1987 - Lim e Ashby - Wear-Mechanism Maps

Acta metal/. Vol. 35, No. I, pp. l-24, 1987 OOOI-6160/87 53.00 + 0.
00
Printed in Great Britain. All rights reserved Copyright c 1987 Pergamon Journals Ltd
OVERVIEW NO. 55
WEAR-MECHANISM MAPS
S. C. LIM’ and M. F. ASHBY’

‘Department of Mechanical and Production Engineering, National University of Singapore,
Kent Ridge, Singapore 0511, Republic of Singapore and ‘University Engineering Department,
Trumpington Street, Cambridge CB2 lPZ, England
(Received 10 March 1986; in revised form II March 1986)
Abstract-The potential of Wear-MechanismDiagrams is explored. Diagrams which show the rate and
the regime of dominance of each of a number of mechanisms of dry wear (delamination, mild and severe
oxidation, melting, seizure, etc.) are constructed empirically (that is, from experimental data alone) and
by modelling (by theoretical analysis calibrated to experiment). The method is applied to steels, and has
wider application as a way of classifying and ordering wear data, and of showing the relationships between
competing wear mechanisms.
Resume-Nous explorons les possibilites des diagrammes de mPcanismesd’usure. Nous construisons

empiriquement (c’est a dire, a partir des seules don&es experimentales) et a I’aide de modeles (grace a
des analyses theoriques Ctalonnies sur l’experience) des diagrammes qui fournissent la vitesse et le regime
de predominance de chacun des mecanismes d’usure s&he (delamination, oxydation legtre et prononcte,
fusion etc). Nous appliquons cette methode aux aciers, mais elle a des applications plus larges en tant
que moyen de classification et de tri des don&es d’usure, ou pour montrer les relations entre les
mecanismes d’usure en competition.
Zusammenfassung-Die Anwendungsmijglichkeiten von Diagrammen der Abriebmechanismen werden

untersucht. Diagramme, die Geschwindigkeit und Bereich des Auftretens einer Reihe von Mechanismen
des trockenen Abriebes (Delamiantion, schwache und starke Oxidation, Schmelzen usw) aufzeigen, werden
empirisch (d.h. aus experimentellen Werten) und im Model1 (d.h. durch theoretische, an das Experiment
angeglichene Analyse) konstruiert. Diese Methode wird auf Stiihle angewendet. Sie hat einen weiteren
Anwendungsbereich, indem sie Abriebdaten klassifizieren und ordnen und die Zusammenhange zwischen
konkurrierenden Abriebmechanism en aufzeigen kann.
Conconius, vile critic, leave me now in peace the details of a single mechanism, not on the re-
and keep your lard-like thoughts for axle- lationship between mechanisms.
grease.. .“t One way of exploring this broader pattern, Tabor
Marcus Valerius Martialis [1] [2] suggests, might be to construct wear-mechanism
maps which (by analogy with deformation-
mechanism maps [3]) summarise data and models for
1. INTRODUCTION wear, showing how the mechanisms interface, and
allowing the dominant mechanism, for any given set
The engineering problem of wear was known to the of conditions to be identified. We have tried to do
ancients who used lubricants such as lard (for the this, and report the results here. We have followed
axles of chariots) to reduce friction and to avoid the two (converging) paths. One is empirical: mechanism
excessive material-loss which is inevitable when dry maps are built up by plotting experimental data for
surfaces rub together. What the modern engineer wear rates on suitable axes, identifying the mech-
knows that the ancients did not is there are many anism at each point by observation. The other uses
mechanisms of wear: wear caused by adhesion, wear physical modelling: model-based equations de-
by abrasion, by oxidation, by delamination, by melt- scribing the wear rate caused by each mechanism are
ing, and more. The outsider to the field is struck by combined (using a numerical procedure) to give a
its complexity, and by the lack of an overall frame- map showing the total rate, and the field of domi-
work or pattern into which the individual obser- nance of each. These models contain constants which
vations can be fitted: research has tended to focus on are poorly known; so the final step is to calibrate each
model-based equation using blocks of data which lie
tcosconi, qui longa pulas epigrammata nostra, Utilis un- in its field. The final version of the map is the result
guendis axibus esse pates.. of this process, and combines, as best as we have been
2 LIM and ASHBY: OVERVIEW NO. 55
able, the experimental data with results of theoretical 2. WEAR-MECHANISM MAPS

thinking.
It is important to emphasise that this study is The wear rate W of a sliding surface is con-
intended as a first, broad, survey. Accordingly, the ventionally defined as the volume lost from the
theoretical treatments are reduced to the simplest surface per unit distance slid. Its dimensions are m2
possible level: heat flow is approximated by the (it is a derivative with respect to distance, not to
equivalent one-dimensional problem; rates of ox- time). For a given sliding geometry, it is a function
idation are calculated from the simplest kinetic of the normal force, F, acting across the sliding
model; and so forth. The justification is that the range surfaces, their relative velocity, v, their initial tem-
of the variables covered by the study-load, sliding perature, TO, and the thermal, mechanical and them -
velocity and wear rate-is enormous; and the experi- ical properties of the materials which meet at the
mental scatter, when the results of many different surface. For each of the mechanisms, (the ith mech-
investigations are combined, is pretty large too. An anism, for instance) one can write
error of a factor of 2-even of l&in the theory does
not seriously change the conclusions. This is not to
thermal, mechanical, chemical props.}. (1)
dismiss the detailed models which abound in the
literature: their precision is of the greatest value in If the mechanisms do not interact, then the dominant
their own context. But it is not needed here, and the mechanism is the one which, for a given F, v, TO, etc
points to be made will emerge most clearly if the leads to the greatest wear rate. Sliding may heat the
calculations are reduced to a minimum. surfaces, of course, but the local temperature does
The symbols used in the paper are defined where not appear among the independent variable of equa-
they first appear; for convenience they are also assem- tion (1) because it, too depends directly on them: if
bled in Table 1, with their definitions and dimensions. they are specified, the local temperature distribution
The reader wishing to get a quick idea of the paper (which we discuss below) is determined also.
might read Sections 2 and 6, and then look at Figs Experience with problems of this sort shows that
16 and 27. much is to be gained by the use of dimensionless
Table I. Symbols, definitions and units
wear rate@G/m) thermal conductivity of oxide (J/msK)

normalised were rate effective thermal conductivity (J/msK)
normal force on sliding interface (N) total number of contacting asperities
normalised pressure on sliding interface true number of contacting asperities
sliding velocity (m/s) probability of forming a new asperity contact
normalised velocity constant used in the calculation of N
nominal area of contact (m2) shear stress (N/m2)
real area of contact (m*) constant used in Tabor’s junction growth equation
radius of pin (m) shear strain rate
radius of an asperity (m) pre-exponential constant in low-temperature dislocation glide
equivalent linear diffusion distance for bulk beating (m) activation energy for low-temperature dislocation glide (J/mol)
equivalent linear diffusion distance for flash heating (m) latent heat of fusion per unit volume for metal (J/m))
dimensionless parameter for bulk heating ( = C/r,,) latent heat of fusion per unit volume for oxide (J/m’)
hardness of sliding surface (N/m*) volumetric rate of production of molten material (ml/s)
room-temperature hardness of metal (N/m*) volume fraction of molten material removed during sliding
room-temperature hardness of oxide (N/m’) constant used in the model for mild-oxidational wear
heat flux (J/m2s) molar gas constant (8.314 J/mol K)
rate of heat input per unit area (J/m’s) activation energy for oxidation (J/mol)
enhanced rate of heat input per unit area at an asperity (J/m2s) arrhenius constant for oxidation (kg2/m4s)
rate of heat conduction into the asperities per unit area (J/m’s) mass of oxygen intake by the oxide film per unit area (kg/m2)
Jaeger’s dimensionless constant parabolic oxidation rate constant (kg*/m%)
heat distribution coefficient mass fraction of oxygen in the oxide film
coefficient of friction thickness of oxide film at an asperity (m)
temperature (K) critical thickness of oxide film (m)
bulk temperature (K) thickness of oxide film due to bulk heating (m)
flash temperature (K) thickness of oxide film due to flash heating (m)
sink temperature for hulk heating (K) pin/disk interaction time (s)
sink temperature for flash heating (K) heat diffusion time (s)
temperature at the metal/oxide interface (K) oxidation time (s)
an equivalent temperature for metal (K) time taken to reach the critical oxide thickness (s)
an effective equivalent temperature for metal (K) critical depth for void nucleation (m)
an equivalent temperature for oxide (K) number of inclusions per unit volume
temperature for martensite formation (K) radius of an inclusion (m)
melting temperature of metal (K) volume fraction of inclusions
melting temperature of oxide (K) area fraction of voids
oxidation temperature (K) critical area fraction of voids
specific heat (J/kgK) cumulative plastic shear strain
density of metal (kg/m’) cumulative plastic shear strain needed to produce failure
density of oxide (kg/m’) plastic shear strain accumulated per pass
specific heat per unit volume (Jjm’K) total number of passes to failure
thermal diffusivity of metal (m2/s) ( = K/PC) distance between asperity contacts
thermal diffusivity of oxide (m2/s) Archard’s dimensionless wear coefficient
thermal conductivity of metal (J/msK)
LIM and ASHBY: OVERVIEW NO. 55
variables. Because many mechanisms are involved (b) medium carbon (0.3 < C < 0.7 wt%)
there is no single ideal choice of variables. But both steels;
the data-plots (shown presently) and the devel- (c) high carbon (C > 0.7 wt%) steels;
opoment of the models (coming presently too) sug- (d) low alloy (2-5 wt% Cr, MO, V and Ni)
gest that data from different sources, using specimens steels
of differing shapes and sizes can best be correlated by (e) high alloy (typically 18 wt% Cr, 8 wt%
using a normalised wear rate, force and sliding Ni) steels; and
velocity, defined by (f) tool (typically 20 wt% of W, Co and Ni,
1.5 wt% C) steels.
WC” The data for fi are plotted on axes of P and d in
n
Figs l-6. Figure 1 (pure iron and mild steels) illus-
F
i;=-.--- trates the main features: the data cover most of the
A,ffO range of the variables; and the normalised axes cause
fi=s data with the same value of I@ to group together.
a The data of Figs l-6 derive from the work of many
research groups (listed on the figures). But the range
Here A, is the nominal (apparent) contact area of the
of load and velocity studied by a given investigator
wearing surface, H,, is its room-temperature hardness,
and his group is usually small, and covers only a
a is the thermal diffusivity and r, is the radius of the
minor part of the figure. The areas studied by a
circular nominal contact area. @ is the volume lost
number of leading groups are shown in Fig. 7. The
per unit area of surface, per unit distance slid; F is the
overlap is small. Often, within one of these areas, a
(nominal) pressure divided by the surface hardness;
single mechanism of wear is dominant; certainly each
and fi can be thought of as the sliding velocity divided
group has its own preferred mechanism, and has,
by the velocity of heat flow.
quite naturally, sought the experimental conditions
All the diagrams in this paper are plotted using F
needed to produce it. The first, crude, indication of
and u’ as axes, with contours (where appropriate)
how mechanisms interrelate can be found (therefore)
which show the value of @. The range of the axes is
by marking the boundaries of the field studied by
large and includes the sliding conditions of almost all
a group, and for which their model was developed.
published work: the normalised pressure F ranges
Such a map of mechanism-schools is shown in Fig. 8.
from 10e5 to 10; and the normalised velocity v^ranges
It shows four main areas:
from 10m2 to 10’. The diagrams show both the
normalised velocity B and the actual sliding velocity (a) seizure;
U, in m/s. (b) melt-dominated wear:
(c) oxidation-dominated wear (mild- and
severe-oxidational wear); and
3. EMPIRICAL WEAR-MECHANISM MAPS

SLIDING VELOCITY v (m/s)
?1
10-L lo-2 1 102
We first examine the data for wear rate and
mechanism without imposing any model-based ideas
on them. To do this we have assembled, from the
open literature, wear rate data and observation of
wear mechanisms for six steels; and we have aug-
mented these with measurements of our own. The
primary data are from dry (that is unlubricated)
pin-on-disk experiments, though data from other
tests (pin-on-ring or cylinder, cylinder-on-cylinder,
etc.) do not, in general, differ much, and can often
usefully be included. Experiments with external heat
input were exluded; so for the most part, were
experiments in environments other than air. We
frequently found that the references did not contain
all the information needed to make the diagrams;
then we had to make reasonable guesses (based, for
instance, on photos of the worn surfaces) to estimate I O-61
10-&
47. .-PC O_n
I
them. 1 102 loL
Data for steels were classified into six subgroups, NORMALISED VELOCITY S
and each is plotted separately. The subgroups are: Fig. 1. The wear-rate data-map for low carbon steel. Data
for pure iron are also ir$uded. The numbers given against
(a) pure iron and low carbon (C < 0.3 wt%) the points are log,,(W). (The references. in order, are
steels; (4201.1
SLIDING VELOCITY v (m/s) SLIDING VELOCITY v (m/s)

.L lo-* 1 IO2
L
‘Q;o-4 1
10"
NORMALISED VELOCITY i? NORMAL&ED VELOCITY v

Fig. 2. The wear-rate data-map for medium carbon steel. Fig. 4. The wear-rate data-map for low alloy steel. The
The numbers given against the points are log,,(@). (The numbers given against the points are log,,,(m). (The refer-
references, in order, are [21,22,6,23-28, 1829,301.) ences, in order, are [15, 10, 26,8,29,32-36, 23, 371.)
(d) plasticity-dominated wear (including ipating the most obvious results of the modelling, we
delamination wear). can sketch them in. Figure 9 shows the results for
steel. Seizure will occur (crudely) when the pin plas-
This may seem more like political than scientific tically indents the disk; that occurs when P = 1. Melt
reasoning. But that is not true. Within an area, wear will occur when the power dissipated against
detailed studies characterise the mechanism (whole friction, pFv/A, per unit area, exceeds a critical
books have been written about some of them-see, value; that defines a triangular field in the top right.
for example, Suh et al. [45]); it is only the boundaries Oxidation starts only when asperity heating exceeds,
between the mechanisms which are uncertain. The say, 400°C; that (we show in the next section) occurs
map suggests where these boundaries lie; and, antic- at a near-vertical boundary, roughly in the middle of

10-Z 1 102
NORMALISED VELOCITY 5 NORMALISED VELOCITY F

Fig. 3. The wear-rate data-map for high carbon steel. The Fig. 5. The wear-rate data-map for high alJoy steel. The
numbers given against the points are log,,(W). (The refer- numbers given against the points are log,,(W). (The refer-
ences, in order, are [7,6,26,3 1, 161.) ences, in order, are [12,23,38,39,4].)
LIM and ASHBY: OVERVIEW NO. 55 5

2
,$O" 10-z
NORMALISED VELOCITY q
NORMALISED VELOCITY v
Fig. 6. The wear-rate data-map for tool steel. The numbers Fig. 8. The mechanism-schools
given against the points are log,,(@). (The references, in
order, are [37,12,40,38].)
calibrated against experiment. The next two sections
describe the extent to which we have been able to
the diagram. And severe-oxidation obviously occurs achieve this.
at a higher velocity than mild-oxidation. That gives
the Jie[d boundaries. Superimposed on them are con -
4. THE TEMPERATURE OF SLIDING
tours of constant normalised wear rate, & They are SURFACES
sketched in by interpolation through the data plotted
in Figs l-6. When two surfaces slide together, most of the work
One could stop here, but it is challenging to done against friction is turned into heat. The resulting
explore the degree to which a diagram like Fig. 9 rise in temperature may modify the mechanical and
might be constructed from theoretical considerations metallurgical properties of the sliding surfaces, and it
alone, or (what is far more profitable) from theory may make them oxidise or even melt; all these things
influence the rate of wear. The frictional heat is really
generated at the tiny contact areas (“asperities”)
which make up the true area of contact at the sliding
16‘ to+ 1 102 interface. The instantaneous temperature of these
contact points (the “local” of “flash” temperature,
Tr) is obviously higher than the average (or “bulk”)
temperature Tb of the surface, sometimes much larger
[30, 46601.
In what follows, the important results for flash and
bulk heating are assembled and combined to give a
diagram showing the surface temperature as a func-
tion of the normalised pressure F and velocity d. In
doing this, one has to assume a geometry for the
sliding surfaces; the pin-on-disk configuration has
been chosen for the reasons given earlier. (Other
geometries lead to diagrams that look only slightly
different.) We have simplified the calculations as far
as possible, replacing three-dimensional patterns of
heat flow by simpler one-dimensional equivalents,
while attempting to keep a consistent level of accu-
racv.
4.1. Bulk heating
Fig, 7. The author-map. (The references, in order, are
[30, 12,4&42, 10,4,13,8,20, 14,9,7,5,21,43,44,33,34,36, When a pin slides on a disk (Fig. IO), frictional
35,23,29,24].) heat is generated at the interface. The heat generated
LIM and ASHBY: OVERVIEW NO. 55
JAxls of sliding velocity Idimension=m/sl

1 IO2
1
Fields identifying S_
range of dominance b SEIZURE PIN-ON-DISK
of mechanism
Contoursof constant
normolised weor
2
Axis of /
normalised
5
n
Rate 6t
ldimensionlessl
pressure f y”
[dimensionless1
tH
10-2 1 102 10‘

NORMALISED VELOfITY ‘j
AXIS of normalised velocity Idimensionless]
Fig. 9. The empirical wear-mechanism map for steel using the pin-on-disk configuration.
per unit area per second (units: J/m2s) is of the total. If the steady-state temperature-
distribution were established in both pin and disk,
PFV
(3=7 (3) frictional heat would be. divided equally between
n them, and then a = f. But a point on the disk surface
where p is the coefficient of friction, F is the normal has heat injected into it only for the interaction time
force on the pin, Dis the sliding velocity and A, is the t,, = 2r,/v (the time taken for the point to slide across
nominal contact area (the area of the end of the the diameter of the pin), whereas the pin has heat
pin). A fraction a of the heat diffuses into the pin; the injected into it all the time. The steady state is
rest goes into the disk. Some of the heat entering the established in the pin, but it may not be established
pin diffuses to its surfaces where it is lost by radiation; in the disk: that depends on whether the interaction
the rest diffuses up the body of the pin where it is time td is bigger or smaller than the heat diffusion
absorbed by the heat sink formed by the clamping time, t,, = r,/a. It is convenient to define the constant
mechanism. Let the mean diffusion distance? be lb. 9 on the ratio of these two terms [46]
After an initial transient, the temperature in the
pin will settle down at the steady-state distribution @2!25.
td 2a
which can be calculated from the first law of heat
flow. Linearising the problem, the heat flux (units: When this is less than 1, a steady state is established
J/m’s) is in the disk and a = l/2; when not, a transient tem-
perature field exists in the disk, and a is less than l/2.
Heat sink (clamping mechanism I.

where K,,, is the thermal conductivity, Tb is the bulk sink temperature = To
surface temperature and To is the sink temperature -r
(which we take to be 20°C). Equating this to the
incident heat flux crq, we obtain [using equation (3)]
T,, = T, + AK. (4)

n m
Heat flows into the disk as well-a fraction (1 - a)
tlb is the “equivalent linear diffusion distance”. The heat Heat input-(l-a)q j/m2s
flow is not, in general, linear; but for dimensional
reasons the temperature must have the form given by
equation (4). We choose I, to have the value which DISK
makes equation (4) equal to the approximate solution to Fig. 10. The bulk temperature generation at the pin/disk
the full, three-dimensional, problem. interface.
LIM and ASHBY: OVERVIEW NO. 55 I
We calculate a from the time-dependent equations SLIDING VELOCITY v (m/s)

-& 1
for heat flow [61]. An incident heat flux (1 - cr)q IO'
1p-2
II
192
produces a surface temperature 0 JcJImScn

et01119L71 STEEL
b run”* 01119801 CCEFFICIENT
OFFRICTION
p
T = T + 2(1 - ah(wP
b 0
k,(n)“* .
Continuity requires that the bulk temperature at the
contacting interface is the same for the pin and the
disk. Equating equation (4) to (5) and solving for a
gives
1
a = 1 + I,(nu /8ar,)“*’
This result reduces to a = 1 when u = 0 because we
have neglected the transition to a steady pattern of
heat flow in the disk when the Jaeger constant 0 is
less than 1. We fix this by writing
1
(6)
a = 2 + &(zv/8ar,,)“*
NORMALISED VELOCITY V
giving an equation for the heat distribution coefficient, Fig. 11. The variations in coefficient of friction with nor-
a, which now has the correct limits. Though approx- malised pressure and normalised velocity for steel. Contours
imate, it is more than adequate for our purposes. of constant p are superimposed on the data points at
At this stage it is helpful to note that the quantity velocities greater than 1 m/s. (The references, in order, are
[62,29, 63, 33,42, 21,64, 5, IO].)
(aHO/&) has the dimension of temperature, and that
1, and rO are both lengths and roughly the same
magnitude. We define
4.1.1. Calibration of the bulk heating equation.
T’=g!Z (74 Figure 13 shows the temperature contour for 1527°C
m
(the melting point for pure iron) obtained using
(for steels, T* z 222K), and equations (8) and (9) and the material properties
given in Table 2 (with /3 = 6). The contour is trun-
Ub) cated at E = 1, roughly corresponding to the seizure
load. Superimposed on the contour are two sets of
(where /3 is a dimensionless number of general data obtained from high velocity sliding experiments
order 1).
The bulk heating of the pin is given by equation (4)
with the value of a given by equation (6). Rear- SLIDING VELOCITY v (m/s)
ranging the result in terms of the dimensionless 16‘ 10-Z 1 102
variables P and r? gives the equation for the bulk
surface temperature :
PT*B
T,, = T,.,+ Fl7.
2 + /II(rru’i8)“’
To calculate it, we need the coefficient of friction,
p. It is often said that p is a constant, but over the
wide range of load and velocity of the diagrams
shown here it is not. Figure 11 shows data for p for
steel rubbing on dry steel, plotted on axes of P and
d. approximate contours of constant p are shown,
drawn by interpolating between the data points. The
thing to note is that g depends on v’ but, except at
the highest velocities, is independent of i? The de-
pendence on 3 is best seen in Fig. 12; it is adequately
described by
p = 0.78 - O.l31og,,(B). (9)
It varied from about 1 at very low velocity to about Fig. 12. The variations in coefficient of friction with sliding
0.1 at very high. This is used in equation (8) to velocity for steel. (The references, in order, are
calculate bulk temperatures. [62,29, 63,33,42, 21,64, 5, IO].)
SLIDING VELOCITY v (m/s) asperities touch, and the heat input per unit area at
an asperity, q’, is larger than that given by equation
(3) by the factor A,/A,. In units of J/m2s, we have
Some authors (e.g. Tabor [65]) have supposed that

the contacting area of an asperity, nrz (where r, is its
radius) increases with load. Others [6668] argue that
the average area of an asperity contact is constant
(with a value of ra between about lo-‘rn and 10-6m)
and that it is their number, not their size, which
increases with load; and indirect experimental evi-
dence [69,70] supports this view. In the following
calculations we have taken the area of the “unit
asperity” to be constant and allowed the real area of
contact to change by changing the number N of
contacting points.
This number enters explicitly in the equation for
NORMALISED VELOCITY “v
the flash temperature, so we need an expression for
Fig. 13. Calibration of the bulk temperature. (The refer- it. When the load is light, N must be small; in the limit
ences, in order, are [21, 51.) of vanishingly small loads, it approaches the value 1.
At the other extreme, when the load approaches the
on steels? [5,21] where large scale melting is reported. seizure load, the “unit” contacts overlap and the true
The contour shows broad agreement with the data number of contacts again becomes small: in the limit
points. of complete seizure it again goes to 1. The problem
is analysed in Appendix 1, where we show that the
4.2. Flash heating true number of contacting points is adequately ap-
Almost always, the real area of contact between the proximated by
sliding surfaces, A,, is much less than the nominal
contact area, A,. Real contact is made only where N= : 2&1-F)+1 (11)
0 d
tThe carbon content of the steels was between 0.1 and where ra is the radius of an isolated “unit” contact
0.52%C, giving melting points between 1500 and (typically 10m5m) and r, is the pin radius. The
1520°C. equation has the proper limits and (as pointed out in
Table 2. Material urooerties

Property Symbol Value
Thermal conductivity, (steel) K (J/msK) 41 ‘a’
Thermal diffusivity, (steel) a @Is) 9.1 x 10-e Cd’
Density, (steel) P (kg/m’) 7.8 x 10’
Volume thermal capacity, (steel) PC(J/m3K) 4.5 x 106 @’
Melting point, (steel) T, (K) I800
Latent heat of melting, (steel) L (J/m) 2.1 x 109 @I
Temperature T* for steel a&I&, (K) % 222 @)
Temperature T: for steel a&i& (K) z 650 lb)
Molecular weight of iron & (kg/KmoU 56
Martensite temperature (A,) T,(K) 1023-I 173
Hardness at 20°C. (steel) H, @N/m*) ,w
Thermal conductivity, (oxide) K., (J/msK) 3.2 (”
Thermal diffusivity, (oxide) a,, @is) 4.5 x 10-G W’
Density, (oxide) P.. &s/m’) 5.2 x 10’ w
Volume thermal capacity, (oxide) (PC+U~~) 7.2 x IO5 w
Melting point, (oxide) 1867 W’
Latent heat of melting, (oxide) L,,m(J/m’) 3.1 x 109 w
Temperature Tzx for oxide a,, &,I&, (K) z 2800 (‘I
Molecular weight of oxygen MO,(kg/KmoU
Activation energy for oxidation Q, (kJ/mol) 1,“,‘b,
Arrhenius constant for oxidation A. lke2/m4s~ 106(b)
‘“‘Data for iron from the Handbook of Chemistry and Physics [146].
““Present work.
@‘M@lgaard and Smeltzer (1451.
‘d’Americon Institure of Physics Handbook [I471
“‘Quinn ef al. [104].
‘“Using Ho, = 2 GN/m’.
the sink temperature T’, is well approximated by the

JF bulk surface temperature Tb (Fig. 14). (The excep-
PIN
tional circumstance when this is not so is analysed in
-V
Sink temperature=T~ aTb Appendix 3.) Further, as we shall see in a moment,
Heat input q + q J/m% in the regime in which the flash temperature plays an
/
important role in wear, the bulk temperature is
low-close to T,,. This allows a further simplification:
there is now no distinction between an asperity on the
pin and one on the disk, and the heat partition factor
c( reduces to l/2.
To get any further we need a relationship between
real contact area and the load. This is a problem
which has been exhaustively discussed in the litera-
ture (e.g., Archard [30]; Bowden and Tabor [73, 741).
Heat input = $q J/m’s saw= zr,
(Total area ol When the contacts are plastic (as we shall assume
ospenty contours=A,l
them to be), equilibrium requires that
(15)
Fig. 14. Flash temperature generation at an asperity contact.
We need, too, an expression for ra/ro. If there are N
asperity contacts per unit area, then
the Appendix) it is in good agreement with the counts
Nnrz A,
of contacts made by Quinn and Winer [70]. --F=-
Consider now heat flow into an asperity (Fig. 14). nr, A,
As an asperity on the pin sweeps over one on the disk, so that
heat is generated, and this heat flows into the two r / b\ I!?
r, 1
surfaces through the circles of contact. The contact

time is
-=-
r. 0N (16)
2r. Finally, if oxide forms on the asperity tips (as it does

t=4 (12) if the flash temperature is appreciable), the effective
V
thermal conductivity of the asperity is altered. The
where r, is the asperity contact radius and v is the T* should be replaced by an effective equivalent
sliding velocity. The local surface temperature at the temperature T,* , which is calculated in Appendix 4.
contact rises, and reaches a peak value, Tr; then the Assembling these results gives the steady-stare ,flash
contacts separate and the local temperature falls temperature
again. The appropriate choice of heat flow equation
depends (Archard 1301, Lim 1711) on whether the (17)
contact time r is large or small compared to the
characteristic time constant for heat flow which (for with Tb given by equation (8) p by equation (9) and
flash heating) is N by equation (11).
L
t,=5.
a
4.2.1. Calibration of the flash heating equation.
The choice is discussed in Appendix 2. The full The equation of flash heating [equation (17)] was
solution (Carslaw and Jaeger [61], Bass [72]) con- evaluated and is plotted in Fig. 15, using the data
tains both limits. But it only makes sense to calculate in Table 2 (together with /J = 6, r, = 10eSm and
r, from heat-flow considerations (ignoring melting) 2 = 10d5m, as before). The axes are the usual ones:
when it is less than the melting point. Then one limit Fand 8. Three flash temperature contours are plotted
is appropriate (Appendix 2) and the equation simpli- corresponding to 750, 900 and 1527°C (the melting
fies to point of pure iron). Superimposed on the contours
are data derived from wear tests in which martensite
was first detected on the sliding surfaces (references
are given on the figure). In these tests, the steel
where /3 is a numerical constant near unity, which is specimens had carbons between 0.12 and 0.78 wt%.
essentially the ratio of the heat-diffusion distance to The minimum instantaneous peak temperature re-
the asperity radius (If/r,), and (as before) tl is the quired to bring them into the fully austenitic field lies
fraction of the heat which flows into the pin, and TL between 750 and 900°C; the subsequent cooling is
is the “sink temperature” for heat flowing away from always fast enough to form martensite [28]. The data
the asperity. Almost always, the asperity size, ra, is points fall between, or close to, the contours for those
much less than the nominal contact radius, r,; then two temperatures.

lo4 IQ” 1 lp’
10 ’
STEEL
TEHPERAT”RE
MAP
PIN-OCDISK
CONFlG”RATlON
NORMALISED VELOCITY v NORMALISED VELOCITY V

Fig. 15. Martensite calibration of the flash temperature. Fig. 16. The temperature map.
(The references, in order, are [34,75,26,23, 24, 15,301.)
at 7 f 3-a number which agrees well with the equa-

A second independent check of the flash tem- tion for N [equation (1 l)].
perature is given by the observations of Quinn and
Winer [70], who slid tool steel pins on a sapphire disk 4.3. The temperature map
with values of P between 1.2 x 1O-4 and 6.6 x 10e4, The results of this section are summarised in
and at sliding velocities between 1.7 and 2.6 m/s (so Fig. 16. It shows, on the usual axes (P and v’),
the data lie almost exactly on top of those plotted in contours of constant bulk (Tb) and constant flash (T,)
Fig. 15). Photographs of the hot spots allowed the temperature. They are computed from the equations
asperity (flash) temperature to be estimated from an derived in the text and summarised in Table 3. In
approximate colour-temperature calibration. Most of simplifying the heat-flow calculations we have taken
the hot spots, they found, were at about 950°C with considerable care to retain the physically-important
one or more in almost every photograph ranging to features of the problems; and each equation has been
1100 or 1200°C. They also found that, when the checked and calibrated against experimental data.
exposure time was reduced below 33 ms, the number We are, therefore, confident that the map gives a
of hot spots per photograph became roughly constant reasonably accurate picture of the temperatures that
Table 3. The simulified temuerature equations

Bulk temperature T,
PT’B Fc
T, = To f equation (8)
2 + /7 (xt7/8)“*
Flash temperature Tf
Tf= T,[I -(;>‘“]+ To(;>‘“+~(;)“‘i equation(AS)
With the followingdefinitions

p=&
n 0
equation(2)
fi=u’”
0
T’ = $ E 222 K for steel equation(7a)

m
T: = ? e 650 K for steel equation(A7)

e
p+; equation (7b)
/I =0.78-0.131og,,(fi) equation (9)
N=
0 2P(1
:
I..= 1.5mm
d
-f)+, equation (1 I)
appear in a steel pin sliding on a steel disk during dry a shear stress s = pF/A, (because of friction) in
wear experiments. And because the normalised axes addition to the normal pressure F/A,, and the junc-
contain some of the geometric and thermal variables tions grow further until
of the problem, the map gives an approximate picture F2
of surface temperatures during dry sliding of steel
surfaces of other geometries. It forms the basis on
which the next section of this paper is built.
0 z
r
+ u,s2 = HZ.
This is Tabor’s equation [65]; he and others (e.g.

(19)
Johnson [80]) find that a value of the constant LX,

= 12
5. WEAR MECHANISMS fits experimental data well; we shall use it here.
Replacing s in the last equation by pF/A, gives a
The second step in constructing wear maps by modified version of equation (19) which now includes
physical modelling is that of identifying and de- the effect of frictional shear
scribing the individual mechanisms that cause loss of
material from the sliding surfaces. We consider four (20)
broad classes of mechanism:
The surfaces seize when the real contact area becomes
(I) seizure;
equal to the nominal one, that is, when A, = A,. This
(2) melt wear,
gives the approximate seizure condition
(3) oxidation-dominated wear, and
(4) plasticity-dominated wear, 1 H
(21)
F = (1 + cc,$)“2H,
For each mechanism, we survey and develop models
and calibrate them against wear-rate data, drawing where F and p are given by equations (2) and (9) as
heavily on published work of the many authorities on before.
wear. As emphasized earlier, each model is reduced For most purposes it is adequate (see Appendix 5)
to the simplest level which still retains the essential to set H = HO in this equation, giving a very simple
physics. These simplified and calibrated models are expression for the seizure line. We have used a
then assembled to give a wear-mechanism map for slightly better approximation. The local hardness
steel. depends not only on temperature but on strain rate
5.1. Seizure
When metal surfaces are placed in contact, the real
area over which they touch, A,, is usually very small.
I
PIN
The large local pressures at the points of real contact
(the asperity contacts) can forge metallic junctions
\
between the surfaces even under static conditions
[76]. When the junctions are sheared the welded areas
grow; under large enough loads, the real area of
contact grows until it is equal to the nominal area,
A,, and the surfaces seize completely [65]. Effects
such as plastic indentation [77], large-scale mass flow
[78] and metallic transfer [79] have been used as the I I
basis of seizure models, but these seem just to be I
I
I I
another way of saying that shear causes catastrophic Total area
junction growth, a condition analysed by Johnson Of ospenty
contact .A PLAN VIEW OF
[80] and Collins [81] who provided a physical expla- lb1 ’
IDEALISED CONTACT
nation for Tabor’s equation [65] for seizure [equation

(19) discussed below].
5.1. I. The model for seizure. When two surfaces

touch, the localised pressure on the asperity contacts
causes plastic flow. Under static conditions, the con-
tacting regions grow until the real area of contact
ICI J
A, is large enough to support the load [Fig. 17(a)].
Fig. 17. An idealised model for the seizure mechanism. (a)
Then the mean pressure on an asperity, F/A,, is The asperity contacts deform plastically, under the normal
simply equal to the local hardness, H, of the material load F, until the real area of contact, A,, is given by F/H,,.
(which may be hot) (b) Further increase in the normal load F and the presence
of a tangential traction (u # 0) will increase the number of
asperity contacts, N, as well as the size of each contact, ra.
f=H. (18) The real area of contact, A,, has increased. (c) Seizure takes
r
place when the real area of contact, A,, equals the nominal
If the surfaces now slide, each asperity is subjected to area of contact, A,.
(10-1000 m/s) the coefficient of friction drops to very

low values because a film of liquid metal forms at the
sliding interface giving “melt lubrication” [63, 87-901.
The liquid layer supports the normal load by the
usual hydrodynamic mechanism [91], but the heat
generated by viscous work in the layer continues to
melt more solid, so the wear rate is high even though
the coefficient of friction is low [21,92]. The metal
removed from the surfaces is ejected as sparks or hot
incandescent particles [64], or squirted out as a
molten stream [5].
5.2.1. The modelfor melt wear. It is well known that
when solids slide on ice or snow, friction-induced
melting generates a lubricating layer of water in just
this way. We adapt the analysis of this problem due
to Evans et al. [93] to the problem of melt wear of
steel. Consider the pin-on-disk geometry in Fig. 19:
part of the heat is conducted into the pin and the
disk, and part is absorbed as latent heat L (J/m3) in
NORMALISED VELOCITY 7
melting a volume V,,,(m3) per second at the interface.
Fig. 18. Calibration of the seizure line. The broken line is Then the heat flow equation of Section 4.1 is modified
described by equation (21) and the solid line by equation to
(22). The lower limit of the normalised-velocity axis has
been extended to accommodate data points from slow-
sliding tests. (The references, in order, are [82,44,5,43, aq= --K,VT+L> (23)
21, 83, 841.) n
where aq is the rate of frictional heat input to the pin,
- K,VT is the rate of heat conduction into the pin
too. As the sliding velocity increases, the bulk tem-
and LV,JA, is the rate of heat absorption by the
perature, T,,, increases, causing the hardness to drop.
melting material. Substituting for q from equation (3)
But a high sliding velocity, v, means a high local
and linearising as in Section 4.1 gives
strain rate, and this raises the hardness. It turns out
that, for steel, the two effects roughly cancel. In W’v Km(Tm-T”)+LV,.
-= (24)
Appendix 5 a seizure equation is derived which 4 4 An
includes. through the simplest kinetic model, these
If all the metal which melts is ejected, the normalised
effects of temperature and strain rate. The result is
>I(22)
wear rate, IV, is simply
Da
1 1 _vr cJIn ‘0”
F = (1 + “,#‘2 2OT,,, (
where Tb is the bulk temperature [equation (8)] and Rearranging equation (24) then gives the equation for
T,,, is the melting point. The seizure line on sub- melt wear
sequent figures is calculated from this equation.
5.1.2. Calibration of the seizure model. Figure 18

shows the seizure line, plotted on the usual axes (P
I’
and a), using data from Table 2, with CI,= 12 and p Sink temperature = To
given by equation (9). Data from a number of high-
load tests on carbon steels (mild steel to 0.95 wt% C) PIN
are plotted? and identified on the diagram. These are
the highest values of i: that we can find; it can be seen
that the seizure line lies, roughly, along the upper
edge of the data, as it should.
5.2. Melt wear

Localised melting can occur even at sliding veloci-
ties as modest as 1 m/s [85,86]. At higher velocities ‘A layer of
molten metal
DISK
tWhen data are plotted, they are normal&d by the actual Fig. 19. An ideahsed melt wear model. A layer of molten
(quoted) hardness of the steel, or an estimate of it based metal is formed at the pin/disk interface when the bulk
on carbon content, not by the “typical” value of HO temperature generated there reaches the melting point of the
given in Table 2. metal.
SLIDING VELOCITY v (m/s) perhaps, surprising: liquid metals have low visco-
sities, and (during melt wear) the pressures and
sliding velocities are enormous. So the liquid is
efficiently moved to the exit surface where it is easily
squirted or thrown out. Oxides (which we come to
next) have higher melting points and much higher
viscosities; they do not flow nearly so easily, and
transfer and ejection of molten oxide may require
much higher temperatures. This, perhaps, is why
there is a sudden drop in wear rate when stable oxides
form; the drop appears as a discontinuity in the wear
contours when they cross the field boundary from
the oxidational wear regime to the melt wear regime.
This field boundary is defined conveniently by the
bulk temperature contour for the melting point of
steel.
5.3. Oxidation-dominated wear

When steel surfaces slide at speeds below about
I m/s, the wear debris is largely metallic; above, it
NORMALISED VELOCITY “v
consists mainly of iron oxides [94-971. A velocity of
Fig. 20. Calibration of the melt wear model. The numbers 1 m/s is just sufficient to give flash temperatures
given against the points are log,,(p). (The references, in which will cause oxidation. The temperature map
order, are [21, 51.)
(Fig. 16) shows that the flash temperature depends
strongly on velocity, but hardly at all on load; at
with LX,p and T* given by equations (6), (9) and (7a), 1 m/s the flash temperature is near 700°C.
as before. A drop in wear rate is associated with the change
The temperature map (Fig. 16) shows that there is to oxide-debris [I 5, 23,26, 30, 31,751 though oxi-
a range of P and v’ in which asperities may melt, but dation alone may not explain its magnitude: part is
the entire surface does not. In this regime it is probably caused by martensite formation (Fig. 15).
obviously wrong to equate the wear rate to the But it is established that an oxide film forms on
volume of metal which has melted because as soon as sliding surfaces [9&100] and that above a critical
the asperity melts it flows, smearing over a cooler part velocity, wear is largely caused by the splitting off
of the surface where it resolidifies. Further, the of this film [loll. Around 1 m/s and at low loads. the
temperatures are high so that, in all but the most inert oxide film is thin, patchy and brittle. At higher
atmospheres, oxidation will compete with local melt- velocities (> 10 m/s) and higher loads the film
ing as the dominant wear mechanism. The experi- becomes thicker and continuous, covering the entire
mental evidence suggest oxidation wins. So we shall surface [9, 14,641. Under these conditions the fric-
postpone discussion of this regime to the next sub- tional heating is considerable; the metal is partly
section, in which oxidation is analysed. insulated by the overlying oxide [9] but the oxide itself
5.2.2. Calibration of the melt wear model. Figure 20 is hot enough that it can flow plastically or even melt.
shows contours of constant normalised wear rate, We therefore distinguish two regimes, both im-
@, calculated from equation (25) using the data portant, and both discussed below. Both are associ-
in Table 2, and plotted on the usual axes. They are ated with the range of sliding velocities in which
compared with data for bulk-melt wear of Mont- surface heating stimulates oxidation (this means
gomery [21] and of Kinsella and Childs [5]. There are o>-lm/sorfi>-lO*).Atthebottomendofthis
no adjustable parameters. Agreement is good. No range Aash temperatures are enough to cause local
special correction is needed to fit the data of Kinsella oxidation, but the oxide is, for most of the time, cold
and Childs [5] despite the fact that they used a and brittle. This is the first regime, which we will call
“disk-edge on block” not a “pin-on-disk” configur- mild-oxidational wear. At higher velocities, the oxide
ation: it is the contact dimension (r,) and the heat- is thicker, more continuous, and almost certainly
flow length (lb) which really matters, not the detailed hotter and more plastic, and the characteristics of the
geometry of the equipment. This adds to the body of wear process change. This is the second regime, which
evidence (of which more later) that the diagrams we will call severe-oxidational wear. The prefixes
broadly summarise wear for most unidirectional slid- “mild” and “severe” refer to the extent of oxidation,
ing configurations. not the wear rate (which is often lower during
One further point should be mentioned here. It severe-oxidational wear).
concerns an assumption of the model: that all the 5.3. I. The model of mild-oxidational weur. Over
metal that melted was lost; and the agreement with the last 25 years, mild-oxidational wear has been the
experiments suggests that this is indeed so. It is not, subject of intensive study. Much of the development
originates in the work of Quinn [102] who marshalls

IF
convincing evidence that, in this regime, flash heating PIN
causes oxidation at contact points; that the oxide -v
thickens until, at a critical thickness (about 10 pm for

steel) it spalls off; and that these oxide-flakes are the
wear debris. (4
Attempts to model mild-oxidational wear were
made as early as 1962 [102]. The model was sub-
to high flash temperature T,
sequently developed and tuned [29,35, 101, 103,-1061
and an iterative technique was developed to deter- DISK
mine, from experimental wear rates, the values of the

parameters which appear in it (e.g. Quinn et al. [106]). b
PIN
The result we need can most easily be derived as -v
follows.
The oxidation of iron has parabolic kinetics [107].t
The mass of oxygen taken up by the oxide film per
unit area, Am, is
Am= = k,t (26) Thinner oxide films critical thickness ZC
ond breaks off
where kP is the parabolic rate constant and
1
DISK
k,= A,exp -3 (27)

[ RT ’ IF
PIN
Here A,, is the Arrhenius constant, Q, is the activation -v
energy for oxidation, R is the gas constant and T is

the absolute temperature. Consider now a sliding
surface with asperities which become hot enough to
oxidise. The oxide thickens and, when it reaches a
critical thickness Z,, it spalls off giving wear (Fig. 21).
and breaks off
It is adequate to assume that the average composition
of the oxide is Fe,O,, so that 1 mol of Fe requires exposed metallic surface.
DISK
2/3 mol of O2 to form the oxide. If a volume AV,, of
iron is oxidised per unit area, the mass gained as a Fig. 21. An idealised mild-oxidational wear model. (a) The
result of the formation of oxide is oxide films formed at the asperity contacts (with a total area
of A,) grow according to a parabolic rate. (b) The critical
oxide thickness Z, is reached at one of the asperity contacts
and the oxide layer breaks off there. (c) The broken off
where pre is the density of iron, Mo, the molecular oxide layer becomes a wear fragment and an oxide layer is
weight of oxygen and M, that of iron. Substituting again formed on the newly exposed metallic surface; while
the oxide film on another asperity contact has reached the
this into equation (26), and noting that the thickness critical thickness Z, and has broken off.
Z of the oxide is simply equal to A VFe, gives
z2 = C=k,t factor [equation (27)] with T = T,, giving
with (29) Zf
(30)
tc=C2A,exp[-Q,/RTfJ’
Wear is caused by this oxide breaking free, so that a
volume A,Z, is lost every t, seconds from the surface.
(using the data in Table 2 gives C = 3.4 x 10e4 During this time, the surfaces have slid through a
m3/kg). distance vt,, so the wear rate is
RT,
1
In the regime of mild-oxidational wear, the bulk A,Z, AC=A
heating is negligible (see Figs 8 and 16). Oxidation is w= =cexp -2%
caused by flash heating and must be confined to the vtc vZc [
asperity tips. The time taken to reach a critical oxide or in dimensionless variables (and using P = AJA,)
thickness Z, at an asperity is obtained by inverting
the last equation and replacing k, by the kinetic
lV=(F)exp[-&]% (31)
tYoshimoto and Tsukizoe [lOS] and Tenwick and Earles

[109] assume linear kinetics in their wear models; but for with Tf given by equation (17).
iron and steel’ the evidence is that oxidation is parabolic This is not exactly the same as Quinn’s equation
(e.g. Kubaschewski and Hopkins [107]). [106] but it is so close that, from the point of view of
this paper, they can be considered identical. A term by giving two values of wear rate for each contour.
(r,/ZJ appears in Quinn’s equation but not in equa- The first are those given by equation (31) with
tion (31); it is due to his use of Archard’s law which H, = 1 GPa; the second (in parentheses) are 100 times
is based on Hertzian contact at the asperities. But smaller, reflecting the increase in H, caused by mar-
since Z,. N ra, the value of the extra term is always tensite formation.
close to unity. With this degree of complexity, and with the use of
5.3.2. Calibration of the mild-oxidational wear A, as an adjustable parameter, one cannot claim the
equation. Everyone who tries to fit equation (31) (or equation (31) is a complete description of mild-
others like it) to wear data encounters the same oxidational wear. But the physical basis of Quinn’s
problem: the kinetic constant A, and Q0 inferred from model (on which the equation is based) has consider-
wear data do not agree with those measured in able experimental support. The equation, though
standard, static, oxidation experiments (see for exam- imperfect, is the best we have got. We will proceed,
ple, Quinn et al. [106]). using equation (31) for mild-oxidational wear.
Perhaps this is not surprising. The wear-oxide films 5.3.3. Severe-oxidational wear. When the sliding
are thin; and they have grown under conditions of velocity is greater than about lOm/s, surface oxi-
severe mechanical loading. In using equation (31) A, dation becomes more severe, though the rate of wear
and Q, have to be treated as adjustable parameters, may actually decrease. This regime has not been
chosen to fit measurements of wear rate. The activa- studied in as much detail as the last; such evidence as
tion energy Q, is not likely (according to Melgaard is available suggests that oxidation is general and that
and Srivastava [l lo]) to be much influenced by the the temperatures are high enough for the oxide to
way the oxide grew, so we set Q, equal to that for become plastic and melt locally to a viscous liquid,
the static oxidation of iron (138 kJ/mol [107]). The though the underlying metal does not melt. We
Arrhenius constant, A,,, on the other hand, shows an cannot find any model describing this form of oxi-
immense spread. Even in static tests, the range is large dational wear. An attempt is made below to develop
(10e2 to IO6kg2/m4s); under sliding conditions it one.
becomes enormous (lo3 to lOI kg2/m4s [106]). Our 5.3.4. The model for severe-oxidational wear.
goal here is a broad survey of wear mechanisms, so Higher velocities generate higher temperatures
we thought it justified to use the single value for A,, (Fig. 16) with two consequences. The extent of
(IO6 kg2/m4s) which gives the best general agreement oxidation is obviously greater; surface observations
between experimental and predicted wear rates in the in this regime show continuous, thicker, oxide
mild-oxidational wear regime. [9,14,64]. And the points of contact between the
The degree of success can be judged from Fig. 22. oxide-coated surfaces are so hot that the oxide there
It shows contours of constant normalised wear rate. must, locally, melt and flow (see Figs 8 and 16).
@‘,calculated from equation (3 l), using equation (17)
for the flash temperature, using data which are SLIDING VELOCITY v (m/s)
summarised in Table 2 and with C = 3.4 x 1O-4
m’/kg and Z, = 10 pm. Superimposed on the con-
tours are data points (for a wide range of steels, tested
at velocities smaller than 5 m/s) from Quinn and his
co-workers and from Archard and Hirst [12].
Behaviour in this regime is complicated by mar-
tensite formation. When a hot asperity loses contact,
it is quenched by conduction of heat into the under-
lying bulk which (as Fig. 16 shows) is still at room
temperature. The quench rate is high+asily high
enough to form martensite if the asperity reached the
A, temperature of the steel. The data plotted in
Fig. 22 shows a sudden drop in wear rate at higher
values of P and d; this, it is generally thought, is due
to martensite (which is observed to form just here-
see Fig. 15). When martensite forms the surface is
suddenly harder (H, increases) and in compression
(because the volume increases). A harder surface
means a smaller value of r’ = F/A, H,, and thus NORMALISED VELOCITY e
[equation (31)] a suddently lower wear rate. This, we
believe, goes a long way toward explaining the tran- Fig. 22. Calibration of the mild-oxidational wear model.
sition, though the role of the oxide in preventing The numbers given against the points are log,,(w). The
normalised wear rates given in parentheses are the values the
direct metallic contact may contribute too. But it predicted wear rates would change to as a result of the
makes a careful comparison of model and experiment switching from severe to mild wear. (The references, in
even more difficult. We have tried to overcome this order, are [12,24,23, 29, 331.)
Without melting, heat transfer is by conduction where I, is the equivalent linear heat-flow length (we
alone; when the asperities melt and smear, fluid flow have set 4 = pr,). [This result can be understood by
transfers heat as well. The most important con- noting that - (Tz - T,)/Z( is the approximate tem-
tribution comes from the latent heat of melting. A perature gradient, AJA, is the fraction of the surface
melting asperity absorbs latent heat; the molten into which heat can be conducted, and K,, is the
material flows and spreads across the adjacent, thermal conductivity of the oxide.] The remaining
cooler, surface and there it solidifies again, releasing heat is available to melt a volume V,,, of oxide, per
its latent heat. So asperity melting is a way of unit area per second, so that
redistributing the heat input to the surface in a more
K,,(TE - TI,) A,
uniform way. Since an asperity spreads as soon as it L,,V,=c(q - (32)
melts, the “dynamic” flash temperature is limited to 4 0 A,
the melting point of the oxide; the high “static” flash where L,, is the latent heat of melting of the
temperatures predicted by equation (17) are not oxide. Using equation (3) for q, and noting that
reached. The metal which lies beneath the oxide sees IV = f, V,/v gives
a uniform, not a localised, heat input, and its tem-
perature is then limited to the bulk temperature, r,. K,,, (T; - T,,) (&V)‘”
ti=f,
That is why there is no melting of the underlying Loxa Pfi
metal even though the “static” flash temperatures
appear to be very high. 27- 11 (33)
’
We will assume that, in melting, spreading and
resolidifying, a fraction f, of all the material which with tl given by equation (6) and T,, given by equation
melts is lost as wear fragments. This we calculate in (8) and N by equation (11). Strictly speaking, the
the following way. The sliding surfaces make contact presence of a thick layer of oxide at the asperity
through asperities (Fig. 23) which are heavily oxi- contact modifies the division of heat there and
dised. Frictional heating causes their tips to melt; changes the parameter CI:it is easier for the frictional
they smear, spreading the viscous, molten oxide over heat to be injected into the cooler and less oxidised
the surface and redistributing the heat. The tem- surface of the disk than into the thicker oxide of the
perature of the contacts is “buffered” at the melting pin. We have considered this and found that it is a
point of the oxide, Tz. Although the heat is initially small effect. The value of tl given by equation (6) is
generated at asperities, the smearing gives a more adequate for our present purposes. We note further
nearly uniform heat input so that the immediate that, in the range of F and B characteristics of this
subsurface is at the bulk temperature, Tb [equation mechanism, bulk heating is small so that T,c TO. We
(S)]. Consider the heat input into one contact at have used Tb in computing the diagram shown below,
some instant in time: an amount gq is injected into the but using TO in its place makes little difference.
surface per unit of (nominal) area per second; some 5.3.5. Calibration of the model for severe-
of this is conducted into the asperities; the rest melts oxidational wear. Figure 24 shows contours of nor-
the asperity tip. Figure 23 shows that the loss of heat malised wear rate, calculated from equation (33) and
by conduction in simply plotted on the usual axes, using the data of Table 2.
There is one adjustable parameter: f,, the fraction of
molten oxide which is lost. We found that f, = 0.01
gives a good fit to the data of Earles and Kadhim [14],
Powell and Earles [9] and Cocks [7] (for low and high
carbon steel) which are plotted on the figure. Later
IF diagrams use equation (33) with this value off,.
PIN
l+ 5.4. Plasticity-dominated wear: adhesion and delami-

Sink temperoture=Tb
nation
Heatinput 5 uq J/m’s \ At sliding velocities below about 0.1 m/s. surface

heating (Fig. 16) is negligible. Then the effect of the
Tf frictional force is to deform the metal surface, shear-
ing it in the sliding direction (e.g. Archard and Hirst
[12]; Suh [41]; Rigney [ill]; and many others). The
1 shearing causes plastic failure: the removal of slivers
of metal from one or both surfaces. When one surface
oxidised
Size = zr, is softer than the other, metal may be transferred
from the soft to the hard surface. When both are
DISK equally hard (as here), flakes or particles of metallic
Fig. 23. An idealised severe-oxidational wear model. The wear debris form and, ultimately, escape from be-
asperities are assumed to have completely oxidised and thin tween the surfaces. In this regime, the wear rate
layers of molten oxide are formed at the asperity contacts. follows Archard’s law [ 1121which, in the terminology
SLIDING VELOCITY v (m/s) component so that a correlation should exist between

wear resistance and fatigue strength.
Any model for plasticity-dominated wear depends
on an understanding of the plastic flow-$eld at and
near the contacting surfaces. This is provided by the
slip-line field analyses of Johnson [80], Bates et al.
[ 1391 and Collins [81]. From these, a point of im-
portance in interpreting experiments emerges: if the
coefficient of friction /* is less than about 0.3, the
plastic flow-field beneath a sliding pin is contained (it
does not break through to the surface); then the
plastic strain accumulated per pass is small. But when
p is greater than 0.3, the flow-field breaks through to
the surface and the plastic strain per pass can be
large. This suggests two regimes of wear, depending
on the value of p. Two regimes are certainly observed
(e.g. Archard and Hirst [12]).
In mild (low load) delamination wear, direct metal-
to-metal contact is thought to be prevented by the
presence of a thin (a few A) but tough layer of oxide.
NORMALISED VELOCITY ^i; Being thin, the layer can deform elastically and is not
Fig. 24. Calibration of the severe-oxidational wear model. ruptured by the light loads, and the coefficient of
The numbers given against the points are log,, (@). (The friction is low. The black oxide powder normally
references, in order, are [7,9,14].) observed in mild wear is generated from this layer
and perhaps from small metallic fragments which are
of this paper, is simply removed and oxidised. The damaged surface quickly
reoxidises, avoiding further metal-to-metal contact
(34)
between the sliding surfaces. At higher loads the
where k, is the Archard wear coefficient (typical oxide is penetrated. The increase in friction causes a
values: 10~5-10-3). dramatic increase in the plastic shear strain accumu-
If slivers of metal are removed, then cracks must lation in the subsurface, and in the rate of wear
nucleate and propagate (on a microscopic scale, of [12, 17, 26,31,40].
course) to allow separation of the wear fragment 5.4.1. The model for plasticity-dominated wear. In
from its parent surface. Models for plasticity- the plasticity-dominated regime, Archard’s law
dominated wear seek to identify and analyse the (w = kAF) is obeyed. Those who attempt to model
nucleation and growth of these cracks. And there is the mechanism (cited above) seek to derive this law,
a host of models: some idea is given by the following and to elucidate the physical content of the dimen-
list. sionless constant k,. Much evidence points to the
(a) Models based on adhesion between asperities nucleation of cracks just below the surface which
date from Archard’s early work [112] and they have propagate, at first parallel to the surface, but later
been further developed [108, 113-I 151. The idea is breaking out to give a metallic wear fragment. The
that touching asperities stick together; subsequent modelling of this delamination is associated, above
sliding shears or plucks off the tip of the softer all, with the name of Suh. The model we now develop
asperity, leaving it adhering to the harder surface, differs in some details from that of Suh (it involves
from which it may later become detached, giving a the formation of a mode II crack by plastic hole
wear fragment. growth, and does not invoke fracture mechanics); but
(b) Delamination-models [41,45, 1161 most nearly the underlying ideas draw heavily on his work and
describe the microstructural observations during the final result is very like his.
severe plasticity-dominated wear. Here the idea is The ideas are these. The frictional traction on the
that unidirectional shearing of the metal surface sliding surface causes accumulating plastic shear
nucleates subsurface voids; these extend and link to [140], as shown in Fig. 25. Voids nucleate at inclu-
give crack beneath and parallel to the surface; and sions below the surface [lo, 411 and grow by the
when crack is large enough it breaks out to give a mechanism of shear-fracture described by Teirlinck et
flake-like wear particle. Fracture mechanics has been al. [141]. The voids lie at a characteristic depth
applied to the problem [117-1251; and microscopic [121, 1421 probably determined by a balance between
ideas based on cell structures have been invoked to the hydrostatic compression beneath a contacting
explain the size of the wear fragments [ 111, 126-l 321. asperity (which tends to suppress void nucleation)
(c) Models based on fatigue crack propagation and the plastic shear strain (which tends to cause it);
have been developed [112, 133-1381. The idea here is the characteristic depth is typically of the same order
that the local plasticity during wear has a cyclic as the size of an asperity (10 pm) [lo, 41, 1431. Figure
depth X0 needed to produce this failure is Y*, then

fX
y*=2f,.
The “life” of the surface is related to the number

n* of passes required to build up the plastic shear
strain to cause failure
y* =n*y, (36)
where y,, is the plastic shear strain accumulated per
pass. After n* passes, a wear fragment of thickness
X0 detaches itself, so the wear rate is
W=---&A,
n*l
where 1 is the distance between asperity contacts. But

(from the equation above)
C-4 n*=- _fz (37)

2f”YO
and
r, A
-_=‘= P
a. An
giving
w
-=- 2KYJ+
A” r,fX
If we take X0 to be equal to the asperity radius r,, we
(4 have
Fig. 25. An ideahsed plasticity-dominated (delamination) @-2Y”f,&k p

(38)
wear model. (a) Voids nucleate around the inclusions at a fX *.
critical depth X0 under the influence of plastic shear strain
This is simply Archard’s law [I 121with k, = 2y,f,/f 2.
(i). The continuous shearing action and compression cause
the shape of the voids to change (ii) and (iii), while the area
fraction of voids increases. (b) Photomicrographs of the 5.4.2. Calibration of the model for plasticity-
void formation around inclusions and crack propagation dominated wear. Calibration of Archard’s law simply
from these voids near the surface in annealed Fe-1.3% MO.
requires a value (or values) of k,. The data-plots
(From Suh et al. [IO]. Copyright American Society of
Mechanical Engineers. Reproduced by permission.) (c) The for steels (Figs l-6) are consistent with two values.
voids around the inclusions on a plane parallel to and at a When wear is mild
depth X0 below the sliding surface. y is the cumulative
plastic shear strain there.
k,= 5 x lO-5 (39a)
and when it is severe
25 shows that the area fraction of voids, fA, on a k,=5x 10-3. (39b)
plane parallel to and at a depth X, below the sliding
Figure 26 shows wear-rate contours, calculated by
surface is
using these values of kA , superimposed on data, most
fA = 6%NV)(2ri1G%Y ) N ‘X Y (35) of it from papers by Archard and Suht.
The model gives insight into the origin of these
where ri is the radius of an inclusion, NVis the number values. Taking y0 = 1 %, f, = 10m3and f 2 = 0.5 gives
of inclusions per unit volume, y is the cumulative k, = 4 x 10m5, which is of the right order for mild
plastic shear strain, and f, is the volume fraction of delamination wear. The corresponding value for the
the inclusions. When this area fraction fA reaches a severe delamination wear is about 100 times greater
critical value, f 2, (say 0.5), fast fracture causes a wear requiring that the plastic shear strain per pass to
flake to break off. If the shear strain at the critical increase from 1 to loo%, which is not an unreason-
able estimate in the severe wear as shown by Archard
and Hirst [12]. The model leads to the additional
TThe data points on Fig. 26 extend into the mild-oxidational
wear regime. In this are of overlap, the two mechanisms conclusion that the wear rate should depend on the
can coexist. inclusion content of the material. In constructing the
SLIDING VELOCITY v (m/s) source, agreement with data from other sources is
good.
The wear-rate equations are expressed in terms of
the normalised variables F, u’ and m. The idea of
normalising is to remove some of the material prop-
erties and geometric features of the experiments,
and make data from diverse tests more nearly com-
parable. It seems to work. We have found that data
from steels (ranging from pure iron to tool steel) and
from a number of test geometries (not just the
pin-on-disk configuration which we took as a stan-
dard) fall together in a sensible way when plotted on
these axes.
It is obvious that the models badly need further
refinement. In particular, the transition from severe
to mild wear during mild-oxidational wear (caused, it
is thought, by martensite formation) is inadequately
modelled (these transitions are discussed elsewhere
[144]); severe-oxidational wear needs more detailed
NORMALISED VELOCITY v study; and the delamination-wear model requires a
better understanding of subsurface plasticity and the
Fig. 26. Calibration of the plasticity-dominated (delami- way it causes subsurface cracks to nucleate and grow.
nation) wear model. Broken lines are used to represent the
estimated normalised wear rates in the ultra-mild wear And certain other mechanisms (such as ultra-mild
regime. The numbers given against the points are log,, (m). wear) are still inadequately modelled.
To avoid congestion, only selected data points from one But despite these reservations, the underlying phys-
source are shown, mainly to indicate a change in wear rate. ical basis of the dominant wear mechanisms is toler-
The normahsed wear rate given in parentheses is the value
ably well understood. We now assemble these, and
for a mild wear condition. The transition between mild and
severe plasticity-dominated wear takes place over a range the equations for bulk and flash temperatures, into a
of normalised pressures as indicated by the shaded region. model-based wear map. The wear equations are
(The references, in order, are [4,6,8,30, l&-13,40, 16, 17, summarised in Table 4.
19, 201.)
6. WEAR-MECHANISM MAPS
wear map shown presently, we have used the two 6.1. The wear -mechanism map
values of k, given above, switching from the smaller There are, then, a number of distinct mechanisms
to the larger at a normalised pressure P of 3 x 10m4. of wear: plasticity-dominated wear, oxidation-
dominated wear, melt-dominated wear, and so on.
5.5. Ultra-mild wear
Each can be divided into sub-mechanisms: mild- and
There is a regime in the low load and low velocity severe-oxidational wear, for example. Each has a
area where very little wear takes place. This is the characteristic range of dominance, that is, a field of
ultra-mild wear regime where the thin but tenacious load F and velocity u in which it contributes more to
oxide film is never penetrated and no metal-to-metal the wear rate than any other mechanism.
contact occurs. No wear model is available for this Each mechanism can be described by a normalised
regime; nor are data adequate to permit calibration wear-rate equation of the form
if there was one. The approximate position of the
W = f (P, 6, material properties) (40)
regime is shown on the wear map, but wear rates are
not computed. where m, P and 5 are defined by equation (2); and
each of these equations can be calibrated by adjusting
5.6. Summary of wear modeIs and equations
parameters in it to give the best fit to data for wear
Five simple analytical models for the dominant within the field (as we did for steel in Section 5).
wear mechanisms have been described and calibrated It is then possible to construct model-based wear-
using experimental wear-rate data obtained primarily mechanism diagrams. The axes are the usual ones we
from pin-on-disk wear tests. Two-those for seizure have used throughout this paper: F and 6. The range
and for melt wear-contain no adjustable parame- of values we wish to cover is large, so it is convenient
ters, and they agree well with the experimental data to use log scales (that is, log F and log 6) and to
in their respective regimes of dominance. The other allow P to range from 10e5 to 10, and B to range from
three-mild- and severe-oxidational wear, and de- lo-* to 105; in this way the diagram covers almost the
lamination wear, do contain adjustable parameters; entire range of all reported wear experiments.
the function of the “calibration” procedure is to set On these axes it is now possible to draw the field
them. But having set them using data from one of dominance of each mechanism by evaluating each
Table 4. The simplified wear-rale equations

Seizure
equation (22)
equation (25)
Mild-oxidational wear
equation (31)
Severe-oxidational wear
equation (33)
Plasticity-dominated wear
@‘=k APwith k =% equation (38)

A fX
With the following definitions
equation (2)
1
equation (6)
a = 2 + /3(nv‘/8)“2
equation (7a)
equation (7b)
p = 0.78 - 0.13 log,, (I?) equation (9)

$ )(I -n+1
NC
0
1
T, = 300 K
equation (1 I)
T, and Tt are given in Table 3
equation in turn, selecting the one which gives the geometries lead to the same map because of the
largest value of I?. Field boundaries are the lines norrnalised axes), together with any information
along which two mechanisms give equal rate: they are about the dominant mechanism. The data are tabu-
plotted by equating pairs of wear-rate equations and lated, listing the load F and velocity v for each wear
solving for P in terms of ~7.Onto these fields are rate W.
plotted contours of constant normal&d wear rate, b?, (b) Information is sought from each study to
obtained by evaluating the wear-rate equation for the identify the nominal area A, (the area of the pin end,
dominant mechanism in each field. for example) and the hardness H,,, together with the
The result of doing this for steel is shown in thermal properties, u, K, L etc.
Fig. 27. The diagram displays the results of the (c) Using this information, values of the nor-
heat-flow analysis and modelling of Sections 4 and 5, malised variables F, 6 and I? are added to the table,
each step of which was separately calibrated to and, as far as possible, a mechanism is assigned to
experiment. It can be regarded as a summary of the each data point.
wear behaviour of dry steel on steel, over all practical (d) The normalised data are plotted in the way
loads and velocities. illustrated by Figs l-6, identifying the mechanisms by
It may be helpful to itemise in slightly more detail appropriate plotting symbols. Approximate bound-
the general procedure for constructing such dia- aries are sketched in by eye enclosing blocks of data
grams. It is laid out below in a way which should with a common mechanism, in order to identify the
allow maps to be constructed for other materials. subsets of data which will be used to calibrate each
wear-rate equation (see Fig. 8).
6.2. General procedure for constructing maps (e) The bulk and flash temperature at the sliding
interface are calculated from equations (8) and (17)
A wear map for a pair of material is constructed
or (if necessary) straight forward modifications of
by the following steps.
these to allow for a drastically different geometry, or
(a) Wear data are collected for the sliding pair, for the division of heat between two dissimilar materials.
the chosen geometry (though, within limits, different The results are checked, and recalibrated if necessary,
SLIDING VELOCITY v (m/s) mechanisms of wear. The method is developed for the
unlubricated wear of steel on steel (Fig. 27); there are
no obvious obstacles to applying it to other materials,
or combinations of materials. The diagrams sum-
/
marise the wear behaviour over a wide range of load
I I I I I I
and sliding velocity, identifying the dominant mech-
anism and showing the overall wear-rate.
The diagrams use normalised variables. So many
mechanisms and material properties are involved in
describing wear that there is no hope of constructing
a universal diagram which describes all materials, or
even all metals. But the normalised axes do allow for
the most important effects of surface hardness, ther-
mal conductivity and sample geometry, and allow
data from different test geometries, and different
variants of the same materials, to be plotted together:
a single diagram does describe, approximately, the
wear behaviour of a wide range of steels.
‘” ,o-2 2
Finally, a warning. Wear diagrams constructed in
1 10‘
the way we have described are only as good as the
NORMALISED “ELO:TV v
models and data used to construct them-and both
Fig. 27. The wear-mechanism map for a steel sliding pair leave much room for improvement. The map shown
using the pin-on-disk configuration. Contours of constant here should be regarded as a first approximation
normalised wear rates are superimposed on fields showing
only. A wear map is a summary of the current,
the regimes of dominances of different wear mechanisms.
There are discontinuities in the contours when they cross the imperfect, understanding (like a phase-diagram), not
field boundaries into the regimes of severe-oxidational wear an exact statement (like a wiring diagram). It must
and melt wear. The wear rates given in parentheses are the never be regarded as precise.
values when mild wear takes place. The shaded regions
indicate a transition between mild and severe wear. ~~~~~~~e~~e~~~f~~ne of the authors (S.C.L.) would like
to thank the National University of Singapore for the award
of a scholarship as well as granting him study leave to
against any available experimental indicators of tem- undertake this piece of work. We also wish to acknowledge
perature (phase changes, direct temperature measure- the numerous discussions with Professor K. L. Johnson,
ments etc) in the way illustrated in Section 4. It is Professor D. Tabor and Professor D. A. Rigney and many
essential that the tem~rature equations are properly others, which have been of the greatest assistance in formu-
lating the mechanisms.
formulated, since they enter most of the wear-rate
equations.
REFERENCES
(f) Each wear-rate equation is now compared with
the appropriate block of data [identified under (d), 1. M. V. Martialis, Epigrams, Book 2 Number 77 (A.D.
above] and calibrated to give an acceptable fit. New 85).
2. D. Tabor, Roe. fnt. Co& Tr~boiogy in the &‘a’~,NASA
material (ceramics, for example) may involve new Lewis Research Center, Cleveland, Ohio, NASA, p.i
mechanisms; and, even for metals, it will be obvious (1983).
that mechanisms exist which we have, in this initial 3. H. J. Frost and M. F. Ashby, Deformation-Mechanism
study, ignored. So it may be necessary to seek new Maps. The Plasticity and Creep of Met& and
Ceramics. Pergamon Press, Oxford (1982).
models and new calibrations to include them.
4. N. Saka, A. M. Eieiche and N. P. Suh, Wear 44, 109
(g) The wear-m~hanism map is then calculated (1977).
from the temperature and wear-rate equations 5. F. H. Kinsella and T. H. C. Childs, Inst. Mech. Engrs.
(Tables 3 and 4). in the way described in Section 6.1. Conf. Publ. 1978, 55 (1978).
This is most easily done numerically (using an IBM 6. S. Bhattacharyya, Wear 61, 13 (1980).
7. M. Cocks, J. appl. Phys. 28, 835 (1957).
PC for instance) by stepping through values of F and 8. S. Jahanmir, N. P. Suh and E. P. Abrahamson II,
6, evaluating all the equations at each point and Wear 28, 235 (1974).
identifying the dominant mechanism as the one which 9. D. G. Powell and S. W. E. Earles, A.S.L.E. Trans. 11,
gives the largest value of i8’. The wear-rate contours 101 (1968).
are plotted by selecting the largest @, and the field 10. N. P. Suh, S. Jahanmir, E. P. Abrahamson II and
A. P. L. Turner, Trans. A.S.M. E.. J. Luh. Tech. 96,63 I
boundaries are plotted by identifying the points at (1974).
which the dominant mechanism changes. Il. M. Antler, Wear 7, I81 (1964).
12. J. F. Archard and W. Hirst. Proc. R. Sot. Ser. A 236.
6.3. ~on~~us~o~s und ~~rnrnar.~ 397 (1956).
13. E. P. Abrahamson, II, S. Jahanmir, N. P. Suh and
It appears practica1 to construct wear-mechanism D. A. Coiling, Proc. Int. ConS. Production Engineering,
diagrams which show the wear rate, and the regime Tokyo, 1974, Japan Sot. of Precision Engng and
of dominance, of each of a number of competing CIRP, Tokyo, Part I, p. 408 (1974).
14. S. W. E. Earles and M. J. Kadhim, Proc. Inst. Mech. Appi. Mech., p 1343. Am. Sot. Mech. Engrs, New
Engrs 180, 531 (1965/66). York (1962).
IS. R. M. Farrell and T. S. Eyre, Wear 15, 359 (1970). 59. F. F. Ling and S. L. Pu, Wear 7, 23 (1964).
16. N. C. Welsh, Proc. Conf: Lubrication and Wear, 60. F. F. Line and J. S. Rice. A.S.L.E. Trans. 9. 195
London, p. 701. Inst. Mech. Engrs (1957). (1966). -
17. R. Ward, Wear 15, 423 (1970). 61. H. S. Carslaw and J. C. Jaeger, Conduction of Heat in
18. T. S. Eyre, Wear 34, 383 (1975). Solids, 2nd edn. ClarendonPress, Oxford (1959).
19. N. Soda and T. Sasada. Trans. A.S.M.E.. J. Lub. Tech. 62. R. L. Johnson. M. A. Swikert and E. E. Bisson.
100, 492 (1978). NACA Tech. Note, No. 1442, 1, National Advisory
20. S. Jahanmir, N. P. Suh and E. P. Abrahamson II, Committee for Aeronautics, Washington (1947).
Wear 32, 33 (1975). 63. F. P. Bowden and P. A. Persson, Proc. R. Sot., Ser.
21. R. S. Montgomery, Wear 36, 275 (1976). A 260, 433 (1961)
22. E. Takeuchi, K. Fujii and T. Katagiri, Wear 55, 121 64. K. Williams and E. Giffen, Proc. Inst. Mech. Engr 178,
(1979). 24 (1963164).
23. T. F. J. Quinn, A. R. Baig, C. A. Hogarth and 65. D. Tabor, Proc. R. Sot., Ser. A 251, 378 (1959).
H. Muller, A.S.L.E. Trans. 16, 239 (1973). 66. E. Rabinowicz, Friction and Wear of Materials. Wiley,
24. P. M. Dunckley and T. F. J. Quinn, Inst. Mech. Engrs New York (1965).
Conf Publ. 1978, 45 (1978). 67. J. A. Greenwood, Trans. A.S.M.E., J. Lub. Tech. 89,
25. K. Endo and Y. Iwai, J. J.S.L.E. Int. Edn 2, 11(1981). 81 (1967).
26. N. C. Welsh, Phil. Trans. R. Sot. Ser. A 257,31 (1965). 68. I. V. Kraghelski and N. B. Demkin, Wear 3, 170
27. P. Grossberg and J. Molgaard, Proc. Inst. Mech. Engrs (1960).
181, 16 (1966/67). 69. C. B. Allen, T. F. J. Quinn and J. L. Sullivan, Trans.
28. N. C. Welsh, J. appl. Phys. 28, 960 (1957). A.S.M.E., J. Tribol. 107, 172 (1985).
29. T. F. J. Quinn, D. M. Rowson and J. L. Sullivan, Wear 70. T. F. J. Quinn and W. 0. Winer, Wear 102,67 (1985).
65, 1 (1980). 71. S. C. Lim. Ph.D. thesis. Cambridee Univ. (1986).
30. J. F. Archard, Wear 2, 438 (1958/59). 72. M. Bass, in Physical Processes in Laser-Material Ioter-
31. N. C. Welsh, Phil. Trans. R. Sot., Ser. A 257, 51 action edited by M. Bertolotti, p. 77. NATO Advanced
(1965). Studies Institute, Plenum, New York (1982).
32. C. L. Harris and D. Wyn-Roberts, Proc. Inst. Mech. 73. F. P. Bowden and D. Tabor, The Friction and Lubri-
Engr 183, 50 (1968/69). cation of Solids, Part I. Clarendon Press, Oxford
33. T. F. J. Quinn, A.S.L.E. Trans. 10, 158 (1967). (1950).
34. T. F. J. Quinn, Proc. Inst. Mech. Engrs 182, 201 74. F. P. Bowden and D. Tabor, The Friction and Lubri-
(1967/68). cation of Solids, Part II. Clarendon Press, Oxford
35. T. F. J. Quinn, A.S.L.E. Trans. 21, 78 (1978). (1964). ”
36. T. F. J. Quinn, Proc. Inst. Mech. Engrs. 183, 129 75. H. Uetz and K. Sommer, Wear 43, 375 (1977).
(1968/69). _ 76. F. P. Bowden and D. Tabor, Proc. Inst. Mech. Engrs
37. W. T. Clark. C. Pritchard and J. W. Midnley, Proc. 160, 380 (1949).
Inst. Mech. &grs 182, 97 (1967168). - _ 77 A. P. Semenov, Wear 4, 1 (1961).
38. D. H. Buckley, M. Swikert and R. L. Johnson, 78. L. Rozeanu, A.S.L.E. Trans. 16, 115 (1973).
A.S.L.E. Trans. 5, 8 (1962). 79. H. Mishina and T. Sasada, J. J.S.L.E. 29, 769 (1984).
39. D. J. Barnes, J. E. Wilson, F. H. Stott and G. C. 80. K. L. Johnson, J. Mech. Phys. Solicis 16, 395 (1968).
Wood, Wear 45, 97 (1977). 81. I. F. Collins, Int. J. Mech. Sci. 22, 735 (1980).
40. J. F. Archard and W. Hirst, Proc. R. Sac., Ser. A 238, 82. L. Ahman and A. Oberg, Proc. Int. Conf. Wear of
515 (1957). Materials, Reston, Virginia, p. 112. Am. Sot. Mech.
41. N. P. Suh, Wear 25, 111 (1973). Engrs (1983).
42. N. P. Suh and H.-C. Sin, Wear 69, 91 (1981). 83. W. A. Glaeser. Wear 73. 371 (1981).
43. T. Kayaba and K. Kato, Proc. Int. Conf Wear of 84. S. J. Calabresd, F. F. Ling and S. Murray, A.S.L.E.
Materials. Dearborn. Michigan, n. 45. Am. Sot. mech. Truns. 26, 455 (1983). -
Engrs (1979). - _ 85. F. P. Bowden and K. E. W. Ridler. Proc. R. Sot., Ser.
44 W. A. Brainard and D. H. Buckley, NASA Tech. Note A 154, 640 (1936).
TN D-7700 (1974). 86. F. P. Bowden and P. H. Thomas, Proc. R. Sot., Ser.
45. N. P. Suh and coworkers, The Delamination Theory of A 223, 29 (1954).
Wear. Elsevier, Amsterdam (1977). 87. E. Saibel R. Aircraft Estab. Trans, no. 391 (1951).
46 J. C. Jaeger, J. Proc. R. Sot. N.S. W. 76, 203 (1942). 88. B. Sternlicht and H. Apkarian, A.S.L.E. Trans. 2,248
47. H. Blok, Proc. Second World Petroleum Gong., Paris, (1960).
Vol. 3, p. 471 (1937). 89. F. P. Bowden and E. H. Freitag, Proc. R. Sot., Ser. A
48. H. Blok, Inst. Mech. Engrs, Proc. General Discussion 248, 350 (1958).
Lubrication and Lubricants, Vol. 2, p. 222 (1937). 90. F. J. Carignan and E. Rabinowicz, A.S.L.E. Trans. 23,
49. J. R. Barber, J. Mech. Engng Sci. 9, 351 (1967). 451 (1980).
50. J. R. Barber, Int. J. Heat Mass Transfer 13,857 (1970). 91. W. R. D. Wilson, Trans. A.S.M.E., J. Lub. Tech. 98,
51. G. A. Berry and J. R. Barber, J. Triboi. 106, 405 22 (1976).
(1984). 92. R. S. Montgomery, Wear 38, 235 (1976).
52. S. W. E. Earles and M. G. Hayler, Wear 20, 51 (1972). 93. D. C. B. Evans, J. F. Nye and K. J. Cheeseman, Proc.
53. S. W. E. Earles and D. G. Powell, A.S.L.E. Trans. 11, R. Sot., Ser. A 347, 493 (1976).
109 (1968). 94. B. Kehl and E. Siebel, Archs Eisenhutt. 9, 563 (1936).
54. S. W. E. Earles, M. G. Hayler and D. G. Powell, 95. R. Mailander and K. Dies, Archs Eisenhutt. 16, 385
A.S.L.E. Trans. 14, 135 (1971). (1943).
55. D. G. Powell and S. W. E. Earles, A.S.L.E. Trans. 15, 96. K. Dies, Archs Eisenhutt. 16, 399 (1943).
103 (1972). 97. K. Nakajima and Y. Mizutani, Wear 13, 283 (1969).
56. F. F: Ling, Z. Angew. Math. Phys. 10, 461 (1959). 98. E. M. Eden, W. N. Rose and F. L. Cunningham, Proc.
57. F. F. Lina. Trans. A.S.M.E.. J. Lub. Tech. 91, 397 Inst. Mech. Engrs, p. 839 (1911).
(1969). - 99. G. A. Tomlinson, Proc. R. Sot., Ser. A 115, 472
58. F. F. Ling and C. W. Ng, Proc. 4th U.S. Natn Cong. (1927).
100. M. Fink, Trans. Am. Sot. Steel Treaf. 18, 1026 (1930). 142. S. Jahanmir and N. P. Suh, Wear 44, 17 (1977).
101. T. F. J. Quinn, Wear 18, 413 (1971). 143. J. W. Ho, C. Noyan, J. B. Cohen, V. D. Khanna and
102. T. F. J. Quinn, Br. J. appl. Phys. 13, 33 (1962). Z. Eliezer, Wear 84, 183 (1983).
103. T. F. J. Quinn and J. L. Sullivan, Proc. Int. Conf on 144. S. C. Lim, M. F. Ashby and J. H. Brunton, /icta
Wear of Materials, St. Louis, Missouri, p. 110. Am. metall. In press.
Sot. mech. Engrs (1977). 145. J. Mslgaard and W. W. Smeltzer, J. appl. Phys. 42,
104. T. F. J. Quinn, J. L. Sullivan and D. M. Rowson, Proc. 3644 (1971).
~nr. Conf on Wear of Materials, Dearborn, Michigan, 146. Handbook of Chemistry and Physics, 52nd edn. Chem.
p. 1. Am. Sot. Mech. Engrs (1979). Rubber Book, New York (1971).
105. D. M. Rowson and T. F. J. Quinn J. Phys. D, Appi. 147. American Institute of Physics Handbook, 3rd edn.
Phys. 13, 209 (1980). McGraw-Hill, New York (1972).
106. T. F. J. Quinn, J. L. Sullivan and D. M. Rowson, Wear
94, 175 (1984).
APPENDIX 1: THE NUMBER OF
107. 0. Kubashewski and B. E. Hopkins, Oxidation of
Metals and Alloys, 2nd edn. Butterworths, London CONTACTING ASPERITIES
(1962). Figure Al shows an idealised pin end with contacts (black
108. G. Yoshimoto and T. Tsukizoe, Wear 1,472 (1957/58).
squares) distributed randomly on it. Each “unit asperity’
109. N. Tenwick and S. W. E. Earles, Wear 18,381 (1971).
has radius ra, and area nra. The sum of these areas is the
110. J. Mslgaard and V. K. Srivastava, Wear 41, 263
real area of contact A,. The number of asperities increases
(1977).
1 Il. D. A. Rigney, Proc. Int. Conf. Fundamentals of Tribo-
logy, MIT, Cambridge, Mass, p. 119. MIT Press,
Cambridge, Mass (1978).
112. J. F. Archard J. appl. Phys. 24, 981 (1953). ‘black’
113. T. Sasada and S. Norose, presented at JSLE-ASLE
= A,
Int. Lubrication Conf Tokyo, Japan (1975).
114. T. Sasada, S. Norose and H. Mishina, Trans. A.S.M.E.
J. Lub. Tech. 103, 195 (1981).
115. J. Halling, Trans. A.S.M.E., J. Lub. Tech. 105, 212
(1983).
116. N. P. Suh, presented at: Am. Sot. Metal. Mater. Sci.
Semin. Pittsburgh, Pa (1980).
117. I.-M- Feng, J. appl. Phys. 23, 1011 (1952).
118. J. R. Fleming and N. P. Suh, Wear 44, 38 (1977).
119. Y. Kimura, Proc. Int. Conf Fundamentals of Tribol-
ogy, MIT, Cambridge, Mass. p. 597. MIT Press,
Cambridge, Mass. (1978).
120. 0. Vingsbo, Proc. Int. Conf. Wear of Materials, Dear- r, >> r.
2 r0
born, Michigan, p. 620. (1979).
121. D. A. Hills and D. W. Ashelby, Wear 54, 321 (1979). Size of pin end, total area = A,
122. A. R. Rosenfield, Wear 61, 125 (1980).
Fig. A 1. An idealised pin end
123. A. R. Rosenfield, presented at: Am. Sot., Metal.
Mater. Sci. Semin., Pittsburgh, Pa (1980).
124. A. R. Rosenfield, Wear 72, 245 (1981). as the normal load F increases. Consider adding an addi-
125. A. R. Rosenfield, Proc. Int. Conf. Wear of Materials, tional number dN of unit asperity to the number N already
Reston, Virginia, p. 390. Am. Sot. Mech. Engrs. there, at random. The probability that a new contact will be
(1983). adjacent to an old one (and thus does not increase the true
126. J. P. Hirth and D. A. Rigney, Wear 39, 133 (1976) number of separated contact Nr) is
127. R. C. Bill and D. Wisander, Wear 41, 351 (1977).
128. N. P. Suh and N. Saka, Wear 44, 135 (1977).
129. D. A. Rigney and W. A. Glaeser, Wear 46,241 (1978).
130. D. A. Rigney and J. P. Hirth, Wear 53, 345 (1979). where r) is a constant
p=q
0 +
n
near unity. The probability of not

131. D. A. Rigney, L. H. Chen, M.G.S. Naylor and A. R. doing this (and thus the probability of creating a true new
Rosenfield, Wear 100, 195 (1984). contact) is, of course, 1-P. Then the true number of new
132. J. Don and D. A. Rigney, Wear 105, 63 (1985). contacts, dN,, is related to the number added dN by
133. L. Rozeanu, Wear 6, 337 (1963).
134. K. Endo and Y. Fukuda, Proc. 8th Japan Gong on dN,=[l -,)(:)]dN.
Testing Materials, Kyoto, Japan, p. 69. Sot. Mater.
Sci., Japan (1965). But Nrtrz = A,. Using this to substitute for dN gives
135. I. V. Kraghelski, Fricfion and Wear. Butterworths,
London (1965). dN,=($)(l ++)d(:)
136. E. F. Finkin, Wear 47, 107 (1978).
137. R. A. Smith, Proc. Int. Conf. Fundamenlals of Tribo- Integrating, using the limiting result that NT = I when
logy, MIT, Cambridge, Mass, p. 605. MIT Press, A,/A, = 1, and substituting e = A,/A, gives
Cambridge, Mass. (1978).
138. Y. Kimura, presented at: Am. Sot. Metal. Mater. Sci. N =N,= 5
‘p(l -F)+ I. (Al)
Semin. Pittsburgh, Pa (1980). 0 rr
139. T. R. Bates Jr, K. C. Ludema and W. A. Brainard, The predictions of this equation are in good agreement
Wear 30, 365 (1974). with the experimental results of Quinn and Winer 170) who
140. J. E. Merwin and K. L. Johnson, Proc. Inst. Mech. found N ranged from 4 to 10 when Franged from 1.2 x IO-”
Engrs, 177, 676 (1963). to 6.6 x 10m4. Equation (Al) gives N = 4 to 16 for this load
141. D. Teirlinck, J. D. Embury and M. F. Ashby, Acfa range, assuming a “unit asperity” to have a radius of 10m3m
metail. (1987). In press. and the pin to have a radius of 1.5 x 10 -‘rn.
APPENDIX 2: THE APPROPRIATE APPENDIX 4: CORRECTION FOR

HEAT-FLOW EQUATION FOR THE EFFECT OF AN OXIDE FILM
FLASH HEATING The presence of an oxide film increases the asperity tem-
The temperature produced by injecting an energy density perature by increasing the resistance to heat flow. In the
aq’ J/m% into an area art (the area of an asperity contact) one-dimensional approximation (Fig. A2), the oxide, of
is well approximated by [61,72] thickness Z, contributes an additional (series) resistance to
metal Sink temperature = TJ
642) hl
where t is the time of injection, K,,, is the thermal conduc-

tivity, and a is the fraction of the energy q’ [equation (IO)]
which enters the pin. We identify the injection time t with the
transit time of an asperity on the pin over one on the disk
2r
tz-2
V
Fig. A2. The oxide film at an asperity contact.
where v is sliding velocity. Using equation (lo), and con-
verting to the dimensionless variables t7, T* [equations (2) the diffusive path, of length I,. At steady state, the effective
and (7a)] gives thermal conductivity K, is given by summing the resistances
of the oxide (Z/K,,) and of the metal [(/, - Z)/k,], giving
Tf= Ti+zr,T*ii tan-l
r.
This expression has two simple limits, one of which is
relevant to the present problem. When 17is sufficiently large
the arctangent becomes equal to its argument, giving where K,,, and K,,, are the thermal conductivity values for
112 metal and oxide respectively, and we have set I, = j?r,. Using
T,= r’, + 2ap T*6”* > . an average thermal conductivity value K,, = 3.2 J/msK for
0 0 magnetite and hematite [145], and a value of X;, = 41 J/msk
When, instead, v”is sufficiently small, the arctangent tends to for steel gives a value of 14 J/msk for the effective thermal
the value n/2, giving conductivity K, when Z = 1Opm. We may then define a
modified equivalent temperature
(A4)
The transition occurs, to an adequate approximation, when

the argument is 1, that is, when
which we have used this, where appropriate, in the calcu-
lations.
APPENDIX 5: THE HARDNESS AS A
The calculation of the flash temperature is relevant only
when it is less than the melting point of the sliding surfaces;
FUNCTION OF TEMPERATURE
when the asperities melt, the local temperature is limited AND STRAIN RATE
to the melting temperatures, and a different calculation
(Section 5.2) becomes relevant. For the values of ~1and T* The local strain rates (i u v/r,) in sliding friction are high:
relevant for steel (Table 2), this requires that the normalised typically they range from 10 to 106/s. At these rates, metals
velocity D be less than roughly 5 x 102. So to a first respond in a way which is best modelled by low-temperature
approximation, equation (A4) is the one we want (Archard plasticity (“dislocation glide”) rather than by creep, even at
[30] reaches the same conclusion). If further sophistication high temperature [3]. We therefore describe the current
is required, it should be replaced by equation (A2). hardness at a surface temperature T and strain rate 1. by
APPENDIX 3: THE “FLASH” /=&exp[-%(I--:)1

TEMPERATURE NEAR SEIZURE
where H is the hardness at temperature T and strain rate i,
The sink temperature for flash heating, Ti, was identified, Ho is the hardness at room temperature, and i, and AF are
in Section 4.2, with the bulk (or average) surface tem- the pre-exponential constant and the activation energy for
perature, T,. Almost always the approximation is comple- dislocation glide. The last are tabulated by Frost and Ashby
tely adequate. But there is, at very high loads, the exception- [3], but for present purposes it is sufficient to note that, for
al circumstance of seizure; then the asperity radius grows most metals, AF/RT,,, = 20 and i,, ‘v 106/s. Inversion gives,
until it is equal to the nominal contact radius, r,,, and the to an adequate approximation
sink temperature for flash heating becomes identical to that
for bulk heating (i.e. To). To include this special case we write X= 1 -~(~)ln@).
0
T:= Tb- ; (Tb-To) Note that increasing the temperature causes the hardness to
0 fall, but increasing the strain rate causes it to rise. In sliding,
where ra/rO is given by equation (16). The flash heating both T and i are determined by the velocity v: for steel the
equation [equation (17)] is then modified to competing effects of Tand i roughly cancel. At the next level
of approximation, we identify T with the bulk temperature,
Tb, and noting that 5 5 v/r,, we obtain the correction factor
for the change of hardness with frictional heating
It reduces to equation (17) under all except the extreme limit
of P % 1, when the modification has the effect of pulling the
W)
flash temperature contours and the bulk temperature con-
tours together. In computing the figures, equation (A5) This has been used in equation (22) of the text to calculate
rather than equation (17) has been used. the seizure line.

1987 - Lim e Ashby - Wear-Mechanism Maps

Uploaded by

Copyright:

Available Formats

1987 - Lim e Ashby - Wear-Mechanism Maps

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

1987 - Lim e Ashby - Wear-Mechanism Maps

Uploaded by

Copyright:

Available Formats

Acta metal/. Vol. 35, No. I, pp. l-24, 1987 OOOI-6160/87 53.00 + 0.

S. C. LIM’ and M. F. ASHBY’

(Received 10 March 1986; in revised form II March 1986)

Resume-Nous explorons les possibilites des diagrammes de mPcanismesd’usure. Nous construisons

Zusammenfassung-Die Anwendungsmijglichkeiten von Diagrammen der Abriebmechanismen werden

able, the experimental data with results of theoretical 2. WEAR-MECHANISM MAPS

Table I. Symbols, definitions and units

wear rate@G/m) thermal conductivity of oxide (J/msK)

3. EMPIRICAL WEAR-MECHANISM MAPS

SLIDING VELOCITY v (m/s) SLIDING VELOCITY v (m/s)

NORMALISED VELOCITY i? NORMAL&ED VELOCITY v

SLIDING VELOCITY v (m/s)

NORMALISED VELOCITY 5 NORMALISED VELOCITY F

SLIDING VELOCITY v (m/s) SLIDING VELOCITY v (m/s)

JAxls of sliding velocity Idimension=m/sl

10-2 1 102 10‘

Heat sink (clamping mechanism I.

T,, = T, + AK. (4)

We calculate a from the time-dependent equations SLIDING VELOCITY v (m/s)

produces a surface temperature 0 JcJImScn

Some authors (e.g. Tabor [65]) have supposed that

Table 2. Material urooerties

the sink temperature T’, is well approximated by the

surfaces through the circles of contact. The contact

2r. Finally, if oxide forms on the asperity tips (as it does

SLIDING VELOCITY v (m/s) SLIDING VELOCITY v (m/s)

NORMALISED VELOCITY v NORMALISED VELOCITY V

at 7 f 3-a number which agrees well with the equa-

Table 3. The simulified temuerature equations

Tf= T,[I -(;>‘“]+ To(;>‘“+~(;)“‘i equation(AS)

With the followingdefinitions

T’ = $ E 222 K for steel equation(7a)

T: = ? e 650 K for steel equation(A7)

p+; equation (7b)

/I =0.78-0.131og,,(fi) equation (9)

This is Tabor’s equation [65]; he and others (e.g.

Johnson [80]) find that a value of the constant LX,

nation for Tabor’s equation [65] for seizure [equation

5.1. I. The model for seizure. When two surfaces

(10-1000 m/s) the coefficient of friction drops to very

5.1.2. Calibration of the seizure model. Figure 18

5.2. Melt wear

5.3. Oxidation-dominated wear

originates in the work of Quinn [102] who marshalls

thickens until, at a critical thickness (about 10 pm for

mine, from experimental wear rates, the values of the

k,= A,exp -3 (27)

energy for oxidation, R is the gas constant and T is

z2 = C=k,t factor [equation (27)] with T = T,, giving

tYoshimoto and Tsukizoe [lOS] and Tenwick and Earles

l+ 5.4. Plasticity-dominated wear: adhesion and delami-

Heatinput 5 uq J/m’s \ At sliding velocities below about 0.1 m/s. surface

SLIDING VELOCITY v (m/s) component so that a correlation should exist between

depth X0 needed to produce this failure is Y*, then

The “life” of the surface is related to the number

where 1 is the distance between asperity contacts. But

C-4 n*=- _fz (37)

Fig. 25. An ideahsed plasticity-dominated (delamination) @-2Y”f,&k p

Table 4. The simplified wear-rale equations