls acting on the inner and outer rings and was discussed
earlier in this chapter. The third is due to damping in the system. In this research
elastrohydrodynamic forces are omitted and the external damping forces due to the
external elastomer dampers will be investigated in Chapter 5.
The equations of motion in five degrees of freedom system can be written as follows (see
Figs.(3.16), (3.17) and Fig.(3.18)):
3
Mi +I ( I C;(3 jicos(a )Lcm(e
1=1
3
K,(831- cos(aiLcos(0,0+ Mg = 0
( (3.46)
104
,
dz.
105
Mechanics of Ball Bearings
Chapter 3
106
Chapter 3 Mechanics of BailBearings
Al y + ±(Ici
1 d COS (a 1, sin (t 9 JO
+i(K (031 cos (cr i b, sin (0)R )+ =0

(3.47)
M + ±(Ki (Og sin (a i0+ t(K4 (83 13t sin ( 4 /1)+ Q =0

(3.48)
Initial
position
Initial
position
Fig.(3.18) Definition of axes and rocking angles
yy if) + g/C4cos(ai)Lcos(8ik)al
3 3
- (K WI cos (cciLcos(0 JR)b, + j(K (8)1. sin (a i)Lcos(0)L)R
m ( 3
- K i (8,)isin(ct jR cos(9 s)R — Q.(a1 + a)— 1 n t k = 0
(3.49)
107

J zz 1.1% 4- Ix (
i=1
3
KO c.os(a L sin(0 ;Oa,
( 3 (K 3
Ki (si )li cos(a) R sin (e)Rjk i M. sin ( ai k sin (0,0R
i=1 i=1
–±(Ki (ajiSin (ai)1 Sin (ei),)R — + a)+ 1.0 co =0

i=1 (3.50)
where Oi is as defined in Equ.(3.43) and therefore equations of motion are time
dependent.
Os= +coct+iy (3.43)
4.4 Dimensionless Groups

For the convenience, the equations of motion are normalised. This is necessary because
otherwise a change in one value of parameters implies a change from one system to
another within the given class of problem and hence the comparison of the results
becomes inadequate. This is particularly true for relatively complicated system such as for
five degrees of freedom in our case.
Redefining the variables in the equations of motion as:
. x A • A •• • K
x =—, y =—, z =—, .;=—, = w, t =cot, A i•= —
(5 0 S o So S o Ko
and the constants as:
= b1 * —
—— — —
R
D D D
Substituting these definitions in Equ.(3.46) through Equ.(3.50) will reveal the equations
of motion in nondimensional form as follows:
n ( 3
MT + K: cos(a,kcos (9,0
i=1
3
+E(K,* (6 T t cos(a jR cos(0 R)+ = 0
1=1 (3.51)
108

Ar y+ I
s3=1
( 3
K:(3a2..cos(a,)Lsin(61)1,)
( 3
K:(8:),icos(ai )1 sin(01)+Q; = 0
i=1 (3.52)
+ E(K(ani, sin ( 01.)+ (K: (8Tr sin (a jr)+ Q: =

1.4
3 3
(3.53)
3
/. + (K:(8:)ICOS (a i)Lcos(EOL)
3
— b1 (K (8:)R cos (cejR cos(9)R)+RI(K.:(45:)1 sin (ajz. cos (el)
R 3
3
- Re / (K :(6:)12t sin (a)it cos (0)11)— T; — =0
i=1 (3.54)
II 3
l e (Ki.(qcos(ai)Lsin(93L)
s=1
, s‘l
- K, (3, gcos(a,)R sin (0,q+
( K; 8, )12., sin (a)L sin (98)L)
3
K:(8:); sin (a i)R sin (011)-7'1:+f:0 e = 0
6.1 ( (3.55)
where dimensionless groups can be defined as follows:
M (02
M=
K0S-oi (3.56)
(dimensionless mass)
Qx
Ko•51-
(3.57)
(dimensionless external force in the X direction)
109
co;.:_gr_r
K081 (3.58)
(dimensionless external force in the Y direction)
(3.59)
(dimensionless external force in the Z direction)
• Mg
g = --T
K08,1 (3.60)
(dimensionless acceleration due to gravity)
re 12,, CO 2 (02
/ = =
D K 0 OcT D K061 (3.61)
(dimensionless moment of inertia about the X or Y axis)
Q (ai -
To.= *
Ko cq D
(3.62)
(dimensionless moment stiffness factor about the X axis)
= Q,(a,+ a)
K o (5,1 D
(3.63)
(dimensionless moment stiffness factor about the Y axis)
0)2
Is -
D K 0 (51 (3.64)
(dimensionless moment of inertia about the Z axis)
110
4.6 Computational Solution of the Equations
The solution of equations of motion (Equ.(3.51) through Equ.(3.55)) is obtained using

the Runge-Kutta Iterative method, which was described earlier, since Equ.(3.51) through
Equ.(3.55) are nonlinear and the direct substitution technique does not hold for them. In
the Runge-Kutta method the initial conditions have crucial importance and have to be
considered carefully.
4.6.1 Initial Conditions
For numerical solutions, the iteration starts from an initial point at which the conditions are
known as initial conditions. Initial conditions and step size are very important for
successive and economic computational solutions. Particularly for nonlinear systems,
different initial conditions mean a totally different system and hence different solutions.
Wrongly selected initial conditions can also cause larger computation time (CPU) or
sometimes wrong results.
The larger the time step, the faster the CPU time. On the other hand the time step should
be small enough to achieve an adequate accuracy. Moreover, very small time steps can
increase the truncation errors and result in wrong results. Therefore an optimisation
should be made between them.
At time t -0 the following assumptions are made:
1. The shaft is held at the centre of the bearing such that there is no radial load on the
balls and all balls are assumed to have equal axial preload and a preloaded contact angle
a The preload, as described earlier in this chapter, is:
( cos (a o
PR. = m K[B db 1) sin(a p)
cos(a p)
(3.18)
and the initial deflection, as described earlier in this chapter, is:
5o = BdbI C0S0 0 1)
cos a
(3.23)
The preloaded contact angle ap can be calculated numerically as descnbed earlier in this
chapter using Equ.(3.18).
111
Chapter 3 Mechanics of BaliBearings
2. For fast convergence the initial displacements are set to the following values:
.- 10 6 • le
xo = —A--, )i) =.___,
,_ 10
4 41
z0 – —, wo* -= 0 and 14=0
3
u0 0 050
The velocities are assumed to be zero:
=O
.4 = 0, ji; = 0, 4 = 0, i)c,* = 0 and 1k;
As the shaft is released, the displacements, velocities and accelerations are obtained.
The initial accelerations are obtained using Equ.(3.51) through Equ.(3.55) and the above
initial displacements and velocities.
3. When t >At the initial conditions have already passed and the normal procedure
commences.
4.6.2 Computer Flow Chart and the Solution Procedure
For the successful solution of the equations of motion, the following procedure is
followed:
1. Read all the necessary data of the shaft-bearing system.
2. All the variables are calculated using the above stated initial conditions.
3. The contact stiffness factor is calculated for these conditions.
4. Calculate the stiffness and damping coefficients of external dampers (i.e. elastomers).
5. Calculate the approximate natural frequencies of the shaft bearing system using a
linearised model.
6. The Runge-Kutta method is commenced to calculate values of x*,r,r i ', Ø, vto and
their derivatives with respect to time.
7. The deflections in the contact direction of each ball and the corresponding contact
angles are calculated from Equ.(3.32) through Equ.(3.35).
8. Considering new contact angles the new dimensionless stiffness coefficients are
calculated for each ball.
112
9. The dimensionless contact load in the direction of each contacting member is next
calculated using Equ. (3.11).
(3.11)
1 0.The dimensionless total restoring forces in the X, Y, Z directions are obtained from:
W'x, = IWi s cos(ai )cos(6 + cm+ y)

i= 1 (3.42)
Wy= Ii=1
Wi s cos(a) sin(/ + co c t + i y)
(3.43)
Wz*, = EWi' sin (as)

:=1 (3.44)
11. Equations of motion Equs. (3.51) through (3.55) are solved to find new values of x*,
y*, z, 0*, and w*
12. Steps 7 to 11 are repeated for the required time steps.
13. The results are recorded and their FFT is computed to obtain the vibrations in the
frequency domain.
Closure
In this chapter the deflection and the force due to this deflection on the i th ball was
calculated from contact mechanics and geometrical considerations as a precursor to
vibrations of the shaft-ball beating system. The method to find the total force on a bearing
was also defined. The equations of motion were derived in terms of dimensionless
groups. The solution procedure was also shown.
Solutions can be represented in time and frequency domains. When the solution is
investigated a series of peaks at different frequencies can be observed. Some of them
will be investigated in more detail in the next chapter.
113
CHAPTER 4
VIBRATION FREQUENCIES
OF A SHAFT-BEARING SYSTEM
4.1 Introductiort
The differential equations representing the shaft-bearing system and their solution were
defined in the previous chapter. The solutions to these equations result in shaft
oscillations that possess a number of peaks at different frequencies. These frequencies
will be investigated in this chapter.
The frequencies resulted from the differential equations may be divided into two main
categories [Gustafsson and Tallian et al, 1964
4.1.1 Frequencies generated by a geometrically perfect (i.e. defect free) bearing.

These frequencies are inherent of the system or in other words they are the
characteristics of the system. This can be further divided into two subcategories. The
first is the natural frequency of the system due to the shaft mass and nonlinear contact
stiffness and the second is the vibrations induced by the ball passage which is one of the
Important system characteristics and occurs at all shaft speeds.
4.1.2 Frequencies generated by geometrical imperfections (i.e. manufacturing

malfunctions). These frequencies may be considered as a result of vibrations that are
forced by the geometrical imperfections. Vibrations due to wavy rolling surfaces or an out
of balance shaft centre can be cited as good examples of these vibrations.
111
Chapter 4 Vibration Frequencies
4.2 The Inherent System Frequencies
4.2.1 System Natural Frequency

The system natural frequency is a function of mass and the stiffness. For linear systems
this relation is defined as:
1
1 (K y.4.—
1.= 27c ii 1—C (4.1)
where C is viscous damping factor and defined as:
I. 1 c
' = 24K-711 (4.2)
Linear
Hardening e Spring
l
Spring I
I
I
I
I Softening
e Spring
ee
e
Displacement, 5
Fig.(4.1) Softening and Hardening Springs
Damping in moderate amounts has little influence on the natural frequency and may be
neglected in its calculations. The effect of damping is mainly evident in the diminishing of
the vibration amplitude with time. For most systems the vibratory mass of the system is
constant for a given system and the stiffness coefficient, K, will therefore be the
predominant cause of the natural frequency. While for a given linear system, the natural
115
frequency is constant, the initial deflection of the vibratory mass will determine the natural
frequency for nonlinear systems where the amount of nonlinearity of the springs will have
crucial importance. An approximate natural frequency for a shaft-ball bearings system
around a constant deflection was given in Appendix 10.
Nonlinear springs are classified into two groups. The first group is hardening springs and
the second is softening springs as shown in Fig.(4.1). For a linear system the restoring
force curve linearly changes with the displacement (see Fig.(4.1)). If the slope increases
as the load increases, the spring is said to be hardening (see Fig.(4.1)) whereas if the
slope decreases as the load increases, the spring is called a softening spring. For
variable stiffness, there are certain regions where the system is unstable. For a softening
spring, with increasing excitation frequency, the amplitude gradually increases until a
point where a sudden jump to a larger value takes place. In the case of hardening spring a
sudden jump to a smaller value is observed [Harris and Credo, 1961]. The shaded region
in Fig.(4.2) is °unstable where a sudden increase or decrease in the amplitude is
observed as the significant feature. The extent of instability depends on a number of
factors such as the amount of damping present, the rate of the change of the excitation
frequency, the extent of spring nonlinearity etc. [Thomson, 1981].
In the case of ball bearings, jump behaviour can only be observed when there Is a very
large mass present which will force the system to oscillate at large amplitudes and hence it
will exhibit the effect of nonlinearity due to Hertzian contact. For small masses the system
will tend to act more as a linear system since the local deflection is smafl (see Fig. (4.3)).
Frequency, j
Fig.(4.2) Jump Behaviour of Nonlinear Systems
116
Another important behaviour of nonlinear vibrations is that permanent oscillations whose

frequency is a fraction (1/2, 113, ..., 1/n) of the exciting force or displacement can also
occur. It is termed as sub-harmonic vibration of frequency. In the same manner, super
harmonic vibration of frequency equal to some muftiple (2,3, ..., n) of the exciting
frequency, can also occur [Stoker, 1950; Den Hartog, 1985; Harris and Crede, 1976].
It should be noted that sub-harmonics may occur only under very special conditions while
super harmonics always occur and damping may completely Inhibit the existence of
subharmonic vibrations for a linear system [Stoker, 19501; but for a nonlinear system it can
occur even in the presence of viscous damping [Harris and Crede, 1976].
i (5 Mai(
Defl ection, (5
u-o
Deflection, S
F7g.(4.3) Large and small oscillations
System stability is always of prime importance for linear as well as nonlinear systems. In
the case of nonlinear systems, due to time dependent variables, slight disturbance from
the equilibrium state causes peculiar vibrations which are referred as 'parametric
vibrations". System can easily become unstable, if this disturbance takes place close to
the jump behaviour region [Harris and Crede, 1976; Rao, 19841.
4.2.2 Cage Speed

Considering pure rolling, the absolute speed of the point i (see Fig.(4.4)), located on the
circumference of a rolling element, is equal to the circumferential speed V i of the inner
ring raceway at the same point i which is on the circumference of the inner ring race way
as well. In the same manner the point o, which is the contact point of a rolling element and
the outer ring circumferences and share them both has the absolute speed of V.. The
absolute speed of each point on the ball is a combination of the speed of the ball around
its spin axis and its orbital speed round the bearing axis.
117
Therefore, the absolute velocities of points between i and o are, as shown in Fig.(4.4),
parallel to each other and vary linearly about the instantaneous axis of zero velocity. The
velocity of inner and outer races respectively can be expressed as follows:
= wi re (4.3)
ve = Co. (4.4)
vc,
Fig.(4.4) The Speeds of Bearings
Two other equations can be written using the angular speed about the instantaneous axis
CL, of zero velocity, as follows:
cou (r..— q — —k COS a)

2 (4.5)
V.= coa (r.,— q+ —11 cos a)

2 (4.6)
where a is the bearing contact angle (see Fig.(4.5)). Solving Equ.(4.5) and Equ.(4.6),
the unknowns q and ati, can be found to be:
118
V +V. d
q=rm --2--L COS a
V0—V, 2 (4.7)
w V—V.
dbcosa (4.8)
The velocity of the ball centre can be calculated from
Vm=cou(rm—q) (4.9)
Substituting Equs.(3 and 5) into Equ.(6):
V =V+V
"' 2 (4.10)
The centre of the ball rotates around the bearing axis at a speed which is known as cage
speed because the cage also rotates at the same speed. The cage speed can be
expressed as
V
(4.11)
From Fig.(4.5)
ri = r — db cos a
2 (4.12)
re = r +—cosa
1° 2 (4.13)
Substituting Equs.(4.3,4.4,4.12 and 4.13) into Equ.(4.11) the cage speed can be
obtained as
1
coc = — co,[1— c cosa]+ 1 41+ cli'cosa]
2 2 dm dm (4.14)
where
2r n. 2 trn
and co °
' 60 ° = 60
119

4.2.3 Ball Rotational Speed

Considering the inner and outer rings have different rotational speeds and the speed
vectors are perpendicular to the rotation plane, k can be shown that the outer and inner
rings and the ball centres, I.e. cage, have their angular speed vectors as shown in
Fig.(4.5).
1'0
Fig.(4.5) Cage Speed for Angular Contact Ball Bearings
The balls rotate around their own axis as well as Instantaneous axes of zero velocity.
Therefore the ball speed can be split into its two components as shown in Fig.(4.5). The
vector cob is the actual ball rotating speed around its own axis and cos is the speed at
which the balls spin about their own axis while they are precessing round the inner race.
Combination of these two speeds is coa which is the ball rotational speed in the cage
speed plane. Hence from Fig.(4.5):
cob = coacos(a)
(4.15)
120
This is caused by the rotation of the ball centre about the bearing centre. Therefore, the
effect of this speed and the cage speed should produce the same velocity with respect
to the instantaneous axis of zero velocity. Hence
= co
(4.16)
Substituting Equ.(4.16) into Equ.(4.15) and considering Equ.(4.7) and Equ.(4.14):
2
id d
2----L" ( CO CO .)[ 1- -L COS 2 a
2 db (4.17)
4.2.4 Relative Speeds
Assuming that both inner ring and outer rings are rotating at speeds of c4 and coo
respectively, cage to inner ring relative speed is
1 [ d 1 [ d
caci .= 1+-Acosa +—o) 1+ - Acos a
2 ds. 2 ° d.,
(4.18)
and cage to outer ring relative speed is
1
0) c/o = — 0) i [1 — gi cos a] + 1 41— -L
( oos a]
2 2 d..
(4.19)
4.2.5 Ball Passage Frequency

The effect of the BPF is present even If the rolling surfaces are perfect. While the shaft is
rotating, applied loads are supported by a few bails restricted to a narrow load region,
Fig.(4.6), and the radial position of the inner ring with respect to the outer ring depends
on the elastic deflections at the ball to race way contacts. Balls are deformed as they
enter the loaded zone where the mutual convergence of bearing rings takes place and
the balls rebound as they move to the unloaded region as shown in Fig.(4.6). As the
positions of the balls change with respect to the applied load vector when they move from
the loaded to the unloaded zone, the load distribution on the shaft changes thus
producing a relative movement between inner and outer rings, I.e. a periodic relative
motion between the rings must occur even though the bearing is geometrically perfect.
121

The situation is best illustrated by considering the change in angular position of a ball set
with respect to a point on the load line as seen in Fig.(4.6).
Fig.(4.6) Variation of the total force for the different positions of ball set
The bearing on the left in Fig.(4.6) show the situation of the balls for a set of 6 balls at time
t =0. Ball 1 is immediately under a gravity load. The bearing in the middle in Fig.(4.6)
shows the situation at the time t =1/2 x (time for a complete cage rotationlm ) where m
is the number of balls. In this case only two balls are supporting the gravity load as the ball
on the right and left become perpendicular to the load line. The radial deflection is
different on account of their being now 2 loaded bearing balls instead of 3 as previously.
The shaft therefore moves downwards in response to the decreased contact spring
stiffness force (which is also nonlinear) as shown in Fig.(4.6). This change occurs
cyclically as shown with the shadowed line in Fig.(4.6). To do this the shaft must
accelerate so a vbration occurs as k is a mass supported by springs.
The shaft and inner raceway have approached closer to the outer race and housing in the
time taken for one half of the ball spacing to pass a point on the outer raceway to reach
the position of the ball set shown in the middle bearings in Fig.(4.6).
When Ball 6 comes to the initial location of Ball 1, the shaft regains its initial position as
shown in Fig.(4.6) for the bearing on the right. The time passed for the shaft's regaining
its initial position is:
time for a complete rotation of the cage

t—
m (4.20)
122
Hence the shaft will be excited at the frequency of m coc where m Is the number of balls
and cec is the cage speed. This frequency is called ball passage frequency (BPF for
short). In mathematical terms, the BPF can be described as the cage speed times the
number of balls.
CObp =M CO, (4.21)
It is relatively more difficult to visualise the case for the inner ring ball passage frequency.
It is described as the passage of ball set from a point on the Inner ring surface and can be
given as:
cObp = m (co,— roc) (4.22)
The vibrations associated with the BPF is called the elastic compliance vibrations and two
dimensional since the rotating ball springs and asymmetry. The vibration is also non-
sinusoidal because of spring nonlinearity.
The effect of the BPF can be worst when it coincides with a natural frequency of the
shaft-bearing system. This is also demonstrated in Appendix 7.
4.3 Manufacturing Malfunctions
In spite of very high standards in the manufacturing of ball bearings, there are still many
errors that are unavoidable. Therefore certain production anomalies in the geometric
shape or dimensions exist even in high precision ball bearings. Some of the malfunctions
are as follows:
4.3.1. Waviness,
4.3.2. Out of round rolling members,
4.3.3. The presence of off-sized rolling elements,
4.3.4. Misalignment of the bearing rings,
4.3.5. Out of balance of the shaft centre.
Some of these anomalies were investigated by different research workers [Wardle, 1988;
Ellis, 1970; Meyer et al., 1980 etc.]. However they need further investigations in order to
understand them better. Some other topics were not thoroughly investigated such as
the flexural vibrations of outer race and the influence of damping and friction on vibration
levels. Particularly for the high speed applications, these effects influence the vibration
of the system but since these anomalies were found to be less related to the aims of the
research, they were left out.
123
In this thesis the available theories for some anomalies such as waviness, off-sized ball,
defects on the running surfaces, misalignment will be extended and some new attempts
will be made in order to understand the effect of these anomalies on the vibration
characteristics of shaft bearings systems.
4.3.1 Waviness
Waviness is a manufacturing imperfection. It may be caused by different manufacturing

malfunctions such as uneven wear of the wheel in grinding operations, variable
interactions between the tool and work piece, vibrations of machine elements or
movements of the work in the fixture.
Fig.(4.7) Waviness of the bearing
Waviness is normally in the form of peaks and valleys of varying height and width. This
causes problems in the mathematical modelling of waviness effect. A statistical approach
is necessary in order to have a complete solution. If the rings are assumed to bend due to
the ball loads then the flexural vibrations of the rings as well as the rigid body motion have
124
to be considered. To avoid these problems the inner and outer rings are assumed not to
bend under the ball loads and a perfect sinusoidal wavy surface is assumed with a regular
peaks and valleys of constant height and width. Furthermore, the wave length Is
assumed to be much greater than the ball to race foot-print width and the wave geometry
itself is assumed to be unaffected by contact distortion.
Ti
5
2
A.
Fig.(4.8) One wave of the inner ring
Waves are described in terms of two important parameters; the wavelength, A, which is
the distance taken up by a single cycle of a wave, and its amplitude, T.
For geometrically perfect surfaces the wavelength is infinite (A - 00). For an imperfect
surface with N waves, the wavelength is inversely proportional to the number of waves,
N.
When a ball is moving round the inner ring, it follows the rolling surface contours
continuously. The point of interest is the amplitude of the wavy surface with respect to
central point at a certain angle from the reference axis. Waviness exists on all surfaces.
To simplify the problem, first the inner ring waviness and then the outer ring and ball
waviness will be considered.
4.3.1.1 Inner ring Waviness
If it is assumed that the inner race surface has a circumferential sinusoidal wavy feature,
the radial clearance consists of a constant part and a variable part. The amplitude at the
angle e' (see Fig. (4.7)) is directly related to the wavelength. This can be clearly seen if
125
one wave is flatten out (see Flg.(4.8)). Since It is assumed the wave is perfectly sinusoidal
the amplitude will vary between zero and the maximum amplitude and can be express as
follows;
l'i = Tpsin(2K—L
A) (4.23)
where rp is the amplitude of the wave. If there is a constant interference or clearance

applied, Equ.(4.23) will take the form of:
11= To + Tpsin(Dril
(4.24)
In the case of angular contact ball bearings, ro is replaced by the deflection due to the
preload. Equ.(4.24) consists of a constant part and a variable part. From Fig.(4.7):
L=r9' (4.25)
111 bs
tifeltO
i th
nIrtt i k
#4101
Mil
;if 1
alai
Fig.(4.9) Relation of the angles
126
Since the wavelength is the length of the inner race circumference divided by the
number of waves on the circumference.
(4.26)
Substituting Equ.(4.25) and Equ.(4.26) Into Equ.(4.24):
= To+ Tpsin(N (4.27)
Since the inner race is moving at the speed of the shaft and the ball centre at the speed
of cage, for the inner race waviness, e' should be replaced with an angle e i for the ith
ball (see Fig.(4.9)).
0,= +(D, — yi (4.28)
If a point a on the circumference of the outer race and a point b at the ball centre are
assumed at the initial time and initial position at an angle 6 apart from a reference axis, as
seen in Fig.(4.9), after the time t taken, the cage, i.e., the point b, will lag the shaft, i.e.,
the point a, and as a result of this, the ith ball will be at the angle of —(c o t— o.)c 0. Hence
the instantaneous height at the angle of interest:
= To + Fp sin [N(19 +(co,— co)t+ yi)] (4.29)
4.3.1.2 Outer race waviness
In the case of outer race since the outer race is assumed to be stationary and the balls are
rotating at the speed of cage, Equ.(4.29) will take the form of:
(I) = To + rp sin [N(i, + am+ yi)] (4.30)
127
Fig.(4.10) Ball waviness
4.3.1.3 Ball Waviness
Balls are free to spin about any axis and this axis may even change during the rotation. In
order to calculate the waviness of balls, a simple case is considered where a ball with a
perfectly sinusoidal wavy surface rotates about an axis normal to the plane containing the
centres of the inner and outer race contacts. In this case the instant amplitude of the ball
waviness is (see Fig.(4.10)):
ai = ap sin (N 01 (4.31)
This will cause a change in the ball diameter through inner and outer contacts:
A db = 2ap sin (N 01 (4.32)
where N .2, 4, 6, ... (for N .1, 3, 5, ... 4db-0.)
The effect of ball waviness on the general clearance can be given from Fig.(4.10) as:

(11b = 2ap sin [N wbt] (4.33)
Note that in Equ.(4.33) the location of ball has no effect on the ball waviness contribution
to the clearance.
128
4.3.2 Off-sized Balls

An off-sized ball causes a different force excitation on the Inner ring from the rest of the
balls. This different force can be larger or smaller than the rest depending on the off-
sized ball diameter. Fig.(4.11) shows that one ball has a greater diameter from the rest of
the balls in the set. This ball will be forced to squeeze more in relation to the other balls
and hence will produce a greater force from the rest. Hence for this ball the force
produced will be:
3
Weer = K(3 + ASP (4.34)
where 48 is the diameter difference of the off-sized ball.
Fig.(4.11) Presence of off-sized ball in a ball set
As explained in Section 4.2.5 under the title of Ball Passage Frequency, each ball inserts
Its force on the shaft in turn. The off-sized ball will insert a different magnitude of force
and hence will disturb the movement of the shaft. As the balls are moving at the cage
speed, this will repeat itself for each cage rotation and produce a peak at the cage speed.
The same thing is true for more than one off-sized ball but this may cause a change in the
129
amplitude of the peak. However, If the position of the balls in a set is arranged
symmetrically, vibrations will be observed at the superharrnonics of the cage speed.
When the cage speed or Its superharmonics coincide with the natural frequency, the
system will resonate. More information on the vbrations associated with the off-sized
balls can be found in Chapter 6.
4.3.3 Misalignment
Ball bearings are normally assumed to be correctly aligned (see Fig.(1.5.a)). Under
operating conditions this is not always true. It is possible for the bearing to operate as
shown in Fig.(1.5.b) where one ring is angularly displaced relative to the other, about an
axis at right angles to the bearing running axis. This is a common fault in bearing
operation and can severely affect the performance of the system. It can cause vibrations
or excite vibrations at other frequencies.
Fig.(4.12) Misalignment of angular contact ball bearings
130

There are many causes of misalignment. These can be placed into five categories [Ellis,
19701. bearing errors, deflection of shaft and bearing housing, poor bearing arrangement
or selection, machining errors and fitting errors.
Since misalignment was not commonly known to cause the severe vbrations and noise
encountered, it was investigated from the view of cage and bearing failure [Barash,1960;
Ellis,1970; Andrirason, 1970; BrAndlein, 19851. More recently Wardle [Wardle and Poon,
1983; Wardle, 1988] identified the misalignment as a predominant effect on bearing
noise and vibration.
The main reason for taking up misalignment into the scope of this work is that the
detection of bearing misalignment before equipment is put into service is desirable. As
the maximum amount of misalignment cannot be seen by the naked eye even if the
bearing concerned is accessible, there is no universal, simple, quick method to examine
misalignment. Even the experienced engineer can hardly detect misalignment without
direct measurement, which is time consuming and expensive, and only used when large
expensive equipment is involved. Moreover, this is only a static measurement and
operational loads may cause misalignment that is unpredictable under dynamic
conditions.
Therefore if the vibration signature due to a certain amount of misalignment is known,

vibration monitoring can tell us of the presence of the misalignment and its magnitude.
4.3.3.1 Deflection due to Misalignment

Misalignment of the bearing is generally defined about an axis at right angles to the
bearing running axis. This can be mathematically transformed to the bearing rotation
centre as a deflection and an angle. Hence misalignment can be expressed in terms of an
angular displacement from the centre of rotation axis.
In the general case, the maximum displacement due to misalignment will be at an angle
ec which depends on the way it is misaligned and can be found using the cosine rule (see
Fig.(4.12)).
,
((91, +a)cos(p)— (91. + a))2 + 2
8,= —Bdb
Pi + a)sin (p)+ Bdb sin (a0))2
(4.35)
131
where p is the misalignment angle. The resulting deflection on the i th ball is (see
Fig.(4.12)):
64 =8, cos O c cos 0i + (5, sin 0, sin 0i (4.36)
4.3.3.2 Contact Angle Variation due to Misalienment
In order to determine the effect of misalignment, the contact angle for each ball must be
calculated. Considering a ball bearing whose inner ring is misaligned, as described in
Fig.(4.12), the maximum contact angle variation is experienced by a ball. at the angle Oc
(see Fig.(4.13)) or right opposite to it. The contact angle at the angle 0, may be written
from the definition of p i as:
arctan ri + a)sin(p) + Bdbsin(a.))

a.=
+ a)cos(p)— (91. + a) (4.37)
where —9t i 5 a 5 9t i for the inner ring misalignment and —91 0 5 a 5 9, for the outer
ring misalignment. From this, the contact angle due to misalignment for the i th ball can
be found using effective misalignment angle for the ith ball, Pei
1(91i + a cos (n +0.— 0,))sin(p.) + Bdbsin(a.))

a i = arctan
(R+ a cos (n + 0 i — e))cos(pa )—(91 0 + a)
(4.38)
where
19c constant (for outer race misalignment) and

and
ec . cot (for inner race misalignment)
From Fig.(4.13) for small p and pee:
88,cos(0,) „I
tan (P. kp cos(8, — c))
.) = 1 — R tan (4.39)
132
Or considering the location (see Fig.(4.13)):
pco0+ (4.40)
Path of ball v.qth

no misalignment
Actual path
of ball
View of surface ABC flattened out
Fig.(4.13) Contact angle variation
Substituting Equ.(4.40) into Equ.(4.38):
1(9ii + a cos(n + e,— ec))sin(p• cos(n + 9,— e c ))+ Bdbsin(a.)

ai = arctan
(91i + acos(x+ 9 6 — 0 ))cos(pcos(n+ 96 -19,))—(91.+ a)
(4.41)
Note that this is only for observing the variation of contact angle due to misalignment,
because otherwise the change in contact angle is automatically included in the equations
via the deflection due to misalignment.
133

4.3.4 Out of Balance of' Mass Centre
In rotating machines a common source of vibration excitation is the off-centre or

eccentricity of the mass centre, where the geometric centre of inner ring, o, Is off-set by
an eccentricity, e, from the geometrical centre of the shaft, as is shown In Fig.(4.14).
Fig.(4.14) Out of balance shaft mass centre
A theoretical investigation of the response of a rotating rigid shaft supported by a pair of

radial deep groove lubricated ball bearings having rotating unbalance shaft centre was
carried out by Rahnejat [1984]. The research showed that rotating unbalance at the shaft
centre, excites the system at the shaft rotating speed and is more severe at high
frequencies. When the shaft speed coincides with the natural frequency, resonance
occurs. Some other researchers also reached the same conclusion [Wardle and Poon,
1983; Eisenbeis, 1985; Franco, 1990].
In modelling this behaviour, it was assumed that the force due to unbalance at the shaft
centre was Mw 2e in the direction o-o' and the gravitational force was In the x direction
due to the mass of the shaft. The angle between the centrifugal force vector and the x
axis is a function of the shaft speed and time. Therefore the out of balance forces acting
134
In the x andy directions are:
(F2)7.= F.+ Mg = M(g+ w2ecos(cot)) (4.42)
(F' )7. = F' = M co2esin(cot)

(4.43)
where subscript T stands for total.
There is also a moment produced by this movement. This is (see Fig.(4.14)):
T = M gesin(cot) (4.44)
4.3.5 Defects on the Rolling Surfaces
The vibration signature produced by bearings can be very informative if its meaning is well
understood. Since failure in rotating machines can be dangerous (in aircraft for example)
and can lead to loss of time and money, machine health monitoring is employed by many
companies to get the necessary information out of the vibration signature . Regular
monitoring of machines point out the change in the state of the the machine immediately
since the condition of ball bearings can be judged by the nature of their vibration pattern
[Karakurt, 1989].
The experimental studies showed that defects on bearing raceways, the cage, or balls
cause unique vibration signals [Taylor, 1980]. However, the theoretical model for defect
detection is not well established although some analytical solutions were presented
[Igarashi and Hamada, 1982; Igarashi and Yabe, 1983; lgarashi and Kato, 1985;
McFadden and Smith, 1984; 1985]. This is mainly because the problem is complex and
more than one defects makes the problem very difficult to analyse. For this reason an
attempt was made to model defects in the present study, in particular to see the effect of
surface defects on system vibrations, since the available computational approach makes it
easy to handle.
First a model is established for a single defect on the outer race surface, then a single
defect on the inner race surface and finally a defect on the ball surface is modelled. The
combination of these defects then is applied in the simulation.
135
4.3.5.1 A Defect on the Outer Race Surface
Let there be a defect on the surface of outer race at the angle from the horizontal axis
X. If coc t + i y (i= 1 , m ) coincides with the defect angle 9, a ball at its contact will
have additional deflection 6d • Hence (see Fig.(4.15)):
Oid=c5;+•54 (4.45)
Fig.(4.15) Defect on the surface of outer race
For the computational convenience, following definitions are made:
1. Since the defect is not at a single point, to define the defect at a certain angle may
cause problems in computation. Thus a tolerance band put on the angles. For example,
for the outer race defect, the defect angle is defined as
Cd = 2r.
11t (4.46)
where w d is the width of the defect.
136
2. Since cod + i y is increasing continuously, its sine and cosine are simultaneously
checked with those of the defect angle.
4.3.5.2 A Defect on the Inner Race Surface
Let there be a defect on the surface of inner race. This defect will rotate at the shaft
speed, co, as the inner ring is forced fitted to the shaft. If the defect angle, on ± w d /
), coincides with one of the balls, roc t + i 7 (i = 1, m), the deflection on that ball will
be (see Fig.(4.16)):
8,d = O . + 8 d (4.47)
cot
Fig.(4.16) Defect on the surface of inner race
4.3.5.2 A Defect on the Ball Surface
From Fig.(4.17) it can be seen that when the cosine of the angle (cob +toc )t equals zero,
for that particular ball the deflection will be:
Sid = Oi + Od (4.48)
137

Fig. (4.1 7 Defect on the surface of a ball
Closure
System frequencies can be classified under two main categories. The first being the
vibrations of a geometrically perfect system and the second is vibrations due to the
imperfections in the system. In this chapter these frequencies were mathematically
defined.
In order to reduce the untoward effects of these frequencies to an acceptable level,

elastomer external dancers will be employed in this thesis. Their dynamic properties will
be investigated in the next chapter.
138
CHAPTER 5
ELASTOMERS AS EXTERNAL DAMPERS
5.1 Introduction
In the last two chapters, a vibratory model of a rigid shaft supported by a pair of angular
contact ball bearings having no external damping was described. In the previous chapter,
it was shown that the solution of the equations of motion would result in a series of
vibration peaks at certain frequencies. One method of reducing untoward effects at
these frequencies, is to introduce elastomers as external dampers. Hence, in this
chapter, the dynamic properties of elastomers will be investigated and the system
characteristic equations including external dampers, will be solved for a three degrees of
freedom system.
Elastomers are an artificial rubber made up of polymer chains pinned together at points
along their length by blobs of polystyrene formed by the coalescence of polystyrene
sections along each chain. Unlike natural rubbers, they can be melted and formed into
any desired shapes. Compared to thermo-setting polymers elastomers do not have
enough cross-links to make them completely rigid, but do have enough to ensure they
do not break under the heavy strains typical of natural rubbers [McCrum at al., 1989]. Like
all polymers, elastomers are viscoelastic in nature, which means that they are very much
rate and temperature dependent, a high strain rate giving a high stiffness in the same way
that cooling the material down to below its transition temperature would do. Therefore,
their internal damping mechanism enables them to be used as external dampers in many
industrial applications.
139
Chapter 5 Elastomers
Elastomers have advantages over oil-squeeze dampers because no oil pump circuitry is
needed. They are suitable for shafts supported by ball bearings which have a lubricant
sealed within them. Elastomers can also be used in engineering applications in order to
keep down the amplitude of the vibration of a shaft running in the vicinity of its critical
speeds, and/or reduce the vibrations of a shaft, and/or reduce or eliminate the noise
problems caused by vibrations of a shaft. For these reasons, in this thesis different
designs of elastomers, mainly cylindrical buttons, rectangular cartridges and 0-rings are
studied as external dampers to shaft-ball bearing systems.
In order to study elastomers one should have a general knowledge of their dynamic
properties. Section 5.2 of this chapter describes the dynamic properties of elastomers.
Since the dynamic properties of elastomers are shape and deflection dependent along
with other dependencies, formulations for obtaining dynamic properties of elastomers
from their material properties (i.e., shear and Young's moduli) is given in Section 5.3 for
various geometries. The material properties of elastomers used in this thesis are
obtained from a series of contractor reports for NASA. The experimental procedure for
obtaining data is summarised in Section 5.4 and the introduction of this data to the shaft-
ball bearing system as a mechanical model (Voigt model) is given in Section 5.5. A brief
description of the subroutine written for this purpose is included in Section 5.6. Finally
the obtained equivalent spring and damping factors are employed in the solution of
equations of the motion for an externally damped shaft-ball bearing system in three
degrees of freedom.
5.2 Properties of Elastomers

Some properties of elastomers have to be well understood before proceeding with the
calculations. Some of the unique properties of elastomers are given below:
5.2.1 Large Deformations

It is important to note that deviations from linearity of the load-compression curves for
elastomers are usually ignored for strains up to 10%. At higher strains, the value of
Young's modulus is to be modified as follows [Allen et al., 1967]:
Eelfr =pEo(l+kS;)
(5.1)
where
p is an empirical constant for a particular geometry,
Eeff is the effective Young's modulus,
140
E0 is Young's modulus,
k is a material constant proportional to hardness and
SI Is a shape factor equivalent to the loaded area divided by the force free area.
5.2.2 Dynamic Properties of Elastomers
In many applications, elastomers are subject to oscillatory or impact stresses. When

examined thoroughly, the factors controlling the dynamic properties of a given
elastomeric product fall into four categories: the composition of the material In the
product; the manner in which the material is processed; the size, shape and configuration
of the product (design factor); the manner in which it is used or tested (strain rate, strain
magnitude and temperature).
ft is important to note that the amplitude effect becomes significant only at relatively high
amplitudes of vibration since it relates to the strain in the elastomer.
Important compounding parameters are the degree and type of cross-linking and the
amount and type of carbon black used to stiffen the compound.
5.2.3 Geometry Effect
The specific value of the dynamic spring rate and damping coefficient in the overall
dynamic response of a elastomeric product, are controlled in a major way by the design of
that product.
It has been proved that the static spring rate of a rubber assembly may be changed by a
factor of over 100 by changing the ratio of free to loaded area (principle reviewed by Allen
et al. [1967]). Similarly the dynamic stiffness changes with shape factor, but data
collected by Puydak and Auda [1967] indicate that the relationship is complex.
According to Sommers and Meyer [1973] damping is even more difficult to treat in design
with the best approach most probably being the use of "tan (6). to calculate the
approximate damping coefficient, since it has been shown to be approximately constant
for a given elastomer.
Care should be taken also when considering the level of preload (governs constant
stiffness K0 (see Appendix 6)), noting that its variations could be important when
operating on the nonlinear portion of the load deflection curve.
141

5.2.4 Frequency Effect
The viscoelasticity of the material is also seen in its time dependent response to stress or
strain. Creep and relaxation are always present to a large degree in elastomers, which
means that their response to continual or sinusoidal loading is further complicated since
the stress and strain are not in phase.
The best way of representing the response to a sinusoidal input is in terms of two rotating
vectors ao and Eo as shown in Fig.(5.1).
Fig. (5.1) Vector representation of stress and strain during oscillatory loading
As can be seen, the stress has two components, one in phase with the the strain and one
90 degrees out of phase. Therefore it would be best to describe the response in terms
of a modulus and a phase angle.
Es = E' -f-iE ff = Ell +itan(3))

(5.2)
where
E' =a
—ecos(8) (in phase modulus)
Co
(5.3)
Eff =LrEsin(3) (out of phase modulus)

E. (5.4)
and
tan(S)=.
(5.5)
142
In Equ.(5.2) E' is called the storage modulus, and tan (8) the loss tangent. Fig.(5.2) is
obtained by plotting these functions: E and tan (8), as functions of frequency. Where,
on these curves, the particular frequency range of interest lies, depends on the material
In question, but overall it shows that the stiffness and loss factor both increase with
frequency.
From Fig.(5.2) the section a (rubbery region) would represent a low damping material,
and the section b (transition region) a high damping material which would be much more
frequency dependent in its properties than the low loss material. The rubbery and
transition regions are those normally encountered in elastomers and otter the best
opportunity for use in vibration control [Smalley and Tessarzik, 19751
Rubbery (a) Transition (b) Glassy (c)

•1A
1.5
E'
E"
tan(S)
1.0
0.5
LOg(Frequency, CO)
Fig.(5.2) Regimes of Elastomer Dynamic Behaviour

(after Smalley and Tessarzik (1975))
An aspect of rubber and elastomers that is not shared by other polymers is their high
dependence on their geometry. Because rubbers have a Poisson's ratio of nearly 0.5,
they are virtually incompressible so the amount of constraint they are subjected to
determines how stiff they will be, up to the point when they are fully constrained they will
be completely rigid and otter no damping at all.
This level of constraint is normally described by the use of a shape factor. This is defined
as:
143
Area of one of the loaded faces

Shape Factor, Sf=
Total area of all unloaded faces (5.6)
With this definition, the actual modulus can be estimated from an equation such as
Equ.(5.1).
5.3. Formulation of Elastomers
The work presented here is derived from a property of rubbers that is not so commonly
known; high extensibility and recovery due to deformation changes in entropy.
The differential of the Helmholtz free energy with respect to the length L of the specimen
is equal to an external tensile force F[McCrum et al., 1989]:
E) _TVS
F.VL
d Ty a idly (5.7)
For an ideal rubber while there are changes in the entropy S with the change in L, there is
no change whatsoever in E. Since the second term on the RHS of Equ.(5.7) is negative
an increase in F is generated by an increase in L.
If the elastomers are made up of chains, for an ideal chain detached from the network with
one end placed at the origin of the coordinate system as seen in Fig.(5.3), the internal
energy is independent of the chain end-to-end vector.
For the specimen undergoing deformation as seen in Fig.(5.3), assuming that the
randomness of change in molecular shape with the passage of time is reproduced by the
Gaussian theory, this will lead to (see Appendix 3):
w.VG(2"2+x2-3)
2 " Y s (5.8)
where
W is the total work done during the deformation of rubber,
V is the volume of rubber undergoing deformation
G is the shear modulus of the rubber
Ax , Ay , and A. are the ratio of deformed dimensions to undeformed dimensions in x, y,
and z axes in Cartesian coordinates respectively (see Appendix 3).
144
By definition, the force applied is linked to the strain energy function by a first order partial
differential with respect to the change in the length. In the same way, the stiffness is
related to the force applied by a first order partial derivative with respect to the change in
length. Hence:
K=a raw)
"C.W) (5.9)
where 1 is the length of the specimen.
Fig.(5.3) The representative chain OA detached from the network

(after McGurn et al. (1989))
Since the change in length, 1, is also involved in the extension ratios, Equ.(5.9) is used in
the following form:
K. a (awanax
aAaA. al ) al (5.10)
If Equ.(5.8) is substituted into Equ.(5.10), the form of resulting function will be as follows:
K = G f (Geometry, deflection) (5.11)
Since the vibrations take place with a small amplitude, It can be assumed that the dynamic
stiffness will also be in the form of [Gabel, 1954; Frederick and Payne, 1978; Smalley and
Tessarzik, 1975; Grassano, 19911:
145
le = G.1 (Geometry, deflection)

(5.12)
where K a is the complex stiffness coefficient and G a is the complex shear modulus.
Hence it will be possible to derive the stiffness of a given elastomer dancer from its shear
or Young's modulus.
5.3.1 Rectangular Cross-section Buttons

The total work to deform a rectangular cross-section button, as seen in Fig.(5.4), is (see
Appendix 3):
w .V G
____x2 .4. 1 2 + 1 2 _ 3]
2 L z ' ' (5.13)
Since the volume is constant k Ay Az . 1. The deformation in the x and y directions

from the deflection in the z direction may be assumed as equal for small deformations
because in this case neither the material, being isotropic, nor the mode of stressing
favours either direction [McCrum et al., 1989]. Hence:
2,2 1, = 1 (5.14)
where A. .A.2 .A„, . Substituting Equ.(5.14) into Equ.(5.13)
Fig. (5.4) Rectangular cross-sectioned Button
146

VG[2 i 2 ,
YV = — - --i-4.--i]
(5.15)
, a (aw al*. V G[ 2 i]
aA, aA, az az
A = — = = -7 +1
(5.16)
Considering z i . h and Az . 1 + Oh I h Equ.(5.16) becomes:
ab [ 2
K =— G +1 ]
2h 1 -4. Lk)
( h) (5.17)
Since
n(n- l) 2
(1 + X )* = 1 + nx+—x +...
2! (5.18)
Ignoring the higher order terms and considering compression and shear cases,
Equ.(5.17) will take the form of:
. 3 ab Oh
K = ---(E +G)(1-2--)
2 h h (5.19)
5.3.2 Cylindrical Buttons
The work done in the axial direction will be (see Appendix 3):
v kT (2)
Wi = -- [VI —1]
2 (?)o (5.20)
The work in the radial direction, considering Equ.(5.21) will be as in Equ.(5.22) [Aklonis
and MacKnight, 1983].
2 2
(r )i = 3. (12); (5.21)
147
v kr (r2).
w= —2.[A 2 —1]
r 2 r2
(5.22)
Hence the total work (see Appendix 3):
VGr
W = — [ 2,1 2 + A2-3]
2 (5.23)
Since the volume is accepted to be constant Equ.(5.23) becomes
, V G[ 2 ,
w= —
2 1, (5.24)
F7g.(5.5) Cylindrical Button
Calculation for a cylindrical button using the Cartesian coordinate system, with the
assumption that 2,, -A, A, ,will also result in Equ.(5.24) [Grassano, 1991]. This is the
same equation as Equ.(5.15) and results in:
3 n d2
K = --(E + G)(1-28÷11)
4 h (5.25)
148
5.3.3 Ring Cartridgg
For ring cartridges, two equations for stiffness were derived. The first is the modified form
of the equation for a rectangular cross-sectional button. It is assumed that during the
deformation, the cross-sectional shape follows the same pattern as shown in Fig.(5.4).
When the bearing outer ring moves in any direction the situation in Fig.(5.6) will arise
where the cartridge outer ring is fixed and the cartridge inner ring (i.e. the bearings outer
race) is displaced by Sr in one direction. In this case, only the half of the cartridge goes
under deformation as seen in Fig.(5.6). Hence from Equ.(5.19):
374R .+R0 )t
K=- s (E+G)(1---L5--
r
4 It—R4
(5.26)
Ri +
( )– Or
2
(5r
cj•
7
R-+R
L=7“--2)
2 •
Fig.(5.6) Ring cartridge
The second equation is derived as follows: For the cartridge, Equ.(5.13) will take the form
of (see Fig.(5.6) and Appendix 3)
w.V2rA2+2,2+2,2_3]
2 L L a (5.27)
where (see Fig.(5.6))
149
Chapter S Elastomers
Or
A.L=1— (5.28)
2a (5.29)
1 1 = 2 a L
IL A, Or' —(L+2a)e r +2a L (5.30)
In these equations (see Fig.(5.6)):
L= 12E (Ri + R.)

(5.31)
a= R.— Ri
(5.32)
Hence
- 1 8
(R.— R,)
27-11 (10- R.)
772---1
96 7t2(R0 — R,) 2 (Ri + R0)2

+
(2(R. R1)— er)2(n(Ri + R.) — 2 Or)4
K— a (aWLV(E+G)
Der WO 2 64 it2(Re — R1 ) 2 (R1 + R0)2
+
( 2 10— 61-)3(7E(R,+ R.)— 2 er)3
24 1t 2(R0 — R1) 2(Ri + R0)2

+,
(2(R.— R,)— Orr + R.) — 2 Orr
(5.33)
5.3.4 0-rings
For 0-rings three equations were derived. The first one is the modified form of the
equation that was derived for cylindrical buttons. It was assumed that during the
deformation the cross-sectional shape remains unchanged but the diameter of the 0-ring
(i.e. R,2 — R 1 ) changes (see Fig.(5.7)). In this case Equ.(5.25) becomes
2 2
= 323 (R.—
n d (Ri + R.) ,
Ri )
41 R.—
15r )
(5.34)
The second procedure followed for 0-rings is very similar to the one for the cartridges. In
150
order to reduce the mathematics Involved, the deformation shape is assumed to be as
shown in Fig.(5.7). In Fig.(5.7) Or is assumed to be vibration amplitude. Following the

same procedure as for the ring cartridges results in:
1 896x2d2(Ri+R.)2
,+ 2+
2c1- n2(R,+ R.)- (2c1- Or) 2(n(R,+ R.)- 2 er)4
V(E+G) 647t2d2(Ri + R0)2

K=
2 (2d- Or)3(n(Ri + R.)-28r)3
24712d2(R4+R.)2
(2d- Or)4(n(Ri + R.)- 2 Orr
(5.35)
2 a=d+ &
Fig. (5.7) Assumed deformation of the 0-ring cross-section
The third method is similar to the beam-column method including radius effects. Basically
It follows the same procedure as in Appendix 4.
dF dr
ercos(g)
J,. E2udg r (5.34)
R.
_dr
Or sin (g) =
r
(5.35)
where
151

1
li
(r — T)2 rf
5TC1
0* )
u = [ 1 f
( d + ercos(c)) 2 )
2) (5.36)
where R = (R i +R 0 )12 (see Fig.(5.8)). Hence Equs.(5.34 and 5.35) becomes

respectively
dFc 1 dr
cos(g) = E _ orcos(g))dc I
r. R 2
-7-?)2
1 )2
(d + Orcos(g)
2 (5.37)
R.
dF 1 dr
ersin ( g) = G (d — 1
Orcos(c))4 2 r
(r_)2
1
d+ Orcos(c))2
2 ) (5.38)
br
2
, Fig.(5.8) 0-ring
152
Chapter S Elastomers
The total force on the element is
dF = dFc cos (C) + dFasin(C) (5.39)
Hence
a I
K = =44 .1@F cos(0+ dF,sin (0)1
aSr s-0 e
(5.40)
Equs.(5.37 and 5.38) were solved using Mathematicao. However, Mathematicas failed
to solve Equ.(5.40) and it is to complicated too solve with the classical methods.
Therefore, Equs.(5.37 and 5.38) are simplified by assuming that, since Sr is small, sine
and cosine of the angle is ignored. For this Equ.(5.40) will take the form of
-1g
a 3 (d— 8r )80 cos2 (c)+ Gsin2(c))dc
K=
a er 4J
RI
c=0 1 1 dr
J
r=R,[ ) 2 r
(r — T2)2
1
(d +)Sr
2 - (5.41)
Equ.(5.41) was solved using Mathematica 6 and it gave the successful results
presented in Chapter 7.
5.4 Experimental Results Used in This Thesis
The experimental data used in this thesis is obtained from five contractor reports [Chiang
et al., 1972; Gupta et al., 1974; Smalley and Tessarzik, 1975; Darlow and Smalley, 1977;
Smalley et al., 1977] prepared for NASA (National Aeronautics and Space Administration)
between 1972 and 1977.
The first report [Chieng et al., 19721 investigates the basic methods. An experimental
arrangement was set up during the preparation of this report and It was used for all reports
153
except the fourth one. it was a forced vibration, resonant mass type of apparatus. It was
designed as a base excitation, electromagnetic shaker driven mass-spring system which
could be brought to resonance for a range of differently sized vibrating masses on top of
the elastomer "springs". One of the major features of this rig was that the tests were
performed at near-resonance conditions. Since phase angle between base excitation
and resonant mass response is an accurate indicator of the amount of damping in the
region of resonance, measurements were preferably made in the phase angle range
between approximately 15 to 165 degrees [Gupta et al., 1974]. This required that the
test frequency be approximately 0.9 to 1.5 times the critical frequency of the elastomer
mass resonant system. In order to cover the frequency range of 100 to 1000 Hz. with
sufficient test points, the size of the mass was therefore changed.
Elastomer Specimen Phase shift
Base excitation resonant mass technique
Resonant Mass
Elastomer Specimen
Rotating excitation resonant mass technique
Fig.(5.9) Experimental techniques used in the reports
Gupta et al. [1974] used this test rig to measure the dynamic characteristics of shear and
compression elements. The research showed that direct power fitting to the
154
experimental data gave satisfactory results. It was also postulated that mechanical models
can represent the elastomer.
Smalley and Tessarzik [1975] investigated the effect of temperature, dissipation level and
geometry. This report also used the same test rig. In particular, the prediction models
presented in this report and the comparison with the experimental results, were useful
tool for this thesis.
However, in this thesis most of the comparisons were made with the experimental data
given in Report 4 [Darlow and Smalley, 1977] and Report 5 [Smalley et al., 1977], since
the experimental rig employed in Report 4 is very relevant to the grinding spindle model
presented in this thesis. While Darlow and Smalley [1977] reinvestigated the previous
test results namely cylindrical buttons, ring cartridges in their new rig, Smalley et al. [1977]
Investigated the dynamic properties of 0-rings employing the rig used in the first three
reports with some modifications and additional 0-ring mounter as seen in Fig.(5.11) (see
also Fig.(2.40)).
The method employed in Report 4 [Darlow and Smalley, 1977] was essentially the same
as the Base Excitation Resonant Mass Technique employed in the other reports (see
Fig.(2.40) and Fig.(5.9)), except that here the elastomer specimen was subjected to a
rotating load rather than a reciprocating load. The rotating shaft was supported by a pair of
preloaded angular contact ball bearings so that the only significant deflection came from
the elastomer support. This test rig employed a forced vibration, resonant mass principle
in which rotating unbalance provided the exciting force. the test rig was designed to allow
the frequency of resonance to be modified by changing the amount of resonant mass
being supported by the elastomer cartridge specimen. Acquisition of data was, however,
not limited to the resonance condition of the system. In fact, data obtained at the
resonance frequency of each mass-spring combinations was just one of several data
points acquired at several vibration frequencies around resonance, where significant
amplitude ratios between shaft and resonant mass vibration existed.
The test method used in Report 5 [Smalley et al., 1977] employs a pair of 0-rings to
support a mass on an electromagnetic shake table so that the mass and 0-rings form a
one degree of freedom, damped system for vertical motion. When the table is shaken
near the resonance frequency of the 0-ring mass system, the relative motion across the
ring is an amplification of the table motion and there is a phase shift across the elastomer
of the order of 90 degrees. The important sensors for the test method are accelerometer
signals, at a particular frequency, and from the known value of supported mass, the
stiffness and damping are inferred.
155
5.4.1 Material Selection
The initial selection of polybutadyne, which Is classed as a Broad Temperature Range

(BTR) material, was primarily based upon the desire to test a material that was not very
sensitive to small variation in ambient temperature variations induced in the sample
through vbration testing [Gupta et al., 1974; Smalley and Tessarzik, 1975; Darlow and
Smalley, 1977; Smalley et al., 1977].
This material, which carried the manufacturers (Nichols Engineering Inc., Shelton,
Connecticut) designation NEX1564, has a nominal hardness of 70 durometers (or Shore
A hardness) and has the highest hardness and damping of those available from the
manufacturer. The values for G', G" and G * were determined from the results of shear
specimen tests conducted over a frequency range from 100 to 1000 Hi., for a constant
energy dissipation rate.
5.4.2 Cylindrical Button Cartridges Test Samples
Each of the test specimens consisted of two rows of three cylindrical elastomer
compression elements (or buttons ) evenly distributed around the origin of the rotating
force vector (see Fig.(5.9)) [Gupta et al., 1974; Smalley and Tessarzik, 1975]. The
buttons were held between the flat surfaces of two shell structures.
Top View Section "AA"
Fig.(5.9) Cylindrical Button Cartridge
In all cases buttons were cemented to the inner housing and in all but one case, the
buttons were also cemented to the outer housing. The buttons were located such that a
line drawn from the origin of the rotating force vector through the centre of any button
156

would form a light angle with either face of that button. Table (5.1) shows the selected
dimensions for the test configuration.
Inner Outer
diameter Button
diameter Button
height
Sample # diameter
1 0.0635 m 0.0699 m 0.0127 m 0.00318 m
2 0.0635 m 0.0683 m 0.0127 m 0.00238 m
3 0.0635 m 0.0730 m 0.0127 m 0.00476 m
Table (5.1) Dimensions of the buttons
Sample 3 is the one that was not cemented to the outer housing hence in theory it
should only undergo compression unlike Samples 1 and 2 which undergo both
compression and shear.
5.4.3 Ring Cartridge Test Samples
Each of the test specimens consisted of two continuous axisymmetric rings of rectangular
cross-section (see Fig.(5.10)) [Smalley and Tessarzik, 1975; Darlow and Smalley, 1977].
Each ring was held between two shell structures and cemented to the inner housing.
The outer housings were split to facilitate the assembly of these test elements and to
allow a small positive radial preload during testing.
Top View Section "AA"
444+
APAM
K.K.
.).:.).
:•:•X
Fig.(5.10) Ring Cartridge
Two sets of dimensions were selected for the test configuration as in Table (5.2).
157
Inner Outer
Sample # diameter diameter Thickness
1 0.064 m 0.073 m 0.0048 m
2 0.064 m 0.073 m 0.0024 m
Table (5.2) Dimensions of the ring cartridges
5.4.4 0-ring Test Sample
A pair of 0-rings with nominal outer diameter of 6.35 cm were tested [Smalley et al.,
1977]. Three different materials (Viton-70, Viton-90, Buna N-70) were used. A
parameter perturbation test programme was executed for each material. Imposed
squeeze and stretch, cross-sectional diameter, 0-ring groove diameter and temperature
were varied in turn, about a nominal value.
-ring housing
A
Bearing housing:,
Bearing
Dg
0-ring
(cross-sectional
diameter, d, inner
diameter, D
Fig.(5.11) 0-ring (after Smalley et al. (1977))
Nominal values are as follows:
Temperature : 25 degrees of Celsius

Vibration amplitude : 7.62x10-6 m
Squeeze :15%
158
Stretch :5%
Cross-sectional diameter : 0.00353 m
Groove width :135% of original cross-sectional diameter
0-ring outer diameter : 0.0635 m
The stretch and squeeze are defined as follows:
D— D.
Stretch =100(-1—d%
Di
(5.42)
Squeeze
d) (5.43)
5.5 Representation of the Experimental Data
Tan (8), the tangent of the phase angle between the real and imaginary part of complex
I II I II
stiffness, K 2 1K 1 in Fig.(5.1), or is the ratio of GIG, E 1E and is often used to describe

the damping characteristics. Various combinations of the above terms may be used to
describe a given rubbery material [Sommer and Meyer, 1973].
G s .= G'+ i G" [Shear Modulus, N/m2]

Es = E'+ i E" [Young Modulus, N/m2]
K s =K1 +i K2 [Stiffness, N/m]
C =K/CO [Damping Coefficient, Ns/m]
where CO is frequency in radians per second.
Sommer and Meyer [1973] gave the brief summary of the changes in the spring stiffness
and damping coefficients produced by variations in rubbers, fillers, oils, level of cross-
linking and processing factors such as mixing, curing, and storage.
Basically it can be said that the real part of the complex stiffness (K1) and the imaginary
part of the complex stiffness (K2 ) (i.e. damping of the elastomer) can be obtained from
the relationships between the force and displacement in the frequency domain.
K s = K i ( co)+ i K2(CD)
where K s is a complex number. This gives rise to discrete data points. For mathematical
159
purposes It is convenient to seek, for an element, a continuous relationship between

force and displacement which closely matches experimental behaviour. There are several
possible approaches available to this problem.
5.5.1 Generalised Viscoelastic Models
An approach to representing the variation of stiffness and damping properties of an

elastomer element with frequency, is to replace the elastomer with a mechanical
analogue, consisting of a series of springs and viscous dash-pots. The simplest of these
models are the Maxwell and Voigt elements (see Fig.(5.12))
However, in practise those two models are generally combined. For more than one
element model, the Voigt element model is more convenient since it readily defines
compliance. In this thesis, for computational purposes, a generalised Voigt model system
is used as shown in Fig.(5.13) (see Appendix 6).

Maxwell Element Volgt Element
Fig.(5.12) Basic Elements of Viscoelastic Model
For one Voigt element [Nahsif, 1985]:
le(to)= K1(co)+iK2(co) (5.44)
Fig.(5.13) General Voigt Model
160
The equivalent complex compliance for the system:
Ha = II: + H;+11;+...+H: (5.45)
where
(5.46)
Substituting Equ.(5.44) into Equ.(5.46) will give the complex compliance for a single
Voigt element:
Into). HI M+ i H2(w)
(5.47)
where
Hi(co)= ic 4.K ic
(5.48)
2
H (w)= —K-7---2-
1q + K2
(5.49)
5.5.2 Generalised Force-Displacement Relationships
An alternative approach for representing the variation with frequency is to assume a

general partial differential relationship between force and displacement:
{a 0 + al(1) + a2(1:-2)+ . . . + a „ig F

a.
.[bo+b1(1)-1-b2()+...+4-w,)]8
(5.50)
For sinusoidal loading and displacement a close connection between this generalised
force displacement relationship and the generalised viscoelastic models exists (see
Appendix 6).
5.5.3 Direct Fit of Data to Mathematical Expressions
A direct approach to the problem is to impose upon the data some correlation function
and to seek those coefficients which give the best match between the function and the
161
data. Polynomial, trigonometrical, or power law relationships are some of the options
available for this approach.
Gupta et al [1974] showed that the variation of complex dynamic stiffness with frequency
can be expressed in the form of a power function of the form shown below:
Y = A cos (5.51)
where A and B are constants.
Smalley and Tessarzik [1975] curve fitted the data with consideration of geometrical
shape factors. Although this is also a prediction method, if K 1 (co) and K 2 (CO) in Equ.(5.1)
are calculated from the functions below, quite accurate curve fitting can be satisfied
[Smalley and Tessarzik, 1975]:
rDN2 r
IC(co)= 3G'(co) I-4 [1+ ())sl
—h (5.52)
irD2 N
K2(co)= 3G" (co) —7t-- [113
+ " (co) s21
h (5.53)
where G ' is the real part and G " is the imaginary part of the complex shear modulus G.
P is the real part and 13 is the imaginary part of the geometrical factor and S is a shape
factor defined as the ratio of the loaded area to the unloaded (see Appendix 5). The
experiment done by Smalley&Tessarzik[1975] and Darlow and Smalley [1977] showed
,j
that if G ,G , p and /3 in Equs.(5.3&5.4) are expressed at constant temperature of 32
degrees Celsius as below, the error for dynamic stiffness would be less than 50% and
with damping the error would be less than 20%.
G' (co) = 3.686 x106 co 6.26" N/ rn2 (5.54)
G"(co)= 8.333 x106 w° N / m2 (5.55)
P'(co)= 12.33 co-6•296 (5.56)
fl" (co) =1.726 0)61'119 (5.57)
162
5.6 Brief Description of the Subroutine
A subroutine called DAMPER was written in Fortran77 to find the coefficients of

damping and stiffness for the equivalent mechanical system. The number of elements in
the Voigt model, the frequency range and the frequency step size are required as inputs.
The subroutine basically minimises the error between the theoretical and experimental
results.
Let a generalised the Voigt model be defined by n Voigt elements connected in series
with a constant stiffness coefficient Ko and the i th the Voigt element have a stiffness
coefficient K and a damping coefficient ci. If this model is applied to solid materials such
as elastomers and there is no preload (K 0 -0), the stiffness coefficients in all elements
must have non-zero values. For this model the equivalent complex compliance is the
summation of the complex compliance of each Voigt element and can be given as:
{-0
11*(co)= Tel +b
i
ml 14151
1 }- iltbi
(5.58)
where n is the number of Voigt elements and
(5.59)
ci (5.60)
The error function can be given as the sum of the squares of the difference between the
theoretical and experimental complex compliance [Gupta et al., 1974]:
E= w±[H, (co) - H1,..(C0)]2 + w2i[ 112 (c0)-112.(0))12

i-I " (5.61)
where k is the number of discrete point used,w / and w2 are weighting functions and
Hi (co) and H2 (co) for the theoretical model are the same as in Equs.(5.48&5.49) for a
H
single Voigt element for example. H1 (co) and 2 (0.) for the experimental results are in
163
the form of stiffness and damping, and can easily be converted into the complex
compliance form as in Equ.(5.47).
To minimise the error function the least square method is used. Detailed information
about the technique can be found in Ref. [Gupta et al., 1974].
The first step In solving the equations, is to obtain the constant stiffness K 0. of the Voigt
element created by preloading. For a perfect solution, the experimental and theoretical
results should be the same. Hence from Equ.(5.47) and Equ.(5.49):
( co)
L 1 = K1 i_Za__
PC0 +tuiTI i ± b, --a '- t -)-x
i=1 1 _a_ inl 1 4. ( CD ) Ki +Ki Ki2+1(22
1 A .n Lii) J (5.62)
Equating real and imaginary parts of Equ.(5.62):
[1+± 1 1+ 12
1 !CI. )_i
K2 fl
KL(1- 2 K2 12 12)}2
(5.63)
Note that this equation always results in a positive solution.
For b 1 the same procedure is followed with a small alteration to be able to introduce the
weighting functions w i and w 2 . The idea is that if a Voigt element being calculated is left
out, the equivalent of the rest of the elements plus the one exempted should be equal to
experimental result. Hence
A
(5.64)
where
A = wi [H, — Ho — 444. + w2 [H 2 — ±bi 021g j(12)

i=1„,
(5.65)
B = ( 11(n))2 + (g1(.0))2
(5.66)
164
in these equations
(.212
f (D ) = `1) )
1+ (fT)
2 (5.67)
and
( 12 )
AD ) = 2
1+(-
a))
(5.68)
Equ.(5.61) is called the sum of the squared deviations of the error and it has to be
minimised. For this:
dE
=0
d.C2 (5.69)
E should be differentiated with respect to all variables in turn and all of them should
simultaneously be equal to zero. Hence;
dE
.p.=0 (i =1,n)
af2i
(5.70)
Since the damping, or in other words 12 in the equations, is nonlinear, the solution must
be obtained by a numerical method. In this study, the Newton-Raphson Method was
employed to solve the equations. For successful solutions below, a Jacobian matrix must
also be solved.
(2C1 .
112•1-12.}=-4P1
t+
j=1. (5.71)
For the Newton-Raphson Method, the initial values are very important as the solution can
converge to different results. Therefore, the initial value for £2 is taken as below.
165
K(0) ay
Q0==
Co K2(tin (5.72)
where
= °min + °max
2 (5.73)
The iterative solution procedure can be summarised as follows: The linear parameters,
K0 , K (where 1Si Sn) are solved for any set of iterates r4 i .,., (where 1 Si Sn, 1Sj Sn
and i ) are obtained by solving the linear algebraic equations (Equ.(5.71)).
It should be noted that such a least squares analysis may sometimes result in negative
coefficients, which are physically meaningless. Under these conditions, it might seem
reasonable to calculate the linear parameters by carrying out a simple linear regression.
5.7 Modelling of The System
The elastomer damper investigated in this thesis may be in the form of cylindrical buttons,
ring cartridges or 0-rings. Such a system with external dampers can be modelled as in
Fig.(5.14) where mo, simulates the mass of outer rings and K e and ce are the stiffness
and damping of elastomer dampers respectively.
Fig. (5.14 Simulation of the system with external dampers
The equations of motion in a three degrees of freedom system can be written as follows:
M -1-EWs cos(ai)cos(6)— Mg =WO

(5.74)
166
9to 9ti
Bdb + bi
or
Fig.(5.15) Deflection of the i th ball with external dampers
my +E IV; cos (a)sin (69 =WO

(5.75)
mi+E Wisin(a )=WO

(5.76)
m 0,12 + Ce .i2 + Kx2 — yWi cos (ai)cos (9 i ) — me,g =a2(t)

(5.77)
M ji2 + C,572 + IC, — I Wi cos (a i) sin (0 i) = Qh(t)

(5.78)
A 1 + C1 1 2 + K1 1471 sin (a)=Q.,(t)

(5.79)
where W denotes the restoring force of the i th ball as given in Equ.(4.1). From
Fig.(5.15) which is the modified form of Fig.(3.10)), the deflection, Si and contact angle,
ai of the i th ball will respectively be
167
(B db sin(a0) + zo+ (z — z2))2

45 i = B db
+((x — x2)cos(0) + (y — x)sin(ei)+ B dbcos(adY
(5.80)
B db sin(a 0) + zo + (z — z2)
a .—
(x — x2)cos(04)+ (y — x)sin(0;)+ B dbcos(a0) (5.81)
The solution procedure is similar to that employed in Chapter 3.
5.8 Excitation Frequency
As was explained earlier the dynamic properties of elastomers vary with the frequency
they are subjected to. This means the current frequency they are subjected to will
determine the system steady-state vibration characteristics. During transient vibrations
period, the excitation frequency the elastomers dampers see will continuously change,
hence the steady-state vibrations may be totally different from the one calculated.
Therefore the instantaneous speed, the elastomer dampers are subjected to, should be
introduced to the system.
time, t
Fig.(5.16) Transient and the steady-state oscillations
However all available data is obtained for the cyclic excitation forces. Since the oscillation
during the transient vibrations are not cyclic, the data available are not useful. Therefore
during the computation the following procedure if followed:
For any motion of the shaft-ball bearing system such as in this case, the amplitude of the
vibrations will have instantaneous maximum and minimum points as shown in Fig.(5.16).
168
In Fig.(5.16) the maximum amplitudes are indicated with 1 and minimum amplitudes are
with 0. It is known that for the points is and Os, ± 2(t)= 0. and from 1 to 0, ±2(t) <0. and
from 0 to 1,12(0 > 0.

0 Amplitdute, y(t)
nn••
Fig. (5.1 Oscillations of the centre of the shaft
Therefore the outer race centre movement is controlled and for the first 1 2(t) = 0., the
time, to is recorded. When the next .±2(t) = 0. Is observed, the time, t is also recorded.
With the assumption that this movement will repeat itself, the first excitation frequency is
found as follows:
tEri =
(5.82)
where 44 = — to. Until the next ..i2(t) = 0. point, this will be the excitation frequency
and it will follow as:
OT
ul =
iti (5.83)
where At, = t, — . When the steady-state oscillations are reached, if at all:
169
i rii_i
Er EE E (0„,
(5.84)
where wor is the speed of the outer race centre. Note that the same result may be
reached using y2(t)oscillations instead of x2(t) (see Fig.(5.17)).
Closure
Since, design-orientated data on the dynamic behaviour of elastomers is limited and
there is no adequate theory describing the dynamic behaviour of elastomers, a method of
obtaining the stiffness and damping characteristics of elastomers from their material
properties (i.e., shear or Young's modulus) for different designs namely, cylindrical
buttons, cartridges and 0-rings was developed in this chapter. The results obtained will
be compared with other researchers' predictions and used in the vibration model available
to see the overall effect of elastomer dampers on the vibration characteristics of a shaft
supported by angular contact ball bearings in the forthcoming Results and Discussion
chapter.
170
CHAPTER 6
RESULTS AND DISCUSSION:

BALL BEARINGS
6.1 Introduction
In the previous three chapters, the necessary theories for studying the dynamics of ball
bearings were established. Employing these theories, a computer simulation of a shaft-
bearing assembly was obtained. In Chapter 4 some anomalies of ball bearings, namely
waviness, off-sized ball, misalignment and local point defects were also simulated. In this
chapter these simulation results will be compared with the other sources of experimental
and theoretical predictions. Results and discussions will be presented in the same order
as in Chapter 4 i.e., natural frequency of the system, ball passage frequency, waviness,
off-sized ball, misalignment and defects on the rolling surfaces.
6.2 Shaft-Ball Bearing Specifications
In order to test the simulation programme a grinding spindle is chosen, as seen in

Fig.(6.1). The shaft (2) is driven by a pulley (5) in order to operate the grinder (1). The
shaft is supported by a pair of precision angular contact ball bearings (3 and 4). The
specifications of the bearings and the shaft were chosen as the Aini's experimental
arrangement [Aini, 1990] since the particulars of both the shaft and bearings were readily
available.
Particulars of the- bearings (see Fig.(6.2)):
0 Type : Angular contact ball bearings

0 Inner race bore : 0.04 m
0 Inner race diameter : 0.046 m
171
Chapter 6 Results and Discussion: Ball Bearings
172
0 Outer race diameter : 0.062 m

0 Outer race outside diameter :0.068 m
0 Inner race radius curvature :0.00408 m
0 Outer race radius curvature :0.00461 m
0 Bearing width :0.015 m
0 Ball diameter (db) : 0.00794 m
0 Maximum number of balls :16
0 Unloaded contact angle :15 deg.
0 Pitch diameter of the ball set : 0.054 m
0 Both bearings are identical.
Fig.(6.2) Ball Bearing
Particulars of the shaft:
0 Diameter of the shaft between the bearings : 0.04 m

0 Length of the shaft : 0.421 m
0 Mass of the shaft : 5.5 kg
0 Distance between the support bearings : 0.215 m
0 Position of the LHS bearing from the C.G. : 0.0875 m
0 Position of the RHS bearing from the C.G. : 0.174 m
0 Moment of inertia of the shaft about x ory- axis : 0.05177 kg m2
0 Moment of inertia of the shaft about z- axis : 0.0044 kg m2
173
6.3 System Natural Frequency
First a bearing with five balls was considered in order to calculate the variation in the
natural frequency of the system due to different positions of ball set. Since for five balls,
the positions of balls repeat themselves after each 72 degrees, a span of 72 degrees of
position change was observed as shown in Fig.(6.3). In order to calculate the natural
frequency of the system the procedure below was followed: the shaft speed was set to
zero and the shaft centre was released from an arbitrary position. The natural frequency
values were calculated from the transient vibrations as the system must vibrate at its
natural frequency because there is not any external force or damping. The natural
frequency values were obtained as described in Appendix 2.
Off-set angle, deg.
Fig.(6.3) The effect of the position of balls on the natural frequency (m.5, Pr.5 N)
As is seen in Fig.(6.3) the change in the natural frequency is cyclic and varies with the off-
set angle (see Fig.6.12 for the definition of off-set angle) as was also observed by
Shimizu and Tamura [1966; 1967; 1968]. Some points in Fig.(6.3) seem to be slightly
misplaced. This is due to an accuracy problem in measuring the natural frequency and
associated with the time step size. When the time increment was decreased, better
results (i.e. a smoother curve) were obtained with the cost of longer computing time.
Then the position of the ball set was varied for 360 degrees in order to obtain a full picture
for a cycle. Fig.(6.4) shows that the natural frequency changes cyclically m times (where
m is the number of balls). This was further observed with a bearing with 8 balls (see
Fig.(6.5)). It was also observed that as the number of balls increases the change in the
amplitude of the natural frequency decreases because the bearing becomes stiffer. This
is particularly correct for for a bearing with relatively higher number of balls (e.g. 16 balls)
174
where the variation in the natural frequency is almost insignificant. This small variation of
natural frequency will be discussed further later In this section.
90 180 270
Off-set angle, deg.
Fig.(6.4) The effect of the position of balls on the natural frequency (m-5, Pr-5 N)
484
482
481
90 180
p v AA)
270 360
Off-set angle, deg.
Fig. (6.5) The effect of the position of balls on the natural frequency (m-8, Pr=5 N)
175
At this point a further discussion on the experimental results from Yamamoto and lshida
[1974] will be useful. They experimentally showed that the natural frequency changes
cyclically with the position of the balls once per cage rotation (see Fig.(2.8)). This seems
likely to be due to wrongly selected experimental position intervals [Xistris et al.,1980]
e.g., as shown in Fig.(2.5). The possible result of wrongly selected intervals is also
shown in Fig.(6.6) for ball bearings. If experimental points are selected as shown with the
diamond shaped points in Fig.(6.6), the results obtained can be misleading since the
actual variation is totally different and shown by dotted line. The larger the number of balls
the greater the possibility of making this mistake, since relatively small changes in the
natural frequency and larger cyclic variations take place.
11. 482
z*g
481

o 90 180
27 0 360
Off-set angle, deg.
Rg.(6.6) The effect of wrongly selected experimental intervals

on the natural frequency (i78, Pr. 5 N)
Due to the nonlinear Hertzian contact conditions, the preload changes the natural
frequency. The effect of axial preload on the natural frequency is shown in Fig.(6.7). For
lower preloads the change is fast since the balls are relatively softer. As the preload is
increased the balls get stiffer and allow smaller deflections for the same amount of
preload. This is clearly observed in Fig.(6.7).
Fig.(6.8) shows the effect of changing the number of balls. As the number of balls are
increased the system get stiffer since larger number of balls support the shaft. For
relatively larger number of balls, the change in the natural frequency is relatively less,
since in this case larger number of balls in the heavily loaded rigion allow relatively small
deflections.
176
The approximate natural frequency for the system (see Appendix 10) can be written in
two forms:
1
co „= A(B 6 + - AO)
4
(6.1)
or
1 1 1
co.=
(1)11= A[ -T + - C2 A8]
Ci 4 (6.2)
where
3 v-,
"
A =f —LK (cos(e i -13)cos2 a i + sin 2 a i)cos a i cos ei)
M (6.3)
Ra
B-
mKsinct,
(6.4)
(6.5)
1 000
900 -
800 -
0
700 -
•
600 -
• •• 71,--V- VV - 11- 1111 g •
5 0 100 150 200

Axial Preload, N.
Fig.(6.7) The effect of the axial preload on the natural frequency (m-8)
177
▪
In Equ.(6.1) the first term on the RHS will be effective for relatively larger values of B,
whereas for small values of B, the second term in RHS will also contribute to the change.
Fig.(6.7) shows the behaviour suggested by Equ.(6.1). The curve shows a big similarity
with that of a curve obtained only by the first term on the RHS of Equ.(6.1). This is
particularly true for preload values greater than 30 N. In our case das being very small
also contributes to making the first term dominant in Equ.(6.1).
In order to study the effect of the number of balls on the natural frequency, Equ.(6.1) can
be written in the form of Equ.(6.2) which is more convenient for this purpose since the
effect of the number of balls has a proportional relation in Equ.(6.2). It is obvious that the
first term on the RHS is less dominant for changing the number of balls as the number of
balls is always greater than 1. It will be more so when the number of balls is increased.
Hence the second term on the RHS of Equ.(6.2) will be more dominant implying the
behaviour in Fig.(6.8). This curve is similar to that obtained only by the second term on
the RHS of Equ.(6.2).
Figs.(6.786.8) were obtained from the simulation model by a similar method to that which
was explained earlier at the beginning of this section. For preloads the number of balls
was assumed to be 8 and their position were kept same during the calculations for both
cases and was such that the first ball was always set to the bottom of the bearings. The
effect of this can be seen in Fig.(6.8) as the natural frequency values flicker slightly for
odd and even numbers of balls.
650 •
•
600 =
550 -
500
450
400 -
350 , • . . - ....... = ...... • - -

4 6 8 10 12 14 16
Number of Balls
Fig. (6.8) The effect of the number of balls on the natural frequency (Pr=5 N)
178
6.4 Ball Passage Frequency

Ball passage frequency is one of the most important frequencies of rolling element-shaft
systems (see Chapter 1 for its definition and Chapter 4 for its calculation). In order to
display its effect, a defect free system was considered with three degrees of freedom.
The particulars of the arrangement was given earlier in this chapter (see Section 6.2). The
axial preload value was set to 5 N, the mass centre was displaced xo .1 .i.un, yo = 0.1 gm
and zo =0.01 gm and oscillations along the x, y and z axes were recorded when the shaft
was rotating at a speed of 5000 rpm. For this speed the BPF is 280 Hz (from Equ.(4.21))
and the natural frequency for the system is about 500 Hz (from Figs.(6.886.9)).
1.06
1.06 -
11
I I I I I
1.04
g_
1 1.02
1.- 1.00 i
0.98 - I I
0.96 - . - -
0.30 0.31 0.32 0.33 0.34
time, s
0.010
d
0.008 -.
b
0.006 -
I
0.004 -
a
0.002 -
C
9
0.000
0 200 400 600 800 1000
frequency, Hz
Fig. (6.9) Vibrations along the x-axis in the time and its FF7'
(m=8, Pr-5 N, n-5000 rpm, (obp-280 Hz, (),-500 Hz)
179
The results obtained in both the time and frequency domains are given below. Fig.(6.9)
shows the oscillations along the x-axis. The effect of the natural frequency along the z-
axis appears at about 200 Hz (a) (see also Fig. (6.11)).
0.15
II I Ihpi I i
0.05 n
r I
-0.05
V i I Ii
-0.15
0.30 0.31 0.32 0.33 0.34
time, s
0.006
b C
0.004 -
I
• a
0.002 -
d a
0.000
0 200 400 600 800 1 000
frequency, Hz.
Fig.(6.10) Vibrations along they-axis in the time and its FFT

(m=8, Pr-5 N, n=5000 rpm, 4 p-280 Hz, con-500 Hz)
The second most dominant peak (c) after the peak at the natural frequency in the
vibration spectrum is at the BPF. Its first superharmonic (24 p) (e) also appears in the
spectrum. Although the ball passage effect is relatively small in amplitude, it should be
noted that when it coincides with the natural frequency the system resonates, as will be
shown later in this chapter.
180
The natural frequency is about 510 Hz. (see Fig.(6.7)) for a shaft supported by bearings
with 8 balls when an axial preload of 5 N is applied on the system. The peak at the natural
frequency (d) is at about 490 Hz. This slight difference comes from the operating
conditions, since in the calculations of natural frequency (see previous section) the
system assumed to be static. As is also seen in Fig.(6.9) the first superharmonic of the
natural frequency (2con )(g) appears in the frequency spectrum.
0.015
0.005
a
i
1
-0.005
-0.015
0.15 0.20 0.125 0.30 0.35
time. is
0.0100
I
0.0075
0.0025
0.0000 rimimilme—flor" • • 1 • • ' I

o 100 200 300 400 500
frequency, Hz.
Fig.(6.11) Vibrations along the z-axis in the time and its FFT
(m.8, Pr.5 N, n.5000 rpm, cobp.280 Hz, con-200 Hz)
181
It is also important to note that the second (c) and third (f) superharmonics of the effect of
the natural frequency along the z-axis also appear in the spectrum. Although the
frequency spectrum for the x-oscillations presented in Fig.(6.9) is the simplest case,
since there are no defects present in the system, the effects of vibrations along the y and
z axes make the spectrum relatively complicated. This is the indicator of the complex
nature of vibrations associated with ball bearings.
Fig.(6.10) is the vibration spectrum for y-oscillations. It is similar to Fig.(6.9) with the peaks
at the natural frequency along the y-axis (c), the effect of the natural frequency in the z
direction (a) and its superharmonic (0), the BPF (b) and its first superharmonic (d). These
frequencies are almost the same as along the x-axis. This is understandable since the
system along the x and y axes are symmetrical with the difference that the force due to
mass of the shaft acts only along the x-axis and hence causes a difference in the vibration
amplitudes.
The z-oscillations produce the frequency spectrum in Fig.(6.11). Due to the small initial
displacement in the z-direction and because opposing bearings were identical, the
resulting vibrations had relatively small amplitudes. The only apparent peak in the
frequency spectrum is at the system natural frequency along the z-axis. The z-oscillations
will be further investigated in this chapter.
In order to study the effect of the BPF in a more detailed form, the angular contact ball
bearing employed in this thesis was reduced to a radial ball bearing for the time being and
modelled as in Fig.(6.12). The angles and reference axes were set as in Fig.(3.15) in
Chapter 3. The off-set angle in Fig.(6.12) is an arbitrary reference point on the cage.
Balls were radially preloaded in order to ensure the continues contact of all balls and the
raceways, since otherwise, a chaotic behaviour may be observed [Gad et al., 19841. The
preloaded deflection for each ball, 450 , was assumed to be 5 gm. The centre of the inner
race was shifted 2 gm in the x direction and 2 p.m in the y direction with respect to the
outer race centre. For different off-set angles (see Fig.(6.12)), the total static net force
and the phase angle were recorded.
Fig.(6.13) shows the change in the total net force and phase angle for a bearing with 5
balls. The off-set angle was changed over a span of 360 degrees. It is shown that the
total force vector and its direction (i.e., phase angle) changes m times (m is the number of
balls) implying vibrations at the BPF. As the number of balls increases the amplitude of
change in the force vector and phase angle becomes small, implying small amplitude
182
variations with relatively larger number of balls. The experimental force monitoring of ball
bearings gave similar results (see Fig.(2.19)) in Chapter 2 and the paper by Gk5dmer [BBJ
225]). This model will be heavily used for the waviness and off-sized ball anomalies later
in this chapter in order to show their effects as well.
Fig. (6.12) Definition of Total Net Force, Phase Angle and Off-set Angle
The change in the phase angle follows the same pattern as seen in Fig.(6.13). This
implies a similar excitation force (i.e. excitation due to the BPF) in the y direction and
hence vibrations at the BPF. The effect of the BPF is apparent in this simplified model.
When the forcing frequency coincides with the natural frequency the system undergoes
severe vibrations. In order to study this effect the results presented in Figs.(6.9 through
6.11) are reinvestigated after the natural frequency effect is damped out. For this,
different arbitrary damping coefficients were introduced to the system and 300 Ns/m was
finally chosen since the smaller coefficients took very long to get to steady-state
vibrations hence expensive computation, whereas larger damping coefficients did not
allow the important peaks to reveal themselves.
183

187. me 187 AC
18 r s i 1111,1111111111t1 18 7.8 . rsalleiMils..

78 NI fillilltillin 187 .78 : Pir .11
187.
E 187. 76 1111111111111111 187 .76 i
187. 74 EINEM : MEE 187 .74
u. 187, 72 187 .72 .
18 F.7 EINIIIIIIIIIIEN 18 7.7 :
187 68 IIIIIIIMITIMIll 187 .68 •: I. .4
187 INIMIlitinfilltill 187 .66 '

Ilhill
187 64 illialL111111111 187

120 180 240 300 3f
0 60
12 8 § 8 § V § s. 1i ,,,
1:1-
45.025 ; 44 4 4 4 v 4 *•
45.02 orionoranwrii
45.015 IML I Illitittillii
45.01 Ini : 111111 : : :
— 45 005 1541111111F1116 :
r 45 Ina : : : ERNI
: :U11: : Hi :
44 995
44.99 itilialt1111121111
44.985 EH : : : ME : NE
44.98
1111111111111•111141111
44.975
U111111111LORMIlil
0 60 120 180 240 300 360
Off-set angle (deg.)
Fig.(6.13) The change in the total net force and phase angle for different off-set angles
(m.5)
184
0.008
a
0.006
I
-
b
0.002
0.000 _.
o 200
• I
400
• • A
600
- • 1
800 1 000
frequency, Hz.
Fig.(6.14) FFT of the vibrations in the x-axis after the natural frequency is damped out
(m=8, Pr=5 N, n=5000 rpm, cobp=280 Hz, c4=500 Hz, c=300 Ns/m)
A time step is an important parameter in the solution of the equations of motion. There
should be sufficient number of points to describe accurately an oscillation of the shaft
centre. On the other hand unnecessarily larger points will cause a delay in reaching to
steady-state vibrations. The longer the time to reach steady state vibrations, the longer
CPU time needed. This means more expensive computation. Therefore before each
run, a check for the number of points was made to ensure that about 40 points will be
present in a complete oscillation. For example for the shaft speed of 10000 rpm the time
step is taken 0.00005 s. Here the BPF for the system with 8 balls is taken as a basis since
the resulting oscillations will be at this speed. Therefore there are 36 points in an
oscillation for the given example.
The shaft was again rotated at a speed of 5000 rpm and the x and y oscillations were
recorded. The FFT of the oscillations are shown in Figs.(6.14 and 6.15) for the x and y
oscillations respectively. The BPF for this system for the given speed is 280 Hz which is
apparent in Fig.(6.14) (a). Its first superharmonic (b) is also present at 560 Hz. For the y
oscillations presented in Fig.(6.15) the same pattern is observed with the BPF (a) at 280
Hz, the first, second and third superharrnonics of the BPF (c, d and a respectively) are
observed at 560 Hz, 840 Hz and 1120 Hz respectively.
185
0.006
0.006 -
0.004-
E
-o
0.003
0.002 -
0.001 -
d e
0.000 - bA . . . , I,' . • 1 - • .
200 400 600 800 1000 1200

frequency, Hz.
Fig.(6.15) FFT of the vibrations in the y-axis after the natural frequency is damped out
(m=8, Pr=5 N, n=5000 rpm, cobp=280 Hz, con=500 Hz, c=300 Ns/m)
6.4.1 Effect of Varying the Number of Balls
Increasing the number of balls means increasing the system stiffness and reducing the
vibration amplitude. This is exhibited in Fig.(6.16) which shows the phase-plane diagram
of the x oscillations for different number of balls, when the shaft is released from an
arbitrary position for free oscillations without damping. When the number of balls are
increased, the centre of oscillations approaches zero implying a stiffer system. From this
it can be predicted that increasing the number of balls will reduce the effect of the BPF.
However, the BPF is the cage speed times the number of balls (see Chapter 4), therefore
It coincides with the natural frequency at a lower shaft speed.
The preload has a significant effect on the system vibrations. When an axial preload is
applied to a small number of balls its effect will be relatively greater. Therefore the axial
preload was chosen to be 10 N (very low), so that its contribution would be negligible.
By introducing damping to the system, the transient vibrations were eliminated and peak
to peak steady-state amplitudes were recorded as shown in Fig.(6.17).
186
Fig.(6.17) shows the effect of varying the number of balls. For a relatively small number of
balls, the peak to peak amplitudes of vibrations at the BPF is more significant. Since a
small number of balls supporting the shaft, the natural frequency of the system is
relatively low. For example, for the system having bearings with 5 balls the natural
frequency coincides with the BPF at a shaft speed of 12700 rpm. This means the system
natural frequency is 445 Hz. This value can also be obtained from Fig.(6.8) by
considering an axial preload of 10 N. The effect of the first subhamionic of the BPF (1/2
co,)
b also appears in a spectrum at the speed of 6350 rpm.
2500
• • • • • • • m-10
• •
• •• • • • • • • m-12
• • • • • • •• • m=15
• • • •o•
• • 0 0 o 0 o •-_
• • •
•
o
o vn-•.•
• • • o
• • • •o
• • •• 0 0 o•
• • •• 000 ••••
• • • • • •• • •
• •
• • • ••• •
• •
• • • _ v •
-2500
0.0 0.5 1.0
x deflection, p.m
F7g.(6.16) The effect of varying the number of balls

(Pr=10 N, ns=0 rpm, the reference ball is on load line)
For a system having bearings with 8 balls, the peak to peak vibration amplitude decreases
and the natural frequency is pushed to a higher value of 510 Hz and it coincides with the
BPF at a speed of 9200 rpm. Although the natural frequency increases with an increase
in the number of balls, the BPF coincides with the natural frequency at a lower shaft
speed as explained earlier. The first superharrnonic of the BPF (2 cobp)also appear at a
speed of 18400 rpm while the first and second subharmonics (1/24p and 1I3cobp
respectively) clearly exhibit themselves at the speeds of 4600 rpm and 3070 rpm
respectively.
187
E 8.00e-7
m-5
1 6.00e-7 •
1 4.00e-7 -
1
9- 2.00e-7 -
1
a 0.000+0 \'%n.„,.,,_....„,
0 5000 10000 15000 20000
E 8.00e-8 .
d
3 6.00e-8 :
=
o.
E
as 4.00e-8 :
%
a 2.00e-8 :
_4c
i
0.00e+0 • - • 1 • •
0 5000 10000 15000 20000
3.00e-8
m-10
2.00e-8 -
1.00e-8 -
0.00e+0
.4.0AruSL...I..".r--.----"—^--'•—'
g i
5000 10000 15000 20000
2.00e-8
m-12
I
0.00e+0
AphAk....../

5000 10000 15000 20000
O 5000 10000 15000 20000

rotational shaft speed, r.p.m
Fig. (6.17) The effect of varying the number of balls on the BPF (Pr. 10 N, c=300 Ns/m)
188

m5
m=8
m=12
10 - 1 ° . . . . .

0 5000 10000 15000 20000
rotational speed, r.p.m
Fig.(6.18) The effect of varying the number of balls (Pr.10 N, c..300 Ns/m)
When the number of balls are further increased, a peak at the BPF always appears in the
vibration spectrum but at relatively lower peak to peak amplitudes as seen in Fig.(6.18).
While the effects of superharmonics disappear, the subharmonic effects get more
influential. In the case of 15 balls, the first, second, third and fourth subhamionic effects
are all apparent in the spectrum with relatively smaller amplitudes as seen in Fig.(6.18).
Effect of Varying the Preload

The preload is assumed to be axially applied through springs. When a preload is applied,
the initial contact angle and deflections will change. Increasing preload therefore means
increasing the difference between the constant initial contact angle (a0 )and the
preloaded contact angle (ap ). This will cause an increase in the initial axial displacement
zo. The larger the initial deflection, the more dominant it will be in the deflection equation
(Equ.(3.26)). Therefore for larger preloads, the vibration amplitudes associated with the
BPF will be reduced.
1
2 i 21
8: = [( 13 db sin(a0 )+ zo + z) +(B db cos(a0 )+ 3,) -2 - B db
(3.26)
189

8.00e-8
I Pr-10 N I
6.00e-8 -
4.00e-8 -
2.00e-8 -
0.000+0 - • • • r—r-1.".71L

0 5000 10000 15000 20000
1.00e-8
8.00e-9-
.
6.00e-9-
4.00e-9-
2.00e-9 -
0.00e+0
0 5000 10000 15000 20000
4.00e-8
IPr-50 NI
3.00e-8 -
2.00e-8 -
1.00e-8 -
0.008+0 A-r ,

. e . 431

0 5000 10000 15000 20000
4.008-8
IPr.100
3.000-8-
2.00e-8-
,
1.00e-8-
. 0.00e+0 .
0 5000 10000 15000 20000
rotational shaft speed, r.p.m.
Fig.(6.19) Frequency response showing the effect of varying the preload

(m-8, c-300 Ns/m)
190
Fig.(6.19) shows the set of results for a constant number of balls (m-8) and mass (M-5.5
kg). When there is an increase in the preload from 10 N to 30 N, there is a sharp decrease
in the peak to peak amplitude at the resonant frequency (i.e. the BPF) as also proven
experimentally [Wardle and Poon, 1983]. This can be predicted because the bearings
get stiffer and they allow lower vibration amplitudes in the radial and axial directions. The
effect of the natural frequency in the z direction is increased as observed at the natural
frequency of 3000 rpm.
In Fig.(6.19) for a preload value of 50 N, which is given as the low preload value by the
manufacturer, the amplitudes of the vibrations at the subharmonics of 6550 rpm and 3275
rpm show no decrease and are slightly higher than the amplitude at the BPF. The effect
of the natural frequency in the z direction is dominant in the spectrum at a shaft speed of
3050 rpm. For a heavy preload value (100 N), as described by the manufacturer, the
effect of the BPF is again dominant with its first and second subharmonics as shown in
Fig.(6.19). The peak to peak amplitude at the resonant frequency shows an increase.
This is because the heavy preload resulted in a high stiffness and the effect of damping
force has lessened because of lower velocities.
191
6.5 Waviness
At relatively low speeds waviness causes few problems with precision bearings [Thomas,
1982]. Waviness only becomes a serious problem when components possess relatively
high-amplitude features such as chatter marks, or where a precision bearing is fitted in a
housing with poor geometry. At high speeds, however, the situation can radically
change, a doubling of speed producing a doubling of vibration amplitude [Thomas,
1982]. A further effect with increased speed is causing the BPF and its harmonics to
appear in the spectrum as the excitation frequency (due to "wave-pass") near the BPF
and can resonate with it. The amplitude of vibration depends on the relation between the
rolling elements and the number of waves of the circumference of the ring.
fia jiiii_ilui.e.r_riag.malinlis
The same procedure of observing the total net force and the phase angle described
earlier in this chapter (see Fig.(6.12)) was applied to a bearing with a wavy outer race
surface in order to study the resulting force variation due to waviness. Since the inner
ring rotation is difficult to model in this manner, the inner ring waviness was not studied
with this model. The waviness amplitude was set to 2 Lan and the number of waves round
the outer race circumference was varied for a bearing with 8 balls. The total net force and
phase angle changes were recorded, as shown in Fig.(6.20).
For a relatively low number of waves (e.g. 1, 2, 3, 4 or 5 waves) the effect of waviness is
small as shown in Fig.(6.20). However, when investigated closely, the total net force and
phase angle are changing at a period equalling of the number of balls (i.e. 8 times in this
case). When the number of waves increased to 6, the change in both the total net force
and phase angle follow almost a sinusoidal pattern with a relatively higher amplitude as
shown in Fig.(6.20).
For seven waves the variation in both the total net force and phase angle are large, the
total net force varies between 100 N and 500 N and phase angle between 0 and 90
degrees. Phase angle almost linearly increases until 90 degrees and suddenly drops to 0
degrees in order to start linearly increasing again. This sharp change take place when the
total net force is about 100 N. Therefore the change of the total net force follows a
relatively smooth pattern as it goes through large forces. When the phase angle is 0
degrees the total force is equal to the force along the x-axis and when it is 90 degrees, it
is equal to the force along the y-axis. However, the total net force and phase angle still
change m times (m is the number of balls).
When the number of balls and the number of waves are equal, the change in the phase
angle is negligible. This was also experimentally observed by Wardle and Poon [1983]
and Wardle [1988]. Since the balls and the waves are in phase, the vibrations produced
will be important. If the change in the phase angle is investigated in detail, minute
changes may be observed as shown in Fig.(6.21). The jump from 0 to 90 degrees
192

▪
disappears and as the system suggest (see Fig.(6.12)) the phase angle is almost steady
at 45 degrees.
600 i
6
2
4cr, 350
100 „

0 90 180 270
600
• FI:31
•
0
I-, 350
INImr
ra •
O.
100 0
90 180 270 360
600 120
2 Cu;
0
• 350 60
Cu
Cu
Cu In
Cu
0.
100
0 90 180 270 360
600 1 20
N=6
• -0
a;
- 60
35° ",.0%Geo.ai%.0o0oOo40%.0* Cu
rt, u
C
Cu
0.
100
0 90 180 270 360
600 ji 120
N_7
Cu J[ 'FS
2
0
..5
c 350 - - 60 F
C Cu
76 co
ct'
co
.c
0.
100

90 180 270 360
Fig.(6.20 )...continued on the next page
193
600
z 1.71-7
e
e
0
-
T)
c 350
ra
o •
100 I I 0
0 90 180 270 360
600 120 .
z N.9
ai I
2o a;
13-, 350 60 E
c
ra
CI
100
0
I A Ai A A A
90 180 270 360
0
caw
.0as
13.
600 120
z ci)
a)
-a
a;
2 a;
o
i 350 60! c
0
C a)
ra cn
15 .0 co
o.
100 0
0 90 180 270 360
600 120 .
z 1
ai
2 6
o
15 350 60 E
c
.0
c73
-6
.
..
0.
100 0
0 90 180 270 360
600 120
Z cla
a)
a; 33
2 a;
o a)
r
z, 350 6010
C 0
ra as in
o .c
o.
100 o
0 90 180 270 360
Fig. (6.20 )...continued on the next page
194
600 120
Z ci)
o
2 a;
o co
•g- 350 c
as
ra °
co
o as
.c
o.
100 0
0 90 180 270 360
600 120
Z- ch
a)
-a
0 a;
60 Cu
a)
u,
cc;
..c
a
100 0
o 90 180 270 360
600 120
it=
z N=15 Co
a)
Gi. Ps -a
e ai
0
co
735 350 -60 c
C co
a)
o)
71-3
o co
.c
0.
100 0

o 90 180 270 360
600 120
Z ci)
a)
ms
o ai
350 o)
60 c
ti
c co
a)
co
Cu
ril c
CI.
100 o

0 90 180 270 360
off-set angle, deg.
total net force, N
phase angle, deg.
Fig.(6.20) The total net force and phase angle due to waviness (tn-8)
For 9 waves in Fig.(6.20) the system behaves almost like the system with 7 waves but this
time shift occurs from 0 to 90 degrees and then it decreases almost linearly to 0 degrees
again. Continuing increasing the number of waves first to 10 and then to 11 causes
195

almost the same sort of effect in a opposite sense to the increasing of the number of
waves from 4 to 6. When the number of waves are 11, 12 or 13 the change in the total net
force and phase angle are small as seen in Fig.(6.20). However, when the case of 11
waves is, for example, investigated in detail, commencement of a radical change is
observed as seen in Fig.(6.22). The number of oscillations of the total net force and
phase angle are shifted to 16 which is twice the number of balls in this case.
400 45.008
42 300 45.003 Tt
5
200 I I 1 44.998
0 90 180 270 360
off-set angle, deg.
Fig.(6.21) The change in the force for different off-set angles (m.8, N-8)
302 48
th
0
296
C
290 • 44
0 90 180 270 360
off-set angle, deg.
Fig. (6.22) The change in the force for different off-set angles (m-8, N=11)
600 45.006
/D
-0
1 350 45.001 13)
a
100 44.996

0 90 180 270 360
off-set angle, deg.
Fig.(6.23) The change in the force for different off-set angles (m-8, N-16)
196
The behaviour of the system for 14, 15 and 16 waves is similar to the system with 6, 7,
and 8 waves. The only difference is that the oscillation of the total net force and phase
angle is twice the number of balls. For 16 waves per circumference the phase angle
change has reduced considerably, as it had for 8 waves. When it is observed closely, a
similar pattern to that of 8 waves is seen as shown in Fig.(6.23).
-3
1.17 1.18 1.20
time, s
0 500 1000 1500

shaft rotating frequency (Hz)
Fig.(6.24) lime domain vibrations due to outer race misalignment and its FFT
(m-8, Pr-10 N, N.8, n-5000 rpm, c-300 Ns/m, cobp=280 Hz)
These results are in agreement with the predictions shown in Table (2.1) in Chapter 2 for
the outer race waviness. The pattern of vibrations for a bearing with a wavy outer race is
given as (see Table (2.1)):
Waviness of orders Vibrations caused by outer ring waviness

k=q m ±p (1 m coc (6.6)
197
10
N=7
8.
0
0
. . I 500 1000
• • I
1500
• •
2000
10
N=8
8-
0 A• . A .k p
0 500 1
1000 1500 2000
10
N=12
0 • • • 1 1
0 500 1000 1500 2000
10
N=16
e.
Tx .
E 2'
al
o 500 1000 1500 2000
Fig.(6.25) ...continued on the next page
198
10
N.17
11-1 - .
0 50015001000 2000
shaft speed (Hz)
Fig.(6.25) The FFT of vibrations due to outer race waviness of different orders
(m-8, Pr-10 N, n-5000 rpm, a-300 Ns/m, co bp-280 Hz)
For example if q=1 and p=2 in Equ.(6.6) the result matches with the Fig.(6.20) quite
favourably. However, for q-1 and p-4 the results barely matches since for k-12 the
variations are transient and changing from 8 cycles to 16 cycle (see Fig.(6.21)). The same
result can also be obtained from q-2 and p-4 where k is again 12 but vibrations are
predicted to occur at twice the BPF. However, this is in an acceptable level since the
amplitudes of the variations are small.
In order to see whether the predicted vibrations would occur in the simulation model,
outer race waviness was introduced to the system. The outer races of both, right and left
hand side, bearings were assumed to be wavy by the same amount, so that the shaft
would move in a cylindrical mode.
The time and frequency domain vibrations of a shaft which is supported by bearings with
8 balls and rotating at a speed of 5000 rpm, are given in Fig.(6.24). In this case the outer
race wave number is 8 with an amplitude of 3 tun As predicted from Fig.(6.20) the most
dominant peak is at the BPF which is 280 Hz for the given shaft speed. Peaks at the
super harmonics of the BPF are also present at 560 Hz, 840 Hz and 1120 Hz.
In order to observe whether the other predicted frequencies would also be observed, a
set of results were obtained for a selected number of waves. The amplitude of the waves
was 2 gm and the number of balls was 8.
When the number of waves is 7, a peak at the BPF appears as seen in Fig.(6.25). The
amplitude of the peak is about 3 gm. For 8 waves the main dominant peak is again at the
BPF (see also Fig.(6.24)) with its super harmonics.
Experimental researchers have pointed out that severe vibrations occur for the outer race
waviness when the number of waves equals to the number of balls [Wardle, 19884
199
Fig.(6.20) suggests that the variation of the total net force and phase angle are maximum
when the order of waviness k=i m ± 1 where i is a positive integer. The amplitude of the
peak at the dominant frequency for N=8 is less than the one for N=7 for the simulation
model as shown in Fig.(6.25). Wardle [1988a] concluded from his linear model that the
radial vibrations occur only when k .i m ± 1 whereas the axial vibrations take place when
k = m. This is partially correct since the radial vibrations have larger amplitudes when k =
i m ± 1.
o 500 1000 1500 2000 2500 3000
0.06
N-12
0.05-
? 0.04-
..s
x
0.03-
-g
aE
I
co .
0.01-
0 .00 - • • 1
0 500 1000 1500 2000 2500 3000
shaft speed (Hz)
Fig.(6.26) The changing of the vibration phase from the BPF to twice the BPF
(m-8, Pr-10 N, N-8, n-5000 rpm, c.300 Ns/m, co,,,-28O Hz)
For N=12 in Fig.(6.25) the vibrations are very small as also predicted by the total net force
change in Fig.(6.20). The pattern of vibration spectrums observed for N=16 and N=17 in
Fig.(6.25) are similar to those that are predicted by the total net force and phase angle
variations in Fig.(6.20).
200
A change of dominant vibration frequency from the BPF to twice the BPF (2 cobp) is of
particular interest. The results presented in Fig.(6.26) shows this in the frequency
domain. As seen in Fig.(6.26) when fs1-11 the dominant vibrations are at the BPF.
However, for N-12 the dominant vibrations are observed at twice the BPF. This
transformation was also predicted from the the total net force and phase angle variations
in Fig.(6.20).
6.51 Inner ring Waviness
Inner ring waviness causes more complicated vibrations than outer ring waviness. In
order to study the inner ring waviness, the bearings of the simulation model were
assumed to have waviness in their inner rings of the same order and magnitude such that
both bearings are identical. The amplitude of the waviness was 3 gm. As the inner rings
of the bearings are fitted to the shaft, they rotate at the shaft speed.
Fig.(6.27) shows the vibration spectrums obtained for different orders of waviness. For
relatively low wave orders (in this case it is also called out of roundness), the vibration
spectrum due to waviness was first experimentally studied in detail by Gustafsson and
Tallian et al. [1963]. They reported that vibration spectrum was dominated by the number
of waves times the inner ring rotational speed (see Fig.(2.24)). This is also predicted from
Table 2.1 in Chapter 2.
The predicted vibrations are clearly observed in the case of two waves in Fig.(6.27). A
dominant peak at 2o) (166 Hz) is observed besides peaks at the super harmonics of the
shaft speed (249 Hz, 332 Hz, 415 Hz, 498 Hz), at inner race to cage relative speed
((co-co) -48 Hz) and the wave passage frequency (384 Hz). Peaks at more complicated
speeds (c+2co -550 Hz and 2 co p+2co -934 Hz) also appear. However, a more
complicated frequency spectrum is observed for 3 waves. A peak at 3w (249 Hz) is
observed at a relatively smaller amplitude implying a shape change in the vibration
spectrum. While two relatively larger peaks are observed at o,,-2o. (218 Hz) and
CO p+co (467 Hz).
For 4 waves there is only one peak at the WPF (384 Hz). This is one of the turning points
for the system behaviour. From this waviness order onwards the vibration frequency
becomes more obedient to the formula given in Table 2.1:
Waviness of orders Vibrations caused by inner ring waviness

k=q m ± p q m (co-co) ±p co (6.7)
201
The vibrations due to the waviness of order 5 are relatively small in amplitude (0.5 p.m) and
do not match with Equ.(6.7). The peaks are at c p +2co (550 Hz) and 2aup-to (685 Hz).
For 6 waves the frequency spectrum has a closer match with the equation. For q.1 and
p=2 a peak at cow1,-20 (218 Hz). However, there are also peaks at cr4coc (244 Hz),
cow,p+co (467 Hz), 2 U,-co (685 Hz) and 2 Cowp-coc (715 Hz) which are not predicted by
Equ.(6.7).
o.s
0.6
zi.
E
as
0.2
0.0

500 1000
shaft frequency (Hz)
N-6
••••••••••••••••
N.7
N-8
MISINSODINIIIIS
N-9
= • MI • MI • = •
Is1-10
1
P.
2- ;!
•i•s
o
o
• -I" k ,
Agio.
500
! i A f.A A
1000
Fig.(6.27) ...continued on the next page
202
0.08
N-11
— N.12
N.13
0.00 ra.445.--. . .,... • . •

0 500 1000 1500 2000
shaft frequency, Hz

0 500 1000 1500 2000
shaft frequency, Hz
Fig.(6.27) The FFT of vibrations due to the inner race waviness of different orders
(m-8, Pr-10 N, n-5000 rpm, c=300 Ns/m, 4p -280 Hz, cowp-384 Hz)
Wardle [1988a] predicted from his linearised equations of motion that the vibrations
would take place only at waviness of order N = i m ± 1. However, the vibrations for
nonlinear systems are relatively more complicated. Vibrations at his prediction speeds are
the most severe ones and the formula given above matches with the simulation model
perfectly for these values of waviness order. For example for 7 waves there is a peak at
ceLp-co (301 Hz) where q=1 and p=1 and for 9 waves the peak is at (4,,p+co (467 Hz)
where q=1 and p=1. From above relationship for 8 balls vibrations are predicted at the
203
I
WPF (384 Hz). Its superharmonics at 2cowp (768 Hz) and 3o.)„,p (1152 Hz) are also
apparent in Figs.(6.27 & 6.28). For 10 waves the relation still holds and for q=1 and p=2 a
peak at co„,p+2co (550 Hz) is observed.
1.2
I
•
•
0.0
0 500 1000 1500 2000
Fig.(6.28) The FFT of vibrations due to the wave passage frequency

(m.8, N=8, Pr-10 N, n-5000 rpm, c-300 Ns/m, c4,,p .384 Hz)
From 11 waves to 13 waves a transformation is observed from q=1 to q=2. Vibration

amplitudes become negligible (<0.08 Luny For 11 waves, peaks at 11 (co -cocj+co (611
Hz), 22 (co -cod-co (973 Hz) and 33 (w-wc) (1585 Hz) are observed. For the waviness
,+4a) or at 2co„,p- 4 co depending on the
of order 12, the predicted peak is either at co 1
parameters chosen for p and q in Equ.(6.7). The peaks for 12 waves are at 2c0,,,,-4w
(436 Hz), 12 ( co -0) (576 Hz), cowp +4a) (716 Hz) and 24( co -00 (1152 Hz). A clear
' transformation from q=1 to q=2 can be observed in the vibration spectrum obtained for
the waviness of order 12.
,--3co (519 Hz) as predicted from the

For 13 waves the dominant peak is at 2a.j,,1
Equ.(6.7). However, there are still peaks at 3 cowp +2co (1318 Hz) and 5cowp- co (1837 Hz).
This indicates that the transformation is not over yet. For the waviness of orders 14, 15,
16 and 17 the peaks in Fig.(6.27) are at the frequencies predicted by Equ.(6.7) as
2C&wp-2co (602 Hz), 2 wwp-co (685 Hz), 20.),,,p (768 Hz) and 20 1
,+CO (850 Hz)
respectively.
There can be found four stages in the results presented in Fig.(6.27). From N=3 to N=5
is a transformation from relatively less vibration amplitude to larger vibration amplitudes.
From N=6 to N=10, there are predictable vibrations for q=1 in Equ.(6.7). From N=11 to
N=13, there is a transformation from q=1 to q=2 in Equ.(6.8) and from N=14 to N=17 is
204
again predictable vibrations for q-2. The same trend for larger orders of waviness can be
expected.
6.5.3 Ball Waviness

In order to study the effect of ball waviness, one ball from each bearing was assumed to
have a wavy surface. The case was simplified as the ball was assumed to rotate around an
axis that passes through its centre and parallel to the bearings axis. The amplitude of the
waviness was 3 gm. The wavy balls were in phase in each bearing.

1.13 1.14 1.15 1.16 1.17 1.18
time (s)
0.5
A,
000 2500
0.3
'.
,....
§. 0.2-
e
=
= .
10.1
M. -
0 +1n9-1--1—v—INIIII 1.11 IN0-11

500 1000 1500
Fig. (6.29) Vibrations due to ball waviness (N-4, m=8, n-5000 rpm, Pr. 10 N,
c-300 Ns/m, c)c-35 Hz, cob -278 Hz, 44-1112 Hz)
205
lime (s)
0.3
N=4
N=6
N=8
0.0 • 1 • -. ° 1 -
500 1500 2000 2500
Fig.(6.30) The time spectrum of vibrations due to ball waviness of different orders and
their FFT (m=8, n=5000 rpm, Pr=10 N, c=300 Ns/m, 4-35 Hz, (4,-278 Hz)
In the case of ball waviness there are two important frequencies. The ball set rotates at
the cage speed round the inner ring and the ball with the wavy surface acts like an
oversized ball, hence since the ball set comes to the same position after one cage
rotation, the system undergoes vibrations at the cage speed (for more information see
off-sized-balls in Section 6.6). Other important frequencies are the ball rotation
frequency. When the ball rotates, the position of the balls will repeat itself after each
21c/N, where N is the number of waves per circumference of the ball. Therefore, the
206
vibrations due to ban waviness will take place at the speed of Nob.
Fig.(6.29) shows the vibrations for the bearing with a ball having waviness of order 4. The
shaft rotates at a speed of 5000 rpm. For this speed, the cage speed is 35 Hz and the ball
rotating speed is 278 Hz. When the time domain vibrations are investigated, there are
two main frequencies: one at 35 Hz (the cage speed) and the other is at about 1110 Hz
(44). The frequency spectrum shows similar results. When the peak at about 1110 Hz is
Investigated in detail, a set of peaks at N cob ticoc is observed as seen in Fig.(6.29).
Relatively higher peaks appear at N cob ±w.
o
L
500
.
1000

. .
1500
Fig.(6.31) The FFT of the vibrations due to ball waviness

(m=8, n-5000 rpm, N-2, Pr-10 N, c=300 Ns/m, co,-35 Hz, 4-278 Hz, 2cob-556 Hz)
Increasing the number of waves means making the ball smoother with a larger diameter.
This can be seen from Fig.(6.30) as well. As the number of waves are increased the
vibration frequency is dominated by the cage speed as clearly seen in the time and
frequency domain vibrations shown in Fig.(6.30). For a more detailed discussion on
vibrations of this sort see Section(6.6).
When the 11/04, ti olc coincides with the natural frequency of the system severe
vibrations take place. This is shown in Fig.(6.31) where a bearing with a wavy ball of order
2 causes vibrations at a speed close to the natural frequency of the system and hence
resulting in severe vibrations.
207
6.6 Off-sized Bait
First, for further simplification of the problem the contact stiffness coefficient between the
balls and the raceways were considered to be linear and constant. For a simple radial ball
bearing as seen in Fig.(6.12) the angles and reference axis were set as was in Chapter 3
(see Fig.(3.15)).
Balls were radially prebaded in order to ensure the continues contact of all balls and the
raceways, since otherwise a chaotic behaviour might be observed. The preloaded
deflection for each ball, 80 , was assumed to be 5 urn. The centre of the inner race is
shifted 2 jun in the x direction and 2 prn in the y direction with respect to the outer race
centre. For different off-set angles the total net force and the angle between the net
force and the x axis (phase angle) as seen in Fig.(6.12) were recorded. •
60000
u_
A
T 50000
•

0 90 180 270 360
Fig. (6.32) The change in the force for different off-set angles for a linear system
(m.5, 3rd ball is 1 jun oversized)
Fig.(6.32) shows the change in the force and the phase angle for such a bearing with 5
balls. In this case the third ball in the set (see Fig.(6.34) for numbering) is 1 tun oversized.
The results show that the changes are periodic with a frequency of cage speed, implying
vibrations at the cage speed. The results also imply that the most severe vibration for this
system will occur when the cage speed coincides with the natural frequency. The shapes
of the change are perfectly sinusoidal with a small shift. This is understandable since the
other forces will be cancelled out by each other and the only force acting on the bearing
will be that of the off-sized ball's. This force will increase as the ball enters to the loaded
zone and decreases as it enters to the relief zone. This is also observed in Fig.(6.32).
Later, similar results were obtained for a bearing with nonlinear ball to race contact
stiffness proportionality as seen in Fig.(6.33). The shape of the curves show slight
208
change due to the nonlinearity. The change in the force remained cyclic with the speed
of cage. Although Tamura [1968] and Meyer et al. [1980] suggested that the off-sized
ball would cause vibrations at the cage speed and its multiples, the study on the ball
passage frequency indicated that the vibrations at the multiples of cage speed would take
place only for special cases. Therefore a series of different cases was tested resulting in
following outputs.
CO
a.
I • I • 1
0 90 180 270 360
Fig.(6.33) The change in the force for different off-set angles for nonlinear system
(m=5, 3rd ball is 1 um oversized)
Fig.(6.34) Numbering of the balls
Figs.(6.35-6.39) show the change in the total net force and phase angle for different
locations of off-sized balls. In Fig.(6.35) a bearing with 5 balls is tested. The second and
the fifth balls (see Fig.(6.34) for ball numbers) are 1 pm oversized from the rest. As
Fig.(6.35) shows the change in the combination of off-sized balls or numbers of off-sized
balls will only cause a change in the shape of the curve. The total net force will change
cyclically at the cage speed. This is also true for a bearing with 8 balls as seen in
Fig.(6.36), in the case of the second, fourth and sixth balls are 1 p.m oversized.
209
▪
210 55
200 50
190 45
•
180
• .•
170 • • 1 • o . 35
0 90 180 270 360
Fig.(6.35) The change in the force for different off-set angles

(m=5, 2nd and 5th balls arel pm oversized)
340 52
2 305 - 45
•
•
•
•
270 • • 38
0 90 180 270 360

(m=8, 2nd, 4th and 6th balls arel pm oversized)
For a bearing with 6 balls, it is shown that vibrations will take place at twice the cage speed
in the case of the second and fifth balls are 1 gm oversized as seen in Fig.(6.37) and
three times the cage speed in the case of the second, fourth and sixth balls are oversized
as seen in Fig. (6.38). These results may explain the experimental studies of Tamura
[1968] and Yamamoto et al. [1981]. The former observed vibrations at twice the cage
speed in the axial direction when there are two neighbouring balls oversized but wrongly
concluded that off-sized balls would cause vibrations at the multiples of the cage speed
and the latter observed vibrations at the cage speed and twice the cage speed.
The results indicate that off-sized balls will cause vibrations at the multiples of cage speed
only when the off-sized balls are symmetrically distributed. In this case, balls will repeat
their pattern as explained in Section (6.4). This was further confirmed with Fig.(6.39)
210
where a bearing with 8 balls produced cyclic forces at four times the cage speed for the
second, fourth, sixth and eighth balls are oversized.
250
z
--- 240
___
220
0 180 270
Fig. (6.37) The change in the force for different off-set angles
(m=6, 2nd and 5th balls arel pm oversized)
238 45.2
235 I •1 44.8

0 90 180 270 360

(6, 2nd, 4th and 6th balls are1 gm oversized)
Table (6.1) shows the possible multiples of cage vibrations for a bearing with different
number of balls. The difference in the size of balls within the off-sized ball set produced
the same effect of having only one off-sized ball. However, the vibration amplitude was
affected by this change.
Having these initial results in hand, further research was carried out in order to obtain the
vibrations due to off-sized balls from the simulation model. The off-sized balls were
located symmetrically in both bearings such that they will move in the same direction
simultaneously (i.e., the balls are assumed to be in phase).
211
315.5 45.03
315.1 1 1 1 44.97
0 90 180 270 360
Off-set Angle (deg.)

(m-49, 2nd, 4th, 6th and 8th balls arel pm oversized)
Possible Speeds at Peaks

No of
Balls coc 2 ok 3 coc 4 coc 5 coc 6 coc 7 wc 8 coc
3 11
4 V
5 1/
6 VV V
7 11
8 VV V
9 V V
10 V 31 11
11 11
12 V 11 V Al 11
13 V
14 11 11 11
15 V 11 V
16 VV V V
17 V
Table (6.1) Frequencies of the possible vibrations for different numbers of balls in a
bearing
212
4.00e-5
E
e 3.00e-5 -
-3
1. 2.00e-5 -
16
$-
16
1.00e-5 -
a
.•nn•011.14°
0.00e+0 i
0 50000 100000
Fig.(6.40) Peak to peak amplitudes of the system vibrations for a bearing
with 1st and 3rd balls are 1 pm oversized (m=8, Pr-100 N, c-300 Ns/m, co n-840 Hz)
First, two balls were assumed to be 1 gm oversized for a bearing with 8 balls. These were
the second and the fourth balls in the set for both right and left bearings. An external
damping of 300 N s/m was introduced to the system in order to eliminate the effect of the
natural frequency and an axial preload of 100 N was applied to ensure the continuous
contact of the raceways and balls. Peak to peak amplitudes were recorded when steady-
state vibrations were reached. Fig.(6.40) shows the peak to peak amplitudes against the
different shaft speeds.
1.00e-5 -
8.000-6 -
1
E
a+
-0
z
= 6.00e-6 -
M.
E '
co

•es 4.00e-6 "
$-
1 2.00e-6 -
&
0.00e+0 . , --t- • . - .n11.
0 50000 100000
Fig.(6.41) Peak to peak amplitudes of the system vibrations for a bearing
with 1st and 5th balls are 1 pm oversized (m.8, Pr-100 N, c-300 Ns/m, con-840 Hz)
From Fig.(6.7) for a shaft supported by angular contact ball bearings with 8 balls and
preloaded with 100 N, the natural frequency can be found to be 840 Hz (s 50400 rpm).
213
As predicted earlier for a bearing with 2 oversized balls (1st and 3rd) the system is forced
at the cage speed. When the cage speed coincides with the natural frequency, the
system resonates as shown in Fig.(6.40). For the shaft speed of about 120000 rpm the
cage frequency is about 50400 rpm and this is clearly observed in Fig.(6.40) as predicted
from Fig.(6.41). The curve indicates jumping behaviour for hardening type stiffness. This
is because the oversized ball goes through a large deflection, hence this change over a
large deflection range causes a clear exhibition of hardening type behaviour and a jump in
the amplitude.
In order to verify the predicted behaviour made by the total net force observations, the
off-sized balls are placed as such, the first and fifth balls are 1 tun oversized for a bearing
with 6 balls. From Fig.(6.37) the system is expected to resonate when the 2 w coincides
with the natural frequency.
Fig.(6.41) shows that at a shaft speed of 60000 rpm, twice the cage speed which is
50400 rpm (840 Hz) coincides with the natural frequency. However, the resonant
amplitude is relatively small at about 2 gm. At the speed twice the off-sized ball passage
frequency ( 4 co),), the cage speed also coincides with the natural frequency and results in
a vibration amplitude of about 10 gm.
20000 40000
Fig. (6.42) Peak to peak amplitudes of the system vibrations for a bearing with 2nd, 5th
and 8th balls are 1 pm oversized (m-9, Pr-10 N, c-300 Ns/m, cen-580 Hz)
In order to show the off sized ball effect further, and to check the predictions of the total
net force investigation, another set of results for a bearing with 9 balls were obtained.
The first, fourth and seventh balls were made 1 tun oversized and the preload value was
set to 10 N. It is predicted by Fig.(6.38) that the resonance would occur at the speed of 3
coc which is 27500 rpm (co. =580 Hz) for the given system.
214
As seen in Fig.(6.42) the system goes into resonance at this speed. The first super
harmonic of the off sized ball passage frequency is at 55000 rpm and its first and second
subharmonics at frequencies 13750 rpm and 9165 rpm respectively are also observed in
Fig.(6.42). Since the super harmonic is close to the natural frequency, the dominant peak
is at this speed.
lAd.0.5 gm'
0.0
0 20000 40000 60000 80000
0.15
I Ad=1 gm I
-0
:a 0.10 -
as
.re
0.05 -
1.
11
0.00
0 20000 40000 60000 80000
0.08
S. Ad=2 gm I
4
:a
0.06 -
0.
n,
E 0.04
0.02
0.00 . •
0 20000 40000 60000 80000
shaft rotational speed, r.p.m.
Fig.(6.o43) Peak to peak amplitudes of the system vibrations fora bearing with 1st, 3rd,
5th and 7th balls are 1 gm oversized (m-8 Pr-10 N, c.300 Ns/m)
215
In order to study the effect of the amplitude of the oversized ball diameter, a case was
studied. The preload was reduced to 10 N and the first, third, fifth and seventh balls were
made 0.5 gm, 1 gm and 2 gm oversized in turn. Fig.(6.39) suggests that the resonance
will occur at the speed of 4 w.
In Fig.(6.43) for the bearing with 0.5 grn oversized balls, the resonance occurs at 21000
rpm which means a natural frequency of 600 Hz. This is relatively higher with respect to
510 Hz of a perfect system in the same conditions. The difference is due to the off-sized
balls. The diameter difference of the balls due to waviness acts as preload to the system.
This point is further proven when the results from 1 gm and 2 gm oversized ball are
investigated. For the former the natural frequency is 630 Hz and for the latter it is 700 Hz
as seen in Fig.(6.43).
Fig.(6.43) shows that increasing the diameter difference for oversized balls decreases
the resonance amplitude resulting in the vibration spectrum becoming more complicated.
For example for 2 gm diameter difference the sub and superharmonics of the off-sized
ball passage frequency also appears in the spectrum. Another important point is that in
this example the balls are symmetrically distributed. If only one ball is oversized, when the
diameter difference increases, the peak to peak amplitude of the vibrations will also
increase.
216
6.7 Misalignment
6.7 1. Both Bearings are Misaligned
6.7.1.1 Outer ring misalignment
In order to see the effect of misalignment a simple case of outer ring misalignment is
studied. In Equ.(4.35) and Equ.(4.36) in Chapter 4, a and O c were both set to 0. The
misalignment angle p was set to be 0.02 rad. and both bearings were misaligned in the
opposite sense in order to eliminate the z oscillations and shaft rotated at a speed of
3000 rpm.
Fig.(6.44) show the FFT of the x oscillations. Vibrations occur at two frequencies: one is
the cage speed (21 Hz) and the other is the system natural frequency (820 Hz). The
outer ring misalignment produces a case that the same position of balls can only be
obtained after one cage revolution and the system is therefore excited at this frequency.
This results in vibrations at the cage speed. Since the superharmonics of the cage speed
excites the natural frequency, vibrations at the natural frequency are also observed. In
the case of misalignment, the system natural frequency is shifted to a relatively higher
frequency since the misalignment acts to preload the system. As studied earlier in this
chapter varying the positions of balls alters the natural frequency. This becomes more
significant when the misalignment is applied. While the balls in the upper half of the
bearing are preloaded more, the balls in the lower half are relieved. This results in a
relatively more nonlinear system. This is not very significant in the z direction since the z
movement is cancelled out.
8E-04
6E-04
—
A
0
I
-0
0 4E-04 -
=
Z.
g
2E-04 -
0E+00

0 500 1000 1500
frequency (Hz)
Fig.(6.44) The FFT of the vibrations due to outer ring misalignment

(m-8, p=0.02, Pr-10 N, n-3000 rpm, c-300 Ns/m, coc-21 Hz, a 1-820 Hz)
217
6.7.1.2 Inner ring misalignment

Fig.(6.45) shows the inner ring misalignment effect on the system vibrations. Three
significant shaft speed were tested. In each case the resulting vibrations were at the shaft
speed. Since the effect of position of the most loaded ball is more dominant than the
position of the ball set itself, the vibrations at the shaft speed overcome the vibrations at
the inner ring ball passage frequency. A small peak was observed at the inner ring ball
passage frequency when the FFT of the oscillations was obtained but the amplitude of
the peak was very small with comparison to the peak at the shaft rotating speed. As
pointed out by the experimental studies of Wardle and Poon [1983] this may excite the
low order shaft resonances.
n-3000 rpm
n-5000 rpm
------ n-7000 rpm
-4 . 1
1.1550 1.1650 1.1750 1.1850
time (s)
Fig.(6.45) Vibrations due to the inner ring misalignment

(m-8, p-0.02, Pr-10 N, c=300 Ns/m)
6.7.2 LHS Bearing is Misaligned
When only the inner ring of the LHS bearing is misaligned, the vibrations produced is
violent. For a shaft speed of 5000 rpm the superhannonics of the shaft speed coincide
with the natural frequency of the system in the x direction and excite it. Due to the
misalignment, the natural frequencies in the x direction and y direction are no longer the
same or are very close. For the x oscillations, small peaks at Con to) appear in the
spectrum. The vibrations in the y direction are at the shaft speed and its superharmonics.
However, the vibrations in the z direction are at twice the shaft speed and its harmonics as
seen in Fig.(6.46).
218

-10
250 500 750 1000 1.17 1.18 1.19 1.2
15
10
-10
-15
0 250 500 750 1000 1.16 1.18 1.2
20
15
10
-5
-10

250 500 750 1000 1.16 1.18 1.2
frequency (Hz) time (s)
Fig.(6.46) Vibrations due to the inner ring misalignment on the LHS bearings only
(n8,10.02, Pr-10 N n-5000 rpm, c.300 Ns/m, co.83 Hz 4.35 Hz, coe750 Hz)
6.8 Defects
A crack or debris was assumed to be located on the running surface. It was assumed to
have 3 gm depth and 1 degree of width.
219
6.8.1 Outer race Defect

When the defect is on the limning surface of the outer ring, vibrations take place at the
BPF and the system natural frequency. When it or Its harmonics coincide with the natural
frequency of the system severe vibrations are observed. At lower shaft speeds, it is
almost certain that one of the superharrnonics of the BPF will coincide with the natural
frequency.
4 1
n - 2000 r.p.m. n - 3000 r.p.m.
—. 3 -
§. .
a)
V 2 - 0.5 -
:a
'
•
o o
o 500 1000 1500 0 500 1000 1500
4 0.3
n -5000 r.p.m.
-
0.2
0.1
-
o- . 1
o
Ai
500
e. 1 . . . .
1000
0
1500 0 500 1000 1500
shaft rotating frequency (Hz) shaft rotating frequency (Hz)
Fig.(6.47) The frequency spectrum of vibrations due to defected outer ring

for different shaft speeds (Pr 10 N, m=8, c-300 Ns/m)
For a shaft speed of 2000 rpm the system vibrates at the natural frequency of 560 Hz as
seen in Fig.(6.47). For this speed the BPF is 112 Hz. 504p coincides with the natural
frequency and the system resonates. For a shaft speed of 3000 rpm the BPF is 168 Hz.
Therefore peaks at i CObp where i -1, 2,.., appear in Fig.(6.47) for this speed. The
dominant peak is at 504 Hz (34 p) since it is the closest harmonic to the natural
frequency. A small peak at the natural frequency appears as well. As for a speed of 5000
rpm the BPF is 280 Hz and 20)bp coincides with the natural frequency, the vibrations take
220
place at the natural frequency for this speed as seen in Fig.(6.47).
In order to confirm the point made above, another shaft speed (7000 rpm) wais chosen
since for it the BPF is 390 Hz and 2 cobp is 780 Hz, both being further away from the
natural frequency. As seen in Fig.(6.47) the dominant peak is at the BPF for this speed.
Peaks at its superharmonics also appear. A peak at the natural frequency Is also present
In the spectrum. However, when the amplitudes are compared, the vibrations for a shaft
speed of 7000 rpm are not significant as the maximum peak amplitude is only about 0.3
gm, compared to 4 pm for n-2000 rpm or n-5000 rpm.
The results obtained from the simulation model are similar to experimental results
reported by Yhland and Johansson [1970], Braun and Danter [1979], Igarashi et al.
[1982; 1983; 1985], McFadden and Smith [1985] and Karakurt [1989]. They all
observed that when a defect exists on the outer ring running surface, the resulting
vibrations are at the BPF and the system resonates when it or its harmonics coincide with
the natural frequency.
6.8.2 Inner ring defect

More complicated vibrations take place when the defect is on the inner ring running
surface since the defect itself rotates at the shaft speed. Some researchers predicted
that it would cause vibrations at the inner ring ball passage frequency [Braun and Danter,
1979; Karakurt, 1989]. Some others [Mc Fadden and Smith, 1985] pointed out that in
the case of inner ring defect, the spectrum would accommodate peaks at the shaft
speed, cage speed, their harmonics and their combinations. The frequency spectrums
shown in Fig.(6.48) exhibits peaks at shaft speed, cage speed, ball passage speeds of
the outer and inner race and their harmonics as also pointed out by Taylor [1980].
For a shaft speed of 1000 rpm peaks are at the shaft speed (17 Hz), the BPF (56 Hz), the
inner ring ball passage frequency (m (co -coc)-77 Hz) and their harmonics. As some of
these coincide with the natural frequency, a resonant peak at the natural frequency (540
Hz) appears as shown in Fig.(6.48).
For a shaft speed of 3000 rpm, peaks at the shaft speed (50 Hz), the BPF (168 Hz), the
inner ring ball passage frequency (232 Hz) and their sub and superharmonics appear in
the frequency spectrum. The most dominant peak is at 504 Hz (3 cobp ) as it is close to the
natural frequency (540 Hz) and excites it. Peaks at the inner ring ball passage frequency
and its harmonics (116 Hz, 232 Hz, 348 Hz, 696 Hz) are less dominant as seen in
Fig.(6.48). When the vibration peaks in Fig.(6.48) for the shaft speed of 3000 rpm and
5000 rpm are observed overall, the peaks form almost a semi-parabola having the centre
at the natural frequency.
221

0.6
0.5
0.4
0.3
0.2
0.1
0 0
0 250 500 750 1000 0 250 500 750 1000
1.5 0.6
0.5
0.4
0.3
0.2
0.1
0 0
0 250 500 750 1000 0 250 500 750 1000
0.3 1.2
0.9
0.6
0.3
0
250 500 750 1000 250 500 750 1000
shaft frequency (Hz) shaft frequency (Hz)
Fig.(6.48) The FFT of vibrations due to defective inner ring

for different shaft speeds (Pr-10 N, m=8, c-300 Ns/m)
For a shaft speed of 4000 rpm the most dominant peak in the frequency spectrum in
Fig.(6.48) is at the natural frequency as it coincides with 19 co, (530 Hz). A relatively small
peak at the shaft speed (67 Hz) and a peak at the BPF (224 Hz) also appear in the
spectrum.
For a shaft speed of 5000 rpm the most dominant peak is again at the natural frequency
since it is close to the 2 Cobp (560 Hz). Other peaks at the shaft speed (83 Hz), the BPF
(280 Hz), the inner ring ball passage frequency (387 Hz). The second dominant peak in
222
the spectrum is at Con—co (457 Hz) and the third dominant peak is at con+4coc (680 Hz).
A similar frequency spectrum was obtained for a shaft speed of 7000 rpm as shown in
Fig.(6.48). However, in this case the natural frequency coincides with the inner ring ball
passage frequency (8 (0)-0))=544 Hz). The second and third dominant peaks are at
ton—co (423 Hz) and con+co (657 Hz). Another dominant peak is at the shaft speed (117
Hz).
The vibration spectrum obtained for a shaft speed of 9000 rpm in Fig.(6.48) is similar to
the one for 1000 rpm or 4000 rpm In Fig.(6.48). The BPF (504 Hz) resonates with the
natural frequency so the vibrations are at the natural frequency.
A close similarity was observed between the frequency spectrums in Fig.(6.48) and the
experimental spectrum due to the inner ring defect presented in Fig.(2.38) in Chapter 2.
6.8.3 Defect on the Ball Surface

One of the balls on each ball bearing was assumed to have a scratch of 3 p.m on its
running surface. The scratch was assumed to be rotating around an axis passing through
its centre and parallel to bearing axis, hence the scratch was passing the same points on
the inner and outer raceways at regular intervals.
The vibrations caused by the defect will have a frequency of twice the ball rotating speed
since it inserts the same amount of force once each half rotation.
Fig.(6.49) shows the frequency spectrum obtained for different shaft speeds. For a shaft
speed of 2000 rpm the ball rotating speed is 111 Hz. Since 5 4 (555 Hz) is in the
neighbourhood of the system natural frequency, it resonates with the natural frequency
and results in vibrations at 555 Hz. Peaks at the ball rotating speed and its
superharmonics (111 Hz, 222 Hz, 333 Hz, 444 Hz, ...) appear in the spectrum. The peaks
at 24 and its superharmonics (222 Hz, 444 Hz, 666 Hz, ...) have relatively higher
amplitudes, implying the actual forcing frequency and its superharmonics. For a shaft
speed of 3000 rpm in Fig.(6.49), the frequency spectrum exhibits peaks at only 24 and
its harmonics. In this case the ball rotating speed is 166.61 Hz. Peaks at 333 Hz (24),
666 Hz (4 cob), 999 Hz (6cob) and 1333 Hz (84) are present in the spectrum. However,
the dominant peak is at 666 Hz since it is the closest one to the natural frequency.
For a shaft speed of 4000 rpm in Fig.(6.49) the dominant peak is at 24 (444 Hz). Peaks
at its superbarmonics (888 Hz and 1333 Hz) also appear in the spectrum. However, since
3c4, (666 Hz) is closer enough to the natural frequency to excite it, a peak at the natural
frequency (540 Hz) and peaks at 222 Hz (4), 666 Hz (34) and 1110 Hz (54) also
appear in the spectrum in Fig.(6.49).
223

1.5 0.06
0.04
0.02
o 0
0 500 1000 1500 13 500 1000 1500
0.15 0.8
0.6
0.4
0.2
0
500 1000 1500 0 500 1000 1500

shaft frequency (Hz) shaft frequency (Hz)
Fig.(6.49) The FFT of the vibrations due to defected ball (Pr=10 N, m=8, c=300 Ns/m)
For a shaft speed of 5000 rpm the ball rotating speed is 277 Hz. 2 cob (554 Hz) is the main
excitation force and is close to the natural frequency. Therefore the resulting vibrations
are at the natural frequency of the system.
6.9 Vibrations in 5 Degrees of Freedom

As was seen in the previous section in this chapter, the vibrations associated with ball
bearings are complicated even with 3 degrees of freedom. Therefore although the
vibrations of a shaft supported by ball bearings were simulated in 5 degrees of freedom,
only the simulation in three degrees of freedom was generally employed in the
investigation.
Fig.(6.50) shows the vibrations for a geometrically perfect bearing in 5 degrees of

freedom. Since shaft also vibrates in the z direction, the shaft centre no longer moves in
synchronisization so the effect of the BPF becomes insignificant. This can be clearly
observed in Fig.(6.51) for the FFT of the vibrations along the x and y axes. Rocking and
yawing angle variations are also very small.
224
2.5
2
1.5
tt
1
0.5
0
130 135 140 145 150
0.4
i
t J

130 135 140 145 150
5E-04
4. 1E-04
\ I stP174141,
a 1,)
(3E-04) I — 1 - I
130 135 140 145 150
2E-04
I
%- 0E+00—
I t
(2E-04) - I ' t
130 135 140 145 150
dimensionles time, r.
Fig.(6.50) The vibrations of a shaft supported by ball bearings in 5 degrees of freedom

(Pr-10 N, m-8, W-31124, n-5000 rpm)
225
0.8
0.6-
I
0.2—
0
500 1000 1500
0.25
0.2-
0.05 — I

o 500 1000 1500
Frequency (Hz)
Fig.(6.51) The FFT of the vibrations of a shaft supported by ball bearings in 5 degrees of
freedom along the x and y axes (Pr=10 N, m=8, M=31124, n=5000 rpm)
Hence in this thesis no further runs were made in 5 degrees of freedom. However, when
actual experimental results of other workers are compared with the simulation model, it is
probable that simulation in 5 degrees of freedom would give closer predictions.
For the perfect system the frequency spectrums of vibrations presented in Fig.(6.9) (or
Fig.(9.10)) show a similarity to the frequency spectrum in Fig.(6.51) implying that since
the system is defect free, the vibrations of the shaft is not disturbed. However, since the
oscillations in the z direction becomes larger, they become dominant in the deflection
equations (i.e. Equ.(3.32) and Equ.(3.34)), hence the effect of the BPF reduces.
226
However, In they oscillations a peak at the BPF (280 Hz) appears in Fig.(6.51) but it is not
significant.
Closure
In this chapter the natural frequency, the BPF, the effects of varying number of balls and
preload on the BPF are studied with the simulation model. It was shown that with a careful
design many problems caused by the effects of the natural frequency and the BPF can
be avoided. The imperfections of the shaft ball bearings namely out of roundness,
waviness, misalignment, defects on the running surfaces, were also investigated. The
conclusions of these studies will be given in Chapter 8. In the next chapter, the dynamic
properties of elastomers will be investigated and elastomers as external dampers will be
introduced to the system in order to eliminate or reduce the untoward effects of the
frequencies discussed in this chapter to an acceptable level.
227
CHAPTER 7
RESULTS AND DISCUSSION:

ELASTOMERS AS EXTERNAL DAMPERS
7.1 Introduction
In the previous chapter, it was shown that even a shaft-ball bearings system equipped
with geometrically perfect ball bearings, undergoes vibrations due to variations in the
overall contact stiffness of the ball bearings. With defects in addition, vibrations can be
quite severe. Therefore in this chapter elastomers as external dampers will be introduced
to the simulation model in order to eliminate the untoward effects of vibrations associated
with ball bearings.
There is limited data available on the dynamic properties of elastomers since the stiffness
and damping coefficients of elastomers depend on their geometry. The experimental
data available in the references were obtained for certain geometries and could not be
directly used in this thesis. There is to date no acceptable, sufficiently correct theory that
gives the dynamic properties of elastomers for different geometries.
Therefore in order to obtain the dynamic properties of elastomer for cylindrical buttons,
ring cartridges and 0-rings used in this thesis, a study was undertaken as discussed in
Chapter 5. In that study dynamic properties of elastomers were obtained from their
material properties (i.e., Young's or shear modulus). In this chapter the predictions of the
formulae obtained in Chapter 5 will be compared with experimental findings and other
researchers' predictions.
228
Chapter 7 Results and Discussion: Elastomers as External Dampers
The properties obtained will later be applied to the shaft-ball bearing system as external
dampers. For this purpose, a three degrees of freedom system will be considered and
the effect of elastomer dampers on reducing the untoward effects of natural frequency,
the BPF and out of balance of the shaft centre will be investigated.
7.1 Dynamic Properties of Elastomers
The formulae developed in Chapter 5 will be employed in predicting the stiffness and loss
coefficients of the elastomer buttons, ring cartridges and 0-rings respectively.
For the buttons and cartridges the material used was polybutadiene which is classed as a
Broad Temperature Range (BTR) elastomer. This material is chosen because of the
availability of the experimental data in the NASA contractor reports [Chiang et al, 1972;
Gupta et al., 1974; Smalley and Tessarzik, 1975, Darlow and Smalley, 1976]. For 0-rings
four different materials are used namely polybutadiene, Viton 70, Viton 90 and Buna—N.
Since the all necessary experimental data for Viton 70, Viton 90 and Buna—N are not
available, the 0-rings used in this thesis are thought to be in the same size as in the
NASA contractor report [Dadow and Smalley, 1976]. This is not a major problem since the
dimensions of 0-rings tested in the report were very close to the ones required in this
thesis.
7.2.1 Cylindrical Buttons
Three different button dimensions are studied as given in Table 5.1. All had the same
diameter of 0.0127 m but each had different height. Each of the button cartridges had six
elastomer buttons arranged in two rows, with three buttons equally spaced
circumferentially in each row as seen in Fig.(5.9).
For this arrangement the compressive and shear stiffness of a ring of three buttons are
both equal to three halves of the corresponding stiffness of a single button for a single
row [Darlow and Smalley, 1977]. Since there are two rows in this case, Equ.(5.25) will be
multiplied by 3. The Young's modulus (E) in Equ.(5.25) is assumed to be about three
times the shear modulus (G) [Smalley and Tessarzik, 1976; Darlow and Smalley, 1977].
Fig.(7.1) shows the prediction values for the button cartridges of buttons with a height of
0.00318 m. Predictions of formulae given by Smalley and Tessarzik [1975] (Equ.(5.52)
and Equ.(5.53)) and the predictions of the formula developed in this thesis are very
close. The predictions for the stiffness and experimental data matches quite favourably.
229
For relatively low speeds the prediction, including shear effect, is closer to the
experimental data while for relatively higher frequencies the predictions, neglecting shear
effect, is closer to the experimental data. In this experiment both ends of the buttons
were glued in place, it was therefore expected that they would undergo shear as well as
compression loading. Experimental results suggest that for higher frequencies the shear
effects are small [Darlow and Smalley, 1976].
1E+08
K' (E) • K' (NASA, E)
....... K' (E+G) -•-•-•- K '(NASA, E+G)
o K' (Exp)
E
2
ca
0 -i- -0 4-
a.
C i
= M.=OMIIMIO=M=8MO.0. n 0
fi
MMMMMMM OMMme4
nn
1
••nn•.

100 1000
Fig.(7.1) Stiffness and loss coefficients of button cartridge no / (0.00318 m)

as a function of frequency
For the loss coefficients it is clear that the predictions from Equ.(5.25) are much closer to
the experimental data than the predictions of NASA (i.e., Equ.(5.53)) as shown in
Fig.(7.1).
230

Chapter 7 Results and Discussion: Elastomers as Eaaernal Dampers
Fig.(7.2) shows the results for a button height of 0.00238 m. Since In this case both
sides of the buttons were glued in place It was expected the predictions including shear
effect would match better with the experimental results. As seen in Fig (7.2) the
predictions for stiffness including shear effect matches with the experimental data almost
precisely.
1E+08
prep-4110k incrudrpri
1 shear froin Equ.(S.5g)
p!ediction-neglectingl
sheaf if0111 Equ-.15.-52)1 ------t
101
OOO aa aaS
i
reilialim iseamalaaMlihmrr 'or .1.••• i ••••=.
1Pr8orFheghg,
prediction incldding shear frori Eqe.(5.25)
shear from Eq0.(5.25)
I

100 1000
-sha-arfrom-Eqty5-.531. + -1--- -4- -L--L--4----t--. -

-÷ r" 1
I 1 - I -
•. . 1 : i 1
i.
et
.
o j. .•• 1
1 :
E 0.1 .______________________ 4 _113000..... 4 :_____ I • _____-_17 4 h_ 4- -4--
8 -4-
1 0 "r t 1 4-
t
-4- --4-- , --t- I-- -4 r---
shear from Equ.(5.25) I I I
i.._ 4. ...4.. 1.----1- .
• Experimental data .
i 1
0.01 ,
1 00 101D O
Fig.(7.2) Stiffness and loss coefficients of button cartridge no 2(0.00238 m)

In this case predictions from Equ.(5.52) including shear effect is slightly higher whereas
the predictions neglecting shear effect matches with the experimental data quite
favourably. Prediction for the loss coefficient are slightly higher than was experimentally
231
observed. However the predictions obtained from the formula developed in this thesis
show the same trend of reduction with an increase In the frequency.
Fig.(7.3) shows the results for a button height of 0.00476 m. In this case buttons were
only glued to the inner ring and hence it was expected that shear effect would be
negligible. However, predictions for stiffness coefficients were slightly higher than the
experimental data although the trend of the data and the prediction from the formula
derived in this thesis were the same. The predictions for the loss coefficients were
favourably close to the experimental data.
1E+08 • - .
_
-4- .
tapertnentat data I i r ' p 4 v-
-----1 predielidnegiterrigli"
prediction including i shear froni EqU.(5;25)
-t --- -- t ---- -
r .
shear from Equ.(5.25)
1 L._ L.......1.,.
s a A
0 NI a
2 E+07 uJ f —4- 1 I
.
AKNANAWANINAN)
sp.wilOTETEVAITMTUTarsmommirmirmrmyt.m.tvolm
-0- 9 vocgr---t 0D
4-
I
t
i I.
if
prediction neglecting- t preOctiorOnclOdinii --1
_ ...___
1 .shear.frnm.EqU45_52)._ i 1 shear iron) Eq_U,(5,P2X_
1
I
I 1
1
1
I
1E+06 i 1 I 1
I nn 10o 0
1
fpnzdirfflortsincludirg:anctring
s-hear -from Erip.(5.-53) 1
1 1 t
i
f----
. r : 4 --i----
, . ----- -4 L -- ---- ----4----
i.
o .
i.
g 0.1 ---- ------------------------- -------,----

_p_:=0:- F ! t 1 !
8 ! !
-f-- --f-- 1 —+—÷- •
4--
0
shear from Equ.(5.25) I ) 1
i
— .
• Experimental data I i 1
n i
!
0.01
100 frequency (Hz) 1000
Fig.(7.3) Stiffness and loss coefficients of button cartridge no 3(0.00476 m)

232
7.2.2 Ring Cartridges

Two different dimensions were tested for ring cartridges. Both cartridges had the same
inner and outer diameter but had different thicknesses as shown in Table(5.2). For the
cartridges the material used was polybutadiene.
The first cartridge had a thickness of 0.0048 m. The results for this ring cartridge is given
in Fig.(7.4). As is clearly seen, the closest predictions are obtained from two formulae
developed in this thesis.
1E+08
0000000
oo
A 00000000000000000000000000
co
Cl) -4_1
ogololossosoessomosetio.w.....10MTWooksjosoloodo.1042meTiosi.•;e.
'
C
Cl)
0 o
00000 00 0
000000 0
1
•1
00000000000
•
I
100 1000
1E+00
-1- -4— 4--
*a all predictions 0 • ,
g 1 E-01 — Q.- ---4--
000
8 -r
= -r
4-- 4- -+
.
_f_ _
-4-
I i •
I I I
1E-02

Fig.(7.4) Stiffness and loss coefficients of ring cartridge no 1 (0.0048 m)

233

Again, for the cartridge with a thickness of 0.0024 m, the closest predictions are obtained
from the formulae developed in this thesis as shown in Rg.(7.5).
For both cartridges the predictions for the loss coefficients were the same from all
prediction formulae.
1E+08 ---h-
,
--1-----1.
! I-1--1------1.----
! I I
. I 1
' r
I .i.
lUm .. Is ... !I
• .±:
: —1: .....
, " 1••••co
----------iriii1461ravs•••4-

, ............. '11= m . 1 .
...".• n • 6" ' a gal:.........star.."°6--
1 -‘°
1E+07 —i— . n 4___i_
m.li.0
a) • n-•.•".• -- -
',1
rI U
.110.1momfwIll
Oserm.sulansese.•• i I 4---
_,0 .0 so ...
1
, ! __I i
— _
_
---- Equ.(5.26) Equ.(5.33)
-
MosOMIMIMI Beam-Column
_ ----- GObel
Grassano o Experiment
1E+06
I 0 0 1 01 0
1E+00 I. — -=—'
-4- --1- --4- A- —a— ---i- -I- 4--
--r
i. --i- --t--- 1---t----t--t--r---
......
I
all predictioris I
I
4 1:6 :::-.10—t:::---- -::::r----- -- --_-_-_-
—1.-
-r-1---
--1.-- -1--- 1-- –1-_----.71:7
tri------i------
0- --t
..i.i -Tr 4 i--
4-
I i i
4 4 . t---
: i
1 1
i I
1
; , 1
1E-02

100 1000
frequency (Hz)
Fig.(7.5) Stiffness and loss coefficients of ring cartridge no 2(0.0024 m)

7.2.3 0-rings
There were no direct experimental results available for 0-rings. Therefore, an 0-ring was
assumed with the dimensions given in Chapter 5 for the 0-ring. The material is assumed
to be the same as that of the ring cartridges (i.e. polybutadiene).
234
Three different formulae were derived in Chapter 5 for 0-rings. As shown in Fig.(7.6) the
predictions for stiffness coefficients formed two groups. The first group are the
predictions from relatively simpler approaches and in the second group there are two
predictions. One of them is the empirical formula of Lindley (Equ.(A5.)) and the second
one is the direct solution of the integral given in Equ.(5.41).
As seen in Fig.(7.6) the predictions from the Lindley's empirical formula and that from the
solution of Equ.(5.41) are almost the same. The predictions for the loss coefficient from
all the formulae were the same as seen in Fig.(7.6).
1E+07 • • .
------- -4- ------ t - ' 1 ---f----i--1---f----
-------
—....4_ A- I- —± b' —•
..... ----- _._...f.._-______:

i
0...,m,.....0n -
• ...,•.....••n••
h-...!.--4-..,
-- -
. .
• 4.
...dm.
: n•••••s : ..... „. r
i .......
i .... „ .... ?..
IMAM A .... •
-- .... ----
-
..........
•i ,
i---;;I:.••11"-ff .
.. . . i '
....21tt....._ ......
L------1-
i 1
I
.i. I- --I- i..- -4
I i
! i :
I
1
I
• ..... -.'
. • t • .-
1E+06 .
i I
I
- -t-- -4------r------1.---1.----!-4— --
,
--i-- : iI. i
t7)
r-- ---t- r-- 1----1.---t---
! ! i I !
i r r 1 I. r Ii f
1 ; ! ! I
I i I !
Eq u.(5.34) riesus........1 Eq u . (5.35)
Equ.(5.41) ••,••• Lindley 1
—o— Lindley (emprical) Grassano

1E+05
1 E+00 -° 1000
1 0 4- —
---:-- -1--- ---4--

— ---t----1----r --
i- -i— . .
-4- -4--------A------4----4—i---
. : .
.20 Iall predtticilit 1
-i-
! i I
-I-- 11 . —f----!---
I I i I :
§ I- ; i
. 1 I i 1
a I I I I
1E-01
Fig.(7.6) Stiffness and loss coefficients of 0-ring as a function of frequency
235
7.3 Simulation of Elastomer External Dampers with a

Mechanical Voigt Model
In the previous section the dynamic properties of elastomer dampers were obtained from
their material properties. In this section a mechanical Voigt model will be introduced in
order to represent the dynamic behaviour of elastomer dampers in the form of cylindrical
buttons and 0-rings. The polybutadiene was used for the button cartridges and Viton
70, Viton 90 and Buna—N were used for 0-rings. For cylindrical buttons the experimental
data was obtained at 32 °C and for the all elastomer 0-rings the testing temperature was
25 °C. The experimental results were obtained from NASA contractor reports [Darlow
and Smalley, 1976; Smalley et al., 1977]
1E-04
Voigt Model

Equ.(5.25)
1
1 1
, -r-
Ko 1_44_
K0=4.086 E7 Nlm 1
K 1=6.974 E71$11m 1
•-144— CI 1
ci=58042 Ns14
1E-05
100 1000
1
— -t
I
-t---r----t----r-
;
----4
.
i i•
' 1
•
1 --1 4--i---1----1---
. ..
..... 000•......•
011ellgall110441114114(fiftilf
n Wiliat, iffi,,,,,
• 1 I I .
: .
.
..-- .L--.•.±:71:= .
4." 1::••
- - ......--4--...-----1
---------; - - -----..F.--...._____t_____F---.
-1::=4.-b-'''4..-
"1-
---7.4------1-1-1.--1
..-----c ___ -i-----h— •
--1------- :
----
Voigt Model -4-- —1-

: 1 1
_ 4- ---4 1
. :
0 Equ.(5-25) : 1
0.01
1
00 101 0
Fig.(7.7) Correlation of theoretical results of Equ.(5.25) with a single Voigt Model for the
compliance and loss angle of the elastomer buttons
236

7.3.1 Elastomeric Cylindrical Buttons
The prediction method developed in this thesis for buttons was employed in order to
obtain the Voigt model elements. A one element Volgt model is employed and the
results shown in Fig.(7.7) were obtained. The figure shows the Voigt element curve-fit
with the theoretical compliance and the loss coefficient.
1E+08
4- -t---
1
Voigt Model
-4--
4 --+ 0 Experimental
i
-
Ki -1
Ko -444e, Ko=4.086 E7 lillm
.444,.. , 1 Ki =6.974 E7 /film
ci=.580.$2 NsItt
)0 101 0
!-- - ' -i
?-. -i -4- -4 4 4---4
i--- -4 -4-- 4 --4- 4--- 4-
i L. -4 L. 4 -4 4- 4
i i ;
i}--- --1— --i- 4- -4— 4---4----
g- . i
i
I
I
I
i
1
I
:
cr,
7--) 1
! i
0.1 ---+- cap 0-
o .
-2t-
co F-- --- -0D----t-----i-----i----t --- -

0 : !-- - - 4 4-- 4- -4- - 4- -i-- •
,
O -i , -l- + —1 —4-
, --1.
Voigt Model -i --3---- -i- --i . .
i i
4- 4- -4 .
0 Experimental i
i
i
0.01

Fig.(7.8) Correlation of experimental data with a single Voigt Model for the stiffness and
loss coefficients of the elastomer buttons
The curve-fit obtained from the Voigt model is almost the same as the theoretical results
for the compliance and very close for the loss angle (i.e. loss coefficient), as seen in
Fig.(7.7). At low frequencies, a slight deviation from the experimental prediction is
observed.
237
This curve-fit is obtained for a single element Voigt model (see Fig.(7.7) for coefficients).
Increasing the number of Voigt elements will give a better curve-fit to the prediction
values.
Fig.(7.8) shows the comparison of the results obtained from the same Voigt element
coefficients for the real part of the complex stiffness and loss coefficients with the actual
experimental data. As seen in Fig.(7.8) the curve-fit is slightly higher than the
experimental data for the loss coefficient while the stiffness coefficients agree with the
experimental data favourably. This is because the curve-fit is obtained from prediction
formulae that result in slightly higher predictions for loss coefficients, as seen in Fig.(7.2).
The loss angle in Fig.(7.7) and the loss coefficient in Fig.(7.8) are the same. The loss
angle is obtained from
tan 8. H(o))
Hi(c0) (7.1)
and the loss coefficient is obtained from
_ K2(o)
71=
Ki(co) (7.2)
Considering (see Chapter 5)
K1(0)) j. i K2(M)
111(0 ) 4- 112( a) )
= K(w)÷K(co) r K12 (0) ) ÷ IC ( 0) )
(7.3)
It can be shown that
H2 (0)) = K2(0))
or tan3=n
Hi (w) K1(0)) (7.4)
It is clearly demonstrated that a close prediction of the stiffness and damping coefficients
of elastomers and hence the coefficients of the mechanical Voigt model is possible from
their material properties for a given geometry. The procedure followed in order to obtain
the coefficients of the mechanical Voigt model elements can also be employed for
rectangular cartridges and 0-rings. Therefore there is no need to repeat the same
calculations any further. Since the objective of the research is to see whether elastomer
dampers are effective in eliminating or reducing the untoward effects of the system
resonance frequencies, as many elastomeric material and geometry as possible will be
investigated. For this reason a mechanical Voigt model is fitted to the experimental data
provided by the NASA contractor report [Smalley et al., 1977] for 0-rings.
238

1E+08 : 4iliilii•ili..iI..................YIW..I.N.V i i
Voigt Model -------r-i-hlitt-
0 Experimental
-i-
r t -1- 1- 41-1-
—r ---1----/- --1.
i
! 1 i '
T
z 1E+07- i 4 ° CI
^ 4-
Se ------t---- --tt ----
------ -----h
"-----1-------1----r-
- r----
r-- -r- -r- r---1-----t---
-1-- ; - t---1----t---
t -1-1 -t f- - 4--
Ko=3:4-P-Nlitt- -i-----1----i----4-----1----
I0006D5--E6-Nlm----
ci=6059 Nslm 1 I
1E+06
1000
10 --4-
.
i---
:,
-h —4---
[
4 !
_
1
tk
• =zy. .7: -_-: _ ______i
8 1
4---_,
co .4_
Voigt Model I
i
0 Experimental
n
1
0.1
1 do 10
1 .)0
loss coefficients of the 0-rings (Viton70)
7.3.2 0—rin es
In the case of the 0-rings, curve-fits are obtained for the empirical formulae given in the
NASA contractor report [Smalley et al., 19771 as
IC(ca) = Al cap' (7.5)
ih( C0 ) = 1420)32 (7.6)
For each materials and the test conditions a set of A 1 , A 2 , B, and B 2 coefficients are
given by Smalley et al. [19771.
239
1E+08
i –4-, 1 ,
Voigt Model ----------i-----4----i----1-1---H-

1 — t– 1 1 i I
0 Experimental ------F.----r---i--1- i 1----

. -----t----t----i-----t---1--i---
, n
I 1 i
LA:JO elr"C:o o oo o o, 0
t- t-- 1-----4------1----
1 Ko=7.0 E7 NItil
KI=3.425 E7 IVIm
ci=143Q0 NsIM
)0 . 101
CO
0
0.01

loss coefficients of the 0-rings (Viton90)
Fig.(7.9) shows the comparison of the curve-fit values with the actual experimental data
for Viton 70. The curve-fit for the real part of the stiffness is in good agreement with the
experimental data. However, a deviation from the experimental data can be observed for
the lower and higher frequencies in the case of the loss coefficient values.
Fig.(7.10) and Fig.(7.11) show similar curve-fits for 0-rings made of Viton 90 and Buna–N
respectively. In both figures, the curve-fit for the real part of complex stiffness agree quite
favourably with the experimental data whereas the loss coefficients show deviations. For
Viton 90, deviations are observed for the low frequencies of the loss coefficient. These
deviations can be reduced by employing more than one Voigt elements (Gupta et al.,
1974].
240
1E+08 7/MaanEgininil
1=v1111•1111111.11111111 1111111111111111111111111111111111111111111
I .
Voigt Model 11111111151111111MO•11111111
0 Experimental IIIIIIIIMIIIIIIIMIIII
i
1E-1-07 ----h ,
- • - •
• • 'n - al
El
-Ki-
-------.444_
— Ko --1:941E7-litlity
1-
; -
----4- -
KO
----- - — --- --- - 11 -4
--Ki.14--E7411,n--
- -•"144". _L
c
Ci =230p Nslm
—rj--
1E+06
100 101)0
1 ------
-
h- --t h- I
-- . ---t+ i
., h-- -i— h-- ---t --i i--- -,
-----O—O
0 _o_ 0 0 -4 -4 4- ,'
I i i
I
i 1 i
1 i I
i--- .4— --4-. --4 --I. 4---4
-- Voigt Model ------i-----i---- ---i-----H-4--

i i
1
0 Experimental ii
!
0.01
1 00 10()0
Fig. (7.11) Correlation of experimental data with a single Voigt Model for the stiffness and
loss coefficients of the 0-rings (Buna–N)
7.4 Elastomer Dampers in Assembly
Elastomer dampers were assumed to be placed round the outer ring of the both left hand
and right hand side ball bearings. For example elastomer buttons in assembly are shown
in Fig.(7.12). The buttons can be preloaded through screw mechanisms shown in the
figure. Similar arrangements can be designed for ring cartridges and 0-rings.
In the previous section, elastomer dampers were modelled in terms of a mechanical Voigt
model. However, it is generally more convenient to model elastomers as an equivalent
241
stiffness and dash-pot system for mathematical purposes. Any number of Voigt element
models can easily be transformed into an equivalent stiffness and dash-pot system by
adding as many two element Voigt models (see Appendix 1) as is deemed necessary.
Equivalent damping and stiffness coefficients for a single element Voigt model, such as
the one employed in this thesis, are given as (see Appendix 1):
(K0+ K1 )K0K1+ Koci2 w2

IC= x2 2
(K0 + K1 ) + CI CO2
(7.7)
V2 ,
/1 0 lei
ceq = (K0 + K1 ) 2 + Ci2 CO2

(7.8)
where co is the damper excitation frequency and an important parameter; co is obtained

as described in Chapter 5.
Fig.(7.12) The elastomer buttons in the assembly
The outer ring of the bearing was assumed to be mounted on an elastomer button
cartridge as seen in Fig.(7.12). The cartridge has two rows of elastomer buttons, each of
which has three elastomer buttons that are equally distributed around the outer ring.
242
In order to see the effect of the elastomer damper more clearly the shaft was assumed to
be 0.5 gm off-centred. This was introduced into the simulation model as described in
Chapter 5.
0 -VVVW
-1 1
0.025 0.05 0.14 0.17 0.2
time (s)
Fig. (7.13) Transient and steady-state oscillations of the centre of the shaft supported by
bearings mounted on elastomer buttons (n.5000 rpm, 13.0.5 pm, m.5.5 kg, Pr-1O N)
The transient and steady-state oscillations for a shaft speed of 5000 rpm are given in
Fig.(7.13) for a shaft-ball bearing assembly supported on elastomer buttons as seen in
Fig.(7.12). The number of balls in each bearing was 8 and a 10 N of axial preload was
applied. The response shown in Fig.(7.13) suggests steady-state vibrations. However,
in order to see the effect of external dampers the vibrations of the system without
external dampers and the system with different external elastomer dampers, having
different geometries and materials should be obtained.
The vibration response curve for the system without external dampers was obtained by
introducing an arbitrary damping of 300 Ns/m in order to eliminate the effect of the natural
frequency. A resonance peak appears when the shaft rotating speed coincides with the
natural frequency of the system, as seen in Fig.(7.14).
The natural frequency of the system is between 38000 and 40000 rpm. It is higher than
the natural frequency of the geometrically perfect system which is about 34000 rpm. The
difference comes from the force exerted by the out of balance of the shaft centre. The
out of balance of the mass centre forces the balls into regions where the stiffness is
greater and therefore might affect the resonance frequency. Therefore the natural
frequency of the system is pushed to relatively higher frequencies. Jump behaviour is
clearly observed in the vicinity of the natural frequency as the system oscillates at large
amplitudes which forces balls to have large deflections over their nonlinear load-
deflection curve.
2e-5
•
I.
••
• •
••
2..
-o
:E •1;1:
• ••
t 1e-5 - ••
•
••
4L
CO
E.
4'
sg
•
a)
•• •
• 3
••
••••••• •
Oe+0 r• • kt •
10000 20000 30000 40000 50000 60000
shaft speed (rpm)
Point 1 (n=20000 rpm)
1.4
_12 EINENIEN NUM=
MAIGNIMEN EMPIMEN
0.8 EMMEN •ENE•
0.6 •MIUSIN 0.6
0.4 MENNIIEI 0.4 NEnIMEN
I
0.2 MENME 1
02 MENNEN
0 •ENNEN 0 EINNEEIM
-0.6 -0.4 -02 0 0.2 04 0.6 -0.6 -0.4 -0.2 0 02 04 0.6
amplitude, y (um) amplitude, y (um)

12 12
PP"4111 8
.P---4\ 4
Iff/MV1III 0
N .E ;IOW/
-12 -12
-12 -8 -4 0 4 8 12 -12 4-4 048 12
amain," y ( pm) amplitude, y (ism)

3
1.5
-1.5 -1.5
-2 -1 0 2 -2 - 0 1
arn0filuclo . y (Um ) anleiitudd. y (pm)
Fig.(7.14) Shaft centre vibration in the x-y plane for the system without external damping
244
The transient and steady-state oscillations of the shaft centre were obtained for three
shaft speeds along the frequency response curve for the system without external
dampers. The speeds were chosen such that the first (n.20000 rpm) was well below the
resonance frequency, the second (n.36000 rpm) was very close to resonance frequency
and the third (n.50000 rpm) was well above the resonance frequency as seen in
Fig.(7.14). As seen in Fig.(7.14), when the steady-state oscillations are reached, the
shaft centre locus for a shaft speed of 20000 rpm is elliptical whereas for a shaft speed of
36000 rpm it is almost circular.

0 20000 40000 60000
Shaft Speed (rpm)
Fig.(7.15) Peak to peak oscillation amplitudes for different shaft rotating speeds when
the bearings are mounted on the elastomer buttons (e=0.5 Am, Pr-10 N)
The procedure described in Fig.(7.13) was followed for different shaft speeds and the
peak to peak amplitudes were recorded as presented in Fig.(7.15) for elastomer buttons
for example.
The resulting vibration spectrum shows slight indication of jump behaviour. Since the
outer ring of the ball bearing is assumed to be 100 grams, only one peak is observed at
about 31000 rpm. The first superharmonic of the natural frequency also appears in the
spectrum. it is at about 59000 rpm. it is slightly less than twice the natural frequency. This
may be because the actual natural frequency was affected large oscillations and pushed
to a slightly higher frequency.
245
The peak to peak amplitudes of the outer ring centre vibrations are in-phase with the shaft
centre peak to peak amplitudes but at relatively lower values. The peak to peak
amplitudes of the outer ring and the shaft centre oscillations are also given in Fig.(7.15).
This behaviour is expected as for the steady-state oscillations the shaft and the outer ring
movements are In-phase.
2e-5
- •. •
e" ;71•7, ''' ' ...... • • •
,..4•••n .. .................................. ;;;;;;;;
Oe+0

10000 20000 30000 40000 50000 60000
Shaft Speed (rpm)
Fig.(7.16) Peak to peak oscillation amplitudes for different shaft rotating speeds for 0-
rings and buttons external dampers
In addition, the peak to peak amplitude for 0-rings of Viton70, Viton90 and Buna-N were
obtained in a similar way. For each one, when the shaft rotational speed coincides with
the natural frequency, the system resonates. Since the data available for elastomer
dampers are between 100 and 1000 Hz, the results are obtained for speeds between
10000 and 60000 rpm. Since the damping becomes very low for the speeds less than
10000 rpm vibration amplitudes become relatively larger. The outer ring mass is small,
therefore the natural frequency associated with this mass does not appear in the vibration
spectrum. Hence only one peak appears as the natural frequency. The system also
resonates at the first superharmonic of the natural frequency. This quite clear for Buna-N.
Vibration spectrums of the system with and without external dampers are given to the
same scale in Fig.(7.16).
246

2e-5
• without damping
10 0-ring (Viton 70) •e:•
••
••
A••
••• •
••• •;.
•:
• ••
•
•
•
B b.
0000 oo0•008300
44.00-
Do
--titmega
&IS
oe+o 06 0 • • • e. 717 —
10000 20000 30000 40000 50000 60000
shaft speed (rpm)
Point A —Without external dampers (n=36000 rpm)
12 12
111/?-TRIE MENNEN
4 111% 11/11-16n11
EINIEMS•
4
4
0
• 4
MINIM
MICZEN
-12 .......... -12 EMMENE
-12 4 -4 0 4 8 12 -12 4 -4 0 4 a 12
amplitude. y (pm) amplitude, y (pm)
Point B -Viton 70 0-rings as external dampers (n=36000 rpm)

12 12
NUE= MINN=
UMW= MENNEN
MINICZINN NUMMI
NININENN •NNE•N
MENNEN NENE=
-12 MENNEN -12 MONNE
-12-8-404812 -12-8-404812
amplitude, y (pm) amplitude, y (pm
Point B -Viton 70 0-rings as external dampers (resealed —n=36000 rpm)
35 3.5
—3
2.5
x- 2
I1.5
a
o5
0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 05 1 1.5
amplitude, y (pm) amplitude, y pm)
Fig. (7.17) The effect of external dampers on the shaft centre vibrations in the x-y plane
247
Fig.(7.16) shows that the correct choice of elastomer external dampers Is important. For
example, button dampers have worsened the vibration amplitudes rather than eliminating
them. The shaft rotating speed is another important parameter to consider in determining
the material for the external dampers. As seen from Fig.(7.16) for a shaft speed of 30000
rpm, or greater speeds, the best material for the 0-ring dampers is Viton 70. Buna-N
should be considered very carefully as it has violent vibrations at the first superharmonic
of its natural frequency. But for lower shaft speeds (less than 25000 rpm) all external
dampers have a worsening effect.
The natural frequency of the system with external dampers can roughly be determined by
a simple technique. If the equivalent stiffness of the system without external dampers is
K wod, considering the Voigt model stiffness and neglecting outer ring mass and dash-pot
system the new equivalent stiffness for the system is given as
1 _ 1 4. 1 + 1
IC — Kwed Ko K1 (7.9)
The calculations of the natural frequency from Equ.(7.9) roughly matches with the natural
frequencies obtained from the simulation model. For example, the equivalent linear
stiffness coefficient for the system without external dampers is 8.7x10 7 NMI (Its
calculation was based on a natural frequency of 38000 rpm). For 0-rings of Buna-N, the
approximate natural frequency can be calculated from Equ.(7.9) as 15000 rpm assuming
that two 0-ring cartridges are employed; one for the LHS bearing and the other for the
RHS bearing. Similar calculations result in a natural frequency of 26000 rpm for elastomer
buttons and 25500 rpm for Viton 90 (see Figs.(7.8-7.11) for the mechanical Voigt model
coefficients). Although these values are lower than the actual values found from the
vibration response curves, they are of correct magnitude and order as seen in Fig.(7.16).
The difference may be due to the calculation of effective number of 0-rings and buttons
or calculation of equivalent stiffness value of elastomers.
In order to see the comparison the shaft centre loci are given for the system without
external dampers and for the system with Viton 70 elastomeric 0-rings as external
dampers at a speed of 36000 rpm which is very close to the natural frequency of the
system without dampers. As seen in Fig.(7.17) a smaller peak to peak amplitude and
more stable and smooth shaft oscillations are obtained by the use of elastomeric 0-rings
as external dampers.
248

4 4
3.5 3.5
t- 3 3
x 2.5 x 2.5
,
C7 2 ai 2
5
E1.5 = 1.5
71.
0.
CD ci 1
0.5 0.5
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5
amplitude, y (pm) amplitude, y (pm)
Fig.(7.18) Locus of the shaft centre for Buna-N 0-rings as external dampers (e- 0.5 pm,
n-36000 rpm)
Finally, the locus of the shaft centre for Buna-N 0-rings is given at a speed of 36000 rpm
as an alternative to Viton 70 0-rings. However, as also seen from Fig.(7.16) Buna-N
resonates at about this speed. Fig.(7.18) clearly shows that the system is on the verge of
instability. For example in this case employing Buna-N Instead of Viton 70 will have
disastrous effect on the system.
Closure
In the previous chapter, the effect of elastomers as external dampers on the shaft-ball
bearing vibratory system was investigated. An out-of-balance of shaft centre was
considered as a forcing function in the system. Jump behaviour due to the ball to
raceways contact nonlinearity and the nonlinearities of elastomer dynamic properties
were also demonstrated. It was concluded that in order to employ elastomers as external
dampers, their dynamic properties have to be known In advance and their material and
geometry have to be chosen at the design stage. Elastomers have to be seriously
investigated before selection as the wrong choice may worsen the vibrations.
The conclusions of the results obtained in this chapter as well as the previous chapter will
be summarised in the next chapter.
249
CHAPTER 8
CONCLUSIONS
8.1 Introduction
In the previous two chapters the results and discussion of the theories presented in
Chapters 3, 4 and 5 were given. Some of the important conclusions reached will be
summarised in this chapter. The conclusions will be given in the same order that was
followed in the results and discussion chapters. At the end of this chapter suggestions
for further work can also be found.
8.2 Conclusions for Ball Bearings

8.2.1 System Natural Frequency
8.2.1.1 The natural frequency varies for different positions of the ball set and
this variation is cyclic.
8.2.1.2 The number of variations per cage rotation is equal to the number of
balls in the bearing.
8.2.1.3 The amplitude of the change considerably reduces for larger number
of balls in a bearing.
8.2.1.4 Preload changes the natural frequency quite considerably. For

relatively low preloads the change is dramatic whereas for larger preload
values the change is almost linear (see Fig.(6.7)).
250
Chapter 8 Conclusions
8.2.2 Ball Passage Frequencx
8.2.2.1 The BPF is one of the important system characteristics and running
speed should be design to avoid from the resonance.
8.2.2.2 The number of balls in the bearings can be significant in the shaft-
bearing dynamics and should be considered from the vibration
perspective as well as mechanics at the design stage.
8.2.2.3 Increasing the number of balls in an angular contact ball bearing will
significantly reduce the peak to peak amplitude at the resonant
frequencies when there is a fixed external damping coefficient.
8.2.2.4 Increasing the number of balls will decrease the resonant frequency
and this has to be carefully considered at the design stage since the
natural frequency coincides with the excitation frequency at relatively
lower speeds.
8.2.2.5 Preload is one of the important parameters in bearing dynamics and can
be useful in controlling the vibrations of the system.
8.2.2.6 Increasing the preload will reduce the vibration amplitude at the
resonant frequencies when there is a fixed external damping
coefficient.
8.2.2.7 Preload values should be carefully selected for the best performance
from a vibration point of view.
8.2.3 Outer ring Waviness
8.2.3.1 The experimental studies of other researchers are confirmed by the

simulation model presented in this thesis.
8.2.3.2 In the case of outer ring waviness the most severe vibrations occur
when the BPF and its harmonics coincide with the natural frequency.
8.2.3.3 Severe vibrations are observed for a order of waviness k -1 m ± 1 and

k =m.. For some orders of waviness, the vibrations produced due to
waviness are negligible (see Fig.(6.20)).
251
8.2.3.4 The frequency of the vibrations due to outer ring waviness depends on
the waviness order and can be given as
Waviness of orders Vibrations caused by outer ring waviness

k=q m ±p q m coc
However, this equation is only applicable to major peaks. Since the

load deflection relation for balls is nonlinear, for some cases this
equation does not hold as explained in Chapter 6.
8.2.4 Inner ring Waviness
8.2.4.1 Available prediction methods are useful to detect peaks at the major
frequencies due to inner ring waviness while peaks at other
frequencies are also present but they cannot be detected by available
prediction methods.
8.2.4.2 For transformations the peaks can be at (see Chapter 6):

i ca„,p ± j CO or i ( co —co) ± j coc.
8.2.4.3 The formula
Waviness of orders Vibrations caused by inner ring waviness

k=q m ±p q m (co—co) ± p co
holds for the major peaks (see Fig.(6.27)).
8.2.4.4 The most severe vibrations occur for waviness of order i m ±1.
8.2.4.5 For the waviness of order i m, superhamionics of the peak at the WPF
are observed.
$.2.5 Ball Waviness
8.2.5.1 A ball with a wavy surface in the set, will cause vibrations at two
frequencies: the cage speed (we) and wave passage frequency of the
ball (N cob).
252
8.2.5.2 Peaks atN 4 ± i (It will also appear.
8.2.5.3 When N 4 ± i ot , or the cage speed coincides with the natural

frequency of the system, a resonance will occur.
8.2.5.4 Increasing the order of waviness will diminish the vibrations at N cob ±
i coc and only vibrations at the cage speed will remain in the spectrum,
implying a similar case to bearings possessing oversized balls.
8.2.5.5 Relatively higher peaks will appear at N cob ±w.
8.2.6 Off-sized Balls
8.2.6.1 A single off-sized ball within a bearing produces vibrations at the cage
speed. This is true for linear and nonlinear ball to race deflection
coefficients.
8.2.6.2 More than one off-sized ball will produce vibrations according to their
arrangement within the bearing. Symmetric combinations will produce
vibrations at the multiples of the cage speed as shown in Table(6.1). All
other combinations will produce vibrations at the cage speed.
8.2.6.3 Difference in the off-sized ball's size for symmetric combinations will
reduce the vibration speed to that the cage speed unless they are
symmetric within themselves.
8.2.6.4 For bearings with 3, 5, 7, 11, 13, ... balls, for any combination or
arrangements of off-sized balls, vibrations due to the off-sized balls will
take place only at the cage speed (see Table (6.1)).
8.2.6.5 For any ball bearing, the highest radial vibrations due to the off-sized
balls is at a speed of half of the number of balls times the cage speed
(112 m cod .
8.2.7 Misalignment
8.2.7.1 When the outer races of both bearings are misaligned, variation occur
at the cage speed as the same position of the ball set can only be
253
obtained after one cage revolution. The system resonates when the
cage speed coincides with the natural frequency.
8.2.7.2 Outer ring misalignment increases the system natural frequency as it

acts to preload the bearings.
8.2.7.3 When the inner rings of both bearings are misaligned, vibrations occur
at the shaft speed. The system resonates when the shaft speed
coincides with the natural frequency.
8.2.7.4 When only the inner ring of the LHS bearing (or RHS bearing) is
misaligned, the natural frequencies of the system along the x and y
axes are no longer the same. For the x oscillations, small peaks at con ±
co appear in the spectrum. The vibrations in the y direction are at the

shaft speed and its superharmonics while the vibrations in the z
direction take place at twice the shaft speed and its superharmonics.
8.2.8 Defective Running Surfaces
8.2.8.1 When the outer race running surface has a defect, the vibrations take
place at the BPF.
8.2.8.2 When a defect is present on the inner ring running surface of a bearing,
vibrations will take place at four frequencies and their harmonics and
combinations. These frequencies are: the shaft rotating speed, the
cage speed, the ball passage speeds of inner and outer races.
8.2.8.3 When any of the above frequencies coincides with the natural
frequency of the system, a resonance occurs. For relatively low
speeds the vibrations will occur at the natural frequency of the system.
That is, the defect will excite the natural frequencies.
8.2.8.4 A ball defect produces vibrations at twice the ball rotating speed. When
cob or 2cob or their harmonics, coincide with or are close enough to the
natural frequency, resonance occurs and the resulting vibrations are at
the natural frequency.
254
8.3 Conclusions for Elastomers as External Dampers
8.3.1 In this research the dynamic stiffness and damping coefficients of

elastomers from their material properties were derived for different
geometries. These are as follows:
For rectangular cross-section buttons:
3 ab
K= (E + 41-211?-1)
h)
For cylindrical buttons:
3 It d2
K = — — (E + G)(1-2±)
4 h h)
For ring cartridges, two equations were derived. The first is the modified
form of the equation for a rectangular cross-sectional button:
3 ic(Ri + Ro )t (E ÷G)(
1 *
(5.
K=
4 R0—R1 ' R.—R,)
The second equation is as follows:
1 8
2(/?0 C) R.)
96 rc2(R0 — Ri)2(Ri + R0)2
(2(R. - er)2(n(Ri + R.)- 2 err
V (E + G)
K=
2 647t2(R0 —R32(Ri + R0)2
+
(2(R0 — R1)— (51-An(Ri + 1?.)— 2 err
24 ic2(R0 — Ri)2(Ri + R0)2
(2 (R.— Ri)— Srr(n(Ri + R.) —2 &I
For 0-rings three equations were derived. The first one is the modified form
of the equation that was derived for cylindrical buttons:
3 x2d2 (Ri + R.) er

K =+ G)(1—
32 (R. — Ri)
255
The second equation is:
1+ 8 + 967E2d2(Ri+ Ro r
2 d2 R.) (2d — er)2(n(Ri + R.)- 2
V(E+G) 641t2d2(R1 +R.)2

K—
2 (2d — 803 (E(Ri + R.) —2 Er)3
24 n2 d2(Ri + R0)2
(2 d — 8r)4(7c(Ri + R0 )— 251)2
The third method is similar to the beam-column method including radius

effects:
a (d — Or )80 cos2 (g) + Gsin2(c))clg

K= 4
a or
c-CO 1 dr
1
mit, [ 15 r
This equation was solved using Mathematica and it gave the successful
results presented in Chapter 7.
8.3.2 All predictions match with the experimental data available quite favourably.
8.3.3 It was shown that in order to employ elastomers as external dampers, their
dynamic properties have to be known in advance and their material,
geometry and working environment have to be chosen at the design stage.
They have to be seriously investigated before selection, as the wrong
choice may worsen the vibrations.
256
8.4 Recommendation for Further Research
8.4.1 In this thesis the cases investigated are generally for steady-state vibrations.
However, for cutting machines, especially for surface processing machines,
transient vibrations are important and of interest. Therefore a careful
investigation of this behaviour is needed. Particularly for the dynamic
properties elastomers the transient vibrations have to be taken into account.
8.4.2 In this thesis the anomalies of the system and the defects are investigated
separately. In real life more than one defect is simultaneously present in the
system. Moreover, different variation of the same defect such as more than
one waviness order, more than one defect on the inner, outer and ball
surfaces, the inner ring and the outer ring misalignment together can be
present in the system. Their interaction has to be investigated.
8.4.3 In this thesis the masses of the balls are not considered. Particularly for high
speed applications they are important and should be modelled. This can
easily be added to the model presented in this thesis.
8.4.4 The shaft was assumed to be rigid. The behaviour of an elastic shaft
supported by ball bearings has to be studied. This can be done by
linearising the load-deflection relation of balls over a narrow band of
deflection as presented in Chapter 2 and Chapter 3.
8.4.5 In this thesis the lubricant film between the rolling elements and the
raceways was not considered. But in real life all ball bearings naturally
include (hopefully) a lubricant film. Investigation of the effect of lubrication
on the misalignment and varying preload may be a subject to study.
8.4.6 Different types of excitation forces are also of interest. For example random
excitation and step force excitation are very important for CNC cutting
machines.
8.4.7 Above all, of course, it is necessary to prove the results presented in this
thesis by more experimental findings. In particular, natural frequency
variation, off-sized balls and misalignment are investigated less than other
characteristics. This is also true for the dynamic properties of elastomers.
8.4.8 Elastomers have to investigated more in detail. Particularly the stiffness and
damping properties of elastomeric 0-rings are not very well known.
257
8.4.9 The response of elastomeric materials to the non-periodic excitation has to

be investigated since little is known about it.
8.4.10 Interaction between material temperature and vibration amplitude has to be

considered as the temperature dependence of elastomers is one of the
main draw-backs of the usage of elastomers.
8.4.11 Available data was for a frequency range of 100 to 1000 Hz. A check has to
be done whether the prediction formulae presented in this thesis are valid
for other frequency ranges.
258
References
REFERENCES
Aini, R. Vibration monitoring and Modelling of Shaft/Bearing Assemblies

under Concentrated Elastrohydrodynamic Conditions
Ph.D. Thesis, Kingston Polytechnic, CNAA,1990, U.K.
Aklonis, J.J. & Introduction to Polymer Viscoelasticity

MacKnight, W.J. John Wiley and Sons, 1983, U.S.A.
AktOrk, N. Vibration of a Shaft Supported by Angular Contact Ball Bearings

M.Sc. Thesis, Imperial Col. of Scien. &Tech. ,1988, England
AktOrk, N.; "Vibration of a Shaft Supported by Angular Contact Ball Bearings"

Uneeb, M. & ASME Paper presented at ESDA Conference
Gohar, R. June 29-July 3, 1992, Istanbul, Turkey
Ariyoshi, S.; "Motion of a Ball in a Deep-groove Ball Bearing under static

Kawakita, K. & Moment Loading"
Hirano, F. J. JSME Int. Ed. No:2, April 1981, pp 99-104
Bachschmid, N.; "Non-linear Vibrations due to Roller Bearing Clearances"

Curami, A. & 5th Int. Con. on Vibrations in Rotating Machinery 7-10 Sept. 1992
Del Vescovo, D. Bath, U.K., I.Mech.E. C432/025 pp 591-596
Beards, C. F. Vibration Analysis and Control System Dynamics

Ellis Harwood Ltd., 1981, England.
Booser, E.R. "Recent Advances in Grease Lubrication of Ball Bearings"

Proc. The Conference on Lubrication and Wear, 1-3 Oct. 1957
London, I.Mech.E., pp.425.
Bratihwaite, E.R. "MoS2, Second Thoughts"

Industrial Lubrication, Vol.21, No:8 (1969) pp.241-247
259
References
Brädlein, J. "Properties of Rolling Bearing Supported Machine Tool Spindles"

Ball and Roller Bearing Engineering, 1985 pp 44-54
Braun S. & "Analysis of Roller/Ball Bearing Vibrations"

Danter, B. Trans. ASME J. Mechanical Design, Vol.101, Jan.1979,
pp 118-125
Cena, K.M. & "The Effects of Ball Quality, Radial Clearance and Grease
Hobbs, R.A. Specification on the Noise and Vibration of an Electric Motor
Tribology Conventions, 1972, IMechE, London.
Chiang, T., "Development of Procedures for Calculating Stiffness and

Tessarizik, J. M. & Damping Properties of Elastomers in Engineering Applications,
Badgley, R. M. Part I: Verification of Basic Methods"
NASA Contractor Report No. CR-120965, March 1972.
Prepared by MTI for NASA-Lewis Research Cenntre Under
Contract NAS3-18546
Dahl, S. R. & "A Derivation of Equevalent Linear Viscous and Elastic

Rice, R. B. Constants for Viscoelastic Materials"
Vib. Damping 1984 Workshop Proceedings, p 0-3
Dareing, D.W.& "Fluid Film Damping of Rolling Element Contacts Vibrations"

Johnson, K.L. J. Mechanical Engineering Science, IMechE. Vol. 17, No: 4,
1975, pp. 214-218
Darlow, M. S. & "Development of Procedures for Calculating Stiffness and

Smalley, A. J. Damping Properties of Elastomers in Engineering Applications,
Part IV: Testing of Elastomers under a Rotating Load"
NASA Contractor Report No. CR-135355, November 1977.
Contract NAS3-18546
Den Hartog, J. P. Mechanical Vibrations

Dover Publication, Inc., 5th Ed., 1985, U.S.A.
Doi, S. "An Experimental Study on Chatter Vibrations in Grinding

Operations*
Trans. ASME 1958 vol.80, pp.133-140
Dowson, D. "Monitoring - 3.Vibration"

Tribology Vol.3 No.4, Nov.1970 pp.216.
260
References
Dowson, D. History of Tribology

Longman, 1979, London
Eisenbeis, H. Vibration of Angular Ball Bearings

M.Sc. Thesis, Imperial Col. of Scien. &Tech. , 1985, England
Ellis, E.E. "Misalignment Ball Bearings"

J. Tribology, Feb. 1970, pp.29

Fersht, R.S.; "Viscoelastic Characteristics of an Elastomeric Isolator

Fersht, S.N. & Damping 1986 Proceedings, Vol.1, pp. AC.1-AC.12
Denice, JR., M.
Frarey, J.L. Discussion of "Vibration Characteristics of Ball Bearings"

Trans. ASME J. Lubrication Technology, April 1977, pp. 288
Franco, J. Vibration of A Rotor Suppoted by Radial Ball Bearings

MSc Thesis, Imp. Col. of Scien., Tech. & Medicine, 1990, London.
Franco, J; "Vibration of a Rigid Shaft Supported by Radial Ball Bearings with

AktOrk, N. & Several Defects"
Gohar, R. ASME Paper No: 92-GT-164 presented at Int. Gas Turbine and
Aeroengine Congress and Exposition, Cologne,
Germany, June 1-4, 1992.
Freakley, P.K. & Theory and Practice of Engineering with Rubber

Payne, A.R. Applied Science Publishers Ltd., London, 1978.
Fukata, S; "On the Radial Vibration of Ball Bearings (Computer Simulation)"

Gad, E.H. et at. Bul. JSME Vol.28, No: 239, May 1985.
Gad, E. H.; "Computer Simulation of Rotor Axial and Radial Vibrations

Fukata, S. & Based on Ball Bearings"
Tamura, H. Memoirs of the Faculty of Engineering, Kyushu University,
Vol 44, No: 2 June 1984
Gad, E. H.; "Computer Simulation of Rotor Radial Vibration due to Ball Bearigs"
Fukata, S.& Memoirs of the Faculty of Engineering, Kyushu University,
Tamura, H. Vol 44, No 1 June 1984
261
References
Gad, E.H.; "Spring Property of Ball Bearing"

Kandou, T. & Memoirs of the Faculty of Engineering, Kyushu University,
Tamura, H. Vol 43, No 3 June 1983
Genertwala, S. N. "Fourier Transform Mechanical Analysis for Determining the

& Rotz, C. A. Nonlinear Viscoelastic Properties of Polymers"
Poly. Eng. and Sci. January 1987, Vol 27, No. 2., pp.165-178
Gent, A. N. & "The Comperission of Bonded Rubber Blocks"

Lindley, R. B. Proc. Inst. Mech. Engrs., Vol. 173, No. 3, L59.
GObel, E.F. Rubber Spring Design

Newnes-Butterworths, 1974
Green, I.& "Pressure and Squeeze Effects on the Dynamic Characteristics

• Etsion, I. of Elastomer 0-rings Under Small Reciprocating Motion"
Trans. ASME J. Tribology, Vol. 108, July 1986 pp.439-445
Gunter, E. J. "Influence of Flexibly Mounted Rolling Element Bearings

on Rotor Response, Part 1-Linear Analysis"
Trans. ASME J. Lubrication Technology, Jan. 1970, pp. 59
Gupta, P. K., "Development of Procedures for Calculating Stiffness and

Tessarizik, J. M. & Damping Properties of Elastomers in Engineering Applications,
Cziglenyi, L. Part II: Elastomer Characteristic under Constant Temperature"
NASA Contractor Report No. CR-134704, April 1974. Prepared by
MTI for NASA-Lewis Res. Centre Under Contract NAS3-18546
Gupta, P. K.; "Frictional Instabilities in Ball Bearings"

Tribology Trans., Vol.31, No:2, 1988 pp. 258-268.
Gupta, P. K. "Dynamics of Rolling Element Bearings, Part I: Ball Bearing

Analysis"
Trans. ASME J. of Lubrication Technology, July 1979, pp.312-318.
Gupta, P. K. "Dynamics of Rolling Element Bearings, Part I: Ball Bearing Results"

Gupta, P. K. "Dynamics of Rolling Element Bearings, Part I: Cylindrical Roller

Bearing Analysis"
262
References
Gupta, P. K. "Dynamics of Rolling Element Bearings, Part I: Cylindrical Roller

Bearing Results"
Gupta, P.K. "Transient Ball Motion and Skid in Ball Bearing"

Trans. ASME J. Lubrication Tech.Vol.97, No.2, April 1975, pp.261
Gupta, P.K.; "Vibration Characteristics of Ball Bearings"

Winn, L.W. & Trans. ASME J. of Lubrication Technology April 1977, pp. 284-289
Wilcock, D.F.
Gupta, P.K. "On the Geometrical Imperfections in Ball Bearing"

Trans. ASME J. Tribology.Vol.110, January 1988, pp.19-25
Gurney, J.P.; "A graphical Analysis of Regenerative Machine Tool Instability'

Tobias, S.A. Tran. ASME J. Engineering for Industry, Feb. 1962.
Gustafsson, 0.G.; Research Report on Study of The Vibration Characteristics of

Tallian, T. et al. Bearings
Dec. 1963, Report: AL63IO23, Reg:585 14:422 3 SKF Ind., Inc.
Hall, L.F. "A Review of The Papers on The Lubrication of Rolling Bearings
and Gears"
Proc. The Conference on Lubrication and Wear, 1-3 Oct. 1957
London, I.Mech.E., pp.430.
Hailing, J. "The Importance of Contact History in the Assessment of

Tribological Systems"
Tribology, August 1971, pp 155-157
Harris, M. C. & Shock and Vibration Handbook, Vold

Crede, C. E. Mc.Graw-Hill Book Comp., 1961, U.S.A.
Harris, T. A. Rolling Bearings Analysis

John Wiley & Sons, 1991, U.S.A.
Harris, T. A. "An analytical Model to Predict Skidding in Thrust-loaded,

Angular Contact Ball Bearings"
Trans. ASME J. Lubrication Technology, Jan. 1971 pp 17-24
Harris, T.A. & "Rolling Element Bearing Dynamics"

Mindel, M.H. Wear, Vol.23 (1973), pp.311-337
263
References
Hedman, T. Elastomers as Damper for Rolling Element Bearings

MSc Thesis, Imperial College, 1987, U.K.
Hemmings, R.C. & "Information from Bearing Vibration"

Smith, J.D Conference on Vibration in Rotating Machinery, Churchill College,
University of Cambridge, 15-17 September, 1976, pp.117-121.
Hertz, H. "On the Contact of Rigid Elastic Solids and on Hardness"

Miscellaneous Papers
MacMillan and Co., 1896, England.
Hess, H. "Ball Bearings"

Trans. ASME 1908, Vol.29, pp. 367-419.
Hite, G.C.; "Vibration and Fretting Corrosion of Instrument Ball Bearings"

Mahie, H. H. & Trans. ASLE, Vol. 15, No: 1 pp 25-36
Hurst, J.
Igarashi, T. & "Studies on The Vibration and Sound of Defective Rolling Bearings
Hamada, H. (First Report: Sound of Ball Bearings with One Defect)"
Bull. JSME, Vol.25, No.204, June 1982, pp 994
Igarashi, T. & "Studies on The Vibration and Sound of Defective Rolling Bearings
Yabe, S. (Second Report: Vibration of Ball Bearings with One Defect)"
IgarashisT. & "Studies on The Vibration and Sound of Defective Rolling Bearings
Kato, J. (Third Report: Vibration of Ball Bearings with Multiple Defects)"
Ishida, Y.; "Effects of Quartic Nonlinear Restoring Forces on 1/2-Order

Ikeda, T. & Subharmonic and Summed-and Differential Harmonic Oscillations
Yamamoto, T. of a Rotating Shaft"
Bul. JSME Vol. 29, No: 258 Dec. 1986, pp 4326-4333
lshida, Y.; "Non-stationary Vibration of a Rotating Shaft with Non-linear

Ikeda, T.; Spring Characteristics During Acceleration Through a
Yamamoto, T. & Critical Speed"
Murakami, S. JSME Int. Journal, Series III, Vol. 32, No: 4, 1989, pp 575-584
Jones, A. B. "Ball Motion and Sliding Friction in Ball Bearings"

Trans. ASME, J. Basic Engineering, March 1959, pp 1-12
264
References
Johansson, L. "Bearing Noise in Electric Motors"

The Ball Bearing Journal 200, pp.28-31
Kanai, H.; Abe, M. & "Estimation of the Surface Roughness on the Race or Ball of
Kido, K. Ball Bearings by Vibration Analysis"
Trans. ASME J. Vibration, Stress, and Reliability in Design
Vol. 109, January 1987 pp. 60-68
Karakurt, M.C. An Investigation into Rolling Element Bearing Diagnostics

via Digital Signal Processing
MSc Thesis, Middle East Tech. University, 1989, Turkey
Kazimierski, Z.& "Stability Threshold of Flexibly Supported Hybrid Gas Journal

Jarzecki, K. Bearings"
Trans. ASME J. Lubrication Tech. Oct. 1979, Vol. 101, pp 451-457
Khan, A.F. Condition Monitoring of Rolling Element Bearings: A Comparative

Study of Vibration Based Techniques"
Ph.D. Thesis, University of Nottingham, 1991.
Kilicay, 0. "Determination of Bearing Deflections"

Proc. I.Mech.E., Vol. 197 C, December, 1983, pp 277-281
Kline, S.J. Similitude and Approximation Theory

McGraw-Hill, 1965, U.S.A.
Kluesener, P.E.M.F. "Comparison of Damping Material Properties by Various Test

Methods"
Damping 1986 Proceedings, Vol. 1, pp AB.1-AB.22
Larsson, J. Mounting of Aerostatic Bearings on Flexible Supports

M.Sc. Thesis, Imperial College, 1989, UK.
Li, X. and lnasaki, I. "Monitoring of Rolling Element Bearing by Using an AE Sensor

Japanese Journal of Tribology, Vol.34, No: 4, 1989.
Matsubara, M.; "Computational Modelling of Precision Spindles Supported by Ball

Rahnejat, H. & Bearings"
Gohar, R. Int. J. Mach. Tools & Manu. August 1988
Matsubara, M. Vibration of An Elastic Shaft Supported by Ball Bearings

M.Sc. Thesis, Imperial Col. of Scien. &Tech. , 1983, England
References
McCrum, N. G.; Principle of Polymer Engineering

Buckley, C. P. Oxford Science Publication, 1988, England
Bucknall, C. B.
Mehdigoli, H.; "Vibration Response of Wavy Surfaced Discs in EFID Contact"

Rahnejat, H. & Submitted for publication to WEAR, 1988.
Gohar, R.
Meldau, E. "Elastische Spielschwingungen konstant belasterer Walzlager

Werkstatt Betr., Vol.85, No. 2 (1952), pp.56.
Meldau, E. "Die Bewengung der Achse von Walzlagem bei geringen

Drahzahlen"
Meyer, L.D.; "An Analytical Model for Ball Bearing Vibrations to Predict
Ahlgren, F.F. & Vibration Response to Distributed Defects"
Weichbrodt, B. Trans. ASME J. Mechanical Design,Vol.102, Apr. 1980,
pp.205-210.
Mitchell, L.D. "Signal Processing and the FFT Analyzer

Experimental Techniques, Oct. 1985.

Moritomo, S. & "Present High Speed Tool Spindles"
Ota, M. Bul. Japan Soc. of Precision Engineering, Vol. 20, No:1
March, 1986, pp 1-6
Nashif, A. D. et al Vibration Damping

John Willey & Sons, 1985, USA.
Newland, D.E. An Introduction to Random Vibrations and Spectral Analysis

Longman, 1975, England
Nicholich, P. "Equipment for Evaluating the imbalance of Rolling Elements for

Aerospace Bearings"
Ball and Roller Bearing Engineering (FAG), 1990-501, pp 39-42
Nishio, K.; "An Investigation of the early detection of Defects in Ball Bearings
Hoshiya, S.; by the Vibration Monitoring"
Miyachi, T. & Trans. ASME J. Lubrication Technology, Sept. 1979.
Matsuki, I.
Olsson, J. Vibrations of Angular Contact Bearings with Dampers

M.Sc. Thesis, Imperial College, 1986, UK.
266
References
Palmgren, A. Ball and Rolling Bearing Engineering

S.H. Burbank & Co., 1946, USA
Perret, H. "Die Lagerluft als Bestimmungsgrqe fOr die Beanspruchung

eines Wälzlagers"
Perret, H. "Elastische Spielschwingungen konstant belasterer Walzlager"

Powell, J.M. & "A study of High Speed Machines with Rubber Stabilised
Tempest, M.C. Air Bearings"
Trans. ASME J. Lubrication Technology Oct. 1968 pp. 701-708
Ragulskis et al Vibration of Bearings

Oxonian Press, Trans., 1979, India.
Rahnejat, H. & "Vibration of Radial Ball Bearings"

Gohar, R. Proc. Instn. Mech. Engrs. Vol. 199 No: C3
Rahnejat, H. Influence of Vibration on the Oil Film in Concentrated Contacts.

Ph.D. Thesis, Imperial Col. of Scien. &Tech., 1984, England
Rao, S. S. Mechanical Vibrations

Addison-Wesley Pub. Co., 1986, U.S.A.
Roman, A.J. & "Robbin Type Retainers for Ball and Roller Bearings"
Gohar, R. I.Mech.E., 1975., pp.755.
Sayles, R & "Surface topography and Rolling Element Vibration"

Poon, S.Y. J. Precision Engineering, Vol. 3, No: 3, 1981, pp. 137-144
Scott, D. "Bearing Failure Diagnosis and Investigation"

Proc. of the First World Conference in Industrial Tribology,
New Delhi, 1972, 11-18 December, pp. 117-121
Shimizu, H. & "VVibration of Rotor Based on Ball Bearing

Tamura, H. (1st Report, Static Stiffness of Ball Bearings)"
Bull. JSME Vol.9, No.35, 1966 pp.524
SKF "Some Contributions by SKF t Rolling Bearing Technology During

the Last Half Century"
The Ball Bearing Journal, 1957, No:1, pp 3-10
SKF "Clearance in Rolling Bearings"
The Ball Bearing Journal, 1959 pp 2-6
267
References
SKF "The Measurement of Vibration in Ball Bearings"

The Ball Bearing Journal, 1961
Smalley, A. J. & "Development of Procedures for Calculating Stiffness and

Tessarizk, J. M. Damping Properties of Elastomers in Engineering Applications,
Part Ill: The Effect of Temperature, Dissipation Level and Geometry"
NASA Contractor Report No. CR-134939, November 1975.
Contract NAS3-18546
Smalley, A. J.: "Stiffness and Damping of Elastomeric 0-Ring Bearing Mounths"

Dariow, M. S. & NASA Contractor Report No. CR-135328, November 1977.
Mehta, R.K. Prepared by MTI for NASA-Lewis Research Cenntre Under
Contract NAS3-19751.
Smalley, A. J.:I. "Testing for Material Dynamic Properties"

Tessarzik, J.M. & ASME Publ.: Vibration Testing Instrumentation and Data Analysis,
Badgley, R.H. AMD Vol.12, 1975. pp.117-141.
Snowdon, J. C. "Representation of the Mechanical Damping Possesed by

Rubberlike Materials and Structures"
J. Acoust. Soc. Amer., 1963 35(6), pp. 821-829
Sommer, J. G. & "Factors Controlling the Dynamic Properties of Elastomeric

Meyer, D. A. Products"
SAE Paper No. 720267, SAE/ASTM Sym.
Detroit, Michigan, January 1973., pp. 109-117.
Stribeck, Prof. "Ball Bearings"

Trans. ASME 1908, Vol. 29, pp. 420-467
Stoker, J. J. Nonlinear Vibrations

Interscience Pub., 1950, U.S.A.
Tallian, T.E. & "Progress in Rolling Vibration Research and Control"

Gustafsson, 0.G. ASLE Transactions Vol. 8, 1965, pp. 195-207.
Tamura, H.& "Vibration of Rotor Based on Ball Bearing (2nd Report, Static
Shimizu, H. Stiffness of Ball Bearings Containing a Small Number of Balls)"
Bull. JSME Vol.10, No.41, 1967 pp.
268
References
Tamura, H.& "Vibration of Rotor Based on Ball Bearing (3rd Report, Static
Shimizu, H. Stiffness of Ball Bearings Containing a Large Number of Balls)"
Bull. JSME Vol.11, No.47, 1968 pp.
Tamura, H; "On the Static Running Accuracy of Ball Bearings"

Gad, E.H.; Bul. JSME Vol.28, No: 240, June 1985.
Kondou, T. et al.
Tamura, H.; "On the Static Running Accuracy of Ball Bearings (2nd Report,
Kondou, T; Results and Discussion)"
Ayabe, T. et al. Bul. JSME Vol.29, No: 257, June 1986.
Taylor, D. L. & "Analysis and Design of Segmented Dampers for Rotor Dynamic
Fehr, V. S. Control"
Trans. ASME J. of Lubrication Tech. January 1982, Vol. 104, p 84
Taylor, R.D. The Prediction of Dynamic Characteristics of Rotating Spindle

Systems
Ph.D. Thesis, The University of Bristol, 1981.
Tecza, J. A.; "Design of Elastomer Dampers for a High-speed Flexible Rotor

Smalley, A.J. & Mech. Eng. Vol. 101, Nov 1979.
Cunningham, R.E.
Tecza, J.A.; "Elastomer Mounted Rotors -An Alternative for Smoother

Dariow, M.S.; Running Turbomachinery"
Jones, S.W.; ASME Paper 79GT149, presented at the Gas Turbine Conference
Smalley, A.J. & and Exhibition and Solar Energy Conference in San Diego
Cunningham, R.E. March 12-15, 1979
Thomas, T. R. Rough Surfaces

Longman, 1982, London
Thomson, W. T. Theory of Vibration with Applications

Allen & Unwin Ltd. 3rd Edi., 1988, U.S.A.
Timoshenko, S.; Vibration Problems in Engineering

Young, D. H. & John Wiley & Sons, 4. Edi., 1974, U.S.A.
Weaver, JR.,W.
Tindale, P. "An Innovation That Became a Science"

Eureka on Campus, Autumn, 1988.
269
References
Tobias, S.A. & "The Vibration of radial drilling Machines under test and working
Fishwick, W. Conditions"
Proc. IMechE, Vol. 170, 1956, pp. 232-247
Tobias, S.A. "The Vibration of vertical milling Machines under test and working
Conditions"
Proc. IMechE, Vol. 173, 1959, pp. 474-494
Tobias, S.A. Machine Tool Vibration

Blackie Pub., London, 1965.
Vierck, R. K. Vibration Analysis

Harper & Row, 2. Edi., 1979, U.S.A.

Watford, T.L.H. & "The measurement of the Radial Stiffness of Rolling Element
Stone, B.J. Bearings under Oscillating Conditions"
J. Mechanical Engineering Science, I.Mech.E., Vol. 22, No:4,
1980, pp.175-181.
Wallin, E. "Preventive Maintenance of Rolling Bearings"

The Ball Bearing Journal No.146, 1966, pp.13-18
Walters, C. T. "The Dynamics of Ball Bearings"

Trans. ASME J. Lubrication Technology, Jan. 1971, pp 1-16
Wardle, F. P. & "Rolling Bearings Noise, Cause and Cure"

Poon, S. Y. Chart. Mech. Engineering, July/August, 1983
Wardle, F.P. "Vibration Forces Produced by Waviness of The Rolling Surfaces

of Thrust Loaded Ball Bearings, Part 1: Theory"
Proc. I.Mech.E. Vol 202 NO C5, 1988, pp305.
Wardle, F.P. "Vibration Forces Produced by Waviness of The Rolling Surfaces

of Thrust Loaded Ball Bearings, Part 2: Experimental Validation"
Proc. I.Mech.E. Vol 202 NO C5, 1988, pp 313.
Wardle, F. P.; "Dynamic Test for measuring the non-repeatable running accuracy
Lacey, S.J. & of rolling bearings"
Haw, C. Precision Engineering, Vol. 10, No:4, Oct. 1988, pp. 221-227.
Weimar, H.C. Discussion part of the paper by Doi [1958]

"An Experimental Study on Chatter Vibrations in Grinding
Operations"
Trans. ASME 1958 vol.80, pp.133-140
270
References
Yamamoto, T. & "The Particular Vibration Phenomena due to Ball Bearings at the
Ishida, Y. Major Critical Speeds"
Bul. of JSME, Vol. 17, No: 103, Jan., 1974, pp. 59-67.
Yamamoto, T; "Oscillations of A Rotating Shaft with Symmetrical Non-linear

lshida, Y. & Spring Characteristics"
Kawasumi, J. Bul. of JSME, Vol. 18, No: 123, Oct., 1975, pp. 965-975.
Yamamoto, T; "Subharmonic and summed-and-Differential Harmonic Oscillations

lshida, Y. in an Unsymmetrical Rotor
Ikeda, T. & Bul. of JSME, Vol. 24, No: 187, Jan., 1981, pp. 192-199.
Yamada, M.
Yamamoto, T; "Vibrations of a Rotating Shaft with Rotating Non-linear

lshida, Y. & restoring Forces at the Major Critical Speeds"
Ikeda, T. Bul. of JSME, Vol. 27, No: 230, August, 1984, pp. 1728-1736.
Yamamoto, T; "Super-summed-and-Differential Harmonic Oscillations in a

lshida, Y. & Symmetrical Rotating Shaft System"
Ikeda, T. Bul. of JSME, Vol. 28, No: 238, Apr., 1985, pp. 679-686.
Yamamoto, T; "Sub-combination Tones of a Rotating Shaft due to Ball Bearings"

Ishida, Y.; Bul. of JSME, Vol. 24, No: 196, Oct., 1981, pp. 1844-1852.
Ikeda, T. &
Yamada, M.
Yamamoto, T; "The Particular Vibration Phenomena due to Ball Bearings at the

Ishida, Y. & Major Critical Speed (2nd report: On the effects of symmetrical
Kawasumi, J. nonlinear spring characteristics)"
Bul. of JSME, Vol. 20, No: 139, Jan., 1977, pp. 33-39.
Yhland, E. & "Analysis of Bearing Vibration"

Johansson, L. The Ball Bearing Journal No.161, 1970.
Young, R.J. Introduction to Polymers

Chapman and Hall, London, 1987.
- The Unknown Leonardo

Hutchinson, 1974
271
Appendix 1 Equivalent Dash-pot System
APPENDIX 1
THE EQUIVALENT STIFFNESS

AND DAMPING COEFFICIENTS
K2
Fig.(A1.1) Two element Voigt model and equivalent dash-pot system
A two element Voigt model is assumed to represent the elastomer. For this, the force-
equilibrium equation can be given as
(x1 . — x2 )K1 + ( II — i2 ) c1 = F1
(A1.1)
( X2 — X3) K2 4- (X2 — X3) C2 = F2
(A1.2)
And the force equilibrium equation for the equivalent dash-pot system can be given as
( X1 — X3) K ( I I — i 3) c F3
(A1.3)
knowing F; = F2 = F3 = O. , X3 = X3 = O. and letting
xl = Aen and x2 = B elau
The equivalent stiffness and damping can be obtained as
1 c . —(K 1 + K2)(c
1 c co
2 — K 1 2 )4-
2 K (K 1 c 2 + K 2 c 1 )(
/ Nc1 + c2) W 2
(K1 + K2 ) 2 + (CI + C2 )2 CO 2
(A1.4)
( CI ÷ C2)(c, c2 co 2 — K1 K2)+(Ki + K2 )(K,c2 + K2 CI)
ciq = \2 i n2 2
(K1 + K2 ) + kC1 + c2
(A1.5)
272
Appendix 2 Phase Plane Representation
APPENDIX 2,
PHASE PLANE REPRESENTATION
Phase plane representation is a very useful and widely used tool in the investigation of
nonlinear systems. In the phase plane representation, the displacement of the shaft
centre, x (t), is plotted against its derivative with respect to time, dxl dt at small time
intervals of At.
The response of the bearing system is initiated with xo and (ax/at )0, and the oscillations
eventually reach a stable amplitude known as the limit cycle. There is only one closed
trajectory for the system which is independent of the initial conditions. Initial conditions
may be inside or outside the closed trajectory. For conservative nonlinear systems, the
limit cycles will respectively be more elliptical than those of respectively less nonlinear
systems to the same scale. The limit cycle occurs over one steady-state cycle, indicating
that the net energy input from the excitation is equal to the energy dissipated within the
system [Rao, 1988].
For a simple equation of motion
I + f(i(t),x(t))= 0 (A2.1)
when &lac . 0 and f(x(t)) * 0, the slope of (ax/at ) becomes
a±(r) . fix(0)
a x(t) i(t) (A2.2)
For example,
a2x(r) ax(r)
M at2
— + c— + K x(t) = 0
at (A2.3)
273
Since axi at - 0, Equ.(A2.3) becomes

a2x(t1
M at2 + K x(t)= 0
(A2.4)
or
at(t) = --
K (x(t))
_
ax(t) Mx(t) (A2.5)
As Equ.(A2.2) suggests the slope becomes infinite when axiat . 0 with the trajectories
of the phase plane crossing the x -axis at right angles. These phase plane trajectories are
symmetrical about the x -axis and the amount of deviation from a circular path trajectory
indicates the extent of nonlinear behaviour in the system.
....... ii
IV
%S.
. A
b
T 1.-

C a
a. .. a.
Fig.(A2.1) Phase-Plane trajectory
Fig.(A2.1) shows the phase plane trajectory for a nonlinear system. For this phase plane
trajectory
11(t)}2 + (x — c) =1
b2 a2 (A2.6)
where 2a is the length of the major axis, 2b is that corresponding to the minor axis, and
c is the centre of the trajectory (see Fig.(A2.1)).
Let V(x) be the potential energy of the system. The energy equation for the
conservative system at any instant can be written as
274
1 M[i(t)12 + V (x(t)) = E = Constant

2 (A2.7)
For the minimum potential energy level
aV(t) = 0
ax(r) (A2.8)
Hence for (a2x /at 2) -0, x(t) - c.
For the phase-plane trajectory, the full period of oscillation can be obtained from the
closed loop as [Rahnejat, 1984
T = idix((:))
(A2.9)
Because of the symmetric nature of the trajectory, the period of oscillations, T, can be
evaluated using
.-.... dx(t)
T = 4 .1.., go
(A2.10)
If the trajectory is obtained for the undamped system for a step load at zero speed, then
(A2.11)
Solving Equ.(A2.6) for axiat will give
1
b [ 2
.40 = T,a — (x — 012 (A2.12)
Substituting Equ.(A2.12) into Equ.(A2.10) will give
d(t)
T=41r1--
bx-.
[a2 — (x — c)2]
(A2.13)
Let
275
x—c=asin4 (A2.14)
dx(t)= a cos 44 (A2.15)

Substituting Equ.(A2.14) and Equ.(A2.15) into Equ.(A2.13)
a 512
T = 4— c14
b 41 (A2.16)
i and 42 in Equ.(A2.16) can be obtained from Equ.(A2.14), considering x = c+a
it
4 1 =0 and 2 -= —2
Hence Equ.(A216), the period of oscillations, will take the form of
a
T = 2ic —
b (A2.17)
276
Appendix 3 Large Deformations
APPENDIX 3
LARGE DEFORMATIONS
OF ELASTOMERS
A specimen of initial dimensions x i , yi and zi is deformed so that the dimensions

become x, y and z. End A of the representative chain at x i , y i and zi before
deformation; after deformation is at x,y and z. The end-to-end vector changes in exactly
the same way as the specimen dimensions (see Fig.(A3.1)). For this single chain, the
work done by external forces to deform it from x i ,y i , zi dimensions to x,y ,z , can be
found from, for example in the x direction:
.,xx
w.= K xdx
f..., (A3.1)
where
K= p2
(A3.2)
Fig.(A3.1) Representation of elastomer deformation
277

Appendix 3 Large Deformations
Hence for a single chain, the work done is
2kT 2 2
Ws =--r Xi (ax-1)
(A3.3)
The total work in the x direction is
v kT
=Ew.= Hat —1)E4
1 P2 1 (A3.4)
By definition [Mc Crum et al., 19891:
Xi2 = 1/(X2).
1 (A3.5)
where <x 2 > i is the mean square value of x i in the undeformed state. The undeformed
state is isotropic, so:
IA
= (z2). /31'
(x2), = (372); (A3.6)
For a Gaussian chain p 2 controls the mean square end-to-end distance <x2 >0 [Mc
Crum et al., 1989]:
/12\ n2
j° (A3.7)
Hence
kT (12)
W.= v- 7--(A 2 — 1)
2 c/21
(A3.8)
The total work will the summation of all work done in three axes:
W =W.+Wy+W. (A3.9)
or
w =VG
_[12 dt2_3]
2 g (A3.10)
where
(12).
G=NkT Tey' and N=—
v
278
Appendix 4 Other Methods for Ring Cartridges
APPENDIX 4
OTHER METHODS OF OBTAINING

THE STIFFNESS COEFFICIENTS
OF RING CARTRIDGES
In Chapter 5 a relatively simple predictive method was given for a practical elastomer
configuration, such as radially loaded ring cartridges. In this appendix a number of
alternative methods of prediction of radial stiffness are developed. Numerical values for
the particular cartridge configurations were obtained and compared with experimental
results and the method developed in this thesis.
Fig.(A4.1) Cartridge showing Beam-Column element

(after Smalley and Tessarzik (19751)
279
The methods developed in this appendix are "static" prediction methods; that is, they
seek a ratio of element stiffness to storage modulus which is a function of the geometry
only and that ratio applies to the determination of dynamic stiffness and damping. The
approach is not rigorously justifiable, but may be given credibility by empirical verification
[Smalley and Terssarzik, 1975].
Beam-Column Method
This method was developed and used in a report by Smalley and Terssarzik [1975]. The
cartridge is assumed to be made up of a series of non-interacting, elemental
column/beams, whose variation of thickness due to radius effects is small. Fig.(4.1)
illustrates the cartridge and the Beam-Column element. Consider the rigid inner member
to be displaced radially by a small radial distance 6r, relative to the outer 'member. For this
the force caused will be:
F=aA (A4.1)
The force on the element is:
df = a da (A4.2)
Considering the element to act independently of adjacent elements, the elemental

compression and shear forces are
df .=(E
. ,)[(12 (R0 + R)dC)(tdcos(C)
(R—Ri) (A4.3)
df G
.
(R— Ri )
)[(1(Ro+
2
Ri)dC)t]sin(C)
=( (A4.4)
The total force exerted at the mean radius is:
2s (R0 + R4)
F i ortrEcos2(0+ Gsin 2 (C)ic/C
s-0 2(R0 — Ri) (A4.5)
Hence
It (R.+Ri) or t
F= G)
2 (R.— Ri) (A4.6)
280
If it is assumed that E =3G (Possion's ratio v.0.5 implied)
(A4.7)
Then
aF ( Ro i- A)
K = — =271 Gt
a Or (R0—R4)
(A4.8)
Beam-Column Method (the radius effect accounted for
When the radius effect is added, a slightly different result is obtained (see also Method of
Gabel below):
Gt
K = 47c
inE,2
R
(A4.9)
However, the reduction in stiffness due to radius effects was found to be insignificant by
Smalley and Tessarzik [1975], and neglected.
Method of Gtihel
As reported by Smalley and Tessarzik [1975] the method of Gebel [Gabel, 1955; Gabel,
1974] initially follows a very similar development to the Beam-Column method accounting
for radius effects.
Total elongation for a body with a varying cross-section is:
Compression or Tension Or.= € dr
Shear Sri= y dr
dF dF
Considering a = -L and = —2 and using mean length t.
da da
r. dF dr
Or cos(C) =
EtdC r (A4.12)
R. dF dr
Orsin(C) = 1,R,
GtdC r (A4.13)
Hence the total force on the element calculated:
dF = dFc cos (C) + dFisin() (A4.14)
Fig.(A4.2) Cartridge showing GObel's element (after Gael [1974])
Hence
aF
= — (G+E)
R
--E
14 (A4.15)
However, &bel then claims that, t I (R.— R,) = 1 the effective Young's modulus is given
by:
Ecff = 6.5G
(A4.16)
and in addition that, for other length to radius ratios, a correction factor, should be
applied, so that
282
Gt f
K =7.5n
(A4.17)
Gabel [1955] presented fi as a graphical function of t 1 (R. — A). Payne and Scott
[1960] present a cubic fit to Gebel's function:
, 0.0097 t3
=1+
(R.— Ri)
(A4.18)
Equ.(A4.18) agrees closely with Gabel's graphical presentation. The basis for fl or for
the statement E, 6.5G is not made clear in Gabel's book [Gabel, 1974]. Smalley and
Tessarzik [1975] reported that although Gabel's method shows good agreement at
higher values of t —Ri) with the test data, at lower t Ri), the Beam-Column
method appeared to come closer, due to a much sharper drop in stiffness of the test
element. This comparison suggested that E gir = 6.5G is not a valid asymptotic condition
for short cartridges.
283
Appendix 5 Prediction Methods
APPENDIX 5
PREVIOUSLY DERIVED PREDICTION

METHODS FOR ELASTOMERS
For the shear element (Fig.(A5.1), the stressed area, A, is its width times its length (a b)
and the strained dimension is its thickness, h. For this element the stiffness and damping
can be related to effective moduli as follows [Smalley and Tessarzik, 1975]:
A
Stiffness Kl (w) = G' —
h
(A5.1)
A
Damping K 2 (w) = G" —
h (A5.2)
Fig.(A5.1) Shear element (after Smalley and Tessarzik (1975))
Hence
A
K=G —
"" h (A5.3)
284
ff
where Gaff. G eff are shear moduli, commonly referred to as the effective storage and
loss moduli, respectively. G eff and G eff are generally functions of the material, the
frequency, the amplitude of loading, the temperature, the initial strain and geometry.
They are, therefore, component rather than material properties. The relationships
between effective and true moduli can be given as follows, as used many researchers
[Gent and Lindley, 1960; Hattori and Takei, 1950; Smalley and Tessarzik, 1975; Freakley
and Payne, 1978]:
(G:il. ). = + p s2)
(A5.4)
where s is a shape factor defined as the ratio of the loaded area to the unloaded are (D 1(4
h) for cylinder) and b is a coefficient. Subscript, o, implies static loading.
Fig.(A5.2) Compression or tension element

(after Smalley and Tessarzik (1975])
For the compression element, stiffness and damping can be related to the effective
moduli as follows:
A
Stiffness (w) = 3
(A5.5)
A
Damping K2(0.0= 3G;fi
(A5.6)
where the factor 3 arises from the classical elasticity relationship E = 2 (1 + v) G and
the fact that elastomer Poisson's ratios are typically in the range 0.4945 to 0.4999
[Holownia, 19701.
Cylindrical Buttons
For cylindrical buttons the real and imaginary parts of the complex dynamic stiffness can
be given as follows [Smalley and Tessarzik, 1975]:
,sc c12 n d2
Kl ( co)=3G —[1 +13'4+ Gi
4h 4h (A5.7)
„std2r
K2 (co)=3G — 11+ rsl+G"Rd2
4h 4h (A5.8)
where the second terms on the right hand side of the Equs.(A5.7 and A5.8) are added in
order to account for the shear forces. G (co) is the real part and G (co) is the imaginary
part of the complex shear modulus G.
Values for G, G ", G, If and fi" are required to make predictions of dynamic stiffness
and damping of the elastomer buttons. These values are determined from experimental
results, as a function of frequency (and temperature). For example, Smalley and
Tessarzik [1975] empirically determined G , G , fi and 13 for polybutadiene as a
function of frequency, by testing at 32 degrees Celsius as follows:
G'(co)= 3.686 x 106 ce 20" N/m

2
(A5.9)
CVO) = 8.333 x106 co4112" N m2 (A5.10)
/3 1 (co) =12.33 co-4129° (A5.11)
= L726 co"299 (A5.12)
where co is frequency in radian per second.
Ring Cartridges
For ring cartridges there are five prediction formulae derivated.
1. GObel's Formula (see Appendix 4)
le(c0).7.57tGe(w)t f
(A5.13)
Appendix S Prediction Methods
where
0.0097t3
=1+
(R.— Ri)3 (A5.14)
2. Beam-Column Formula (see Appendix 4)
(R0+ .
Ks(co)=27r Re) G (co)t
Ri) (A5.15)
3. Beam-Column Formula (the radius effect accounted for- see Appendix 4)
K(o)=47t
0co)t
111(12;;) (A5.16)
4. Lindley's Formula
Lindley [Lindley, 1966] started his derivation from a simple case that under static
conditions
DF aaA
K= =
as as (A5.17)
which leads to
Ri2)
K1(o)=37c Gviii+k(L---112]
2t (A5.18)
5. Grassano's Formula [Grassano, 1991]
4 V
Ga(COil 1
K:7(co)=—
9 (R.— Ri) (A,;+ 2A,A) (A2,+2A.,A)
(A5.19)
where A, = 1 — and A = R°
R —R. R.2 — Ri2
0-Rings
There are two prediction formulae for 0-rings.
287
1. Lindley's Formula
A similar treatment to the one which was applied to ring cartridges is applied to the 0-rings
to result in Equ.(A5.25) (see Fig.(A5.3) for notation):
(2 d)
Inca) = 40(0) 81-
d – Sr (A5.20)
Lindley has also given an empirical formula for 0-rings [Freckley and Payne, 1978]:
dDE0
.1.25(y +50(-816
(A5.21)
Hence
K s (co) = 47c(R0 + R1 ) G *( 0)) 1.875(A1 ) + 3001—

()5
a
(A5.22)
Fig.(A5.3) Assumed deformed cross-section of 0-ring
2. Grassano's Formula [Grassano, 1991]
4 V ( 1 1+ 6(2, -
Kff(co)=
9 (R0–R1)2 –.‘ wf ( 2 +22..,A) (+2A)4
(A5.23)
Or Or
where A =1 and A=
' Ro – R, R0– R,
288
Appendix 6 Voigt Model
APPENDIX 6
VOIGT MODEL
The Voigt model is a way of representing the material in terms of springs and dash-pots. It
can be traced back to the generalized Hook's law which can be mathematically
represented as [Moore, 1972]
[4+ di(1)+d2VF...+0)] a
= + b +b ;()A- + b (L)] e
(A6.1)
Or
[a° + al(1)+a4Y .+4L)] F
= [bo + b (1) + b2(y

a22)+...+b4L)]
(A6.2)
where ai, bi and b are constants (1 =0, 1, 2, ..., n ). If the force and the
deflection vary sinusoidally,
F b'(co)+ j b"(co)
a'(co)+ j alto) (A6.3)
where a'(co), a"( co), bico) and b"(co) are function of frequency and can be obtained
from Equ.(A6.2) as
a'(co)= a0 — a2 co 2 — a4co4—... (A6.4)
a"(co)= cli c° + a3 co 3 + a5co5... (A6.5)
b'(co) = 1)0 — b2 co 2 — bsco4 (A6.6)
289
Appendix 6 Voigt Model
b"(co) = bi co + kw' + bs cos ... (A6.7)
Equ.(A6.3) can be reduced to
F
= Ki(co)+ j Ki(co) = Ina))
3 (A6.8)
where
a' (co) b'(co)+ a"(co) b"(co)

Ki(c0)=
[a' (co)]2 +fre(co)12 (A6.9)
a'(co) b"(co)— a"(co) b'(co)

K2(0)) =
[a' (co)]2 +[a"(w)]2
(A6.10)
Fig. (A6.1) Modified Voigt Model
For a modified Voigt model as seen in Fig.(A6.1)
F = Ko (3 — 31) = K1 31 + cfr
a St')
(A6.11)
Hence
[1 4- (K
K0 +1 j(g F = [( KIC0 °+KK11 ) + (K 0 °+cK
K i 1)(:)]8
(A6.12)
From the similarity of Equ.(A6.2) and Equ.(A6.12)
Ki(w)= (K+
0 K1 ) K0 K1 + K0 4 o) 2
(K0 + K1)2 + 4 co2
[(K0 + Kj K 0 — K 0 Kil c1 0)
K2(co) =
(K0 + K1)2 + ci2 co 2
290
Appendix 7 Ball Passage Frequency
APPENDIX 7
BALL PASSAGE FREOUENCY
The ball passage frequency can be exhibited from the basic theory of vibrations as done
by Meyer et al. [1980]. General differential equations for a vibrating system in x and y
directions can be written:
M+c A +Kx= Qz(t)

de at (A7.1)
d'y
iv] -57 + + K y = Qy(t)
(A7.2)
Assuming that;
1. balls to race way contacts have linear spring characteristics,
2. the shaft is not flexible,
3. system does not accommodate any damping,
4. the forcing functions are in the form of;
Q(t) = Po cos(n 0.) (A7.3)
Qy(t)= Po sin(n i) (A7.4)

where
ei = am+ y
The differential Equ.(A7.1) and Equ.(A7.2) will give successive solutions to any
xi. = zo cos(n co c t + ni (A7.5)
291
= Yosin(nco c t + ni 7) (A7.6)
Substituting Equ.(A7.5) and Equ.(A7.6) into Equ.(A7.1) and Equ.(A7.2):
F'
X0 = Y = O 2
n2 M co[( ct ' ) -1
nco,
(A7.7)
Substituting Equ.(A7.7) into Equ.(A7.5) and Equ.(A7.7):
F'0
cos(nc.oct + ni
n 2 m coc2[( ffl pg )2 _
n co,
(A7.8))
X.
PO
2 sin(n c t + ni
n 2m coe2[( W it ) _1
n co,
(A7.9)
For a single mode, the deflection in one of the ball to race way contacts can be given as:
Si. = xia cos(nco,t + niy)+ yi. sin(n (a c t + ni (A7.10)
For the total deflection, the effects of other modes should also be considered, hence the
total deflection in the direction of i th ball is given as the summation of all deflections due
to the different modes:
(Si = Eki. cos(nco c t+ ni + ksin(nco c t + ni

(A7.11)
A deflection 8r of the inner ring relative to the outer ring in the radial direction with respect
to the reference axes at an arbitrary angle i3 can be obtained by summation of the ball to
race way contact deflections for all balls:
or = E Pocos(ncoct + ni y P sin(ncoc t + ni
cos(6)+ ° sin(0)
i=1 4=1
n 2m CO 2
co
H2 1 n2 mco 2[( co. )2_1
n co, n co,
(A7.12)

Hence,
m E-
= 1, 2icos(n a )c t + ni y —
i=1 m=1 — a ),)
(A7.13)
Since the balls are assumed to be linear, unless n is equivalent to m or its multiples, all
summations will be zero. Hence,
0
r=li
m 2 cos(rim(oc t + nDri— .6)
i1 i (nm a ) 1
(A7.14)
As a result;
mPo
= 2 cos(nmcoc t— .6)
M[c0,2, — m co c)]
(A7.15)
The term cos (n m tac t— ) in Equ.(4.35) implies that the frequency band of deflection
will have a series of peaks at, and multiples of, ball passage frequency with a trend of
reduction in the amplitudes. If the shaft is assumed to be linear as in our case, only the
first mode will be effective. Therefore, there will be only one peak in the frequency
spectrum at the ball passage frequency as long as the system is not disturbed by any
other external excitations. Hence,
mP0
= COS (M (O c t — 6)
M[a): — (m co c)1
(A7.16)
is infinity when m coc = con
293
Appendix 8 Fast Fourier Transform
APPENDIX 8
FAST FOURIER TRANSFORM
A signal of a vibrating system can be described either in the time domain or in the
frequency domain. The amplitude of the signal, h, will be a function of time or frequency
depending on the domain. The Fourier Transform can easily relate these two domains.
Since the equations in this thesis are nonlinear, a time domain solution has to be
performed. Therefore it is quicker and more accurate to estimate the spectra directly from
the original time series. However, it is sometimes easier and more convenient to
investigate the results in the frequency domain.
It should be remembered that the Fourier series is an approximate one. Nevertheless it

can reproduce the original discrete-data-point time series exactly [Mitchell, 1985].
Fig.(A8.1) shows the relationship between time and frequency domains. The signal in
the time domain is the sum of many ideal sinusoidal waves at different frequencies and
amplitudes. In other words the amplitudes in the frequency domain coincide with the
peak to peak amplitudes of sinusoidal waves at that frequency. Hence the signal in the
time domain is the sum of the signals in the frequency domain:
1 +-
h(t) = H(co)e-l'cico
2,r
(A8.1)
or in other words
H(w) = h(t)ele"dt
(A8.2)
H is generally a complex number. The Fast Fourier Transform (FFT for short) is a
remarkably efficient way of calculating the Fourier transform of a time series [Newland,
1975]. The FFT algorithm saves a considerable time and increases the accuracy by
reducing the errors caused due to truncation.
294

Appendix 8 Fast Fourier Transform
Frequency
Fig.(A8.1) The relation between the frequency and time domains
In this thesis the displacement histories of the shaft centre are obtained by the solution of
the differential equations in the time domain. The sampling interval, At (constant) and the
discrete values of x(t), at time t=k At , Xr are used and the sequence (X }, r=...,-1, 0, 1, r
2,..., N-1, is obtained and called a "discrete time series". From this sequence a new finite
k
sequence (X ) is defined as
irrArk r)
1 N-1
Xk = — EX,.e
N r=0 (A8.3)
which is obtained by using
1[ 1 pv.,T1x __i
2k k r)
w
1 N/2-1 (2 Nir ;211
XL = —
2 N I 2 .() 1
2r.
„ ( N/2 )
k N/2
r.9:2 ,+1 e
E (A8.4)
1 N/2-1 k ( 1 N 12-1 ir 2z kr)

1 (W))
N/2 )
X k+N 12 = — —
2 NI2 ,..c, 2 re EX i —W —
N/2 E
r=o
X2r+1 e
(A8 .5)
in half of the computing time of that of a normal discrete Fourier transform.
In these equations
k . 0, 1, 2,—, (N/2-1) and W =
More detailed information about the FFT and derivation of these formulas may be found in
Ref. [Newland, 19751
295
Appendix 9 Runge-Kutta Method
APPENDIX 9
RUNGE-KUTTA METHOD
The Runge-Kutta Method is a numerical solution technique that is employed in the

solution of equations of motion. It is widely used in computational solutions to differential
equations since the truncation error of the method is of the order five of the interval h
(therefore it is called 4 th order) it is a relatively accurate method.
In this method, the differential equation has its solution extended forward from known
conditions by an increment of the independent variable without using information outside
of this increment. The solution is obtained from the weighted average of a number of
estimates of the change in the variable. Hence the next iteration will be:
= + + K2 + K3 + K4) (A9.1)
where
K1 = h f (xoti)
(A9.2)
K2 = h f (Xi + h)
(A9.3)
1
K3 = hf(xj+- K2,tii-lh)
2 2 (A9.4)
K4 = h f (xi + K3,ti.1) (A9.5)
where h is the time increment.
296
Appendix 10 Natural Frequency Variations
APPENDIX 10
THE EFFECT OF PRELOAD AND

THE NUMBER OF BALLS ON
THE NATURAL FREQUENCY
=&coi-fl)
al*
StegV:)r =sri; cos ai
Fig.(A10.1) The force on the i th ball due to the shaft centre deflection
Let us assume that the bearing in Fig.(A10.1) is an angular contact ball bearing. The
maximum deflection will caused at the angle p due to the displacement of the inner ring
centre in the x, y and z directions. This can be given as:
8 =.42"2+z2
(A10.1)
and the angle p can be calculated from (see Fig.(A10.1)):
297
tanfi =
(A10.2)
and the radial deflection on the I th ball can be calculated from (see Fig.(A10.1)):
Si= 6 (cos (9i — p)cos 2 a i + sin' cci) (A10.3)
If the total forces in the x, y and z directions are calculated:
3m
7.K Ai cos a i cos ei
Wx = 3 -22
i=1 (A10.4)
3m
Wy = (PIK Ai cos a i sin 0
i=1 (A10.5)
3m
= (PIK Ai sin a
i=1 (A10.6)
where m is the number of balls and Ai = cos (0 i — fl)cos2 ai + sin2ai
From these forces the stiffness coefficients in the x, y and z directions can be derived.
For brevity here only the x direction will be derived. The force in the x direction can be
written as:
(A10.7)
where
IR
S= K (cos (ei -13)cos2 a i + sin2 cc i )cos a ; cos 0;

1=1 (A10.8)
As Equ.(A10.7) suggest for a given time t =t1 ,S is constant. Therefore the derivation of
Equ.(A10.7) with respect to the deflection will give an instantaneous linear stiffness
around a value of 60 as
aw= 3 is(616.
1
(K.).=
as (A10.9)
The approximate natural frequency of the system in the x direction, around the deflection
60 will be (see Equ.(4.2)):
298
i
(3Sy
_ 3)4( 1
(co.),.,.= m % 18= 80
(A10.10)
If the deflection is considered in terms of a constant deflection 3 0 and a small variation 48
around As„ Equ.(A10.10) will take the form of:
1
( co \ 5.5 =(3.5)i (501(1+ 1 AS)
1 4 , 0,1) 4 15 0 (A10.11)
3
PRa = m K (3 0 )i sin a,
(3.16)
Considering Equ.(3.16), Equ.(A10.3) will become:
_ I \—
1
i ita )6
+ 1 t18
(6).) =(—m 3S )i ( P
mKsinap 4
r
5=do
( PR.
m Ksina P
\ (A10.12)
299

Nizami Akturk 1993 PHD Thesis

Uploaded by

Copyright:

Available Formats

Nizami Akturk 1993 PHD Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nizami Akturk 1993 PHD Thesis

Uploaded by

Copyright:

Available Formats

Imperial College of Science, Technology and Medicine

DEPARTMENT OF MECHANICAL ENGINEERING

DYNAMICS OF A RIGID SHAFT

A Thesis submitted for the Degree of

In this thesis, the dynamics of a rigid shaft supported by angular contact

The method used to investigate the vibration characteristics of angular

It is shown that the vibration characteristics of the shaft-bearing

Finally, the bearing outer races were supported on elastomer buttons,

(Niany students for over a

First and foremost I am thankful to The Creator of the

I would like to express my deepest gratitude to my supervisor

Thanks are also due to my colleagues and friends Dr. M.

I wish to thank Dr. R. Dwyer-Joyce for proof reading some

The help provided by the Mechanical Engineering Library staff

1 am profoundly grateful to my family members and relatives, in

Finally, I would like to express my sincere thanks to the

CHAPTER 1General Introduction

CHAPTER 2 Previous Work

CHAPTER 3 Mechanics of Ball Bearings

CHAPTER 4 Vibration Frequencies of a Shaft-Bearing System

4.3.3.1 Deflection due to Misalignment 131

CHAPTER 5 Elastomers as External Dampers

CHAPTER 6 Results and Discussion: Ball Bearings

6.4.1 Effect of Varying the Number of Balls 186

CHAPTER 7 Results and Discussion: Elastomers as External Dampers

CHAPTER 8 Overall Conclusions and Suggestion for Future Work

8.2.6 Off-sized Balls 253

APPENDIX 1 Equivalent Voigt Model 272

APPENDIX 2 Phase-Plane Representation 273

APPENDIX 3 Large Deformations of Elastomers 277

APPENDIX 4 Other Methods of Obtaining the Stiffness Coefficients of 279

APPENDIX 5 Previously Derived Prediction Methods for Elastomers 284

APPENDIX 6 Voigt Model 289

APPENDIX 7 Ball Passage Frequency 291

APPENDIX 8 Fast Fourier Transform 294

APPENDIX 9 Runge-Kutta Method 296

APPENDIX 10 The Effect of Preload and Number of Balls on 297

symbol Description Units

a the distance between the misalignment centre and bearing centre m

E. static Young's modulus (Chapter 5) N/m2

k material constant, proportional to hardness

K* complex stiffness N/m

K0 outer raceway/ball contact stiffness factor N/m3r2

/ length of the specimen (Chapter 5) m

no outer race speed Ep.m.

N number of data points

Po the amplitude of forcing function N

Oc deflection due to misalignment m

3i total instantaneous contact deflection for the i th baU m

0 angular position of balls rad.

Yo the amplitude of vibration in y direction m

n the amplitude of the waves at the contact angle m

FO initial wave amplitude m

j relating to the Voigt element being calculated

relating to the first derivative with respect to time

A general introduction to the shaft-angular contact ball bearings assembly as a part of

1.2 Angular Contact Ball Bearings as a Source of Vibrations

In modern machine applications, ball bearings have become indispensable since,

Fig (1.1) Assembled Angular Contact Ball Bearing