Wear 225–229 Ž1999.

1–17

Wear modelling: analytical, computational and mapping: a continuum

mechanics approach
)
J.A. Williams
Cambridge UniÕersity Engineering Department, Trumpington Street, Cambridge, CB2 1PZ, UK

Abstract

As a result of its widespread and pervasive economic consequences wear, the almost inevitable companion of friction, has been the
subject of much scientific and empirical investigation; the literature is rich with ‘wear equations’ so that, on occasion, the practitioner can
seem to have too much information rather than too little. Some guidance and reconciliation between analytical and computational models
and empirical observations can be provided by plotting wear maps for specific materials so that the dominant wear mechanism for
particular operating conditions can be established and some indication provided on probable wear performance. A challenge facing the
research community is the production of a sound theoretical framework to underpin such design aids. However, because of the variety and
complexity of the surface conditions it is not straightforward to relate tribological performance to more easily established material
parameters. These difficulties are illustrated by looking at the models available for two particular classes of wear involving metallic
materials—severe abrasive wear, when surface life is likely to be short, and lubricated mild wear when very much longer component
histories can be anticipated. q 1999 Published by Elsevier Science S.A. All rights reserved.

Keywords: Wear; Modelling; Mapping; Abrasion; Severe wear; Mild wear

1. Introduction display very different forms of behaviour from those ob-

served under less arduous testing conditions; in particular,
Wear, the progressive damage and material loss which because of the intense local compressive stress fields
occurs on the surface of a component as a result of its materials which are usually classified as brittle Žsuch as
motion relative to the adjacent working parts, has far ceramics. can show significant plastic deformation while
reaching economic consequences which involve not only those that are ductile can show greatly enhanced strains
the costs of replacement but also the expenses involved in prior to failure. The practical difficulty of simulating the
machine downtime and lost production. As a result, con- physical environment Žprincipally this high containing
siderable efforts have been expended on the development stress which can be of the order of several GPa. within the
of theories and deterministic models; for example, Meng zone immediately under a point or line contact under more
and Ludema w1x have identified nearly 200 ‘wear equa- carefully controlled test conditions has meant that there is
tions’ involving an enormous spectrum of material proper- a dearth of good material data which can be used with
ties and operating conditions. However, despite the best confidence in some of the mechanical models referred to
efforts of their authors, there is still no way of predicting, above.
with confidence or certainty, the tribological performance No simple and universal model is applicable to all
of a loaded pair of surfaces, whether dry or lubricated, situations. In the dry, unlubricated or perhaps marginally
even if all of their physical and chemical properties have lubricated sliding of two, usually dissimilar, loaded sur-
been independently established. faces, so-called two-body conditions, the rate of surface
The response of a material to mechanical loading can be degradation or damage of each depends Žat least. on the
broadly classified as being either ductile or brittle. How- factors of Table 1.
ever, under the peculiar conditions generated under in- When a third body is present at the interface wear may
tensely loaded point or line contacts, a material may be inhibited, though not entirely eliminated Žfor example if
the third body is a lubricant or low shear strength film with
a thickness dimension at least comparable with the mean
)
Tel.: q44-01223332641; fax: q44-01223332662 surface roughness. or enhanced as in the case of contami-

0043-1648r99r$ - see front matter q 1999 Published by Elsevier Science S.A. All rights reserved.
PII: S 0 0 4 3 - 1 6 4 8 Ž 9 9 . 0 0 0 6 0 - 5
2 J.A. Williamsr Wear 225–229 (1999) 1–17

Table 1 Several points emerge from this elementary table. First,

Factors influencing dry wear rates the numerical values of K are always lower than unity,
. Normal load usually very much lower. Secondly, in the tests of Table 2
. Relative sliding speed
. all the values of the coefficients of friction lay in a
Geometry Žboth macroscopic and local or topographic.
. Initial temperature comparatively narrow range between 0.18 and 0.8 whereas
. Local environment, and the range of wear coefficients was very much greater,
. The thermal, mechanical and chemical varying over 4 orders of magnitude. There is no simple
properties of the materials involved correlation between friction and wear: although in a quali-
tative way it is reasonable to expect situations in which
there are higher frictional forces to be also those involving
nation by entrained dirt Žor even just the retained debris relatively high wear, it is quite possible for material com-
from previous wear events.. Contamination by debris both binations to produce very similar frictional forces but very
harder and softer than the opposing solid surfaces is likely different wear behaviour.
to be detrimental to the life of the contact though may Whatever the nature of the particular materials in-
involve different detailed mechanisms: the distribution of volved, the simplest classification of unlubricated surface
sizes, shapes and mechanical properties of the third bodies interactions is into those involving either mild or severe
are all influential variables. wear. This is not based on any particular numerical value
of wear rate but rather on the observation that increasing
1.1. The Archard wear equation the severity of the loading Že.g., by increasing the normal
load, sliding speed or bulk temperature. for any pair of
A common starting point in the analysis of wear is often materials leads at some stage to a comparatively sudden
the Archard Žor Rabinowicz. wear equation w2,3x which jump in the wear rate. The commonly observable differ-
asserts that the wear volume w is directly proportional to ences between the two regimes of mild and severe wear
the product of the load P on the contact and the sliding are summarised in Table 3.
distance s but inversely proportional to the surface hard- Such jumps or nonlinearities in wear behaviour or in
ness H of the wearing material, so that material response have significant practical consequences
for engineering designers. From a practical engineering
Ps
wsK= . Ž 1. point of view, mild wear might well be considered accept-
H able whereas the transition to severe conditions often
The dimensionless constant K is the non-dimensional represents a change to commercially unacceptable values.
wear coefficient. When interpreting experimental situa- The engineer would thus like to know first for a given
tions, the hardness of the uppermost layer of material in material pair and particular set of operating conditions,
the contact may not be known with any certainty and, how close to this step change in outcome is the contact
consequently, a rather more useful quantity than the value operating; secondly, supposing that there is no danger of
of K alone is the ratio KrH; this is known as the an abrupt change into catastrophic conditions is it possible
dimensional wear coefficient or the specific wear rate and to estimate the rate of wear in terms of the operating
is usually quoted in units of mm3 Ny1 my1 . For a material conditions and readily measured material parameters.
with hardness of 1 GPa, a non-dimensional wear rate of
unity is equivalent to a measured wear rate of 1 mm3 Ny1 1.2. Wear maps
my1 . Some values of K taken from relatively early wear
literature, but still characteristic of such unlubricated sys- To provide guidance to the first of these questions, the
tems, are shown in Table 2. operating point of the design can be located in a map

Table 2
Table 3
Typical values of dimensionless wear coefficients for various materials
Distinction between mild and severe wear
against tool steel sliding in dry, unlubricated pin-on-disc tests in air: from
Ref. w4x Mild wear Severe wear
Material K Results in extremely Results in rough,
y3 smooth surfaces—often deeply torn surfaces
Mild steel Žon mild steel. 7=10
smoother than the original much rougher than
a r b brass 6=10y4
the original surfaces
PTFE 2.5=10y5
Debris extremly Large metallic wear
Copper–beryllium 3.7=10y5
small typically less debris up to 0.01 mm
Hard tool steel 1.3=10y4
than 100 nm diameter diameter
Ferritic stainless steel 1.7=10y5
High electrical Low contact resistance,
Polythene 1.3=10y7
contact resistance true metallic
PMMA 7=10y6
junctions formed
 J.A. Williamsr Wear 225–229 (1999) 1–17 3

whose coordinates are chosen from the list in Table 1 and the way indicated in which changes in service conditions
which is split into various territories each associated with might be expected to influence wear response, by plotting
some dominant mechanism of degradation w5x. Wear rates changes of this sort on the appropriate wear map. The
in successfully operating industrial equipment, as opposed simplest imposed variables are clearly load and speed and
to imposed experimental conditions Žsuch as those of using these as the ordinate and abscissa gives rise to the
Table 2. can vary greatly from high values under particu- now relatively familiar Lim and Ashby maps w6x. The
larly aggressive conditions to very low values in more challenge, as these authors themselves pointed out, is to
benign circumstances; for example the designed ‘wear extend such plots from those based on empirical observa-
allowance’ in continuously running mechanical seals, tional data and to construct such maps either from theory
which at best run under conditions of marginal lubrication, alone or, more profitably, from theory calibrated against
often permits them to lose material from their operating experiment.
surfaces at a rate of no more than 1 mm per annum, The Lim and Ashby form of map plots the severity of
equivalent to a non-dimensional rate of the order of 10y9 , the load expressed as the applied normal pressure p Ži.e.,
well below any of the values of Table 2. A consequence of the normal load divided by the nominal area of the contact
wear rates of these very small magnitudes is that in order A nom . vs. the sliding speed V. Such a map, which is
to achieve dimensionally reliable measurable amounts of specific to a particular material Žsurface topography, grain
wear in laboratory tests these must be run for several size, hardness etc.. has the merit that it also incorporates
hundred hours. Reducing the time scale of wear tests to regimes of different chemical or surface reaction behaviour
speed up the production of data, for example by increasing which are associated with temperature effects—although
loads or speeds, is a potentially hazardous procedure as this is at the expense of detail on the mechanical forms of
there is always the danger of moving from a regime of wear associated with different states of surface topography
operation within which one specific form of wear mecha- and asperity interaction. The general form of such a map
nism is dominant to another controlled by different physi- for a metal is illustrated by that for medium-low carbon
cal phenomena. Such pitfalls can often be prevented, and steels shown in Fig. 1. Pressure has been expressed non-di-

Fig. 1. A load–speed mechanism wear map for medium carbon steel based largely on pin-and-disc data. Load and speed both normalised as described in
the text. Thick lines delineate different wear mechanisms and thin lines are contours of equal wear rates. Chain lines represent constant values of the pV
factor.
4 J.A. Williamsr Wear 225–229 (1999) 1–17

mensionally as p˜ s prH where H is the indentation during the initial stages of component history a mild and
hardness of the softer surface, so that p˜ is always less than acceptable wear regime is established.
1. Regions of the map associated with different wear
mechanisms are traversed by contours of equal wear rate, 1.3. Mechanical wear maps
i.e., values of K. These can be established by extrapolation
from test data available in the literature. An often quoted maxim in tribology is the warning that
Characteristically, the map is divided into two regions it makes no sense to speak of the friction of a single
by a roughly vertical field boundary. To the left of this material without specifying both the nature of the counter-
division, wear is controlled by essentially mechanical pro- face as well as the operating conditions of the overall
cesses; here the wear rate depends on the normal pressure contact. At first sight, the Ashby map appears to conflict
Žor load. but is not greatly dependent on sliding velocity so with this admonition as there is no mention of the material
that contours of wear rate are more or less horizontal. On of the counterface. Features of this, the second body, have
the other hand, to the right of this division, thermal and differing levels of significance depending on the imposed
chemical effects Žinvariably under normal atmospheric conditions. For example, when the sliding speed is large,
conditions these involve oxidation. become the dominant so will be the energy input, and the conductivity of both
influence and the contours of wear rate become functions first and second bodies will be influential in controlling the
of both load and velocity. For steel, this fundamental conduction paths of the thermal energy generated at the
mechanism transition corresponds to dry sliding speeds of interface; counterface thermal conductivity thus has a sig-
about 0.1 m sy1 ; below this, surface heating, and so nificant effect on interface temperatures. At lower sliding
oxidation, is relatively insignificant. speeds, within the mechanical range of surface interac-
In the regime dominated by mechanical wear, there are tions, conductivity will be less important but now counter-
regions within which the spacing between contours of face topography and interfacial shear stress will play more
wear rate is equal to that between changes of load of the significant roles.
same factor, for example, as K goes from 10y9 to 10y8 or The conventional categories of mechanical wear pro-
from 10y5 to 10y4 , the load also changes by a factor of cesses are summarised in Table 4. A surface exposed to
10. In each of these cases wear rate and load are directly hard particulate contamination can deteriorate by both
proportional to each other, i.e., in accord with Eq. Ž1.. But erosion Žwhen the particulate have a significant velocity
this is not the case throughout the region; the gap between normal to the surface., and abrasion and the latter can
the contours for K equal to 10y8 to that for 10y5 , either be of the three-body form Žif the particles roll and
covering three orders of magnitude corresponds to a load tumble through the constriction., or two-body if they be-
change of only one order. This very rapid increase in wear come embedded in the opposing surface, and when the
rate is an example of the transition from ‘mild’, and hence situation will be more akin to the wear by a hard rough
usually acceptable levels, to ‘severe’ Žusually unaccept- counterface. Under these circumstances, an alternative map,
able. values brought about by an increase in load. Dry or of a form suggested by Childs w7x, which gives more
marginally lubricated contacts in machines between steel emphasis to the mechanical aspects of surface damage but
surfaces often operate at nominal pressures of a few less to the thermal Žsince velocity is not considered as an
megapascals and at sliding speeds of the order of some- independent variable. is more appropriate: such a map is
thing less than a metre per second; in other words just at sketched in Fig. 2. This shows the regimes of wear on axes
this level of wear uncertainty. Note that at a given level of representing the shear strength of the interface t between
load but at high sliding speeds, surface oxidation can be the two materials plotted linearly Žand usually normalised
protective and provide reductions in loss of material, i.e., by k the shear strength of the weaker material. vs. the
dips in the contours of K, when the softened oxide film logarithm of some roughness parameter such as the angle u
acts as a protective almost lubricant layer over the metallic representing the average slope of asperities on the harder
substrate. abrading surface. Rough counterfaces, i.e., those with high
An important design guide to the consequences of dry values of surface slope, lead to severe wear of the softer
rubbing is the energy input per unit area, i.e., the product surface by abrasion; this always involves severe plastic
of pressure and sliding speed, the ‘ pV ’ value. Since the deformation and can take the form of a combination of
map is drawn on log-log scales constant pV values will ploughing Žin which, although the surface topography is
plot as straight lines with a negative slope and three such much modified, only a small proportion of the displaced
lines are shown in Fig. 1 for specific energies of 0.1, 1 and material is actually detached from the surface. and micro-
10 MW my2 which are typical of dry running machine machining or cutting Žwhen a much higher proportion of
contacts. Very often the machine designer will have very the plastically deforming material is lost as wear debris..
little control over the nominal contact pressure and the Processes of abrasive machining—grinding, polishing, lap-
sliding speed for a particular application. He or she will be ping etc.—can be thought of as being analogous to abra-
obliged by making recommendations about materials and sive wear. Asperity slopes of grinding wheels and abrasive
secondary operating conditions to try and ensure that papers are likely to range from a few degrees to 108 or
 J.A. Williamsr Wear 225–229 (1999) 1–17 5

Table 4
The classification of mechanical wear processes

more and so generate wear by a combination of ploughing situation in most tribological devices, depends on some
and cutting mechanisms. form of fatigue or damage accumulation mechanism. Wear
Once significant volumes are lost by micro-machining, rates are much lower, ‘mild’ rather than ‘severe’, and their
reducing the interfacial shear stress has the effect of prediction, in terms of material and operational parameters,
increasing the volumetric loss because the efficiency of the presents a formidable challenge to those working in the
cutting operation is improved; a situation in which reduc- field.
ing friction enhances wear. This is in contrast to the Observation by scanning electron microscopy and fer-
circumstances over the rest of the map in which reducing rography have made it apparent that metallic wear parti-
the surface shear stress lowers the wear rate. This switch in cles, other than those generated under relatively severe
the effect of local frictional reductions can be used as the abrasive conditions or pitting fatigue, commonly take the
defining feature of truly abrasive wear. When the abrading form of thin platelets. In addition, metallographic sections
surface is much less rough, so that the value of u is small, of wearing surfaces often reveal near-surface cracks lying
elastic deformations cannot be neglected—indeed they parallel, or very nearly parallel, to the free surface. This
may be sufficient to accommodate the applied loads alone led Suh w8x to introduce the term ‘delamination wear’ and
—so that wear in repeated traversals, which is the usual to initiate their analysis by the techniques of fracture

Fig. 2. Wear mechanism map for a soft surface abraded by a harder, rough counterface. The ratio trk represents the relative strength of the interface and u
the mean slope of the rough surface Žor the attack angle of a single asperity.. Thicker lines delineate different wear mechanism and thin lines represent
contours of equal wear rate.
6 J.A. Williamsr Wear 225–229 (1999) 1–17

mechanics: a review of this work is available by Jahanmir

w9x. However, these cracks present mechanics with a prob-
lem in that under contact loadings they appear on planes
which carry very large compressive stresses w10x. To over-
come this difficulty it was suggested that they were, in
fact, mode II Žshear. cracks and several analyses were
carried out on this basis Žsee Ref. w11x.. While these
models have been of value in relation to pitting failure in
rolling contact fatigue, where the cracks penetrate rela-
tively deeply into the material, it became clear that in
shallower cases, the forces of friction at the highly com-
pressed faces of any such defect effectively suppress the
local mode II stress intensity factors. It now seems much
more likely that in many cases these near-surface failures
are of a ductile nature and result from an accumulation of
plastic strain very close to the component surface rather
than being driven by elastic stress intensity factors. This
view of the metallic wear process suggests that in situa-
tions of repeated sliding, modelling should concentrate on
plastic deformation of the near-surface material.

2. Modelling abrasive wear

2.1. Multiple contacts

The classical experimental work on abrasive wear and Fig. 3. The effect of hardness of metals on their abrasion resistance
on the abrasion resistance of metals Že.g., Kruschov w12x Žproportional to the reciprocal of wear rate.: from Kruschov w12x.
and Richardson w13x. involved experiments in which speci-
mens of each material, usually in the form of cylindrical
pins, were rubbed against a ‘standard’ abrasive—conven- lurgy is essential to optimise material performance. Experi-
tionally an abrasive paper or cloth carrying silicon carbide ments producing data of the sort displayed in Fig. 3 are
or quartz particles; the volume lost from the test samples examples of multiple wear events, i.e., situations in which
increased in proportion to both the sliding distance and the the resultant abraded surface is formed by the superposi-
normal load applied, as indicated in Fig. 3. This plot also tion of many interactions between individual abrasive
demonstrates that the relative wear resistance Žthat is the grains or asperities on the harder surface and the wearing
ratio of the of the wear volume of some standard material surface of the softer specimen.
to that of the sample material tested under the same Setting to one side the difficulties encountered with
experimental conditions. of a wide range of pure metals is two-or multi-phase materials, it has become conventional
proportional to hardness. However, the relationship be- to interpret such data in terms of an Archard wear equa-
comes more complicated for engineering materials which tion, i.e., Eq. Ž1., notwithstanding the fact that this was
consist of two or more phases: for example, steels of proposed initially to describe adhesive rather than abrasive
different alloy compositions which have been heat treated wear and that over the course of a complete test Žor the life
to give a range of hardness values, still give linear plots of a component. linearity between the worn volume and
but of different slopes from that of pure single phase load may not be maintained, generally being higher both
metals. Metallurgical investigations have suggested that as during the initial running-in Žor break out. phase and in the
a general rule, the abrasive wear resistance of steels in- final stages of component life preceding some form of
creases with their carbon content but the actual numerical catastrophic failure. Typical values of the constant K in
value is very dependent on micro-structural details. In two-body abrasion lie in the range 0.005 to 0.05; values in
pearlitic steels, for example, considerable improvements in three-body abrasion Žwhen the abrasive can roll and tum-
abrasion resistance can be achieved by refining the struc- ble between the sliding surfaces. tend to be lower perhaps
ture of the pearlite. Martensite, by virtue of its hardness, is down to 0.0005. Most research into abrasive wear Žsee for
very resistant to wear, but even in a martensitic steel the example the review by Zum Gahr w14x. has concentrated
volumes and precise compositions of the carbide phases on the two extremes of the process. One approach has been
can be extremely important in determining the value of the to experimentally study multiple asperity contacts of the
relative wear resistance, and careful control of the metal- sort described above. While these can provide much useful
 J.A. Williamsr Wear 225–229 (1999) 1–17 7

data on the relative wear resistance of groups of materials have larger values of the angle u and so will be expected
they can give comparatively little information about the to produce more wear damage. Although of the desired
actual deformation processes which take place as the worn form, the non-dimensional constant 2tan urp in Eq. Ž3.
surface is created: without this sort of detailed information greatly over-estimates the wear observed in real circum-
it is difficult to predict likely wear rates for new materials stances. A rough engineering surface might have a man
or coatings, or even existing alloys in unconventional value of u of say 58 Žat most. so that the predicted
circumstances. If the material is brittle then the situation is non-dimensional wear constant in this most severe of
more complex because material can be lost not only circumstances is equal to 6 = 10y2 , barely within the
through plastic flow but also by additional mechanisms of range experienced experimentally. One explanation for this
inter- or intra-granular fracture. There will be a sensitivity overestimate is the assumption that all the material from
to scale in a way not present in ductile metals and conse- the groove is lost from the surface: observations show that
quently on the geometry or conformity of the contact; the in general only some of the displaced material is actually
dependence of wear rate upon load is unlikely to accord detached Žin the form of a curly slivers or chips by a
with the simplicity of Eq. Ž1. w15,16x. micro-cutting process. while the remainder is simply piled
up at the edges of the resultant groove as ploughed ridges;
2.2. Single contacts this can reduce the wear rate very considerably; the other
factor is, of course that not all the asperities on the rougher
A starting point for most modelling of wear processes is surface in such a multiple encounter are as aggressive as
the consideration in more detail of the damage arising the single event which has been examined.
from a single isolated asperity moving across a softer If the abrading asperity is regarded as having the shape
previously undeformed surface: the literature contains ac- of a simple symmetrical pyramid rather than a cone Žboth
counts of a large number of investigations into the me- are probably equally reasonable approximations to the
chanical and materials aspects of the formation of single shape of real asperities. with one of its sloping edges
wear grooves, usually in carefully prepared and undam- leading, then the general appearance of the deformation is
aged metallic surfaces w17–26x. There is an obvious simi- as indicated in Fig. 4Žb. with a wedge or prow of deform-
larity between these circumstances and those of the scratch ing material extending ahead of the leading edge of the
hardness test w27x. The interaction between successive wear asperity. The relative volumes appearing in the ridges and
events or their statistical combination to generate observed the machined chips depends on the geometry of the asper-
final surface topographies has been far less well studied. ity which can be conveniently described by the two angles
shown in Fig. 4Žc.. The angle between the two leading
faces of the asperity 2 c , is known as the dihedral angle.
2.2.1. Three-dimensional models
For a given value of the dihedral angle, there is a signifi-
The simplest situation, illustrated in Fig. 4Ža., is that of
cant contribution to wear by machining only when the
a conical asperity with semi-angle Ž908 y u . carrying a
attack angle exceeds some critical value. If the angle of
normal load P ploughing a groove of depth h in a soft and
attack is less than this then the process is essentially one of
ductile surface. The deforming material is assumed to be
ploughing or rubbing. The dihedral angle can also influ-
homogeneous in composition and isotropic in its properties
ence the mode of material response: if 2 c is small then
and thus conform with the classical model of an engineer-
the asperity is rather like a knife moving through the
ing continuum. The angle u might represent the ‘average’
surface and deformation is principally that of ploughing; as
surface slope or roughness of the abrading surface which
c increases so the likelihood of significant wear increases.
carries an array of such asperities Žand so corresponds to
In the limit 2 c is equal to 1808 so that effectively the
the abscissa in Fig. 2.: it is often known as the attack
pyramid is moving with one of its flat faces forward rather
angle. The asperity now slides through a unit distance so
than one of its edges.
that the volume of displaced material or the effective wear
The effect that both these angles have on the form of
rate will be given by h 2 = cot u . The depth of penetration
the resultant deformation has been investigated on both the
h can be related to the indentation hardness of the material
comparative macro-scale and in a size regime more com-
H although, since it is only the front half of the asperity
parable to that of real asperity contacts by carrying out
that carries the load P, the supporting projected area is
small-scale tests within the chamber of a scanning electron
semi-circular and of radius hcot a , and so P s pr2 =
microscope; a summary of these and other observations
Ž hcot u . 2 = H. This means that
can be usefully displayed on the wear mode diagram
w 2tan u P suggested by Kato et al. w28x whose general form is shown
wear volume per unit sliding distance, s = in Fig. 4Žd.. When a square based pyramid or asperity
s p H
Ž 3. moves edge forward Ž2 c s 908. there is a fairly clear
transition from cutting or machining to ploughing or rub-
Thus, we have an ‘Archard’ equation with a wear bing; however, if the indenter moves face forward Ž2 c s
coefficient K equal to 2tan urp . Rougher surfaces will 1808. then, as the angle of attack increases, there is a
8 J.A. Williamsr Wear 225–229 (1999) 1–17

region of unstable prow or wedge formation before pure of plastic working. Strictly speaking, the application of this
ploughing is established. theorem requires that the surface configuration of the
Most of the attempts to model this apparently simple plastic zone be known ab initio. In the case in question this
situation have depended an application of the upper bound is not the case, as it is bounded by a stress free surface
method to estimate the rate of plastic working and so the whose shape is not controlled by the indenter. This means
associated force terms. The asperity being modelled as that solutions arrived at on the basis of energy minimisa-
either a pyramided or a cone. The upper bound method tion are not necessarily those that will in fact be favoured;
assumes that the material, which is characterised as a in the terminology of Avitzur and Talbert w29x this is a
rigid-perfectly plastic continuum, accommodates the ‘casual’ application of the limit theorem and the predic-
movement of the indenter by moving internally as a series tions are not so much upper bounds as ‘physically reason-
of rigid blocks so creating discontinuities in the tangential able’ solutions. This difficulty has been addressed for this
components of velocities at the interfaces between them. particular geometry recently in a more rigorous analysis
The number and general form of these blocks is specified which confirms that although these ‘quasi-upper-bound’
in the model chosen: their displacements must be compati- solutions can be significantly in error as far as geometric
ble with incompressibility of the material and the imposed predictions are concerned the values of ploughing forces,
boundary conditions, but the exact position of the inter- and the associated expenditure of energy are likely to be
faces, as well as the ploughing component of the force, is accurate to a few percent w30x.
found by minimising the rate at which energy is expended The production of a wear groove in an initially feature-
at the sliding which takes place at the inter-block bound- less surface is clearly a great simplification of what actu-
aries. This method of analysis is based on the upper limit ally occurs in practice when wear events are superposed
load theorem which states that, provided certain conditions upon one another to give a statistically stable but develop-
are met, the true velocity field minimises the overall rate ing surface profile, and there are clearly problems in

Fig. 4. Ža. A simple conical wear indenter presenting an attack angle u and ploughing a groove in a soft material of depth h. Žb. A pyramidal indenter
moving through previously undeformed material. Žc. The geometry of the indenter is described by the two angles c and u . Žd. The wear mode diagram
after Ref. w28x, ‘real’ asperities will have 908 - 2 c - 1808. Že. Ži. A single 50 mm high diamond asperity produced by deposition Žii. densely and Žiii.
˚
sparsely packed arrays of such model asperities Žafter Gahlin w33x..
 J.A. Williamsr Wear 225–229 (1999) 1–17 9

angles found in single wear track investigations which are

typically of the order of 208 or more. Consequently, if we
consider their asperities to be acting separately and take
into account the effect of the critical attack angle minimal
wear losses are predicted. A cursory examination of an
abrasively worn surface leaves no doubt as to the direction
of relative sliding as the original features of the surface are
gradually replaced by a series of parallel wear tracks. In a
very short time, the load carrying asperities are abrading
an already abraded and profiled surface. The problem of
allowing for the interaction of later events with the profile
left by earlier asperity passes remains unsolved. In their
work on the statistical development of surfaces, Jacobson
et al. w31x assumed that all the displaced material was lost
and, not surprisingly, thereby overestimated the resultant
wear coefficient. Garrison w32x and Zum Gahr w14x amongst
others have attempted to account for the extent to which
ploughed material is removed by subsequent abrasion by
the introduction of an empirical factor Žwhich depends on
the work hardening exponent of the metal. and recent
work by Gahlin˚ et al. w33x using artificially abrading
surfaces with remarkably precisely controlled geometry
Žsee Fig. 4Že.. makes experimental interpretation much
easier.
The work by Williams and Xie w34x has shown that the
proportion of the material displaced by the passage of the
hard asperity is affected by its geometry, the ability of the
material at the surface to strain harden and the proximity
of any near-by pre-existing tracks; these have the effect of
reducing the angle of attack at which micro-cutting be-
comes the preferred mechanism of material response Žsee
Fig. 5.. This analysis was also based on a quasi-upper
bound argument and while comparison of the predicted
traction or friction force with that measured showed a
good agreement it was, as anticipated, much more difficult
to make predictions about the distribution of deformations,
and thus the wear rates themselves.

2.2.2. Simplified two-dimensional models

Because of the geometric complexities inherent in
three-dimensional plasticity solutions, no complete analyti-
cal model Žin the sense of simultaneously satisfying the
requirements of equilibrium, compatibility and a plasticity
flow rule. of the ploughing or micro-machining of a flat
surface by an indenter with the comparatively simple
shape of either a cone or a pyramid has yet been produced.
To facilitate the analysis, an alternative simplification to
Fig. 4 Žcontinued..
the use of an upper bound technique is to consider the
analogous two dimensional problem. If the material is
again imagined to exhibit perfect plasticity then the prob-
transferring the results of this form of investigation to lem is tractable by the technique of slip line fields which
more practical circumstances involving multiple asperity have the advantage of satisfying both the demands of
interactions such as produce the data of Fig. 3. The equilibrium of stresses and compatibility of strains. How-
majority of real surfaces, even very rough ones, have mean ever, in such two-dimensional or plane strain problem, the
values of slope very much less than the critical attack asperity is obliged to take the form of an infinitely wide
10 J.A. Williamsr Wear 225–229 (1999) 1–17

Fig. 5. Ža. The effect on wear mode of a pre-existing wear track produced in a surface by a previous pass of the asperity; L is the lateral displacement of
the two tracks and h the dimension indicated. The wear map is for such a test on a copper surface comparing predicted and experimentally observed
regions of behaviour: open symbols: predominant mechanism ploughing and closed: significant amount of material was lost by cutting. The solid line is the
prediction from the upper bound model referred to in the text.

wedge and for steady state conditions to be achieved, some normal force per unit length of asperity and the material
material must either actually be removed by the cutting hardness H as
process or the asperity must ride up vertically during the
initial stage of the deformation until its apex is at the same hH 3'3
Ks s Ž 4.
level as the undeformed surface; deformation will now be P 1
associated with a plastic wave pushed ahead of it—like a
ruck in a carpet. A possible slip line field for this deforma-
½ ž
1q2
4
pygyu /5 cot Ž g y u . y 1

tion mode, originally suggested by Green w35x but much where g is the shear angle made by the shear plane with
developed by Petryk w36x and Challen and Oxley w37x, is the direction of relative motion. Estimating a wear rate for
illustrated in Fig. 6Ža.; a simple possible shear mechanism the case of wave propagation is clearly more difficult than
for the machining mode is shown in Fig. 6Žb.. for the cutting mode because an inherent feature of
Each field provides, from consideration of the stresses steady-state wave formation is that the asperity is taken to
within the plastically deforming material, expressions for ride up during the initial phase of the motion to the level
the ratio of the tangential or traction force on the asperity of the undeformed free surface—in one real sense, there-
to the normal loading, i.e., the overall or global coefficient fore, it is a ‘no wear’ model.
of friction m , in terms of the attack angle u and the However, although the slip line field of Fig. 6Ža. allows
strength of the interface between the deforming material for no immediate loss of material, every element of mate-
and the indenter expressed as a friction factor f. These are rial which lies at a depth below the free surface of no more
shown in Fig. 6Žc.; notice that increasing the value of f at than the greatest extent of the arc between points B and C
a given value of u provides a greater value of overall Ži.e., dimension h., experiences a sequence of plastic strain
friction for the wave model but reduces m during chip increments as it moves through the zone of deformation.
formation. When f s 0 then m s tan u . It is clear that for Each element is compressed by a direct strain D ´ x x paral-
the majority of asperities on engineering surfaces Žfor lel to the surface and subsequently extended by the same
which u - 58. the wave mechanism will operate. The amount Žsince there is no net surface extension. but, in
lowest value of u for which machining will be preferred is addition, is given a uni-directional increment of shear
21.28. For very severe asperities and non-zero friction, i.e., strain D ´ zx : these shear components accumulate with each
f ) 0, it may seem as if there are two possible responses, passage of the indenting wedge. The effective cycle of
both cutting and wave formation: however, since the over- plastic deformation is thus not closed but has both a
all coefficient of friction is a measure of the global energy reversing component and an incremental, monotonic com-
expenditure, in this overlapping region, the mechanism ponent—the so-called ratchetting element. These strain
with the lower value of m , i.e., cutting, will be favoured. components can be related to the geometry of the slip-line
The wear rate for the cutting model is relatively easy to field and Fig. 7 shows some typical stain history cycles
establish by evaluating the dimension h in terms of P the experienced by materials passing through a slip-line field
 J.A. Williamsr Wear 225–229 (1999) 1–17 11

Fig. 6. Slip line fields for the interaction of a hard asperity with a soft surface. Ža. Steady-state plastic wave ŽABCDE. whose geometry is defined by there
angles u , f and h. Žb. Chip formation or micro-machining at larger values of attack angle u . Žc. Resulting map of overall coefficient of friction vs.
asperity attack angle for differing values of interface friction factor f Žafter Ref. w37x..

of this geometry for various values of the roughness or generation of wear particles under these circumstances
attack angle u . Notice that these strain values are large— appears to be intimately associated with the magnitude of
even when the asperity attack angle is 2.58 a single pass of these strain components although recent literature contains
the indenter imposes a shear strain of nearly 15%. The a lively debate about the relative significance of the revers-
12 J.A. Williamsr Wear 225–229 (1999) 1–17

Fig. 7. Strain cycle imposed by wave slip line field for f s 0.25 and a s 108, 58 and 2.58 Žfrom Ref. w39x..

ing and the ratchetting components w38x. It may be that of f, while increasing the asperity attack angle u leads to a
some circumstances, e.g. in reciprocal sliding, the revers- transition in the particular mechanism of wear Žfrom
ing component is of most significance, while in uni-direc- ploughing to cutting. there is no dramatic change in wear
tional sliding the ratchetting element is of more relevance. rate. In theory at least, very high values of f can produce
On the basis that the rate determining mechanism is that values of K greater than unity, resulting from the in-
of ratchetting, and that material failure occurs when the creased depth of penetration of the plastic zone into the
accumulated strain reaches some fixed value, the wear rate softer material. In practice, engineering contacts will be
of the surface will be proportional to D ´ zx , the ratchetting restricted to the range 0 - f - 0.2 so that K - 1. Within
strain experienced by an element of material as it passes this range the even spacing of the contours for K Ži.e.,
through the field. An equation for K, equivalent to Eq. Ž4., K s 10y3 , 10y2 and 10y1 . suggest that to a reasonable
but for the plastic wave mechanism has been derived w39x approximation K is proportional to the square of the
as asperity slopes.
However, there is limitation to this wear mode; smooth,
'3 wsin f y sin u x D´zx well lubricated surfaces with attack angles of less than a
Ks = Ž 5.
Acos u y sin Ž 2 f y u . C few degrees and f - 0.2 give very much lower wear rates
than the extension of this rigid-plastic model would sug-
where A s 1 q pr2 q 2 f y 2h y 2 u and the angle f gest, e.g., values of 10y7 or 10y8 rather than 10y4 Žthe
Žshown on Fig. 6Ža.. is related to the friction stress t at the ‘severe’ to ‘mild’ transition of Figs. 1 and 2.. One option
interface between the asperity and the wave by the equa- is to explain this discrepancy between the predicted and
tion observed values of K in much the same way that Archard
f s 0.5 arccos f . Ž 6. reconciled his much earlier estimates of wear rates with
observation, namely that in a surface covered with an array
There are three contributions to D ´ zx , the first when the of asperities, the statistics will be such that only a very
material enters the plastic zone by crossing the line AB, small proportion of the harder asperities are ‘active’—in
the second within the fan region BCE and a third as the the sense of being sufficiently aggressive to contribute to a
material leaves the plastic zone across CD and expressions wear mechanism directly. Effectively, this is to say that
for these, in terms of the angles of the field, are available real asperities are unlikely, at their tips, to be of the
w39x. angular form the analysis supposes but rather of a more
We can now begin to plot a wear map, of the form of cylindrical profile Žif we retain the notion of a strictly
Fig. 2, for a specific material with known material proper- two-dimensional contact., and thus for those relatively
ties; as an example taking the value C in Eq. Ž5. to be 2.9 lightly loaded of a very small effective angle. Under these
Žas in Ref. w39x. generates the right hand section of the more benign circumstances of mild wear, we have to begin
map of Fig. 2 with the contours of K as shown. At high to consider the elastic as well as Žor even instead of. the
asperity angles micro-machining or cutting is preferred to plastic response of the contact. As the proportion of asper-
wave formation and, as anticipated, within this zone reduc- ity interactions that generate only elastic stresses increases,
ing interfacial friction, i.e., the value of the friction factor we expect a very steep reduction in the rate of surface
f, increases the wear rate. Within the zone associated with damage and a new mechanism of material loss to be
wave formation the reverse is true. Note that at low values established.
 J.A. Williamsr Wear 225–229 (1999) 1–17 13

3. Mild wear half-space has been assumed to be elastic-perfectly plastic

with a yield strain equal to 0.01%. The asperity was first
loaded to produce a vertical displacement equal to 0.5% of
3.1. Effects of elasticity— shakedown its radius and then has been slid, to the right, through a
distance equal to three asperity radii. The surface layer is
The unit event of interest now becomes the passage of left with a pattern of residual compressive stresses. If these
an asperity with a summit which has a curved profile are sufficient for subsequent applications of the load to be
Žeither cylindrical or spherical. sliding over an elastic-plas- carried entirely elastically then the surface is said to have
tic half-space. Such a component can respond to the shaken down and the maximum pressure for which this is
conditions of repeated loading, as will occur in a continu- possible is known as the shakedown limit or shakedown
ous sliding contact with a harder rough counterface, in a pressure, ps w10,41x. In such a case, since the steady-state
number of different and distinct ways. stresses are again elastic, we also expect very low wear
At sufficiently small loads where no element within the rates.
surface reaches the yield point the stress field beneath the At loads above the shakedown limit, there will be some
contact is perfectly elastic and so fully reversible. Any plastic deformation at each contact and what happens in
damage or wear would be associated with some form of the steady-state will be influenced by the way in which the
high cycle fatigue and expected wear rates would be very softer surface responds to the imposition of repeated plas-
low. Fig. 8Ža., for example, illustrates the elastic stresses tic straining, in other words, the way in which it hardens.
Žplotted as maximum Tresca values. beneath a compara- In the case of linear kinematically hardening material Ži.e.,
tively lightly loaded cylindrical moving asperity with a one with a hardening modulus independent of the mean
local coefficient of friction of 0.3. The peak stress is stress level in the deformation cycle. the steady-state
generated beneath the surface at point C but as the asperity situation will consist of a closed loop of reversed plastic
translates, say from left to right, so the elastic field sweeps strain; in the more realistic case of a material exhibiting
through the half-space; material behind it is left stress free nonlinear kinematic hardening Ža hardening modulus which
w40x. decreases with increasing peak stress. ratchetting is ex-
At somewhat higher loadings, although plastic deforma- pected. Calculations of the plastic strain cycle at the
tion may take place on the first passage of the load, the surface under a sliding cylinder in plane strain Žfor exam-
process of elastic unloading leads to the generation of ple by Bower and Johnson w42x. show that again this
residual stresses Žand perhaps also an increase in the local consists of a combination of reversing and ratchetting
value of the strength of the materials through some form of components in many ways similar to that arising from the
strain hardening.. Fig. 8Žb. shows the Tresca stress con- plastic wave described above for rigid-perfectly plastic
tours for a cylindrical asperity carrying a higher load than materials—except that the magnitude of the deformation
in Fig. 8Ža. and in which the material of the deforming per cycle is now very much less.

Fig. 8. Distribution of equivalent Tresca stresses beneath a traversing cylindrical asperity. Ža. Elastic conditions for the case m s 0.3. The most heavily
loaded element of material is at the depth of point C. Žb. The half-space is now elastic-plastic with a yield strain Ži.e., YrE . of 0.0001 and the indenter has
moved through a distance of 3 R under a load which is sufficient to exceed the elastic limit of the material. Local value of m s 0.1. A system of protective
Ži.e., compressive. residual elastic stresses is left in the surface layers of the half-space.
14 J.A. Williamsr Wear 225–229 (1999) 1–17

Fig. 9. The different forms of component response to cyclic loading. Ža. perfectly elastic, Žb. elastic shakedown, Žc. plastic shakedown or cyclic plasticity,
Žd. incremental collapse or ratchetting failure.

These different responses of a structure, in this case the Very thin slivers or lamellae of material, typically less than
component surface, are illustrated in Fig. 9. At sufficiently a micron in thickness, are extruded from the edges of the
small loads such no element reaches yield the response Ža. contact patches on the softer surface. This behaviour has
is perfectly elastic and reversible. In Fig. 9Žb., although been termed ‘filmy wear’ and is quite distinct from the
plastic flow takes place in the first cycle, the generation of much more severe forms of surface damage associated
residual stresses Žtogether with strain hardening. have en- with more aggressive values of u . Kapoor and Johnson
abled the structure to shakedown to a purely elastic re- w46x have explained how such thin slivers can be formed
sponse in the cyclic steady-state. If the elastic shakedown by a mechanism in which a thin surface layer of the softer
limit is exceeded, then there will be an element of plastic material is compressed and sheared, or ‘pummelled’, by
deformation with each passage of the load though whether contact with the asperities on the harder and rougher
of the closed loop form of Žc. or ‘ratchetting’ form of Žd. counterface. When the islands or patches of contact are
depends on details of the material behaviour. aligned transversely to the direction of relative sliding, as
Under lubricated, and so low friction, sliding conditions illustrated in Fig. 10, then the compressed material appears
involving a harder surface with low values of roughness or as an extrusion on the downstream side of each asperity
slope, a characteristic from of mild wear is often apparent contact. If the lay of the harder surface is aligned parallel
and has been described by, among others, Akagaki and to the sliding direction Ži.e., parallel to the x-axis rather
Kato w43x, Kuo and Rigney w44x and Martin et al. w45x. than the y-axis in the figure. although extrusion in the

Fig. 10. A mechanism for the formation of ‘filmy’ wear particles: the surface layer of the contact patches on the softer surface are ‘pummelled’ by the
asperities on the harder Žafter Ref. w46x.. When the ‘lay’ of the softer surface Žshown here parallel to the y-direction. is at right angles to the direction of
relative motion Žthe x-direction. filmy wear is extruded on their downstream side. If the relative motion were parallel to the lay, i.e., in the y-direction,
wear particles are extruded on either side of each of the contact patches.
 J.A. Williamsr Wear 225–229 (1999) 1–17 15

Fig. 11. The non-dimensional wear coefficient K as a function of the plasticity index for repeated sliding cs for various values of the mean surface
separation drs for the mild wear regime illustrated in Fig. 10 Žafter Ref. w47x..

sliding direction is prevented, a ratchetting mode of incre- of asperity contacts exceeding shakedown is negligibly
mental collapse involving lateral extrusion is still possible. small as is indicated by the very rapid reduction in the
Estimates of wear rates based on this mechanism can be value of K ; this is effectively the shakedown state associ-
made if the harder, now assumed spherically tipped asperi- ated with the region of ultra-low wear rate in Fig. 2. The
ties, of radius R, on the counterface are assumed to have fact that the model shows some dependence of K on the
some appropriate distribution of heights. Kapoor et al. w47x mean spacing of the two surfaces implies some depen-
have shown that for a Gaussian distribution with standard dence of this non-dimensional wear parameter on load
deviation s , the non-dimensional wear coefficient K is Žapproximately proportional to 'load . unlike the Archard
principally dependent on a non-dimensional group cs de- equation in which K is taken to be independent of load
fined as intensity. The hardness of the wearing surface and the
EU s coefficient of friction influence the wear rate through their
cs s
ps
( R
; Ž 7. effect on the value of the shakedown pressure ps . In
principle a model such as this enables values of wear rate
EU is the elastic contact modulus of the contact conven- to be calculated—although at the moment there remain
tionally defined as Ž1 y n 12 .rE1 q Ž1 y n 22 .rE2 4 y1 and significant uncertainties in achieving this goal: for exam-
ps is the shakedown pressure. The group cs can be thought ple, the wear rate is related to the density of asperities on
of as either the ratio of the roughness of the harder surface the hard counterface as well as their radii of curvature and
to the shakedown strain of the softer surface or as a rms heights, but these quantities are known to change with
measure of the proportion of the asperity contacts that are the sampling interval at which they are measured. The
responding plastically, by ratchetting, compared to those wear rate likewise depends on the ratchet strain per cycle
that have been shaken down elastically. To this extent the but its reliable estimate poses serious problems. Theoreti-
group cs is a ‘plasticity index appropriate for repeated cal values relating the rate of damage accumulation to the
contacts’ w48x analogous to the plasticity index introduced extent to which shakedown pressures are exceeded at
by Greenwood and Williamson w49x and Mikic w50x for individual asperity contacts are sensitive to strain harden-
static contact. ing, and experimental fatigue data, under these very high
Fig. 11 shows how the wear coefficient K varies with hydrostatic pressures, when ductility is much enhanced,
cs for a particular case Ž H s 3 GPa, EU s 115 GPa other are scarce w51x.
details in Ref. w46x., the three curves representing different
values of mean surface separation. It is clear that K is
very dependent on cs ; as the value of this index varies 4. Conclusions
from 1.0 to 3.5 so the wear coefficients vary over several
orders of magnitude from very mild Ž- 10y8 . to very However measured, wear in conditions of sliding con-
severe Ž) 10y2 .. If the value of cs - 1, then the number tact can vary over many orders of magnitude. There is no
16 J.A. Williamsr Wear 225–229 (1999) 1–17

universal mechanism of wear and no simple correlation References

between rates of wear or surface degradation and values of
friction coefficient. Identifying the operating point of a w1x H.C. Meng, K.C. Ludema, Wear models and predictive equations:
sliding contact in an appropriate wear map can assist in their form and content, Wear 181 Ž1995. 443–457.
establishing the possible or likely modes of surface dam- w2x J.F. Archard, Contact and rubbing of flat surfaces, J. Appl. Phys. 24
age and how close the operating conditions are to any Ž1953. 981–988.
w3x E. Rabinowicz, Friction and Wear of Materials, Wiley, 1965.
transition between mild and severe regimes of wear. Quan-
w4x J.F. Archard, W. Hirst, The wear of metals under unlubricated
titative models are restricted to limited areas of any such conditions, Proc. R. Soc. A 236 Ž1956. 397–410.
map and are material-specific. They often require material w5x S.C. Lim, Recent developments in wear mechanism maps in New
properties or process ‘constants’ which cannot be estab- Directions in Tribology, I.M. Hutchings ŽEd.., MEP London, 1997.
lished other than in a wear test itself and so are self-refer- w6x S.C. Lim, M.F. Ashby, Wear mechanism maps, Acta Metallurgica
ential. Nevertheless, there has been some progress in mod- 35 Ž1987. 1–24.
w7x T.H.C. Childs, The mapping of metallic sliding wear, Proc. Instn.
elling specific examples of mechanical wear and here we Mech. Engrs. C 202 Ž1988. 379–395.
examine recent work in the cases of severe abrasive wear w8x N.P. Suh, An overview of the delamination theory of wear, Wear 44
and mild sliding wear. Ž1977. 1–16.
w9x S. Jahanmir, On the mechanics and mechanisms of laminar wear
particle formation, Advances in Mech. and Phys. Surfaces 3 Ž1986.
261–332.
5. Nomenclature w10x K.L. Johnson, Contact Mechanics, Cambridge Univ. Press, 1985.
w11x K.L. Johnson, Contact mechanics and the wear of metals. Proc.
Austrib ’94 Int. Trib. Conf. December 1994, Perth, 53–58.
A Area w12x M.M. Kruschov, Resistance of metals to wear by abrasion, as related
C Material constant to hardness, Proc. Conference on Lubrication and Wear, Instn Mech.
Engrs, London, 1957, 655–659.
E Elastic modulus w13x R.C.D. Richardson, The wear of metals by hard abrasives, Wear 10
´ Strain Ž1967. 291.
f, h Angles within slip-line field w14x K-H. Zum-Gahr, Micro-structure and wear of materials, Elsevier,
g Shear plane angle Amsterdam, 1987.
H Indentation hardness w15x S.M. Hsu, M.C. Shen, Ceramic wear maps, Wear 200 Ž1996.
154–175.
h Wear or layer depth w16x Y.S. Wang, S.M. Hsu, Wear and wear transition mechanisms of
K Non-dimensional wear coefficient ceramics, Wear 195 Ž1996. 35–46 and 112–122.
k Flow stress in shear w17x T.O. Mulhearn, L.E. Samuels, The abrasion of metals: a model of
L Lateral displacement the process, Wear 5 Ž1962. 478–498.
m Coefficient of friction w18x T. Tsukizoe, T. Sakamoto, Friction in scratching without metal
transfer, Bull. Jap. Soc. Mech. Engrs. 18 Ž1975. 65–72.
n Poisson’s ratio w19x H. Sin, N. Saka, N.P. Suh, Abrasive wear mechanism and the grit
P Normal load size effect, Wear 55 Ž1979. 163–190.
p Pressure w20x K.H. Zum Gahr, Modelling of two-body abrasive wear, Wear 124
ps Shakedown pressure Ž1988. 87–102.
u Asperity attack angle w21x R.F. Scrutton, N. Youssef, The action of build-up when scraping
with rough conical tools, Wear 15 Ž1970. 411–421.
R Radius of curvature w22x A.A. Torrance, A three dimensional cutting criterion for abrasion,
s Sliding distance Wear 123 Ž1988. 87–96.
s Standard deviation w23x T.H.C. Childs, The sliding of rigid cones over metals in high
V Sliding velocity adhesion conditions, Int. J. Mech. Sci. 12 Ž1970. 393–403.
w Wear volume w24x P. Gilormini, E. Felder, Theoretical and experimental study of
ploughing of rigid-plastic semi-infinite body by a rigid pyramidal
c Dihedral angle
indenter, Wear 88 Ž1983. 195–206.
cs Plasticity index w25x Abildgaard, T., Prediction of force components acting on a plough-
f Friction factor ing cone by means of three-dimensional upper bound theory, in
Proc. Int. Conf on Advanced Technology of Plasticity, Tokyo, 1984,
121–126.
w26x Abebe, M., Appl, F.C., Theoretical analysis of the basic mechanics
Acknowledgements of abrasive processes. Wear 126 Ž1988. 251–266 and 267–283.
w27x J.A. Williams, Analytical models of scratch hardness, Trib. Int. 29
Ž8. Ž1996. 675–694.
I am grateful to those who have contributed to the work w28x K. Kato, K. Hokkirigawa, T. Kayaba, Y. Endo, Three dimensional
described above or with whom I have been able to discuss shape effect on abrasive wear, J. Trib. 108 Ž1986. 346–351.
w29x B. Avitzur, S.H. Talbert, in: T. Blazynski ŽEd.., Plasticity and
some of these issues, in particular: Professor K.L. Johnson,
Modern Metal Forming, Elsevier, Amsterdam, 1989.
Dr. A. Kapoor, Professor A.R.S. Ponter, Dr. I.N. Dyson, w30x A. Azarkhin, O. Richmond, A generalisation of the upper bound
Dr. Y. Xie, Dr. A.A. Torrance, Professor T.H.C. Childs method with particular reference to the problem of a ploughing
and Dr. G.M. Genin. indenter, J. Appl. Mech. 56 Ž1992. 10–14.
 J.A. Williamsr Wear 225–229 (1999) 1–17 17

w31x S. Jacobson, P. Wallen, S. Hogmark, Correlation between groove w41x K.L. Johnson, The application of shakedown principles in rolling
size, wear rate and topography of abraded surfaces, Wear 115 Ž1987. and sliding contact, Eur. J. Mech., ArSolids 11 Ž1992. 155–172,
83–93. Special Issue.
w32x W.M. Garrison, Abrasive wear resistance: the effect of ploughing w42x A.F. Bower, K.L. Johnson, The influence of strain hardening on the
and the removal of ploughed material, Wear 114 Ž1986. 239–247. cumulative plastic deformation in rolling and sliding line contact, J.
w33x R. Gahlin,
˚ Micro-scale studies of Wear. PhD Dissertation, Univ. of Mech. Phys. Sol. 37 Ž1980. 471–493.
Upsala, 1998. w43x T. Akagaki, K. Kato, Plastic flow processes in flow wear under
w34x J.A. Williams, Y. Xie, The generation of wear surfaces through the boundary lubricated conditions, Wear 117 Ž1987. 179–196.
interaction of parallel grooves. Wear 155, 363–379. w44x S.M. Kuo, D.A. Rigney, Sliding behaviour of aluminium, Mater.
w35x A.P. Green, The plastic yielding of metal junctions due to combined Sci. Eng. A 157 Ž1992. 131–139.
shear and pressure, J. Mech. Phys. Solids 2 Ž1954. 197–201. w45x J.M. Martin, J.L. Mansot, I. Bebezier, H. Dexpert, The nature and
w36x H. Petryk, Slip line field solutions for sliding contact, Proc. Conf. origin of wear particles from boundary lubrication with ZDTP, Wear
Friction, Lubrication and Wear, Instn of Mech. Engs, London, 1987, 93 Ž1984. 117–125.
987–994. w46x A. Kapoor, K.L. Johnson, Plastic ratchetting as a mechanism of
w37x J.M. Challen, P.L.B. Oxley, An explanation of the different regimes metallic wear, Proc. R. Soc. A 445 Ž1994. 367–381.
of friction and wear using asperity deformation models, Wear 53 w47x A. Kapoor, K.L. Johnson, J.A. Williams, A model for the mild
Ž1979. 229–243. ratchetting wear of metals, Wear 200 Ž1996. 38–44.
w38x A. Kapoor, A re-evaluation of the life to rupture of ductile metals by w48x A. Kapoor, K.L. Johnson, J.A. Williams, The steady-state sliding of
cyclic plastic strain, Fatigue and Frac. Engng. Mat. and Struct. 17 rough surfaces, Wear 175 Ž1994. 81–92.
Ž1994. 201–219. w49x J.A. Greenwood, J.P.B. Williamson, Contact of nominally flat sur-
w39x A.A. Torrance, T.R. Buckley, A slip-line field model of abrasive faces, Proc. R. Soc. A 295 Ž1966. 300–330.
wear, Wear 196 Ž1996. 35–45. w50x B.B. Mikic, Thermal contact conductance: theoretical considera-
w40x J.D. Bressan, G.M. Genin, J.A. Williams, The influence of pressure, tions, Int. J. Heat and Mass Transfer 17 Ž1974. 205–214.
boundary film shear strength and elasticity on the friction between a w51x Y. Jaing, H. Sehitoglu, Rolling contact stress analysis with the
hard asperity and a deforming softer surface, Proc. 24th Leeds-Lyon application of a new plasticity model, Wear 191 Ž1996. 35–44.
Symposium on Tribology, in press.