1991 Book Tribology PDF

Tribology : Principles and
Design Applicatio ns
Tribology
Principles and Design Applications
R. D. Arnell
P. B. Davies
J. Halling
T. L. Whomes
M
MACMILLAN
© R. D. Arnell, P. B. Davies, J. Halling and T. L. Whomes 1991
All rights reserved. No reproduction, copy or transmission

of this publication may be made without written permission.
No paragraph of this publication may be reproduced, copied

or transmitted save with written permission or in accordance
with the provisions of the Copyright, Designs and Patents Act
1988 or under the terms of any licence permitting limited
copying issued by the Copyright Licensing Agency,
33 - 4 Alfred Place, London WC1E 7DP.
Any person who does any unauthorised act in relation to

this publication may be liable to criminal prosecution and
civil claims for damages.
First edition 1991
Published by
MACMILLAN EDUCATION LTD
Houndmills, Basingstoke, Hampshire RG21 2XS
and London
Companies and representatives
throughout the world
British Library Cataloguing in Publication Data

Tribology.
1. Tribology
I. Arne!, Derek
621.89
ISBN 978-0-333-45868-6 ISBN 978-1-349-21387-0 (eBook)
DOI 10.1007/978-1-349-21387-0
Contents
Preface vi
Notation viii
1. Surface Properties and Surface Contact 1
2. Friction 27
3. Wear 66
4. Tribological Properties of Solid Materials 96
5. Fluid-lubricated Thrust Bearings 124
6. Fluid-lubricated Journal Bearings 161
7. The Lubrication of Highly Loaded Contacts 194
8. Bearing Selection 216
9. Lubricating Systems 230
References 245
Index 249
v
Preface
In 1975 we and four of our former colleagues wrote the textbook Principles
of Tribology. That book was very successful both in the United Kingdom
and overseas and it is still being sold today. However, 13 years after the first
publication of the book, the publishers suggested that we should consider
revising the text, to include the more important developments which had
taken place in the meantime.
Although we have remained active in tribological research, it was only
when we started work on this revision that we realised that the developments
which have taken place have been so extensive that they could not be
accommodated simply by revision of our existing text. As a result, although
we have reused some of our original material, the present book has resulted
from a complete reappraisal of our ideas on the necessary content of a modern
introductory textbook on tribology. As the title of the book implies, it has
been our intention to write a book which will provide a thorough grounding
in the principles of our subject for undergraduates and new research workers
and will also act as a basic reference book for practising designers.
In the areas of friction and wear, although there has been a great deal of
high-quality research on the underlying physical and mechanical processes,
there are no current, or indeed foreseeable, models which will allow
quantitative predictions of friction coefficients or wear rates from basic
physical, chemical and mechanical properties. Such predictions will continue
to be based on experimental measurements. Therefore, in these areas, we
have attempted to set out as clearly as possible the underlying processes and
thus to assist the reader in making sensible initial choices of materials.
In the areas of fluid film lubrication, as well as covering the fundamental
mechanisms, we present simple design procedures for hydrodynamic bearings
and incorporate the more important aspects of the growing amount of
information on elastohydrodynamic lubrication, with examples of its use in
design.
vi
PREFACE vii
We have also included in the book introductions to more practical

considerations such as lubricant selection, lubricant-supply systems, condition
monitoring and failure analysis.
We should like to record our thanks to Mrs Marise Ellidge for her help
in preparing the typescript and to Miss Jenny Shaw, formerly Engineering
Editor of Macmillan Education, for her great tact and patience.
May, 1990 R. D. A.
P. B. D.
J. H.
T.L.W.
Notation
Symbol Definition
A Real area of contact
AF Area of filter
a Semiaxis of contact ellipse in transverse direction, or radius of
contact circle (Chapters 1, 7)
Width of axial supply groove (Chapter 6)
Width of bearing
Length of journal bearing (Chapter 6)
Semiaxis of contact ellipse in direction of motion (Chapter 7)
Dynamic load capacity of rolling bearing (Chapter 8)
Static capacity of rolling bearing
k,/k
Specific heat
Diametral clearance
Radial clearance
Diameter of gear wheel
Diameter of journal bearing
Mean diameter of rolling bearing
E' .
Elastic modulus
Elastic modulus of contact= 2 I( -
El
vi + -
1- - v~ )
1- -
E2
1-v 2 1-v2)
E* Elastic contact modulus = 1 /( - -1 + - -2
Et E2
Shaft eccentricity
Shear or friction force
Axial force
Rolling friction
viii
NOTATION ix
Frad Radial force

Fs Sliding friction
Fx, Fy, Fz Force in coordinate directions
F F/(tfU 0 BL/hd
f Fraction of scratched volume to become wear debris
G Materials parameter ( = E'a.)
g Acceleration due to gravity
=ar
-ars/3
0112 =(Jl
ar rectangular elliptical
=on contact =(~Yn contact
-arJ/2 G-ar 3
= 0112 G = ----rJ2
H Hardness (Chapters 2, 3)
H Power to rotate shaft (Chapter 6)
H Dimensionless power coefficient
n Dimensionless film thickness ( = hI R)
h Dimensionless surface separation (Chapter 1)
h Film thickness (Chapters 6, 7, 8)
he Central film thickness
hm Film thickness at maximum pressure
hmax Maximum film thickness in a journal bearing
hmin Minimum film thickness in a journal bearing
h0 Minimum film thickness
hP Film thickness at pitch line
h. Starved film thickness
h1 Outlet- film thickness of slider
h2 Inlet film thickness of slider
h 00 Film thickness with fully flooded inlet
I Tou
Inlet shear heating parameter = ( 11
2
)
J Inlet factor ( = bxJ(2Rh 0 ) 2 ' 3 )

j Coordinate in X direction
K Wear coefficient (Chapter 3)
K Bearing inclination (Chapter 5)
K Thermal capacity constant (Chapter 6)
K Thermal conductivity (Chapter 7)
KF Filter characteristic
K0 Grease lubrication constant
k Yield shear stress (Chapters 1, 2, 3)
k Specific wear rate (Chapter 4)
k Proportion of heat convected by oil (Chapter 6)
x NOTATION
k. Ellipticity parameter ( = a/b)

k0 Constant specific wear rate
kr Contaminant shear stress
L Length of sliding bearing (Chapter 5)
L Life of bearing in revolutions (Chapters 3, 8)
1 Length of axial supply groove (Chapter 6)
1 Centre distance of gears (Chapter 7)
M Mass supported by journal bearing (Chapter 6)
M Moment of frictional force (Chapter 8)
m Dimensionless distance to inlet boundary ( = xI b)
mi Critical inlet boundary distance
N Speed (rev/s)
NF Number of cycles to failure
N1 Number of ordinates of spacing 1 in sample length
n Exponent in stress-strain equation (Chapter 1)
n Number of contacts (Chapter 3)
n Speed (rev/min) (Chapter 9)
P Load per unit length (Chapter 1)
P Power, rate of heat generation (Chapters 5, 8)
P Dynamic equivalent load on rolling bearing (Chapter 8)
P0 Static load on rolling bearing
p Pressure
Po Yield pressure
Pr Pressure of fluid in supply groove
Ph Maximum Hertzian pressure
Pmean Mean contact pressure
Pmax Maximum pressure
p Normalised pressure ( = p/ H)
Q Tangential force in rolling contact (Chapter 2)
Q Wear volume (Chapter 3)
Q Volume rate of flow (Chapters 5, 6, 9)
Q Q/(Bh 1 U 0 )
Q Dimensionless flow rate
QP Pressure-induced flow
Qv Velocity-induced flow
Qd Depth wear rate
q Tangential traction in rolling contact
R Equivalent radius of curvature (Chapter 7)
R Relubrication factor (Chapter 9)
RA Mean roughness of a surface
R. Reynolds number
R(1) Autocorrelation function
Rx Radius of curvature in direction of rolling
Ry Radius of curvature in transverse direction
NOTATION xi
R Ratio between plastic and elastic area

R1 Radius of journal bearing
R2 Radius of shaft
r(l) Normalised autocorrelation function
s Dimensionless asperity height (Chapter 1)
s Friction force/unit area (Chapter 2)
s Slid distance (Chapter 3)
T Temperature
T. Effective temperature of oil
Ti Temperature of oil at feed groove
T1 First transition load
T2 Second transition load
T Dimensionless temperature coefficient
t Thickness of surface film
U Rolling speed ( = ( U 1 + U 2 )/2)
0 Dimensionless speed parameter ( = 17 U j E 'R)
U0 Velocity of moving surface or shaft surface
Uc Velocity of surface
u Speed
V Scratch volume
VP Velocity at pitch line
V. Entraining velocity at pitch line
V. Viscosity ratio
V. Viscosity ratio at operating temperature
W Load
W Dimensionless load parameter
W WJ(rJU 0 B(L/h) 2 ) (Chapter 5)
W Dimensionless load coefficient
W0 Load/unit length
x Coordinate in direction of rolling
X; Distance from inlet to edge of Hertzian contact
xm Coordinate for maximum pressure
Y Load factor
y Transverse coordinate
z Coordinate normal to plane of contact
r.x Parameter defining accuracy of Hertz equations (Chapter 1)
r.x Constant (Chapter 2)
r.x Pressure viscosity coefficient (Chapter 7)
p Radius of asperity tip (Chapter 1)
p Mechanical constant defining load capacity (Chapter 3)
y Surface energy (Chapter 2)
y Ry/ Rx (Chapter 7)
AT Temperature rise
{J Normal approach of surfaces (Chapter 1)
xii NOTATION
Normal approach at limit of elastic deformation

Temperature-viscosity coefficient
Hysteresis loss coefficient (Chapter 2)
Eccentricity ratio
'1 Number of asperities per unit area (Chapter 1)
'1 Dynamic viscosity (Chapters 5, 6, 7, 8, 9)
1'/e Effective dynamic viscosity
1'/r Dynamic viscosity of oil in supply groove
1'/o Viscosity at ambient temperature and pressure
'ls Viscosity at surface temperature, ambient pressure
'lx Viscosity at surface temperature, actual pressure
() Semiangle of conical asperity (Chapter 2)
() Polar coordinate (Chapter 6)
A Parameter defining amount of asperity contact (Chapter 2)
A. Parameter governing normal approach (Chapter 1)
A. Coefficient of rolling resistance (Chapter 2)
A. Specific film thickness ratio ( = h/ R) (Chapters 7, 8, 9)
Jl Coefficient of friction
v Poisson's ratio (Chapters 1, 7)
v Kinematic viscosity (Chapters 6, 7, 8, 9)
e apcV /2rx.
p Density
(J Standard deviation
Composite roughness
Uniaxial yield stress
Shear stress
Interfacial shear strength
Plasticity index ( =! (~Y 12

) (Chapter 2)
Attitude angle (Chapter 6)
(1 +2y)- 1 (Chapter 7)
Normal pressure angle of gear teeth
Surface height distribution function
Subscripts
1, 2 Refer to the two surfaces of the contact
Chapter 1
Surface Properties and
Surface Contact
1.1 INTRODUCTION
Tribology is defined as 'the science and technology of interacting surfaces in

relative motion and ofrelated subjects and practices'. Therefore, if we are to
understand tribological processes, we must understand how surfaces interact
when they are loaded together, and it is the primary purpose of this chapter
to develop such an understanding.
However, before we can consider surface interactions, we need to have a
clear idea of what we mean by 'surface', and the ineaning in the context of
tribology is by no means as simple as we might expect. Traditionally, we
might think of the surface of a solid body as simply the geometrical boundary
between the solid and its environment: i.e. a surface in space which has
molecules of the solid to one side of it and molecules of the environment to
the other. However, such a definition is far too limited for our purposes.
Tribological behaviour is influenced by the physical, chemical and mechanical
properties of not only the surface material, but also the near-surface material;
and when we talk to surfaces, we shall implicitly include the material to a
significant depth below the .actual boundary.
It will become clear in the remainder of this chapter and in the following
chapters that the interactions between two surfaces are very dependent not
only on the materials of the two contacting surfaces, but also on their shapes,
and we therefore need an appropriate method of describing and measuring
surface shape.
Thus, we start this chapter by describing the nature of the surface and
near-surface regions of a solid; we shall then show how we describe and
measure the shape of a surface; and, finally, we shall briefly describe those
aspects of the mechanical interaction of surfaces which underlie our current
understanding of tribological behaviour.
2 TRIBOLOGY
Primarily because of their high strength and toughness, metals have been
the most common tribological materials for many years. For this reason,
their tribological behaviour has been studied in greater detail, and is
consequently more fully understood, than that of other materials. Therefore,
we shall start by concentrating on the nature of the metal surface, but in
later chapters we shall extend all our discussions to include more recent
tribological materials such as ceramics, elastomers and polymers.
1.2 NATURE OF METALLIC SURFACES
A surface of a typical metal might appear as shown in Figure 1.1. On top of

the normal crystalline structure of the bulk material lies a layer of deformed
material resulting from the processes used in the manufacture of the surface.
This deformed layer is itself covered by a compound layer resulting from
chemical reaction of the metal with its environment. (With metals in air, this
reacted layer will usually be an oxide, and, as we shall see, such oxide layers
can critically affect both the friction and the wear of metals.) This reacted
layer will, in a normal environment, often be covered with a layer of
contaminants such as condensed oil vapours and particles of smoke and dust.
Finally, machining processes may also cause contaminants such as cutting
lubricants to be trapped in heavily deformed regions of the surface.
It should now be obvious that a metal 'surface' in a normal environment
is a very complex structure containing varied amounts of different materials
with very different properties, and we are unlikely to be able to make accurate
predictions of the ways in which two such surfaces will interact when loaded
together. It may, therefore, be thought that we should take as our baseline
two completely clean surfaces- i.e. two surfaces from which all contaminants
and reacted chemical films have been removed. Such surfaces can be generated
in ultra-high-vacuum conditions, either in highly specialised laboratory
apparatus or in outer space, and, as we shall see in the next chapter, the
study of such surfaces has helped to elucidate some tribological processes.
Figure 1.1 Typical surface layers

SURFACE PROPERTIES AND SURFACE CONTACT 3
However, apart from this particular case, other surfaces which we describe
as clean will still retain the oxide films which are of paramount importance
in determining tribological behaviour.
There are now many techniques available for studying the chemical
nature of the surface and near-surface layers of materials. However, for the
purposes of this text it is only necessary to make use of the relevant results
of such studies: it is unnecessary to describe the actual techniques, and readers
interested in these should consult specialised texts such as Walls ( 1989) and
Woodruff and Delchar (1988).
1.3 SURFACE GEOMETRY
The geometric shapes of ordinary surfaces are controlled by the characteristics

of the finishing processes by which they are produced. Close analysis of these
surfaces shows that, even after the most careful polishing, they are still rough
on a microscopic scale. Statistical analysis of the undulations of a surface
normally shows that a very wide range of wavelengths is present, ranging
from a fraction of a micrometer to many millimetres. Indeed, it is likely that
the range of wavelengths detected is limited only by the resolution of the
measuring instrument and that the wavelength spectrum actually extends
from atomic dimensions to the full length of the surface being measured.
It has been common practice to describe different wavelength bands in
different terms, referring to wavelengths of the order of micrometres as
roughness and the longer wavelengths as waviness. However, such a
sub-division is quite arbitrary, and the only practical necessity is to be aware
of which undulations are functionally important in any particular application.
In tribology the individual points of contact between two solid surfaces have
dimensions of the order of micrometres, and we are, therefore, concerned
with the amplitudes of wavelengths of the same order - i.e. with those
wavelengths which have been conventionally described as roughness. The
peaks of such surface undulations are known as asperities, and friction and
wear both arise from the contacts between such asperities on opposing
surfaces.
It is reasonable to assume that both friction and wear will depend in some
way on the stresses arising at such asperity contacts and that the stresses
themselves will depend on the heights of the asperities above the general
level of the surface; in a later section we shall show more rigorously that this
is indeed the case. We can, therefore, see that the tribological behaviour of
two extended surfaces in contact will depend on the statistical distributions
of asperity heights relative to the general level of the surfaces, and if we are
to understand such behaviour, we must be able to measure and describe
these distributions. The ways in which we do this form the subject of the
following section.
4 TRIBOLOGY
1.4 SURFACE TEXTURE ANALYSIS
A typical surface might have ~ 10 5 peaks. The problems of measuring the

height and location of each of these peaks on each of two surfaces would be
formidable. The problem of then analysing the interactions between the two
sets of peaks when the surfaces are loaded together under normal and
tangential loads would be overwhelming.
Any realistic approach to this problem must, therefore, use measurements
from a small but representative sample of the surface - i.e. a sample chosen
to be of such a size that there is a high probability that the surface lying
outside the sample is statistically similar to that lying within it. In the
remainder of this section we describe, first, the instrument used to obtain
such measurements and, second, the ways in which the measurements are
interpreted to give us descriptive parameters for the surfaces.
1.4.1 Surface Profilometry
The requirement for measuring a statistically representative sample of the

surface is partly met by using a profilometer to sample a representative length
of the surface with a high resolution in the plane normal to it. In the
profilometer a very fine diamond stylus, with a tip radius of 2 Jlm or less, is
drawn over the surface, as shown in Figure 1.2(a). The vertical movement
of the stylus as it traverses the surface is then amplified and recorded to give
Stylus
(a)
Profile trace
(b)
Figure 1.2 (a) Schematic view of profilometer; (b) typical surface profile
a surface profile, as shown in Figure 1.2(b ). In examining such profiles, some

limitations of the method must be borne in mind, and these are discussed
below.
The horizontal magnification is simply the ratio between the speed of the
recording paper and the traverse speed of the stylus, and is typically x 100.
The vertical magnification is controlled electronically, and can normally be
varied from x 500 to x 100000, according to the resolution required. This
difference between the two magnifications is useful in giving greater emphasis
to the important height characteristics of the surface, but it does mean that
the resulting record is distorted. In Figure 1.3 a typical record is shown
together with a record from the same surface using equal horizontal and
vertical magnifications. The latter record shows that the asperities are
undulations with slopes of a few degrees rather than the sharp peaks shown
in the distorted profile.
Figure 1.3 also shows the size of the stylus to the same scale as the record,
and it can be seen that some recording errors must occur due to the finite
size of the stylus: the valleys will be recorded as being slightly narrower than
their actual size, while the peaks wll be recorded as slightly broader.
The main disadvantage of profilometers of the type described above is
that they are restricted to a single-line sample which may be unrepresentative
of the whole surface if the texture has characteristics which are dependent on
the orientation of the record. To overcome this disadvantage, there are now
instruments available which can produce a three-dimensional map of the
surface by recording a series of parallel and closely spaced profiles. However,
for any surface which is approximately isotropic, the simpler instrument can
provide measurements of the various parameters which are necessary for the
characterisation of the surface.
It is also necessary to appreciate the difference between a peak on a
profilometer trace and a true summit on the surface. The profilometer will
register a peak each time it traverses a shoulder of a summit, and the number
of peaks recorded will, therefore, be much greater than the number of true
summits. This is clearly illustrated by the three-dimensional records of the
type described above. This limitation is mitigated by the fact that any asperity
on an opposing sliding surface will follow a similar path to that of the
traversing stylus and the undulations over which it will have to pass are
quite well represented by the profilometer trace.
Stylus radi~ I
x200 ~ X 200
L x200 ' , , __ .,.."' 1 Lx20
Figure 1.3 Illustration of the profile distortion due to unequal vertical and
horizontal magnifications
6 TRIBOLOGY
Finally, in profilometry the profile height must be measured with respect

to some datum. The two most common methods of representing such data
are the use of datum-generating attachments, which ensure accurate horizontal
movement of the stylus support system, and the use of large-radius skids or
flat shoes which rest on and traverse the surface being measured, and thus
generate the general level of the surface. Each of these techniques creates
some experimental errors, although the former is rather more precise.
In spite of the limitations discussed above, the profilometer remains the
outstanding instrument for studying surface geometry and for evaluating its
descriptive parameters. These parameters and the ways in which they are
evaluated are described in the following sections.
1.4.2 Traditional Roughness Parameters
Until recent years, the analogue output of the profilometer was usually
analysed within the instrument to give a graphical record of the profile and
a single-parameter description of the surface roughness, which was displayed
on an analogue meter. The two parameters in most common use were: ( 1) the
roughness average (RA) value, defined as the mean vertical deviation of the
profile from the centre line, treating deviations both above and below the
centre line as positive; (2) the root mean square (RMS) value, defined as the
square root of the mean of the square of these deviations. The RA has
traditionally been the most common parameter in use in the UK, while the
RMS value has been the most common in the USA.
These parameters are seen to be concerned only with relative departures
from the centre line in the vertical direction; they do not provide any
information about the shapes, slopes and sizes of the asperities or about the
frequencies of their occurrence. It is, therefore, possible for surfaces of widely
different profiles to give the same RA and RMS values, as shown in Figure 1.4.
These single-parameter descriptions are, therefore, mainly useful for
comparing surfaces which have been produced to different standards but by
similar methods: for example, lapped surfaces, differing only in the grade of
lapping compound with which they have been finished.
1.4.3 Statistical Analysis of Surface Profiles
Surface profiles often reveal both periodic and random components in their
geometric variation, and such components are not revealed by the RA and
RMS values. In recent years it has become common to digitise the surface
profile, to give a record which is more amenable to statistical analysis and
to separation of its periodic and random elements. In this technique the
analogue output from the amplifier is digitised at discrete intervals, as shown
in Figure 1.5, to give a sequence of height readings relative to the profile
centre line. A typical record would consist of 4000 digital height readings
Figure 1.4 Various surfaces with the same RA value
taken at intervals of 2 Jim along the surface. The sequence can then be
analysed by an inbuilt microcomputer, to extract any chosen statistical
information. The two most common types of information to be extracted are
the height distribution of the surface texture and the amplitudes and
wavelengths of periodic variations along the surface; we shall consider these
in turn.
The Height Distribution of the Surface Texture
As shown in Figure 1.5, the profile is sampled as height readings z 1 , z 2 , etc.,

at some discrete interval l. The resolution in the vertical direction is not
infinite, and each height reading, z, will actually fall into one of perhaps 200
bands. The whole profile is then scanned by the computer, to determine the
number, N ., of ordinates detected in the band of mean height z, and a histogram
of these occurrences can then be plotted, as shown in Figure 1.6. The profile
can then be described by its all-ordinate distribution curve, ¢(z), which is the
best-fit smooth curve drawn through the histogram, and this distribution, in
turn, can be converted into a probability distribution by normalising the
readings so that the total area enclosed by the distribution curve is unity.
Figure 1.5 Surface height readings taken at discrete intervals

8 TRIBOLOGY
distribution curve
Figure 1.6 Method of deriving the all-ordinate distribution function
Many statistical parameters are available to describe the shape of this

probability curve, and readers requiring further information on these can
consult Halling ( 1975). However, for the purposes of this text, we shall use
only the standard deviation of the distribution, which is identical with the
RMS value defined earlier. In terms of the digitised height readings, this is
defined by
u = RMS = [ -1 n (zY L Jt ( 1.1 )

n i=l
Note that we can also calculate the RA value from the statistical data, as it
is simply given by
1 n
RA =- L lz;l ( 1.2)
n i=l
Other parameters have been used in other countries, and in recent years
some of these, together with the RMS and RA values, have been precisely
defined by the British Standards Institution in BS 1134. However, although
several of these parameters are still widely used to specify required standards
of surface finish on manufactured items, they are not as amenable to
mathematical manipulation as is the u value, and in the following chapters
we shall use this single parameter to characterise height variations.
Periodic Variations along the Surface
Periodic height variations along the surface can be displayed by plotting the
variation of the autocorrelation function R{l) of the statistical data against
the sampling interval l. R(l) is obtained by delaying the profile relative to
itself by some fixed interval I, multiplying the delayed profile by the original
one, and averaging the resulting product value over a representative length
of the profile. Thus,
1 N-1
R(l)=- L z(x)·z(x+l) ( 1.3)
N-Ix= 1
where l is the distance between the original and the delayed profiles and
N is the total number of ordinates in the total sample length L.
~:Jww, F-l.~~
~------- ~---------
Surface A Surface B
"'~"'L,
Figure 1. 7 Typical surfaces and the resulting autocorrelation functions
For a continuous profile, Equation ( 1.3) can be rewritten as

1 fL/2
R(l) = lim - z(x) · z(x + /) dx (1.4)
L-+00 L -L/2
It can then be seen that when l = 0, R(l) reduces to the square of the u value
of the profile. The autocorrelation function is therefore usually plotted in its
standardised form r( /), where
r(l) = R~) (1.5)

CT
and the r(l) value always has its maximum value of unity when l = 0.
Typical plots of autocorrelation functions for two different surfaces are
shown in Figure 1.7. The shapes of these plots reveal both the random and
the periodic characteristics of the profiles. The general decay indicates a
decrease in correlation as l increases, and is an indication of the random
component, while the oscillatory component of the function reflects a similar
periodicity in the profile. Figure 1.8 shows results from some machined
surfaces, where the plots can be seen to demonstrate the general features of
the profiles.
It can thus be seen that it is possible to describe the general features of
any surface profile by two characteristics: the height distribution function
cf>(z) and the standardised autocorrelation function r(l).
When we come to discuss the nature of contact between two surfaces,
we shall also see that the heights, curvatures and frequencies of distribution
of the surface asperities are of critical importance. All these data can be
extracted from the digitised profile.
In practice, it is often found that both the all-ordinate distribution and
the peak height distribution for engineering surfaces are good approximations
to Gaussian distribution curves - i.e. curves of the form
(1.6)
10 TRIBOLOGY
Shaped surface,
RA 1.6 ~m
¢ (z) A r(/) ~
~_l JJJ'J/JJV I
" ... ''"l '·

'¢:fl__~
L A .AIL.~f\JV'. Fine-turned surface,
'···~
I
5 L
100
¢~
r4.....,_ ... ,A.o_ •
r(I)L
_..~...... AJ
••-"V'Yfv-v.,.........-.- .,
.-.A Ground surface,
RA 1.0 ~m
0.2 L ~JIN.. n.JI,... ~ 1\ r'""'\t Superfinished surface,

20 . ,. • v r VVU V . RA 0.18 ~m
¢(z) ( \ r(l)~
ll_~ I
Figure 1.8 Examples of engineering surfaces, their distributions and their auto-
correlation functions
This is by no means universally true, as full distributions often depart

markedly from the Gaussian shape. However, surface contact normally
involves no more than the outer 10% of the asperities, and even when the
full distribution departs radically from Gaussian, this outer tail of the
distribution is often a good approximation to the tail of a Gaussian
distribution. This is a very important point, as we shall show later that it is
this Gaussian or near-Gaussian distribution of asperities in the regions of
contact which leads to the observed frictional behaviour.
1.5 SURFACE CONTACT
Both friction and wear are due to the forces which arise from the contact of
solid bodies in relative motion. Therefore, any understanding of friction or
wear must be based on an understanding of the mechanics of contact of solid
bodies.
Solid bodies subjected to an increasing load deform elastically until the
stress reaches a limiting value, known as the yield stress ay, and at higher
stresses they then deform plastically. In most contact situations some
asperities are deformed elastically, while others are deformed plastically: the
loads will induce a generally elastic deformation of the solid bodies but at
the tips of the asperities, where the actual contact occurs, local plastic
deformation may take place.
In this section we shall start by considering the elastic deformation of
cylinders and spheres. This study is valuable for two reasons: first, many
engineering situations involve the non-conforming contact of bodies defined
by smooth curves; second, all solid bodies have surface asperities which can
be considered as small spherically shaped protuberances, so that the contact
of two macroscopically flat bodies can be reduced to the study of an array
of spherical contacts deforming at their tips.
It is clear that a detailed study of surface contacts would involve a detailed
understanding of the elastic and plastic deformation of contacting surfaces.
However, it is the purpose of this chapter simply to set out the basic knowledge
which we need to understand the material of the following chapters: it is not
necessary fully to develop the theory, and readers requiring more detail are
referred to the definitive book by Johnson (1985).
We shall be concerned with effects within the outermost layers of bodies,
typically within a millimetre of the surface, and can ignore effects at greater
depths. This allows us to treat the surfaces as surfaces of semi-infinite bodies,
ignoring overall shape, and leads to mathematical simplification.
We are particularly interested in problems of contact between bodies
whose geometries are defined by circular arcs. The problems of elastic contact
between such bodies were first solved by Hertz in 1882, and such situations
are referred to as Hertzian contact. We shall initially concentrate on the
stresses in such contacts.
1.5.1 Stresses in Non-conforming Contacts between Curved

Bodies
We consider first the contact of two identical cylinders under conditions of

plane strain. From symmetry, we can see that compression of the cylinders
will produce a plane-rectangular contact zone, as shown in Figure 1.9(a).
(In the remainder of this chapter, to avoid ambiguity in use of the word
'contact', the shape of the contact zone will be referred to as the footprint.)
For unequal cylinders, the footprint is no longer truly plane, but the
discrepancy is negligible and plane contact can be assumed. As we increase
the load, the width of the footprint will increase, and, as the deformation at
the centre of the footprint is greater than that at its edges, we expect a contact
pressure which varies with position. Therefore, to determine the stress
distribution within the surface, we must define both the size of the footprint
and the pressure distribution within it for any applied load. A detailed
analytical treatment can be found in Johnson ( 1985), but we shall use a
simple physical argument.
12 TRIBOLOGY
(a) (b)
Figure 1.9 Pressure distribution due to contact of two cylinders
For two identical elastic cylinders in contact under a normal load, P,

per unit length, let the resulting plane footprint have a width of 2a, as shown
in Figure 1.9(b ). It is apparent that the stresses in such a system would be
such that stress oc (PI a). The actual pressure distribution is given by
p=(2P/na)(1-x 2 ja 2 )t (1.7)
We also note that an increase in load would cause increases in both the
footprint half-width, a, and the strain, so we expect that strain oc (a/ R), where
R is the radius of the cylinder. From these relations for stress and strain, we
have (P/a) oc E(a/R) or a 2 oc PR/E. The actual solution for this case is
a 2 =4PR(1-v 2 )/nE (1.8)
The solution defined by Equations ( 1.7) and ( 1.8) is almost true for both
identical and non-identical cylinders as long as the angle subtended by the
contact width at the centre of the cylinder is less than 30°. It is also
approximately true, over the same angular range, for contacts between
different materials if we substitute E* and R', respectively, forE and R, where
1/E* = (1- vi}/E 1 + (1- vD/E 2
and
1/R' = 1/R 1 + 1/R 2
where a 2 = 4PR' /nE*. For the case of a cylinder on a plane, the plane has
infinite radius, so that R' becomes the radius of the cylinder only; for concave
curvatures the radius is negative.
We know that the onset of plastic deformation will be associated with
the maximum shear stress in the material reaching a critical value, k, and we
therefore wish to know the distribution of maximum shear stress for a body
loaded with the pressure distribution given by Equation (1.7) acting over the
footprint extending from -a to + a.
The maximum shear stress in plane strain conditions is given by
rmax={(ax-O"z) 2 /4+r;z}t (1.9)
where expressions for ax, az and rxz can be found in Johnson (1985). Therefore,
Equation ( 1.9) defines the values of 'max at all points. The evaluation of this
expression allows us to plot the maximum shear stress distribution beneath
the footprint, and the rather surprising result is shown by the isochromatics
of the photoelastic stress pattern of Figure 1.10. The greatest value of rmax
occurs at a distance of0.79a below the surface, and we find that it will attain
the critical value of k when the maximum pressure, Pmax• at the centre of the
footprint is 3.3k. If we had simple compression, we should expect the surface
material to yield when Pmax attained a value of 2k, but this does not happen
in the present case, because the surface elements are subjected to compressive
stresses in all three orthogonal directions, giving a hydrostatic stress
component which cannot contribute to plastic deformation.
This very important result means that higher loads than expected can
be carried elastically in these non-conforming contacts. Furthermore, even
when sub-surface yielding has taken place, very little plastic deformation can
occur, because the plastic zone is constrained by the surrounding elastic
material.
As the load is increased further, the plastic zone will increase in size and
ultimately spread to the surface, so that plastic indentation can occur. This
happens when the mean pressure is approximately 6k - i.e. approximately
twice the pressure at which yield first occurred. The mean pressure at this
point is essentially the indentation hardness value, H , of the material, which
p..,.,
L.nes ?I
._~-i-- constant
r....,.
Figure 1.10 Actual isochromatics obtained for the contact of a cylinder and a
plane due to normal load alone
14 TRIBOLOGY
Fully plastic
behaviour
Plastic behaviour
constrained by
surrounding elastic matetial
Puft-ly ekJsr•c
behovoour
Figure 1.11
is why, for metals, we find that H ~ 6k ~ 3ay, where O"y is the uniaxial tensile
yield strength of the material.
The behaviour described above is summarised in Figure 1.11.
We have, as yet, considered the effects of normal loads only, and to
understand tribological phenomena, we must consider the effects of both
normal and tangential loads. As we shall see in the next chapter, the tangential
force, F, required to cause sliding under a normal load, W, is given by J1. W,
where J1. is a constant known as the coefficient of friction. The stress field due
to a tangential force can be derived by a method similar to that described
above for a normal force, and combining the stress distributions due to the
normal and tangential loads leads to a distribution of rmax of the type shown
in Figure 1.12. 1t can be seen that the location of the point of maximum shear
stress is now much closer to the surface, so that plastic deformation can take
place more readily than in the previous case. Thus, macroscopic plastic
deformation becomes easier in the presence of friction forces. A full solution
of this problem has been produced by Hamilton ( 1983 ).
In many practical situations, where friction is used to prevent slip
between mating components, contacting bodies are subjected to tangential
forces which are less than JLW. It can then be shown that within the contact
area there exists a central area in which no slip occurs, while outside this
area a small amount of slip can take place. This coexistence of zones of
sticking and microslipping is possible because the contacting materials are
deformable and the stress distribution is such as to allow slipping in the outer
regions of the contact zone. The subject of microslip will be covered in more
detail in the next chapter, when we discuss the nature of friction in rolling
contacts.
Lonn of
Qli'ISianl
r,...
Figure 1.12 Actual isochromatics obtained for the contact of a cylinder and a
plane due to combined normal and tangential loads, where T = 0.5 1-1P
For simplicity, we have concerned ourselves so far with two-dimensional

non-conforming contacts, although in many practical situations we are
concerned with three-dimensional problems. In such situations the patterns
of behaviour are similar to those already described, but some of the equations
describing the behaviour must be modified. For contact between two similar
spheres we have a plane circular footprint and a hemispherical pressure
distribution, as shown in Figure 1.13, given by
(1.10)
where the value of a is given by
a= (3WR/8E*)t (1.11)
Figure 1.13 Pressure distribution due to contact of spheres

16 TRIBOLOGY
Table 1.1 Elastic contact stress formulae
Suffixes 1, 2 refer to the two bodies in contact.

1 1 1
Effective curvature - =- +-
R R1 R2
where R 1 and R 2 are the principal radii of curvature of the two bodies (convex positive).
1 1- vf 1- v~
Contact modulus - = - - + - -
E' E1 E2
where E and v are Young's modulus and Poisson's ratio, respectively.
Line contact Circular contact

(width, 2a; load, P'junit length) (diameter, 2a; load, P)
Semi-contact
width or a=2 - -
nE'
(P'R)t a(~
4 E'
WR)t
contact radius
Maximum contact
pressure -(P'E')l
Po- - -
nR
_1 (6WE'
Po--
n
---
R 2
2 )t
('Hertz stress')
Approach of
centres ~_2P(1-vf(
u -- - - ln4R
- + -1) 1
n E1 a 2
+ 1- v~ (In
E2
4R 2
a
-~))
2
Mean contact P' n p 2

pressure Pmean = 2a = 2 Po Pmean =- 2 = -3 Po
na
Maximum shear 'l:max ~ 0.300 Po !max ~ 0.310 Po

stress at x = 0, z = 0.79a at r = 0, z = 0.48a
for v = 0.3
Maximum tensile 1
stress zero -(1-2v)p 0 atr=a,
3
z=O
Mildly elliptical contacts
If the gap at zero load is z = !Ax 2 + !By 2 and A/ B < 5, take the ratio of semiaxes
bja~(A/Bl
To find the contact area or Hertz stress, use the circular contact equations, with
R. = (AB)-t.
Although the footprint for dissimilar spheres is not a plane circular area,
Equation ( 1.10) still holds with substantial accuracy, but in this case we have
a=(3WR'/4E*)1 (1.12)
For contact between a sphere and l!plane, R' is simply the radius of the sphere.
For general contact between two curved surfaces, the footprint is an
ellipse. This is the case in many important practical situations, particularly
the contact between the balls and bearing races in rolling contact bearings.
In such situations the equations become more complex; it is beyond the scope
of this text to derive them or to state them in full, and readers requiring
more information should consult either Johnson ( 1985) or Halling ( 1975).
Fortunately, for mildly elliptical footprints, the equations for circular
footprints can be used in a slightly modified form.
In Table 1.1 we summarise the most commonly required formulae for
rectangular footprints, circular footprints and mildly elliptical footprints.
1.5.2 Surface Deformations in Non-conforming Contacts
In considering asperity contacts, where we approximate the asperity tip as

a hemisphere, we shall find that it is necessary to know the normal approach
of a sphere to a plane, and the resulting deformation, due to the application
of a normal load.
Consider the contact between a sphere and a plane, as shown in Figure
1.14. The separation, u, of the surfaces at a distance r from the centre of the
circular footprint is given by
u = R- (R 2 - r 2 )t
which, when r is small compared with R, is given by
u = (r 2 j2R) (1.13)
The normal approach is defined as the distance by which points on the
two bodies remote from the deformation zone move together on application
Figure 1.14 Elastic contact between a sphere and a plane

18 TRIBOLOGY
of a normal load; it arises from the flattening and displacement of the

surface within the deformation region. If a is the radius of the footprint
and w is the displacement of the sphere at the footprint boundary, then
the normal approach, D, will be given by
D=u+w=(a 2 /2R)+w (1.14)
We know that at the centre of the footprint D is the degree of deformation

and it is reasonable to assume that the normal approach will be proportional
to the flattening of the sphere - i.e.
Doc a 2 j2R
Substituting the value for a given by Equation ( 1.12), we see that

Doc ( w2 I E*2 )t
The exact results show that
or
W = (4/3)E*RW (1.15)
Combining Equations ( 1.12) and ( 1.15), we see that the area of contact, A,
will be given by
A =na 2 =nRD (1.16)
Equation ( 1.16) indicates that the surface outside the footprint is displaced
in such a way that the area of contact is exactly half the area 2nRD which
would be obtained by plastic flattening of the sphere.
Equation ( 1.16) also shows that for spherical contacts on a plane
D= a 2 jR.
As before, plasticity first occurs when the maximum elastic Hertzian
contact pressure reaches 0.6H, and we know that for this geometry
Pmax = 1.5Pmean• so that, at the onset of plasticity, (3/2) W jna 2 = 0.6H.
Substitution for Wand na 2 from Equations (1.15) and (1.16) shows that the
maximum normal approach before the onset of plastic deformation is given by
(1.17)
We shall see later that this result is of major significance in the contact
of rough surfaces.
The results described above apply to perfectly elastic/plastic materials,
while many real materials, particularly metals, exhibit strain hardening.
Although no analytic solution exists for such materials, Halling and Nuri
(1974) have developed a simple treatment which shows the effect of strain
hardening. They consider the deformation behaviour of any material to be
represented by a stress-strain law of the form

a= Pe" (1.18)
When n = 0, we have rigid/perfectly plastic behaviour, where p = ay; when
n = 1, we have elastic behaviour, where p = E. They also assumed that the
normal approach, <5, could be defined by
a2 = A.R£5 ( 1 .1 9)
with)., having the values 1 when n = 1 and 2 when n = 0. They were able to
make experimental measurements justifying these assumptions and, given the
necessary experimental data, were thus able to define the normal approach
and footprint size for the non-conforming contact of strain hardening
materials.
1.5.3 The Effects of Surface Films
So far, we have considered contact situations involving, at the most, two

different materials. However, tribological contacts often involve surfaces
covered by films of other materials. These films may be adventitious - for
example, due to chemical reaction with the environment; they may arise as
a result of the rubbing process itself; or they may be deliberately formed using
one of the processes described in Chapter 4. Clearly, as the film thickness
tends to zero, the behaviour is determined entirely by the properties of the
substrate, while, when the film thickness becomes very large, the behaviour
is determined entirely by the properties of the film. Between these extremes
the behaviour is a function of the properties of both the film and substrate
materials. This is illustrated by the results of Figure 1.15, which show the
2.5
Film thickness (J..!m)
Figure 1.15 Microhardness values of lead films on steel substrates as a function

of film thickness
20 TRIBOLOGY
£'"
1.4 K = __!_
E'"
2
0.05
1.2
1.0
.
0
~
0.8
0.6
0.4
10-3 10-2
tlao
Figure 1.16 Effect of film thickness on the Hertzian contact of coated surfaces
microhardness values for lead films of various thicknesses on mild steel

substrates.
A study of the Hertzian contact of a sphere on a substrate coated with
a different material (Sherbiney and Halling, 1976) gave the results shown in
Figure 1.16. Here
1/ ET = (1 -vi)/ E 1 + (1 - vD/ E 3
and
1/ E! = (1/vn/E 2 = (1- vn/ E 3
where the suffixes 1, 2 and 3 refer to the substrate, film and sphere materials,
respectively. In this figure, a is the actual footprint radius, a 0 is the value of
the contact radius for bulk film material and t is the film thickness. As would
be expected, as t increases, a tends to a 0 , and a is approximately equal to a 0
when t is equal to a0 . These results are valid for any constant values of load
and sphere radius.
As yet, there is no satisfactory description of the non-conforming plastic
contact of coated surfaces. Halling ( 1986) has suggested a relationship of the
form
( 1.20)
where He is the effective hardness of the coated surface and the subscripts
1 and 2 refer to the substrate and film materials, respectively. It has been
shown that this equation fits the results for lead films on mild steel substrates
for a range of R and t when r has a value of 125.
The most important observation to be made from these results is that,
as would be expected, for soft films on hard substrates the load capacity
decreases as film thickness increases, while for hard films on soft substrates
the converse is true.
1.5.4 The Contact of Rough Surfaces
Although all surfaces are rough, we can simplify the study of two rough
surfaces in contact if we consider the situation to be that of a single rough
surface in contact with a perfectly smooth surface. We can further simplify
the problem by modelling the asperities on the rough surface as sectors of
spheres, so that their elastic deformation characteristics can be modelled by
the Hertz theory for the contact between a sphere and a plane. Finally, we
shall also assume that there is no interaction between the separate asperities,
so that vertical displacement due to a load on one asperity does not affect
the heights of the neighbouring asperities.
We consider first the simple, but unrealistic, model illustrated in Figure
1.17, where all the asperities on the plane of unit area have the same height,
z, relative to the reference plane XX'. As the surfaces are loaded together,
we see that the normal approach is (z- d), where dis the current separation
of the smooth surface and the reference plane in the rough surface. Each
asperity is deformed equally, and carries the same load, W;, so that for '1
asperities per unit area the total load W will be '7W;. For each asperity, the
load, W;, and the area of contact, A;, are known from the Hertzian theory.
Thus, if p is the asperity radius, we have
W; = (4/ 3)E*P(z- d)t
and
A;= nP(z- d)
The total load will then be given by
W = (4/ 3)E*{P(A;/np)!
so that the load is related to the total area of contact, A(= '7A;), by
W = (4E/ 3nt'ltp) ·A! ( 1.21 )
This result shows that, for this particular model, the real area of contact is
proportional to the two-thirds power of the load when the deformation is
elastic.
_x t-"F-~~ ·-¥3-
Refer~e
plone on
rrugh surface
Figure 1.17 Contact between a smooth plane and an idealised rough surface
22 TRIBOLOGY
If the loads are such that the asperities are deforming plastically under
the constant flow pressure, H, each individual contact area, A', will be given
by 2npo. The individual load, w;, will then be given by
w; = HA; = 2Hnp(z- d)
Thus,
W' = 11 w; = 11 HA; = HA' = 2HA (1.22)
so that the real area of contact is proportional to the load.
We shall see in Chapter 2 that it is basic to observed frictional behaviour
that the real area of contact of rough surfaces should be proportional to the
load. While our simple model gives this proportionality for plastic deformation,
it does not do so for elastic deformation.
However, we know that our model of the rough surface is unrealistic.
As pointed out earlier in this chapter, on real surfaces asperities have different
heights following some probability distribution, and we must, therefore,
modify our surface model to take this into account.
Therefore, we consider the contact of a rough surface, having a height
distribution cf>(z), with a smooth reference plane, as shown in Figure 1.18. If
the separation of the two surfaces under load is d, then there will be contact
of any asperity with an original height greater than d. The probability that
any asperity has a height between z and (z + dz) above the reference plane
will be cf>(z) dz. Thus, the probability of contact for any asperity of height z
is given by
prob(z>d)= Loo cf>(z)dz (1.23)
If we consider unit area of the surface containing '1 asperities, the number of
contacts, n, will be given by
n = '1 Loo cf>(z) dz ( 1.24)
Since the normal approach is (z- d) for any asperity and W; and A; are
known from Equations (1.15) and (1.16), the total area of contact and the
Figure 1.18 Contact between a smooth plane and a rough surface

corresponding load will be given by
A= rt'lP LX> (z- d)c/>(z) dz (1.25)
and
W=(4/3)'1PE* f., (z-d)lcJ>(z)dz (1.26)
It is convenient and usual to express these equations in terms of standardised

variables by putting h = d/a and s = z/a, where a is the standard deviation
of the peak height distribution of the surface. Thus,
n = '1 f'' c/>*(s) ds
A= 1r11Pa loo (s- h)c/>*(s) ds

W = (4/3)'1tfza1E• loo (s- h)lcJ>*(s) ds
c/>*(s) being the probability density scaled to unit standard deviation.
Using these equations, the number of contact spots, the real area of
contact and the load can be evaluated for any given height distribution.
An interesting situation arises for an exponential height distribution -
i.e. c/>(s) = e-•. In this case we have
n = 'le-h
A= nt~Pae-h
W = nt'l ptatE*e- h
These equations give W = C 1 A and A= C 2 n, where C 1 and C2 are constants
of the system. Thus, for this distribution, the area of contact and the number
of contact spots are both proportional to the load, even though the asperities
are deforming elastically.
For other distributions such a simple relationship will not apply, but it
will be very nearly true for any distribution which approaches an exponential
shape for the highest 10% of the asperities. As we stated in Section 1.4.3,
many surfaces have distributions which are Gaussian or near-Gaussian in
the outer 10% of the texture; with such surfaces the area of contact is found
to be almost proportional to the load over several orders of magnitude ofload.
1.5.5 Criterion of Deformation Mode
We shall see in the following two chapters that both friction and wear are
very strongly influenced by the degree of plastjc interaction of the contacting
24 TRIBOLOGY
asperities. In most engineering situations some asperities will deform plastically,

while others will deform elastically, and we thus have a mixed elastic-plastic
contact where the greater the load and, hence, normal approach, the greater
the number of plastic contacts. Thus, the normal approach should be an
indicator of the degree of plasticity.
From Equations ( 1.15) and ( 1.16) we can see that the mean contact
pressure, Pmean• for an elastic asperity contact is given by
Pmean = (4£*<51)/( 3ntp)
or
( 1.27)
For a spherical contact we know that the transition from first yield to full
plasticity takes place as the contact pressure increases from approximately
0.5H to H. To a good approximation, therefore, we can calculate the normal
approach for the onset of plasticity by
<5~ =(HI E*)p!
It is convenient to standardise this by dividing both sides by at, so that
<5*1 = (be/ a)1 = (H / E*)(P / a)t
However, this parameter decreases as the surface roughness increases, and
it is, therefore, usual to define a function t/1, the plasticity index, as the inverse
of <)*!, so that
t/1 = (E* /H)· (ajp)! ( 1 .28)
This index, first defined by Greenwood and Williamson ( 1966), is very
indicative of the degree of plastic deformation. When the index is greater
than unity, surface deformation is largely plastic, while when it is less than
0.6, the deformation is largely elastic; these statements are true whatever the
load, as long as the asperities continue to deform independently.
The parameter depends on both the mechanical properties and the
surface topography of the contacting surfaces; we shall return to this point
when comparing the tribological properties of different types of materials in
the following chapters.
Recently Halling et al. ( 1988) have extended this theory to allow
predictions of the relative amounts of elastic and plastic deformation.
In the original formulation by Greenwood and Williamson ( 1966) the
upper limit of the distribution curve was taken as infinity, as it must be, if
the curve is truly Gaussian. However, this makes it impossible to define the
surface separation at which contact first occurs (with a true Gaussian
distribution, there is a finite, though very small, probability that first contact
will occur when the mean separation of the surfaces is some metres). Halling
et al. ( 1988) overcame this problem by truncating the distribution so that
the probability of finding an asperity fell to zero at some finite distance, ba,
above the mean line. This is open to some reservations; in particular, the
expected height of the highest asperity, as measured by b, will not have a
constant value but will increase as the surface area increases. However, even
for Gaussian surfaces, about 99% of all the asperities lie within a distance
3a of the centre line and, as we shall see, truncation of the distribution at a
value of 3a - i.e. at a value of b = 3 - gives a very good fit of theory to
experimental observations.
Halling et al. then considered the situation shown in Figure 1.19. It can
be seen that the asperities from d to (d +be) are deforming elastically, while
the asperities from (d +be) to ba are deforming plastically. By integrating
the distribution curve between these two sets of limits, it is then possible to
define the ratio between plastically deforming area and elastically deforming
area for different contact pressures and different values of the plasticity index,
1/J. The results for truncation at 3a are shown in Figure 1.20; where R is the
ratio between plastic area and elastic area, and p is the non-dimensional
nominal pressure (N I AH). For R = 0, we have the result for purely elastic
deformation. It can be seen that this approach allows us to determine the
relative amount of plastic deformation for different values of the plasticity
index and different nominal pressures. As both friction and wear are very
dependent on the degree of plastic deformation, it is likely that this approach
will prove to be very useful in the future in helping us to predict friction and
Figure 1.19 Elastic-plastic contact of a smooth surface and a rough plane with
a peak height truncated at bu
Figure 1.20 Ratio between plastic and elastic contact area as a function of plasticity
index, t/J, and normalised pressure, jj
26 TRIBOLOGY
wear behaviour. The current difficulties in making such predictions will

become abundantly clear in the following chapters.
Finally, it should be noted that with Hertzian contacts, such as bearings,
gears and cams, the effect of roughness at low loads is to lead to contact
sizes much larger than those predicted for smooth surfaces. A useful guide
to the effect of roughness on Hertzian predictions has recently been proposed
by Greenwood (1984). A parameter a is defined as
a=u'R'ja 2 (1.29)
where u' is the combined roughness of the two surfaces, R' is their relative
radius of curvature and a is the predicted Hertzian contact radius. Then, if
a< 0.05, the error in using the Hertzian equations to calculate the contact
radius will be less than 7%.
1.6 PROBLEMS
(In solving these problems, the following properties may be assumed. Young's
modulus( GPa): steel, 200; copper, 124; tungsten carbide, 550. Poisson's ratio:
0.3 in each case.)
1 A surface profile is sinusoidal, with unit amplitude and wavelength A..
The profile is sampled at equal intervals, with the origin on the centre
line at a position of zero amplitude. Calculate the computed RA
values for this profile for sampling intervals of A.j2, A./4, A./8 and A./16.
Calculate also the RA value which would be derived from the analogue
signal for the same profile.
[Computed RA values: zero; 0.5; 0.603; 0.628. Analogue value: 0.637]
2 A hard steel cylinder 100 mm long and 10 mm in diameter is loaded
under an increasing normal force against a flat copper plate having a
hardness of 6000 MPa. At what load does the material of the plate first
yield; what is the corresponding contact width; and what are the mean
and maximum pressures in the contact zone?
[Py = 2200 N; 2a = 0.084 mm; Pmean = 261 N/mm 2 ; Po= 393 N/mm 2 ]
3 A 2 mm diameter tungsten carbide ball is pressed into a 5 mm diameter
hemispherical recess in a steel plate under a load of 50 N. Calculate
(a) the minimum hardness of the steel plate for the contact to remain fully
elastic; (b) the diameter of the contact zone.
[(a) H=5245 MPa; (b) 2a=0.056 mm]
4 Two similar, macroscopically flat steel surfaces are ground to give an
RMS value of 0.2 11m and a mean asperity radius of 300 Jlm. Calculate
(a) the maximum value of hardness at which the deformation of the
asperities is predominantly plastic and (b) the minimum value of hardness
at which the deformation of the asperities remains predominantly elastic.
Take the mixed elastic-plastic range to be defined by 0.5 < 'P < 1.
[(a) H=4750 MPa; (b) H=7920 MPa)]
Chapter 2
Friction
2.1 INTRODUCTION
Friction is the resistance to motion which occurs whenever one solid body
slides over another. The resistive force, which acts in a direction directly
opposite to the direction of motion, is known as the friction force. The friction
force which is required to initiate sliding is known as the static friction force,
while that required to maintain sliding is known as the kinetic friction force.
Kinetic friction is usually lower than static friction.
2.1.1 The Laws of Friction
It has been observed experimentally that there are two basic 'laws' of
friction, which are usually obeyed to a good approximation. The laws are
entirely empirical and no physical principles are violated in those cases where
the laws are not obeyed.
The first law states that the friction is independent of the apparent area
of contact of the two bodies, and the second states that the friction force is
proportional to the normal load between them. Thus, a brick can be slid as
easily on its side as on its end, and if the load between two bodies is doubled,
then the friction force is also doubled. These laws are known as Amon ton's
laws, after the French engineer who first presented them in 1699.
Coulomb in 1785 proposed a third law, that the kinetic friction is almost
independent of the speed of sliding, but this law has a smaller range of
applicability than have the first two.
2.1.2 The Coefficient of Friction
Amonton's second law states that the friction force, F, is proportional to the
normal load, W - i.e.
( 2.1 )
27
28 TRIBOLOGY
where Jl, the constant of proportionality, is known as the coefficient of friction.

As with the friction force, we can define coefficients of static and kinetic
friction, but, unless stated otherwise, the coefficient of friction is normally
taken to mean the kinetic coefficient.
It must be stressed that the coefficient of friction is not a constant for
any particular material, but is typical of two materials, which may be similar
or dissimilar, sliding against each other under a given set of surface and
environmental conditions. The often-posed question 'What is the coefficient
of friction of steel?' has no meaning.
2.1.3 The Basis of Amonton's Laws
To explain the form of Amonton's laws, we start by making the following

two assumptions.
( 1) During sliding, the resistive force per unit area of contact is constant, so
that
F=A·s (2.2}
where F is the friction force, A is the real area of contact and s is the
friction force per unit area.
(2) The real area of contact, A, is proportional to the normal load, W- i.e.
A=q·W (2.3}
where q is the constant of proportionality.
Eliminating A from these two equations gives

F=q·s·W (2.4}
If we can justify our assumptions, then these equations explain' the form of
Amonton's laws: (2.2) and (2.3) tell us that the friction force depends upon
the real area of contact and that this is independent of the apparent area,
while (2.4) tells us that friction force is proportional to the normal load. It
therefore remains to justify the assumptions.
Assumption ( 1) is easily justified; it involves no assumptions about the
nature of the specific friction force, but is simply a statement that any part
of the contact area is statistically representative of the whole.
Assumption (2) could not be justified under all circumstances, but we
haye already shown in Chapter 1 that it is justified: (a) whenever contact is
wholly plastic, regardless of the surface topography; (b) whenever the
contacting surfaces have exponential asperity height distributions, regardless
of the mode of deformation; (c) to a very good approximation, whenever the
contacting surfaces have a Gaussian asperity height distribution, again
regardless of the mode of deformation.
FRICTION 29
Until the pioneering work of Greenwood and Williamson ( 1966 ), which

demonstrated the critical importance of surface topography as summarised
in (b) and (c) above, it was assumed that Amon ton's second law implied
that frictional contacts must be plastic. However, it is now clear, as explained
in Chapter 1, that almost all rubbing surfaces will simultaneously contain
both elastic and plastic contacts, and that it is the generally Gaussian or
near-Gaussian surface topographies of most engineering surfaces which
explain the wide range of validity of this law.
2.1.4 The Prediction of Friction Coefficients
The process of relative sliding of two materials appears to be a simple one,

but this appearance is extremely deceptive. There is, as yet, no theoretical
method of predicting the coefficient of friction (or, as we shall see in the next
chapter, the rates of wear) when any two materials are in sliding contact. It
should now be clear from a reading of Chapter 1 that both friction and wear
are the results of extremely complex interactions between the surface and
near-surface regions of the two materials. These regions differ from the bulk
of the materials in their physical, chemical and mechanical properties.
Furthermore, these properties themselves can change radically as a result of
interactions of the surface atoms with their environments and with each other.
It would be difficult to better the remark of Pauli, recently quoted by Tabor
(1985), that 'God made solids, but surfaces were made by the Devil'.
As a result of this complexity, if it is necessary to know the coefficient
of friction of a particular pair of materials under a particular set of operating
conditions, then the safest procedure is to measure it experimentally, under
conditions as close to the operating conditions as is feasible.
However, although accurate quantitative predictions of friction coefficients
are not yet possible, the conceptual understanding of friction processes is
now quite advanced, and this should help the practising tribologist to make
sensible choices of pairs of materials for given rubbing conditions. It is the
purpose of this chapter to impart this basic understanding. However, before
turning to this, it will be useful to describe briefly how friction coefficients
are measured.
2.2 THE MEASUREMENT OF FRICTION
In any apparatus for measuring friction, two specimens are placed together
under a known normal load, one of the specimens is caused to slide relative
to the other, and the tangential force resisting sliding is measured. There are
many methods available, of varying degrees of complexity and cost.
Probably the simplest method is to use the inclined-plane test illustrated
in Figure 2.1. In this test a plate of one of the chosen materials is securely
30 TRIBOLOGY
Figure 2.1 Measurement of friction using the inclined-plane test
fastened to a horizontal but tiltable plane surface; a block of the second

material, of weight S, is placed upon it; and the plane is then gradually tilted
until it reaches some angle () to the horizontal at which the block begins to
slide. At this point we have
J.1 = F I W = S sin ()IS cos () = tan ()
This method measures the coefficient of static friction and is obviously
unsuitable in those cases where an investigation of the variation of friction
with continuous rubbing is required. However, it can be used to make rapid
measurements on many pairs of materials, and it is also a useful aid in
teaching, where it can be used, for example, to demonstrate the validity of
Amon ton's laws.
Where continuous friction and wear measurements are required over a
period of time, alternative tests must be used. In all such tests one specimen
is driven continuously, while a second specimen, which is nominally stationary,
is loaded against it. The loading of the stationary specimen can be by simple
dead weight or, if the experimental conditions demand it, by some more
complex mechanism such as spring loading, or hydraulic or pneumatic
pressure. The measurement of the friction force is usually accomplished by
mounting the nominally stationary specimen so that the friction force causes
a very small tangential movement of either a capacitive or an inductive
transducer, which is calibrated to give a continuous measurement of friction
force.
The testing geometries for these more complex tests can be divided into
two groups: conformal geometries, where the contact pressures are moderate
and normally constant throughout the test, and non-conformal geometries,
where the contact pressures are initially high and often vary with time.
Examples of conformal tests and non-conformal tests are illustrated
schematically in Figures 2.2 and 2.3, respectively. The conformal geometries
can be used to simulate the tribological behaviour of devices such as: brakes,
thrust washers, plain bearings or face seals, whereas the non-conformal tests
can be used either to simulate the behaviour of high-pressure contacts, such
as those occurring between gear teeth, or, for example, to provide accelerated
tests of the friction and wear behaviour of a number of candidate material
pairs.
FRICTION 31
w
w
+
(a) (b)
I w~
-~- I
--_)
w
(c) (d)
Figure 2.2 Measurements of friction using conformal geometries: (a) flat pin on
disc; (b) thrust washer; (c) pin on cylinder; (d) shaft in bush
w
w
+
(a) (b)
Figure 2.3 Measurements of friction using non-conformal geometries: (a) hemi-

spherical pin on disc; (b) crossed cylinders
2.3 THE ORIGINS OF FRICTION
The purpose of this section is not to give a state-of-the-art review of the large
amount of research that has taken place in this area, but rather to give a
clear understanding of the concepts involved.
We know that the friction force is due to interactions between the
opposing asperities of the two sliding surfaces. Each asperity interaction will
contribute to the friction force, so that the total friction force at any time will
be the sum of the forces at the individual contacts. Furthermore, as the
surfaces move relative to one another, the energy supplied by the forces
causing the motion is continuously dissipated at or near the sliding surfaces
and, again, the energy dissipated in unit time will be the sum of the energies
32 TRIBOLOGY
dissipated in the individual processes occurring during the same time. The
observation that friction forces are often almost constant is due to the fact
that the number of asperity interactions taking place at any time is so large
that the statistical distribution of the contact processes is almost constant.
Clearly, in equilibrium the energy dissipated per unit time is equal to
the product of the friction force and the sliding velocity, so we can approach
the problem of explaining the nature of friction by considering either the
origins of the forces at the contacts or the processes of energy dissipation. In
what follows it will be convenient to use both approaches.
In considering asperity interactions, it is conventional to describe two
types of interaction: adhesion and deformation. This convention will initially
be followed here, but as the discussion is developed, it will be stressed that
their contributions to friction are not simply additive, but interactive.
2.3.1 Adhesion
When two surfaces are loaded together, they can adhere over some part of
the true contact area, to form friction junctions, and these junctions must
then be broken if relative sliding is to take place. The break will occur at
the weakest part of the junction, which may be either at the original interface
or in the weaker of the two materials. If the break occurs at the original
interface, the interaction has simply subjected the two asperities to a stress
cycle, although, as we shall see in Chapter 3, the accumulation of such cycles
may ultimately cause the creation of a wear particle by a fatigue mechanism.
If the break occurs in the softer of the two materials, then a fragment of this
material will be transferred to the harder counterface. In either case it is clear
that adhesion is one form of surface interaction which would cause frictional
resistance.
Before considering the contribution of adhesion to friction, it is useful
to describe the adhesion of two solids placed in static contact. In doing this,
use is made of the thermodynamic concepts of surface and interfacial energies.
The surface free energy of a solid is defined as the thermodynamically
reversible energy involved in creating unit area of new surface. Similarly, if
two surfaces are brought into contact, the interfacial free energy is defined
as the thermodynamically reversible energy involved in creating unit area of
interface. Thus, if we take two surfaces having surface energies y1 and y2 and
place them in contact, to form an interface with interfacial energy y1 , 2 , then
the energy released is
ll.y = 1'1 + Y2- 1'1,2 (2.5)
This quantity is known as the thermodynamic work of adhesion as, in
principle, a similar amount of work is needed to separate the bodies.
Although Equation (2.5) has been applied very successfully to liquids,
it has been shown by Tabor (1985) that it cannot be applied precisely to
FRICTION 33
even ideally elastic/brittle solids. For solids of other types, the energy
calculated from Equation (2.5) is only a small fraction of the energy actually
required to separate the surfaces; the reasons for this will be made clear later.
However, although the thermodynamic work of adhesion is of little relevance
in many contacts, the actual work and force needed to separate the surfaces
are highly relevant. Tabor (1985) -proposed for these the apposite terms
'pull-off work' and 'pull-off force' and these will be adopted here.
2.3.2 Deformation
If no adhesion takes place, then the only alternative interaction which would
result in a resistance to motion would be one in which material is deformed
and displaced during relative motion. We need consider only two interactions
of this type: the microscopic interaction illustrated in Figure 2.4(a), where
deformation and displacement of the interlocking surface asperities are
required, and the more macroscopic interaction illustrated in Figure 2.4(b ),
where the asperities of the harder material plough grooves in the surface of
the softer. Other examples of the second type of interaction would involve
...
- Ia)
....
(b)
Figure 2.4 Sliding interactions of surfaces: (a) asperity interaction only; (b)
macroscopic ploughing interaction
34 TRIBOLOGY
the ploughing of one. or both surfaces by wear particles trapped between

them, and truly macroscopic ploughing of the softer material by the harder,
with the dimensions ofthe ploughed groove being orders of magnitude greater
than those of the asperities on either surface.
Asperity interactions will always be present and, together with adhesion,
will often be the major cause of friction. The ploughing contribution may or
may not be significant; its magnitude will depend on the surface roughnesses
and relative hardnesses of the two surfaces, and on the size, shape and hardness
of any wear debris and reaction products trapped between them. In the
following section we shall describe how the three processes adhesion,
ploughing and asperity interaction contribute to friction.
2.4 THEORIES OF FRICTION
2.4.1 Simple Adhesion Theory
This theory, due to Bowden and Tabor (1950), was the first modern
explanation of the existence of friction. It was developed for ideal elastic-
plastic metals and the same approach will be followed here, with the
importance of the theory to other types of materials being covered later in
the chapter.
The theory can be summarised as follows.
( 1) When surfaces are loaded together, they actually make contact only at
the tips of asperities.
(2) Even at low loads, the real contact pressures are so high that the asperity
tips of the softer material deform plastically.
(3) This plastic flow causes the total contact area to grow, both by growth
of the individual initial contacts and by initiation of new contacts, until
the real area of contact is just sufficient to support the load elastically.
(4) Under these conditions, for an ideal elastic-plastic material
W=A·p 0
where W is the normal load, A is the real area of contact and p0 is the
yield pressure of the softer of the two materials. The yield pressure is
very nearly the same as the hardness, H, measured in an indentation
test, so that this can be rewritten as
W=A·H (2.6)
( 5) As a result of the severe plastic deformation, the asperity junctions cold
weld - i.e. strong adhesive bonds are formed.
(6) The specific friction force, s, is then simply the force required to cause
shear failure of unit area of asperity junction, so that
F=A·s (2.7)
FRICTION 35
Neglecting, for the moment, any ploughing contribution, Equations (2.6)

and (2.7) can be combined to give
F/W=p.=s/H (2.8}
It can be seen that the simple theory provided what was the first theoretical
explanation of Amonton's laws: that friction is independent ofthe apparent
area of contact and that friction force is proportional to the normal load.
In the above analysis we considered an ideal elastic-plastic material
and ignored the effect of work hardening. For the case of failure in the softer
of the two materials, we can take s equal to k, the critical shear stress of this
material, so that
p. = k/H (2.9}
This ratio k/ H is fairly constant for most materials, and Equation (2.'.:l)
indicates why many material pairs have similar friction coefficients despite
the fact that the individual values of k and H vary very widely. As shown in
Chapter 1, a typical value of k/ H is ~ 0.16, which is of the same order as
the observed values of p. for many material pairs in normal atmospheres.
According to this simple theory, the effect of ploughing was taken into
account simply by adding a term Jl.ploughing to the coefficient calculated from
Equation (2.9). The derivation of such ploughing terms is described in
sub-section 2.4.3.
Although this simple theory is attractive, it is inadequate in several
important respects.
( 1) The friction coefficient given by Equation (2.9) depends only on the
mechanical properties of the softer of the two materials, so we should
expect a particular material to have the same friction coefficient against
any harder counterface: this is not found to be the case.
(2) Actual values of p. in normal atmospheres, particularly for metal pairs,
are normally of the order of 0.5, rather than the value of ~ 0.16 which
is given by Equation (2.9).
( 3) Many materials, particularly ductile metals, exhibit friction coefficients
much higher than those quoted above when their surfaces are free of the
normal contaminant and oxide films- a condition which can be achieved
in ultra-high vacuum. Friction coefficients much greater than unity are
often observed, and in extreme cases gross sliding does not occur at all
and a weld is formed over much of the nominal contact area.
These deficiencies led Bowden and Tabor to re-examine some of the
assumptions of the simple adhesion theory, and to present a more realistic
description of friction in terms of adhesion.
2.4.2 Extension of the Adhesion Theory
In the simple adhesion theory the effects of the normal and tangential loads
were considered separately; the true area of contact was assumed to be
36 TRIBOLOGY
determined by the normal load only (Equation 2.6) and the friction force
was then taken to be the force required to cause shear over this area (Equation
2.7). However, the very high friction coefficients observed between clean metal
surfaces indicate that the true area of contact must be far greater than that
predicted by the simple adhesion theory. This was explained when Bowden
and Tabor considered the combined effect of the normal and the shear
stresses.
To illustrate this, we consider first the simple two-dimensional stress
system shown in Figure 2.5(a), and assume that yielding occurs when the
maximum shear stress attains the critical value k. At this point k is given by
the radius of the Mohr's circle shown in Figure 2.5(b ). (The Mohr's circle
construction is an illuminating technique for illustrating compound stresses,
and readers who are not familiar with it should consult any standard text
on strengths of materials, e.g. den Hartog, 1952). Using this construction it
can be seen that yield occurs when
(p/2)2 + !2 = k2
or
(2.1 0)
We can now examine the effects of the combined normal and shear stresses
on the real area of contact in an asperity junction.
Consider first a single asperity junction under a normal load W, having
an area of contact A, defined by A = WI H. Note that the material is at its
yield point, defined by H, and that the maximum shear stress in the material
is equal to k, so that any increase in either the normal stress or the shear
stress in the junction will cause further plastic flow. Thus, if we impose an
increment of tangential force,f, some plastic flow will occur and this will lead
Shear
stress
llllllll
Normal
s srress
(a)
(b)
Figure 2.5 Mohr's circle construction for finding the maximum shear stress for
an idealised two-dimensional junction under normal and tangential stress
FRICTION 37
to an increase in the contact area. This process is known as junction growth.

As the area of the junction grows, both the normal and the shear stresses
will decrease, and this process will continue until the maximum shear stress
in the junction has fallen to its previous value of k, when the load will again
be just supported elastically.
If we then impose further increments of tangential force, the above
process will be repeated each time - i.e. the junction will continue to grow
and no gross sliding will occur. In such cases the term 'coefficient of friction'
is not strictly applicable, but, as the normal force remains constant while the
tangential force continues to increase, the ratio between tangential stress and
normal stress can attain a very high value, limited only by the gross shear
strength of the softer of the two materials. Indeed, once strong adhesion has
occurred, a strong resistance to shear can still be experienced if the normal
load is removed, or even reversed.
We would expect the combined normal and shear stresses to obey an
equation of a form similar to that given in Equation (2.10) for a simple
two-dimensional junction. The exact form for a three-dimensional case is not
known, but we assume that it is of the form
( 2.11 )
where IX and r are constants which we can determine by considering two
boundary conditions.
( 1) When there is no tangential force, r is zero and p = H. Therefore, r 2 = H 2
and
(2.12)
(2) When large-scale junction growth has taken place, the friction force
is much greater than the normal load, and, as these loads both act
over the same area, the normal stress must be much less than the
shear stress - i.e. r » p. Therefore, 1Xt 2 ~ H 2 • In this case r must be
approximately equal to k, so that
IX~ H 2 jk 2 (2.13)
From Equation (2.13)we should expect IX to have a value of approximately
36. However, experiments suggest that the value of IX is lower than this, and
Bowden and Tabor assumed a value of9; as we shall see below, the exact
value of IX has only a secondary effect on the amount of junction growth
which takes place.
Before moving on, it is useful to re-examine the second boundary
condition described above - i.e. that when large-scale junction growth has
taken place, the shear stress is very much greater than the normal stress. As
the shear stress cannot exceed the shear strength of the softer material, this
tells us that the normal pressure must now be very much lower than the
yield pressure of that material. This is clearly the case, as the normal load
38 TRIBOLOGY
has remained constant, while the area over which it acts has greatly increased,
but it is worth emphasising that junction growth continues under the
increasing shear stress, even when the normal stress is negligible.
This revised treatment was still based on the assumption that all junctions
initially deform plastically under the applied normal load. However, this is
an unnecessary restriction, as the same type of reasoning can also be applied
to the many asperity contacts which, as shown in Chapter 1, deform elastically
under the applied normal load. In such contacts, although the initial contact
may be well within the elastic range, the same shear stress will still be required
to shear the junction, and for clean surfaces this will be the full shear strength
of either the interface or the weaker of the two materials. Thus, a contact
which was initially elastic will be subjected to plastic shear at a low level of
normal stress - a condition which is similar in all respects to the shearing
of an initially plastic contact after some junction growth.
We have now explained the very high friction coefficients which can be
obtained with clean metals in vacuum. We have not yet explained why such
high coefficients are not observed in normal atmospheres, but it is clear that
junctions formed in such atmospheres must be weaker in shear than those
formed between clean surfaces, and that this relative weakness must be due
to the presence of the contaminant films which are always present in normal
atmospheres. Following Bowden and Tabor, we therefore assume that at any
junction there is a thin contaminant film of shear strength r., and that r. = ck,
where c is less than unity. While the frictional stress, F I A, is less than r.,
junction growth will proceed as described above; but when F /A= r., the
contaminating film will shear, junction growth will end and gross sliding will
occur.
Thus, the condition for sliding is
p2 + ar; =H 2
But it has already been shown that
H 2 =ak 2
Therefore,
or
Hence,
(2.14)
FRICTION 39
According to this theory, as c tends to unity, then Jl tends to infinity,

in agreement with the previous result for clean metals. The friction coefficient,
Jl, is plotted against c for various values of ex in Figure 2.6, and we see that
Jl falls rapidly as c reduces from unity. Thus, a small amount of weakening
at the interface produces a drastic reduction in Jl. It is also clear from
Figure 2.6 that, as mentioned earlier, the exact value of ex is not of major
importance.
Note that when cis small,
w:::::. cI ext
so that we can obtain a low friction coefficient by arranging to have a film
of low shear strength separating the surfaces. This is the principle underlying
lubrication by soft metal films and boundary lubricants, which will be
discussed later.
20
Figure 2.6 Variation of Jl with c for different values of IX. It can be seen that, except
at large values of c, the exact value of IX is not of major importance
40 TRIBOLOGY
It is worth noting that this modified theory will also apply when the
interface is inherently weaker than either of the asperity materials, even in
the absence of contaminating films: in this case -r. is simply the shear strength
of the interface.
In this final modification to the adhesion theory, unlike the case for
clean surfaces, there can be a very marked difference in behaviour between
the junctions which are initially plastic and those which are initially elastic.
In the former case there will always be some continued plastic deformation
and junction growth until the interface shears, whereas in the latter case the
interface may shear while the deformation of the asperities is still within the
elastic range. The importance of this fact will emerge in Chapters 3 and 4.
In the foregoing discussion we have presented the various versions of
the adhesion theory· as first put forward by Bowden and Tabor. Johnson
(1968) has suggested that Equation (2.11) should be modified to
(2.15)
so that Equation (2.14) becomes
c
11 - -----=-·
]l
-[(J(2-(J(1C 2
( 2.16)
Then, if 0( 2 exceeds 0( 1 by more than 1, 11 is less than or equal to 1, even

when c has a value of 1. Johnson has shown that as c approaches 1, sliding
causes (0( 2 - (J(d and, hence, 11 also to approach 1. It is then possible to
produce friction coefficients greater than 1, not by further junction growth,
but by internal shear in the junction to form a chip or prow with adhesive
transfer to the opposing surface.
More recently, Halling has suggested that ()( is not a constant but is
itself a function of c. He considers the case of a surface element subjected to
a normal stress p and a shear stress -r, and kept in equilibrium by lateral
internal stresses a 1 p and a 2 p, as shown in Figure 2.7(a). In the general case,
where 0 < c < 1, the plane strain situation can then be represented by the
Mohr's stress circle of Figure 2.7(b ). Then, from the geometry of this figure,
it can be seen that
and
Therefore,
p2 + O(C2k2 = O(k2
where
(2.17)
FRICTION 41
(a)
a,
(b)
a,p
Figure 2. 7 Stresses on a surface element in plane strain
Hence,
fJ. = ckjp = c/[1X(1- c 2 )]! (2.18)
It can be seen that Equation (2.18) is of the same form as Equation (2.14),
but the constant IX has been replaced by the variable given by Equation (2.17 ).
The values of IX for the extreme cases of c = 0 and c = 1 can be calculated
from the Mohr's circles shown in Figures 2.8(a) and 2.8(b), respectively. It
can be seen,' for a rigid-plastic material in plane strain, that when r = 0,
t.
a 1 = a 2 = while when r = k, a 1 = a 2 = 1. Inserting these values into
Equation (2.17) shows that IX has a minimum value of 36 when c = 0 and
tends to infinity as c tends to 1. The variation of IX between these limiting
conditions is not known, and the equations would need to be modified for
the more realistic case of general triaxial stress and for either fully elastic
deformation or plastic deformation with work hardening. However, this
treatment does bring out the important point that a friction coefficient will
depend on both the interfacial shear strength, as represented by ck, and the
deformation characteristics of the material.
42 TRIBOLOGY
p=H
(a)
(b)
Figure 2.8 Mohr's circles for extreme values of surface shear stress: (a) c = 0;
(b) c = 1
2.4.3 Ploughing
The effect of a hard asperity ploughing a groove in a softer material is

illustrated here for the example of a hard conical asperity of semiangle 0, as
shown in Figure 2.9. During rubbing, only the front surface of the asperity
is in contact with the softer surface. Therefore, the normal load, W;, which
W;
Figure 2.9 Ploughing of a soft surface by a conical asperity

FRICTION 43
is supported by the horizontal projection of the asperity contact, is given by

ltj = !nr 2 •H
and the friction force, Fi, which is supported by the vertical projection, is
given by
Fi=r·h·H
Therefore,
11 = FJltf = 2h/nr = 2 cot() /n (2.19)
Similar expressions can be calculated for other asperity shapes. For most
engineering surfaces, the asperity angles are large and the ploughing
component of friction due to asperity interaction is correspondingly small.
For example, Equation (2.17) shows that for an asperity angle of 85°, which
corresponds to an unusually rough surface, the ploughing coefficient of
friction would be only 0.056. However, abrasive material, including work-
hardened or oxidised wear debris may be very angular, and when significant
amounts of such material are present between the rubbing surfaces, the
ploughing component of friction may be much higher.
2.4.4 Deformation Theories
In the adhesion theory described above the normal and yield stresses on a
single asperity were assumed to be representative of the stresses on all
asperities. In deformation theories, it is recognised that the normal and shear
stresses on the asperities will vary during the lifetime of a junction. The
physical basis of the analysis is that, in the sliding of macroscopically flat
surfaces, motion is parallel to the interface and the separation of the surfaces
remains constant. This must be so, to maintain the area of contact at the
constant level which will support the constant normal load, and the significant
consequence of this is that the contacting asperities must deform to allow
movement to continue.
This type of theory was first advanced by Green (1955) and has been
extended by other workers. To illustrate the principles of such theories, the
treatment of Edwards and Halling (1968) is summarised here.
Edwards and Halling considered two wedge-shaped asperities of
semiangle ()moving as illustrated in Figure 2.10. By assuming that the solids
are ideally elastic-plastic and that the asperities deform plastically, it is
possible to calculate the instantaneous values of shearing force Fi and normal
load ltj over the complete life of the junction. This was done both for the
case of intimate contact, where the shear strength of the junction was taken
to be k, and for the case of asperities separated by a film of shear strength
ck, where c and k have the same meanings as in the Bowden and Tabor
theory. Calculations were made for various values of() and c. The variations
44 TRIBOLOGY
Figure 2.10 The idealised wedge-shaped asperities studied in the 'plastic interaction
theory'
of Fi and W; with displacement are shown in Figure 2.11 for an asperity

angle of 10° and two values of c. The coefficient of friction can then be
computed as the instantaneous value of :EF;/:EW;, where the summation is
made over all contacting asperities, or, alternatively, as the mean value of
F ;/ W; for a single pair of asperities over the full life of the junction. Calculated
values of Jl. for various values of c and () are shown in Figure 2.12.
\
\
,... _ _ c•0·8
/
"' "' ..............
..........
I '.._
I ' ...
',J
Figure 2.11 Variation of the normal force, P, and the friction force, F, throughout
the junction life for a junction of angle 10°
FRICTION 45
2<l
1·5
,,,
\
\
June lion
0·5 \ angle
'' 30"
--- --- --
.....
c ..........
li = .....,,(...,-1--c""2~)•
o•
1·0 0·5 0
c
Figure 2.12 Variation of Jl with c for various junction angles
The general equation for J.l. given by the Edwards and Halling theory is
J.l.=[a(l~c 2 )t+c/> ]/[l- a(l~c 2 ic/>] (2.20)
where cf> is a function of c and the geometry of the junction, and is equal to
zero when () is zero.
It is interesting to note that when cf> is zero, this equation reduces to a
form which is identical with the modified Bowden and Tabor theory
(Equation 2.14).
It is a characteristic of all asperity interaction theories of friction that
the energy consumed in plastic deformation increases with the sharpness of
the asperities. This is illustrated in Figure 2.12 for the Edwards and Halling
model, and similar trends can be shown to occur with, for example,
hemispherical asperities of decreasing curvature. In the case of elastic
deformation of asperities, there is no deformation component of friction, but
the real contact area of a junction and, hence, the force required to shear it
46 TRIBOLOGY
increase with asperity slope. These two facts account for the now widely
accepted observation that friction between macroscopically smooth surfaces
increases with the mean absolute surface slope.
2.5 COMPARISONS OF THEORIES OF SLIDING FRICTION
2.5.1 Energy Dissipation at Asperity Junctions
There are only three basic mechanisms which could possibly cause appreciable
energy dissipation at asperity junctions - i.e. fracture, elastic deformation
and plastic deformation.
New fracture surfaces will be created whenever an asperity junction is
broken, whether at the original interface or within one of the asperities, and
whenever a wear particle is formed. However, as will be shown in Chapter 3,
except in the case of very severe wear, asperities are normally subjected to
many thousands of stress cycles before a wear particle is formed. Therefore,
the energy losses associated with wear particle formation will, in general, be
negligible, and the contribution of the fracture mechanism to friction will be
largely determined by the energy dissipated in breaking the adhered junctions,
either at the original interface or within one of the contacting asperities. For
brittle materials this will correspond to the energy required to create the two
new surfaces, but for ductile metals, which we are considering here, it will
be almost entirely determined by the energy consumed in plastic deformation
prior to fracture.
Elastic hysteresis and viscoelastic losses can be dominant factors in the
friction of elastomers and polymers, and they will be discussed later in this
chapter. However, the above discussion of friction theories was restricted
to ideal elastic-plastic solids and there can be no energy loss during the
elastic deformation of such materials. As many metals are almost perfectly
elastic, it is normally assumed that elastic hysteresis can also be ignored as
a contributory factor to the observed sliding friction of metals.
Therefore, we must conclude that plastic deformation is the principal
cause of metallic friction. At first sight this is a rather surprising conclusion,
since, in general, the outermost asperities will always undergo fully plastic
deformation during, at most, the first few contacts; they will then have
flattened sufficiently to remain nominally elastic. However, on subsequent
contact the required elastic stress will be similar to that attained on the first
contact - i.e. close to the yield pressure. If we regard the asperities as having
spherical geometries, we can see that we now have highly loaded Hertzian
contacts, and the surface pressure will be sufficiently high for the maximum
shear stress to exceed the critical shear stress for the onset of plasticity. Under
normal loading only, the critical shear stress will occur below the surface
and the plastic zone will be constrained. This will also be the case if the local
FRICTION 47
friction coefficient - i.e. the ratio between shear stress and normal stress on
the asperity - is less than 0.25. In these cases plastic deformation must be
very limited, but the energy consumption will still make some contribution
to the total friction force. However, if J1 is greater than 0.25, the plastic zone
will be at the surface, the deformation will not be constrained and it is likely
that a more significant amount of plastic work will be done on each pass of
an opposing asperity.
Thus, even when the contact geometry is such that deformation
would be expected to be elastic, there may be a significant plastic
deformation contribution to the observed frictional work, and, as stated in
sub-section 2.4.2, this must be the case when the interfacial shear strength
is high, even when the direct stress is very low.
The above arguments will apply to some extent to any asperity which
is subjected to an element of plastic deformation (i.e. to a normal pressure
greater than 0.6H), so that a high proportion of the contacting asperities
will be subjected to the repeated plastic strain described above.
To assess the relative importance of the various contributions to the
total friction, we remember that it takes many thousands of contacts to
produce a wear particle. Only very few of the early contacts involve gross
plastic deformation and only one - the final one - involves fracture, so that
at any time the overwhelming majority of asperity contacts will be nominally
elastic, with about half ofthem experiencing stresses close to the yield pressure.
It is inconceivable that the very small fraction of fully plastic contacts will
make a significant contribution to the overall friction force, so the deformation
contribution to friction will arise almost entirely from localised plastic
deformation of the nominally elastic contacts.
The energy dissipation mechanism described above is identical with that
described by Johnson for rolling friction under high stresses. We shall refer
to it again in Section 2.8, when we discuss rolling friction.
2.5.2 The Interaction of Deformation and Adhesion
As we know, when any two surfaces are placed together under a normal
load, the surfaces approach each other until the elastic reactive force at the
interface is just sufficient to support the normal load. During sliding, a similar
situation obtains, with the reactive force being required to support both the
normal and the tangential forces.
Force equilibrium demands that each ofthe asperities forming a junction
must be subjected to the same stress, and behaviour during sliding is therefore
very dependent on the relative yield pressures of the two materials. If the
materials are identical, then they will be strained in the same way, either
elastically or plastically, depending on their degree of overlap. However, if
their yield pressures are significantly different, the softer material may deform
plastically but the deformation of the harder one must be nominally elastic.
48 TRIBOLOGY
As deformation continues, the softer material may work harden; but as long
as its current yield pressure is lower than that of the opposing surface, the
deformation of the harder material must remain nominally elastic.
It is important to note that the softer material may sometimes cause
localised plastic deformation in the harder material. This is because the
pressure to cause full plasticity of an asperity is approximately twice as great
as that required for initiation of subsurface plasticity. Thus, an interfacial
pressure which is sufficiently high to cause full plasticity of the softer asperity
will also initiate plastic deformation in the harder asperity as long as the
hardnesses do not differ by a factor of more than about 2. This will be
seen to be particularly important when we come to discuss fatigue wear in
Chapter 3.
If one material is very much harder than the other, the condition of
stress similarity can only be obtained if the harder asperity penetrates the
softer surface, with a consequent ploughing contribution to the friction force.
Ploughing is the extreme form of plastic deformation: in what follows we
shall consider the less extreme forms of plastic deformation during asperity
contact.
For clarity, we shall consider the total life of an asperity junction to
comprise three stages: formation, shearing and separation. This is a rather
unrealistic picture, because the surfaces are macroscopically in continuous
relative motion; therefore, the processes of formation and shear will overlap,
as will the processes of shear and separation. Nevertheless, it is a useful way
of illustrating the way in which adhesion and deformation interact. The three
stages can be characterised by the sense of the direct stress normal to the
interface: during junction formation the direct stress is compressive; during
shear it progressively falls to zero; and during separation, in the presence of
adhesion, it becomes tensile.
During junction formation the only significant process of energy
dissipation will be plastic deformation of one or both asperities; the actual
amount of energy dissipated will depend on the yield pressures and
work-hardening characteristics of the plastically deformed asperities, and on
the plastic strains which occur. This plastic deformation will result from
the combined effects of the normal and shear stresses, as described in
sub-section 2.4.2, and must continue until the total deformation is sufficient
to allow the asperities to pass at constant surface separation. The plastic
deformation can result in significant junction growth, and may also disrupt
any protective surface films, so that strong adhesive bonds are formed over
part of the junction.
If the interface is weak, sliding may occur at the interface and the only
adhesive contribution to the friction will be the small force required to shear
this weak interface. However, if the interface is strong, owing to the formation
of strong adhesive bonds, plastic deformation may continue, initially under
compression, then in pure shear and finally, as the asperities begin to separate,
FRICTION 49
under an increasing tensile force. As pointed out in sub-section 2.4.2, this

plastic shear will occur when the interface is strong, even if the normal stress
is well within the elastic range.
It is during junction separation that the surface pull-off force and pull-off
work become important and may make an important adhesive contribution
to the total friction.
It is interesting to note that interfaces having the same pull-off force
may have very different values of pull-off work. One reason for this is
illustrated in Figure 2.13, where the stress-strain curves to failure are plotted
for a brittle junction and a ductile junction of the same strength - i.e. having
the same pull-off force. It can be seen that far more energy will be consumed
in breaking the ductile junction and, although both junctions will break at
the same force, the ductile junction must make a greater contribution to the
observed friction force. This apparent contradiction is resolved when it is
realised that the ductile junction will persist over a much longer distance
than the brittle one, and therefore its integrated effect on the total observed
force will be comparatively great.
It should be clear from the above discussion that the adhesion and
deformation theories of friction are not competitive but complementary.
There will usually be both adhesion and deformation contributions to any
observed friction force. Furthermore, the energy consumed in deformation
during junction separation will depend on the strength of the adhesive bond,
so the two components are not simply additive, but interactive.
Finally, the products of wear are often abrasive; they may comprise
oxide or highly work-hardened metal, or mixtures of both, and they may
exist as particles trapped between the surfaces or transferred patches on the
rubbing surfaces. In any case, they may cause continuous roughening of one
or both surfaces, so that the surfaces never settle down into the nominally
elastic condition described above; and if they are very angular, they may
give a very significant ploughing contribution to the total friction force.
Ductile
Figure 2.13 Comparison of pull-off work for brittle and ductile materials having
the same pull-off force
50 TRIBOLOGY
The above description of possible forms of behaviour illustrates why it

may never be possible to predict accurately the tribological behaviour of a
previously untried pair of materials, particularly in the case of unlubricated
sliding of metals.
In recent years work has begun on explaining the separation of friction
junctions using the ideas of fracture mechanics. This work is very interesting
and may well be fruitful, but it is still in its early stages and is beyond the
scope of the present chapter.
2.6 FRICTION OF NON-METALLIC MATERIALS
2.6.1 Elastomers
Elastomers consist of long-chain organic molecules linked at various points

along their lengths by strong chemical bonds, the elastic modulus increasing
with the number of cross-linking bonds. The free surface of an elastomer
consists almost entirely of fully saturated hydrocarbon groups, so that the
surface energy is very small. As all the primary interatomic bonds in the
surface are fully saturated, the forces between elastomer surfaces are limited
to weak van der Waals bonds, rather than the strong cross-linking bonds
which exist within the bulk of the material. Thus, in adhesion experiments
separation always occurs at the original interface. However, although the
thermodynamic work of adhesion, as defined in Equation (2.5), is very low,
the pull-off work can be high. This is because there are substantial viscoelastic
hysteresis losses as an elastomer is deformed, and the contribution of these
losses to the observed work of adhesion is similar to that described for plastic
deformation of metals. However, unlike metals, the losses in elastomers are
strongly dependent on both temperature and strain rate.
The pull-off force is also critically dependent on the surface roughness
of the counterface, and decreases rapidly as the roughness increases. In
essence, the pull-off force is the resultant of two competing forces: an attractive
force due to the work of adhesion and a repulsive force due to elastic reaction
of the elastomer against penetration by the higher asperities. It can be shown
that the ratio between these forces becomes very small at surface roughnesses
as low as 1 J.lm RMS.
If a clean smooth specimen of elastomer is slid against a clean smooth
hard surface, it is found that the friction may be high but there is no transfer
of material between the surfaces. Observations of sliding against transparent
substrates show that there is no true sliding at the interface; rather, a ruck
forms towards one edge of the contact and travels through the interface,
thus resembling the movement of a caterpillar. Continuous movement
proceeds through a sucession of these waves, which are known as Schallamach
waves, so that the frictional work is the net effect of the work of adhesion
and the pull-off work (Briggs and Briscoe, 1975).
FRICTION 51
As the degree of cross-linking in an elastomer increases, both the elastic

modulus and the mechanical strength increase very rapidly, while the
interfacial forces, which are still dominated by weak van der Waals forces,
remain approximately constant. Under these conditions the Schallamach
waves decrease in size, and it is probable that at some critical point true
sliding begins.
Under unlubricated conditions the friction is dominated by the
viscoelastic effects associated with the movement of Schallamach waves, the
deformation by counterface asperities having little effect. However, in the
presence of lubricant the adhesion and the associated viscoelastic losses
become very low and the friction is largely dependent on deformation losses.
This has led to the use of rubbers of high hysteresis loss in the manufacture
of tyre treads, to increase skid resistance on wet roads (Bowden and Tabor,
1964).
2.6.2 Polymers
As with elastomers, it has been observed that when a polymer adheres to a

hard counterface, the pull-off work is proportional to the thermodynamic
work of adhesion, but very much greater. The high values of pull-off work
are also associated with deformation losses, but in this case the losses are
due to flow of the polymer rather than hysteresis losses.
As with metals, the friction of polymers involves both adhesion and
deformation, but in the case of polymers there is little or no junction growth.
Also, as with elastomers, the deformation losses during sliding are much
more dependent on temperature and strain rate than are those in metals.
Polymers generally have low melting or softening points and poor
thermal conductivity, so, when there is significant frictional heating, the
surface layer usually softens or melts. The friction then largely arises from
viscous losses in the surface layers. At low sliding speeds, where frictional
heating is negligible, the frictional characteristics depend on the roughness
of the counterface.
When the counterface is smooth, polymers may be divided into two
groups. With strongly cross-linked polymers, and glassy polymers below
their glass-transition temperatures, the internal bonds within the polymer
are usually stronger than the interfacial bonds, so that sliding takes place at
the original interface; with softer, semicrystalline polymers, such as PTFE
and high-density polyethylene, a thin film or highly oriented polymer is
drawn out and transferred to the counterface, and subsequent sliding takes
place on this transfer film. In the latter case, if the sliding speed exceeds a
low critical value, it becomes easier to rupture the polymer than to draw
out a film; the transferred material and loose wear debris then become lumpy
and the friction increases.
When the counterface is rough, the asperities can penetrate the polymer
and sliding causes the polymer to shear at an angle to the interface. The
52 TRIBOLOGY
shear angle correlates with the energy-to-rupture of the polymer, and the
frictional work is largely accounted for by the associated energy input.
2.6.3 Ceramics
The adhesion of ceramics to each other or to metal counterfaces is not well

understood.
The type of primary bonding within the ceramic is clearly important:
if it is ionic, then the bonding across the interface will be Coulombic and,
in principle, amenable to calculation; however, if the bonding within the
ceramic is covalent, it is by no means certain what, if any, valence bonds
will be available for interfacial bonding.
Furthermore, as ceramics are elastic solids, they are subjected to
the same effect of surface roughness as that described for elastomers in
sub-section 2.6.1 -i.e. there is a repulsive force due to elastic reaction of the
ceramic against indenting asperities. Thus, the adhesion of ceramics to each
other would be expected to be negligible at surface roughnesses as low as a
few atomic spacings. However, surface roughness would not have the same
effect on the adhesion between a ceramic and a ductile metal, as the effect
of asperity penetration would be accommodated by plastic deformation of
the metal.
The same problems arise in trying to understand frictional behaviour.
If all surface valences are saturated, by either internal bonding or bonding
to adsorbed films, it is likely that sliding will take place at the interface and
both adhesion and friction will be low. However, it is possible that the sliding
process itself could create interfacial bonds which would cause the adhesion
and friction to increase.
Despite this lack of fundamental understanding, ceramics and ceramic-
based materials are coming into increasing use for tribological applications.
It is, therefore, likely that the friction and wear of ceramics will become a
major field of research in the near future.
2.7 BOUNDARY LUBRICATION
So far in this chapter we have considered sliding between unlubricated

surfaces. Later chapters will consider different types of full fluid film
lubrication, where sliding behaviour is determined largely by the lubricant
viscosity, and elastohydrodynamic lubrication, where sliding behaviour is
determined by lubricant properties and the elasticities of the opposing
surfaces.
Boundary lubrication is defined in 0 ECD ( 1968) as 'a condition of
lubrication in which the friction and wear between two surfaces in relative
motion are determined by the properties of the surfaces, and by the properties
FRICTION 53
of the lubricant other than bulk viscosity'. It is, therefore, a rather ill-defined
regime which covers the operating range between dry sliding and elasto-
hydrodynamic lubrication: a fluid film is present but it is not sufficiently
thick to prevent contact between the higher asperities.
The situation can be characterised by the value of the A ratio, hfa,
where h is the separation of the mean planes of the two surfaces and a is
the composite roughness (af +ant. As we saw in Chapter 1, for surfaces
having Gaussian or near-Gaussian asperity height distributions, surface
separation is virtually complete if A~ 3. For values of A less than 3, some
asperity contact will occur, and both friction and wear will be expected to
increase as A decreases.
It should be clear from the earlier parts of this chapter that, if we wish
to ameliorate the effects of this asperity contact, we need to interpose a film
of low shear strength between the sliding surfaces. Therefore, oils intended
for boundary lubrication contain additives which will deposit such films on
the surfaces.
The most common type of additive is the aliphatic 'fatty' acid of the
general formula CnHln+ 1 COOH, a typical example being stearic acid. Each
molecule of such an acid consists of a long covalently bonded hydrocarbon
chain with a single reactive acid group at one end. The acid groups of these
molecules react with surface metal atoms to form a strongly bonded metallic
soap, and the hydrocarbon chains align themselves normal to the surface,
thus providing a very effective barrier to metal-metal contact. The process
is illustrated schematically in Figure 2.14.
H
I
H- C -H
H- C- H
I
I
I
I
H- C- H
0 # " OH
(a)
(b)
Figure 2.14 Boundary lubrication by long-chain fatty acid molecules: (a) molecular
structure; (b) surface protection by reacted layer
54 TRIBOLOGY
These films can provide excellent protection at high pressures and at

high sliding velocities, and, as long as the oil does not become depleted of
the additive, the soap films should form at any area of exposed metal. The
factors which limit the use of these films are not well understood, but it is
probable that the most critical factor is the local temperature, which, in turn,
depends on both the bulk temperature and the rate of frictional heating.
When the operating conditions become too severe for these boundary
lubricants, recourse must be had to extreme pressure (EP) additives. These
are aggressively reactive compounds, usually containing sulphur, chlorine or
phosphorus, which, again, form protective reaction products at exposed metal
surfaces. The components are so aggressive that they cause significant rates
of metal removal by corrosion (corrosive wear), and the additive should be
selected with care to ensure that its net effect is beneficial.
2.8 FRICTION IN ROLLING CONTACT
Lubricated rolling contact will be discussed in detail in Chapter 7. It is the

purpose of this section to describe frictional effects during rolling. These
effects have been described in detail by Johnson (1985) and only the more
salient points will be covered here.
To describe the various effects which can arise, we shall use the coordinate
system adopted by Johnson and illustrated in Figure 2.15. This figure shows
two non-conforming bodies making contact under a negligibly small force,
so that they touch at a single point, 0. This point is taken as the origin of
a rectangular coordinate system Oxyz, with the axis Oz coinciding with the
common normal to the two surfaces at 0, and the axes Ox and Oy being
chosen, where possible, to coincide with axes of symmetry of the surface
profiles. In the special case of line contact between two cylinders, we choose
the y axis to be parallel to the axes of the cylinders.
Common
normal
Figure 2.15 Non-conforming surfaces in contact at 0

FRICTION 55
The frame of reference as defined above moves with the linear velocity
of the contact point 0 and rotates with an angular velocity which maintains
its orientation relative to the tangent plane and the common normal at 0.
We can then state the following definitions.
Sliding is the relative linear velocity between the two surfaces at 0; as
there can be no component of relative velocity along the normal, sliding is
always in the tangent plane.
Rolling is the relative angular velocity between the two bodies about an
axis lying in the tangent plane.
Spin is the relative angular velocity between the two bodies about the
common normal through 0.
Any relative motion between two contacting bodies can be regarded as
a combination of sliding, rolling and spin. In the previous sections we have
been concerned only with sliding, but in the remainder of this section we
shall describe frictional effects due to combinations of sliding, rolling and spin.
To understand such effects, we must also understand the stresses which
can arise in the contacting bodies due to the forces acting at the interface.
In Chapter 1 we considered in some detail the stresses which arise at
contacting non-conforming surfaces due to normal forces acting at the
interface, but we referred only in passing to the stresses which arise due to
tangential forces. We now need to consider in rather more detail the combined
effects of normal and tangential forces.
The resultant force transmitted across any interface can always be
resolved into a normal force, P, acting along the common normal and a
tangential force, Q, acting in the tangential plane. The tangential force Q is
reacted by frictional resistance and its magnitude must therefore be less than
or equal to Jl-P. When Q attains the limiting value of Jl-P, gross sliding occurs
and the direction of Q is directly opposed to the direction of sliding.
The interfacial forces lead to a contact area of finite size, and the
distribution of forces over this area can, therefore, lead to the transmission
of a moment across the interface. The components of this moment about
axes lying in the tangent plane are known as rolling moments, while the
component about the common normal is known as the spin moment.
Foil owing Johnson ( 1985 ), we define free rolling as rolling motion in
which the tangential force Q is zero, and tractive rolling as rolling motion
in which the tangential force Q is non-zero but less than the limiting value
of Ji-P at which gross sliding will occur.
The interfacial forces and moments described above are transmitted
across the interface by surface tractions, the normal traction, which is simply
the contact pressure, being denoted by p and the tangential traction by q.
In general, both p and q vary over the contact area, so they can be written,
respectively, as p(x, y) and q(x, y), where (x, y) denotes position in the tangent
plane. It is important to realise that tangential tractions can be present even
56 TRIBOLOGY
when the tangential force Q is zero; in such cases it is simply necessary that
the tractions integrated over the contact area should sum to zero.
As p(x,y) andq(x,y) have different distributions over the contact area,
the ratio q(x, y)/p(x, y) also varies with position within the contact area.
As a result, two different contact conditions may obtain: ( 1) wherever
q( x, y) < Jl.P( x, y ), local sliding is prevented by friction and the corresponding
regions of the interface are said to be regions of 'stick'; (2) wherever
q( x, y) = Jl.P( x, y ), local sliding is possible and the corresponding regions are
said to be regions of 'slip' or 'microslip'.
A difference between the tangential strains in the two bodies in the
'stick' area leads to a small apparent slip known as creep. Creep can be
understood by considering the example of a deformable cylinder rolling under
load on a relatively rigid surface. If the tangential strain in the cylinder is
tensile, the cylinder is stretched in the 'stick' area. The cylinder then behaves
as though it has an enlarged circumference, and therefore in one revolution
it moves forward by a distance greater than its actual circumference. The
ratio between the distance traversed in one revolution and the actual,
unstretched circumference is known as the creep ratio.
The actual location of the boundaries between the areas of 'stick' and
'slip' is quite difficult and need not concern us here. Readers requiring a
fuller treatment of such problems are referred to Johnson ( 1985). In the
following sections we shall simply describe in qualitative terms the locations
and sizes of the zones of stick and slip for various rolling situations, starting
with the most simple situation of free rolling between identical elastic bodies
and progressively introducing more variables.
2.8.1 Free Rolling between Identical Elastic Bodies
The simplest form of free rolling occurs between two bodies which have the
same elastic properties and are geometrically identical. When such bodies
roll freely under the action of a normal force, no tangential tractions or slip
can occur and the contact stresses and deformations can be calculated from
the Hertz equations for normal loading, which are given in Chapter 1.
2.8.2 Tractive Rolling of Non-conforming Bodies
We shall largely confine our attention to tractive rolling between bodies

having the same elastic properties. However, before considering rolling, it is
useful to discuss briefly the tractions which arise at the interface between
two stationary cylinders when they are submitted to a tangential force, Q,
which is less than the force required to cause gross sliding. In this case, within
the contact zone, there is a central 'stick' area, where q(x, y) < Jlp(x, y), and
two outer 'slip' areas, where q(x, y) = Jlp(x, y). The coexistence of zones of
stick and slip is possible because of the deformable nature of the materials.
FRICTION 57
Thus, if P is kept constant and Q is increased steadily from zero, microslip

begins at the edges of the contact zone and spreads inwards as Q increases;
when Q reaches the limiting value, Jl.P, the two zones of slip meet in the
centre and gross sliding occurs.
The conditions which obtain in tractive rolling of elastically similar
cylinders are qualitatively similar to those described above, with regions of
both 'stick' and 'slip' as long as Q < Jl.P. However, if we examine the position
in detail, we find that, in the forward region of slip, the slip would be in the
same direction as the tangential traction. This is inadmissible, as it would
contravene the basic law that friction must act to oppose the motion.
Therefore, in practice, the boundary of the stick region coincides with the
leading edge of the contact area and the slip is confined to a single zone at
the trailing edge. Again, as Q increases, this slip zone extends forward until,
when Q = Jl.P, the whole area becomes a slip zone and gross sliding occurs.
The problem of rolling between general three-dimensional bodies
becomes mathematically very complex. However, the technologically important
problem of rolling of spheres has been studied in some detail, and the situation
is found to be qualitatively similar to that described for rolling cylinders,
with a single stick zone at the leading edge of the contact and slip over the
remainder.
2.8.3 Sphere Rolling in a Grooved Track
So far we have considered situations where the contact area lies substantially
in a single plane. However, in many engineering situations, particularly ball
bearings of all types, this situation does not obtain. For a ball rolling along
a grooved track, as illustrated in Figure 2.16, the contact area will be an
A A
Figure 2.16 Heathcote-type slip

58 TRIBOLOGY
ellipse and, provided that the angle pis less than 30°, the dimensions of the
ellipse are still predicted with reasonable accuracy by the Hertz equations.
If we now consider the situation when the ball has rolled forward through
one complete revolution, we can see that in the centre of the groove the ball
will have measured out its full diameter along the track, while points on the
ball at the edges of the contact zone will have measured out a smaller distance
corresponding to the circumference of the circle of radius ri. These
differences can be accommodated by the ball slipping in its track, and we
might expect the contact area to consist of three zones of slip, as shown in
Figure 2.16- i.e. a single central zone of backward slip and two outer zones
of forward slip. In this situation the axis AA, which passes through the two
boundaries between the regions offorward and reverse slip, can be considered
to be the instantaneous axis of rotation. This situation, which is known as
Heathcote slip, was first put forward by Heathcote (1921) as it is described
above. However, Johnson (1985) has shown that some of the tendency to
slip can be accommodated by elastic deformation of the contact areas.
For problems involving both rolling and spin, as in thrust ball bearings,
and for problems of materials having dissimilar elastic properties, the
solutions become much more complex than those described above. The
problems have been treated in detail by Johnson ( 1985), but are beyond the
scope of this book.
2.8.4 Rolling Resistance
In all rolling contact situations there is some resistance to rolling, although

the magnitude of the resistance is usually very much lower than the resistance
to sliding. As with sliding friction, this resistance can arise from only two
sources: adhesive losses, due to resistance to sliding in the slip regions, and
deformation losses during stress cycling of the contacting surfaces.
It has been found that the rolling resistance in any free-rolling situation
is influenced to only a small extent by the presence of lubricants, which
would act to lower the adhesive contribution to the rolling resistance, and
it can, therefore, be concluded that the contribution of adhesion is very small
and that resistance to free-rolling arises mainly from deformation losses. The
same is usually true in tractive rolling and rolling in grooved tracks, although,
as the tractive force approaches the limiting value or if the spin moment
becomes high, adhesive losses become more significant.
In considering deformation losses we shall consider separately losses in
elastic contacts and losses in the presence of plastic deformation.
As an example of an elastic rolling contact, we consider a cylinder rolling
freely on a plane under a normal load P per unit length, as illustrated in
Figure 2.17. The local pressure, p, in the contact zone, and the contact
semiwidth, a, are known from the Hertz equations. Considering a strip of
width dx at point x, we can see that, during forward compression of the
FRICTION 59
Expons1on
f- 2a -j
Figure 2.17
contacting material, the normal force per unit length of the roller will be
p dx, and this will exert a resistive moment about the centre of the contact
zone of magnitude px dx. Thus, the total moment exerted by all the elements
on the front half of the roller will be
M =I: px dx = 2Pa/3n
If the cylinder rolls forward through a distance x, the elastic work done by
this couple is given by
cp = Mx/ R = 2Pax/3nR
Under ideal conditions the cylinder would be subjected to an equal and
opposite moment arising from the pressure distribution in the rear half of
the contact. However, owing to elastic hysteresis losses, not all the work is
recovered and there is some dissipation of energy. If we define the hysteresis
loss by a coefficient e, we can see that the energy dissipated in rolling a
distance x will be scp, and if this loss represents the rolling resistance, it
follows that
Fx = scp = s·2Pax/3nR
where F is the force required to overcome the resistance. By analogy with
sliding friction, the ratio F I P can be defined as the coefficient of rolling
resistance, A, where
A= 2ea/3nR (2.21)
Using a method similar to the above, it can be shown that the coefficient of
rolling resistance for a sphere rolling on a plane is given by
A= 3ea/16R (2.22)
60 TRIBOLOGY
Equation (2.20) can also be used for elliptical contacts if a is taken as the
half-width of the contact in the direction of rolling.
If the contact pressure on first contact exceeds a critical value, then, as
described in sub-section 1.5.1, some subsurface plastic deformation takes
place and the energy dissipated in such deformation can be substantially
higher than that due to elastic hysteresis. However, almost all rolling contacts
are subjected to repeated stress cycles and, owing to the accumulation of
residual stresses, the degree of plastic deformation can decrease with repeated
cycling until, after a relatively small number of cycles, the deformation has
become fully elastic. This process is referred to as shakedown. However, if
the contact stress exceeds a second critical value, which is higher than the
value for first yield, shakedown does not occur and there is some plastic
deformation on each stress cycle. Johnson ( 1985) has calculated the ratio
between the maximum shakedown pressure (p 0 ). and the pressure for first
yield (p 0 )y for a variety of situations. For free rolling of an elastic cylinder
on an elastic-perfectly plastic half-space, this ratio is 1.29, and in the presence
of tangential tractions the ratio decreases. For rolling of a sphere on an
elastic-plastic half-space, the ratio is 2.2.
When the load exceeds the shakedown limit, some plastic deformation
occurs on each cycle and the surface layers are displaced forward relative to
the deeper layers. The rolling resistance due to this effect can then be much
greater than that due to elastic hysteresis. At even higher loads, when the
plastic zone extends to the surface, the rolling resistance increases to even
higher values.
2.8.5 Tyre-Road Contacts
An important example of rolling wheels subjected to tangential tractions

occurs in the wheels of tyres and locomotives. Whether such wheels are b.eing
driven or braked, control of the vehicle requires that limiting slip (skidding)
does not occur and that the slip behaviour is limited to zones of microslip.
For locomotive wheels the previous sections describe the microslip behaviour,
but for pneumatic-tyred wheels the situation is more complex and cannot
be treated in detail here. However, the principles of the behaviour are given
below.
We consider the contact patch between a tyre and a road, as shown in
Figure 2.18. We can represent the total friction traction by a single force
Q < p.P, acting through some point such as X. This force through X can
then be resolved into three components acting at the origin 0:
( 1) The force Qx, the driving or braking traction which produces the
type of microslip described earlier.
(2) The transverse tangential traction QY, which will produce transverse
microslip; this is the cornering force which produces changes in direction
of the vehicle.
FRICTION 61
M, I
u~--f--o,
X
0 X -
Figure 2.18 Resolution of the frictional effects in the tyre-road contact
(3) The torque MZ' the self-aligning torque which acts to straighten the
course of the vehicle.
When the vehicle is travelling on a straight line, both QY and M z are
zero, but if we consider the car being driven round a left-hand turn, the force
system and the corresponding microslip velocities are as shown in Figure 2.19.
Now the path of motion lags behind the centre line of the tyre by the slip
angle (), which can be interpreted as the angular microslip due to the
self-aligning torque Mz. The relationships between QY, Mz and () are then
shown in Figure 2.20 (Gough, 1954). In the region ABC, the self-aligning
torque and the slip angle increase with the cornering force. In the region CD
there is little change in the self-aligning torque as the cornering force and
the slip angle increase, and the feel of the steering is therefore lost. In the
Vehicle
polh
Figure 2.19
62 TRIBOLOGY
A~ Position depends on p.P

ComeriiiQ - 4"'_/
fc:wce ----~
a, £ ' \.
\ Effect of
~ increasing P
I
I
I '
c ,'
"
Self-aligning torque ~
Figure 2.20 Relation between the major variables during cornering- Gough plot
region DE, which approaches the point of skidding, there is a reversal in the
feel of the steering as the car drifts around the corner. Although the effects
of the normal load and the friction coefficient are modest in the region ABC,
they become increasingly important as the point E is approached.
2.9 THERMAL EFFECTS IN SLIDING FRICTION
Almost all frictional energy is dissipated in the form of heat. As this heat
release is a continuous process, temperature gradients will be set up in the
contacting bodies, with the highest temperature at the heat source- i.e. the
contact surface. The overall temperature rise can, in principle, be calculated
from the rate of input of frictional energy and the thermal properties of the
contacting bodies and the environment; a theoretical treatment and applications
to many practical problems have been discussed by Carslaw and Jaeger
(1959). However, as we now know, the bodies make contact only at the
asperity tips, and all the heat must be generated at these small areas of
contact. As a result, the local temperatures may be instantaneously very
much higher than the bulk temperatures of the bodies. These high temperatures
of short duration are known as 'flash temperatures', and they can be very
important in changing, for example, the microstructures of alloys or the rate
of oxidation and the types of oxide formed. A full analysis of the magnitudes
of flash temperatures has been made by Jaeger (1942), but this analysis is
very complicated and we shall summarise the much simpler treatment given
by Archard ( 1959).
The temperature effects will depend on whether a body is stationary or
moving with respect to the heat source. Archard (1959) based his analysis
of the various problems on the model shown in Figure 2.21, with a
proturberance on the surface of body B forming a circular area of contact
A = na 2 and moving with velocity V over the flat surface of body C. Body B,
therefore, receives heat from a stationary heat source and body C from a
FRICTION 63
(B)
Figure 2.21
moving heat source. Then, for the two bodies, Q8 and Qc are the quantities
of heat supplied per second, K 8 and Kc are the thermal conductivities, c 8
and cc are the specific heats, p 8 and Pc are the densities, and XB and Xc
(given by X = K I pc) are the thermal diffusivities. (]mean is the mean temperature
of the contact area, taking the temperature of a point far removed from the
contact as zero.
Then Archard derives the following temperature rises for the stationary
heat source (body B) and a slow-moving heat source (body C) to be,
respectively,
( 2.23)
and
(}mean = Qc/ 4aKc ( 2.24)
At higher speeds Equation (2.24) does not apply, as there is insufficient time
for the temperature distribution of a stationary contact to be established in
body C. The speed at which this occurs is determined by the dimensionless
parameter
L= Va/2x (2.25)
Equation (2.24) applies to a reasonable approximation for L < 0.1. For L > 5
the temperature rise is given by
(]mean= (0.31Q/ Kca)·(xc/Va)t (2.26)
We can define a parameter N = nq/ pc V, where q is the rate of heat supply
per unit area. Then, for a stationary heat source or a very slow-moving heat
source (L < 0.1 ),
(]mean= 0.5NL (2.27)
64 TRIBOLOGY
10
10-1
10 10 2 103 104
Figure 2.22 Average temperature as a function of the dimensionless parameter L:

A, stationary heat source; B, moving heat source
while for a fast-moving heat source (L> 5)

(}mean= 0.435NL! ( 2.28)
For speeds in the intermediate range (0.1 < L < 5),
(}mean = 0.51Y.N L ( 2.29)
where IY. ranges from about 0.85 at L = 0.1 to about 0.35 at L = 5. These
results can be summarised in graphs of (}mean/Nasa function of L, as shown
in Figure 2.22.
To apply the results shown in this figure, the proportion of the total
heat which is supplied to each of the bodies must be taken into account. To
do this, the flash temperature is calculated for each body on the assumption
that all the heat flows into it, using the equation which is appropriate for
the speed ofthe body. Then the true temperature is obtained from the equation
(2.30)
Archard applies this theory to show that the maximum flash temperature
which can be attained at a steel-steel contact, assuming that the total load
is borne by a single plastically deforming area, is of the order of 10 000 oc.
2.10 PROBLEMS
(In solving these problems, assume that the friction coefficient is given by
Jl = f.ladhesive + J.iploughing ·)
1 When a hard steel ball of diameter D is rubbed across a soft metal surface,
it ploughs out a groove of width d. If d « D, derive an expression for the
ploughing contribution to the friction coefficient.
[Jl = d/nD]
2 A hard steel ball is slid across a soft metal surface under two different
loads. In one case the friction coefficient is 0.22 and the groove width is
FRICTION 65
0.3 mm; in the other case the friction coefficient is 0.25 and the groove
width is 0.8 mm. Calculate the diameter of the slider and the adhesive
contribution to the coefficient of friction.
[d = 5.3 mm; Ji = 0.202]
3 Two hard conical sliders of sell}iangles 75o and 80° are slid across an
unlubricated metal surface and the measured friction coefficients are in
the ratio 11:10. The experiment is then repeated with the surfaces
lubricated and the measured friction coefficients are then in the ratio
13:10. Find the coefficients of adhesive friction in the two cases. If the
effect of the lubricant is to reduce the surface shear stress by a factor
of 3, what are the values of c and oc for the unlubricated surface?
[Jidry = 0.5938; JiJub = 0.0866; C = 0.905; OC = 13.96]
Chapter 3
Wear
3.1 INTRODUCTION
Wear is defined (OECD, 1968) as 'the progressive loss of substance from

the operating surface of a body occurring as a result of relative motion at
the surface'. This 'loss of substance' is of enormous economic importance.
As an example, it was estimated that its cost to the British economy in 1988
would probably be about £2000000000.
Unfortunately, as with friction, there is no reliable way of predicting
wear a priori, given the two materials to be used and the proposed rubbing
conditions. The reader should not find this surprising, as it arises from exactly
the same complexity in surface properties and interactions as we have already
described in Chapters 1 and 2. However, in the case of wear, the results of
making the wrong choice of materials are usually much more serious, owing
to the much greater variation in wear rates.
The coefficients of friction observed for most pairs of materials in normal
atmospheres are usually in the range 0.1-1, and in the presence of a little
lubricant the range is usually narrower than this. Furthermore, if the observed
friction coefficient is greater than expected, this usually manifests itself as no
more than a small and tolerable increase in power consumption. It is
sometimes important that the friction coefficient remain approximately
constant over a wide operating range (for example, in braking), but this is
a separate problem which is always solved empirically.
In contrast, wear rates can vary over many orders of magnitude, and
may change catastrophically because of a relatively small change in operating
conditions. Thus, the wrong choice of materials can have disastrous
consequences.
It is generally agreed that the only safe way to select materials to resist
wear in a particular situation is on the basis of tests - first, realistic bench
66
WEAR 67
tests, and then field tests. Fortunately, because of the economic importance
of wear, many results of such tests have already been published (see, for
example, Peterson and Winer, 1980), and potential users can refer to these
to make an initial, or sometimes final, choice of materials.
Although there is much current research which is increasing our
understanding of wear processes, it is probable that material selection will
remain empirical for the foreseeable future. It is true that there are theoretical
wear equations, some of which will be described later in this chapter, but
they always contain factors which have to be determined by experiment.
Furthermore, even these experimentally determined factors are inherently
imprecise, and in many cases the user must be satisfied if an observed wear
rate is within an order of magnitude of that predicted on the basis of empirical
data. The reason for this lack of precision can be described as the dependence
of wear rates on extreme values, rather than mean values, of experimental
variables. Examples are the domination of the surface chemistry by elements
which are only minor impurities in the bulk (Miyoshi and Buckley, 1985);
the presence of a few relatively hard inclusions in what is nominally a softer
material; the probable disproportionate importance of the heights of the few
outermost asperities of a hard material, rather than its average surface
roughness; and the presence or absence of contaminants from the environment,
particularly in the form of abrasive dust.
However, there are now generally accepted qualitative explanations of
the various processes underlying wear behaviour, although there is much
that remains to be clarified, particularly the relative importance of the various
processes. It would not be possible in a book of this length critically to
review the vast amount of work which has taken place in these areas, and
in discussing the relative importance of the various processes the views
expressed must necessarily be the subjective views of the author. No claim
is made for the infallibility of these views, but it is hoped that they do
represent a clear and consistent picture of wear, which will provide the
interested reader with a background on which to base further reading.
We shall initially concentrate on metals, as they have been subjected to
more detailed analysis than have other materials and have a wide range of
mechanical properties. In Chapter 4 we shall show how this basic understanding
can be extended to the behaviour of other types of materials, particularly
ceramics, polymers and elastomers, how it then allows an informed selection
of candidate pairs of material to be made and, where possible, how wear
behaviour can be predicted on the basis of published data.
3.2 TYPES OF WEAR
In a recent review paper Godfrey ( 1980) states that there are about twelve
kinds of wear, each with symptoms which must be recognized for the wear
68 TRIBOLOGY
mechanism to be diagnosed. He also states that some authors group several

of these types together and that in his opinion this is not justified, as each
type has its own mechanism and symptoms. Despite this stricture, for the
purpose of clarifying the processes underlying wear behaviour, we shall
initially divide these processes into only two types- mechanical and chemical
- and shall then follow the conventional approach of subdividing these into
fewer than twelve separate processes. Subsequently, we shall attempt to show
that this subdivision is somewhat arbitrary, particularly in the case of
mechanical wear, and that it is reasonable to think of a continuous spectrum
of mechanical wear processes. This is not to belittle the type of approach
taken by Godfrey, which may well be a valuable aid to diagnosing and
correcting a problem after wear has taken place, but we believe that the
approach described here will lead to clarity of thought in making a prior
selection of materials.
3.3 MECHANICAL WEAR PROCESSES
In his excellent 'Survey of possible wear mechanisms', Burwell ( 1957) listed

four major mechanisms: (1) adhesive wear; (2) abrasive wear; (3) surface
fatigue; (4) corrosive wear. He also included a fifth classification under the
heading 'Minor types of wear', which covered erosion and cavitation. In this
· section we shall discuss the three mechanisms of mechanical wear listed
above, and also the delamination theory of wear, which was put forward
some years after the publication of Burwell's review. Corrosive wear is
discussed separately in Section 3.6.
3.3.1 Adhesive Wear
The theory of adhesive wear has the same basis as the adhesion theory of
friction which was described in sub-section 2.3.1. We shall initially follow
the historical development of the theory, as it was originally proposed by
Bowden and Tabor and formulated as a semi-empirical law by Archard ( 1953),
but in Section 3.4 we shall examine the theory rather more critically and
show that some phenomena currently ascribed to an adhesive wear process
are explained rather more satisfactorily by a fatigue mechanism.
The basic mechanism underlying the theory has already been explained
in sub-section 2.3.1 - i.e. strong cold welds are formed at some asperity
junctions and these welds must be sheared for sliding to continue. The
amount ofwear then depends on where the junction is sheared: if shear takes
place at the original interface, then wear is zero, whereas, if shear takes place
away from the interface, a fragment of material is transferred from one surface
to the other. This transfer of material is observed in practice, normally from
the softer material to the harder, but occasionally from the harder to the
WEAR 69
softer. The theory at this stage has explained only material transfer and not
the formation ofloose wear debris; in the original theory it was simply stated
that subsequent rubbing causes some of the transferred particles to become
detached. We shall return to this point in Section 3.4.
Laws of Adhesive Wear
Here we follow the approach of Archard ( 1953 ), who first derived a theoretical
expression for the rate of adhesive wear.
It is assumed that the area of contact comprises a number of circular
contact spots each of radius a. The area of each contact spot is na 2 and,
assuming, as before, plastic deformation of an ideal elastic-plastic material,
each contact supports a load of na 2 H, where H is the yield pressure, or
hardness. The opposing surface will pass over the asperity in a sliding distance
2a and it is assumed that the wear fragment produced from each asperity is
hemispherical in shape and of volume 2na 3 /3.
Then the wear volume !5Q produced by one asperity contact in unit
sliding distance is given by
!5Q = (2na 3 /3 )/2a = na 2 /3
and the total wear volume, Q, per unit sliding distance is Q = nna 2 /3, where
n is the total number of contacts. But each contact supports a load of na 2 H,
so the total load is W = nna 2 H, and nna 2 = W/ H. Hence,
Q= W/3H (3.1 )
This equation suggests that there should be three 'laws' of adhesive wear.
( 1) The volume of material worn is proportional to the sliding distance.
(2) The volume of material worn is proportional to the load.
(3) The volume of material worn is inversely proportional to the hardness
of the softer material.
The first of these 'laws' is found to be true for a wide range of conditions.
The second 'law' is often true from very low loads up to some point at
which the rate of wear increases catastrophically. At the point where this
transition occurs the apparent pressure (the load divided by the apparent
area of contact) is approximately equal to or slightly less than one-third of
the hardness. We know from Chapter 1 that H/3 is approximately equal to
the bulk yield stress. Therefore, this transition clearly corresponds to the
point at which the whole surface begins to deform plastically, so that the
asperities can no longer be considered to be independent. We also know
from Chapter 1 that in the presence of frictional tractions the material will
yield at a lower normal stress, and we should, therefore, expect the wear
transition to occur at a pressure somewhat lower than H /3; this has been
confirmed experimentally (Amell et al., 1975).
70 TRIBOLOGY
The third 'law' is supported by much experimental work, notably that

of Kruschov ( 1957), although, as will be shown later, material properties
other than hardness are also important in determining wear rates.
However, there is one important respect in which Equation (3.1) is very
misleading. As it stands, it says that the volume of material worn per unit
slid distance is equal to the normal load divided by three times the hardness,
whereas experimental work shows that the worn volume is always less than
this, and often by several orders of magnitude. The equation was derived on
the assumption that a wear particle was produced at each asperity encounter,
and Archard therefore reconciled the theory with experimental observations
by postulating that wear particles were produced at only a fraction k of such
encounters, so that
Q=kW/3H (3.2)
or
Q=KW/H (3.3)
where
K=k/3
Thus, the uncertainty in predicting a wear rate is due to the uncertainty in

the value of K, and this must be found experimentally for different
combinations of sliding materials and different conditions of rubbing.
The parameter K is known as the wear coefficient, and examination of
Equation (3.3) shows that it is dimensionless. However, it can be seen that,
even when K is known, it is also necessary to know the hardness before the
rate of wear at a particular normal load can be calculated, and the hardnesses
of some materials may vary, owing to factors such as work hardening and
point-to-point variations in composition. Furthermore, it can be seen that
the wear rate depends on K/H, rather thanK alone, so that tables of K
alone must be interpreted with care. For example, a hard material with a
high value of K may have a lower wear rate than a softer material with a
lower value of K. Thus, it is often more convenient to use a wear equation
of the form
Q=CW (3.4)
where the wear constant C can be found experimentally without any
independent determination of H, and the lowest wear rate will be given by
the material with the lowest value of C.
3.3.2 Abrasive Wear
The term 'abrasive wear' covers two types of situation, known, respectively,
as two-body abrasion and three-body abrasion, in each of which a soft surface
is ploughed by a relatively hard material. In two-body abrasion a rough
WEAR 71
____ _L
Figure 3.1 Abrasive wear by a conical indicator
hard surface slides against a relatively soft opposing surface, whereas in

three-body abrasion rough hard particles trapped between the two sliding
surfaces cause one or both of them to be abraded.
We initially obtain a semiquantitative expression for the rate of abrasive
wear by assuming a simplified model in which a particle of abrasive has a
conical shape, defined by the angle ()shown in Figure 3.1, and the abraded
softer surface is flat.
Figure 3.1 shows a single abrasive particle ploughing a track through
the softer surface. In traversing unit distance it displaces a volume of material
v = r·d. But d = r·tan (); therefore, v = r 2 ·tan 0. For simplicity, we assume
that the softer material has yielded under the normal load only and therefore
the abrasive particle transmits a normal load of nr 2 H /2, where H is the
hardness of the softer material.
Thus, if there are n asperity contacts, the total normal load, W, is given
by W = nnr 2 H /2, and the total volume of material displaced in unit sliding
distance is Q, where Q = nr 2 ·tan 0.
Eliminating n,
Q = 2Wtan OjnH (3.5)
As with adhesive wear, it is found that not all traversals of abrasive particles
produce loose wear debris, so Equation (3.5) is modified to
Q = k·(tan 0/n)· W/H (3.6)
where k is the proportion of events which actually result in the formation
of a wear particle, or
Q=K·W/H (3.7)
where K = k tan() jn.
Equation (3.7) is of the same form as the adhesive wear equation (3.1)
and, according to this simple derivation, the 'laws' of wear which are derived
from Equation (3.1) and given in sub-section 3.3.1 should apply equally well
to abrasive wear.
In the derivation described above we assumed that all the material
displaced by the abrasive particle becomes loose wear debris, but examination
of abraded surfaces shows that this is a gross oversimplification. We
shall discuss more modern views of the processes of material removal in
Section 3.4. It is sufficient to note here that much of the displaced material
72 TRIBOLOGY
can be simply piled up at the sides of the grooves and not lost to the surface.
As a result, typical values of k are ~ 0.1.
Two-body abrasion is typified by the action of a file or emery paper
on a softer metal surface. There are two requirements for this process to take
place: one surface must be harder than the other; and the asperities on the
harder surface must be 'sharp' - that is, have high values of tan fJ. The
roughness of the surface does not appear in Equation (3.7) but, although
the worn volume does not depend on roughness, in practice it is also necessary,
if the effects of abrasive wear are not to be too severe, that the hard surface
should have a low roughness; this ensures that the abraded surface has a
myriad of very fine abrasion marks rather than relatively few very deep and
severe ones. Luckily, most finishing processes which reduce asperity slopes
also reduce surface roughness.
Severe two-body abrasive wear of lubricated surfaces has been largely
eliminated from modern machinery, owing to greater awareness of the
importance of material properties and surface finish, and the availability of
surface profilometers to ensure that engineering surfaces are of the required
quality. Studies of this form of wear are now more related to polishing and
grinding processes, where two-body abrasion is deliberately employed.
However, in the absence of effective lubrication, when adhesive transfer can
take place, both the worn surface and the transferred material can be
roughened and work hardened so that two-body abrasion becomes a
significant problem.
Three-body abrasion is still an important cause of wear. Many moving
parts operate in environments containing significant amounts of adventitious
dirt and dust; the products of corrosive wear are more often than not abrasive
in character, and metallic wear particles are often fully work hardened, and
thus abrasive to their parent surfaces. The practical solution to the problem
of three-body abrasion is to flush away abrasive particles and permanently
to remove them by filtration.
3.3.3 Fatigue Wear
Fatigue wear can be an important phenomenon on two scales: macroscopic

and microscopic. Macroscopic fatigue wear occurs at non-conforming loaded
surfaces, such as those found in rolling contacts, whereas microscopic fatigue
occurs at the contacts between sliding asperities. We shall discuss these two
situations in turn.
Rolling Contact
Adhesive and abrasive wear depend on direct solid-solid contact and these
wear processes operate throughout a period of rubbing. However, if the
moving surfaces can be separated by a lubricating film, as in hydrodynamic
and hydrostatic bearings, these mechanisms of wear cannot operate. In the
WEAR 73
absence of any abrasive particles, hydrostatic bearings can be expected to

last indefinitely, and hydrodynamic bearings wear only for a fraction of the
first revolution after each start-up, so that, for example, the crankshaft
bearings in a modern car can be expected to run for ~ 10 9 cycles under
severe fluctuating loading without suffering unacceptable wear.
However, in well-lubricated rolling element bearings there is, again, no
progressive visible wear due to adhesion or abrasion, but bearing life is
limited by fatigue. Such rolling contact fatigue is characterised by the
formation of large wear fragments after a critical number of revolutions.
Prior to this time there is almost no detectable wear, but as soon as the first
wear fragments are formed, the bearing life is at an end. Thus, it is not
appropriate to talk of the wear rate of a rolling element bearing; a more
relevant term is 'the useful life'.
Although direct solid-solid contact does not occur in a well-lubricated
rolling element bearing, the opposing surfaces are subjected to very large
stresses, as described in Chapter 2, which are transmitted through the oil
film. As we know, in the presence of such stresses the maximum compressive
stress occurs at the surface but the maximum shear stress, and therefore the
position of first yield, occurs at some distance below the surface. As rolling
proceeds, any subsurface element is subjected to a stress cycle for each passage
of a ball or roller, and we have already shown in sub-section 2.8.5 that it is
this stress cycling which is almost entirely responsible for energy dissipation
in rolling contacts. If the stress amplitude is above the fatigue limit for the
bearing material, as is almost invariably the case under the high contact
pressures used in rolling element bearings, fatigue failure will eventually occur.
The position of failure in a perfect material subjected to pure rolling
contact would be defined by the position of maximum shear stress given by
the Hertz equations. If some sliding were superimposed on the rolling motion,
as in many gear teeth, then the sliding traction would move the position of
maximum shear stress closer to the surface. In practice, materials are never
perfect and the position of failure would normally depend on microstructural
factors such as the presence of inclusions or microcracks. In recent years the
lifetimes of rolling element. bearings have been improved considerably by
using only materials of very high quality to avoid such microstructural effects.
The bearing life, as specified by a manufacturer, is the number of cycles
which will be reached or exceeded by 90% of similar bearings under identical
operating conditions. It has been found empirically that the life, L, is inversely
proportional to the cube of the applied load, W - i.e.
(3.8)
Sliding Contact
Wear coefficients in most engineering situations are found to be of the order

of 10- 3 -10- 7 , with the upper value of 10- 3 normally corresponding to
unacceptably high wear. In adhesion theory the wear coefficient is simply
74 TRIBOLOGY
defined as the probability of a particular asperity contact resulting in the

formation of a wear particle, but this cdefinition leaves unanswered two
important questions:
( 1) Why, when adhesion theory assumes that all asperities are deformed
plastically, are wear coefficients so low?
(2) Why, when all gross plastic deformation of an asperity must usually
take place during its first few contacts, do wear particles form after many
thousands of further cycles of nominally elastic deformation?
Both these questions can be answered by invoking fatigue as the
predominant wear mechanism.
As we now know, when any two solid surfaces are placed in contact
under a normal load, the surfaces approach each other until the elastic
reactive force at the interface is just sufficient to support the normal load.
A similar situation obtains during sliding, with the reactive force being
required to support both the normal and the tangential forces. Because of
their height distribution the contacting asperities are strained to different
extents, so that, in general, the contact stresses will vary from a very low
level of elastic stress up to the fracture stress of the weaker of the two
contacting materials.
Truly elastic stresses will cause no damage, while the fracture stress will
cause the formation of a wear particle during a single interaction. However,
at intermediate stresses - i.e. at any stress above the fatigue limit but below
the fracture stress - the asperity will be subjected to a single fatigue cycle,
and the accumulation of such cycles with continued sliding will eventually
cause the formation of a wear particle by local surface fatigue fracture.
(At this point it is useful to discuss briefly the concept of elastic fatigue.
To an engineer this is simply taken to mean fatigue of a material at a stress
below its yield point. However, to a physical metallurgist true elastic fatigue
does not exist: after a cycle of truly elastic deformation, each atom in the
specimen would return to its original position, and this stress cycling could
be repeated indefinitely without causing fatigue failure. Fatigue failure
requires plastic deformation in the form of dislocation movements which
eventually lead to the nucleation and growth of fatigue cracks. However,
these two views can be reconciled when it is realised that in the polycrystalline
materials which are normally tested by engineers some favourably oriented
grains can undergo microscopic plastic deformation when the specimen as
a whole is subjected to a stress well below the yield stress. The minimum
stress at which this localised plastic deformation occurs is known as the
fatigue limit - i.e. the stress below which fatigue failure will never occur -
and when such a stress exists, it is often only about a half of the bulk yield
stress.)
Previous authors have described fatigue theories of wear. Halling ( 1975)
derived a wear equation based on the Manson-Coffin (Tavernelli and Coffin,
WEAR 75
1959) fatigue equation, which relates fatigue life to the plastic strain increment
during each cycle, and others (Kragelsky, 1965; Soda et al., 1977) have
suggested fatigue as a predominant process at low wear rates. In what follows
we hope to show that fatigue is probably one of the more important
contributory factors in almost all mechanical wear processes.
To do this, we return to the phenomenon of shakedown (Johnson, 1985)
which we described in Chapter 2. We recall that, in rolling contact of
non-conforming solids, the type of deformation which occurs on repeated
loading is critically dependent on the stress level. At stresses below the
shakedown stress, contact will become entirely elastic within a few stress
cycles, whereas, at any higher stress level, there will be an increment of plastic
deformation on each loading cycle.
We should, clearly, expect qualitatively similar behaviour in the sliding
contact of non-conforming asperities. There will be a limiting stress below
which deformation on repeated cycling will be entirely elastic. However, this
stress may be closer to the stress for first yield than is the case in macroscopic
rolling contact, because of the difficulty of accumulating in the microscopic
contact zones the residual stresses which are necessary for shakedown. At
stresses only marginally above the shakedown stress, plastic deformation will
be subsurface and constrained, but at higher stresses the zone of plastic
deformation will extend to the free surface and be unconstrained. Again,
owing to the much higher friction tractions which occur in sliding contacts,
the zone of plastic deformation may reach the free surface at much lower
stresses than those required in rolling contacts. Indeed, as we showed in
Section 2.4, in the presence of strong interfacial bonds plastic flow may occur
in shear under very low normal stresses.
In any event, each cycle of plastic deformation will correspond to
one fatigue cycle, and the accumulation of such cycles will eventually cause
a fatigue fracture and the generation of a wear particle.
Clearly, in assessing the contribution of fatigue to mechanical wear, we
need an estimate of the proportion of the contacting asperities which will be
subjected to stresses above the fatigue limit, which, for simplicity, we take
to be equal to the stress for first yield. This estimate can be made from a
more detailed examination of the data summarised in Figure 1.20, which
shows the ratio between plastic and elastic contact area, under normal loading
only, for different values of the plasticity index and different normalised
contact pressures. A selection of the data which were computed to plot these
graphs is tabulated in Table 3.1, where R is the ratio between plastic and
elastic contact area; ljJ is the plasticity index (E/H)·(ujp}!; and pis the
normalised pressure, pI H.
The conclusions to be drawn from the data in Table 3.1 are similar to
those of Greenwood and Williamson ( 1966) - i.e. for all realistic loads the
deformation of surfaces is elastic when ljJ is less than 0.6, and at least partially
plastic when ljJ is significantly greater than unity. However, the additional
76 TRIBOLOGY
Table 3.1
p= w-4 p=5·10- 4 p=l0-3 p=5·10- 3 p=l0-2
1{1 R 1{1 R 1{1 R 1{1 R 1{1 R
2 0 2.3 5.3 5 200
1.3 0 1.5 2 5 500
1.2 0.5 1.3 3.6 1.8 13
1.1 0 1 0 1.2 3
1 2
0.6 0
important point to be noted here is that most engineering surfaces as

machined have values of tjJ which are substantially greater than unity, and
when such surfaces are first placed in contact, the deformation of the
contacting asperities will be predominantly plastic. Furthermore, as already
noted, the presence of tangential tractions during sliding will exacerbate this
situation. It is true that subsequent rubbing may cause the surfaces to run
in, so that tjJ decreases to a point where the deformation will be nominally
elastic but, as long as this nominally elastic deformation is above the fatigue
limit, the asperities will be subjected to fatigue cycling. Furthermore, it is
clear that surfaces which are machined to a very fine finish, so that the initial
value of tjJ is less than unity, could still be subjected to similar levels of
nominally elastic stress and, therefore, similar levels of subsurface fatigue.
Thus, we should expect a high proportion of the contacting asperities on
engineering surfaces to be subjected to fatigue wear.
The operation of the fatigue mechanism answers both the questions
raised above. First, tne wear coefficient can now be simply interpreted in
terms of the number of fatigue cycles needed to cause the detachment 9f a
wear particle. For example, if the wear coefficient indicates that only 1 in
10 6 asperity contacts results in the formation of a wear particle, this suggests
that the stress situation is such that, on average, it takes 10 6 stress cycles to
produce a fatigue fracture. Second, although almost all contacts are nominally
elastic in the engineering sense, in that no observable irreversible deformation
takes place, cyclic, subsurface plastic deformation causes many asperities to
fail by fatigue after many cycles of such nominally elastic deformation.
Clearly, it is not necessary for the same asperity pairs to make repeated
contact for the above mechanism to operate. In general, the higher asperities
will always undergo gross plastic deformation during, at most, the first few
contacts, after which they will have flattened sufficiently to remain nominally
elastic. However, as before, a high proportion of the contacting asperities
will be subjected to stresses which will cause eventual fatigue failure.
Furthermore, as the original heavily loaded contacts are worn away, the
load is transferred to lower asperities, which, in turn, become subject to fatigue.
WEAR 77
In Section 3.4 we shall expand this discussion to include the effects of

rubbing dissimilar materials and to take account of interfacial adhesion.
3.3.4 Delamination Wear
The delamination theory of wear was first put forward by Suh ( 1973) and
has since been elaborated by Suh (1977) and others (Jahanmir et al., 1974).
The theory involves detailed consideration of subsurface dislocation interactions,
and readers interested in the details should consult the original papers. For
the purposes of this text it is sufficient to set out the physical basis of the
theory, which may be summarised as follows (Barwell, 1983).
( 1) When two sliding surfaces interact, the asperities on the softer surface
are flattened by repeated plastic loading, until the sliding condition
corresponds to that of a harder, rough surface sliding against an
approximately plane softer surface. Thus, at each point on the softer
surface there is repeated subsurface cyclic shear loading, as already
described in discussing fatigue wear.
(2) As this cyclic loading continues, voids and cracks are nucleated below
the surface, crack nucleation at the surface being prevented by the triaxial
compressive stress which exists immediately below the surface.
(3) Further cycling causes these voids and cracks to link up, to form long
cracks at an approximately constant distance below the surface.
(4) When the cracks reach some critical length, the stress situation at the
crack tip causes the crack to break through to the free surface, resulting
in the formation of the thin plate-like wear particles which are often
observed in practice.
3.4 DISCUSSION OF MECHANICAL WEAR PROCESSES
In discussing mechanical wear processes we must offer explanations of the

full range of wear coefficients, from the maximum possible value of unity to
values as low as 10- 7 . Although the division is somewhat arbitrary, for
clarity we shall initially divide the wear processes into two groups: the
higher-wear processes, where the deformation of at least one of the surfaces
is seen to be plastic, and the lower-wear processes, where the deformation
is nominally elastic.
3.4.1 Wear Involving Observable Plastic Deformation
Clearly, the highest wear coefficients are found under conditions of abrasive
wear and, under these conditions, the value of K would be unity if every
hard asperity removed a chip of material on each contact with the opposing
78 TRIBOLOGY
1•1
lbl
Figure 3.2 Effect of asperity shape on the form of abrasive wear (courtesy, Professor
T. H. C. Childs)
WEAR 79
lei
Figure 3.2 (cont.)
surface. In sub-section 3.3.2 we simply stated the fact that abrasive wear
coefficients are normally much lower than this, because much of the displaced
material is simply piled up at the sides of the groove and not lost to the
surface. In a recent paper Childs ( 1988) has presented a much more realistic
picture of abrasive wear, and this is summarised below.
Studies of metals scratched by model hard asperities generally show
that the surface damage is of three types, illustrated in Figure 3.2. The blunter
asperities allow the softer metal to flow round them as a wave, as shown in
Figure 3.2(a); this does not cause any removal of metal on the first pass,
although, as we shall see later, metal may be removed on subsequent passes.
The sharper asperities cause chips to be machined from the softer surface,
as shown in Figure 3.2(b ). Asperities of intermediate shapes cause a wedge
or prow of deformed metal to form ahead of them at the start of sliding, as
shown in Figure 3.2( c). This wedge effectively blunts the asperity, so that
further metal may flow round the wedge without removal (cf. Figure 3.2a ),
or it may intermittently displace the wedge, to form a new one. Childs mapped
the form of material displacement in terms of two variables: the surface slope,
(),and the interfacial shear strength parameter, r:./k, where r:, is the interfacial
shear strength and k is the plastic yield stress in shear of the softer material.
(Note that r:,/ k is identical with the parameter c defined in sub-section 2.3.3,
but in this discussion we shall use the form chosen by Childs.)
Childs cites the work of Challen and Oxley ( 1979), who have developed
plane strain slip line field models of wave-forming and cutting flows in friction
80 TRIBOLOGY
and wear. They showed that, for a wedge-shaped asperity of slope (), wave
flow should occur when
()::::;; 0.5 cos - 1 ( •.! k)
while chip formation should occur when
() ~ n/2- 0.5 cos - 1 ( r./ k)
The limiting conditions represented by the equalities in these equations are
mapped as the lines AB and BC, respectively, in Figure 3.3. Within the region
ABC no steady flow is possible and this is the region of initial wedge formation
as illustrated in Figure 3.2( c).
Further work by Challen and Oxley ( 1984) has shown that the situation
is far more complicated than that illustrated in Figure 3.3. However,
discussion of this work would not be appropriate here, and readers requiring
more information are referred to the original papers.
Childs expresses the severity of wear by a wear coefficient K in an
equation similar to the Archard equation for adhesive wear - i.e.
V=KWs/H (3.9)
where V is wear volume, W is the normal load, s is the slid distance and H
is the hardness of the softer material. However, in this case the K value
cannot be interpreted as the probability of forming a wear particle, which
would have a limiting value of unity: it is simply a coefficient relating the
volume of material removed per unit slid distance to the load and the hardness,
and can have a value greater than unity.
When metal is removed as a chip, the wear process is described as
e (deg)
Figure 3.3 Map of flow regimes (after Childs, 1988)
WEAR 81
1.0
0.8
0.6
!:!.
k
0.4
0.2
0
10 20 50
9 (deg)
Figure 3.4 Calculated values of abrasive wear coefficient K for various values of
r:sfk and fJ (after Childs, 1988)
machining wear. In this regime upper limits to wear coefficients can be

calculated, using slip line field theory, and such wear coefficients, for various
values of r./ k and 0, are shown in Figure 3.4. It can be seen that the calculated
values of K for the machining process fall to zero at a surface slope of
approximately 20°, which is very rough by normal engineering standards.
Other wear processes must, therefore, take over at lower slopes, and Challen
and Oxley have shown that in the region of plastic wave flow the observed
wear rates can be explained on the basis of high strain fatigue.
Childs points out that, in the fully plastic conditions to which his work
applies, calculations of the forces associated with the flow fields automatically
combine the effects of adhesion and deformation, and may therefore be used
to calculate friction coefficients. Such calculations have been made, and they
show that for wave flows friction coefficients are less than 0.4, while with
cutting wear they are greater than 0.4.
The theoretical work described above was based on plane strain
deformation by wedge-shaped asperities. However, real engineering situations
involve th:J:ee-dimensional asperities having a wide variety of shapes and
orientations, and Childs cites many sources to show that the rate of abrasive
wear and the transitions from wave flow to prow formation and from prow
formation to cutting are very dependent on the shape and orientation of the
abrasive particle and on the friction stress.
The transitions are also strongly dependent on the mechanical properties
of the abraded surface, the critical parameters being its hardness and
work-hardening characteristics. In addition to altering the flow regime
boundaries, work hardening also affects the wear rate within the cutting
regime.
82 TRIBOLOGY
To illustrate these effects, Childs rewrites Equation (3.9) as
V=fKWs/H (3.1 0)
where K is primarily dependent on abrasive shape and friction stress, and

relates the scratch volume to the load and hardness, whereas f is the fraction
of the scratch volume which becomes wear debris. Childs cites work by
Hokkirigawa and Li (1987) and Zum Gahr (1987) which shows that f
decreases as surface hardening increases.
For hard brittle metals fracture toughness is thought to be more
important than work hardening in influencing flow mode. The work of Zum
Gahr shows that the ranking of fk/ H for martensitic and austenitic irons
changes with abrasive slope. With rounded abrasives the higher hardness of
martensitic irons gives them a better wear resistance than that of austenitic
irons, but with sharper abrasives their poorer toughness reverses that order.
3.4.2 Wear under Nominally Elastic Contact Conditions
In sub-section 3.3.2 we have already shown that, during nominally elastic

contact, a high proportion of the contacting asperities will be subjected to
repeated subsurface shear stresses in the plastic range, and, consequently, to
eventual fatigue failure. Thus, during nominally elastic contact, which obtains
on all well-run-in surfaces, fatigue will be the predominant wear mechanism.
As shown in sub-section 3.3.3, when the plasticity index is less than 0.6,
all asperities deform nominally elastically, and in many cases, particularly
at low nominal pressures, the deformation may be truly elastic. In such cases
wear rates are extremely low and are often due to corrosive wear only or to
the removal of material at the atomic level rather than as particles.
Recently Halling ( 1988) has developed a criterion for elastic contact
which has already been described in sub-section 2.3.1, and implicitly used in
Table 3.1. The criterion gives the limiting normalised pressure for elastic
behaviour as
p=p/H=4( 11 pmjJ) fb (s-h)t<p(s)ds (3.11)
J(b-(I/Vt 2 ))
where p is the nominal applied pressure; b defines the height, ba, of the
highest asperity; sis the non-dimensional height, z/ a, of an asperity; his the
non-dimensional separation, d/ a, of the surfaces; and ljJ is the plasticity index.
This criterion was developed for surface contact under normal load only,
but it would also apply to frictional loading with friction coefficients less
than 0.1. The criterion would need to be modified for situations of higher
friction; nevertheless, it still provides a useful insight into the factors which
must be considered in choosing materials to resist wear. In particular, it
stresses the great importance of the plasticity index.
WEAR 83
3.4.3 The Form of the Wear laws
We know from experiment that wear is proportional to the normal load and
slid distance, and, although we have shown that this is predicted by theories
of adhesion and abrasion, we have not shown that it would be expected if
fatigue were the predominant wear process. Fatigue wear equations predicting
such relationships have been derived by Halling ( 1975) and Kragelsky ( 1965).
However, the following simple argument shows that the observed linear
relationships of wear rate to both normal load and slid distance derive simply
from surface topography and would be expected for any form of mechanical
wear involving asperity contact only.
( 1) As shown in Chapter 1, when contacting surfaces have exponential height

distributions, the mean contact pressure and the mean individual contact
area are independent of the load: the number of contacts, and thus the
total contact area, are proportional to the load. The same is true, to a
very good approximation, for surfaces having Gaussian or near-Gaussian
height distributions.
(2) Most surfaces have Gaussian or near-Gaussian height distributions, at
least in the highest 10% of the surface to which contact is normally
limited, so that surfaces under load are subjected to contact stresses and
contact stress distributions which are almost independent of the load.
( 3) Whatever the mechanical wear process, the probability that an individual
asperity interaction will lead to the formation of a wear particle will
depend on the local contact stress, and is therefore also independent of
the load.
( 4) The effect of an increase in the load is simply to give a proportionate
increase in the true contact area and, thus, in the number of contacts in
each stress state, including the number of wearing contacts.
( 5) Similarly, the effect of an increase in slid distance is simply to give a
proportionate increase in the number of wearing contacts.
Statements (4) and (5) can be combined to write down a general
mechanical wear law as
Q=K·A (3.12}
or
Q=K·W//3 (3.13}
where f3 is a mechanical constant defining the load capacity. For plastic
asperity contact this constant is simply H, in line with Equation (3.3), whereas
with elastic contact its value will be somewhat different from H. However,
we now know that even in elastic contact most asperities are very close to
their yield stresses, so that f3 may be taken as being approximately equal to
H for all practical situations.
84 TRIBOLOGY
3.4.4 Wear of Dissimilar Materials
When two materials having different hardnesses are loaded together, the two
opposed asperities involved in any particular contact must be loaded, and
therefore stressed, to the same level. If the hardnesses are significantly different,
the softer material may deform plastically, but deformation of the harder
material must remain at least nominally elastic. As the load is increased, the
plastically deforming material may work harden; but as long as its current
hardness is lower than that of the opposing surface, the harder asperity will
remain nominally elastic. However, we now know that subsurface plastic
deformation will occur when the pressure at the surface is about half the
hardness. Therefore, a softer material can cause subsurface plastic deformation
in a harder one as long as its hardness, in the fully work-hardened condition,
is more than about halfthe initial hardness of the harder surface. To emphasise
this important point, it can be seen that the harder surface must be at least
twice as hard as the softer if it is to remain fully elastic.
Thus, behaviour during sliding is very dependent on the difference in
hardness between the two materials.
If one material is very much harder than the other, the condition of
stress equality can only be obtained if the hard asperity penetrates the softer
surface, as in a hardness test. Relative sliding then requires the hard asperity
to plough a groove in the opposing softer surface. Hence, the softer surface
is subjected to abrasive wear. However, it must be remembered that if the
hardness ratio is less than about 2, the abrasive surface may still be subjected
to subsurface plastic deformation and fatigue wear.
If the two hardnesses are such that both surfaces can remain nominally
elastic (a condition that often obtains after running in), it is probable that
each asperity of any contacting pair will be subjected to subsurface plasticity
and, hence, fatigue. Indeed, both will be subjected to identical cycles offatigue
stress, and repeated contacts will cause fatigue failures at both surfaces.
However, the harder asperity will require more stress cycles to produce a
wear particle, and will therefore wear at a lower rate than the softer one.
This is illustrated schematically in Figure 3.5, which shows the S-log N fatigue
curves (stress-log number of cycles to failure) for the two materials. For the
example given, a horizontal line through the stress S 0 shows that the numbers
of cycles to failure for the harder and softer materials at the stress S0 would
be in the ratio N HI N s and the relative wear rates would thus be in the ratio
Nsf NH (remembering that N is shown on a log scale). This oversimplifies
the true situation, where the stress cycles would be of varying amplitude, but
it clearly illustrates the principle involved.
Thus, fatigue wear explains the observation that a harder material can
be worn by a softer one, a phenomenon which has never been explained
satisfactorily by adhesion theory.
By invoking fatigue as the basic wear process we can also explain the
formation of loose wear debris.
WEAR 85
Ns
log N
Figure 3.5 Illustrating simultaneous fatigue ofboth hard and soft rubbing surfaces
It will be remembered that the adhesion theory explains the transfer of

particles of the softer material to the harder surface by postulating that the
interfacial bond is so strong that the junction breaks in the softer material
rather than at the interface. The theory then goes on to suggest that loose
debris is formed by the transferred particle becoming detached during a
subsequent contact. However, it is difficult to see how the transferred particle
can be detached in this way. For the initial transfer to occur, the bond of
the softer material to the harder must have been stronger than the cohesive
strength of the softer material in its fully work-hardened condition. Therefore,
the softer material should never be sufficiently strong to break the bond
between the transferred particle and the counterface. However, if the
transferred particle is removed from the softer material by fatigue cycling,
its bond to the counterface may be very weak, corresponding only to the
force required to cause final fracture during the last fatigue cycle. Indeed,
there is no requirement for the observed loose debris ever to have been
attached to the counterface: it may have suffered final detachment simply
because of a collision with an opposing asperity.
In summary, with the exception of the cutting wear process described
by Childs, it is evident that fatigue is a contributory factor over the full
spectrum of wear behaviour. This is so by definition, as fatigue failure is
defined as failure caused by repeated applications of stresses which are
insufficiently high to· cause failure in a single application- a situation which
manifestly obtains in all wear situations where the wear coefficient is less
than unity. Thus, the very high wear rates observed in the milder forms of
abrasive wear - i.e. wave-forming and prow-forming - can be ascribed to
high strain fatigue, in line with Challen and Oxley, while the low wear rates
observed in nominally elastic contact can be ascribed to low strain fatigue.
86 TRIBOLOGY
However, the fact that fatigue is always a contributory factor should

not be taken to minimise the importance of adhesion. The effect of adhesion
will invariably be to increase the stress required to break an asperity junction,
either by increasing the r:.fk ratio in the form used by Childs or, more
obviously, by requiring the application of a tensile pull-off force. Thus,
whenever adhesion is present, the stress amplitude at each asperity contact
will be increased, and the number of cycles required to cause junction failure
will be correspondingly decreased. It is, therefore, clear that fatigue and
adhesion should not be regarded as separate wear processes: fatigue is always
the basis of wear particle fonnation, but in the presence of adhesion the rate
of fatigue wear will increase, as will the amount of transfer, and the stronger
the adhesion the greater will be the wear rate. As, on the above explanation,
the process which has been previously described as adhesive wear is simply
fatigue wear made more severe by adhesion, in the remainder of this text we
shall refer to this process as adhesion-enhanced wear.
Delamination wear also conforms to the definition of fatigue given above
- i.e. failure resulting from many stress cycles at a stress level which cannot
cause failure in a single cycle. The distinctive feature of delamination theory
is that it invokes fatigue cycling over relatively large areas at an approximately
constant distance below the surface, and thus explains the occurrence of the
plate-like wear particles which are commonly observed. It is very likely that
the fatigue process described by delamination theory does indeed take place.
However, an alternative explanation of the plate-like wear particles is that
they are formed by severe repeated compression of initially equiaxed wear
particles. There is no doubt that when such an equiaxed particle was first
formed, it would instantaneously separate the sliding surfaces and thus take
a large fraction of the total applied load. The resulting stress would certainly
cause very severe plastic compression of the particle in a plane parallel to
the interface and would thus deform the initially equiaxed particle into a
plate-like shape (Sasada et al., 1979).
3.5 CORROSIVE WEAR
When rubbing takes place in a corrosive environment, either liquid or gaseous,

then surface reactions can take place and reaction products can be formed
on one or both surfaces. These reaction products can be removed by rubbing,
and the cycle of reaction and removal can be repeated indefinitely.
By far the most common corrosive medium is the oxygen content of
air. Therefore, we shall largely confine our attention to corrosive wear in an
oxidising environment, otherwise known as oxidative wear, but the principles
described would apply to wear in any other corrosive medium.
The first important point to note is that oxidation of sliding surfaces is
usually beneficial. The oxide film prevents metal-metal contact and thus
WEAR 87
mitigates against the severe adhesion-enhanced wear which would otherwise

occur. Very often, when surfaces are oxidised, the wear debris is finely divided
oxide, the rubbing surfaces remain smooth and the rate of loss of material
is low.
The effects of oxidation depend not only on the oxidation rate, but also
on the mechanical properties of both the oxide and the substrate, and the
adhesion of the oxide to the substrate. As with the mechanical wear processes
discussed earlier, it is unlikely that there will ever be a theory of oxidative
wear which will allow accurate predictions of wear rates over a wide range
of conditions. Any such theory would require, as before, precise knowledge
of the conditions necessary to produce a wear particle. In addition, it would
require knowledge of the surface temperature distribution, as a function of
both position and time; of the relationship between this distribution and the
oxide film thickness; and of the stabilities of films of different thicknesses
under the variable thermal and mechanical conditions to which they would
be subjected. In particular, as oxidation is a thermally activated process, the
rate of oxidation can increase exponentially with temperature. This makes
prediction particularly difficult, as a change of only a few degrees in surface
temperature can change the rate of oxidation by an order of magnitude.
Simplified models of oxidative wear have been produced (Quinn, 1971;
Earles and Powell, 1971) and they probably provide good qualitative
descriptions of the wear process. However, to give good agreement with
observed wear rates, it is necessary to invoke, for example, activation energies
for oxidation which are different from those found in static oxidation
experiments. We can conclude that the prediction of oxidative wear rates
will remain empirical, and we shall restrict ourselves to a qualitative
description of the process.
The oxides of most metals, including iron, have volumes which are
significantly different from those of their parent metals. Therefore, during
static oxidation, oxide formation causes the development of stresses in the
oxide which increase with film thickness. At some critical film thickness the
oxide film can then fail, either by blistering, when the stress within the film
exceeds the strength of the adhesive bond between oxide and substrate, or
by cracking, when the oxide fails in tension. The critical film thickness thus
depends on the metal/ oxide volume ratio, on the adhesive bond between the
oxide and substrate and on the strength of the oxide.
The complexity of the process increases substantially when sliding occurs.
In this case failure does not depend only on the stresses resulting from the
film formation, but also on those caused by the normal and tangential external
loads, including the thermal stresses induced by flash temperatures at asperity
contacts.
Finally, the stability of the film will be dependent on the hardness of
the substrate: a hard substrate will provide good support for the film, whereas
a softer substrate is likely to deform plastically, at least locally, and thus
effectively leave the thin brittle film unsupported.
88 TRIBOLOGY
Archard ( 1980) has given a simple qualitative development of corrosive

wear theory which shows that the rate of wear can again be described by
an equation of the form Q = kWI H, where the wear coefficient, K, is given by
(3.14}
where K 3 is the proportion of events which produce a wear particle, A. is the
critical film thickness at which the film becomes unstable and 2a is the width
of an asperity contact. The values of both K 3 and A. are dependent on the
various thermal, chemical and mechanical variables described above.
As factors such as reactivity and film thickness are functions of both
time and temperature, the effectiveness of oxidation in providing a protective
film can vary dramatically for relatively small changes in rubbing conditions.
The best-known example of this is in the mild-to-severe wear transitions
observed in the unlubricated sliding of many metal pairs. This effect is
illustrated schematically in Figure 3.6, which shows a log-log plot of the
variation of wear rate with load for carbon steel sliding on carbon steel.
At low loads, below the first transition load, T1 , the wear rate remains
low, the surfaces remain smooth and the wear debris is finely divided and
almost entirely oxidised. It is thought that in this regime the wear debris is
removed in the form of oxide, and that the worn areas have time to reoxidise
before they next make contact. (Remember that the number of contacts in
any stress state is proportional to load, so that the time available for oxidation
between contacts is inversely proportional to load.)
At the transition load T1 , the wear rate changes by more than two orders
of magnitude, the wear debris consists of relatively large metal particles and
the surfaces are severely damaged and roughened. In the range T1 - T2 there
is insufficient time for the contact spots to reoxidise between contacts, so
that adhesion-enhanced wear takes place.
0
E'
logW
Figure 3.6 Illustrating the mild-to-severe wear transitions of an unlubricated steel

WEAR 89
The rate of oxidation and, hence, the film thickness attained between
successive asperity collisions are very dependent on the local temperature,
and this temperature, in turn, depends on the input of frictional work and
thus on the load. Therefore, all other things being equal, the rate of oxidation
increases approximately exponentially with the load, while the time between
collisions decreases only linearly. Thus, at the second transition load, T2 , the
temperature effect becomes predominant and the wear process returns to
one of 'mild' oxidative wear.
Steel surfaces examined after rubbing above T2 also have a hardened
surface layer, due to the repetitive rapid cooling from high flash temperatures.
This hardened layer also favours the mild wear regime, owing to the reduced
incidence of plastic deformation and the improved support given to the oxide
film. This effect of hardness is illustrated more dramatically by rubbing
steel-steel pairs which have been previously hardened and tempered to
different bulk hardnesses. As the initial hardness is increased, it is observed
that the severe wear regime becomes narrower until, at some critical value
of hardness, it disappears entirely and mild wear persists over the full load
range.
In the above discussion we have considered only the effect of increasing
the load. However, it is clear that similar effects would be seen as a result
of increasing the speed. Increasing speed would both increase flash temperatures,
owing to the increased input of frictional work, and decrease the time available
for reoxidation between collisions.
It has been pointed out by Endo and Fukuda ( 1965) that the presence
of oxygen at crack tips will tend to increase the rate of fatigue crack growth
and, thus, the fatigue wear. This is a further example of the interactions
between, and complementary effects of, wear processes which are often
described as though they are quite independent.
Finally, Barwell (1983) has pointed out that the adjectives 'mild' and
'severe' can be very misleading, in that the rate of material loss in a 'mild'
wear regime can be substantially higher than that in a 'severe' wear regime.
The terms 'mild' and 'severe' apply to the form of surface damage rather
than the rate of material loss, and Barwell suggests that the term 'smooth
sliding wear' should be substituted for 'mild wear'.
As mentioned earlier, the above discussion also applies to corrosive
media other than oxygen. For example, we have already described, in
Section 2.7, the use of extreme-pressure additives to deliberately generate
significant corrosive wear in preference to the much more damaging severe
wear which would occur in their absence.
3.6 WEAR MAPS
Recently Lim and Ashby ( 1987) have suggested that the various regimes of
mechanical and corrosive wear for any particular pair of rubbing materials
90 TRIBOLOGY
Seizure
10-1
J!
H
V (m/s)
Figure 3.7 Wear map for soft carbon steels (after Lim and Ashby, 1987, and
Childs, 1988)
could be shown on a single wear map plotted on axes of normalised pressure

pI H and sliding velocity. As an example of this technique, a wear map for
soft carbon steels sliding in air at room temperature is shown in Figure 3.7.
It can be seen that, in principle, the map can be divided into areas
corresponding to different wear regimes, with boundaries of sliding speed
and contact pressure beyond which a purely mechanical view of wear would
be grossly in error, owing to the effects of oxidation. Childs ( 1988) has
criticised this approach, at its current stage of development, on two grounds:
that it takes no account of differences of surface slope or surface shear
strength; and that it is only of the same practical use as quoting a wear
coefficient. Nevertheless, although the plotting of such maps for a wide range
of environmental conditions would require the analysis of enormous amounts
of wear data, they do offer the possibility of condensing such data into a
form which would be readily understood by designers, and, together with
the alternative type of map proposed by Childs (Figure 3.3), may offer the
first real opportunity of developing simple design guides for sliding wear
behaviour of metals.
3.7 MINOR TYPES OF WEAR
3.7.1 Fretting
The wear process known as fretting is not a separate process, but it is

convenient to treat it separately. It has been described in detail by Waterhouse
( 1988 ), and only the principles of fretting behaviour are summarised here.
WEAR 91
Fretting can occur whenever low-amplitude vibratory sliding takes place

between two surfaces. This is a common occurrence, since most machinery
is subjected to vibration both in transit and in operation.
Fretting can combine many of the wear processes already described.
The oscillatory sliding causes fatigue wear, which may be enhanced by
adhesion. Most commonly, this wear is combined with the effects of corrosion,
by oxygen or some other medium, and, as many corrosion products are
harder than their parent metals, this can also lead to abrasion. The fact that
there is no macroscopic sliding at fretting contacts often means that the
fretting wear debris cannot escape but is trapped between the surfaces.
In fretting there may be no true macroscopic sliding: the surfaces may
be in static contact over the central region of a contact zone, where the
contact pressure is relatively high, and subject to microslip at the outer
regions of the zone, where the pressure is relatively low and the tangential
tractions are sufficiently high to overcome the static friction. Under such
circumstances, fretting wear takes place over a narrow region at the edge of
the contact zone; in this region fatigue cracks may be nucleated and
propagated by the oscillatory stress, and this fatigue can lead to gross
component failure.
3.7.2 Erosion
Erosion is the form of damage experienced by a solid body when liquid or

solid particles impinge on a solid surface. In describing erosion, it is useful
to describe separately the effects of impacting liquids and solids, although,
in practice, the two processes often occur simultaneously.
Fluid Erosion
There are two basic types of fluid erosion - i.e. liquid impact erosion and
cavitation erosion. The current state of understanding of these processes has
recently been comprehensively reviewed in excellent papers by Hammitt
( 1980) and Schmitt ( 1980 ), and it is the purpose of this sub-section
to give only an outline of the more salient features of the processes.
The processes of removal of material are rather less well understood
than are the processes occurring in sliding wear, but many studies have been
made of the damage which occurs. It appears that there are different damage
processes for brittle solids, ductile metals and thermoplastic polymers,
thermosetting polymers and elastomers.
Liquid impact on a brittle material generates momentary stresses which
are sufficiently high to cause cracking in the form of initially unconnected
annular ring segments. With continued impact, further cracking occurs and
the cracks eventually join, so that material is removed in the form of chips.
The erosion damage of ductile materials, such as many metals and
thermoplastic polymers, initially takes the form of surface depressions with
92 TRIBOLOGY
upraised edges. These edges are then subjected to high stresses during the
radial flow of subsequent impacting drops, until they eventually fail, with
consequent surface roughening. The surface roughening itself then changes
the flow patterns of impacting liquid, with a possible exacerbation of the
erosion.
The erosion behaviour of thermosetting polymers is similar to that
already described for brittle materials, with the coalescence of ring cracks
leading to the formation of relatively large wear particles.
Elastomers are often used in the form of surface coatings to reduce
erosion damage. Some of these coatings fail by erosive wear which is
phenomenologically similar to that described for thermosetting plastics, while
others remain superficially almost undamaged but suffer loss of adhesion to
their substrates.
Cavitation damage occurs when bubbles entrained in a liquid become
unstable and implode against the surface of a solid. This is thought to produce
a combination of shock waves in the liquid and liquid microjet impact on
the solid surface. Cavitation damage closely resembles damage by liquid
impact erosion, but it is on a finer scale. It can be a very important and
damaging form of attack on solids which are moving in a liquid, examples
being ships' propellers and pump impellers.
Solid Erosion
When a stream of solid particles impinges on a solid surface, it is found that

the rate of erosive wear is dependent on the angle of incidence of the particles,
and that the wear rates for ductile and brittle materials follow different curves,
as shown in Figure 3.8.
For ductile materials the maximum erosion occurs at an apgle of
approximately 20° to the surface. Several workers (Bitter, 1963; Finnie et al.,
1979) have attempted to model this process but there is, as yet, no accepted
explanation of the form of the curve. Perhaps the closest approach to an
accepted analysis is that due to Bitter. He suggests that there are two wear
processes: cutting wear, which predominates at low angles, and deformation
wear, similar to that described for liquid erosion, which predominates at
high angles. The observed peak at 20° is then the point at which the sum
of the two effects has its maximum value.
For brittle materials the erosion process seems to be similar to that
described for liquid impact, with wear particles forming due to the coalescence
of partial ring cracks.
There are various theories of erosion damage but there are no reliable
methods of predicting erosive wear rates. Erosion damage in brittle materials
has been modelled as Hertzian impact between elastic solids (Hutchings,
1979) and some degree of agreement between theory and experiment has
been obtained. With ductile materials it is well established that there is an
WEAR 93
Angle of attack (deg)
Figure 3.8 Dependence of rate of erosion on angle of attack of impinging particles
'incubation' period during which no loose debris is formed, and this suggests
that erosive damage of ductile materials is a cumulative process, with fatigue
again being a likely cause.
3.8 EFFECTS OF SURFACE FILMS ON WEAR
It should now be clear that the wear resistance of any surface is dependent
on both the hardness and the elastic modulus of the contacting surface. We
have already shown in Section 1.5 that a relatively thin surface film can
modify both the hardness and the modulus of a surface, and thus have a
very significant effect on its tribological properties. This has recently been
treated in detail by Halling ( 1986), who shows that the critical factors in
determining the tribological properties of any substrate-coating-counterface
combination are the mechanical properties of the three materials, the
roughnesses of the substrate and counterface, and the thickness of the surface
film. As both film thickness and surface roughness are very likely to change
with continued rubbing, there is unlikely to be a quantitative method of
predicting the long-term wear behaviour of coated surfaces in the near future.
However, the principles established by Halling are very useful and can guide
the selection of coated materials for tribological applications. We shall return
to this topic in Chapter 4.
94 TRIBOLOGY
3.9 WEAR OF LUBRICATED SURFACES
Fluid lubricants have three important effects in reducing both friction and
wear of sliding surfaces.
( 1) They provide a film of low shear strength between the surfaces, thus
mitigating the effects of adhesion. This effect is particularly marked if
boundary lubricants or extreme-pressure additives are present in the
lubricant.
(2) They cool the surfaces, both by reducing the input of frictional energy
and by removing heat, thus reducing such deleterious thermal effects as
surface softening.
(3) In recirculating systems they carry the products of wear away from the
interface, thus preventing three-body abrasion of the surface by work-
hardened or oxidised debris.
Thus, wear coefficients are almost invariably reduced if the surfaces can
be continuously lubricated with a suitable oil. However, the prediction of
wear rates, if anything, is made more difficult than in dry rubbing. This is
because the oil film will normally protect some unknown fraction of the
surfaces, with the remaining fraction, comprising the higher asperities, making
solid contact through the film. Thus, the effective wear coefficient is the
resultant of one coefficient on the unprotected fraction of the surface, and
of a second coefficient on the protected fraction.
As a result of this additional complexity, hardly any wear coefficients
have been published for conditions of lubricated sliding. Recently Rowe
(1980) has suggested that wear coefficients can remain approximately
constant under some sliding conditions, and has made a plea for the
publication of more measured coefficients, with precise descriptions of the
conditions under which they were measured. However, it is likely that wear
prediction in lubricated sliding will rely on realistic laboratory tests and tests
under field conditions for the foreseeable future.
One particularly damaging type of wear under lubricated conditions is
scuffing, a form of wear in which localised solid-phase welding occurs between
sliding surfaces, resulting in damage and roughening of the surfaces. The
original area of scuffing is almost invariably roughened to such an extent
that it is likely to penetrate the oil film on subsequent contacts. Some scuffed
areas are self-healing - i.e. they undergo a period of mild wear which results
in smoothing of the roughened surface. However, in other cases the problem
becomes rather worse on each contact, with both the severity and the extent
of roughening increasing. In this latter case the damage can lead very rapidly
to gross damage and fracture of the sliding components. Thus, it is
characteristic of scuffing failure that it occurs very rapidly: mechanisms which
have been running very smoothly can suddenly become noisy, and failure
often ensues within minutes.
WEAR 95
Scuffing failure is often observed on components which are expected to

be separated by hydrodynamic or elastohydrodynamic films, particularly
important examples being piston rings and gear teeth. The oils used in such
situations always include boundary lubricants, which are expected to provide
protection on the occasions when the full-film protection breaks down. It is,
therefore, clear that the onset of scuffing is associated with the breakdown
of these boundary lubricants, but the factors which cause this breakdown
are by no means clear.
The subject has been admirably reviewed by Dyson ( 1975), who states
that the onset of scuffing is probably associated with the attainment of a
critical temperature at the contacts. However, Dyson states that the questions
which remain to be answered include: how scuffing is initiated; how initiation
proceeds to propagation; whether lubricants control scuffing; which chemical
reaction products are important; and what the functionally important
characteristics of the reaction products are. As Dyson concludes, these
questions give a good summary of the present state of ignorance of the subject.
3.10 PROBLEMS
1 In a wear test, a bronze annulus having an outside diameter of 25 mm

and an inside diameter of 15 mm is placed with its flat face resting on a
flat carbon steel plate under a normal load of 100 N and rotated about
its axis at 5 Hz for 20 h. At the end of the test, the specimens are separated
and weighed and it is found that the mass losses of the bronze and steel
are 250 mg and 8 mg, respectively. Using the materials data given below,
calculate wear coefficients for the bronze and the steel.
(Hardnesses (MPa): steel, 2400; bronze, 800.
Densities (Mg/m 3 ): steel, 7.8; bronze, 8.4.)
[Kbronze = 7.4 X w- 4 ; Ksteel = 6.57 X w- 5 ]
Chapter 4
Tribological Properties of
Solid Materials
4.1 METALS
Much of the material in the previous chapters has been concerned with the
tribological properties of metals, and it is not necessary to reiterate it here.
In particular, the effect of hardness on resistance to wear was mentioned
frequently in Chapter 3 and the importance of oxide films was stressed in
Section 3.5. However, in covering the basic mechanisms of wear some other
properties which can have a marked effect on wear behaviour were mentioned
only in passing, or not at all. The effects of these properties are described in
the following sections.
4.1.1 Work Hardening and Ductility
One effect of work-hardening characteristics was mentioned in Section 3.4.1,

where it was stated that the resistance to milder forms of abrasive wear can
increase with the work-hardening capacity of a metal. However, work
hardening combined with high ductility can lead to very poor tribological
properties. This is because high ductility can allow plastically deforming
asperities to undergo large amounts of plastic deformation before fracture.
If the plastically deforming asperities are work hardened to a hardness level
considerably above that of the opposing surface or, in the case of transferred
material, above that of the parent metal, they can then cause severe abrasive
wear. Furthermore, high levels of plastic deformation almost invariably cause
disruption of protective oxide films, thus leading to strong interfacial adhesion
and exacerbation of the situation described above. Therefore, materials which
combine high ductility with the ability to work harden often exhibit both
high friction and severe adhesion-enhanced wear, where material loss may
be slight but damage to the sliding surfaces is very severe.
96
TRIBOLOGICAL PROPERTIES OF SOLID MATERIALS 97
Important examples of such materials are austenitic stainless steels and

alloys of titanium. These materials have poor tribological properties in many
situations, particularly where the loads are very high. For this reason, the
materials are notoriously difficult to machine, owing to pick-up (adhesive
transfer of the material to the cutting tool) and consequent blunting of the
tool and damage to the workpiece:However, the austenitic steels may have
good tribological properties under lighter loads, particularly at high
temperatures in oxidising environments. This is because such steels naturally
form tenacious, protective films of chromium oxide and, if the loads do not
cause disruption of the films, can thus provide an excellent sliding surface.
4.1.2 Effect of Crystal Structure
In Chapter 3 we implicitly treated metals as isotropic materials which would

yield and plastically deform at a certain stress level. We took no account of
anisotropy (i.e. the variation of mechanical properties with crystallographic
direction) or of the direction of the stress axis relative to the crystal axes.
These simplifications do not introduce any significant error in describing
the tribological behaviour of many common metals, particularly those with
face-centred cubic (FCC) and body-centred cubic (BCC) crystal structures.
These materials are almost isotropic and have a multiplicity of slip systems,
which allows them to yield at similar stress levels, whatever the crystallographic
orientation of the stress axis. (Readers who are unfamiliar with the terminology
used here should consult one of the standard texts on the plastic deformation
of metals, such as Honeycombe, 1968.)
However, there is one important class of metals for which these
assumptions are far from justified, and this leads to important, and usually
beneficial, consequences for their tribological properties. These metals are
those with close-packed hexagonal (CPH) crystal structures which deform
predominantly by basal plane slip, the major examples being cobalt and
rhenium.
As these metals have only one slip plane, they cannot deform by slip
when the stress axis is either .perpendicular or parallel to that plane, and can
thus support very high loads elastically. Furthermore, even when the stress
axis is oriented so that slip can occur, the anisotropy precludes the isotropic
plastic deformation which is necessary for junction growth. Thus, such
materials can have acceptably low friction coefficients, even in ultra-high
vacuum, when their surfaces are perfectly clean and there is no possibility
of the gross seizure which can occur with more isotropic metals in such
conditions.
The effect of crystal structure is demonstrated very clearly by the example
of cobalt, which undergoes an allotropic change from CPH to FCC at 417°C.
As shown in Figure 4.1, at the point where this transformation occurs both
the friction coefficient and the wear rate increase dramatically. (Note that
98 TRIBOLOGY
1·6
1·2
g
c
£
0
c!!. 0·8 Wear rate
~ 3·5xi0- 7 mm 3/m
8
u
0·4
0 100 200 300 400

Ambient temperature (°C)
Figure 4.1 Friction coefficient and wear rate for cobalt on cobalt in vacuum at
various temperatures: pressure, 10- 7 N/m 2 ; sliding speed, 2 mjs; load, 9.81 N
in Figure 4.1 the ambient temperature at the transition is less than the quoted
temperature of 417 oc; this is because frictional heating causes the actual
surface temperature to be higher than the ambient temperature.) Fortunately,
this allotropic transformation temperature can be increased by additions of
suitable alloying elements. For this reason many wear-resistant materials are
based on cobalt-30% chromium, which retains the CPH structure up to a
temperature of > 800 oc.
4.1.3 Effects of Microstructure
In Chapter 3, where we described the basic mechanisms of wear, we largely

confined our attention to homogeneous, single-phase materials. We did this
because, with such materials, each asperity contact will be between the same
two materials and the mechanical properties of all the asperities can be
assumed to be similar to those of the bulk parent materials. This allows us
to develop clear descriptions of the effects of properties such as material
hardness. However, by adding other elements to a material - i.e. by alloying
- it is possible to change material properties in various ways, often with
beneficial effects on wear behaviour. It would be beyond the scope of this
book to detail the effects of alloying, but for an excellent introduction to the
subject the reader is referred to Ashby and Jones ( 1986). However, for our
purposes, we can divide the mechanisms into two groups: those resulting in
single-phase alloys, and those resulting in multiphase alloys.
There are two principal effects which can cause a single-phase alloy to
be more wear-resistant than the parent metal: solution hardening and
transformation hardening. These effects have recently been reviewed by
Glaeser ( 1980) and (particularly for the technologically very important

ferrous metals) by Borik ( 1980 ). The more important of the two effects is
transformation hardening, particularly in the heat treatment of steels to cause
a martensitic transformation. This transformation can lead to increase in
hardness of almost an order of magnitude and a correspondingly great
reduction in ductility, both of which will cause improvements in wear
behaviour. However, the brittleness of martensitic materials means that they
are prone to tensile failure. Therefore, they must either be used in compression
or tempered to improve their toughness, at the expense of some reduction
in hardness and increase in ductility.
Multiphase alloys for wear-resistant applications usually consist of a
relatively soft matrix containing a fine dispersion of relatively hard particles.
This type of structure can confer two principal beneficial effects on tribological
properties. First, the fine dispersion can limit the area over which junction
growth can occur and can thus reduce both friction and adhesion-enhanced
wear; for this reason it is a rule of thumb that the improvement in tribologicaJ
properties increases with the fineness and concentration of the dispersed
second phase. Second, the hard particles can confer excellent abrasion
resistance when they are harder than any potentially abrasive particles in
the environment. However, they can significantly worsen the situation if they
are softer than the external abrasive, as they can be loosened from the parent
matrix and cause additional abrasive wear.
The cobalt-based superalloys combine the crystal structure effect
described in sub-section 4.2.2 with precipitation effects due to a fine dispersion
of hard carbides; these materials have excellent tribological properties in a
wide range of applications.
4.1.4 Mutual Solubility of Rubbing Pairs
Rabinowicz ( 1970) has suggested that pairs of materials with a high degree
of mutual solubility will have poor tribological properties, as a consequence
of high interfacial adhesion resulting from their mutual interdiffusion and
bonding. He attempted to correlate mutual solubility, as indicated by binary
alloy equilibrium diagrams, with adhesion, friction and wear. The analysis
showed zero correlation with adhesion; some positive correlation with
friction; and a greater correlation with wear. This suggests that pairs of
metals for boundary lubricated or unlubricated sliding should be chosen to
have low mutual solubilities. This is a generally accepted rule of thumb but,
as with most tribological rules, there are some notable exceptions. For
example, the rule suggests that the same material should not be specified for
each component of a rubbing pair, but some materials, particularly the
cobalt-based superalloys mentioned above, have excellent tribological
properties against themselves. Recently Tabor ( 1985), Buckley ( 1985) and
Ferrante et al. ( 1985) have shown that the situation is much more complex
100 TRIBOLOGY
than is suggested by this simple rule, and that further understanding of

interfacial adhesion will need to be based on a better understanding of such
factors as the electronic structure of the interface and the degree of misfit
between the crystal lattices of the two surfaces. It is worth noting that two
lattices with a high pegree of misfit are likely to have low mutual solubilities,
so it may be difficu'lt to separate these effects.
4.~ .5 Effects of Temperature on the Above Properties
It is clear that a change of temperature could have a profound effect on most

of the properties described above. As the temperature is raised, any
single-phase materials, including those with martensitic structures, will tend
to become softer and more ductile; rates of oxidation will increase and the
type of oxide formed may change; hard precipitates may be dissolved, or
grow and become less effective; crystal structures may change; and mutual
solubility may increase. Therefore, it is obvious that, when choosing rubbing
pairs, it is necessary to choose them on the basis of the material properties
obtaining at the predicted operating temperature. As an example, for
operation at elevated temperatures it is usually necessary to specify materials
with high hot hardness, such as tool steels or superalloys.
4.2 BEARING ALLOYS
We use the term 'bearing alloys' to describe alloys developed for use in
well-lubricated journal bearings. As will be shown in Chapter 6, under normal
running the bearing and its counterface- often a crankshaft- are completely
separated by a thin oil film which is generated by hydrodynamic lift. Under
such conditions the tribological properties of the solid material are, of course,
irrelevant. However, the oil film is not present for a short time after each
start of the engine, or when the engine is running slowly at high loads. Under
these circumstances the surfaces are protected only by boundary lubrication,
as described in Chapter 2, and as boundary lubrication involves some
solid-solid contact and high asperity stresses, the tribological properties of
the materials are then very important.
4.2.1 Required Properties of Bearing Alloys
When we use bearing alloys, we are not usually in the fortunate position of
being simply concerned with two chosen materials rubbing together in the
presence of a film of clean oil containing boundary lubricants. For example,
the inside of an internal combustion engine is a very hostile environment,
containing corrosive combustion products, abrasive wear debris and usually
-particularly in new engines- cast-iron dust left behind after machining the
engine block, and wear debris from the piston, cylinder and piston rings
formed during running in. The rubbing materials must, therefore, be able to
withstand most of the wear processes described in Chapter 3. Furthermore,
the bearing must sustain very high fluctuating dynamic loads, which are
inherent in the operation of an engine, without suffering fatigue failure. These
various, and often conflicting, requirements have led to the development of
ranges of bearings which often contain several components. The compositions,
properties and methods of manufacture of the various types of alloy cannot
be described in detail here and the reader requiring more detail should see,
for example, the excellent review by Morris ( 1967) or The Tribology Handbook
(Neale, 1973). We shall simply describe the general requirements of such
bearings and briefly mention some examples. The required characteristics
are described in turn below.
Embeddability
As mentioned above, engines often contain abrasive particles, and, if these

are larger than the oil film thickness at its thinnest, they cannot pass through
the bearing. Thus, unless both the bearing and the crankshaft are considerably
harder than the abrasive particles, there is the potential for abrasive wear
to occur. This problem is solved by making the surface of the crankshaft
hard enough to resist abrasion (using one of the processes described below
in Section 4.7) and making the bearing alloy so soft that the abrasive particles
can be pushed into the bearing surface, where they are permanently trapped
and rendered harmless. This ability of the alloy to trap abrasives is known
as embeddability.
The reader may wonder why the abrasion problem is not solved by
making both rubbing surfaces harder than the abrasive. This is explained by
the need for other characteristics, as described below.
Conformability
It would be extremely difficult and expensive to manufacture a crankshaft

and a set of bearings with no significant misalignments between the bearings.
This problem is also solved by making the bearing alloy so soft that it can
deform plastically on assembly and early running, and thus accommodate
any initial misalignments. This attribute of the bearing is known as
conformability.
Protection of the Crankshaft
A crankshaft is a relatively expensive component to manufacture and install,

compared with a bearing shell, and the crankshaft-bearing system should,
therefore, be designed so that wear is largely confined to the bearing shell,
102 TRIBOLOGY
with the crankshaft remaining relatively unscathed. It should now be clear

from reading Chapter 3 that this is best accomplished by making the rubbing
surface of the bearing shell much softer than that of the crankshaft.
Furthermore, the soft metals chosen for the bearing surface will also have
low melting points, so in the extreme case of complete oil starvation the
surface of the bearing can melt and lubricate the rubbing surfaces, thus
preventing catastrophic engine seizure.
Load-carrying Cap{Jcity and Fatigue Strength
So far, all our requirements have led to the choice of a soft material for the
bearing shell and, at first sight, this conflicts with the need for the bearing
to carry very high fluctuating loads without gross plastic deformation or
fatigue failure. However, we can have all the properties listed above and good
mechanical strength by making the soft metal very thin. The reason for this
is that when a very thin film of soft metal is supported on a hard base, it
can carry very high normal pressures without yielding. For example, El-Shafei
et al. ( 1983) have shown that thin lead films on hard steel substrates can
sustain the same pressure as the uncoated steel before yielding. This is because
the film is trapped between two extensive hard surfaces and it can deform
plastically only by flowing outwards so that some material leaves the contact
zone. As the film becomes thinner, the pressure required to cause this outward
flow increases very rapidly, and tends to infinity at zero film thickness. This
effect of plastic constraint is used in bearing design by supporting a thin and
soft bearing material on a relatively strong and hard steel backing strip.
4.2.2 Compositions and Structures of Bearing Alloys
The two metals which combine best the required properties of softness, low
friction against steel and low melting point are tin and lead. However, in
order for these materials in the unalloyed form to have the required
load-carrying capacity they would need to be no more than a few microns
thick, and this would both limit the lifetime and restrict the embeddability
and conformability of the bearing. These metals are therefore universally
used in two types of alloy: first, as lead-based or tin-based alloys, in which
properties such as hardness and strength are enhanced considerably by small
alloying additions; second, as dispersions in a matrix of a stronger material.
The two most common materials of this second type are copper-lead and
aluminium-tin alloys; in each of these the zones of the softer material
interconnect, so that any lead or tin lost from the rubbing surface can be
continuously replenished from the underlying material. Typical micro-
structures of alloys of such materials are shown in Figure 4.2.
The copper-lead and aluminium-tin alloys are stronger and harder
than those based principally on lead and tin. However, the lead component
Figure 4.2 (a) Microstructure of an aluminium-tin-based bearing alloy (courtesy,

Glacier Metal Co. Ltd, Kilmarnock); (b) microstructure of a cast copper-lead bearing
with an electroplated lead-indium overlay (courtesy, Vandervell Ltd, Maidenhead)
104 TRIBOLOGY
of the copper-lead bearing is prone to corrosive attack by both decomposition

products from the lubricating oil and combustion products from the fuel. In
lead-based alloys these effects can be reduced by alloying additions of tin or
antimony, but the same techniques cannot be used with copper-lead alloys,
as the additions alloy preferentially with the copper and leave the lead
unprotected. This problem is normally overcome by applying a thin coating
or 'overlay' of a corrosion-resistant lead-based alloy. Such coatings are not
necessary for corrosion protection of aluminium-tin bearings, which have
inherently good corrosion resistance, but they are normally applied to impart
the other beneficial effects of low friction and embeddability. The overlay
coating, though only about 30 11m thick, normally lasts for much of the
required bearing life, because of the plastic constraint effect described above.
However, when the coating does wear away, the underlying material still
allows a further period of useful life.
In summary, a bearing shell for an application involving high loads
normally consists of a strong steel backing strip, a copper-lead or
aluminium-tin layer ~0.4 mm thick and a soft overlay ~30 11m thick. For
lower loads, the bearing may simply consist of the steel backing strip with
a lead-based or tin-based alloy coating.
4.3 SOLID LUBRICANTS
There is no doubt that the ideal way to reduce both friction and wear of
sliding surfaces is to arrange for them to be completely separated by a fluid
film, and the various ways in which this can be done will be described in the
following chapters. However, even the best fluid lubricants have limitations
which preclude their use under certain circumstances. Their main limitations
are described below.
4.3.1 Limitations of Liquid Lubricants
The main limitations of liquid lubricants are as follows.

( 1) They can be used only over a limited temperature range; they become
thin, and decompose or oxidise, at high temperatures, while at low
temperatures they become very viscous or even solidify.
(2) Unless they are externally pressurised, they cannot provide complete
surface separation at high loads or low speeds; under such circumstances
they can provide only boundary lubrication.
( 3) They can often not be used in environments where cleanliness is essential,
such as in food processing equipment.
(4) They are prone to attack in hostile environments such as acids and
solvents.
(5) They cannot be used to provide permanent lubrication for mechanisms

which are inaccessible after assembly.
(6) They are degraded by radioactive environments.
These limitations can often be overcome by using solid lubricants.
Furthermore, when solid lubricants can be used, they remove the need for
expensive lubricating systems and they can greatly reduce the need for periodic
maintenance.
4.3.2 Requirements of Solid Lubricants
The basic requirement of any solid lubricant is exactly the same as that for
a liquid lubricant. It should completely separate two solid surfaces, even
under a very high normal load, while allowing the surfaces to slide over each
other under a relatively low shearing force: in other words, a solid lubricant
must meet the twin requirements of having a high normal load capacity and
a low shear strength.
We have already described one type of solid lubricant in sub-section
4.2.2, where we showed that these twin requirements can be met by a thin
film of a soft metal supported on a relatively hard substrate. However, this
type of solid lubrication is a property of the multicomponent system, rather
than an inherent property of a single material, and in this section we intend
to describe single materials which can act as solid lubricants.
It is clear that the twin requirements described above are unlikely to be
met by any isotropic material, in which the shear strength will be about 16%
of the yield pressure, and it has been found that the best solid lubricants are
very anisotropic materials known as lamellar solids.
4.3.3 Lamellar Solids
A lamellar solid is one in which the atoms are bonded together in parallel
and comparatively widely spaced sheets, the two best-known being graphite
and molybdenum disulphide, which have the crystal structures shown in
Figures 4.3 and 4.4, respectively. Under many circumstances these are
excellent lubricants, as are other less well-known lamellar solids such as
tungsten diselenide and cadmium iodide. However, not all lamellar solids
are good solid lubricants, and there is as yet no theory which will tell us
whether or not any particular lamellar solid will be an effective lubricant.
The lamellar solids which can lubricate effectively do have certain
characteristics in common. First, they form a strongly adherent transfer film
on the surface being lubricated, so that, after the short running-in period
during which this film is formed, the actual interface is between lubricant
and lubricant. Second, the lubricant on both surfaces develops the preferred
orientation shown schematically in Figure 4.5, and this reduces the mechanical
106 TRIBOLOGY
-tf~~?1,
~~~,,r ll
f
I
tf
I I I I I I I
I I I I 1 I
~. ~Jff-:;t.
ljl
I1 1I I
I
I
1
I I
11 I
I
I I
I
I I I . I I I
~r?tit 7:J
Figure 4.3 Crystal structure of graphite
e Molybdenum
0 Sulphur
Figure 4.4 Crystal structure of molybdenum disulphide
interaction between the surfaces, as can be demonstrated by reversing the

direction of motion, when the friction increases significantly. -
Although there have been many experimental investigations of the
lubricating properties of graphite and molybdenum disulphide, there is no
universally accepted explanation of their behaviour.
The earliest explanation of the properties of graphite (Bragg, 1924) was
that its shear strength parallel to the atomic sheets was inherently very low.
This explanation was accepted until World War II, when it was found that
graphite brushes in the electrical generators of high-flying aircraft wore out
very rapidly. This effect was investigated by Savage ( 1948), who showed that
Figure 4.5 Schematic illustration of the surface orientation developed during the
rubbing of lamellar solid lubricants
graphite has very poor tribological properties in the absence of condensable

vapours. This effect is very marked and the amount of vapour required to
give effective lubrication can be very low. The effectiveness of any vapour
seems to depend on the size of its molecules. Nitrogen is ineffective, even at
atmospheric pressure; water provides very effective lubrication at a pressure
of 400 Pa; and many long-chain hydrocarbons are effective at pressures as
low as ~ 1 Pa. The transition in behaviour is startling: at the critical pressure
the wear rate can change by about three orders of magnitude and the friction
coefficient by a factor of 5. It is now thought that the condensable vapours
saturate any available valence bonds at the surfaces and edges ofthe individual
crystallites, so that the adhesion between neighbouring crystallites is very
low and they can easily slide over each other. This explanation has been
supported by the work ·of Arnell et al. (1968), who have shown that even
strong and brittle mechanical carbons and pyrolitic carbon exhibit the same
low-friction behaviour as graphite.
Unlike graphite, molybdenum disulphide appears to be an intrinsically
good lubricant, and its friction coefficient under high-vacuum conditions is
lower than that in air. There are alternative suggested explanations of this
inherent low friction. The first is that it is a true low-shear-strength material,
with shear taking place between the neighbouring layers of sulphur atoms
shown in Figure 4.4. The second, which is similar to that now accepted for
graphite, is that any available surface and edge valences are saturated by
reaction with oxygen atoms but, unlike the vapours on graphite, the oxygen
is not volatile and persists up to the decomposition temperature of the solid.
108 TRIBOLOGY
4.3.4 Methods of Use
Solid lubricants can be used in several different ways, the most common ones
being described below.
Dry Powders or Dispersions in Fluids
Surfaces to be lubricated can be rubbed with dry powder, to form burnished

films. Alternatively, the lubricant can be applied as a dispersion in a volatile
solvent which subsequently evaporates, to leave a lubricant film. Such films
would have only a short lifetime under continuous rubbing, but they are
used extensively to facilitate assembly of close-fitting parts, to lubricate metal
working components and as parting agents for screw threads.
Solid Blocks
Graphite and graphitic carbons are often bonded together in solid blocks
which can, for example, be made into solid thrust bearings and journal
bearings. The blocks are usually made from a mixture of finely divided coke
and a carbonaceous binder such as pitch. This mixture is fired to a very high
temperature, to cause it to graphitise, and by varying the heat treatment a
wide range of materials can be made. These materials range from highly
crystalline electrographite, which is generally used for low-load applications
such as electrical generator brushes, to almost amorphous mechanical
carbons, which are used for bearings.
Additives to Oils and Greases
The major use of molybdenum disulphide is probably as an additive to oils

and greases. In this form it is very beneficial during running-in operations
and it increases the effectiveness of lubrication under heavy loads. When
used in engine oil, it is deposited as a thin solid film at the piston-cylinder
interface, and this can offer some short-term protection in the event of
complete oil loss. If such additives are used in lubricating systems containing
a filter, it is essential that the particle size be sufficiently small to allow the
particles to pass through the filter, which would otherwise quickly become
clogged.
Resin-bonded Films
Solid lubricants are frequently bonded to metal surfaces by use of resin

binders, and such films can have wear lives some orders of magnitude higher
than those of dry powder films. The only drawback of these films, compared
with dry powder films, is that thermal decomposition of the resin imposes
z.o
ME 1·5
E.
1!
g
~
ill
~
...~
10 15 20 25
Weight % MoS 2
Figure 4.6 Wear of powder metallurgy specimen containing molybdenum

disulphide
a limit on the maximum rubbing temperature, and therefore on the load and
sliding speed.
Metal-based Composites
The thermal limit mentioned in the previous paragraph can be overcome by

incorporating the lubricant in a metal matrix. The relative proportions of
matrix and lubricant are quite critical, as can be seen from Figure 4.6. This
figure illustrates the necessity of careful control of materials content if the twin
requirements of high load-carrying capacity and low shear strength are to
be met. If the composite carries too little lubricant, it has a very high
load-carrying capacity but also a high shear strength, whereas, if it carries
too much lubricant, it has a low shear strength but also a reduced
load-carrying capacity.
4.4 POLYMERS AND COMPOSITES
Polymer-based bearings can have all the advantages of solid lubricants listed
in the introduction to Section 4.3, plus the following additional advantages.
( 1) They absorb vibrations well and are quiet in operation.
(2) They readily deform to conform to mating parts, so that machining
tolerances and accuracy of alignment are usually less critical than for
metal parts.
110 TRIBOLOGY
(3) They are easily formed into complex shapes, by either machining or
moulding.
( 4) They are cheap.
There is a very wide range of commercially available polymer-based
bearing materials, and they cannot be described in detail here. The aim of
this section is to describe in principle the tribological properties of polymer-
based materials and the various operating factors which can influence these
properties.
4.4.1 Tribological Properties of Pure Polymers
The frictional behaviour of polymers has already been discussed in

sub-section 2.6.2. It is known that, for most polymers, the friction force
between a polymer surface and a steel slider is as predicted by the simple
Bowden and Tabor theory of friction (Section 2.4)- i.e. the friction force is
approximately equal to the bulk shear strength of the polymer multiplied
by the effective area of contact. However, shear strength is not a materials
constant for a polymer. Polymers are viscoelastic, and their deformation is,
therefore, dependent on strain rate; consequently, friction forces can vary
with such parameters as sliding speed and surface roughness.
However, the friction coefficients of most polymers, against metals and
against themselves, are usually in the range 0.2-0.4. Polytetrafluoroethylene
(PTFE) is an outstanding exception: the friction coefficient for PTFE sliding
on itself can be as low as 0.04, which is the lowest known value for any solid.
There is no firm explanation for this phenomenon, although it is usually
ascribed to the intrinsically low chemical bonding forces between PTFE
molecules, which consist of relatively stiff chains of carbon atoms surrounded
by larger fluorine atoms. Therefore, molecules at the surfaces can slide over
each other under very low shear forces, while mechanical interlocking of the
molecules in the bulk material gives it a relatively large load-carrying capacity
and bulk shear strength. As already described in sub-section 2.6.2, when
PTFE slides against a counterface of normal roughness, a thin film of the
polymer is transferred to the counterface, so that the observed friction
coefficient is due to the polymer sliding on itself.
We can see that, with the exception of PTFE, polymers do not have
particularly low coefficients of friction. The tribological property of polymers
which causes them to be so widely used is that they wear at low and reasonably
predictable rates, thus allowing a designer to select with some confidence a
bearing which will give the required operating life under specified conditions
of load and speed. We shall return to this subject in sub-section 4.4.3.
4.4.2 Effects of Reinforcements and Fillers
There are only five basic polymers in common use in tribology: PTFE,
polyacetal, Nylon, polyethylene and polyimide, although other materials are
now coming into use. However, many reinforcements and fillers have been
incorporated into these basic materials, so that a wide range of bearings is
commercially available.
As their name implies, the purpose of reinforcements is to increase the
mechanical strength of the bearing. Reinforcement is carried out in one of
three basic ways. First, a mechanically strong, normally fibrous material is
mixed with the polymer before moulding, the most common example being
chopped glass fibre, which increases strength, stiffness and creep resistance.
Second, the polymer is incorporated into a woven matrix; with thermosetting
resins the woven matrix is soaked in the resin, which is then allowed to
harden, whereas with thermoplastic poly~ers, such as PTFE, the polymer
fibre and reinforcing fibre are woven together in such a way that the actual
bearing surface is predominantly the lubricating polymer. Finally, the polymer
can be supported as a relatively thin layer on a metal backing plate, to give
the plastic constraint and increased load-bearing capacity already described
for bearing alloys in sub-section 4.2.2. If the polymer is also supported around
its edges, so that it is effectively held in a metal container, its load-carrying
capacity can be increased enormously. Pure PTFE has very poor creep
resistance and, if unconstrained, will creep under very low loads, but if fully
constrained, as described above, it can be used to carry very high loads -
for example, in civil engineering applications such as bridge bearings.
A metal liner on a thin polymer bearing has the additional advantage
of greatly assisting heat dissipation from the sliding interface. In general,
polymers have low thermal conductivities and high thermal expansion
coefficients when compared with metals, and, therefore, if solid bushes of
pure polymer are used against steel journals, frictional heating can cause
bearing seizure. To avoid this, it would be necessary to make the polymer
bearing to a much looser fit than the designer would otherwise wish. However,
the problems of low conductivity and high expansion coefficient can both
be avoided by using the polymer as a thin film on a metal liner.
Fillers, which are particulate materials, are used to improve strength,
thermal properties and frictional behaviour. Metal powders of high conductivity
and relatively high strength are used to improve thermal and mechanical
properties, while graphite, molybdenum disulphide and soft metals such as
lead are used to improve frictional behaviour.
4.4.3 The PV Factor
The rate of wear of a polymer will obviously be some function of the load
and sliding speed, and it is found that we can define for any polymer bearing
a design criterion known as the PV factor.
The theoretical derivation of the relationship between the wear rate and
the PV factor starts from the reasonable assumption that the rate of wear
will be proportional to the rate of energy dissipation at the sliding interface.
On this basis, we can then derive a relationship between the rate of wear
112 TRIBOLOGY
-
Load
Shdtng veiOCtly V
Beanng
, , ..
area A
, ,,
,,
Figure 4. 7 Schematic diagram of flat bearing for derivation of P V factor
and the PV factor for the two basic bearing configurations.

( 1) Flat bearings If we have a flat bearing surface of area A, as shown in
Figure 4.7, subjected to a normal load W and sliding against its
counterface under a coefficient of friction J.l, then the energy dissipated
in sliding through a distance dx is given by J.l W · dx, and the rate of energy
dissipation is given by J.l W( dx I dt) = J1. WV, where V is the sliding speed
at the interface.
We can then state our assumption, that the volume wear rate, Q,
is proportional to the rate of energy dissipation, as
Q=kWV { 4.1)
where the coefficient of friction, J.l, which, for the purpose of this
derivation, we have assumed to be constant, has been subsumed into
the constant of proportionality, k.
By rearranging Equation (4.1 ), we can see that the constant of
proportionality, k, is the volume wear per unit load per unit slid distance;
therefore, k is known as the specific wear rate of the material.
We are normally interested in the depth of wear, or the rate oflinear
wear normal to the sliding surface, and we can see that the depth of
wear at any time is simply the volume of wear divided by the bearing
area. Therefore, the rate of linear wear, Qd, is given by Q/ A, so that
{4.2)
where P is the apparent pressure on the bearing.
(2) Journal bearings Figure 4.8(a) shows a journal bearing supporting a
shaft, under a normal load W, on the shaded bottom half of the bearing.
We assume that we have a normal reaction, R, per unit area of the load-
Shaft
Load supported
on shaded hoff
!al of bearor>g
(b)
Figure 4.8 Schematic diagram of solid lubricated journal bearing and forces acting
bearing surface. By considering the equilibrium of vertical forces, we see

that the normal load W must be kept in equilibrium by the vertical
component of this reaction integrated over the bearing area. From Figure
4.8(b) we can see that the reactive force over an element of surface of
length ds will be given by Rl·ds, where I is the axial length of the bearing,
and the vertical component of this reaction will be Rl cos O·ds. But
ds = D ·dO 12, so that the total vertical reaction is given by
12
(RIDI2)f" cos OdO =RID
-n/2
Therefore R = wI Dl. But the rate of energy dissipation per unit area of
interface is p.PV, so that the total rate of energy dissipation is given by
p.RV·nD/12
= (p. WI Dl)· V·( nDl/2)
= p.WVn/2
Therefore, the volume wear rate, Q, is given by Q = kWVn/2, where, as
before, p. has again been subsumed into the constant.
114 TRIBOLOGY
But the radial wear rate, Qd, is again given by the volume wear rate
divided by the contact area - i.e.
Qd = Q/(nDl/2) = kWV I Dl
so that Qd = kPV ( cf. Equation 4.2 ), where Pis the load per unit projected
area.
Thus, we can see that we expect the same proportionality between linear
wear rate and the PV factor for both types of bearing, provided that P is
taken as load per unit projected area, where the constant of proportionality,
k, is the specific wear rate of the material.
Equation (4.2) suggests that the rate of wear of a polymer bearing should
be proportional to the product PV, and to a good approximation it is, indeed,
often possible to specify a PV value which must not be exceeded, to give a
certain wear rate or life expectancy. Also, for any material there is a limiting
PV factor above which the material would fall very rapidly, owing to melting
or thermal decomposition, and this is known as the PV limit of the material.
The use of single PV factors is open to some criticism, since it assumes
the same sensitivity of wear rate to changes of P and V. This assumption is
frequently unfounded, and a more exact practice is to give plots of acceptable
wear rates on PV diagrams, as shown in Figure 4.9. This diagram shows
that the PV factor for the chosen wear rate is almost constant, except at the
extremes of load and speed. This is often the case, although some materials
have PV diagrams which are continuous curves over the whole PV range.
It is also found that there is a limited range of low pressures and
temperatures over which k can be taken to be constant, and within this
region k is given the symbol k 0 . Thus, within its range of applicability,
Equation (4.2) can be used to predict wear rates; it is only necessary to
measure the wear rate at one value of PV to determine the value of k 0 , and
this value can then be used to predict wear under other PV conditions.
At higher temperatures or pressures the value of k increases, and for
accurate data recourse must be had to detailed design manuals. In particular,
the ESDU Guide to the Design and Material Selection for Dry Rubbing Bearings
( 1987) should be regarded as an indispensable guide to any designer involved
in the specification of such components. This guide gives not only PV data
of the type described above, but also detailed information on the effects of
other variables such as the friction coefficient, which we initially assumed to
be constant, the counterface composition and roughness, the presence of
lubricants, and the thermal properties of the bearing assembly.
The PV factors or PV curves which are published or supplied by bearing
manufacturers normally refer to unlubricated sliding against a counterface
of a specified material having a specified surface finish. When the sliding
conditions depart from these standard conditions, the designer must take
into account such changes when specifying a bearing, and to do this it is
essential that a more detailed guide, such as the ESDU guide, be consulted.
Velocity lm/s)
Figure 4.9 Typical limiting PV curve of PTFE-based material for wear rates of
25~tmin 100 h
4.5 Materials for Brakes and Clutches
The materials used in the manufacture of brakes and clutches are known as
friction materials. Braking is usually by far the more demanding of these
applications, and in what follows we shall concentrate on brakes, although
the same principles will also apply to clutches.
In contrast to all other tribological applications, high friction coefficients
are desirable when using friction materials, and the materials are normally
developed to give friction coefficients, against the chosen counterface, of the
order of 0.35-0.55. The principal frictional requirement is that the friction
coefficient should be stable with respect to sliding speed, time and temperature,
to avoid the brakes fading during a prolonged application, but a high friction
coefficient ensures that the required braking force can be obtained under a
relatively low normal load.
116 TRIBOLOGY
In addition to the relatively high friction coefficient, the materials must

also meet the following requirements.
( 1) Thermal properties The purpose of a brake is to convert kinetic energy
into heat. The brake must be able to absorb all the heat dissipated
without suffering catastrophic damage or losing its effectiveness. Therefore,
the brake materials must have the thermal stability to withstand the
highest temperature rise which is likely to be caused, and this temperature
rise is minimised by selecting materials with high thermal capacity and
conductivity.
(2) Wear resistance Ideally, the wear of both rubbing components in a
brake should be low, but, in practice, materials are often selected so that
one material, which is relatively cheap and easily replaced, suffers almost
all of the wear, while the relatively expensive counterface suffers little
damage.
( 3) Coriformability Very high local temperatures are generated at the points
of contact, and to ensure that the general heating of the brake materials
is uniform, it is necessary that the contact spots be uniformly spread
over the apparent contact area. Therefore, the brake material should
conform well to its counterface, and this requires the material to have
a low modulus.
(4) Mechanical properties The materials must have the required strength
and toughness to take the highest anticipated load without suffering
mechanical failure.
(5) Weight The brake should have the lowest weight which is compatible
with the other properties listed above. This is particularly important in
aircraft braking, to avoid weight penalties, and in railway braking, where
the brakes are usually part of the unsprung mass, to ensure a smooth ride.
By far the most common use of friction materials is in the brakes of
road vehicles. These comprise a polymer-based composite friction material
rubbing against a cast-iron counterface. The friction material component,
the brake shoe or brake pad, is cheap and easily replaceable and is normally
replaced several times during the life of the vehicle, while the counterface,
the drum or disc, is normally not replaced.
The friction material is a multicomponent composite with a formulation
which is developed empirically. The material will normally comprise:
A matrix, of a thermosetting polymer, which confers conformability.
A fibrous reinforcement, to give the required strength and thermal
stability. This has traditionally been asbestos but, for health and
environmental reasons, this is now being superseded by alternative
materials such as chopped steel wool.
A filler, to give frictional stability, usually barium sulphate.
High-temperature lubricants, which are designed to melt and lubricate
hot spots, examples being lead sulphide, zinc and aluminium.
A scavenger, which keeps the counterface free of transferred material,

usually brass chippings.
Such materials are in a constant state of development and it is not
uncommon to find friction materials comprising more than ten components.
In railway braking the traditional method has employed a cast-iron
shoe rubbing on the steel tyre of the wheel. However, there has been a
tendency in recent years to use composite materials, of the type described
above, in preference to cast-iron. In either case the demands on the brakes
are so severe that it is normal to replace the brake shoes after some hundreds
of miles, rather than the tens of thousands of miles which are expected in
road vehicles.
The most demanding area of application for friction materials is in
aircraft braking. For example, to abort a take-off of a modern airliner, the
brakes would be expected to dissipate all the kinetic energy of the aircraft,
perhaps 100 MJ, in about 30 s. Because of the short time involved, very little
of this heat can be lost from the brake, and the materials must be capable
of absorbing it all. Therefore, polymer-based materials are out of the question
for this application. Most modern aircraft use multiple brakes consisting of
alternate plates of steel and sintered iron. However, there is a large weight
penalty involved in the use of such materials, because the large thermal mass
required means that the brakes must have a mass of =::: 1 t. To reduce this
mass, friction materials are now being manufactured from a carbon-carbon
composite. The use of this new material can reduce the weight of the brakes
by about 40%, and can increase the replacement intervals. Nevertheless, the
materials are still very expensive and they are still used only where weight
saving has a very high priority.
4.6 CERAMICS AND CERMETS
The present state of knowledge on the friction and adhesion of ceramics has
already been discussed in sub-section 2.6.3, where it was made clear that, as
yet, little is known of the nature of interatomic bonding across ceramic-metal
and ceramic-ceramic interfaces. This is because, particularly with covalently
bonded ceramics, it is unclear how many valences are fully saturated and
how many are free to form interfacial bonds. However, it has been observed
that friction coefficients for ceramic-ceramic sliding are not particularly low,
being of the same order as those between metals in normal atmospheres.
On the other hand, it is well known that ceramics can exhibit very low
wear rates, compared with metals, and this property is causing ceramics to
be used increasingly in tribological situations. Furthermore, ceramics are
refractory materials which retain qualitatively similar properties up to high
temperatures, and they are also chemically very unreactive, so there is no
118 TRIBOLOGY
doubt that they will find increasing use in hostile environments, particularly
those involving high temperatures.
We shall discuss specific examples of the uses of ceramics a little later
in this section, but before doing that it is useful to discuss the reasons for
the good wear behaviour. This is best done by referring again to the plasticity
index, t/t, where
t/t=(E*IH)·(ulfJ)! ( 1.28)
As we stated in sub-section 3.4.2, surface contact between metals will be
almost entirely elastic when t/t < 0.6 and almost entirely plastic when t/t > 1.0.
However, while ceramics have similar elastic moduli (E) to those of metals,
they have very much higher hardness values (H). Thus, for similar surface
finishes (represented by u and /3) ceramics are much more likely than metals
to be subjected to fully elastic surface contact.
To clarify this point, it is useful to consider the ratio HIE from a different
standpoint. As H is a simple multiple of the yield stress of a material, this
ratio is a measure of the yield strain of the material - i.e. the strain which
can be accommodated by fully elastic deformation. Thus, materials with a
high HIE ratio, such as ceramics, can accommodate much higher deformations
elastically than can those with lower HIE ratios, such as metals; this is
obviously critically important in determining the mode of deformation as
asperities interact.
There is another crucial difference between ceramics and metals: ceramics
are brittle, whereas most metals are ductile. Thus, when most metals reach
the limit of elastic deformation, they can accommodate additional deformation
plastically, whereas ceramics are likely to fracture. It is true that ceramics
are capable of undergoing some plastic deformation when there are local
hydrostatic stresses, and such conditions often obtain at regions of asperity
contact. Nevertheless, if a ceramic asperity is subjected to an increasing load1
it is likely to fracture without significant plastic deformation. The higher
asperities on a sliding ceramic surface will, of course, be subjected to the
highest forces, and if the resulting wear debris can escape from the interface,
the surfaces should therefore become progressively smoother, the surface
interactions should become more elastic and the wear rate should decrease.
Despite the elastic deformation processes described above, the wear rates
of ceramics are never zero and we must ask what causes the residual wear.
The first point to note is that we must consider the situation in terms of
fracture mechanics. Ceramics are not homogeneous materials and they
invariably contain surface cracks and internal flaws. Any flaw above a certain
critical size is able to extend if the stress exceeds a critical level, which may
be well below the theoretical fracture stress of the material, and thus a short
stress cycle may cause a small incremental crack extension. On repeated
stress cycling, such incremental extensions can cause the flaw to extend to
the point where a loose wear particle is formed. Furthermore, this process
is likely to be exacerbated by the presence of a corrosive environment, which
causes more interatomic bonds at the tip of the flaw to be broken on each
stress cycle.
The propagation of a crack as described above may be responsible not
only for the creation of local fractures, but also for large-scale fracture of
the bulk material, and it is this which has, until now, restricted the use of
bulk ceramics in tribology. However, the situation is now being improved
in three ways.
First, ceramics technology is improving rapidly, so that materials which
are more homogeneous and have fewer and smaller flaws are now being
produced. Such bulk materials will be used increasingly in future and are
already being actively investigated for various applications. For example,
manufacturers of automobile engines are aiming to design all-silicon nitride
engines in which the engine block, pistons and connecting rods will be made
from bulk silicon nitride. If a suitable ceramic with the necessary mechanical
properties can be developed, the resulting benefits will be: excellent wear
resistance at the piston-cylinder interface; the possibility of a higher operating
temperature and therefore higher efficiency; and lower inertial forces, owing
to the low density of the ceramic.
Second, problems of bulk fracture are being avoided by the use of cermets
- i.e. materials which consist of mixtures of metal and ceramic. An example
of this is metal-bonded tungsten carbide- a mixture which is approximately
90% tungsten carbide with a metal binder. The ceramic grains are coated
with the metal, usually cobalt or cobalt-chromium alloy, and are then
sintered, to form a dense solid. The metal at the grain boundaries then
prevents any crack being propagated across a ceramic-ceramic grain
boundary and leads to the composite material being much tougher than a
pure ceramic. Such materials are already used extensively as, for example,
sealing rings in mechanical seals.
Third, methods have been developed in recent years for depositing
ceramic coatings onto metal surfaces; systems of this type will be described
in more detail in Section 4.7.
In closing this section, we should note that, even when the stresses are
very low and the flaws are unimportant, ceramics, in common with other
materials, will still wear at finite rates. It is thought that this residual wear
probably takes place on an atom-by-atom basis, as a combination of
mechanical, chemical and thermal energies makes it energetically favourable
for an atom to leave the surface.
4.7 SURFACE TREATMENTS TO REDUCE WEAR
It should now be clear that the tribological behaviour of any component

which is not completely separated from its counterface by a fluid film is
critically dependent on the properties of the contacting surfaces, and the
120 TRIBOLOGY
choice of contacting materials is therefore of paramount importance. However,

the material selected for its tribological properties may be very unsuitable
for all but the surface region of the component.
We now know that the wear resistance of a surface tends to increase
with hardness, and we should therefore choose materials of high hardness,
to resist wear. However, high hardness is often associated with brittleness
and low impact strength, and these properties may make any hard material
which would be chosen for wear resistance unsuitable for the bulk of the
component.
Conversely, materials which have low friction coefficients often have
low mechanical strength, and a material which would be selected as a
self-lubricating solid might be too weak to be used for the bulk of many
components.
Finally, materials with good tribological properties are often very
expensive and very difficult to fabricate, and the cost of using them to
manufacture bulk components could be prohibitive.
These factors lead us to the conclusion that many components to be
used in tribological applications should ideally be made from two materials:
a bulk material which meets requirements such as strength, ease of manufacture
and cost; and a surface material which provides the required tribological
behaviour.
We have seen examples of this independent choice of materials earlier
in this chapter, in the use of steel backings to support polymer-based bearings,
resin-bonded lamellar solid lubricants and soft bearing alloys. In the
remainder of this section we shall describe other ways in which these
independent choices can be made. We cannot cover the various processes in
detail, and readers requiring more information are referred to an excellent
review by Wilson ( 1975) and a more recent paper by Ramalingan ( 1980).
Following the treatment of Wilson, we shall separate the surfacing treatments
into two groups: short-term treatments, which are used to assist running in
and for short-term processes such as metal forming; and long-term treatments,
which are expected to last the full life of the component.
4.7.1 Short-term Treatments
Short-term treatments are treatments which interpose a thin, non-welding

layer between solid surfaces and thus allow the surfaces to run in without
adhesion-enhanced wear or the formation of hot spots which could impair
future performance.
Some films which have inherently short lives are self-regenerating. They
include: metal oxides; films formed by reaction of the surface with boundary
lubricants and EP additives; and graphite and molybdenum disulphide
deposited from oils and greases.
Other films are permanently lost after a short time, but during that time
they have the required beneficial effect, which may then last the life of the
component. The predominant film of this kind is formed by 'phosphating'.
Many commercial processes exist for forming layers of metal phosphate
crystals on the surfaces of metal, particularly steel and cast-iron, and these
films give good corrosion resistance, good short-term wear resistance and
good oil retention. The oil retention is particularly good on grey cast-iron
surfaces, as the phosphate film forms only on the metal parts of the surface
and leaves oil-retaining channels above the graphite flakes.
A second short-term film in very common use is the very thin film of
electroplated tin which is used on aluminium pistons to assist running in.
4.7.2 long-term Treatments
Long-term treatments can involve modification of the existing surface, with

or without changing its composition, or deposition of a second material onto
the existing surface. For clarity, we shall divide the treatments into three
groups.
Surface Treatments without Change of Composition
Steels containing more than ~0.3% carbon can be hardened by quenching.

(Readers requiring more information on the heat treatment of steels are
referred for an excellent introduction to Ashby and Jones, 1986, or for a
fuller treatment to Honeycombe, 1980.)
For wear resistance combined with toughness and strength, it is usually
beneficial to harden only the surface layers of a steel component, so that we
have the required combination of a hard, wear-resistant surface on a tough,
strong core. This can be accomplished by heating only the surface of the
component, either in a flame or by electrical induction, and then quenching.
The induction-heating process, in particular, is easily automated and cheap,
and is applied to many components such as gears. The process has the
additional beneficial effect that the outer hard martensitic layer is left in
compression and thus has improved resistance to processes such as fatigue
and stress corrosion.
As an alternative to heat treatment, the surfaces of many metals can be
hardened mechanically by shot peening - i.e. bombarding with small metal
shot at high speed. This is essentially a work-hardening process which also
has the advantage of leaving compressive stresses in the surface layers.
Surface Treatments with Change of Composition
There are two basic ways of changing the composition of the surface layers
of a material: by diffusion or implantation.
In diffusion treatments the component is held at high temperature in a
specified environment, so that atoms from the environment diffuse into the
122 TRIBOLOGY
outer layers of the component. The principal atoms which are used to modify
the surfaces of steels are carbon and nitrogen.
Carburisation involves holding low-carbon steel in a carbon-rich
environment at high temperature. Carbon from the environment then diffuses
into the outer layer and transforms it into high-carbon steel, which can then
be heat treated to improve its hardness. Nitriding involves the diffusion of
nitrogen into a steel surface to form small crystals of metal nitrides, and
nitriding steels usually contain alloying elements which react readily with
nitrogen. The nitrogen can be diffused in from gaseous atmospheres, fused
salts or argon-nitrogen plasmas (Bell et al., 1983 ), and nitriding temperatures
are usually much lower than those used for carburising. After nitriding, the
component often does not require any further heat treatment. Cyaniding and
carbonitriding cause bpth carbon and nitrogen to diffuse into the surface.
Cyaniding is a salt-bath process carried out at a nitriding temperature,
whereas carbonitriding is a gaseous process carried out at higher temperature.
Many commercial processes exist under particular trade names, for
carrying out the processes described above and for diffusing in other elements
such as boron, aluminium and chromium. Readers requiring more detail of
these are referred to the review by Wilson ( 1975).
In ion implantation processes, ions with very high energies, of the order
of MeV, are fired into the component surface and, as a result of their high
energies, penetrate the surface up to distances of ~ 1 JJ.m. Very high
concentrations of the implanted ions can be created in this way and the
resulting surfaces often have excellent wear resistance (Dearnaley and Grant,
1985). One beneficial, and as yet unexplained, effect is that the improved
wear resistance often persists up to depths well below the original depth of
implantation.
Application of Surface Coatings
Surface coatings can be applied over a range of thicknesses from several

millimetres to a few micrometres. At the thicker end of this range, layers of
wear-resistant metal can be deposited by welding techniques, and this process
can be used to reclaim very badly worn components such as the teeth of
excavators. Popular materials for this type of application are the cobalt-based
superalloys described in Section 4.1.
Coatings of the order of 1 mm thick can be deposited by flame spraying,
which is usually used for metallic coatings; plasma spraying, which can be
used for a wide range of metals and ceramics; and spray-fuse coating, where
a metal is initially sprayed as a powder and subsequently fused, to form a
fully dense and highly adherent coating.
Coatings of the order of0.1 mm thick can be deposited by electroplating,
electroless deposition, plasma spraying or detonation-gun plating. Electro-
plating is used to deposit hard and abrasion-resistant chromium plating, and
also to deposit the lead alloy overlays on bearing alloys. Electroless deposition
can be used to deposit a range of hard and wear-resistant nickel-based
coatings. Both plasma spraying and detonation-gun plating can be used to
deposit a wide range of ceramic and cermet coatings, which are finding
extensive use as wear- and corrosion-resistant materials in hostile environments
such as nuclear reactors. Both types- of coating are slightly porous, the
plasma-sprayed coatings being rather worse in this respect, and this can
allow some aggressive corrosive species to penetrate the coating and cause
it to separate from its substrate.
Finally, coatings only a few micrometres thick are being increasingly
deposited by physical vapour deposition (Teer and Arnell, 1985) and chemical
vapour deposition (Hitchman, 1985). Both these techniques can allow the
deposition of fully dense and highly adherent coatings of a wide range of
compositions, including metastable compositions which cannot be formed
by any other means. Such coatings are already being spectacularly successful
in a variety of applications, which include: the coating of cutting tools with
titanium nitride to give a fivefold increase in life; coating of the races of
satellite bearings to give a coating ~0.5 .urn thick which provides soft metal
film lubrication for the full life of the satellite; and coating of other satellite
bearings with highly adherent films of molybdenum disulphide.
4.8 PROBLEMS
No problems have been set for this chapter. This is because exhaustive design
procedures for dry rubbing bearings have been set out recently in ESDU
Item No. 87007: Design and Material Selection for Dry Rubbing Bearings.
For commercial reasons neither the data nor the procedures can be
reproduced here, and under these circumstances it is not possible to set
meaningful design problems. However, the reader is strongly recommended
to study the ESDU unit and to work through the example design problems
therein.
Chapter 5
Fluid-lubricated Thrust
Bearings
5.1 INTRODUCTION
So far in this book the contact of bodies in relative transverse or sliding

motion has been considered for the important case of sensibly dry contact.
In such situations, which are common in many everyday applications, there
is by definition intimate contact between the sliding, or rubbing, pairs of
surfaces, and the consequences of this physical fact have been discussed at
length in the previous chapters.
However, sliding surfaces may be completely separated by the creation
of relatively thin films of fluid which are capable of transmitting the normal
loads between the surfaces while allowing sliding to occur in the transverse
direction. The study of such films is based on hydrodynamic theory as applied
to the analysis of the viscous flow of fluids in gaps or clearances which are
relatively small in comparison with the other dimensions of the components
involved. The application of the results of these studies leads to the design
and development of hydrodynamic, or self-acting, fluid film bearings.
In this chapter much simplified analyses will be made of relatively simple
geometric configurations in order to gain an understanding of the principles
involved, but the results of much more complicated solutions will be presented
and used in the design of hydrodynamic thrust bearings. This approach will
be extended in the next chapter to show how the design of hydrodynamic
journal bearings may also be approached.
Fluid films may also be created between sliding surfaces by supplying
fluid from an external supply at a relatively high pressure. Such externally
pressurised, or hydrostatic, bearings are usually used in special circumstances
and have been extensively studied (Opitz, 1967; Halling, 1975a). This class of
bearing will not be dealt with in this book.
124
FLUID-LUBRICATED THRUST BEARINGS 125
5.2 FLUID PROPERTIES
5.2.1 Dynamic Viscosity
The physical property of state known as dynamic viscosity was recognised

by Newton, some 300 years ago, as an important characteristic of fluids
(Massey, 1989). Dynamic viscosity may be defined, as far as the requirements
of this chapter are concerned, by reference to Figure 5.1, where a plate of
area A is propelled with a force F at a velocity U so that a film of fluid of
uniform thickness h exists between the plate and a plane stationary surface.
The axes x and z are introduced as shown, to describe the geometry of
the problem.
The shear stress, t, which must act on the underside of the plate is given
by t = F I A. The velocity gradient, ouj oz, is assumed to be constant
throughout the thickness of the fluid film and is given by oujoz = u /h.
Newton proposed that for an ideal fluid the shear stress, t, is proportional
to the velocity gradient ouj oz. Thus,
Toc-
au or
au
t=l'f- (5.1)
oz oz
where the constant of proportionality, l'f, is known as the coefficient of dynamic
viscosity. The dimensions of this coefficient may be found as '1 = t/(oujoz)
and it follows that t /( ou/ oz) has dimensions of
(F/L 2 ) FT
(L/T)/L =u
In the SI system the coefficient of dynamic viscosity has units of
(newton·second)/(metre) 2 , but as a newton is defined as (kilogram)
(metrejsecond 2 ), the units of dynamic viscosity may also be given in
dimensions of (ML/T 2 )T/L2 = M/LT or (kilogram)/(metre·second). For
example, water at room temperature and atmospheric pressure has a dynamic
viscosity of about 10- 3 Ns/m 2 or 10- 3 kg/m·s.
Another common and important unit of dynamic viscosity is the poise,
which has units of (dyne· second)/( centimetre )2 or (gram)/( centimetre·second)
in the CGS system of units, but a more useful unit is the centipoise (cP).
The relationships between these units is as follows:
dynamic viscosity (kg/m·s)= / 0 dynamic viscosity (P)
= 10100 dynamic viscosity (cP) (5.2)
zL;
Velocity U Area (A)
L,.c;:: :~ . Force F
f 777777777777777777777
Film thicknessh
X
_ F l u i d film
Figure 5.1
126 TRIBOLOGY
5.2.2 Kinematic Viscosity
The kinematic viscosity of a fluid is simply defined as the ratio between its
dynamic viscosity, 1J, and its density, p. So the kinematic viscosity, v, is simply
v = 11/ p and its dimensions in the SI system are (metre) 2 /(second). The
corresponding unit in the CGS system has units of (centimetre) 2 /(second)
and is known as the stoke, although a more useful unit is the centistoke
(eSt). This latter unit is very important, because most of the internationally
3.0 22
4.0 37
46
5.0
68
7.0 100
9.0
12.5
20
~ 30
40
·~0
&I 75
·;
·~ 150
E
~ 250
c
i2 400
750
1500
3000
10000
50000
200000
-20 0 20 40 60 80 100 120
Temperature (°C)
Figure 5.2 Typical characteristics of Shell Tell us oils:
Shell Tel/us oil
22 37 46 68 100
Viscosity, at ooc 188 340 535 960 1650

kinematic,
eSt, IP 71 at 20oc 60 93 170 196 325
at 40°C 22.8 36.2 46 68 100
at 100°C 4.38 6.1 6.97 9.05 11.6
Density at 15°C,
kg/1, IP 160 0.866 0.875 0.878 0.880 0.885
recognised scientific instruments which are used to measure viscosity are

calibrated in centistokes (eSt).
However, it should be noted that there is also a range of instruments,
known as efflux viscometers, which measure viscosity by simply giving the
time required for a given volume of fluid to flow through an orifice of known
characteristics. Thus, the viscosity of engine oils, for example, is often quoted
in units of 'seconds' as measured by a No. 1 Redwood viscometer. It is
usually possible to convert such measurements in 'seconds' into the more
technically useful units of centistokes, for example (Massey, 1989).
5.2.3 Variation of Viscosity with Temperature
A major difficulty in the design of hydrodynamic bearings is the rapidity

with which the viscosity of mineral oils, with and without additives, changes
with temperature. This feature is important, as mineral oils are used
extensively to lubricate such bearings. The caption and graphs of Figure 5.2
show this variation for a family of commercially available oils, the Shell
Tell us oils. It should be noted that the scales of the graphs have been distorted,
to produce straight-line characteristics, as these are of practical use. A more
revealing presentation of the viscosity-temperature characteristics for two
of this family of oils is given in Figure 5.3.
The variation of density with temperature of mineral oils is negligible
in comparison with the variation of viscosity, and the values of density given
in the caption of Figure 5.2 may be taken to be constant over normal ranges
of temperature. Similarly, the changes in pressure experienced by oils in
hydrodynamic bearings are normally not large enough to affect their viscosity,
but this is not the case in highly loaded non-conforming contacts, as discussed
in Chapter 7.
200
60 80 100
Temperature (°C)
Figure 5.3 Viscosity-temperature characteristics on linear scales

128 TRIBOLOGY
5.3 LAMINAR FLOW THEORY
5.3.1 Flow through a Parallel Gap
Consider the flow of a viscous fluid (Massey, 1989) through a wide parallel
rectangular gap of relatively small, but uniform, clearance h under the action
of a pressure drop from p 1 to p 2 over a length L and the action of a moving
surface which has a velocity Uc• as shown in Figure 5.4. The axes x, y and z
are set up as before and it is assumed that the edges at y = 0 and y = B,
corresponding to the width of the gap, are sealed so that the flow of fluid
proceeds only in the direction of the x axis.
Consider the equilibrium of an element of fluid oflength ()x at a distance x
from the origin. Let this element have a thickness (jz and be positioned at
a height of z above the lower surface which moves to the right with a velocity
Uc. The local pressure at x = 0 is p 1 , and at x = L, p2 •
The element of fluid of Figure 5.4 is shown enlarged in Figure 5.5,
together with the pressures and shear stresses which are assumed to act on
it. Pressure p is assumed to increase with x and the shear stress r with z.
In the steady state the sum of the forces acting on the element is zero.
In each case the local pressure, or shear stress, may be multiplied by the
appropriate area, to give the corresponding force. The sum of the forces
acting to the right on the element is thus:
p(B<5z) + ( r + ~; ·Jz )(B<5x)- ( p + ~~ ·<>x )<B<>z)- r(Bbx) = 0
Figure 5.4
T+-·/jz
OT
az j z
~- op
P~j:~j~ p+a;·6X
x 6x
Figure 5.5
or
or op (5.3)
oz ox
This is the equation of equilibrium for the element of fluid. Development
of the solution now proceeds on the basis that the flow throughout the
gap is laminar, or viscous, so that the relationship between the shear
stress and velocity gradient given in Equation ( 5.1) may be substituted
into Equation (5.3) above, to yield the following expression:
o u op
2
'1 oz 2 =ox
As op I ox is not a function of z, this equation may be integrated twice,

to obtain:
u=(~op)z 2 +C 1 z+C 2
I]OX 2
where C 1 and C 2 are constants to be determined. Now, when z = 0, we know
that u = u c> and when z = h, u = 0.
Thus
c = -(~op)~- uc
1
I]OX 2 h
and C 2 = Uc. These values for the constants may be back-substituted into
the above expression, to give the following solution for the velocity of the
fluid in the gap:
U=- 1(
21]
-- op)
ox (hz-z )+Uc
2 ( 1--
h
z) (5.4)
It is clear that the pressure gradient must be negative if the flow is to proceed
to the right and that the total velocity at any value of z is given by the sum
of the pressure-induced, or Poiseuille, velocity and the shear velocity induced
by the movement of the lower surface.
The total volumetric flow rate of fluid through the gap is given as
Q = J~uBt5z. The expression for u, a function of z, may be substituted from
Equation ( 5.4) into this expression to give
or
Q=Bh 3 ( - op)+ Bhu (5.5)

121] ox 2 c
130 TRIBOLOGY
In the case of the geometry of Figure 5.4, it is clear that the volumetric flow
rate given by Equation (5.5) does not vary with x, so that
op =
OX -12n(g-
., B 2~u)
c
Integration with respect to x gives
p= 1211(~- ~·Uc )x + C 3
Now, when X= 0, p = Pl, so c3 = Pl• and when X= L, p = p2; thus,
P2= -12'7(~-~uc)L+pl
In this case the pressure gradient, opjox, is a constant equal to (p 2 - p 1 )/ L;
note that this is negative, so that for this special case Equation ( 5.5) becomes
Q = Bh3 (Pl - P2) + Bh U (5.6)

12'7 L 2 c
5.3.2 Example of Flow Rate Calculatio n
Tellus 37 oil flows at a mean temperature of 40°C through a rectangular

channel or gap of uniform thickness 0.1 mm, width 15 mm and length 200 mm
under a pressure drop of 4 Njmm 2 • What is the flow rate of fluid?
From Figure 5.2 the kinematic viscosity of the oil is 36.2 eSt and its
density0.875 g/cm 3. Its dynamic viscosity in SI units is, from Equation(5.2),
36.2 x 0.875 = 0.0317 kg/m. s

1000
In this case Uc = 0 and the flow rate follows from Equation (5.6) as
= (15 X 10-3}(0.1 X 10-3)3 {4 X 106- 0} = 0.807 X 10-6 m3/S
Q 12 X 0.0317 200 X 10- 3
= 0.0481/min
5.4 HYDRODY NAMIC THRUST BEARINGS
5.4.1 Plane-incli ned Slider Bearings- Reynolds's Equation
This elementary geometry is shown in Figure 5.6 and it is assumed that the
width of the bearing, B, is much greater than its length, L. This restriction
has the effect of ensuring that most of the flow through the gap between the
Figure 5.6
bearing and the moving surface is 1n the direction of the x axis. It will only
be near the ends of the bearing, where y = 0 and B, that there will be significant
flow in the direction of the y axis. This restriction is discussed further in
sub-section 5.6.1.
There is no change of gap or film thickness in the direction of the
y axis, but an uniform inclination of the undersurface of the bearing is
assumed to exist in the direction of the x axis, as shown in the cross-section
of Figure 5.7. The gap is shown to vary uniformly from a minimum of h1
to a maximum of h2 over the length, L, of the bearing. The stationary origin
is taken to be at the intersection of the produced upper stationary and the
lower moving surfaces of the bearing.
Now let there be a local pressure p, a pressure gradient op I ox and a
film thickness h at the section of the bearing located by the ordinate x. If
the fluid is incompressible, then the volumetric flow rate of fluid through
the bearing must be a constant. This is the continuity of flow statement,
which means that as much fluid flows into the bearing at one end as flows
out at the other, so the flow rate, Q, is a constant and independent of x;
thus, oQ/ ox = 0. Substituting for Qfrom Equation ( 5.5) in this expression gives
!__{~(- op)+~uc}=o
ox 121] ox 2
This expression may be put in the following form, as U cis not a function of x:
!___ {h3 op} = 6Uc oh (5.7)

ox '1 OX OX
.. Uo
x 2 = L (K+ 11/K
Figure 5.7
132 TRIBOLOGY
This is Reynolds's equation in one dimension, and it has been derived in

this very simple manner by assuming that expression (5.5) applies within the
film even though its thickness changes with x. Reynolds's equation is of
fundamental importance in the study of hydrodynamic lubrication and its
derivation is discussed in more detail elsewhere (Greenwood and Halling,
1971 ; Cameron, 1981 ).
In the case of Figure 5. 7 it is clear that
oh = h 2 -h 1 = h1(h2 _ 1) = h1·K
ox L L h1 L
where K = (h 2 /hd- 1. Study of the geometry also shows that h = h 1 when

x 1 = L/ K and h = h 2 when x 2 = (L/ K)(K + 1).
Putting h=(ohfox)x=(h 1K/L)x and Uc= -U 0 in Reynolds's
equation (5.7) gives the following equation:
!__{(h x) xop} = _ 6u (h K)
1 3 3
0
1
ox L '7 ox L
or
Note that the terms in the first parentheses may be assumed to be constant
for a given bearing if the variation of viscosity with temperature is ignored
at this stage. The above expression may now be integrated twice with respect
to x, to obtain the following expressions:
(5.8)
(5.9)
This is the solution for the pressure distribution in the bearing, and the two
constants of integration, C 3 and C4 , may be determined by putting p = 0
(atmospheric) at x 1 = L/ K and x 2 = (L/ K)(K + 1). The constants have the
following values:
A typical shape of this pressure distribution has been superimposed in

Figure 5.7.
Of interest is the value of the maximum pressure which occurs at xm

where opfox = 0. It follows from Equation (5.8) that:
x =
m
{~(K +
K K+2
l).}L
and that the corresponding value of the film thickness is hm, where
The maximum value of the pressure in the film at xm is found from

Equation (5.9) as
Pm = ( ~~fL) (2(K + ~~K + 2))

Now the total load carried by the pad, W, is given by the following integral:
W= J:' pBdx = { 6~uo(~Y} : J~~~KJ<K+lJ{~- ~32 + C4) dx

2
By substituting for the constants C 3 and C 4 and completing the above

integral, it may be shown that
The rate of flow of fluid through the bearing may be found conveniently
at xm, as there is no pressure-induced flow here, because the pressure gradient
is zero. The velocity- or shear-induced flow is then given by the second term
of Equation (5.4) as Q=(U 0 /2)Bhm and putting
h =2(K+l)h
m K+2 1
gives the flow rate as:
Last, the lower sliding member experiences a shear force due to the
shear stress r which is distributed over its surface as it passes under the
inclined pad. Remembering that the shear stress is given by r = ~ ou/ ox and
that u is given by expression (5.3), it follows that, at z = 0,
ou hop U0
r=~-= ---+~-
ox 2ox h
134 TRIBOLOGY
The total shear force on the sliding member is given by
F= f x,
x1
rBdx=B
f(L/K)(K + 1)
L/K
( h Op U 0)
---+17- dx
2 OX h
Integrating the first term by parts and the second by putting h = (h 1 K I L) x
gives
_ h 1 (K) + (1JU 0 B)(log.(K + 1))

F-W-- --
L 2 h1 K
Substituting for the previously determined expression for W,
5.4.2 Summary of Results- Dimensionless Coefficients
The detailed analysis presented in the previous section has shown that, within
the inherent assumptions, it has been possible to derive expressions for four
of the main parameters which would be of interest in the design of a bearing
of the geometry under discussion. These parameters are the maximum
pressure in the film, the volumetric flow rate of fluid required to operate the
bearing, the load capacity of the bearing and the shear force necessary to
move the sliding member under the bearing.
It will be noticed that the fluid film converges in the direction of sliding
and that all of the four parameters listed above are proportional to the
sliding velocity. Thus, the pressures under the bearing are developed by the
action of the sliding member dragging fluid through the converging gap -
i.e. the bearing is self-acting.
The four expressions are reproduced in a convenient form in expressions
(5.10):
r}[;
Q = (Bh 1 U0 )(K + 1 )
K+2
(5.1 0)
W= { 1JUoB(~ 2{ log.(K + 1)- K2~ 2 } J

F= ('1UoBL){~1og.(K + 1)- _6_}
h1 K K +2
Each of the four practically important parameters is given as the product of
two separate terms. In each case the first set of enclosures contains all
dimensional information for the particular bearing, while the second term
contains information which depends only on the shape or geometry of the
bearing in dimensionless form.
In this simple case the geometry of the bearing (Figure 5.6) is described
by a single dimensionless variable, K, which is a function of the ratio between
the entry and exit film thicknesses, h2 and h1 . Therefore, the following four
dimensionless parameters or coefficients are defined and used to simplify the
presentation of the corresponding expressions of (5.10):
Pm={2(K+~~K+2J
( 5.11 )
- { Kloge(K+1)-
F= 4 6 }
K+ 2
The expressions of (5.10) then become:
(5.12)
The four important parameters for a wide hydrodynamic thrust bearing

under discussion are therefore given by the straightforward expressions ( 5.12 ),
where the dimensional details of any particular bearing are given by the
contents of the enclosures, while the shape or geometry of the bearing is
reflected in the values of the four dimensionless coefficients fim, Q, Wand F,
which multiply the sets of enclosed terms.
In this case the values of the dimensionless coefficients Pm• Q, Wand F
depend only on one geometric parameter, K, which itself is dependent on
the ratio h2 / h 1 • The values of these four coefficients may be evaluated from
expressions ( 5.11) and tabulated for a useful range of K, as shown in
Figure 5.8, but it is more usual to express these results in the graphical form of
Figure 5.9. It may be shown from the third expression of ( 5.11) that the
maximum value of W occurs when K = 1.1889.
136 TRIBOLOGY
5.4.3 Example of Thrust-bearing Calculations
A thrust bearing of width 200 mm and length 50 mm operates with a minimum

film thickness of 0.1 mm and a sliding velocity of 3 m/s. If the effective
viscosity of the lubricating oil is assumed for simplicity to be 40 cP and the
film thickness ratio is adjusted to produce the maximum load capacity,
determine the operating characteristics of the bearing.
As the maximum load capacity is produced when K = 1.1889, we get
h 2 /h 1 = K + 1 = 2.1889, and as h 1 = 0.1 mm, h 2 is 2.1889 mm. The value of
Pm• (J, Wand F corresponding to this value of K may therefore be evaluated
from Equations (5.11) or read from Table 5.1 and Figure 5.8.
The value of viscosity to be used in the calculations may be found in
SI units from expression ( 5.2) as 1 ~8 0 = 0.04 kg/m · s.
Maximum pressure
From Figure 5.8, Pm = 0.2555 and expressions (5.12) give

3 )} = 153 x 10 3 N/m 2 = 0.153 N/mm 2
Pm = 0.2555{ 0'04 x 3 (SO ~ 1 ~-3
(0.1 X 10 )
Flow rate
From Figure 5.8, (J = 0.6804 and expressions (5.12) give

Q = 0.6804{ (200 X 10- 3 )(0.1 X 10- 3 )3}
= 41.184 x 10- 6 m 3 js = 2.471/min
ILl..
0.6
I~
..
-g
0.410:
E
lcl: 1<>.
Figure 5.8 Variation of dimensionless coefficients, Pm• Q, W and F for a wide

hydrodynamic thrust bearing with film thickness ratio h2 /h 1
Table 5.1 Variation of dimensionless coefficients Pm• Q, Wand F for a wide hydrodynamic thrust bearing with film thickness ratio h2 /h 1
h2
- I 1.25 1.5 1.75 2.0 2.1889 2.25 2.5 2.75 3.0 3.25 3.5 3.75 4.0
hl
,r-
K 0.25 0.5 0.75 1.0 1.1889 1.25 1.5 1.75 2.0 2.25 2.5 2.75 3.0 c
0.1333 0.2000 0.2338 0.2500 0.2555 0.2564 0.2571 0.2545 0.2500 0.2443 0.2381 0.2316 0.2250 0
Pm '
r-
Q 0.5556 0.6000 0.6364 0.6667 0.6804 0.6923 0.7143 0.7333 0.7500 0.7647 0.7778 0.7895 0.8000 c
OJ
:0
w 0.0884 0.1312 0.1511 0.1589 0.1602 0.1601 0.1577 0.1533 0.1479 0.1420 0.1360 0.1300 0.1242 (")
)>
F 0.9036 0.8437 0.8028 0.7726 0.7542 0.7488 0.7292 0.7122 0.6972 0.6836 0.6711 0.6594 0.6494 -1
m
0
-1
:I:
:0
c
en
-1
OJ
m
)>
:0
z
C)
en
~
w
-..J
138 TRIBOLOGY
Load capacity
Again Figure 5.8 gives W = 0.1602 and, using expressions (5.12),
w = 0.1602{0.04 X 3(200 X 10- 3 ) ( 50 X

10 - 3 ) 2
} = 960 N
0.1 X 10- 3
Shear force
Figure 5.8 gives F = 0.7542 and, using expressions (5.12),
F = 0.7542{0.04 X 3(200 X 10- 3 )(50 X 10- 3 ) } = 9.05 N

(O.t x to- 3 )
5.4.4 Further Characteristics of Plane-inclined Slider Bearings
The coefficient of 'friction' of a bearing is sometimes defined asp., the ratio

between the shear force and load capacity - i.e. p. = F I W In the previous
example we get p. = F I W = 9.051960 = 0.0094. This low value is typical of
self-acting hydrodynamic bearings.
Of great practical interest is the power required to drive the sliding
member under the bearing and this is simply the product of the shear force
and the sliding velocity - i.e. P = FU 0 . In the previous example the power
is P=FU 0 =9.05 x 3=27.15W.
Because of the sensitivity of the viscosity of mineral oils to temperature,
it is again of great practical importance to assess the temperature rise of the
oil as it passes through the bearing. The power consumed by the bearing
appears as heat which causes a rise in the temperature of oil. It will be
assumed in the first instance that all of the oil which flows through the
bearing is equally effective in absorbing the heat which is generated in the
bearing. It will also be assumed that no heat is conducted through the metallic
components of the bearing, so that all of the heat generated is absorbed by
the oil.
A simple power balance may then be written as
where pis the density of the oil, CP its specific heat and AT the temperature
rise of the oil. Most mineral oils have a specific heat of about 1880 Nmlkg·°C.
Hence,
AT=__!_____= 275 =0.4°C

QpCP (41.184 X 10- 6 )(880)(1880)
Figure 5.9 has been produced to summarise the discussion of this bearing.
960 N
0.923
N/mm 2
0.1 mm 0.219 mm
Figure 5.9
5.4.5 Further Dimensionless Coefficients for a Plane-inclined

Slider Bearing
Other useful coefficients may be derived from those which have already been
presented. For example, the power balance may be modified by recognising
that only a proportion, k, of the power is convected out of the bearing by
the oil. Then
SO,
Putting T = PI Q,
ilT = kT( pCPhi

1JUoL) ( 5.13)
Thus, 'the temperature rise is given directly as the product of a

dimensional group, in parentheses, and a dimensionless coefficient which is
dependent on the geometry of the bearing. In the previous example,
T = FI Q= 0.754210.6804 = 1.0988, and taking k = 1, although a more typical
value is about 0.8, we get from expression (5.13)
ilT=kT('1UoL)=1xl.0988{ 0.4x3(50x10-3) }
pCPhi 880 X 1880(0.1 X 10- 3)2
= 0.4 oc, as before
Values of T for the bearing in question have been added to Figure 5.9.
140 TRIBOLOGY
5.5 EFFECTS OF HEATING ON BEARING BEHAVIOUR
5.5.1 Effective Viscosity and Temperature
As the temperature of the lubricant increases as it passes through the bearing,

it is obviously necessary to compensate for the inevitable decrease of viscosity
which results in all mineral oils. It is assumed that oil enters the bearing at
an inlet temperature 7i and leaves at a temperature 7;,. The temperature rise,
.:\ T, is thus found from the power balance as
L\T=T.-T.=kT
o I
-c-
0-{rtU L}
h2
p p 1
The effective temperature, T., of the lubricant is conventionally taken as the

mean of the inlet and outlet temperatures:
It is clearly necessary to relate the effective temperature, 7'., of the oil

to the operating characteristics of the bearing, and it is normal to take the
effective viscosity of the oil as that at the effective temperature, r_. It is
assumed that for the case of a wide plane-inclined slider bearing all of the
oil which enters the bearing travels through the whole length, L, of the
bearing and is therefore all equally effective in cooling the bearing, as
comparatively little oil escapes from the bearing in the sideways direction.
The following examples illustrate the methods which may be used to
include the viscosity-temperature characteristic of an oil in the assessment
of a bearing's behaviour.
5.5.2 Example of Thrust-bearing Calculations Including Effects

of Temperature
A plane-inclined pad thrust bearing is 160 mm wide and 40 mm long in the

direction of sliding. The sliding velocity is 6 m Is and the bearing is supported
so that the inlet film thickness is always three times the outlet film thickness.
A load of 18 kN is carried by the bearing and Tellus 37 oil is supplied at an
inlet temperature of 40°C. Assuming that 80% of the heat generated in the
bearing is carried away by the oil, describe the operating characteristics of
the bearing.
From Figure 5.2 the required table and graph may be derived
(Figure 5.10).
As h2 /h 1 = 3, we get K = 2, Pm = 0.25, Q= 0.75, W = 0.1479, F = 0.6972
and T = 0.9296 from Figure 5.8.
"=>
~
c.
b 40 E
.e !!!
~ 30
~ ~--+-~~_,~~~
·~ Effective viscosity= 0.0263 kg/m·s
~ 20 ...._..___,___.__.....____._.-J
35 40 45 50
Mean temperature rcl
Figure 5.10
From expressions (5.12) and (5,J3) we may write
and
~T= kT( 1fUoL)

pCPhi
Inserting all of the known information yields
18000 = 0.1479{ 1J X 6X 0.16( 0~~4 Y}

and
~T = 0.8 X 0.9296( 'I X 6 X 0.04 )

875 X 1880hi
which reduce to the following expressions:
18000 = 2.272 X 10- 4 (h1JJ
and
~T = 10.85 X 10- 6 C,i)
Eliminating 1J I hi from these two expressions gives
~T= 10.85 X 10- 6 X

18000 =8.59°C
2.272 X 10- 4
So the outlet temperature is 40 + 8.59 = 48.59°C and the effective temperature

is given as
~T 8.59 o
2 = 40 +2- = 44.3 C
T.e = T1 +-
142 TRIBOLOGY
- ....
118 kN'
4.87 N/mm 2
I '
I \
- 6 m/s
39.1 N
O.Q18 mm
I
0.055 mm
Figure 5.11
Reference to Figure 5.2 gives the effective viscosity at the effective

temperature of 44.3 oc as 0.0263 kg/m · s. The corresponding value of the film
thickness h 1 then follows from the above as
h = ( 10.85 X 10 _6 _.!!._)t = ( 10.85 X 10- 6 X 0.0263)t

1 tJ.T 8.59
hl = 1.82 X l0- 5 m=0.018mm
So,
h2 = 3h 1 = 3(0.018) = 0.054 mm
The important operating characteristics then follow from expressions ( 5.12) as
= 0. 25 {0.0263 X 6 X 0.04} = 4.87 X 106 Njm 2

Pm (0.018 x 10- 3 ) 2
Q = 0.75(0.16 X O.Q18 X 10- 3 X 6) = 12.96 X 10- 6 m 3 /s
_O (0.0263 X 6 X 0.16 X 0.04) _

F- .69n o.ot 8 x 10 _ 3 -39.1 N
The operating characteristics of the bearing are shown in Figure 5.11.
5.5.3 Bearings with a Known Taper
In the previous example it was assumed that the dimensionless parameters

of the bearing were known; as it was stated that the manner in which the
bearing was supported ensured that the ratio between the inlet and outlet
film thicknesses was always equal to 3. This restraint may be imposed on a
plane-inclined pad by supporting it with a pivot at the centre of pressure
corresponding to the desired film thickness ratio. This aspect of the simplified
theory has not been presented in sub-section 5.4.1 but it has been described
elsewhere (Cameron, 1981 ).
A more general case is that in which a known taper is machined in a

pad- i.e. only h 2 - h 1 is known at the outset. The technique required to
deal with this practical situation is outlined in the following extended example.
5.5.4 Example of Calculations for a Thrust Bearing of Known

Taper
An inclined thrust bearing of rectangular form, 90 mm wide by 30 mm long,

is machined with a uniform taper of 0.03 mm over its length of 30 mm. It
carries a load of 13 kN at a sliding speed of 10 m/s when using Tellus 68
oil supplied at an inlet temperature of 40°C. Determine the operating
conditions of the bearing.
In this example neither the film thicknesses, the effective viscosity nor
the temperature rise of the oil is known. For convenience, the enlarged
viscosity-temperature characteristic of the relevant oil has been replotted in
Figure 5.12 from Figure 5.2 over the temperature range 40-60°C.
60
55
~ 50
E
"'~"' 45
0
::
-~ 40
·;;
u
-~
~ 35
0
30
40 45 50 55 60
Oil temperature (°C)
Figure 5.12 Temperature-viscosity characteristics for Shell Tellus 68 oil. From

data sheet on Tellus 68 oil:
Temperature ( C) 0
v (eSt) 'I (cP) 'I (kg/m·s)
40 68 59.7 0.0597
45 56 49.2 0.0492
50 45 39.5 0.0395
55 36 31.6 0.0316
60 30 26.1 0.0261
144 TRIBOLOGY
Table 5.2
K h 1 (mm) w F Q T=F/Q
0.5 0.06 0.1312 0.8437 0.6000 1.4062
0.75 0.04 0.1511 0.8028 0.6364 1.2615
1.00 0.03 0.1589 0.7726 0.6667 1.1589
1.25 0.024 0.1601 0.7488 0.6923 1.0816
1.50 0.02 0.1577 0.7292 0.7143 1.0209
2.00 0,015 0.1479 0.6972 0.7500 0.9296
2.50 0.012 0.1360 0.6711 0.7778 0.8626
3.00 0.010 0.1242 0.6484 0.8000 0.8105
Now the geometry of the oil film is described by the variable K, where
K = (h 2 - h 1 )/h 1 , but in this example only h 2 - h 1 is known as 0.03 mm, so
we may write
h 1 = 0.03 x 10- 3 m
or
K
This relationship may be used in conjunction with Figure 5.8 to produce
Table 5.2. Thus, the dimensionless coefficients Q, W, F and T, which describe
the operation of the bearing, are given as functions of the as yet unknown
dimensional film thickness h 1 . It is useful to plot the values of W and T
against the film thickness h 1 as shown in Figure 5.13.
0.16
1.6
1.5
1>-
0.15 ,,.
1.4
~
·u ~
"'~ ·u
ill
1.3 "'~
·~
]
~ 1.2
0.14
0
-'
~
c.
,_"E 1.1
1.0
0.13
0.8 IL----'---------'---L----'---------'---L----'---------'--___jL_______J
0.01 0.02 0.03 0.04 0.05 0.06
Minimum film thickness of oil (h 1 mm)
Figure 5.13 Load and temperature coefficients as functions of minimum film

thickness
Now the known values may be inserted into the third expression of
(5.12), to give
_ { 17X10x0.09 (30 Xh 10- 3 )

13000=W
2
}
1
or
'7 = 16.05 X 10 6 (~) (a)
Similarly, putting k = 0.8, p = 878 kg/m 3 and cp = 1880 J/kg·OC into (5.12)
gives
dT=0. 8 X r('l X 10 X 30 X 10- 3 )

878 X 1880 X hi
dT=0.145 x 10- 6 (:~) (b)
Again
dT
T.e = T+- (c)
1
2
Thus, there are six unknown quantities to be found from the three
derived expressions (a), (b) and (c) above, together with the three relationships
given in Figures 5.12 and 5.13. The following procedure may be used to
obtain solutions for these six relationships:
( 1) Guess an initial value of h1 .
(2) Obtain values of Wand T from Figure 5.13.
( 3) Evaluate 17 from expression (a).
( 4) Evaluate T. from Figure 5.12.
(5) Evaluate dT from expression (b).
(6) Evaluate I;, from expression (c).
Now, if the initially guessed value of h 1 is close to the actual value, then
the effective temperature, I;,, found in step (4) above will be very close to
the value of I;, found from step (6). Usually further adjustment to h 1 will be
necessary in order to obtain a consistent set of values for all of the six
unknowns listed above.
Take an initial guess of h 1 = 0.03 mm = 0.03 x 10- 3 m and proceed
through the six steps described above:
(1) h 1 =0.03 x to- 3 m
(2) W = 0.1589 and T = 1.1589 from Figure 5.15
( 3) 17 = 0.0909 from expression (a)
( 4) I;, = 34 oc from Figures 5.13 and 5.2
146 TRIBOLOGY
(5) AT= 16.97°C from expression (b)

(6) T. = 48.5°C from expression (c)
The values for the effective temperature, 'f., obtained in steps (4) and
(6) are clearly inconsistent, and a trial will show that a reduction in the value
of h 1 is necessary to cause convergence of these values. Subsequent results
obtained by repeating the above set of calculations for various values of h 1
are shown in Table 5.3. The convergence is clearly sensitive to small changes
in h 1 but the difference of only 0.56 oc between the values of T. obtained in
the fourth line of the table is sufficiently accurate. The bearing will therefore
operate with a minimum film thickness, h1 , of0.021 mm and a temperature
rise, AT, of 15.35°C, so that the effective temperature of the oil, 'f., becomes
!( 47.1 + 47.66) = 47.4 oc. Reference to Figure 5.12 shows that the effective
viscosity of the oil is 0.0443 kg/m·s.
The load capacity of the bearing may now be checked as follows from
the third expression of (5.12):
0.09( 30 X 10 - 3 )
3 2
W = 0.1586{0.0443 X 10 X }
0.021 X 10-
= 12905 N
which is close to the given load of 13 000 N.
Again the temperature rise may be confirmed from expressions ( 5.12) as
L.lT=08· 1
A Q4{ 0.0443 X 10 X 30 X 10- 3 }
. . 878 X 1880(0.021 X 10- 3 ) 2
= 15.2oC
and
T. = 40 + 15·2 = 47.6oC
2
which is very close to the value of 47.4 oc used above.
Table 5.3
T., dT T.,
h 1 (mm) w f 11 (kg/m·s) ("C) (OC) (OC)
0.03 0.1589 1.1589 0.0909 ~34 16.97 48.5

O.Q25 0.1600 1.1000 0.063 ~37 16.0 48.0
0.023 0.1598 1.0700 0.053 43.1 15.58 47.8
0.021 0.1586 1.0400 0.045 47.1 15.35 47.66
0.020 0.1576 1.022 0.0408 49.4 15.10 47.55
0.0205 0.1581 1.030 0.0426 48.3 15.27 47.63
7.74 N/mm 2
0.021 mm 0.051 mm
Figure 5.14
Remembering that K = 0.03010.021 = 1.428 allows Q to be interpolated

from Table 5.2 as 0.71 and the flow rate follows from the second expression
of ( 5.12) as
Q=0.71{0.09(0.021 X 10- 3 )x 10}=13.4x 10- 6 m 3 ls
= 0.81llmin
Similar calculations to those presented previously allow the maximum
pressure in the bearing to be found as 7.74 Nlmm 2 and the shear force as
42 N. The operating characteristics ofthe bearing may therefore be summarised
by the diagram shown in Figure 5.14.
5.6 THRUST BEARINGS OF FINITE WIDTH
5.6.1 Two-dimensional Form of Reynolds's Equation
It was shown in sub-section 5.4.1 that the pressure distribution in a

plane-inclined slider bearing of high width-to-length ratio, Bl L, is given by
the solution of the one-dimensional form of Reynolds's equation (5.7):
!___{h3 op} = 6Uo op

ox 1'/ ox ox
Inherent in this equation is the assumption that there is no flow of fluid in
the direction of they axis, so that op I oy = o. Considering the inclined slider
bearing of high BIL ratio shown in Figure 5.15(a), it is clear that this
assumption will be approximately true away from the ends of the bearing.
However, near the ends of the bearing the lubricant is able to flow in the
direction of the y axis and to escape from the bearing in this direction;
consequently, in these regions op1oy -:F o.
The effects of the flow of fluid in the direction of the y axis, at right
angles to the direction of sliding, is to diminish the pressures in the fluid at
the ends of the bearing, as shown in Figure 5.15(b ). Away from the ends of
148 TRIBOLOGY
(a)
Figure 5.15
the bearing it has been assumed that the one-dimensional form of Reynolds's
equation is applicable and that the load capacity, for example, may be
calculated by neglecting the reductions in pressure which occur near the ends
of the bearing. Effectively, this means that the pressure profile deduced from
the one-dimensional form of Reynolds's equation is assumed to exist over
the full width, B, of the bearing.
This is a remarkably good assumption, as in the case of a bearing of
B I L ratio equal to 2 the actual load capacity is typically only about '30%
less than that calculated on the basis of one-dimensional theory. However,
if a square bearing, of B I L ratio equal to unity, is considered, then the flow
of fluid in the direction of they axis may not be ignored and a two-dimensional
solution is necessary. Such two-dimensional solutions are obtained from the
following two-dimensional form of Reynolds's equation (Cameron, 1981):
i_{h3op}+i_{h3op}=6Uooh (5.14)
ox 11 ox oy 11 oy ox
The addition of the second term on the left-hand side involves derivatives
with respect to y and therefore includes the effects of flow in the direction
of the y axis.
The flow pattern of fluid within a square bearing will be of the form
shown in Figure 5.16( a), while the corresponding pressure distribution will
be as indicated in Figure 5.16( b). It will be noticed that there is no extended
(a) (b)
Figure 5.16
central region in which iJpjiJy = 0, and as a consequence the maximum

pressure generated in a square bearing is considerably lower than that in a
wide bearing which operates under the same conditions.
5.6.2 Characteristics of a Square Bearing
Solution of the two-dimensional form of Reynolds's equation enables the

important operating parameters of a square inclined slider bearing to be
evaluated and expressed in the same form as that developed from first
principles for a wide bearing in sub-section 5.4.2. Thus, the flow rate, Q, the
load capacity, W, the shear force, F, and the temperature rise, llT, for a square
bearing will also be given by expressions (5.12) and (5.13), provided that the
appropriate values of the non-dimensional coefficients Q, W, F and T are
used. These expressions are restated below:
Q = Q(Bh 1 U 0 )
W= w{ ~U0 B(~Y}
F=F-(~U 0 BL)
hl
llT= kT( pCPh~

~UoL)
Again the dimensionless coefficients depend only on the geometry of the film
as described by the film thickness ratio h 2 /h 1 or the parameter K. Values
of the four coefficients for a square bearing are tabulated in Table 5.4 and
plotted in Figure 5.17 as functions of h 2 / h 1 and K. These results were obtained
by numerical analysis (Jakobsson and Floberg, 1958) and should be
compared with the results for a wide bearing given in Table 5.1 and Figure 5.8.
It now follows that a square bearing may be analysed and designed with
the aid of the above four expressions and Figure 5.8, without any detailed
knowledge of how the values of the four coefficients were obtained.
150 TRIBOLOGY
Table 5.4 Variation of dimensionless coefficients Q, W, F and T for a square

hydrodynamic thrust bearing with film thickness ratio h 2/h 1 (after Jakobsson and
Floberg, 1958)
h2
1.5 2.0 2.5 3.0 4.0 5.0
hl
K 0 0.5 1.0 1.5 2.0 3.0 4.0
Q 0.5000 0.6796 0.847 3 1.0080 1.165 0 1.4700 1.7690
w 0 0.055 75 0.06894 0.06997 0.06701 0.058 36 0.05011
F 1.0000 0.8249 0.7276 0.663 3 0.6163 0.5496 0.5026
T 2.0000 1.2140 0.858 8 0.6579 0.5291 0.3739 0.2842
'!1.
h,
L-----~----_L_____ J_ _ _ _ _ _L-----~-----L------L---~0
0 4
K
Figure 5.17 Variation of dimensionless coefficients Q, W, F and T for a square

hydrodynamic thrust bearing with film thickness ratio h2 /h 1
Similar sets of data have been obtained for bearings of other width-to-
length ratios, but as most of the individual bearings used in practical designs
of thrust bearings are of an approximately square form, the information
presented in Table 5.4 and Figure 5.17 is particularly useful.
5.6.3 Example of Calculations for a Square Bearing
A thrust bearing of square form works with a uniform sliding velocity of

56 m/s and supports a steady load of 25 kN. The bearing is to be lubricated
with Tellus 68 oil, which is available at an inlet temperature of 40°C. The
bearing must be designed to meet the following three conditions: ( 1) the

minimum film thickness is not to be less than 0.025 mm; (2) the flow rate
of oil is not to exceed 10 1/min; and (3) the outlet temperature of the oil is
not to exceed lOOoC.
In order to assign an initial value to the length and width of the bearing,
it is remembered that the mean pressure in the bearing of sub-section 5.5.4
was about 4.8 N/mm 2 , so that in the present case
25000)1
L=B= ( - - =72mm
4.8
Thus, as a conservative guess, the initial value of L and B is taken to be

80 mm. Study of Figure 5.17 shows that the value of W in the region of
K = 1.5 does change greatly from the tabulated value of 0.069 97, with the
corresponding value ofT= 0.6579. These two values will be taken as constant
for the purposes of this example.
Now substitute all known values in the expressions for Wand /:iT from
expressions (5.12) and (5.13):
25 X 10 3 = 0.06997 { 1]56 X 80 X 10- 3 (

8o xh;w- 3 ) 2
}
or
(a)
Again
1]56 80 X 10- 3 }
X
11 T = 0.8 X 0.6579 { 2
ci)
880 X 1880hl
/:iT= 1.425 X w- 6 11 (b)
Substitute for 11/hi from (b) into (a), to give

/:iT= 1.425 X 10- 6 X 12.461 X 10 6 = 17.8°C
The effective temperature, 7'., follows as
T. = 40 + 17·8 = 48.9°C
2
and the outlet temperature, 7;,, follows as 57.8 oc.
The effective viscosity of Tell us 68 at the temperature of 48.9°C may be
obtained from Figure 5.13 as 0.0413 kg/m · s. Substitution of this value into
152 TRIBOLOGY
(a) allows the value of the minimum film thickness, h 1 , to be found:

h2- 0.0413 h1 = 5.75 x 10- 5 m or 0.058mm
or
1 - 12.461 X 10 6
From Figure 5.12, Q= 1.008 at K = 1.5 and the flow rate, Q, follows
from expressions (5.12) as
Q = 1.008(80 X 10- 3 X 0.058 X 10- 3 X 56)
= 0.2619 x 10- 3 m 3 /s = 15.711/min
This flow rate is in excess of the allowed value of 10 1/min and the choice
of L = B = 80 mm is clearly unsatisfactory. By taking three additional values
of L = 40 mm, 50 mm and 65 mm and repeating the calculations outlined
above, Table 5.5 may be constructed. As limits have been set on h1 , Q
and 7;, in the design specification, the results of Table 5.5 were plotted
in Figure 5.18 to show the variation of these critical quantities with
the length and width of the bearing.
Now the limitations which have been set on h 1 , Q and 7;, may be
superimposed on the curves of Figure 5.18, to give the bounded characteristics
shown.
( 1) As h 1 is limited to a minimum value of 0.025 mm, the horizontal line A

is drawn through the curve for h 1 at a value of 0.025 mm.
(2) As Q is limited to a maximum value of 10 1/min, the line B is drawn
through the curve for Q.
(3) As 7;, is limited to a maximum value of 100°C, the line C is drawn
through the curve for 7;,.
Study of Figure 5.18 reveals that:
( 1) Bearings of length less than 53.5 mm operate with values of h1 less than
0.025mm.
(2) Bearings of length greater than 69.3 mm require flow rates in excess of
10 1/min.
( 3) Bearings of length less than 44 mm generate outlet temperatures greater
than 100°C.
Table 5.5
L(mm) .1T (oC) Y/ (kg/m·s) h 1 (mm) Q (1/min) T. (OC)

80 17.8 0.0413 0.058 15.7 57.8
65 26.9 0.0326 0.037 8.23 66.9
50 45.6 0.0239 0.022 3.72 85.6
40 71.0 0.0156 0.012 1.69 111.0
16 120 0.07
14 0.06
12 100 c 0.05
44
eE
c
10 p 0.04 ~
>-.:'
~ ~
9: "'
~ ~
0 8
rn
80 0.03 ~
0 "'
c.
§
"'rn ~ "'E
~
0
u::
6 "
;
0
0.02 .§
~
<:
~
4 60 0.01
0.08
0 40~----~----~----~----~L-----~-----L----~----~ 0.006
40 50 60 70 80
Length and width of pad (L mm and 8 mm)
Figure 5.18 Bounded characteristics of a square bearing using Tellus 68 oil
It is, therefore, clear from Figure 5.18 that an acceptable design of bearing
must have a length of between 53.5 mm, line A, and 69.3 mm, line B. If the
boundary of line A is not violated, then that of line C is also satisfied.
As the relative importance of the limits set on h 1 and Q, by lines A
and B, has not been stated, it is reasonable to select a length of bearing
approximately midway between the lengths corresponding to lines A and B
- say 60 mm, as indicated by line D in the figure.
The values of h 1 , Q and 1;, corresponding to a length of 60 mm may be
read directly off Figure 5.18 or recalculated as a check in the manner employed
previously. It then follows that h 1 has a value of 0.032 mm, Q a value of
6.561/min and 1;, a value of 71.6°C. The table of Figure 5.22 gives the
coefficient F as 0.6633, so that the shear force becomes, from expressions
(5.12),
F = 0.6633(0.0308 X 56 X 0.06 X 0.06)

o.o32 x w- 3
= 127N
and the power consumed by the bearing is simply FU o = 127 x 56= 7112 W.
The operating conditions of the bearing are shown in Figure 5.19.
154 TRIBOLOGY
7777777
. 56 m/s
1
777777/r/777~77777 127 N
6.561/min
0.032 mm 0.080 mm
Figure 5.19
5.7 Practical Designs of Thrust Bearings
Consider the use of a number of square thrust bearing pads of the type and
size discussed in the previous section to support the axial thrust of 250 kN
which acts on a shaft of 200 mm diameter rotating at 3600 rev /min. The
tangential velocity of a circle of 300 mm diameter at this rotational velocity
is n(3600/60)0.3 =56 m/s, and if ten pads of length 60 mm are spaced
around a circle of 300 mm, then the arrangement of Figure 5.20 evolves.
There is adequate space for the shaft of 200 mm diameter to pass through
the 'ring' of separate thrust pads.
As ten pads of the final design deduced in sub-section 5.6.3 are used in
the complete bearing of Figure 5.20, its specification is as follows:
Load capacity= 25 x 10 3 x 10 = 250 kN
Flow rate of oil= 6.56 x 10 = 65.6 1/min
Inlet temperature = 40°C
Outlet temperature= 71.6°C
Lubricant = Tellus 68 oil
Shaft velocity= 3600 rev /min
Mean velocity over pads= 56 mfs
Power consumed= 7112 x 10 = 71 kW
Minimum film thickness = 0.032 mm
Length and width of pads = 60 mm
The analysis presented here is typical of the methods used in bearing
design, but the schematic arrangement shown in Figure 5.20 is a gross
simplification, and details of the construction of more realistic bearings are
shown in Figure 5.21. An important feature of practical thrust bearings is
that the necessary convergent geometry of the pads is created by allowing
them to tilt about a fixed pivot rather than by attempting to build a known
taper into the assembly. Consequently, the analysis and design of such
practical thrust bearings is more complicated than outlined here, and more
details are available elsewhere (Martin, 1970).
156 TRIBOLOGY
Section A-A
Figure 5.21 One-piece thrust ring
5.8 PROBLEMS
1 A thrust bearing of the Rayleigh Step type is modified as shown in

Figure 5.22( a) to include a centrally placed recess of a depth much greater
than the film thicknesses which exist under the lands of the bearing when it
is in operation. The two lands of the bearing and the recess are of equal
length L/3 and the minimum and maximum film thicknesses are h 1 and
h 2 , respectively. If the sliding velocity of the working surface is U and
the working fluid has a dynamic viscosity of 17, show that the load capacity
of the bearing per unit width, W, is given by
W= i{ '1~r}LK + ~) 3 + ~}
where K = h 2 /h 1 -1.
(a)
30mm 30mm 30mm

3000 N
(b)
Figure 5.22
A bearing of this type lies with one. of its lands in contact with its
stationary working surface, as shown in Figure 5.22(b ), under a load of
3000 N. The gap between the other land and the stationary surface is
0.07 mm, the length of the bearing is 90 mm and its width is 240 mm. If
the working fluid has a dynamic viscosity of 0.05 kg/m · s, estimate the
sliding velocity necessary to separate the bearing from the working surface.
[0.15 m/s]
2 The inclined hydrodynamic bearing pad shown in Figure 5.23(a), of
length L and width B, operates with a fluid of dynamic viscosity 17 and
is positioned relative to a plane slider, which has a velocity U in the
direction shown, so that the film thickness, h2 , at entry is k times the
film thickness, h1 , at exit. During operation the pressure in the fluid at
exit is maintained at a value of p 1 avoe atmospheric, while that at
entry, p2 , remains at the atmospheric value. On the assumption that the
width, B, is much greater than the length, L, solution of Reynolds's
equation shows that the load capacity of the bearing, W, is given by the
expression
W=Cll { qUB(~Y} +Cl2{BLpt}
0.1 mm 0.05 0.2 mm

kg/m·s
(b)
0.05 mm 0.15 mm
(c)
Figure 5.23
158 TRIBOLOGY
and the volumetric flow rate, Q, of fluid through the bearing is given by
the expression
Bhip 1 }
Q=a 3 { BUh1 } -a4 { ~
When k = 2, a 1 = 0.1589, a 2 = t, a 3 = f and a4 =~;when k = 3, a 1 = 0.1479,

tX2 = t, a3 = i and a4 = i; where k = h2 /h 1 •
(a) The thrust pad shown in Figure 5.23 (b) has a length of 30 mm and
a width of 100 mm, and is machined so that a uniform taper ofO.l mm
exists over a length of 15 mm, while a parallel step of height 0.05 mm
exists over the remaining length of 15 mm. A load of 4000 N is carried
by the bearing when the slider has a velocity of 21.5 mjs and an oil
of dynamic viscosity 0.05 kg/m · s is used. Show that the bearing will
operate with a minimum film thickness of 0.05 mm, as shown in
Figure 5.23(b ).
(b) If the load of 4000 N is imposed on .the bearing when it starts from
rest, as shown in Figure 5.23 (c), estimate the sliding velocity at which
the bearing will separate from the sliding member.
[13.72 m/s]
3 The inclined hydrodynamic thrust bearing shown in Figure 5.24(a) of
length L and width B operates with a fluid of dynamic viscosity 11 and
is positioned relative to a plane slider, which has a velocity U0 in the
direction shown, so that the film thickness, h2 , at entry is 1.5 times the
film thickness, h 1 , at exit. During operation the pressure in the fluid at
exit is maintained at a value of p 1 above atmospheric, while that at entry
is maintained at a value of p 2 above atmospheric. Solution of Reynolds's
equation on the assumption that the width, B, is much greater than the
length, L, shows that the load capacity of the bearing is given b-y
and that the volumetric flow rate, Q, of fluid through the bearing is given by
Q =~(U Bh )+ 3Bhi(P2-Pl)
5 ° 1
2011 L
The thrust bearing shown in Figure 5.24(b) has a length of 50 mm and

a width of 200 mm, and is machined so that a uniform taper of 0.04 mm
exists over a length of 25 mm and another uniform taper of 0.06 mm
exists over the remaining 25 mm. The bearing is designed to operate with
a plane slider which moves at a steady velocity of 4 mjs, an oil of dynamic
viscosity 0.08 kg/m ·sand a minimum film thickness of0.08 mm, as shown
in Figure 5.23(b).
Monitored pressure
h,
0.08 mm 0.12mm 0.18mm
Ia) (b)
Figure 5.24
It is proposed to monitor the operation of the bearing by measuring

the pressure at the mid-section of the bearing, where the film thickness
is expected to be 0.12 mm. Find the value of this pressure for the
configuration shown in Figure 5.24(b ), and, hence, estimate the total load
capacity of the bearing.
[0.571 x 10 6 N jm 2 , 4040 N]
4 The inclined hydrodynamic bearing pad shown in Figure 5.25(a), of
length L and width B, operates with a fluid of dynamic viscosity '1 and
is positioned relative to a plane slider, which has a velocity U in the
direction shown, so that the film thickness, h 2 , at entry is twice the film
thickness, h 1 , at exit. During operation the pressure in the fluid at exit
is maintained at a value of p 1 above atmospheric, while that at entry,
p 2 , remains at the atmospheric value. Solution of Reynolds's equation,
on the assumption that the width, B, is much greater than the length, L,
shows that the load capacity, W, of the bearing is given by
and that the volumetric flow rate, Q, of fluid through the bearing is given by
The mineral dressing mill shown schematically in Figure 5.25(b)

weighs 8 · 10 6 N, rotates about the vertical axis at 360 rev /min and is
supported on 49 identical thrust pads which are equispaced around the
circumference of an annular surface of the mill at a mean diameter of
2 m. Each pad has a length of 90 mm and a width of 240 mm and is
machined so that a uniform taper of 0.07 mm exists over a distance of
60 mm, as shown in Figure 5.25( c).
160 TRIBOLOGY
~ (a)
c-1) 360 rev/min
1 r~- 49 pads
"""E'l~--.7.r 117.~
r I
' '
L 2m dia.
(b)
~
0.07 mm
~~ (c)
Figure 5.25
Three different oils of dynamic viscosity 0.02, 0.04 and 0.06 kg/m · s
are available. Select one of these oils so that the pads operate with a
minimum film thickness of about 0.07 mm.
[0.06 kgjm·s]
Chapter 6
Fluid-lubric ated Journal
Bearings
6.1 INTRODUCTION
In Chapter 5 bearings of plane form only were considered and these were
shown to be of use in supporting loads which act at right angles to their
plane. Bearings of this type are generally recognised as thrust bearings.
However, another important type of bearing is the journal bearing. Journal
bearings are used to support rotating circular shafts which are loaded in a
radial direction.
In many applications it is found that a hydrodynamic journal bearing
is most suitable for the range ofloads and speeds involved, and it is fortunate
that the shape of the oil film which is formed between the bore of a circular
bearing and an eccentric rotating shaft is conducive to the generation of
hydrodynamic pressures which are, in practice, able to support considerable
radial loads.
6.1 .1 Geometry
Consider the geometry of the mutually eccentric circles of similar diameter

shown in Figure 6.1. In this figure the outer circle of radius OA, equal to
R 1 , represents the bore of the journal bearing and the inner circle of slightly
smaller radius CB, equal to R 2 , represents the periphery of the rotating
circular shaft. The shaft is assumed to rotate clockwise with a peripheral
velocity U 0 within the journal bearing, as shown in Figure 6.1. The eccentricity
between the shaft and the journal bearing is represented by the distance, e,
between the centre of the journal 0 and the centre of the shaft C.
In the study of journal bearings the straight line EOCF is known as the
'line of centres' and the distance CO as the eccentricity of the shaft in the
journal bearing.
161
162 TRIBOLOGY
Figure 6.1 Basic geometry of a journal bearing
It is required to obtain an expression for the varying distance, h, which

exists between the surfaces of the shaft and journal bearing at a location
given by the angle (), which is measured clockwise to A from the line of
centres at location E, where the maximum value of h exists. In an operating
hydrodynamic journal bearing, the space between the two surfaces is filled
with a fluid and h will therefore correspond to the varying film thickness in
the bearing.
The perpendicular OD is constructed from 0 onto AC and it is clear that
AC = R 2 + h = e cos () + R 1 cos ct
where ct is the angle OAC. From the triangle OCA it follows that
OD = (CO) sin()= (AO) sin ct
or
0 (CO) . () = ( - e ) sm
smct=--sm 0 ()
(AO) R1
In practical journal bearings eI R 1 is about 10- 3 , so that the above

expression for cos ct may be expressed as
cos ct = { 1- (;J 2
sin 2 () }t = 1- ~(;J 2 sin 2 ()
Put this identity in the above expression for AC, to obtain
A C = R 2 + h = e cos () + R 1 { 1 - ~(; J
2
sin 2 ()}
or
FLUID-LUBRICATED JOURNAL BEARINGS 163
as the term involving sin 2 () is very much less than unity. So

h=(R 1 -R 2 )+ecos()
Now R 1 - R 2 is the difference between the radii of the journal bearing and
of the shaft. This difference is also recognised as the uniform radial clearance
between the bearing and the shaft in the concentric case. Putting R 1 - R 2 = c
gives the following expression for h:
h = c + ecos ()
When () = 0, h = c + e, the maximum film thickness at E, and when () = n,
h = c- e, the minimum film thickness at F. Conventionally, h is expressed
as follows:
or
h = c( 1 + e cos()) ( 6.1 )
where e = ejc is known as the 'eccentricity ratio'. This is a dimensionless
geometric parameter which describes completely the shape of the fluid film
which exists between the surfaces of the journal bearing and shaft.
6.1.2 Hydrodynamic Action in Journal Bearings
Remembering that the radial clearance, c, and therefore the film thickness,
h, are very much smaller than the radius of either the journal bearing or the
shaft, then the shape of the film between the two surfaces may be developed
into the equivalent orthogonal form shown in Figure 6.2(a). The clockwise
rotation of the shaft shown in Figure 6.1 produces a tangential surface
velocity, U 0 , which appears as the velocity of the lower surface in
Figure 6.2(a). In Figure 6.2(a) the upper surface is fixed in space and
corresponds to that of the journal bearing. The coordinate () varies from
zero, where th,e film thickness is c + e, to n, where the film thickness is c - e,
and then to 2n, where the film thickness is again c + e.
Inspection of Figure 6.2(a) shows that a convergent film exists between
() = 0 and () = n, while a divergent film exists between n and 2n. As the
direction of sliding of the lower surface, at a velocity U0 , is in the direction
of convergence between () = 0 and () = n, it is clear that hydrodynamic
pressures will be developed in the fluid as it flows through the bearing. This
important deduction follows from the general nature of the results which
were obtained in Chapter 5 for the plane convergent geometry of Figure 5.7.
It must be remembered that the sections corresponding to () = 0 and
() = 2n in Figure 6.2(a) do, of course, define the same section because of the
actual circular shape shown in Figure 6.1.
164 TRIBOLOGY
Bore of journal
R,
(a)
Circumferential profile
of hydrodynamically
Pressure generated pressure
(b)
(c)
Figure 6.2 Film shape and pressure profile between shaft and bore of a hydrodynamic
journal bearing
The existence of the combined 'convergent-diverge nt' geometry of the

film shown in Figure 6.2(a) produces a hydrodynamic pressure profile of
the general shape given in Figure 6.2(b ). It is of importance that the generation
of pressure extends into the divergent position of the film beyond the section
where the film thickness has its minimum value at 0 = n.
Now the profile shown in Figure 6.2(b) is typical of the variation of
hydrodynamically generated pressure in the circumferential direction. However,
as most bearings have an axial length b which is usually less than their
Surface of bore of
journal_bearing
Line of centres
Figure 6.3 Operation of a hydrodynamic journal bearing
diameter, a considerable flow of fluid occurs in the axial direction, so that

pressure profiles of the shape given in Figure 6.2( c) exist in the axial direction.
The actual shape of the hydrodynamic pressure distribution in a journal
bearing is represented by a combination of the pressure profiles given in
Figure 6.2(b) and (c).
It may appear that the fluid in the bearing simply recirculates under
the action of shaft rotation, as the sections at () = 0 and 2n represent the
same actual section in Figure 6.1. However, complete recirculation of fluid
does not occur, as some fluid is able to escape from the bearing by flowing
in the axial direction. Abearing which completely encircles a shaft as shown
in Figure 6.3 is known as a complete journal bearing.
It is, therefore, necessary to replace continually the fluid which escapes
in the axial direction from the clearance by introducing replacement fluid
into the bearing in the divergent region between () = n and () = 2n.
The presence of the converging- diverging geometry and the need to
replace fluid within the completely circular shape of the typical journal
bearing introduce complications which were not present in the simple case
of plane thrust bearings. These and other important aspects of journal
bearings are considered further below.
6.1.3 Operation of Practical Journal Bearings
By combining the geometry of Figure 6.1 with the deduced pressure profile
shown in Figure 6.2(b ), the general features associated with the operation
166 TRIBOLOGY
of a practical journal bearing may be seen in Figure 6.3. Fluid is drawn into
the converging clearance at E by the rotating shaft and some of this fluid
flows completely through the bearing, to emerge at G, where the pressure in
the fluid has returned to the ambient value.
The pressure distribution described in Figure 6.2 does, of course, act
around the surface of the shaft and integrates to support the radial load, W,
which is applied to the shaft as shown in Figure 6.3. As the pressure
distribution acts over the arc between E and G, the direction of the load W
which acts on the shaft cannot coincide with the line of centres EOCF. The
relationship between the direction of the load and the line of centres is given
by the 'attitude angle', which is shown as rjl in Figure 6.3. As the load
increases, the position adopted by the centre of the shaft within the available
radial clearance, c, is as indicated in Figure 6.3. The relationship between
the attitude angle, rjl, and the eccentricity ratio, e, is such that a trajectory
of approximately semicircular shape is formed (Cameron and Wood, 1949).
Fluid may be introduced into a bearing thnmgh various configurations
of supply grooves. The usual position of supply groove adopted in many
common bearings is shown schematically in Figure 6.3, although many other
variations exist (Martin, 1983).
The analysis of the flow of fluid within the confines of a journal bearing
is clearly much more complicated than in the case of plane thrust bearings.
Consequently, a two-dimensional form of Reynolds's equation must be solved
(Cameron, 1981 ), and additional complications are encountered in the
establishment of the correct boundary conditions to be applied at the
termination of the pressure profile in the divergent part of the clearance,
section G in Figure 6.3.
As, in general, flow of fluid into a convergent clearance causes an increase
in pressure, it follows that its continued flow into a divergent clearance-will
lead to a decrease in pressure. However, when the pressure approaches the
ambient value, air may be drawn into the divergent clearance, as there will
be insufficient fluid available to fill the clearance, which increases in the
direction of shaft rotation. Similarly, dissolved air or other gases may come
out of solution in the fluid as the pressure decreases. Again, if the pressure
falls below the vapour pressure of the fluid, then cavities composed of the
fluid's vapour will form in the film. It is, of course, physically impossible for
the pressure in the film to fall below zero absolute (Dowson and Taylor, 1975).
These physical limitations are generally referred to as cavitation, and
result in streamers or striations of fluid being formed in a mixture of air and
vapour in the divergent clearance (Cole and Hughes, 1956). Much effort has
resulted in useful sets of boundary conditions being postulated (Dowson and
Miranda, 1975; Etsion and Pinkus, 1975).
A more complete description of a typical complete journal bearing is
given in Figure 6.4, where a shaft of diameter d is supported in a bearing of
axial length b. The shaft rotates with an angular velocity N and supports a
Lubricant
supply groove
Stationary journal
H w bearing
Figure 6.4 Nomenclature of a hydrodynamic journal bearing
load W, while a power H is required to rotate the shaft against the viscous
resistance which is produced in the clearance of the bearing. Fluid is supplied
at a flow rate Q with viscosity '1r through a supply groove of axial length 1
and width a, which is machined in the wall of the bearing, but the bearing
operates with an effective viscosity '1· The difference between the diameters
of the shaft and the journal bearing is Cd, so that the previously defined
radial clearance, c, is simply Cd/2.
Other important parameters associated with the arrangement of
Figure 6.4 are listed in the Notation section (pp. viii-xii) and introduced in
the following sections.
6.2 SOLUTIONS FOR HYDRODYNAMIC JOURNAL BEARINGS
It is an experimentally observed fact that the behaviour of a hydrodynamic

journal bearing may be described by using an effective or mean viscosity to
take account of the variation of temperature and, hence, viscosity which
occurs within the bearing under steady running conditions.
The effective temperature is determined by the flow rate of fluid, which
is itself dependent on the pressure in the supply groove, the form and
dimensions of the supply groove, the viscosity of the fluid, the speed of the
shaft and the geometry of the bearing.
A certain minimum flow rate of fluid will be required for the bearing
to form an adequate load-carrying film of full extent. In the design procedure
outlined here, the flow rate required by the bearing is first determined, and
the dimensions of the supply groove and pressure of fluid in it are then
selected to provide the required flow rate.
168 TRIBOLOGY
6.2.1 Dimensionless Coefficients
In principle, the analysis of the behaviour of hydrodynamic journal bearings

is the same as that of plane thrust bearings, but the details of the necessary
mathematical analysis required to solve the appropriate form of Reynolds's
equation leads to great complication. Inclusion of realistic boundary
conditions makes it essential that numerical methods be used, and only in
the case of bearings of very short length are closed-form solutions of any
practical use(Dubois and Ocvirk, 1953; Woolacot, 1965): many contributions
have been made (Cameron and Wood, 1949; Hays, 1959).
The important operating variables for a hydrodynamic journal bearing
are the radial load supported, W, the flow rate of fluid required to operate
successfully, Q, and the power required to rotate the shaft against the viscous
resistance of the bearing, H. It is found that these variables may be expressed
as follows:
W= w{ qNbd(~J 2 } ( 6.2a)
Q= Q(NbdCd) (6.2b)
H=H( qN~3b) (6.2c)
Therefore, it is seen that, within acceptable accuracy, the load capacity,

required flow rate and power consumption of a hydrodynamic journal bearing
may be expressed as products of dimensionless coefficient and combinations
of each bearing's dimensional features. The dimensionless coefficients, W,
Q, k and H have been shown to be functions of the geometry or shape of
the fluid film under a particular set of conditions. Furthermore, for a complete
journal bearing only two geometric variables are required to determine the
values of the various dimensionless coefficients. These variables are the
length-to-diameter ratio of the bearing, bId, and its operating eccentricity
ratio, e.
These results should be compared with the similar solutions presented
for thrust bearings of square form in sub-section 5.6.2.
Again, as in sub-section 5.4.5, a power balance for a journal bearing
may be written down as follows:
or
AT=kT{ qN (~) 2 } ( 6.2d)

pCP Cd
where T = HI Q as in sub-section 5.4.5.

It now follows that a complete hydrodynamic journal bearing may be

analysed and designed with the aid of the above four expressions without
any detailed knowledge of how the values of the four dimensionless coefficients
were derived. Values of these coefficients may be found in published work
(Cameron and Wood, 1949; Jakobsson and Floberg, 1957), but some
variations in the form of the dimensional groups have led to differing values
of the coefficients given above.
6.2.2 Example of Journal-bearing Calculation
A hydrodynamic journal bearing of length 160 mm operates with a shaft of

200 mm diameter which rotates at 1200 rev /min. If the effective viscosity of
the lubricating oil is assumed for simplicity to be 40 cP and the diametral
clearance is 0.18 mm, determine the operating characteristics of the bearing
at an eccentricity ratio of 0.7.
In this case the bjd ratio of the bearing is seen to be 160/200 = 0.8.
References then show that the important dimensionless coefficients for a
bearing of this length-to-diameter ratio at an eccentricity ratio of 0.7 are as
follows: W = 9.9, Q= 1.20 and H = 79. The value of viscosity to be used
in the calculations may be found in SI units from expression (5.2) as
40/1000 = 0.04 kg/m·s.
As in sub-section 5.4.5, the oil density will be taken as 880 kgjm 3 , the
specific heat as 1880 Nm/kg·oc and the proportion of heat convected out
of the bearing by the oil as 0.8. The consistent measure of angular velocity
is revolutions per unit time, which in this case is 1200/60 = 20 rev js.
Load capacity
With W = 9.9 in expression (6.2a),
W=9.9{0.4 X 20(160 X 10- 3 )(200 X 10- 3 ) ( 200 X 10- 3 ) 2

}
0.18 X 10- 3
W= 312900 N
Flow rate
With Q= 1.2 in expression (6.2b ),

Q = 1.2{20 X (160 X 10- 3 )(200 X 10- 3 )(0.18 X 10- 3 )}
170 TRIBOLOGY
Power
With H = 79 in expression (6.2c ),

(160x10- 3 )}
H = 79 { 0.04 X 20(200 X 10- 3 ) 3
(0.18 X 10- )
H = 8988 W = 12 h.p.
Temperature rise
With k = 0.8 and T = HI Q = 79 I 1.2 = 65.833 in expression (6.2d ).

L\ T = 0.8 X65.833 { 0.04 X20 ( 200 X10- 3)2}
880 X 1880 0.18 X 10- 3
L\T = 31.4°C
Minimum film thickness
Withe=0.7andc=Cdl2=(0.1 8 x 10- 3 )12=0.09 x 10- 3 inexpression(6.1),

hmin = 0.09 X 10- 3 ( 1 + 0. 7 X COS 7t)
hm; 0 =0.027 X 10- 3 m=0.027mm
6.3 ALTERNATIVE FORMS OF DIMENSIONLESS

COEFFICIENTS
It is clear from the previous section that the operating characteristics of a

hydrodynamic journal bearing may be determined easily, provided that the
appropriate values of the dimensionless coefficients are known. However, the
values of the dimensionless coefficients may only be found for a bearing of
known length-to-diameter ratio if the eccentricity ratio is also known. Again
evaluation of the load capacity, for example, also requires that the effective
viscosity of the lubricant be known. The effective viscosity is, in turn,
determined by the temperature rise which occurs in the lubricant as it flows
through the bearing.
The dimensionless coefficients which describe the operation of a
hydrodynamic journal bearing are, therefore, seen to be related to its
operating characteristics as the eccentricity ratio adopted by the shaft
is related to these dimensional characteristics. The proper design of a
hydrodynamic bearing is, therefore, more complicated than a simple evaluation
of a number of characteristics at a known value of eccentricity ratio. The
proper evaluation of the effective viscosity of the lubricant, as determined
by its associated temperature rise, is a crucial aspect of the design process.

A similar situation was revealed in the study of plane thrust bearings
(sub-section 5.5.4 ).
6.3.1 Dimensionless Coefficients for Design Purposes
As explained earlier, the dimensionless coefficients W, Q, H and T presented

in sub-section 6.2.1 for a complete hydrodynamic journal bearing are
functions of the length-to-diameter ratio, bId, and the eccentricity ratio, e.
Although values of these coefficients have been presented in either tabular
or graphical form, and may therefore be easily obtained from the previously
cited references, it has been found that other forms of presentation are more
suitable for design purposes. One such alternative presentation (Woolacot
and Macrae, 1967) is outlined in the following derivations.
Initially put K=(klpCP) in expression (6.2d) and rearrange it to get
the following:
rJNK(_!_)2 = (11T)
AT Cd
Multiply this expression by expression (6.2a ), to obtain
WK b -- -
d 2 AT = d(WQIH)
Substitute for rJN from expression (6.2c) into expression (6.2d ), to get
Finally, rearrange expression (6.2b ), to obtain
Nd 3
Q(d)
cd = d(Q)
b_
The coefficients on the right-hand sides of the above expressions are functions
of only the length-to-diameter ratio, bId, and the operating eccentricity ratio,
e, for a particular bearing. These three coefficients are now redefined as
follows:
- b -- -
W0 =d(WQIH)
T0 = (11T) (6.3)
- b -
Qo = d(Q)
172 TRIBOLOGY
so that the rearranged dimensional groups may be expressed as

WK -
d2ilT=Wo
(6.4)
HK
Nd 3 ilT Cd
(d) = - Qo
The values of W0 , T0 and Q0 have been evaluated from knowledge of

the appropriate values of W, Q and il as functions of the length-to-diameter
ratio, bId, and the eccentricity ratio, e. The important coefficients defined in
(6.3), and shown on the right-hand sides of the expressions of (6.4), enable
the full design of a hydrodynamic journal bearing to be approached in a
routine manner, as shown in the examples which follow.
The values of the coefficients W0 , T0 and Q0 are shown for practically
useful ranges of the length-to-diameter ratio, bId, and the eccentricity ratio,
e, in Figures 6.5, 6.6 and 6.7, respectively.
6.3.2 Example of Bearing Design with a Known Temperature

Rise
A journal bearing must be designed to support a radial load of 1750 N which

is placed on a shaft of 50 mm diameter. The shaft rotates at 10 000 rev /min
and the minimum film thickness is 1.5 X w- 2 mm. Experience shows that an
eccentricity ratio of not less than 0.7 is appropriate and the desired
temperature rise of the lubricant is to be 10°C. As the bearing will be well
cooled, it may be assumed that only 0.7 of the heat generated will be convected
from it by the oil. Design the bearing by determining its length, the diametral
clearance and the type of oil required.
Assuming that the previously taken values of oil density and specific
heat are applicable here (i.e. p = 880 kg/m 3 and CP = 1880 J /kg· 0 C}, then
K =~= 0•7 = 0.423 X 10- 6 m 2 • oc;N

pCP 880 X 1880
To determine the length of bearing consider the first expression of( 6.4 ):
w, = wK = 1750(0.423 x w- 6 ) = 29 .6 x 10 _ 3
0 d 2 tlT (50 X 10- 3 ) 2 X 10
Reference to Figure 6.5 shows that this value of W0 corresponds to a number
of different combinations of bId and e. However, as it is prescribed that e
1.0 bid-f -
8 1.5
6
5 == --
1.
-
-
v
1-· f-. 0.9
- - · -- - - -~0.8
2
3
- = 0.7
"'~ 0.6
0.
..
~
-·
0.4
10- '
8
0.3 -
6
5
4
-
3
0.
-
I-=
10- 2 = _, == - /
,-; 8
~ 6
;g 5
4 0.1 f-
] 3
1--
E
c0
·~ ..
e
0 10- l
8
6
5
4
3
1--
1o-•
8
6 - -
5
4
3
2
=
0.1 0.2 0.3 0.4 0.5 0.6 0 .7 0.8 0.9 1.0
Eccentricity ratio, c:
Figure 6.5 Variation of dimensionless coefficient W0 with eccentricity ratio e and

length-to-diameter ratio bId
be at least 0.7, it is seen from this figure that a value of e = 0.71 occurs when
bId= 0.4. It then follows that the length of the bearing is 0.4 x 50= 20 mm.
The diametral clearance now follows from expression ( 6.1) by putting
e = 0.71 and h = 1.5 x 10- 2 mm at 0 = n; thus,
2(1.5 X 10- 2 )
Cd= =0.103mm
1-0.71
174 TRIBOLOGY
0.03
= - - i...--
- - - - -
--- - - -
- -- -
- ~
- - - - -
0.02
-
- -
//./. ~
' ' """'...... -
-
~- ~
-
r--
·"""' f\\ - 1--
- """
0 -~
1>- ~~ ~
~'\2 ~\\ 1--
~
0.0
f-- [0..'\.\ \\\\\ -
/
............ ~ ~\W · b /d -
- - ~ ~ -
~""
~-:~-
~~
- -
~ 1\~~ '1- -
-
- - - - - \. \'~\
g:~ - -
\~
O.QI
0.7= --
n ,_ -
_1.0=
0.008
0.007
1-------j
f--- - -
--1
t--
- f- 1-
--
n rvv:
0 0.1 0.2 0.3 0.4 0.5 0.6 07 0.8 0.9 1.0
Eccentricity tatio, E
Figure 6.6 Variation of dimensionless coefficient 7'0 with eccentricity ratio e and
length-to-diameter ratio b / d
The type of oil necessary may now be found from Figure 6.6 by
determining the value of T0 corresponding to bId = 0.4 and a = 0.71; thus,
2
T_0 =rJNK(
-- -
d ) = 17.6 x 10 _ 3
!iT cd
The angular velocity is 10 000 I60 = 167 rev Is and the required dynamic
viscosity, rJ, follows as
rJ = 17.6 x 10- 3 {~~ ( ~d Y}

= 17.6 x 1o-3{ 10 (0.103 x 10-3)2}
167 X 0.423 X 10- 6 50 X 10- 3
rJ = 10.5 x 10- 3 kgl m · s = 10.5 cP
The corresponding kinematic viscosity is 10.510.88 = 11.9 eSt and
reference to Figure 5.2 shows that Tellus 22 oil has this viscosity at a
3.0 F - -
-- -
= 1- - -
2.0
b/ d -
- -
>----- - - 1.5-
f.-
- -
1.5 ~
>--- - 1-- 1 - -- L-
1.0_
;:::::;:;;;- ~ 0,9 _ -
-~ 0 .8 -
1.0
~ -~ O.J -
0 .9 0.6
::;;;
0.8
0.5 ====::
0.7
I~ 0.4
-~ 0.6
u 0.5 +---
"§ 0.3
E o.4
c0
·~ 0.2 -
E 0.3
i5
0.2
-
0.2
0.1 5
v
~
~ = - -""'
0.1 _
0.1
0 .09
...........
0.08
0.0 7
0.06
0.05
0 0.1 0.2 0.3 0.4 0.5 0.6 0. 7 0.8 0.9 1.0
Eccentricity ratio, t
Figure 6.7 Variation of dimensionless coefficient Q0 with eccentricity ratio e and

length-to-diameter ratio b/d
temperature of 59°C. As the temperature rise is 10°C, the inlet or feed

temperature of the oil must be 59- 10 = 49°C. The kinematic viscosity of
the oil at 49°C is seen to be 16.2 eSt, so that the dynamic viscosity of the
oil as it enters the bearing, tfc, is 16.2 x 0.88 = 14.3 cP.
Thus, the length of the bearing has been determined as 20 mm, the
diametral clearance has been determined as 0.103 mm, and a suitable oil has
been identified.
Two other parameters of interest may also be determined - the power
required to rotate the shaft and the flow rate required to ensure proper
operation of the bearing.
176 TRIBOLOGY
It is seen from Figure 6.7 that when b/d = 0.4 and e = 0.71, the value
of Q0 is 0.518. Thus, from the fourth expression of (6.4),
-
Qo= Hk (d)- =0.518
Nd 3 AT Cd
or
H =0.518{ Nd:AT( ~d)}
=0.518{167(50 X 10-3)3 X 10(0.103 X 10-3)}

0.423 X 10- 6 50 X 10- 3
H = 526 W = 0.7 h.p.
Similarly, from the third expression of (6.4),
- 0 =Q
Q -3 -
Nd cd
(d)
=0.518
or
Q = 0.518{ Nd 3 ( ;J}
= 0.518{ 167(50 X 10- 6 ) 3 ( 0· 103 X 10 - 3 ) }
50 X 10- 3
Q = 22.28 x 10- 6 m 3/s = 1.341/min
6.3.3 Further Aspects of Bearing Design
Consideration of the previous example reveals that it is a little unrealistic,

in that the temperature rise of the lubricant was initially fixed and the
necessary inlet or feed temperature of the selected oil then determined as an
operational requirement. However, in most practical cases it is the inlet
temperature of the oil that is known, perhaps approximately, so that the
temperature rise must be determined. A similar situation was considered in
the case of thrust bearings in sub-section 5.5.4, where an iterative procedure
was used to determine the initially unknown temperature rise as one of a
number of interrelated variables.
The application of a similar procedure to the design of hydrodynamic
journal bearings is described in a later example (sub-section 6.5.1).
The flow rate of fluid necessary to ensure proper operation of the journal
bearing considered in the last example was easily determined. However,
further consideration must be given to the associated problem of specifying
the details of the supply groove in order that the required flow rate of fluid
be able to enter the hydrodynamically generated film. Common grooving

arrangements are discussed in the next section.
6.4 OIL-SUPPLY GROOVES
In sub-section 6.1.3 the general shape and position of an oil-supply groove

in a bearing was shown schematically in Figure 6.4. In most cases the actual
groove is machined in the cylindrical liner which constitutes the bearing
proper and provides the stationary surface of the clearance. The type of
groove shown in this figure is known as the rectangular feed groove, but
several other widely used geometries exist (Martin, 1983a). Some of the more
commonly encountered grooving arrangements are discussed in the next
sub-section.
6.4.1 Geometry of Oil-supply Grooves
The purpose of an oil-supply groove is to supply the necessary quantity of

oil to the start of the convergent section of the clearance - i.e. to section E,
where () is zero, in Figure 6.3. A full specification of an oil-supply groove
demands a description of its basic geometric shape, its dimensions and its
position in the journal bearing relative to the line of action of the load. In
this sub-section it will be assumed that the direction of the load does not
vary, but some of the grooving arrangements discussed here are suitable for
loads which vary in direction or for shafts which are able to rotate in both
directions.
In all cases it is necessary for the lubricant to be pumped by external
means to the supply groove, and the determination of the supply pressure,
Pr• which is to be maintained in the supply groove is part of the design
procedure.
Four basic geometries of oil-supply grooves are shown in Figure 6.8
and discussed below.
The provision of a single circular hole in a bearing, as shown in
Figure 6.8(a), represents the simplest form of grooving. The presence of a
complete circumferential groove of the type shown in Figure 6.8(b) effectively
provides two separate bearings which are able to carry loads applied in any
direction and to accept shaft rotation in both directions. A variation of this
arrangement is shown in Figure 6.8(c), where the circumferential groove
extends over about 180°. This arrangement allows loads to be applied over
a limited range of direction but still accepts shaft rotation in either direction.
Figure 6.8( d) illustrates the general arrangement of a rectangular groove in
its basic form, but variations occur because of the choice of position available.
With reference to Figure 6.2, it would seem possible to locate a
rectangular oil supply groove in any position between sections G and E in
178 TRIBOLOGY
(a) Single circular hole
(b) Complete circumferential groove
(c) Partial circumferential groove
(d) Single rectangular groove
Figure 6.8 Types of lubricant-supply grooves
the divergent portion of the clearance. However, the practically acceptable

positions correspond in most cases to those shown in Figure 6.9.
Two common locations for a single rectangular oil-supply groove are
at 45° to the load line, as shown in Figure 6.9(a), and at 90° to the load
line, as shown in Figure 6.9(b ). Naturally, the direction in which these angles
are measured from the load line must correspond to the direction of shaft
rotation. It follows, therefore, that bearings equipped with only a single
~
-~-
(a) Single groove at 45°

to load line
~0"
~)1
~~-
T (b) Single groove at 90"

to load line
(c) Two grooves at 90°

to load line
Figure 6.9 Positions of rectangular lubricant-supply grooves
rectangular groove are not suitable for shafts which are able to rotate in
both directions or for loads which vary significantly in direction.
Provision of two rectangular grooves at right angles to the load line
overcomes these limitations to some extent, as shown in Figure 6.9( c). The
arrangement of Figure 6.9( c) is commonly adopted for constructional reasons,
as bearings are often split along a line at right angles to the load line in
manufacture.
Further consideration of oil-supply grooves will be limited to a
discussion of the flow of fluid from and the design of a single rectangular
groove in the locations shown in Figures 6.9(a) and (b).
180 TRIBOLOGY
6.4.2 Flow Rate from a Single Rectangular Groove
The evaluation of the flow rate from an oil-supply groove is complicated

(Cameron, 1981; Martin, 1983a), and in many cases only approximate
analyses are available and experimental experience is used to provide
adequately accurate means for the assessment of proposed designs. One such
method is described here, in which the necessary flow rate, Q, which has
already been established in sub-section 6.3.1, is considered to be composed
of two distinct components. The first of these components is that due to the
effect of the moving shaft in dragging fluid out of the supply groove by shear
action (sub-section 5.3.1). This component will be identified as the velocity-
induced flow rate, Qu. The second component of flow rate from the supply
groove is that due to the pressure which is maintained in the groove by
external means, e.g. a pump. This component of flow rate is termed the
pressure-induced flow rate, QP"
The object of designing an oil-supply groove is to ensure that the sum
of the velocity-induced and pressure-induced flow rates exceeds, or at least
equals, the flow rate required to operate the bearing. Thus,
Qu+QP~Q
In parallel with the second expression of (5.12), analysis shows that the
velocity-induced flow rate of fluid from a rectangular groove may be expressed
as
(6.5)
where the dimensionless coefficient Qu is dependent on the ratio between the
length of the supply groove and the length of the bearing, ljb and the
operating eccentricity ratio of the bearing, e, as shown in Figure 6.10.
Similarly, the pressure-induced flow from a rectangular groove may be
given as follows:
Qp=Qp{p~~~} (6.6)
Many assumptions have been made in the analyses which are summarised
by expressions (6.5) and (6.6 ), as the geometry of a rectangular groove results
in complicated flow patterns. Simplifications have made the curves of
Figure 6.10 independent of the length-to-diameter ratio of the bearing, bjd,
and the angular position of the groove, but the length of the groove, l, does,
of course, appear in the ratio ljb.
The value of the dimensionless coefficient QP is conveniently expressed
as the product of two other coefficients, Q1 and Q2 , which are given by the
curves of Figures 6.11(a) and 6.11(b ), respectively. Thus, QP = Q1 Q2 , so that
QP = Ql Qz { p~~~} (6.7)
2.0
1.5 1/b=
0.9 -
_;.... 0.~=
1.0 0.7=
0.8 0.6=
o.s=:
0.6 0.4::::=
0.5
0.3::;.::;:
0.4 _...,..,
1---C 0.25=
0.3
-- - -A < - / / . /
_.... - .2=
"'
~
10
~
;g 0 .2
f - - 1- //
~E o.25 =
f--
~c
e 0.08 =0.1 7
0 -
0.06
0.0 5
=
0.04 1 - - -
·-
0.0 3 F==
f- 1-
0.02
0 .025 1--
1------ - 1---
0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0 .7 0.8 0.9 1.0

Eccentricity ratio. t
Figure 6.10 Variation of dimensionless coefficient Q. with eccentricity ratio e and

length ratio 1/ b
The value of Q1 is affected by the dimensions of the rectangular groove, a

and I, as they are introduced through the ratios aj b and 1/ b, while the two
selected angular positions of the groove and the eccentricity ratio are reflected
in the value of Q2 •
Acceptance of the above presentation now enables the dimensions of
the supply groove required to ensure operation of the bearing of sub-section
6.3.2 to be determined along with the necessary supply pressure, as shown
in sub-section 6.4.4.
6.4.3 Effective Temperature and Viscosity in Hydrodynamic

Journal Bearings
It was explained in sub-section 5.5.1 that account is usually taken of the

decrease in the viscosity of oil as it passes through a thrust bearing by basing
calculations on an effective viscosity. The effective viscosity was defined for
182 TRIBOLOGY
1.5
r-- 1/ b
1.0
- ·- ·-
-=F=- 0.9- -
O.B o.8 =
10 0.7 =
~ 0.6
o.s:
;g 0.5 0.5 - -
] 0.4 - -
~
0.4
-- .. ~
0.3 - · -
"
-~
0 .3 0.2
e 0 .2 --
c
·:
0 - c:-
. - --
1-
0.15
-- - - ·-
0.1
0 0.1 0.2 0.3
Width ratio, alb

(a)
1.0
0.8 · -1-- ·
10 0.6
~
·c;
0.5
;;: 0.4
8 · -1-···
"~ 0.3 __ _._-
c - r-- - f--- - .
·~
e 0.2
c
0
0. 15
0 0.1 0.2 0.3 0.4 0.5 0.6 0. 7 0.8 0.9 1.0
Eccenuicity ratio, t
(b)
Figure 6.11 Variation of dimensionless coefficients (2 1 and Q2 with eccentricity

ratio e, length ratios af b and 1/ b, and groove position
a thrust bearing as that at the effective or mean temperature of the oil during
its passage through the bearing. Most of the oil which enters a thrust bearing
passes completely through the converging clearance and is nearly all available
to convect heat from the bearing. Consequently, the effective temperature
was defined as the sum of the inlet temperature and a half of the temperature
rise as given by the power balance (sub-section 5.5.4 ).
However, in most journal bearings the width of the load-bearing film
is much less than its length. For example, in a journal bearing with the
relatively high length-to-diameter ratio of unity the axial width of the

load-bearing film is only about ~ of its circumferential length, as it extends
over an arc of only about 180°C. Therefore, a considerable proportion of
the oil which enters the convergent, and load-bearing, clearance escapes
axially, so that not all of the oil which enters the bearing is equally effective
in convecting away the heat which is generated in the load-bearing film.
It follows that the temperature of the oil at exit from the load-bearing
arc (section Gin Figure 6.3) attains a higher value than that which would
have occurred if all of the oil which entered at section E had passed completely
through the bearing.
This higher exit temperature results in the effective temperature of the
oil in a journal bearing being more accurately represented by the sum of the
inlet temperature and the whole of the temperature rise found from the power
balance rather than a half of it, as in the case of a thrust bearing. This way of
assigning the value of the effective temperature was used in sub-section 6.3.2
and will be used later in another example (sub-section 6.5.1)
6.4.4 Example of Determination of the Dimensions of a

Supply Groove
This example simply completes the design of the bearing commenced in

sub-section 6.3.2 by determining the dimensions of the rectangular oil-supply
groove required, subject to the condition that the supply pressure in the
groove is limited to a maximum of 0.4 Njmm 2 •
It has already been shown in sub-section 6.3.2 that the flow rate required
to operate the bearing is 1.341/min and that the viscosity of the oil in the
supply groove is 14.3 cP. It is required that Q" + QP ~ Q.
Study of Figure 6.10 reveals that an estimate of the length of the groove,
l, must be made at this stage by choosing a value for the ratio 1/b. Take the
geometrically reasonable value of 1/b = 0.65 and B = 0.71 in Figure 6.10, to
get llu = 0.7.
The velocity-induced flow rate, Q", then follows from Equation ( 6.5) as
Qu = iJu(NbdCd)
=0.7{167(20 X 10- 3 )(50 X 10- 3 )(0.103 X 10- 3 )}
Qu= 12 x 10- 6 m 3 /s=0.721/min
The pressure-induced flow, QP, may now be determined by use of

expression (6.7) and Figures 6.1l(a) and 6.11(b). From Figure 6.1l(a), with
the assumption that the ratio ajb is 0.25 and 1/b = 0.65, the value of Q1 is
seen to be 0.59. If it is then decided to place the supply groove at an angle
of 45° to the load line, the value of Q2 is given in Figure 6.1l(b) as 0.62 for
an eccentricity ratio of 0.71.
184 TRIBOLOGY
The pressure-induced flow rate then follows from Equation (6.7) as
Qp = Ql Q2( p~~J)
= 0.59 X 0.62{(0.4 X 106)(0.103 X 10-3)3}
14.3 xw- 3
Qp= 11.2 X 10- 6 m 3ls=0.67llmin
Thus,
Qu + QP = 0.72 + 0.67 = 1.39llmin
This flow rate is greater than the required value of 1.34llmin and is therefore
acceptable.
The length of the axial groove is therefore
l = 20 x 0.65 = 13 mm
and its width is
a = 20 x 0.25 = 5mm
6.5 DETERMINATION OF TEMPERATURE RISE
The form of the dimensionless coefficients introduced in sub-section 6.3.1

provide a means by which the temperature rise in a journal bearing, relative
to a given inlet temperature, may be determined, along with appropriate
values of length-to-diameter ratio, eccentricity ratio and diametral clearance.
Such evaluations inevitably involve the viscosity temperature characteristics
of oils and these, too, may be included directly in the design process.
Iteration is usually necessary to achieve a satisfactory design, and one
such scheme of iteration is illustrated in Figure 6.12 and described briefly
below.
In general, the diameter of the bearing and the load applied to it are
known, so that W0 may be evaluated for an initially assumed value of the
temperature rise, ~ T, from the first expression of ( 6.4 ). This value of W0 may
be entered in Figure 6.5 and corresponding values of bId and e selected.
Once e is known, the diametral clearance, Cd, may be calculated, as the
minimum film thickness is usually specified. Figure 6.6 may now be entered
with the known values of bId and e, to determine the value of '1"0 • The second
expression of ( 6.4) then enables the necessary value of the effective dynamic
viscosity to be calculated.
AT, Initial value
AT
I::J. T, Derived value
AT
T, AT
T,
Figure 6.12 Iterative scheme for design of hydrodynamic journal bearings
The corresponding value of the kinematic viscosity, v, is then entered

in Figure 5.2, along with the now known effective temperature of the oil, as
given by the sum of the inlet temperature and the assumed value of
temperature rise, AT.
At this stage the necessary kinematic viscosity should correspond with
the characteristic of the selected oil at the calculated effective temperature.
If no correspondence exists, then the process may be repeated by choosing
another value of the temperature rise, AT, until the variables involved give
consistency in all three of the characteristics represented in Figure 6.12.
Once this agreement has been established, the other important operating
parameters of the bearing may be studied. The following example illustrates
this method of iteration for determining a consistent set of variables.
186 TRIBOLOGY
6.5.1 Example of Bearing Design with a Known Inlet

Temperature
A steam turbine rotor is supported by a pair of hydrodynamic journal

bearings, each of which carries a steady radial load of 200 kN. The bearings
have a diameter of 200 mm and the rotor revolves at a steady speed of
3000 rev /min. Tellus 46 oil is supplied to the bearings at an inlet temperature
of 50°C and a maximum supply pressure of 0.9 Njmm 2 • The minimum oil
film thickness allowable in bearings of this type is 0.02 mm and their
length-to-diameter ratio should be between 0.55 and 0.7. Design suitable
bearings for this rotor and specify the axial length of the bearings, the
diametral clearance, the dimensions and locations of the oil-supply grooves,
and the supply pressure of oil in the grooves. Give the following details of
the operating characteristics of the design: the operating eccentricity ratio,
the required rate of delivery of oil for each bearing and the power dissipated
by rotation of the shaft within each bearing.
Clearly, the temperature rise of the oil above the value of 50°C at inlet
is to be determined, along with the bjd ratio, the diametral clearance, Cd,
and the operating eccentricity ratio, e. It will therefore be necessary to use
the iterative technique described on pp. 184-185.
Assume that the value of k is 0.8, as no further information is given on
the way in which the bearings are cooled. The value of K then follows from
sub-section 6.3.2 as
k 0.8 6 2
K=-= =0.484x 10- m ·°CjN
pCP 880 X 1880
and the value of W0 follows as

_ wK 200 x 10 3 x 0.484 x w- 6 2.418
(a)
wo = d 2 AT = (200 x w- 3 ) 2 AT AT
The required diametral clearance is, from expression (6.1 ),
2(0.02 x w- 3 ) o.o4 x w- 3
Cd= = m (b)
1-B 1-B
and '1"0 is, as N = 3000/60 =50 rev /min:
f. = 1'fNK (!!__) 2 = 50 X 0.484 x 10- 6 (200 x 10- 3 ) 2
0 AT Cd '1 AT Cd
or
-
T0 = 0.967 x 10 -6(-ATq- '1 ) (c)
In addition, the effective temperature of the oil will be (50+ AT)°C. The
sequence of calculations outlined in Section 6.5 is now carried out for
various values of AT until convergence occurs between Figure 6.5 and
Figure 6.6 and the viscosity temperature characteristics for Tellus 46 oil given
in Figure 5.2. Take AT= 20°C; then, from expression (a) above,
- 2.418
W0 = - - = 0.129
20
Reference to Figure 6.5 shows that this value corresponds to bid= 0.7 and
6 = 0. 76 or bId = 0.5 and 6 = 0.81. Take the first pair of values and proceed
to find cd from (b) as
o.o4 x w- 3
C - 0.167 X 10- 3 m
d- 1-0.76
Put this value in (c) and recall that, as both bId and 6 are now known, the
value of T0 may be read off Figure 6.6 as 14.52 x 10- 3 ; hence,
14.52 x w- 3 = o.967 x w- 6 { 20(0.16717x w- 3 )2 }

from which 17 = 0.0084 kglm ·s or v = 9.55 eSt, a value which does not
correspond with the viscosity of 15 eSt given for Tellus 46 at an effective
temperature of (50+ 20) = 70oC in Figure 5.2.
Repetition of this process for different values of the temperature rise,
AT, in accordance with the above specimen calculation and the method of
Section 6.5 yields the following set of consistent results for AT= 25 oc.
From (a)
W - 2-
-0 - 18 -
.4- - 0 •096 7
25
which corresponds to bid= 0.6 and 6 = 0.77 in Figure 6.5.
The value of Cd then follows from (b):
o.o4 x w- 3
C= =0.174x10- 3 m
d 1-0.77
With bid= 0.6 and 6 = 0.77, the value of T0 may be read off Figure 6.6
as 15 X 10- 3 , SO that (c) becomes
15 x w- 3 = o.967 x w- 6 { 25(0.17417x w- 3 )2 }
for which 11 = 0.0117 kglm · s or 13.3 eSt.
The effective temperature of the oil is (50+ 25) = 75°C and Figure 5.2
shows that the kinematic viscosity of Tellus 46 at 75°C is about 13.5 eSt, so
that the values corresponding to a temperature rise of 25°C are consistent.
188 TRIBOLOGY
With b/d = 0.6 and e = 0.77, the value of Q0 may be read off Figure 6.7
as 0. 78, so that the power loss follows from the fourth expression of( 6.4) as
H= Qo{ Nd~AT( ~d)}

= Q. 78 .{50(200 X 10- 3 ) 3 25 (0.174 X 10- 3 ) }
0.484 X 10- 6 200 X 10- 3
H= 14 X 10 3 W= 18.8h.p.
Again the required flow rate follows from the third expression of ( 6.4) as
= 0.78{50(200 X 10- 3 ) 3 (QJ20074X 10 - ) }

X 10- 3
3
Q = 0.271 x 10- 3 m 3 /s = 16.281/min

The velocity-induced flow, Qu, may be found from Figure 6.10 with an initially
assumed value of 1/b as, say, 0.6, so that Qu becomes 0.7 and Qu is then
Qu = Qu(NbdCd)
= 0.7 {50(0.6 X 200 X 10- 3 )(200 X 10- 3 )(0.174 X 10- 3 )}
Qu = 0.146 x 10- 3 m 3 /s = 8.761/min
Thus, the minimum pressure-induced flow rate is Q- Qu = 16.28-8.76 =
7.52 1/min = 0.125 x 10- 3 m 3 /s.
Now assume that the supply groove is placed at 90° to the load line,
with the ratio a/b set at 0.25. Figures 6.11 then give Q 1 = 0.54 and Q2 = 0.35,
so that the pressure-induced flow becomes, from expression (6.7),
QP = Q1Q2{Pr~:}
Now the viscosity of the oil at an inlet temperature of 50°C is, from
Figure 5.2, 30 eSt or 30(0.88)/1000 = 0.0263 kg/m·s. The supply pressure
required to provide the necessary flow rate of oil may now be found from
the above expression as
0.125 X 10- 3 = 0.54 X 0.35 {Pr(O.l0.0263

74 x 10- 3 3
) }
Pc = 3.31 x 10 6 N/m 2 = 3.31 N/mm 2

This value of Pc is in excess of the maximum available supply pressure of
0.9 N/mm 2 • Changes in the various assumptions may now be made in order
to evolve an acceptable design. By taking l/ b = 0.8 it is found that Qu becomes

0.9 and Qu = 0.187 x 10- 3 m 3 js. The required pressure-induced flow then
becomes
QP = 0.271 X 10- 3 -0.187 X 10- 3 = 0.084 X 10- 3 m 3 js
With the supply groove placed at 45° to the load line and ajb maintained
at 0.25, it is found that Q1 = 0.78 and Q2 = 0.67; thus,
0.084 x 10- 3 = 0.78 X 0.67{Pr(0. 174 X 10- 3 ) 3 }

0.0263
Pr = 0.8 x 10 6 N/m 2 = 0.8 N/mm 2
This value of Pr is less than 0.9 and therefore acceptable. The length of the
bearing is 0.6(200) = 120 mm and the dimensions of the oil-supply groove
are as follows:
l = 0.8( 120) = 96 mm
a= 0.25( 120) = 30 mm
6.6 INSTABILITY IN JOURNAL BEARINGS
Rotating shafts in hydrodynamic journal bearings are subject to two possible

forms of instability.
The first, often called synchronous whirl, is caused by a periodic
disturbance outside the bearing, such as in balance of the rotor, which excites
the bearing-shaft system into resonance. Since it is impossible to produce
a perfectly balanced shaft, the centre of the shaft will not normally be
stationary but will describe a tiny circle or other closed orbit around the
equilibrium position. If this orbit is constant, the whirl is stable. However,
if the speed of rotation is progressively increased, at some stage the rotational
frequency of the rotor in balance will coincide with the first natural frequency
of the rotor system and the whirl orbit will increase. At this speed either the
whirl will become stable at a bigger amplitude or failure will take place. The
magnitude of this first critical speed depends on the rotor stiffness and inertia
and on the stiffness and damping characteristics of the bearings. On further
increasing the speed, the whirl subsides until the second critical speed is
reached and resonance will once more occur. If the shaft is accelerated rapidly,
these critical speeds can be passed through before the whirl orbit has time
to grow large enough to damage the bearing. Clearly, it is desirable that the
running speed be selected to avoid these critical speeds.
The other form of instability is called 'half-speed whirl' or 'oil-whip'
and is induced by the lubricant film itself(Holmes, 1963). We have seen that,
when loaded vertically downward, the centre of the journal does not simply
190 TRIBOLOGY
move down in the direction of the load but also moves around the bearing
in the direction of rotation (Figure 6.3). In some cases this motion can
continue, so that the shaft centre describes a circular orbit. If this rotation
takes place at half the rotational speed, it will coincide with the mean
rotational speed of the lubricant and no relative motion between the film
shape and the lubricant will occur, so that the hydrodynamic mechanism
will be destroyed.
This phenomenon will also take place at a certain threshold speed but,
unlike synchronous whirl, there is no possibility of running through half-speed
whirl, further increases in speed only accelerating the failure of the bearing.
6.6.1 Prediction and Suppression of Instability
Synchronous whirl is usually avoided by either increasing the bearing stiffness,

so that the first critical speed is well above the commonly used running speed
of the rotor, or decreasing the stiffness, so that the critical speeds are quickly
passed through and normal running takes place where the attenuation is large.
Another means of suppressing or allowing for whirl is the introduction
of extra damping into the system. This may be done by flexibly mounting
the bearings in, for example, rubber or metal diaphragms.
Half-speed whirl is most likely to become a problem in lightly loaded
bearings and it is often possible to raise the threshold speed by redesigning
the bearing to run at a higher eccentricity ratio. It is, therefore, very
troublesome in vertically mounted spindles, where the horizontal load is
negligible and the operating eccentricity ratio correspondingly small.
By mounting the bearings flexibly, as described earlier, extra damping
may be incorporated into the system and some of the energy generated by
the whirling will be absorbed in the damping medium. This form of instability
is very common in bearings lubricated by a gas, because of the very-poor
damping properties of the lubricant itself. Indeed, this is often the limitation
on the use of gas journal bearings, rather than the lack of load capacity
(Marsh, 1964 ).
The criterion given for the onset of whirl (ESDU 84031, 1984) can be
written as MeN 2 I W < 0.2, where c is the radial clearance, M the mass of
the rotating shaft, W the load carried by the bearing and N the angular
velocity. For a bearing with a purely gravitational loading due to the mass
of the shaft, this becomes cN 2 I g < 0.2. More information on bearing
instabilities may be found in the references (ESDU 77013, 1977).
An effective way of preventing or delaying half-speed whirl is to interfere
with the circumferential symmetry of the bearing. An axial groove in the
low-pressure region, two or more lobes in the bore or even accidental
out-of-roundness errors in manufacture can be sufficient to delay the onset
of whirl (Malik et al., 1982; Lanes et al., 1982).
6.7 PRACTICAL DESIGN OF JOURNAL BI;ARINGS
This chapter has described the operation of hydrodynamic journal bearings

and presented a method of evolving complete designs of bearings for relatively
uncomplicated cases. Other methods of bearing design have been derived
(Barwell, 1979; ESD U 84031, 1984) and methods of optimisation proposed
(Moes and Bosma, 1971). Many considerations enter into the full design of
journal bearings, and the choice of materials to be used in their construction
is of importance (ESDU 88018, 1988). Special considerations must be taken
into account in some applications where, for example, dynamic loading is
of significance (Martin, 1983b ).
6.8 PROBLEMS
1 For a range of eccentricity ratio 6 between 0.2 and 0.7, the dimensionless
load coefficient, W (defined in expression 6.2a ), is found to be related to
the eccentricity ratio by the following expressions:
log 10 (W) = 2.24966- 1.194, when bjd = 0.3
log 10 (W) = 1.9186-0.347, when bjd = 0.8
A complete journal bearing has a diameter of 80 mm, a length of 64 mm
and a diametral clearance of 0.08 mm, uses an oil of effective dynamic
viscosity 0.04 kg/m ·sand supports a shaft which rotates at 600 rev /min.
(a) The radial load imposed on the bearing in service varies between
3000 Nand 18 000 N. Estimate the corresponding values of eccentricity
ratio.
[0.267 and 0.673]
(b) The bearing is now modified by machining in it a complete centrally
located circumferential groove of width 16 mm, as shown in
Figure 6.8(b ). Find the value of the diametral clearance required to
ensure that this bearing operates at an eccentricity ratio of 0. 7 under
the full load of 18 000 N.
[0.036 mm]
2 It is found that in the range of eccentricity ratio 6 between 0.2 and 0.7
the dimensionless load coefficient, W (defined in expression 6.2a ), may
be related to the eccentricity ratio and length-to-diameter ratio of the
bearing, bId, by the following expression:
W = 0.68 (d
b)1.76 e 4 · 2'
192 TRIBOLOGY
A journal bearing has a diameter of 80 mm, a length of 64 mm and a

diametral clearance of 0.08 mm, uses an oil of effective dynamic viscosity
0.04 kg/ m · s and operates at 600 rev I min.
(a) The load carried by the bearing in service varies between 3000 N
and 15 000 N. Estimate the corresponding range of eccentricity ratio
and the minimum film thickness under the larger load.
[0.276-0.66; 0.014 mm]
(b) What should be the length of the bearing if it is to operate at an
eccentricity ratio of0.5 uader a load of 3000 N? Its other dimensions,
the viscosity of the oil and the speed of rotation remain unchanged.
[45.6mm]
(c) The length of the bearing is now fixed at 48 mm while carrying a
load of 3000 N. Investigate the variation of eccentricity ratio and
minimum film thickness when the diametral clearance varies between
0.050 mm and 0.070 mm. Hence select the diametral clearance which
gives a minimum film thickness of 0.02 mm.
[0.058 mm withe= 0.312]
3 A short rotor of total mass 1000 kg is supported by two identical

hydrodynamic journal bearings of length 50 mm, diameter 100 mm and
diametral clearance 0.14 mm. The rotor is driven at a steady speed of
1200 rev /min and the effective dynamic viscosity of the oil is 30 cP. It
may be assumed that for a journal bearing with a length-to-diameter
ratio of 0.5 the load coefficient, W (defined in expression 6.2a ), is related
to the eccentricity ratio, e, by the following expression:
W = 0.18e 4 · 8 '
(a) Show that the stiffness of the bearing for small displacements about
any steady operating eccentric position defined by the eccentricity,
e, may be found as:
where oWjoe is the stiffness, or rate of change of load with small

displacement, '1 is the dynamic viscosity of the lubricant, N is the
rotational speed of the rotor, b is the length of the bearing, d is its
diameter and C d is the diametral clearance.
(b) Show that the bearings of the assembly described above operate with
a steady operating eccentricity ratio of about 0.6.
(c) Estimate the stiffness of each bearing and the fundamental undamped
free frequency of vibration of the rotor in its bearings.
[335 x 10 6 N/m, 130 c/s]
4 A circular shaft rotates at 2400 rev /min and is supported in two

hydrodynamic journal bearings, each of which carries a steady radial
load of 60 kN. The shaft has a diameter of 240 mm and one of the Tell us
range of oils may be supplied to the bearings at an inlet temperature of
30°C and a maximum supply pressure of 0.8 N jmm 2 . Experience shows
that the minimum film thickness in the bearings must not be less than
0.04 mm. Each bearing is to have an axial length of 96 mm, and as the
bearings' housing is very well insulated, it is estimated that 90% of the
heat generated by rotation of the shaft is convected away by the oil.
Design a suitable bearing for this shaft and give the following details of
the full design:
(a) The operating eccentricity ratio in the bearings.
(b) The diametral clearance between the shaft and bearings.
(c) The dimensions of the oil-supply groove and the pressure in it.
(d) The required rate of delivery of oil to each bearing.
(e) The mean temperature rise of the oil as it flows through the bearings.
[One consistent set of answers is as follows:
(a) 0.7, (b) 0.27 mm, (c) 77 mm X 24 mm, (d) 19.21/min, (e) 20.3 °C]
Chapter 7
The Lubrication of Highly
Loaded Contacts
7.1 INTRODUCTION
In Chapter 5 the hydrodynamic build-up of a lubricating film between the

moving surfaces was examined. However, if the pressure between the surfaces
becomes very high, there are two extra effects which have to be taken into
account when assessing the performance of the lubricated contact: the increase
of viscosity with pressure and the local elastic deformation of the surfaces.
Both these effects lead to an increase in film thickness compared with the
simple hydrodynamic theory, and the.mechanism of film generation in this
case is known as elastohydrodynamic lubrication (EHL).
This situation is most likely to occur either where the load is concentrated
over a small area, owing to the surfaces being non-conforming, such as in
rolling element bearings, gears and cams, or where at least one of the surfaces
is of a material oflow elastic modulus, such as a rubber tyre rolling on a road.
7.2 EQUIVALENT CYLINDER
Two different forms of contact are identifiable.

( 1) The contact such as between a ball and the race of a ball bearing, where
the area of contact is nominally a point but, because of elastic deformation,
becomes elliptical or, in the special case of a sphere and a flat plate,
circular. This area of contact is sometimes referred to as the footprint.
(2) The contact, such as that between a cylindrical roller and a flat raceway,
which produces an essentially rectangular footprint.
Most EHL rectangular contacts, such as that shown in Figure 7.1, can
be represented with sufficient accuracy by an equivalent cylinder against a
194
THE LUBRICATION OF HIGHLY LOADED CONTACTS 195
(b)
---Uz
Figure 7.1 Equivalent cylinder
flat plane, where the radius of the equivalent cylinder is calculated as in

Hertzian contact (Section 2.2 ). This is achieved by putting
RlRz
R = -=-----==- (7.1 )
Rl +Rz
where the surface velocities of the two components are preserved. In the case
of a concave surface the radius would be negative.
The same procedure is adopted for the elliptical footprint, where an
equivalent radius of curvature is calculated in the two principal directions,
Rx being along the line of fluid entrainment and Ry being perpendicular to it.
7.3 RIGID CYLINDER THEORY
The film shape between the cylinder and plane has the form
h = h0 + R( 1 -cos fJ)
Solution of the Reynolds equation must clearly include the possibility of
cavitation in the divergent outlet region, as described in Section 6.13. An
approximation to the film shape is given by assuming a parabolic profile
xz
h=h 0 +-
'2R
Solutions were obtained for a parabolic cylinder (Martin, 1916; Purday,

1949) and for a circular cylinder (Floberg, 1961). Since Reynolds's equation
cannot really hold over the entire 90° inlet zone that Floberg considered,
the results of this analysis can only be regarded as accurate for relatively
thin films, where the pressure generated is largely concentrated in a small
area near the centre. In this case the solutions coincide. The load capacity
of the contact can be written
196 TRIBOLOGY
F,
- u,'
Figure 7.2 Cylinder and plane
The value of f3 depends on R/h 0 , but for R/h 0 greater than about 10 4 may
be taken as 2.45.
The sideways force on the cylinder due to the horizontal component of
pressure is given by
Fx=4.S17(U1 + Uz)B {R (7.2)

-Jho
for large values of R/ h0 • For the plane, of course, Fx = 0. If the real situation
consisted originally of two discs, clearly there will be a horizontal force on
each one which is given by
R
Fx, =-F
R X
1
R
Fx2 =-F
Rz x
where R 1 and R 2 are the radii of the original cylinders.
The viscous drags along the plane surface and around the cylinder
circumference are
(7.3)
Fx
F 2 =--+(U 1 -U 2 )AB
2
The value of A, assuming a full contribution from the cavitated region, is
found to be 3.84 for cases where R/ h 0 is large.
The forces described above are shown in Figure 7.2.
7.4 ELASTOHYDRODYNAMIC MECHANISM
When a load of significant size is imposed on a non-conforming contact,

both surfaces will suffer elastic deflection to change the nominal point or
line contact into a footprint having an area which is significant compared

with the dimensions of the lubricant film. This produces a 'flat' on balls and
rollers, with the load carried by a pressure spread over this flattened area in
'Hertzian' contact, as described in Chapter 2.
When the surfaces move, they will drag lubricant into the convergent
inlet zone in the familiar hydrodynamic fashion. The lubricant will then pass
through the flattened zone, to emerge in the divergent outlet, where cavitation
will normally take place. For many cases the presence of the film does not
significantly affect either the size or the shape of the flattened area and,
indeed, the Hertzian pressure distribution can remain essentially unchanged.
As the lubricant passes through the gap, it is subject to an increasing
pressure. Most fluids, with the notable exception of water, exhibit a marked
increase in viscosity with increasing pressure. Indeed, in very-high-pressure
contacts the viscosity can increase several hundredfold, rendering the fluid
almost solid in its effect. The result of both these phenomena is drastically
to increase the load-carrying abilities of the lubricated contact relative to
that calculated by rigid body, isoviscous theory.
7.5 ELASTOHYDRODYNAMIC THEORY
To produce results of use to the designer of highly loaded contacts where

viscosity increase and elastic deflection are important, the theory described
above for rigid cylinders gives film thickness values which are severely
underestimated. In practice the local elastic flattening of the curved surface
spreads the load over a larger area and the viscosity increase enhances the
film-building properties of the lubricant. To analyse this situation it is
necessary to carry out the simultaneous solution of the Reynolds equation
(7.4 ), the elastic deformation equation (7.5) and the equation relating viscosity
to pressure (7.6):
!_(ph3. op) + !_(ph3. op) = 12Vo(ph) (7.4)

ox 11 ox oy 11 oy ox
v=-
1 fx, p(s)log(x-s)·ds (7.5)
E' x,
(7.6)
If isothermal conditions no longer prevail, the energy equation and the
conduction equation for the heat passing to the surfaces must be incorporated.
The cavitation boundary condition must also be employed as described in
sub-section 6.1.2.
Although a full solution had to wait until the introduction of the
high-speed digital computer, inspired assumptions produced very good
198 TRIBOLOGY
approximate solutions as early as 1949 in the work by Ertel (Grubin, 1949)

and in the next decade by several workers (Petrusevich, 1951; Weber and
Saalfeld, 1954 ). Even today analytical methods can give a valuable insight
into the mechanisms at work (Archard and Baglin, 1986). However, the more
accurate computer results, beginning with those of Dowson and Higginson
in 1959 and extended and refined subsequently by many workers, are used
as a basis for the design procedures described in this chapter.
7.6 RESULTS
7.6.1 Film Shape and Pressure Distribution
Results from the foregoing analyses permit the plotting of the shape of the
distorted cylinder and the pressure distribution" for a given set of conditions.
Typically, for a relatively slow-speed, rectangular footprint these will have
the form illustrated in Figure 7.3.
Considering first the film shape, we see the convergent inlet zone, followed
by a virtually parallel section. At outlet there is a constriction which can
amount to a reduction of 25% of the parallel film thickness.
The pressure curve is very close to the Hertzian dry contact pressure
shown in Chapter 2. There is a build-up in the inlet and towards the outlet
there is a pressure spike and the cavitation boundary. The spike is difficult
to find experimentally, because it is very narrow (Kannel, 1965), but it can
be shown theoretically to exist and is important because of the possibility
of high subsurface stresses. There is evidence that, if a more accurate
pressure-viscosity relationship is used (Roelands, 1966), the calcUlated
severity of the spike may be reduced. Of course, the total area under the
curve must be the same as that under the Hertzian half-ellipse. The length
of the parallel section will be approximately 2{(8RW/nBE')}l and the
maxirimm pressure (WE' /2nR )f.
As the speed increases, the pressure distribution will depart more and
more from the Hertzian, as shown in Figure 7.4.
Pressure , - - ,
/ \ ,,P'
II '''
·.1
I 'j
I I
I ~
-/i
I
.
--- r:
Figure 7.3 Pressure distribution and film shape in EHL

Figure 7.4 Effect of velocity on pressure distribution
For the elliptical footprint the results are very similar, as shown in
Figure 7.5, which is an interference pattern for a circular EHL footprint.
There is still an essentially parallel central section, while the constriction at
outlet now extends around about 180° of the contact circle (Gohar and
Cameron, 1966).
Ex11 ~m
Qron<;~e
Purple
Green
I• Gr~:~
1 Yellow
Yellow
Figure 7.5 Interference pattern in elliptical contact, k = 1

200 TRIBOLOGY
7.6.2 Film Thickness
Formulae for film thickness are the result of fitting an algebraic expression
to a large number of computed results. Such a formula only applies to the
particular range of results from which it was derived and must not be
extrapolated into wildly different areas. For this reason we need to define
'regimes' for film thickness calculation and to use the formula appropriate
to each regime (Johnson, 1970). The regimes are defined as follows.
( 1) Rigid-isoviscous In this case the pressures are not high enough to cause
appreciable viscosity change or elastic deformation.
(2) Rigid-piezoviscous The elastic deformation is negligible, but there is
significant viscosity increase with pressure.
(3) Elastic-isoviscous Considerable elastic deformation, but negligible
viscosity change.
( 4) Full EHL solution, including deformation and viscosity change
In order to determine the appropriate regime, it is necessary to refer to
the relevant chart (Figures 7.6-7.9). The horizontal axis of the charts is the
criterion for elastic deformation, g., and the vertical axis is the criterion for
significant viscosity increase, gv. To evaluate these criteria, it is necessary to
examine the dimensionless groupings of the next sub-section.
7.6.3 Dimensionless Groups
The results from the computer analyses are conveniently presented in the
form of dimensionless groups. This has the advantage of ease of plotting,
independence of the system of units used and, if the groups are chosen
intelligently, the possibility of physical insight.
The most popular scheme is to use the following four groups:
the dimensionless film thickness fl = h0 1R or fl = h0 1Rx for elliptical
contacts
the load parameter W= WI E' RB for rectangular contacts or
W = WI E' R; for elliptical contacts
the speed parameter 0 = 1'/o U IE' R for rectangular contacts or
0 = 1'/o U IE' Rx for elliptical contacts
the materials parameter G = rx.E'
For the elliptical contact we also need the ellipticity parameter
k=alb (7.7)
where a is the semiaxis of the contact ellipse in the tranverse (y) direction
and b is the semiaxis in the ( x) direction of motion. It is sufficiently accurate
10000 9h ~ 500
9h ~ 300
I
I
9h ~ 200 I
1000
I
9h ~ 100 Rigid
9h 70 piezoviscous
~
I
9h ~50 I
I
., 100
9h ~ 30
I
~
9h ~ 20
E I I
~ I
Q_
~
Uh = 10 I
'iS I
I;!
> 10
9h ~ 7 I
;,~
~c¥0
·~0
I I ~v
9h = E -5'
0
I
1.0 I
0.1 1.0 10 100 10000
Elasticity parameter 9e
Figure 7.6 Regimes of rectangular contact
10 7
10"
g.~ 10000 /
10• g. ~84000
.;; Piezoviscous /
.;; 10' Piezoviscous 40000 /
~.,-
!
~
rigid 2500 /
rigid 20000 0..>'
~
·,r.,V /
105
1000 / 0~ 106 10000 ~"" /
~'~?1
/
~'~·11
~
Q_
5000
~ Q_
-~
10 4 105 1832 /
0
I;! .8~ /
> I;! / / lsoviscous
10 3 > 104
lsoviscous
rigid rigid
10 2 103
10 10 2 10 3 104 10 5 10•
10 2 10 3 104 105 10• 10 7
Elasticity parameter 9e Elasticity parameter 9e
Figure 7.7 Regimes for k = 1 Figure 7.8 Regimes for k = 3

202 TRIBOLOGY
109 ~,qP "'

9h=450000
~,~"':~c.
<l
ti; 1o" PiezoviscOus /
I
rigid 150000 /
/
10 7 65000 / /
0.
20000 /
~
/
'§ 10" /
/
~ 6562/ / lsoviscous
> 105
104
10 3 104 10 5 106 10 7 10 8
Elasticity parameter Ue
Figure 7.9 Regimes fork= 6
to derive k from the simple formula (Hamrock and Brewe, 1983)

_ (Ry)o.64
k- - (7.8)
Rx
7.6.4 Film Thickness Equations
Rectangular Footprints
Solutions are available for each of the regimes: rigid-piezoviscous (Blok,

1952), elastic-isoviscous (Herrebrugh, 1968) and full EHL. All these results
can be expressed in terms of the dimensionless groups given above. However,
it can be shown that only three groups are really necessary, thus rendering
the results easy to plot. By choosing sensible groups, the relative effect of
viscosity increase and elastic deflection may be seen.
If
w
ge = O!
the film thickness may now be expressed by the equation
(7.9)
for all the regimes, the regime being identified from Figure 7.6. However, the
values of the constant Z and the exponents m and n are different for each
regime, as indicated in Table 7.1.
Table 7.1
Regime z m n
(1) 2.45 0 0
(2) 1.05 l. 0
3
(3) 2.45 0 0.8
(4) 1.654 0.54 0.06
Note the very low value of the exponent n in the full EHL solution.
This indicates the effect of elastic modulus, since g. is the only term containing
E'. This suggests that the film thickness is relatively insensitive to changes
of elastic modulus and it is found to be relatively insensitive to changes of
load. This is confirmed in practice, where an increase in load or decrease in
modulus will produce a resulting increase in the flattening of the surfaces
and a bigger effective load-carrying area. The above values are for the
minimum film thickness at the constricted outlet. The film thickness over
the essentially parallel region is given by the ratio
ho/hc = 0.72-0.81
The error incurred by assuming a ratio of 3/4 is negligible for most cases.
Elliptical Footprints
These may be approached in the same way. The general equation is

(7.1 0)
where
The 'constants' Z, m and n will be different for each regime, as will the
function f(k). The regime is identified from Figures 7.7-7.9. The values of
Z, m and n are shown in Table 7.2.
Equation (7.10) gives the value of the minimum film thickness, i.e. at
the outlet constriction. The ratio between this thickness and that in the
parallel region shows considerably more variation than the elliptical case.
204 TRIBOLOGY
Table 7.2
Regime z m n f(k)
(1) 128 0 0 rtf>{ 0.131 tan- (0 + 1.683 J1
(2) 1.66 2 0 [ 1 _ e- o.68k]

3
(3) 8.70 0 0.67 [1-0.85e- 0 · 31 k]
(4) 3.42 0.49 0.17 [ 1 _ e-o.6sk]
7.7 DESIGN AIDS FOR EHL FILM THICKNESS
A valuable manual for the calculation of EHL film thickness in line contact
is published as ESDU Item No. 85027 ( 1985). A similar publication is in
preparation for three-dimensional (point) contacts.
7.7.1 Examples of Film Thickness Calculation
( 1) A cylindrical roller in a roller bearing is 32 mm in diameter and 35 mm

long. The outer race has a concave radius of 103.5 mm and it is this race
that rotates at 100 rev /min. The load on the roller is 3 kN and it is
lubricated with an oil of viscosity 0.003 Nsjm 2 and pressure viscosity
coefficient 2.3 x 10- 7 m 2 jN. Estimate the minimum film thickness in the
contact.
First we need to calculate the equivalent cylinder radius, with
R 1 = 16mm and R 2 = -103.5 mm:
R = 16( -103.5) = 18.9 mm

16-103·5
The linear speed of the two surfaces is the same, assuming no slip, and so
U = U = 207 X 7r X 100 X 2X 7r = 6.81 mjs

60
Using the rigid, isoviscous solution and assuming that Rjh 0 is going to
be greater than 10 4 gives
h = 2.45( U 1 + U 2 )17RBjW
= 2.45 X 2 X 6.81 X 3 X 10- 3 X 18.9 X 10- 3 X 35 X 10- 3 /3000
h=0.022J.lm
This is, of course, a very small film thickness, much smaller than the
surface roughness of the roller, and this theory does not predict full film
lubrication.
Now we shall introduce elastohydrodynamic theory into the
problem.
Since the two surfaces are steel, E' = E/(1- v2 ), and assuming
E = 208 x 10 9 Njm 2 and v = 0.3 gives E' = 228 x 10 9 N/m 2 • Calculating
the operating parameters,
W= 3000/(228 X 10 9 X 18.9 X 10- 3 X 35 X 10- 3 ) = 1.99 X 10- 5
0 = 3000 X 6.81/(228 X 10 9 X 18.9 X 10- 3 ) = 4.74 X 10- 12
G= E' = 2.3 X 10- 7 X 228 X 10 9 = 524 X 10 2
The criteria for the effect of viscosity change and elasticity are as
follows
= (1.99 X 10- 5 )~524 X 10 2 = 2137
gv (4.74 X 10- 12 )!
g.=l.99x 10- 5 /(4.74x 10- 12 )!=9.14
Plotting these values on Figure 7.6 shows a point just in the rigid-
piezoviscous regime. Using the appropriate values from Table 7.1 in
Equation (7.9),
gh = 0.99(gv)f = 0.99(2137)f = 164.2
0 4.74 X 10- 12
11=-;;::-gh= 5 x 164.2=391 x 1o-
7
w 1.99 x 1o-
h0 = 391 X 10- 7 X 18.9 X 10- 3 = 0.739 ,Uffi
Note that the film thickness has been increased by a factor of more
than 30 by the inclusion of variable viscosity. This is a much more
realistic thickness.
(2) A standard deep-groove ball bearing has balls of 8.5 mm diameter. The
inner race rotates at 5000 rev /min. The groove in which the balls roll
on the inner race has a radius in the direction of motion of 10 mm and
a radius of curvature of 4.6 mm at right angles to the motion. The load
on the bearing is 300 N and it is lubricated with an oil of viscosity
0.6 Nsjm 2 and viscosity index 2.1 x 10- 8 m 2 jN. Estimate the film
thickness.
The load on the most heavily loaded element in a rolling bearing
is given as a pessimistic estimate by 5IN times the total load, where N
is the number ofrollers or balls (Stribeck, 1901). If in this case there are
12 balls,
W = 5(300/12) = 125 N
206 TRIBOLOGY
The radii of curvature fire

R = 4.25( 10/( 4.25 + 10) = 2.98 mm
R = 4.25( -4.6)/( 4.25- 4.6) = 55.8 mm
k = (55.8/2.98)! = 6.46
0= 0.6 X 5.235 = 4.623 X lQ-9

228 X 10 9 X 2.98 X 10- 3
W= 125 = 6.174 X 10- 5

228 X 10 9 X 2.98 2 X 10- 6
(; = 2.1 X 10- 8 X 228 X 10 9 = 4.788 X 10 3
This gives the criteria as
4.788 X 10 3 X 6.174 3 X 10- 15 7

gv = 2 18 = 5.27 X 10
4.623 X lO-
ge= 2.786 X 10 5
This is just in the piezoviscous, rigid regime on Figure 7.9, fork= 6.

Using Equation (7.10) and the appropriate values from Table 7.2,
gh = 1.66(5.27 X 10 7 )i( 1- e-O.OS x 6·46 ) = 1.66 X 1.406 X 10 5(0.9876)

= 2.305 X 10 5
{J =
gh
( 0)2 = 2.305
W
X 105(4.629
6.174
X lQ-9)2
X 10-S
= 1.292 X 10-3
h0 = 1.292 X 10- 3 X 2.98 X 10- 3 = 3.85 j..tm
7.7.2 Inlet Shear Heating
As the fluid is dragged into the inlet zone, the shearing which takes place
increases the temperature of the lubricant and, therefore, reduces its viscosity.
This results in a reduction of film thickness, the magnitude of which is
dependent on the parameter I, given by
where o is the temperature-viscosity coefficient and K is the thermal

conductivity of the lubricant. For rectangular contacts and a fully flooded
inlet, the film thickness as calculated from sub-section 7.5.4 must be multiplied
by a thermal reduction factor obtained from Figure 7.10.
Thermal
1.0~
reduction 10-1
factor
10-2 L
_ __J__ _L . __ _
_ J __ __.__ __J
0.1 10 102 10'
Inlet shear heating parameter I
Figure 7.10 Inlet shear heating effect
7.7.3 Starvation
All the results so far have assumed that the inlet to the contact is full of
lubricant. If this is not the case, the contact is said to be starved. This results
in a delay to the start of pressure build-up and a consequent reduction of
film thickness.
For rectangular footprints the reduced film thickness, h., can be
expressed approximately in terms of the fully flooded film thickness, h 00 , by
the formula
h. 2
=- arctan[l.37(J + 0.5) ]
2
-
h 00 n
where J = b1x;/(2Rh 0 )l and X; is the distance from the edge of the Hertzian
zone to the inlet boundary, which is normally considered to be the position
where the films of lubricant on the two surfaces come together; and b is the
Hertzian half-width (Wymer and Cameron, 1974).
For the elliptical case we first determine the dimensionless distance m;.
As the inlet position is moved progressively inward from infinity, it reaches
a position where it begins to affect the film thickness. This distance from the
centre, made dimensionless by dividing by b, the Hertzian half-width, is m;.
If h 00 is the minimum film thickness in the fully flooded case,
m; = 1 + 3.34[ R;~oo T.56

and we can then calculate the dimensionless minimum film thickness for the
starved condition from
!2_ = (~)0.25
hoo m;-1
where m is the actual dimensionless distance to the inlet boundary (Hamrock
and Dowson, 1981).
Clearly, in a case of a starved contact the effect of the shear heating in
the inlet is reduced. The two effects are, therefore, interdependent. A useful
review of the combined effect can be found in Gohar ( 1988).
208 TRIBOLOGY
Examples of Starved Contacts
( 1) Taking the roller described in Example ( 1) on p. 204, assume that

observation of the inlet meniscus indicates that the gap is full at a distance
of 5 mm from the centre line.
The Hertzian half-width is given by
b= [8WR]t
nBE'
8 X3000 X 18.9 X 10- 3 Jl
[
= 1t X 35 X 10- 3 X 228 X 10 3
b = 1.35 X 10- 4 m= 0.135 mm
Therefore, the distance of the inlet boundary from the edge of the Hertzian
zone is
X;= 5-0.135 = 4.865 mm
J= 0.1351 X 4.865 I=
19 _3
(2 X 18.9 X 1.23 X 10- 3 )3
h. 2 2
-=-arctan( 1.37 x 19.8 ) = 0.9988
hoo 1t
In other words, the film thickness is virtually unaffected by the delay in
pressure build-up.
Only if the contact suffers much more starvation than this will the
film thickness be diminished. For example, if the inlet meniscus were
0.5 mm from the centre line,
X;= 0.5- 0.135 = 0.365 mm
0.1351 X 0.365
j = I = 1.448
(2 X 18.9 X 1.23 X 10- 3 ) 3
h 2
-• =- arctan(1.37 x 1.948 2 ) = 0.88
hoo 1t
Even in this case only 12% of the film is lost.
(2) Using Example (2) on p. 205, let us suppose that there is a film
adhering to each surface of thickness 1 Jim as they enter the contact. If
we neglect the film thickness and the flattening at the contact, simple
geometry shows that the film will be complete at a distance x from the
centre line, given by x = (2Rh)1, where his the film thickness at the inlet
meniscus - in this case 2 x 10- 3 mm. This is, of course, only a guess at
where the pressure build-up begins.
X=(2 X 2.98 X 2 X 10- 3 )i=0.109mm
For an elliptical contact, the Hertzian half-width, b, is given with sufficient

accuracy by
b= [6ZWR"]
nkE'
where
z= 1.0003 + 0.5968
Ry/Rx
= 1.0003 + 0.5968 = 1.032

55.8/2.98
b = (6 X 1.032 X 125 X 2.98)! = 0.0793

1t X 6.46 X 228 X 10 3
The maximum distance for starvation to have a noticeable effect is,
therefore,
_ (2.98 X 1.94 X 10- 5) 0·56 _
m;- 1 + 3.34 0.0 7932 - 1.242
The actual dimensionless distance is

m = 0.109/0.0793 = 1.37
Since m > m;, no starvation effect will be seen.
If the film were reduced to 0. 75 p.m on each surface, the inlet meniscus
would move to
x=(2 X 2.98 X 1.5 X 10- 3 )l=0.0945
giving
m = 0.0945/0.0793 = 1.19
In this case m < m;, and so the reduced film thickness is
~= ( 1.19- 1 )0.25 = 0.94

h 00 1.242- 1
a reduction of 6%.
7.7.4 Effect of Film Thickness on Fatigue Life
In many rolling contacts, especially rolling element bearings, the primary

cause of failure is fatigue due to the repeated application and removal of
stress on the material as it passes through the contact. It is not immediately
obvious how the presence of a lubricant film helps to delay the fatigue failure,
as these stress cycles are still present. However, the improved life achieved
with a 'thick' lubricant film can be explained in several ways.
210 TRIBOLOGY
The Hertzian pressure profile is smoothed by the lubricant, so that the

application of the stress is not as sudden: see Figure 7.4. The spike can be
viewed as a cause of stress, but it seems unlikely that it is nearly as pronounced
in practice as in these theoretical profiles.
In all rolling contacts microslip takes place (Halling, 1976), resulting in
shear stresses at the surface. These combine with the contact stresses already
present, to move the maximum shear stress, previously below the surface,
closer to the surface, which is where cracks can be initiated. The lubricating
effect of the fluid will greatly reduce these stresses.
Fatigue cracks will often propagate from surface defects. A thick film
will prevent asperity contact and its associated microscopic stress fields close
to these defects.
The effectiveness of the film is quantified by the specific film thickness,
A, where A= h/ Ra and the mean roughness, Ra, is the square root of the sum
of the squares of the two surface roughnesses. Harris ( 1984) gives a value of
A~ 3 for a complete film for rolling bearings. As A is reduced below this
value, the fatigue life will be progressively reduced.
7.8 TRACTION
In EHL, as in all lubrication mechanisms, including dry rolling, surface

tractions are present. Even if pure rolling is taking place, energy is required
to compress the fluid as it enters the contact, the only possible source of this
energy being the motion of the surfaces. This will be seen as a retarding force
on the surfaces, which we will call the rolling friction, FR.
If the surfaces are moving at different speeds, the slower surface will
attempt to retard the faster one and vice versa, this constituting the sliding
friction, F8 • Hence, the retarding force on each surface may be written
for the slower surface, FR- Fs
for the faster surface, F R + F s
Typical traction curves are shown in Figure 7.11, from which several points
can be noted.
( 1) At constant rolling speed F R is virtually unaffected by load, but rises
with increasing rolling speed.
(2) For low sliding speeds Fs is proportional to sliding speed, indicating an
essentially Newtonian behaviour.
(3) At constant rolling speed F 8 increases with load. This is to a small extent
due to the reduction of film thickness, but to a much greater extent to
the increase of viscosity with pressure.
(4) At constant load F s decreases with increasing rolling speed. This is
because of the increase of film thickness.
~
2
• 1000 rev/min
45kN/m
21kN/m
9·6kN/m
2·8kN/m
0·5 1·0
Slop 2 cu,-u2 1
u,tu2
(a)
W = 44·8 kN/m
Rolling speed
(RS) • U,+U2
2
2500
0·5f-l---t---+---l------+----l
0 0·2 0·4 0·6 0·8 1·0

Shp 2 cu,-u2 1
U 1 +U2
(b)
Figure 7.11 Traction curves for rectangular contact: four-disc machine results
(Dowson and Whomes, 1967)
212 TRIBOLOGY
10-4
H=8oo·10- 6
10-5
----------
10-6
:::.
.gto:
10-7
"
10-B
10--9
0 4 8 12 16 20 24 28 32 36 40 44 48
O<W
Figure 7.12 Regimes of traction: oil, HVI 650; steam rollers, R = 20 mm (after
Evans and Johnson, 1986)
(5) There seems to be a limiting value of traction as sliding is increased.

This limit is due to the increase in temperature of the lubricant and
departure from Newtonian behaviour. This area is a complicated one
and performance will depend very much on the conditions and the
individual lubricant. There are four possible modes of behaviour for the
lubricant: linear viscous (Newtonian); non-linear viscous; non-linear
viscoelastic; elastic-plastic. To determine which of these regimes is
appropriate, it is necessary to plot a map for the lubricant as shown in
Figure 7.12, which is taken from Evans and Johnson (1986).
For a case where the fluid remains essentially Newtonian, the traction
can be calculated by considering just the parallel zone and using an effective
viscosity, given by
4K
ij= 2 [cxW +2bln(U 2 -Ud+bln("·(j)]
M(U 2 - Ud B 2K
where {J is the temperature coefficient of viscosity, K is the thermal
conductivity and 'Is is the viscosity of the lubricant at the temperature of the
surfaces at ambient pressure. This approximation is valid for
.:(U2-Ut)2 1
'lxu »
{JK
where 'lx is the viscosity at the surface temperature and at a pressure

appropriate to the value of x.
7.9 CAMS AND GEARS
The discussion so far has been concerned with the sort of contact encountered
in rolling element bearings, wheels, etc. The situation with cams and gears
can be somewhat different, with the inclusion of large velocities of approach
and other dynamic effects. Some of the film thickness formulae suggested for
these cases are given below.
For involute gears the contact at the pitch line may be modelled as two
rollers, having radii of curvature D 1 sin r/J/2 and D 2 sin r/J/2, with the rollers
rotating at the same angular velocities as the gearwheels themselves.
Holmberg (1982) gives the pitch-line film thickness, hP in microns, for
gears on parallel shafts as
hp = 1.7 X w- 3 ('7o Vp)i[t
where VP is the velocity at the pitch line in m/s, 'lois the viscosity in cP and
l is the distance between centres in mm. In this case a standard value has
been taken of the pressure- viscosity coefficient and the piezoviscous regime
has been assumed.
An earlier approximation is given in The Tribology Handbook as
hp=3.5~
where V. is the entraining velocity in m/ s, R is the relative radius of curvature
on the pitch line in metres and '1 0 is the viscosity in poise.
D 1 D 2 sin rjJ
R = ---------=--
2(D1 + D 2 ) cos 2 rjJ
and
V. = 0.4 V sin rjJ cos u
where rjJ is the normal pressure angle and u is the helix angle.
The minimum film thickness encountered during the contact is hP for
gears of approximately equal size, but reduces to around 50% of hP for large
gear ratios.
Experience suggests that, in order completely to avoid surface distress,
we need a specific film thickness, A., of more than 2. However, the situation
is so complicated by running-in and transient effects that results appear like
those shown in Figure 7.13, from Tallian ( 1967).
1.63- 6.5- Pitch

5.5 13 line
~ 80
~
t:
'C 60
0
g 40
:c
20
20
<t
0 0.4 0.8 1.2 1.6 2.0
Specific film thickness A
Figure 7.13 Probability of surface distress

214 TRIBOLOGY
7.10 PROBLEMS
1 Calculate the minimum film thickness for a steel ball of 30 mm diameter

rolling along a flat steel plate at a linear speed of 3 m/ s. It can be regarded
as fully flooded with oil of viscosity 80 cP and pressure viscosity coefficient
3 x 10- 8 m 2 /N. It is loaded only by its own weight. (Steel weighs
78 kN/m 2 .)
2 Repeat Problem 1 with the lubricant changed to water. Water has a
viscosity of 0.6 cP and the pressure viscosity coefficient may be taken as
zero.
3 A cylindrical rubber tyre of 800 mm diameter and 250 mm wide rolls
along a flat road at 50 km/h under a vertical load of 2 kN. Assuming
that a film of water is present on the road to a depth of 2 mm, estimate
the film thickness between tyre and road. For rubber E = 4 Njmm 2 and
Poisson's ratio= 0.5.
4 A steel strip 60 em wide is transported on cylindrical rollers of 10 em
diameter rotating at 600 rev /min. At any instant the strip is carried by
ten rollers, each carrying 100 N. If the contacts are fully lubricated by
an oil of viscosity 5 x 10- 2 Ns/m 2 and the strip moves at 50 cmjs, calculate
the minimum film thickness at the contacts, the traction force on the strip
and the energy consumption of the rollers.
5 Pushing the theory to its limit, calculate the film thickness in a journal
bearing, 50 mm long, having a shaft diameter of 65 mm and a diametral
clearance of 1 mm, lubricated with an oil of viscosity 90 cP and coefficient
8 x 10- 8 m 2 /N. The journal is steel and the bush is of plastic having
E = 200 Njmm 2 and Poisson's ratio= 0.33. The shaft rotates at
600 rev /min. Are there any reasons why the theory is not applicable to
this case?
6 A cylindrical roller of 30 mm diameter and 30 mm long rides with its axis
horizontal on a well-lubricated metal foil, which is inclined at an angle
of 1o to the horizontal. The foil is stretched between rollers which drive
it at a speed of 120 mm/min in the upward direction. The roller rotates
about its axis without slip relative to the foil, while otherwise remaining
fixed in space. Estimate the viscosity of the lubricant, assuming negligible
elastic deflection of the foil.
7 Two meshing involute spur gears have pitch circle radii of 60 mm and
120 mm, respectively, a pressure angle of 20° and a tooth width of 20 mm.
The speed of the larger wheel is 1500 rev /min and the power transmitted
is 40 kW. The gears are steel and the lubricant has a viscosity of
0.08 Ns/m 2 and a pressure coefficient of 2.4 x 10- 8 m 2 /N. If the c.l.a.
roughness of each surface is 0.4 J.lm, comment on the possibility of
satisfactory operation of the pair. Use each of the special formulae given
in Section 7.9 and also the general theory of sub-section 7.5.4.
8 A steel ball of 20 mm diameter rolls at a speed of 2 mjs in a straight steel
groove of 12 mm radius, carrying a load of 30 N. What is the minimum

thickness of the lubricant film on each surface, to avoid starvation effects?
If the inlet were fully flooded, would inlet shear heating be important?
The lubricant has a viscosity of 95 cP, a pressure-viscosity coefficient of
3 x 10- 8 m 2 jN and a temperature-viscosity coefficient of0.575;oc, and
the thermal conductivity of the fluid is 0.13 Wm/m 2 oc. Note that
although Figure 7.10 is derived for rectangular contacts, an indication
of the effect of inlet shear heating in elliptical contacts may be derived
from it.
9 In Section 7.8 two little formulae are given for the tractions on the two
surfaces. Explain why these do not contravene Newton's Third Law and
draw sketches showing all the forces acting on the two rollers.
Chapter 8
Bearing Selection
8.1 INTRODUCTION
In the preceding chapters we have examined many kinds of tribological

contact. Advice has been given on the theory and design of the types of
bearing available, but we have yet to establish the criteria used to decide
which solution to adopt for any tribological problem. In order to select the
appropriate bearing, we must first consider the options available.
8.2 BEARING TYPES
The possible main bearing types are listed below.

( 1)Plain rubbing bearing with a sensible choice of compatible materials.
(2) As ( 1), but including a solid film either attached to a surface or interposed.
( 3)Rolling elements inserted between the surfaces.
(4) A fluid film maintained between the surfaces by an external pressure
source (hydrostatic lubrication).
(5) A fluid film maintained between the surfaces by the motion of the contact
(hydrodynamic lubrication).
(6) A magnetic field to separate the surfaces.
(7) The use of a mechanism or flexible material in bearings of limited travel.
In this case the tribological problem has been designed out.
8.3 LIMITS OF TRIBOLOGICAL SOLUTIONS
The nature of the limitations encountered is shown in Figure 8.1

Strength considerations Considering the geometry of the contact and
the materials used, we can define the maximum load which may be safely
216
BEARING SELECTION 217
- logW
llogW
W =canst.
log V log 1
Strength limit lnert10 Ioree 11m1t lnstablhly 11m1t
(a) (b) (c)
Rollers W 1013 V =canst.
log W log W ~
~
log V log V log V
Thermalllm11 Wear llm1t FotiQUe hmtt
(d) (el (f)
log W log W
HydrostatiC, ,
------~ , ...,
,, , '
,,
, 'HydrodynamiC
log V log V
HydrodynamiC f1lm Hydrostatic and hydrodynamic
ltmlt limit
(g) (h)
Figure 8.1 Load and speed limits on performance
applied to the contact. Thus, for a cylindrical or spherical contact Hertzian

theory will define. the safe working load, while for other geometries even
simpler calculations will be appropriate.
Inertia limit Since tribological components are in motion, they are
subject to stresses arising from their inertia. This defines a maximum speed
above which failure will ensue. This could be the bursting of a shaft due to
centrifugal force at high rotational speed or the combined centrifugal and
gyroscopic forces in rolling element bearings, the bursting of the cage or
similar effects.
Frictional instabilities At low operating speeds some tribological
systems suffer from stick -slip instability.
Thermal effects The heat produced by friction and fluid shear will
cause a temperature rise in the components. There will be a maximum
permissible temperature defined somewhere in the system.
218 TRIBOLOGY
Wear limit A permitted wear rate will be defined for the system. As
shown in Chapter 4, the wear rate in many wear mechanisms such as adhesive
or abrasive wear is proportional to the product of load and velocity.
Fatigue limit In some mechanisms, notably rolling element bearings,
it is often fatigue failure which limits the life of the bearing. Practical tests
show that the relationship between load, W, and life, L, in revolutions for a
ball bearing is L oc 1I W3 and L oc 1I wlf for a roller. Since Lis clearly related
to speed, V, we can define the fatigue limit by W3 V =constant for ball
bearings and wlf V = constant for rollers.
Hydrodynamic limit When load is carried by the pressure generated
by hydrodynamic action, the film thickness is some function of '7 VI W, where
'1 is the viscosity. In such cases the operational limit is defined by the need
for a continuous film to be present to carry the load at a given speed using
a particular lubricant. This defines the limit as a line VI W =constant.
However, at higher sliding velocities the heat generated due to viscous
shearing results in a reduction of viscosity of the fluid. This causes a departure
from linearity of the limit at high velocity.
Elastohydrodynamic limit As we have seen in Chapter 7, in the
lubrication of very highly loaded contacts the dependence of film thickness
on load is very slight and so a nearly vertical characteristic is appropriate.
Hydrostatic limit The load capacity of a hydrostatic bearing is
determined entirely by the external pressure available. This leads to a limit
in the form of a horizontal line, with a slight reduction at higher speeds due
to the effect of shear heating on viscosity. However, it is almost inevitable
that, as the velocity increases, some hydrodynamic action will take place and
so the total load capacity will be the sum ofthe hydrodynamic and hydrostatic
effects.
8.4 SELECTION OF JOURNAL BEARINGS
In order to see how the above limits may be combined to produce the
characteristic limit curve for a bearing solution, we consider three journal
bearing types: the dry rubbing bearing based on PTFE compounds, the
rolling contact bearing and the hydrodynamic journal bearing. These
characteristics are shown in Figure 8.2.
If we plot these characteristics for different shaft sizes we obtain a chart
as shown in Figure 8.3. These curves cover the majority of the range of
engineering applications. From the plot we can see clearly the advantages
of rolling bearings at speeds in the range 1000-2000 rev /min, which explains
their use in such situations - e.g. small electric motors. With larger shafts
the enhanced load capacity of hydrodynamic bearings is clearly demonstrated.
This explains why these bearings are so often used in such applications as
In W In W
Max.
speed
limit
Max.
speed
limit
In V In V In V
(rev/m1n) (rev/min) (rev/min}
Dry rubbing bearing Rolling contact bearing Hydrodynamic bearing
Figure 8.2 Limits on three types of journal bearing
- - Hydrodynamic bearing
- - - Rolling bearing
- .. - Dry rubbing bearing
-g
0
...J
'-·---
d= 50 mm
. "'-·;--
·,·~
_..-->:( I . )'!
•,.--d= 5mm
Speed
Figure 8.3 Comparison of three bearing types
large steam turbines, where it is common practice to use a form of hydrostatic

jacking to preclude metallic contact during start-up and stopping.
A more complete chart for journal bearing selection based on a nominal
life of 10000 h, assuming a viscosity grade of 32-100 (BS 4231) and a
length/ diameter ratio of unity, except for the rolling bearings, is shown in
Figure 8.4. This is based on Neale (1967) which forms the basis of ESDU
Item No. 65007 ( 1965). Externally pressurised bearings will operate over the
entire range up to shaft-bursting speed. These give very high stiffness and
low friction torque, especially at start-up and low speeds, and have a load
capacity only limited by the size of the bearing and the pump pressure.
However, because of the complexity of design and cost of installation, they
are only used where these properties are essential. For example, externally
220 TRIBOLOGY
10 7 r---,-----,-~~--~--,-----,----,----,
"- d• 500 mm
105
10
0.01 0.1 10 100 1000 10000
Shalt speed (rev/s)
- - Rolling bearings
- · - · - Dry rubbing bearings
- - - - Hydrodynamic bearings
Figure 8.4 Journal bearing selection chart
pressurised bearings are found on massive telescopes, which need to rotate

with great precision at very low speeds.
For cases where even lower frictional forces are needed, gas lubrication
can be used. In this case the load capacity is very limited, but, because of
the low friction and consequent low heat generation, gas (usually air)
lubrication can be used for very-high-speed contacts. For example, air-
bearings are used in instruments, such as gyroscopes, in high-speed grinding
and drilling spindles and, of course, in the dentists' air-powered drill.
Example Find a suitable bearing for a 50 mm diameter shaft, rotating
at 1500revjmin and carrying a load of 10kN.
1500 rev /min= 25 rev js. Enter Figure 8.4. The point indicated is well
above the 50 mm line for rubbing bearings, and so these are not viable. It
lies just above the rolling bearing line. A rolling bearing is possible, but
nothing is in reserve on fatigue life. The point lies well below the line for
hydrodynamic bearings, indicating that such a bearing would be very suitable.
The final bearing selection will also involve practical considerations, such as
space available, lubricant supply, temperature and environmental effects, etc.
8.5 SELECTION OF THRUST BEARINGS
Similar considerations may be applied to the selection of thrust bearings.

The results are summarised in Figure 8.5, which is based on Neale (1967)
10 7
106
10 5
~
a:
0
104
..J
103
102
/
/
/ /
10
w-' w-' 10 10 2 104
Rotational speed (rev/s)
- - Rolling bearings
· - · - Dry rubbing bearings
- - - - Hydrodynamic bearings
Figure 8.5 Thrust bearing selection chart

which reappears in ESDU Item No. 67033 (1967). This chart is based on a
nominal life of 10000 hand, with the exception of the rolling element bearings,
assumes a 'typical' ratio between outside and inside diameter. This ratio will
depend on the particular application, but is unlikely to exceed 2. Theoretically,
higher load capacity can be obtained by increasing the outer diameter, but
tolerance to misalignment would be less and excessive heating could impair
the performance.
Example Select a thrust bearing type to carry a load of 10 kN at
1500 rev I min if the shaft size is 50 mm.
1500 rev /min= 25 rev js. Entering Figure 8.5 with 10 kN and 25 rev js
gives a point which is almost on the speed limit for rolling element bearings,
but is well under the interpolated line for hydrodynamic bearings.
8.6 SELECTION OF ROLLING ELEMENT BEARINGS
It is most unlikely that the reader will ever be called on to design a rolling
element bearing, but the selection of such a bearing is a common design
task. Each bearing manufacturer produces a guide to the selection and use
of his particular range of bearings. For more than a decade all the major
manufacturers world-wide have conformed to the recommendations of
ISO 281/1:1977 (BS 5512 Pt 1:1977). This has been reviewed recently to
reflect the widespread availability of cleaner steels, but the selection procedures
222 TRIBOLOGY
remain very similar for the major manufacturers. The material in the
discussion that follows is drawn from the RHP data published in the reference
given (RHP, 1977).
8.6.1 Selection of Bearing Type
The type of bearing to be used will depend on three major criteria: the radial
load, the axial load and the degree of axial location to be provided by the
bearing.
The loads on the bearing may be calculated by normal means, although
help is provided in the manufacturers' guides for certain common situations,
such as shafts with heavy rotors and ~ear trains of different types. The degree
of axial location is, of course, a matter for the designer to decide. The most
common bearing types are shown in Figure 8.6.
R ~ ~
!al (bl (C)
~ (d )
~ ~
!e) (f)
~
(g)
~~
(h) Iii
A (j)
~
(k)
~ (II
(m)
Figure 8.6 Rolling bearing types
(a) Single-row Radial Ball Bearings- Deep Groove
Radial load Good capacity.

Axial load Moderate capacity.
Axial location Both directions, with a little movement.

Comments Versatile, with a wide range of load, speed and mounting
conditions.
(b) Single-row Radial Ball Bearing -Filling-slot Type
Radial load Greater than the deep-groove equivalent.

Axial load Not for use with axial load only. Axial load should be less than
20% of radial in order to prevent damage to the balls by the edges of the
filling slots.
Comments In this bearing slots are cut in the raceways to allow more balls
to be inserted into the bearing.
(c) Double-row Radial Ball Bearings
Radial load Higher than the single-row bearing.

Axial load Not for axial load only. Axial load limited to 20% of radial load.
Comments More susceptible to misalignment than single-row.
(d) Double-row Self-aligning Ball Bearings
Radial load Moderate.

Axial load Low capacity, owing to the wedge action of balls in spherical
outer race.
Axial location Both directions, with some movement.
Comments Used where there is a need for misalignment allowance.
(e) Single-row Separable Ball Bearings- Magneto Bearings
Radialload Low.
Axial load Low.
Axial location One direction only.
Comments The bearing races are detachable and interchangeable. They are
often used in pairs, one bearing being adjusted against the other.
(f) Single-row Angular Contact Ball Bearings
Radial load Good.

Axial load Good.
Comments Better axial load capacity than deep-groove balls. Often used in
pairs, face-to-face or back-to-hack.
224 TRIBOLOGY
(g) Double-row Angular Contact Ball Bearings
Radial load Good.

Axial load Good in both directions, except in the type with filling slots.
(h) Single-row Cylindrical Roller Bearings
Radial load Better than the equivalent ball bearing.

Axial load None for basic types. Some bearings include specially designed
ribs to carry very small, intermittent axial force. These rely on effective
lubrication and mounting design.
Axial location None for basic types.
(i) Double-row Spherical Rollers
Radial load Very good.

Axial load Not for use with axial load only. Axial load is limited to 50%
of the radial load.
Comments For use where allowance for misalignment is desirable.
(j) Taper Roller Bearings
Radial load Good.

Axial load Good in one direction.
Axial location In one direction.
(k) Single-row Thrust Ball Bearings- Flat Race
Radial load None.

Axial load Light loads only.
Comments The flat races give low friction, but correspondingly low load
capacity.
(I) Single-row Thrust Ball Bearings- Grooved Race
Radial load None.

Axial load Good.
Comments The effect of centrifugal force limits the speed range.
(m) Needle Rollers- Thrust and Journal
Comments The rollers have a high length/diameter ratio. They are often
used in situations where there is not sufficient space for the equivalent ball
or roller bearing.
8.6.2 Selection of Bearing Size
Having decided on the bearing type, the size must be determined. For each
bearing the manufacturer will quote:
C - the dynamic load capacity of the bearing
C0 - the static load capacity of the bearing
Each of these must be checked against the operating conditions.
Dynamic Capacity
The dynamic failure of rolling bearings takes place through various fatigue
mechanisms. It is in the nature of fatigue that the life of samples shows a
large scatter and can only be expressed statistically. The 'life' of the bearing
is usually defined by the L 1 0 life, which is the number of millions of revolutions
that 90% of the bearings are expected to exceed before failure. If P is the
dynamic equivalent load, which is in most cases the applied load while
running (but check with the manufacturer's recommendations), the life is
r
given by
L 10 = ( ~ for ball bearings
L10 = (pc)-'f- for roller bearings

Or, for a constant speed of n rev /min,
16667(C)
L 1o = - n - p
3 .
hours for ball beanngs
16667(c)-'
L 10 = -n- p f- hours for roller beanngs
.
Modifications to this life should be made to get a more realistic life, L, by
L=a 1 a 2 a 3 L 10
Reliability factor a 1 There may be some applications where 90%
reliability is not good enough. In this case we introduce the factor a 1 , as given
226 TRIBOLOGY
in the table below.
Reliability (%) Life factor a 1
90 1.0
95 0.62
96 0.53
97 0.44
98 0.33
99 0.21
Material and lubrication factor a 2 , 3 This factor is used as the product

of a 2 and a 3 . It is used to include the improved materials (a 2 ) and the
lubricating and operating conditions (a 3 ). The procedure for calculating a 2 , 3
is given as follows:
( 1) Calculate the basic L 10 life.
(2) Obtain the recommended lubricant viscosity at 37.8 oc from Figure 8. 7.
n = speed (rev/s)
d = bearing bore+ o/d (mm)
m 2
104 ~-------------L--~-L-L----~-----U----~
10 50 100 500 1000
Viscosity at 37.8'C (eSt)
Figure 8. 7 Choice of lubricating oil for rolling bearing

(3) Determine the actual lubricant viscosity at 37.8°C.

(4) Divide the viscosity in (3) by that in (2), to give the viscosity ratio, V.-
(5) Use Figure 8.8 to obtain the viscosity ratio at the operating temperature,
~-For speeds given by (dmn) less than 10 or V.less than 1, use~= V.-
(6) Read off the factor a 2 •3 from Figure 8.9.
This procedure is valid for normal temperatures and lubricants and
assumes a viscosity index greater than 90. Above 150 oc there is a loss of
bearing hardness, which will adversely affect bearing life, and for non-
conventional lubricants, such as water-oil emulsions, further reductions in
life are to be expected.
10 10
8 8
~ ~
.J .J
v,
4
v,
4
+
3 3
3 4 5 6 8 10 3 4 5 6 8 10
V, v,
Figure 8.8 Viscosity ratio at operating temperature
3.5
3.0
V = actual lubricant viscosity {eSt)
1 recommended viscosity (CSt)
2.5 at the operating temperature
2.0
;::
~
1.5
1.0
0.5
0
0.1 0.5 1.0 5.0 10.0
v,
Figure 8.9 Material and lubrication life adjustment factor, a 2 •3
228 TRIBOLOGY
Static Loading
The manufacturers' tables give a static load capacity C0 . This is the load
which, when applied to the stationary bearing, produces a permanent
deflection of 0.0001 of the rolling element diameter. This is known as
brinelling. For satisfactory performance the maximum equivalent static load,
which is in most cases the actual value of the static load, must not exceed C 0 .
Other Considerations
The preceding data assume that the bearing is installed correctly, with the
appropriate tolerances, and is adequately lubricated. Instructions on fitting
and lubrication are to be found in the bearing manufacturers' catalogues
and technical literature, together with maintenance schedules. Additional
information is to be found in The Tribology Handbook (Neale, 1973), which
includes an interesting section on the identification of bearing failures. This
is also dealt with in Section 9.9.
8.6.3 Example
Determine the life to give 97% reliability for the deep-groove ball bearing
described below.
C = 11.2 kN; load, P = 1 kN

bore diameter= 25 mm; outside diameter= 47 mm
speed = 600 rev I min
lubricant viscosity at 37.8°C = 60 eSt
dm = (25 + 47)/2 = 36 mm
dmn = 36 X 600 = 216 X 10 2
basic life, L 10 = ( 11.2/1 )3 = 1.4 x 10 9 revolutions
The reliability factor for 97% = 0.44.
Enter Figure 8.7, which gives a recommended lubricant viscosity of

55 eSt at 37.8°C. The actual viscosity is 60 eSt, giving a viscosity ratio, V.,
of 60/55 = 1.091. Although, in general, the increased viscosity will enhance
the life, it will also generate more heat and, therefore, some of this
enhancement will be lost. Using Figure 9.8, we can calculate a revised viscosity
ratio, ~. Since in this case dmn < 10 5 , we take ~ = V. = 1.091. Using this
value in Figure 8.9 gives a 2 , 3 = 1.64.
The bearing life for 97% reliability is given by
L = 0.44 x 1.64 x 1.4 x 10 9 = 1.01 x 10 9 revolutions

8.7 PROBLEMS
1 Select a suitable type of journal bearing to carry a radial load of 100 kN

at 6000 rev /min, assuming a shaft diameter of 100 mm.
2 A journal bearing of shaft diameter 25 mm is to carry a load of 9 kN at
a speed of 30 rev /min. Decide on a suitable bearing type.
3 Select a thrust bearing type to carry 300 Nat 6000 rev /min, if the shaft
diameter is 10 mm.
4 A thrust bearing of 400 mm shaft diameter is. to carry a load of 500 N
at 30 rev /min. Choose a suitable bearing type.
5 A single-row cylindrical roller bearing having a dynamic capacity of
523 kN, a bore of 110 mm and an outside diameter of280 mm is lubricated
by an oil having a viscosity of 40 eSt at 37.8°C. Calculate the life for a
reliability of 96% if the speed of rotation is 1500 rev I min and the load
is 20 kN.
6 Determine the necessary value of C for a ball bearing carrying a radial
load of 20 kN on a shaft of diameter 50 mm rotating at 1000 rev /min,
lubricated with an oil of viscosity 50 eSt at 37.8°C, to give an L 10 life of
1000 h.
7 A single-row roller bearing having a bore of 160 mm, an outside diameter
of 340 mm and a value of C of 1.32 MN has been selected to carry a load
of 200 kN at 260 rev /min. The lubricant has a viscosity of 40 eSt at
37.8 oc. The required life is 65 · 10 7 revolutions. Comment on the choice
of bearing and lubricant.
Chapter 9
Lubricating Systems
9.1 INTRODUCTION
In the preceding chapters we have seen the beneficial effect of the presence
of the lubricant on the bearing performance. The lubricant has three main
functions: ( 1) to provide a coherent film between the surfaces; (2) to remove
heat from the contact; ( 3) to prevent the ingress of dirt and other contaminants.
The following sections deal with the volume and means of lubricant supply,
the monitoring of the system and the lubricant, and the likely results of poor
performance in this area.
9.2 FLOW REQUIREMENT
The necessary rate of lubricant supply to a hydrodynamic slider or journal

bearing can be determined by the method given in Section 5.3. The inlet to
the bearing must be constantly supplied with a sufficient quantity to generate
the film between the surfaces and to keep the temperature to within
satisfactory limits.
On the other hand, the volume of lubricant necessary to generate an
elastohydrodynamic film is extremely small. In almost every case a film of
fluid attaches itself to each surface and passes repeatedly through the contact,
the only additional supply required being that to make up the almost
negligible losses to these films. It follows, therefore, that any requirement for
a large volume will result from the need to remove large quantities of heat
from the vicinity of the bearing. The energy generated in the bearing is given
by P = FV for a linear bearing and P = M2nN for a rotating bearing.
The heat generation in sliding bearings has been described in Chapter 5.
For rolling element bearings the friction moment, M, may be calculated from
230
Table 9.1
fo
Oil-bath Oil bath
Bearing type Oil mist grease vert. shaft ft g!Po R
Deep-gve. ball 0.7-1 1.5-2 3-4 0.0009Z (2.3)F.-0.1F, 0.85-1

Self-al. ball 0.7-1 1.5-2 3-4 0.0003Z 1.4YF.-0.1F, 0.4-0.5
Ang.-cont. ball 1 2 4 0.0013Z F.-0.1F, 0.4-0.5
single-row
double-row 2 4 8 0.001Z 1.4F.-0.1F, 0.6
Cyl. and needle Rs. 3-6 6-12 12-24 0.00025-0.0003 F, (F.= 0) 0.5
single-row
double-row 6-10 12-20 24-40 0.00025-0.0003 F,(F.=O) 0.5
spherical Rs. 2-3 4-6 8-12 0.0004-0.0005 1.2YF. 0.3-0.4 r
taper Rs. 1.5-2 3-4 6-8 0.0004-0.0005 2YF. 0.3-0.4 c
Ill
thrust balls 0.7-1 1.5-2 3-4 0.0012Z 0.25 :0
F.
cyl. thrust 2 4 0.0018 F. 0.25 0
)>
needle thrust 2-3 4-6 0.0018 F. 0.25 :::!
3-4 6-8 z
spherical thrust 0.0005-0.0006 F.(F, < 0.55F.) 0.25 C)
(/)
Z = (P 0 / C 0 ). Lower values are for light series, higher for heavy series. -<
(/)
--1
m
s::
(/)
N
~
232 TRIBOLOGY
the manufacturers' literature (e.g. SKF, 1984 ): M = M 0 + M 1 , where

M 0 =/0 (601'fN)d! x 10- 7 and M 1 =f1 g 1 P 0 N·mm. M 0 is the moment
arising from hydrodynamic losses, while M 1 is produced by the elastic
deformation and partial localised slip between the contact surfaces. The value
of fo depends on the design and lubrication, and / 1 and g 1 depend on the
load and its direction. These values are given for selected cases in Table 9.1,
but a fuller list is available in SKF ( 1984 ).
The energy that we require to remove from the bearing, as calculated
above, will be dispersed by the usual heat transfer mechanisms of conduction,
convection and radiation. The relative importance of these will depend very
much on individual circumstances, but there will inevitably be some cases
where we have to rely on the lubricant to carry away most of the heat. The
heat thus removed can be calculated from P = yQcT, where T is the
temperature rise of the lubricant through the bearing and c is the specific
heat of the lubricant. For popular oils at 15°C, c can be taken as 1.82 kJ /kg· oc;
for water it is about 4 kJ /kg· oc.
Thus, the flow rate, Q, can be determined. Obviously a conservative
design would assume all heat removal by the lubricant.
Thermal problems are very important in bearings, especially in those
rolling bearings and gears where large lubricant flows are not possible. Any
increase of temperature has a detrimental effect on the viscosity, which, in
turn, reduces the film thickness and may lead to an increase of heat generation.
This situation may reach a stable condition or it may become disastrously
unstable. This latter condition is the basis for the concept of'ftash temperature'
as a criterion for failure (Blok, 1937).
Having decided on the lubricant and the necessary flow rate, we must
select a supply system from the main alternatives listed below.
9.3 METHODS OF LUBRICANT SUPPLY
This section is confined to the supply of grease and oil to rolling bearings.
These form the vast majority of lubricants used today. Which is used will
depend upon the circumstances. Grease is good for keeping out contamination,
operating in any attitude, and for low speeds, and is extensively used in
rolling element bearings, especially in sealed units where continuous lubricant
supply is not possible or is undesirable. Oil lubrication is suitable for heat
removal and for hydrodynamic bearings operating at higher speeds.
9.3.1 Grease Supply
There are three forms of grease supply: ( 1) bearings packed on assembly;

(2) grease nipple to each bearing for periodic replenishment; (3) grease piped
to a number of bearings either by hand or by automatic pump.
LUBRICATING SYSTEMS 233
"'tlc
~
c
8
c 106
0
·~ 10 3
107
~w
10"
~
"'
105
10 2 L----------------L--------~L---~
10' 105
dmn
(mm x rev/s)
Figure 9.1 Grease lubrication constant
For rolling element bearings the initial charge of grease should reach
the working surfaces, but the housing should not be overcharged as this can
cause excessive churning and high temperatures.
The life of the bearing may be limited by the degradation of the grease
unless it is replenished occasionally. The interval between regreasing is given
by RHP ( 1977), T = K 6 R h, where K 6 is the grease lubrication constant
derived from Figure 9.1 and R is the relubrication factor from Table 9.1.
9.3.2 Oil Supply
There are many methods of supplying oil to a bearing. The most common
are illustrated in Figure 9.2. Any of these methods can be found in use for
rolling bearings, but for plain journal bearings and sliders, oil mists, capillaries
and ring oilers are unlikely to provide sufficient flow rate.
Pumped Supply
In this case the oil is supplied to the bearing by a pump under pressure. The
flow rate can be as large as the pump can produce and therefore the heat
removal can be very effective. The capital cost, maintenance costs and initial
filling cost can be high and, of course, energy is required to operate the
pump. However, where a large, reliable flow is required, this system is most
suitable.
234 TRIBOLOGY
Pumped
8
Oil mist
i
g
Capillary
Gravity
Ring oiler
Dipping
Figure 9.2 Oil-supply systems
Pumped Oil Mist
Mist systems (aerosol systems) produce a mist of fine lubricant droplets

which travel at low velocity through the piping to a nozzle, where they are
speeded up to over 40 m/s so that they wet the surfaces. Clearly, this is of
use in enclosed spaces, such as gearboxes, where the mist will permeate the
entire space. The air is vented to atmosphere and the oil is usually collected
and recirculated. Since the thermal capacity of air is very low and the volume
of oil quite small, the heat removal capability is also small. A detailed design
guide to such systems is found in The Tribology Handbook. The initial cost
and maintenance costs are relatively high.
Gravity Supply
In cases where no great flow is necessary, gravity feeding, normally in the

form of droppers, can be sufficient. This can be made automatic by collecting
the oil and pumping it back to a reservoir. This system is simple and cheap,
but is limited to lightly loaded, low-speed bearings.
Capillary Systems
These systems use a wick or pad of felt in which the capillary action draws
the fluid from a reservoir on to the bearing surfaces. The oil flow increases
as the cross-sectional area of the wick, and decreases as the length of the
wick increases and with viscosity increase. The design and construction of
such systems is very simple and the costs small. However, the supply rate is
quite limited and is only sufficient for lightly loaded bearings.
Dipping Systems
If a sump is included in the assembly containing a horizontal rotating shaft,

it is possible to arrange for part of the rotating mechanism to dip into the
lubricant. This will carry oil up to a higher point, where it will drop onto
the bearing, splash the oil onto the bearing, create an oil mist, or a
combination of these actions.
It is possible for rolling element bearings to dip below the oil surface,
but it is recommended that immersion be only as far as the centre of the
lowest rolling element, in order to avoid excessive churning and consequent
temperature rise.
The ring-oiler consists of a ring of much larger diameter than that of
the shaft, which dips below the surface and as it rotates (more slowly than
the shaft, obviously), carries oil by viscous lift up to the shaft. These are used
for rotational speeds between 50 rev /min and 3500 rev /min and shaft sizes
up to 50 mm, since the ring speed is limited to about 9 m/s. The oil supplied
depends on the ring size and speed, but if more is required, it is possible to
use more than one ring or to use a chain, which has a bigger surface area.
In some cases the ring can be replaced by a disc, which is convenient
for low-speed bearings and high-viscosity oils.
If components dip intermittently into the oil, as in crank systems, these
can be used to splash oil onto other components or to create a mist. This
reduces the drag which continuously immersed components experience.
Clearly, all these methods are simple and cheap. However, it is necessary
to maintain the sump level within quite close limits.
9.4 FILTRATION
When supplying the lubricant to the bearing, it is usually necessary to

incorporate a filter in the system, to remove dirt, debris, contamination and
water. The filter( s) may be located in various parts of the circuit, as shown
in Figure 9.3 (which is based on The Tribology Handbook), according to
their main purposes.
The oil reservoir vent filter is to prevent the entry of air-borne
contaminants, while that on the filler is to exclude coarse solids. The pump
may have a medium filter on the suction side to protect the pump and a
finer one on the outlet to protect the bearings. A medium return line filter
may be included to stop wear debris from entering the reservoir. In systems
236 TRIBOLOGY
Figure 9.3 Filtration
where cleanliness is paramount a separate bulk cleaning operation, perhaps

by centrifuging the oil, may be used.
The size of the particles trapped by the filter will depend on its
construction. This varies from perforated metal strainers, which will trap
particles from 100 J.lm to 1000 J.lm, through woven fabrics and felts at
10-100 J.lm, down to ceramics and non-woven sheets at 0.5-50 J.lm. For
specialised applications membranes are used which will remove particles
down to 0.005 J.lm.
In addition to the normal filters, magnetic plugs can be inserted in the
lines to pick up appropriate material. These consume very little energy and
will collect very fine particles.
Water can be removed by using a hydrophobic element, such as
silicon-coated papers or Teflon-coated meshes, and allowing the water to
fall by gravity into a collection sump.
It is necessary to decide the degree of filtration necessary for any bearing.
The finer the filter the higher the penalty to be paid in pressure loss through
the filter or in flow rate if the pressure is preserved.
Most hydrodynamic bearings have minimum film thicknesses in the
region of 20 J.lm upwards. Clearly, if particles of this size and above are
precluded from entering the bearing, those smaller ones that remain should
pass through the bearing clearance without interfering with the performance.
With externally pressurised bearings the criterion will be not only the film
thickness, but also the restricter diameter.
In rolling element bearings the thickness of the lubricant film is frequently
much less than 1 J.lm. It is usually uneconomic to try to remove all the
particles down to this size. The effect of degree of filtration on performance
is not easy to see, but evidence suggests that filtration down to about 25 J.lm
is beneficial; further refinement to the order of 2-3 J.lm produces no
statistically significant improvement (Loewenthal and Moyer, 1979).
The pressure loss through a filter can be written (Wells, 1967)

~ - Q2KFy
p- 2gA 2
where KF is a characteristic of the filter, which is essentially constant in the

laminar region, falling off as the flow in the filter becomes turbulent at a
transition Reynolds number of 4-15, where the Reynolds number is given by
VD
Re=-
v
D being usually taken as the diameter of the filter inlet port. K F is a function
of the filter construction and is determined by the Kozeny -Carman relation:
e is the void fraction or porosity, Sis the specific surface of solid/unit volume
(m 2 jm 3 ) and k is a constant for the filter material (for cellulose k = 5.55).
The relationship between flow rate and pressure drop is shown for a
typical case in Figure 9.4, which represents the results from tests to BS 6277.
This graph is for the case where the filter is equipped with a bypass which
opens when the pressure drop through the filter reaches a predetermined
value. This could happen if the filter were blocked.
As the pressure drop increases with Q, if the flow rate is high, it is often
advantageous to use bypass filtration. In this case the filter is connected in
parallel with the bearing system and the pressure differentials are arranged
so that only a small fraction of the lubricant passes through the filter on
each circuit. This means that not all the fluid is cleaned on every pass, but
filtration is still quite efficient.
A more common arrangement is a combined full-flow /bypass filter,
where all the lubricant is filtered to some degree, but a small proportion still
goes through a bypass to remove finer contamination. Cartridges including
both types of filtration in parallel in one element are available.
Since most filters operate by trapping particles within the medium, there
will come a time when the trapped material will seriously interfere with the
performance. With fibrous filters, where filtration operates through the depth
of the element, this will be noticed as a reduction in the degree of filtration
and the possible release of previously collected material due to pressure
surges. For a surface filter, which operates by trapping the particles in
well-defined openings on the surface, trouble is indicated by a decrease in
flow rate or increase in pressure drop. In either case the filter must be serviced,
by either replacing the element or cleaning it in a solvent, by steam backwash
or ultrasound.
238 TRIBOLOGY
1.4
1.2
~ 1.0
e
Q_
'0
~ 0.8
~
6::
0.6
0.4
0.2
0 10 20 30 40 50 60 70 80
Flow rate (1/min)
Figure 9.4 Pressure drop plotted against flow rate for a filter with integral bypass
9.5 MACHINERY HEALTH MONITORING
The traditional method of keeping machinery in a good operating condition

has been through planned preventative maintenance - i.e. regular scheduled
servicing involving oil change or replenishment, replacement of filters,
readjustment of machine settings, etc. However, this type of procedure suffers
from two drawbacks: first, to ensure good general reliability in the intervals
between services, many of the procedures are carried out more often than is
necessary; second, if a serious fault should develop between services, it could
lead to disastrous consequences before the next scheduled service. Therefore,
there is an increasing movement to overcome both these drawbacks by
replacing planned preventative maintenance by maintenance on condition.
Under a regime of maintenance on condition the health of the machinery is
monitored, without shutdown, either continuously or at very frequent
intervals, and maintenance procedures are effected only when necessary. The
two most effective and widely used methods of machinery health monitoring
are oil analysis and vibration monitoring. These procedures will be described
in turn.
9.6 OIL ANALYSIS
Techniques now exist for monitoring the condition of any oil-lubricated

machine by investigating at regular intervals the amount and types of wear
debris present in the oil. Samples of oil can be easily taken and the techniques
do not require the machinery itself to be stripped down and examined. A
sudden change in the amount or type of debris between successive oil samples
may then indicate the need for immediate remedial action.
There are two types of analysis procedure in common use: spectroscopic
oil analysis procedure (SOAP) and ferrography. The methods are to some
extent complementary; each has its own advantages and disadvantages, which
are described briefly below.
9.6.1 Spectroscopic Oil Analysis Procedure (SOAP)
In this procedure a small sample of oil, typically 0.1 ml, is vaporised, and its
atomic constitution is determined both qualitatively and quantitatively by
spectroscopic analysis. Three types of analysis are in common use: differential
infrared spectroscopy; atomic emission spectroscopy; and atomic absorption
spectroscopy. All give essentially the same information.
SOAP gives the total concentrations of the elements in the sample but
it gives no information on whether the atoms were previously combined in
a chemical compound or an alloy. Furthermore, it does not give any
information on the sizes or shapes of th~ particles, and it detects particles
only in the size range 0-3 Jlm. Its major advantage lies in the detection and
measurement of wear debris produced by corrosive or oxidative wear, which
is normally of submicron size.
9.6.2 Ferrography
Ferrography (Scott et al., 1975) is a technique which allows metallic wear

debris in an oil sample to be magnetically separated and arranged according
to size on a transparent slide, so that it can be subsequently examined by
optical and scanning electron microscopy and chemically analysed by electron
probe microanalysis. The amount of debris and its particle size distribution
can be determined by optical density measurement, and a rapid increase in
the total quantity of debris or in the ratio between large and small particles
is indicative of a transition to a severe wear situation.
Ferrography detects particles in the size range of 1-20 J.lm, and may
therefore miss the very fine debris which often results from corrosive wear.
The shapes of the particles are good indicators of the types of wear that are
occurring. In particular, long chips of metallic debris indicate severe abrasive
cutting wear, which almost certainly necessitates machine shutdown and
component replacement.
240 TRIBOLOGY
The technique is under development to give on-line continuous monitoring

(Centers, 1983) and to relate the particle size distribution to the state of wear
(Roylance and Pocock, 1983; Roylance and Vaughan, 1983).
The information given by both SOAP and ferrography can be misleading
when samples are taken from filtered systems, as the filter may permanently
trap some components of the debris and thus preclude their detection and
analysis. Further work is required on this and other problems before the
techniques become completely reliable indicators of machine condition.
9.7 VIBRATION ANALYSIS
Many different types of problem can be revealed by changes in the vibration

pattern of a machine, including: unbalance of rotating parts; misalignments
of couplings and bearings; worn or damaged gears or bearings; looseness;
and rubbing between mating surfaces.
Vibration can be measured in terms of displacement, velocity or
acceleration. The two most common methods are monitoring of acceleration
and displacement, for which reliable and robust accelerometers and non-
contacting displacement transducers are readily available. Measurements are
normally made using one or other of these systems at each bearing housing
in three orthogonal directions, including the direction parallel to the axis of
the bearing. However, such transducers are normally only useful up to
frequencies of ~ 2 kHz and cannot be used to detect higher-frequency
vibrations which can occur in high-speed rolling element bearings or
gearboxes. In such cases shock pulse measurements must be made.
In general, there will be slow but progressive changes in either the
amplitudes or frequencies of vibration of various parts of a machine, and
such changes often do not indicate serious faults. The aim of vibration
monitoring is to detect either new frequencies, sudden changes in amplitude
at any of the existing frequencies or significant departures from previously
established trends. In any case, the frequencies at which changes are observed
are often indicative not only of the onset of a fault, but also of the component
in which the fault has developed. This often allows the necessary spares,
tools, etc., to be made available before the machinery is shut down, thus
minimising downtime.
The level of sophistication of vibration monitoring is usually dependent
on the probable level of cost of any unscheduled shutdown.
At the lowest level, it usually consists of periodic, perhaps weekly or
monthly, checking of the vibrations of each bearing housing with portable
transducers. It is essential that such measurements are made at the same
place each time, and this is usually accomplished either by marking permanent
reference points on the housing or by using permanently attached blocks
into which the portable transducer can be fitted reproducibly. In this type
of monitoring, the results are usually recorded and plotted manually, and
there is an interval of some hours, or even days, between the measurements
being taken and any changes being noted and acted upon.
At the second level of sophistication, transducers are permanently
mounted at critical points on machinery and the output of each transducer
is monitored continuously. This output is then used in two ways. First, alarm
levels are set on each output so that immediate warning will be given if any
potentially dangerous vibration develops; second, the output levels are
sampled at frequent intervals and trends identified by the data acquisition
and processing system.
At the most sophisticated level, vibration levels are both recorded and
trended continuously so that at any time the maintenance engineer has at
his disposal a fully up-to-date display of the machine condition.
9.8 FAILURE ANALYSIS
Inevitably there will be occasions when bearings fail. It would be wrong to

assume that this is always the fault of the bearing itself or the way it has
been used. Problems in other machine components can produce increases
of load, temperature, etc., which result in conditions for the bearing far more
arduous than those for which it was designed.
When analysing bearing failure, it is essential to collect every piece of
evidence available. This will be greatly helped by records from bearing
monitoring devices and oil analysis, as described above, but there is no doubt
that very often the most vital clues are given by examination of the failed
bearing itself. The most valuable aid to this is a collection of photographs
of bearing failures, with their causes. There are several such collections
available to which the reader is referred for cases where failure analysis causes
any difficulty (Neale, 1973; Harris, 1967; Nisbet and Mullet, 1978; SKF,
1974 ). In this section a brief summary will be given, to allow the more obvious
cases to be .identified.
9.8.1 Sliding Bearings
Dirt in the Bearing
Dirt can be introduced into the bearing either during assembly or with the
lubricant during running. This will be shown by scoring and scratching of
the surface in the direction of motion. Larger particles introduced during
assembly can cause localised overheating and distortion.
242 TRIBOLOGY
Inadequate Clearance in Journal Bearing, Insufficient Lubricant,

Excessive Load
Any of these will cause the film thickness to be too small for satisfactory
performance. Severe damage to the bearing may result, with surface melting
of the bearing material, especially if a low-melting-point alloy is used, and
smearing or wiping of the surface.
Inadequate Interference Fit, Flimsy Housing
This will permit small oscillatory movements between the surfaces of the
bearing or bearing and housing. This is a common cause of fatigue failure
and is characterised by the production and oxidation of fine debris, usually
reddish-brown in colour for ferrous metals, with welding or pick-up of
material at the joint between the housing and bearing shell.
Misalignment, Excessive Journal Deflection
This is shown by uneven wear. In journal bearings it will be on diagonally

opposite top and bottom halves.
Fatigue Failure
Fatigue damage is recognised by the pits or spalls created, where eventually

pieces fall out of the surface. It can be caused by excessive dynamic loading
or by cavitation.
Corrosion
Corrosion will differ in appearance, depending on the bearing material used.

The usual sources are water, fuel combustion products, additives in the
lubricant or oxidation of the lubricant.
9.8.2 Rolling Element Bearings
Since the predominant mechanism in rolling element failure is fatigue, the

life of the bearing is defined by the number of hours that 90% of the bearings
should survive (see sub-section 8.6.2 ). It follows, therefore, that one premature
bearing failure does not indicate a bad selection or application. Neither does
failure at or after the life predicted.
It is very rare for failure to result from a faulty bearing; it is almost
always from poor fitting, selection or lubrication.
Fatigue Failure
This is the normal, expected mode of failure. It is characterised by pits of

detached material. With a unidirectional load this will occur in the stationary
raceway at the point directly under the load, since this is where the material
suffers the highest number of high-stress applications. If this occurs repeatedly,
it is likely that the load is too high. This may arise from insufficient bearing
clearance to allow for thermal expansion or centrifugal effects.
Insufficient Interference
This causes fretting between the race and the housing, with the production
of the characteristic reddish-brown debris. It is very important to get right
the diametral interference of raceways, as axial clamping is an inadequate
substitute.
Misalignment, Excessive Shaft Deflection
This will cause the normal track of dulled material on the raceways to be
skewed. It does not necessarily lead to immediate failure, but can be damaging
if allowed to continue. In cylindrical roller journal bearings this can lead to
excessive loading on the flange, which has only very limited axial load
capacity.
Dirt in the Bearing
Hard particles will be rolled into the bearing surface to produce dents. Grit
will be crushed to powder, making the raceways become frosted in appeatance.
Softer dirt will produce marking of the tracks and may interfere with the
free running of the elements. This will produce noise and possible wear.
Damage During Fitting
Since raceways are an interference fit on the shaft and in the housings, there
is a great temptation to abuse them by overenthusiastic use of hammers.
This can dent the races or the rolling elements and in extreme cases fracture
the flanges. This damage is often visible after the bearing has run.
Inadequate Lubrication
This is shown by evidence of overheating, such as the parts of the bearing

being blackened and showing the temper colours. The hardness of the parts
will have reduced and in extreme cases the cage may melt. In grease-lubricated
244 TRIBOLOGY
bearings the cage pockets and rims may be worn and the grease will be hard
and dry.
Brine/ling
Dents may be present in the rolling tracks, which conform to the shape of
the rolling elements. This is symptomatic of a sudden overload being applied
to the bearing. This can happen during assembly, in which case the grinding
marks will remain visible or when the bearing is stationary (false brinelling),
which will obliterate the grinding marks.
Corrosion
This is usually the result of the presence of water. If a small percentage of

water is present in the lubricant, this can greatly reduce the fatigue life,
possibly due to hydrogen embrittlement of the surfaces. Great care must be
taken in the storage of bearings before use to exclude moisture.
9.8.3 Gears
Gear failure is not a common problem. Although damage to the surfaces

may take place, running will often continue quite satisfactorily. The marking
produced on the teeth by running should be examined periodically during
the early life to see if a pattern is established. This can be very informative
in giving clues to misalignment, excessive temperatures, etc.
Mechanical Failure
The most dramatic failure is when a tooth breaks off. The obvious caQse of
this is that the tooth load produces bending stresses at the tooth root in
excess of the tensile or fatigue strength of the material. For brittle materials
a sudden shock load may cause the failure. The tooth may also be weakened
by other forms of surface damage.
Surface Fatigue
This is caused by the load on the tooth being too high. Pits form in the
contact zone which may join together to affect large areas of the tooth.
Scuffing
Scuffing is a result of inadequate lubrication. The lubricant film is not

maintained, the temperature of the surface rises above the 'flash temperature'
(Blok, 1937) and the tooth surfaces become roughened and torn. This process
can lead to severe welding and complete destruction of the surfaces.
References
Archard, J. F. (1953). J. Appl. Phys., 24, 981

Archard, J. F. (1959). Wear, 2(6), 438
Archard, J. F. ( 1980 ). Wear Control Handbook, New York, ASME, 35
Archard, J. F. and Baglin, K. P. (1986). Proc. I. Mech. E., 200, 281
Arnell, R. D., Herod, A. P. and Teer, D. G. ( 1975). Wear, 31, 361
Arnell, R. D., Midgley, J. W. and Teer, D. (1968). Brit. J. Appl. Phys., 237, 1543
Ashby, M. F. and Jones, D. R. H. (1986). Engineering Materials, vol. 2, Oxford,
Pergamon Press
Barwell, F. T. (1979). Bearing Systems- Principles and Practice, Oxford; OUP
Barwell, F. A. (1983). Wear, 90(1), 164
Bell, T., Staines, A.M., Fairhurst, W. and Semchysen, M. (1983). Molybdenum Mosaic,
6(3), 14
Bitter, J. G. A. (1963). Wear, 6(5), 169
Blok, H. ( 1937). Second World Petroleum Congress, Paris
Blok, H. (1952). J. Inst. Petr., 38, 673
Borik, F. (1980) Wear Control Handbook, New York, ASME, p. 327
Bowden, F. P. and Tabor, D. The Friction and Lubrication of Metals, Pt. 1 (1950)
and Pt. 2 (1964); Oxford, 0 UP
Bragg, W. L. ( 1924 ). Introduction to Crystal Analysis, London, Bell
Briggs, G. A. D. and Briscoe, B. J. (1975). Wear, 35, 337-364
Buckley, D., see Miyoshi and Buckley ( 1985), p. 18
Burwell, J. T. (1957). Wear, 1, 119
Cameron, A. (1981). Basic Lubrication Theory, Chichester, John Wiley
Cameron, A. and Wood, W. L. (1949). Proc. I. Mech. E., 161, 59-72
Carslaw, H. S. and Jaeger, J. C. ( 1959). Conduction ofHeat in Solids, Oxford, Clarendon
Centers, P. W. (1983). Wear, 90,1-9
Challen, J. M. and Oxley, P. L. B. (1979). Wear, 53, 229
Challen, J. M. and Oxley, P. L. B. (1984). Int. J. Mech. Eng. Science, 26, 403
Childs, T. H. C. (1988). Proc. I. Mech. E., 202(C6), 379
Cole, J. A. and Hughes, C. J. (1956). Proc. I. Mech. E., 170, 499-510
Dearnaley, G. and Grant, W. A. ( 1985). In Recent Developments in Surface Coatings
and Modification Processes, London, Mechanical Engineering Publications, p. 43
Dowson, D. and Higginson, G. R. (1959). J. Mech. Eng. Science, 1(1), 6
Dowson, D. and Miranda, A. A. S. ( 1975). Conference on Cavitation and Related
Phenomena, University of Leeds/Mechanical Engineering Publications
245
246 REFERENCES
Dowson, D. and Taylor, C. M. ( 1975). Conference on Cavitation and Related

Phenomena, University of Leeds/Mechanical Engineering Publications
Dowson, D. and Whomes, T. L. (1967). Proc. I. Mech. E., 182, Pt. 1, No. 14
Dubois, G. B. and Ocvirk, F. W. (1953). NACA Report No. 1157, Washington DC
Dyson, A. ( 1975). Trib. Int., June, 117
Earles, S. W. and Powell, D. G. (1971). Wear, 18, 381
Edwards, C. M. and Halling, J. (1968). J. Mech. Eng. Science, 10, 101
El-Shafei, T. E. S., Arnell, R. D. and Halling, J. ( 1983 ). ASLE Trans., 26( 4 ), 481-487
Endo, F. and Fukuda, Y. (1965). Proc. 8th Japan Congress on Testing and Materials,
Tokyo, p. 51
Evans, C. R. and Johnson, K. L. (1986). Proc. I. Mech. E., 200, 313
ESDU ( 1965). General Guide to the Choice of Journal Bearing Type, Item No. 65007,
London
ESDU ( 1967). General Guide to the Choice of Thrust Bearing Type, Item No. 67033,
London
ESDU ( 1977). Low Viscosity Process Fluid Lubrication of Journal Bearings, Item
No. 77013, London, revised 1982
ESDU (1984). Calculation Methods for Steadily Loaded Axial Groove Hydrodynamic
Journal Bearings, Item No. 84031, London
ESDU ( 1985). Film Thickness in Lubricated Hertzian Contacts (EHL): Part I, Item
No. 85027, London
ESDU (1987). Guide to the Design and Material Selection for Dry Rubbing Bearings,
Item No. 87007, London
ESDU ( 1988). Selection ofAlloys for Hydrodynamic Bearings, Item No. 88018, London
Ferrante, J., Smith, J. R. and Rose, J. H., see Miyoshi and Buckley ( 1985), p. 138
Finnie, 1., Levy, A. and McFadden, D. H. (1979). In Adler, W. F. (ed.), Erosion:
Prevention and Useful Applications, Philadelphia, ASTM-STP-664, p. 36
Floberg, L. (1961). Trans. Chalmers Univ. of Tech., 234
Glaeser, see Miyoshi and Buckley ( 1985), p. 214
Godfrey, D. (1980). Wear Control Handbook, New York, ASME
Gohar, R. ( 1988). Elastohydrodynamics, Chichester, Ellis Horwood
Gohar, R. and Cameron, A. (1966). Proc. ASME/ASLE Conf, Minneapolis
Gough, V. E. ( 1954 ). Auto Eng., April
Green, A. P. (1955). Proc. Roy. Soc., 228A, 191
Greenwood, J. A. ( 1984 ). Proc. Roy. Soc., 393A, 133
Greenwood, J. A. and Halling, J. (1971). I. J. Mech. E. E., 1(2), 35
Greenwood, J. A. and Williamson, J. B. P. (1966). Proc. Roy. Soc., 295A, 300
Grubin, A. N. ( 1949). Central Scientific Research Institute of Technology and
Mechanical Engineering, Book No. 30, Moscow, DSIR Translation No. 337
Halling, J. (1975a). Principles of Tribology, London, Macmillan Education, p. 308
Halling, J. ( 1975b ). Wear, 34, 239
Halling, J. (1986). Proc. I. Mech. E., 200(C1), 31
Halling, J., Arnell, R. D. and Nuri, K. A. (1988). Proc. I. Mech. E., 202(C4), 269
Halling, J. and Nuri, K. A. ( 1974 ). JUT AM Symposium, University of Delft Press
Halling, J. and Nuri, K. A. (1985). Proc. I. Mech. E., 199(C2), 139
Hamilton, G. M. (1983). Proc. I. Mech. E., 197, 53
Hammitt, F. G. (1980). Wear Control Handbook, New York, ASME
Hamrock, B. J. and Brewe, D. E. (1983). J. Lub. Tech., 105, 171
Hamrock, B. J. and Dowson, D. (1981). Ball Bearing Lubrication, John Wiley
Harris, J. H. ( 1967). Lubrication of Rolling Bearings, London, Shell-Mex and BP
Harris, T. A. ( 1984 ). Rolling Bearing Analysis, Chichester, John Wiley
den Hartog, J.P. (1952). Strength of Materials, New York, McGraw-Hill
Hays, D. F. (1959). J. Basic Eng., March, 13-23
Heathcote, H. L. (1921). Mech. World, 70(1804), 79
REFERENCES 247
Herrebrugh, K. (1968). Trans. ASME Series F, 262

Hitchman, M. L. ( 1985). In Recent Developments in Surface Coatings and Modification
Processes, London, Mechanical Engineering Publications
Hokkirigawa, K. and Li, Z. Z. ( 1987). Proc. Conf Wear Mat., Houston, ASME, p. 585
Holmberg, K. ( 1982). Trib. Int., June, 123
Holmes, R. (1963). Proc. I. Mech. E., 177, 291-307
Honeycombe, R. W. K. (1968). The Plastic Deformation of Metals, Arnold
Honeycombe, R. W. K. (1980). Steels: Microstructure and Properties, Arnold
Hutchings, I. ( 1979 ). Erosion: Prevention and Useful Applications, ASTM-STP-664, p. 59
Jaeger, J. C. ( 1942). Proc. Roy. Soc. NSW, 56(203 ), 378
Jahanmir, S., Suh, N. P. and Abrahamson, E. R. ( 1974 ). Wear, 28, 235
Jakobsson, B. and Flo berg, L. ( 1957). Trans. Chalmers Univ. of Tech., No. 3
Jakobsson, B. and Flo berg, L. ( 1958). Trans. Chalmers Univ. of Tech., No. 5
Johnson, K. L. (1968). J. Mech. Phys. Solids, 16, 395
Johnson, K. L. (1970). J. Mech. Eng. Science, 12(1), 9
Johnson, K. L. (1985). Contact Mechanics, Oxford, Clarendon
Kannel, J. W. (1965). Proc. I. Mech. E., 180(3B), 135
Kragelsky, I. V. (1965). Friction and Wear, London, Butterworths
Kruschov, M. M. ( 1957). Proc. Corif. Lub. Wear, London, I. Mech. E., p. 655
Lanes, R. F., Flack, R. D. and Lewis, D. W. (1982). Trans. ASLE, 25(3), 289-298
Lim, S.C. and Ashby, M. F. (1987). Acta Meta/1., 35, 1
Loewenthal, S. W. and Moyer, D. W. ( 1979). J. Lub. Tech., 101, April, 171
Malik, M., Chandra, M. and Sinhasan, R. (1982). Trans. ASLE, 25(1), 133-140
Marsh, H. (1964). The Stability of Aerodynamic Gas Bearings, Mechanical Science
Monograph No. 2, London, Institute of Mechanical Engineers, pp. 1-44
Martin, F. A. ( 1970). Proc. Inst. Mech. Eng., 184, Pt. 3L, 120-138
Martin, F. A. ( 1983a). Trib. Int., 16(2), 64-65
Martin, F. A. (1983b). Trib. Int., 16(3), 147-164
Martin, H. M. ( 1916). Engineering, 102, 199
Massey, B.S. ( 1989). Mechanisms ofFluids, London, Van Nostrand Reinhold, Chapt. 6
Miyoshi, K. and Buckley, D. H. (1985). In Loomis, W. R. (ed.), New Directions
in Lubrication, Materials, Wear and Surface Interactions, New Jersey, Noyes
Publications, p. 282
Moes, H. and Bosma, R. (1971). Trans. ASME, J. Lub. Tech., 93, 302-306
Morris, J. A. (1967). In Lubrication and Lubricants, Lausanne, Elsevier, p. 310
Neale, M. J. (1967). Proc. I. Mech. E., 182(3A), 547
Neale, M. J. (ed.) (1973). The Tribology Handbook, London, Newnes-Butterworths
Nisbet, T. S. and Mullet, G. W. (1978). Rolling Bearings in Service, London,
Hutchinson Benham
OECD ( 1968). Glossary of Terms and Definitions in the Field of Friction, Lubrication
and Wear, Paris
Peterson, H. B. and Winer, W. 0. (eds) (1980). The Wear Control Handbook, New
York, ASME
Petrusevich, A. I. (1951). Izd. Akad. Nauk SSSR, OTN 2, 209
Purday, H. F. P. (1949). Streamline Flow, London, Constable
Quinn, T. J. F. ( 1971 ). Wear, 18, 413
Rabinowicz, E. ( 1970). In Friction and Lubrication in Deformation Processing, New
York, ASME, pp. 90-102
Ramalingan, S. (1980). In Wear Control Handbook, New York, ASME
RHP ( 1977). Standard Bearings Technical Catalogue, Chelmsford
Roelands, C. J. A. ( 1966 ). Druk. V RB Groningen
Rowe, C. N. (1980). In The Wear Control Handbook, New York, ASME, p. 143
Roylance, B. J. and Pocock, G. (1983). Wear, 90, 113
Roylance, B. J. and Vaughan, D. A. (1983). Wear, 90, 137
248 REFERENCES
Sasada, T., Norose, S. and Mishina, H. (1979). Proc. Int. Corif. Wear Materials,
Dearborn, MI, ASME, p. 72
Savage, R. H. ( 1948 ). J. Applied Physics, 19, 1
Schmitt, G. F. (1980). In The Wear Control Handbook, New York, ASME
Scott, D., Seifert, W. W. and Westcott, V. C. (1975). Wear, 34, 251
Sherbiney, M. A. and Halling, J. ( 1976). Wear, 40, 325
SKF (1974). Rolling Bearing Damage: A Morphological Atlas, King of Prussia, Pa.,
The Centre
SKF ( 1984 ). Principles of Bearing Selection and Application, King of Prussia, Pa.,
The Centre
Soda, N., Kimura, Y. and Tanaka, A. (1977). Wear, 43(2), 165
Stribeck, R. (1901), Z. ver. D. I., 45(3), 73-125
Suh, N. P. (1973). Wear, 25, 111
Suh, N. P. (1977). Wear, 44, 16
Tabor, D. (1985). In Loomis, W. R. (ed.), New Directions in Lubrication, Materials,
Wear and Surface Interactions, New Jersey, Noyes Publications, p. 2
Tallian, T. E. (1967). Trans. ASLE, 10, 418
Tavernelli, J. F. and Coffin, L. F. (1959). 1rans. Am. Soc. Metals, 51, 438
Teer, D. G. and Arnell, R. D. ( 1985). In Recent Developments in Surface Coating and
Modification Processes, London, Mechanical Engineering Publications, p. 21
Walls, J. M. (ed.) (1989). Methods of Surface Science, Cambridge, CUP
Waterhouse, R. B. ( 1988 ). Fretting Corrosion, Oxford, Pergamon Press
Weber, C. and Sallfeld, K. (1954). Z. angew. Math. Mech., 34, Nos. 1-2, 54
Wells, R. M. (1967). Proc. I. Mech. E., 182, Pt. 3A, 443
Wilson, R. W. (1975). Proc. Eurotrib. 1, J. Mech. Eng., 165
Woodruff, D.P. and Delchar, T. A. (1988). Modern Techniques of Surface Science,
Cambridge, CUP
Woolacot, R. G. (1965). National Engineering Laboratory Report No. 194, East
Kilbride, NEL
Woolacot, R. G. and Macrae, D. (1967). National Engineering Laboratory Report
No. 315, East Kilbride, NEL
Wymer, D. G. and Cameron, A. (1974). Proc. I. Mech. E., 188, 221
Zum Gahr, K. H. (1987). In Microstructure and Wear of Materials, Amsterdam,
Elsevier, Chapt. 5
Index
Abrasion Ball bearings

three-body 70 deep groove 222
two-body 70 double-row radial 223
Abrasive wear 68, 70-72 filling-slot type 223
Accelerometers 240 single-row, magneto 223
Additives, extreme pressure single-row angular
54 contact 223
Adhesion 32, 86 Bearings
and friction junctions centre of pressure 142
32 coefficient of friction
in static contact 32 of 138
thermodynamic work of effect of dirt in 241
32, 50-51 effect of temperature
Adhesion theory 34 rise of oil on
Adhesion-enhanced wear 86, 138-142
120 effective viscosity and
Adhesive wear 68-69 temperature 140
laws of 69-70 externally pressurised
Aircraft braking 117 219
Alloys hydrodynamic journal 161
multiphase 98 hydrostatic 124
single phase 98 life of 225
Amonton's laws 27-29 plain rubbing 216
basis of 28-30 rolling element 242-244
Anisotropy 97 self-acting 134
Asperities 3 sliding, heat generation
heights of, statistical in 230
distribution of 3 square, characteristics
Attitude angle 166 of 149-150
Autocorrelation function, thermal problems in 232
standardised r(1) 9 thrust, practical designs
Autocorrelation function of 154-155
R(1) 8-10 Bearing alloys 100-104
249
250 INDEX
boundary lubrication 100 Cobalt-based superalloys

required properties of 99
100<-102 Coefficient of dynamic
Bearing failures 241 viscosity 125
Bearing selection 216-229 Coefficient of friction
elastohydrodynamic limit 14, 27-28, 29
218 of a bearing 138
fatigue limit 218 prediction of 29
hydrodynamic limit 218 Composites 109-117
hydrostatic limit 218 Conformability 101
influence of thermal Contacts
effects on 217 between similar spheres
journal 218-220 15
in relation to size 225 Hertzian 11, 18, 20, 26
rolling element 220, non-conforming, see
221-228 Non-conforming
strength considerations contacts
216 of rough surfaces 21-23
thrust 220-221 Convergent films 163
Bearing size, selection of Convergent-divergent
225 geometry 164
Bearing types 216 Converging gap 134
Boundary lubrication Corrosion 242, 244
52-54, 100 Corrosive or oxidative wear
Brakes 115-117 239
Braking Corrosive wear 68, 86, 239
aircraft 117 CPH, see Close-packed
railway 117 hexagonal structures
Brinelling 228, 244 Creep 56
Critical speeds 189
Crystal structure, effect
Cams and gears 212-213 of 97-98
Capillary systems 234 Cutting flows 79
Cavitation 68, 166 Cylinder
Cavitation erosion 91 equivalent 194
Centipoise 125 parabolic 195
Centistoke 126 rigid, theory of 195-
Centre of pressure of 196
bearing 142 Cylindrical roller
Ceramics bearings, single-row
adhesion and friction of 224
52
tribological properties
of 117-119 Deep-groove ball bearings
Cermets 117-119 205
Chip formation 80 Deformation theories 42-46
Churning 235 Delamination theory of wear
Circumferential symmetry 68
190 Delamination wear 77, 86
Close-packed hexagonal Design aids, for EHL film
structures 97 thickness 204-210 ,
Clutches 115-117 Dimensionless coefficients
Cobalt 97 of hydrodynamic journal
INDEX 251
bearings 168, 171 Energy dissipation at

of hydrodynamic thrust asperity junctions 46
bearings 134-135 Equival&nt cylinder
Dimensionless groups in EHL 194-195
200-202 Erosion 68, 91-93
Dipping systems 235 cavitation 91
Dirt 243 fluid 91-92
Displacement transducers impact 91
240 solid 92-93
Distribution functions 7-8
standard deviation of 8
and ~A value 8 Failure analysis 241
and RMS 8 False brinelling 244
Double-row angular contact Fatigue
ball bearings 224 rolling contact 72-73
Double-row radial ball sliding contact 73-77
bearings 223 Fatigue failure 242-43
Double-row spherical Fatigue life, effect of
rollers 224 film thickness on 209
Ductility 96 Fatigue strength 102
Dynamic load capacity 225 Fatigue wear 72-77
Dynamic viscosity 125 Ferrography 239-240
centipoise 125 Fibrous filters 237
coefficient of 125 Fillers 110-111
shear stress 125 Film shape, and pressure
velocity gradient 125 distribution 198-199
Film thickness, effect on
fatigue life 209-210
Eccentricity ratio 163 Film thickness equations
Effective viscosity 212 202-204
Effective viscosity and Filters
temperature 140 bypass 237
EHL, see Elastohydrodynamic fibrous 237
lubrication pressure loss in 237
Elastic deformation surface 237
equation 197 Filtration 235-238
Elastic-isoviscous regime Fine debris 242
200 Fitting 228
Elastic-plastic regime 212 Flash temperature 62, 232
Elastohydrodynamic Flow requirements 230-232
lubrication 194-215 Fluid erosion 91
Elastohydrodynamic Fluid film bearings 124
lubrication film hydrodynamic 124
thickness, design aids self-acting 124
for 204-210 Footprints
Elastohydrodynamic definition of 11
lubrication solution, formulae for 16
full 200 Free energy
Elastomers 50-51 interfacial 32
Elliptical footprints 199, surface 32
203-204 Free rolling 55
Ellipticity parameter 200 between identical elastic
Embeddability 101 bodies 56
252 INDEX
Fretting 90 bearings 124

Friction Hydrodynamic journal
adhesion theory of 34-41 bearings 161
coefficient of 27-28 Hydrodynamic pressures 163
deformation theories of Hydrodynamic thrust
42-46 bearings 130-134
junctions 32 pressure distribution
laws of 27 132
measurement of 29 Hydrogen embrittlement 244
of non-metallic materials Hydrostatic bearings 124
50-52 Hydrostatic jacking 219
origins of 31-34
ploughing 34 Impact erosion 91
rolling 210 Inadequate clearance 242
in rolling contact 54-62 Inadequate interference fit
sliding 210 242
Friction force 27 Inadequate lubrication 243
kinetic 27 Inclined-plane test 29
static 27 Induction-heating 121
Friction materials 115 Inertia limits 217
Friction moment, M 230 Inlet boundary 207
Frictional instabilities Inlet shear heating
217 206-207
Full EHL solution 200 Insufficient interference
243
Insufficient lubricant 242
Gas lubrication 220 Interfacial free energy 32
Gaussian distribution Interfacial shear strength
curves 9 parameter 79
Gaussian distributions 23 Involute gears 213
near 23 Iteration 184
Gaussian surfaces 25
Gear failure 244 Journal bearings
Gears complete 165
and cams 212 eccentricity in 161
causes of failure in 244 fluid lubricated 161-193
involute 213 hydrodynamic, effects of
Graphite 105 temperature rise on
Gravity supply 234 181-183
Grease lubrication constant hydrodynamic action in
233 163-165
Grease supply 232-233 instability in 189-191
Gyroscopic forces 217 line of centres of 161
practical design of 191
selection of 218
Half-speed whirl 189 Journal deflection,
Heathcote slip 58 excessive 242
Height distribution, of Junction growth 37
surface texture 7-8
Hertzian contacts 11, 18,
20, 26 Kinematic viscosity
Hertzian impact 92 126-127
Hydrodynamic fluid film centistoke 126
INDEX 253
L 10 bearing life 225 Normal approach, of a

Lamellar solids 105-107 sphere to a plane 17
Laminar flow theory Notation viii-xii
128-130
Liquid lubricants, limit-
ations of 104-105 Oil, temperature rise of
Load, excessive 242 138
Load-carrying capacity 102 Oil analysis 239-240
Lobes 190 Oil supply 233-235
Local elastic deformation Oil-supply grooves 177-179
194 Oil-whip 189
Lubricant supply 232-235 Oxidative wear 86, 239
Lubricants, temperature
rise of 232
Lubricating systems Parabolic cylinders 195
230-228 Particles, size of 236
Lubrication, elastohydro- Plane-inclined slider
dynamic 194-215 bearings 130
Reynolds's Equation 130
Plastic deformation, onset
Machinery health moni taring of 12-14
238 Plasticity index 24, 118
Machining wear 81 Ploughing 34, 42-43
Magnetic fields 216 Poiseuille velocity 129
Magnetic plugs 236 Polymers 51-52, 109-117
Maintenance on condition pull-off work of 51
238 Pressure distribution 132
Martensi tic transformations and film shape 198
99 Pressure spike 198
Material and lubrication Pressure-induced velocity
factor a 226 129
Mechanical f~ifure 244 Preventative maintenance
Metallic surfaces 238
clean 2 Probability distribution 7
nature of 2-3 Profilometers 4-6
Microslip 14, 56, 210 Pull-off force 33, 49, 86
Misalignment 239, 243 Pull-off work 33, 49
Molybdenum disulphide 105 Pumped oil mist 234
Multiphase alloys 98 Pumped supply 233
Mutual solubility, of PV factor 111-114
rubbing pairs 99-100 flat bearings 112
journal bearings 112
limit of 114
Needle rollers 225
Non-conforming contacts
11-17, 196 Radial clearance 163, 167
and coefficient of Railway braking 117
friction 14 Rectangular footprints
and microslipping 14 202-203
surface deformations in Regimes, in EHL 200
17-19 Reinforcements 110-111
Non-linear viscoelasticity Reliability factor a 1 225
212 Reynolds's equation
254 INDEX
in one dimension 132 Sliding bearings, heat

plane-inclined slider generation in 230
bearings 130 Sliding friction 210
two-dimensional form of thermal effects in 62-64
147-154 Solid erosion 92
Rigid-isoviscous regimes Solid lubricants 104-109
200 Solid materials, tribolog-
Rigid-piezoviscous regimes ical properties of 96-
200 113
Ring-oiler 235 Solution hardening 98
Road vehicles 116 Specific film thickness
brakes of 116 210, 213
Rolling 55 Specific wear rate 114
Rolling contact fatigue 73 Spectroscopic oil analysis
Rolling element bearings, procedure 239
selection of 221 Sphere rolling, in a
Rolling friction in EHL grooved track 57-58
210 Spherical rollers, double-
Rolling resistance 58-60 row 224
Roughness 3 Spin 55
tradi tiona! parameters of Standardised variables 23
6 Starvation, in EHL 207-209
root mean square (RMS) Static load capacity, of
value 6 rolling bearings
roughness average (RA) 225-228
value 6 Static loading 228
Structures, close-packed
hexagonal 97
Schallamach waves 50 Supply grooves 166
Scuffing 94, 244 Surface contact 10-26
Self-acting bearings 134 and surface properties
Self-acting fluid film 1-26
bearings 124 Surface distress, in gears
Shaft deflection, excessive 213
243 Surface fatigue 68, 244
Shakedown 60 Surface films, effects of
Shear force 138 19-20
Shear stress 125 Surface filters 237
Shear velocity 129 Surface free energy 32
Shot peening 121 Surface geometry 3
Single-phase alloys 98 Surface properties, and
Single-row angular contact surface contact 1-26
ball bearings 223 Surface slope 79
Single-row cylindrical Surface texture
roller bearings 224 analysis of 4
Single-row separable ball height distribution of
bearings, magneto 223 7-8
Single-row thrust ball and convertion into a
bearings probability
flat race 224 distribution 7
grooved race 224 profilometry 4-6
Skidding 60 statistical analysis of
Sliding 54 6-10
INDEX 255
Synchronous whirl 189 Viscosity, increase of,

with pressure 194
Taper roller ~arings 224 Water-oil emulsions 227
Temperature rise, effects Wave-forming flows 79
of, on oil 100 Waviness 3
Temperature rise, of 138 Wear 66-95
Thermal effects, in sliding abrasive 68
friction 62 adhesion-enhanced 120
Thermodynamic work of adhesive 68
adhesion 50-51 corrosive 68, 86-89, 239
Three-body abrasion 70 delamination 77, 86
Thrust ball bearings delamination theory of
single-row 68
flat race 224 of dissimilar materials
grooved race 224 84-86
Thrust bearings effect of surface films
fluid lubricated 124- on 93
160 fatigue 84
selection of 220 form of laws of 83
Traction 210-212 of lubricated surfaces
Tractive rolling 55 94-95
of non-conforming bodies mechanical processes of
56-57 68-77
Transformation hardening minor types of 90-91
98 under nominally elastic
Transition load 88 contact conditions
Tribology, definition of 1 82
Two-body abrasion 70 oxidative 239
Tyre-road contacts 60-62 reduction of, using
surface treatments
119-123
van der Waals bonds 50 Wear coefficient 70
Velocity Wear limit 218
Poiseuille 129 Wear maps 89-90
pressure-induced 129 Wear transitions, mild-to-
shear 129 severe 88
Velocity gradient 125 Work hardening 96
Vibration analysis 240
Viscoelastic hysteresis
losses 50 Yield stress (ey) 10

1991 Book Tribology PDF

Uploaded by

Copyright:

Available Formats

1991 Book Tribology PDF

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

1991 Book Tribology PDF

Uploaded by

Copyright:

Available Formats

Tribology : Principles and

All rights reserved. No reproduction, copy or transmission

No paragraph of this publication may be reproduced, copied

Any person who does any unauthorised act in relation to

First edition 1991

British Library Cataloguing in Publication Data

1. Surface Properties and Surface Contact 1

4. Tribological Properties of Solid Materials 96

5. Fluid-lubricated Thrust Bearings 124

6. Fluid-lubricated Journal Bearings 161

7. The Lubrication of Highly Loaded Contacts 194

8. Bearing Selection 216

9. Lubricating Systems 230

We have also included in the book introductions to more practical

Frad Radial force

J Inlet factor ( = bxJ(2Rh 0 ) 2 ' 3 )

k. Ellipticity parameter ( = a/b)

R Ratio between plastic and elastic area

Normal approach at limit of elastic deformation

Plasticity index ( =! (~Y 12

Tribology is defined as 'the science and technology of interacting surfaces in

1.2 NATURE OF METALLIC SURFACES

A surface of a typical metal might appear as shown in Figure 1.1. On top of

Figure 1.1 Typical surface layers

1.3 SURFACE GEOMETRY

The geometric shapes of ordinary surfaces are controlled by the characteristics

1.4 SURFACE TEXTURE ANALYSIS

A typical surface might have ~ 10 5 peaks. The problems of measuring the

1.4.1 Surface Profilometry

The requirement for measuring a statistically representative sample of the

a surface profile, as shown in Figure 1.2(b ). In examining such profiles, some

L x200 ' , , __ .,.."' 1 Lx20

Finally, in profilometry the profile height must be measured with respect

1.4.2 Traditional Roughness Parameters

1.4.3 Statistical Analysis of Surface Profiles

Figure 1.4 Various surfaces with the same RA value

The Height Distribution of the Surface Texture

As shown in Figure 1.5, the profile is sampled as height readings z 1 , z 2 , etc.,

Figure 1.5 Surface height readings taken at discrete intervals

Figure 1.6 Method of deriving the all-ordinate distribution function

Many statistical parameters are available to describe the shape of this

u = RMS = [ -1 n (zY L Jt ( 1.1 )

Periodic Variations along the Surface

For a continuous profile, Equation ( 1.3) can be rewritten as

r(l) = R~) (1.5)

" ... ''"l '·

0.2 L ~JIN.. n.JI,... ~ 1\ r'""'\t Superfinished surface,

This is by no means universally true, as full distributions often depart

1.5 SURFACE CONTACT

1.5.1 Stresses in Non-conforming Contacts between Curved

We consider first the contact of two identical cylinders under conditions of

Figure 1.9 Pressure distribution due to contact of two cylinders

For two identical elastic cylinders in contact under a normal load, P,

For simplicity, we have concerned ourselves so far with two-dimensional

where the value of a is given by

Figure 1.13 Pressure distribution due to contact of spheres

Table 1.1 Elastic contact stress formulae

Suffixes 1, 2 refer to the two bodies in contact.

Line contact Circular contact