387

Archard wear and component geometry

J J Kauzlarich1 and J A Williams2¤

1
Department of Mechanical Engineering, University of Virginia, Charlottesville, Virginia, USA
2
Department of Engineering, University of Cambridge, UK

Abstract: The wear performance of tribological material pairs is commonly assessed by experimental
investigations which make use of rigs incorporating a pin-on-disc geometry. The information so generated
is conventionally interpreted in terms of a linear relation between wear volume, applied load and sliding
distance in accord with the Archard wear equation. An appropriate material parameter is thus the so-called
dimensional wear rate usually measured in mm3 =N m. However, there are many practical situations in
which, as components wear, the area of apparent contact changes so that, although the Archard relation
may still be applicable on the microscale, the relation between either the macroscopic wear dimension, or
the total wear volume, may be other than a linear function of sliding distance or load. The following
examples are considered: a spherically ended pin sliding or rotating against a flat, a pair of crossed
cylinders, a radially loaded journal bearing and a radially loaded spherical bearing. Data from the wear
literature are examined in the light of the analyses presented.

Keywords: Archard wear, component geometry

NOTATION whereas premature wear-out of the same element is highly

undesirable. In order to accommodate this behaviour,
b radius ratio ˆ R=r
engineering designers need access to predictive governing
B bearing or chord length
equations which take account of practical material combi-
Kw dimensional wear constant
nations, realistic component geometries and appropriate
L sliding distance
operating conditions.
N number of revolutions
There has been controversy concerning wear theory for
P nominal bearing pressure
over a century and it is improbable that a single governing
r small cylinder or sphere radius
equation will be established to cover all eventualities.
R larger cylinder or sphere radius
However, it has been demonstrated that for a wide range of
t time
conditions, both ‘adhesive’ and ‘abrasive’, the wear volume
V volumetric wear
V, load W and sliding distance L are related by an equation
W load
of the form
á, â, ç surface contact angles
á0 angle where the sphere and the bearing coincide V ˆ K w WL (1)
â0 angle where the journal and the bearing coincide
¢ dimensional wear where Kw is termed the dimensional wear coefficient and is
ì stress coefficient conventionally quoted in units of mm3 =N m. An equation
ó stress of this form was postulated by Preston [1] in relation to the
ö rotational speed abrasive polishing of glass and is consequently known as
the Preston equation to those involved with this and similar
technologies. It was likewise proposed by Archard and
1 INTRODUCTION Hirst [2] and Archard [3] in relation to adhesive wear and
extended by Rabinowicz [4] to more general cases of
Wear is a commonplace phenomenon. In the field of abrasive wear and is often known as the Archard equation.
machine design and operation, the ‘running-in’, or ‘wear- The numerical value of the ‘constant’ Kw can vary
ing-in’, of a rubbing machime element is often beneficial enormously; situations of very mild wear generate values
less than 10¡9 mm3 =N m while values of 10¡4 or more
The MS was received on 26 July 2000 and was accepted after revision for would generally be considered unacceptably severe for
publication on 6 December 2000.
¤ Corresponding author: Department of Engineering, University of Cam- most tribological situations.
bridge, Trumpington Street, Cambridge CB2 1PZ, UK. Although some progress has been made in relating the
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
388 J J KAUZLARICH AND J A WILLIAMS

value of Kw to more fundamental or intrinsic material harder or more aggressive material (so concentrating the
properties and relevant operating conditions [5], the most wear on the softer lower member at a single wear scar) or
significant sources of guidance to practical values are can be the softer material itself so that the test procedure
experimental investigations. Most of these are carried out generates a circumferential wear track. Figure 1c illustrates
using a simple pin-on-disc arrangement illustrated in Fig. a simple journal bearing arrangement in which the rotating
1a where the prismatic pin is a specimen of the wearing journal is gradually producing an axial groove in the softer
material and the disc represents the more wear-resistant bearing sleeve and Fig. 1d represents a spherical bearing in
counterface. Wear volume is assessed either by the change which a hard sphere rotates in a spherical housing or cup.
in the length of the pin or by its loss in mass or volume, the In each case, although on the local scale the wear
load W is usually applied as a dead weight and the length behaviour of the softer material might be governed by a
of the sliding path is simply related to the number of relationship between local loads and sliding distances of
revolutions of the disc. Provided that the wearing surface the form of equation (1), it is not immediately obvious that
remains parallel to the abrading disc this simple arrange- this linear relation will be exhibited by the contact as a
ment has the advantage that the apparent contact area and whole. In addition, to the design engineer or end user it is
thus the nominal pressure on the wearing pin remain often the rate at which the dimensional tolerance of the
constant during the course of the test. However, in order to contact grows (rather than the volumetric loss) that is of
ease alignment problems during setting up, the pin is often significance. It is this that dictates the rate at which the
given a spherical tip which gradually wears away; during bearing loosens with use, so generating a consequential
this ‘running-in’ process the area of the contact is growing loss of location accuracy in any device of which the
while load remains the same so that, even if the Archard wearing element is a part.
law is obeyed on the very local scale, the overall di-
mensional loss will no longer grow linearly with increasing
sliding distance.
There are many other situations, of both potential test
geometries and practical devices in the field, where the 2 SPHERICAL ENDED PIN OR PIVOT
apparent contact area changes with wear and/or sliding
distance. Three examples are illustrated in Figs 1b to d. Consider a spherical slider of radius r carrying a normal
Figure 1b represents a crossed-cylinder test geometry in load W sliding against a plane surface. Suppose that it is
which the upper cylinder rotates. This either can be of the the sphere that wears and that at any point on the interface
between the two materials the Archard wear law is obeyed
with a constant dimensional wear coefficient of magnitude
Kw .
Working in terms of the angle â0 defined in Fig. 13 of
Appendix 1,

¢
ˆ 1 ¡ cos â0 (2)
r

and

V ð
3
ˆ (2 ¡ 3 cos â0 ‡ cos3 â0 ) (3)
r 3

where V is the volume lost by wear from the sphere.

Furthermore, it can be shown (see Appendix 1) that

K w PN ˆ 13(2 ¡ 3 cos â0 ‡ cos3 â0 ) (4)

in which the nominal bearing pressure P ˆ W =(ð r 2 ) and

N ˆ L=r where L is the total sliding distance.
If â0 is small, then trigonometrical simplifications
reduce these equations to

¢ p
º K w PN (5)
r
Fig. 1 Wear test geometries: (a) pin-on-disc; (b) crossed cylin-
ders; (c) journal bearing; (d) spherical or revolute bearing while
Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 389

V the parameter Kw PN both computed from equations (2) to

ˆ ð K w PN (6)
r3 (4) (with angle â0 increased in 58 steps from 58 to 908) and
from both the small- and the large-angle asymptotes, i.e.
at all times. While there is a linear relation between volume equations (5) and (6) and equations (8) and (9). In this
(or mass loss) and sliding distance, namely particular case, â0 can be eliminated between equations (2)
and (4) to provide the exact relation
V ˆ K w WL (7)
 ´ ´ 2
¢=r ¢
the relation between the dimensional wear ¢ and either 1¡ ˆ K w PN
3 r
load or sliding distance is, initially, one of a square root
dependence.
Values of â0 > ð=2 correspond to conditions when the The validity of this approach can be tested by comparing
upper hemisphere of the slider is wearing away. If the the outcome both with direct experimental observation and
situation is one in which the slider is prismatic, then, once with results computed by numerical simulation. In a recent
it has worn flat, the corresponding equations become publication, Pödra and Andersson [6] have examined the
initial rate of dimensional loss in a simple pin-on-disc test
¢ in which a sacrificial pin with an initial tip radius of 5 mm
/ K w PN (8)
r was slid under a constant normal load of either 21 or 50 N
against a harder steel substrate. This process was modelled
with the transition occurring when â0 ˆ ð=2, while by an incremental finite element analysis (FEA). Figure 4
shows a comparison of both their computed dimensional
V wear curves and their measured data points for the two load
ˆ ð K w PN (9)
r3 values with that generated by equation (5) above. The mean
wear coefficient used in both the FEA simulation and the
Note that equations (6) and (9) are identical. Figures 2 and application of equation (5) was K w ˆ 1:33 3 10¡13 Pa¡1 ,
3 show the variations in ¢=r and V =r 3 respectively with i.e. 1:33 3 10¡4 mm3 =N m.

Fig. 2 Dimensional wear ¢=r as a function of the load parameter Kw PN for a sphere or spherically ended pin
sliding against a flat counterface: s , computed from equations (2) and (4) for values of â0 from 58 to 908 at
58 intervals; - - -, small- and large-angle asymptotes [equations (5) and (8)] referred to in the text

J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
390 J J KAUZLARICH AND J A WILLIAMS

Fig. 3 Volumetric wear V =r 3 as a function of the load parameter Kw PN for a sphere or spherically ended pin
against a flat counterface: h , computed from equations (3) and (4) for values of â0 from 58 to 908 at 58
intervals; - - -, the asymptote equations (6) and (9) referred to in the text

2.1 Spherical pivot V

º 2K w PN (13)
Br 2
If the pin rotates about its own axis instead of sliding, and
if N is now the total number of revolutions completed by where P ˆ W =(2Br).
the pin, equations (5) and (6) are replaced by

¢
º 0:5(3ð)2=3 (K w PN )2=3 (10)
r
3 CROSSED CYLINDERS
and
Of the many ways to determine wear and friction of
V =r 3 º 0:25(34 ð7 )1=3 (K w PN )4=3 (11) materials, the crossed-cylinders geometry is one of the
simplest. As wear progresses, both the area of contact and
These are identical with the corresponding equations the rubbing distance change; this situation can be analysed
derived by Kauzlarich et al. [7] for this geometry and given in an analogous fashion to that of a sliding wear pin (see
as equations (10) and (13) of reference [7]. Appendix 2). The situation considered is shown in Fig. 14
in which the upper harder cylinder of radius r rotates with
2.2 Sliding cylindrical contact angular speed ö wearing the lower softer cylinder which
has radius R. The dimension ¢ now represents the
In the case of a softer cylinder, of length B, sliding in a approach of the axes of the cylinders and V is the worn
direction normal to its axis then the appropriate expressions volume.
for the initial dimensional and volumetric wear rates are In terms of the angle â0 shown in the figure,
respectively

¢ ¢
º 0:5 3 91=3 (K w PN )2=3 (12) ˆ 1 ¡ cos â0 (2)
r r

and but now

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 391

Fig. 4 Comparison of experimental data and FEA analysis (both from reference [6]) with simplified small-angle
asymptotes at loads of (a) 21 N and (b) 50 N; in both cases, K w ˆ 1:33 3 10¡4 mm3 =N m

… â0 W W
V Pˆ ˆ
ˆ 2b2 f (â) cos â dâ (14) 4rR 4br 2
(15)
r3 0

then
where … â0 … â0
1 B
b ˆ R=r K w PN ˆ cos2 â sin â0 dâ dâ0 (16)
2ð b 0 0 2r
and
where B=(2r) ˆ [b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2 and N is
f (â) ˆ arcsin (1=b)[b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2
the number of revolutions completed by the rotating
¡ (1=b)2 (b ‡ cos â0 ¡ cos â) cylinder.
3 [b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2 If the angle â0 is small, then equations (2), (14) and (16)
can be simplified, so leading to elementary analytical
If P is again the nominal bearing pressure here defined expressions for ¢ and V:
as the normal load W divided by the projected area ¢ pp
according to º 2 2b1=2 K w PN (17)
r
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
392 J J KAUZLARICH AND J A WILLIAMS

and obeys the local Archard wear equation. Referring to Fig. 15

again to define the angle â0 and setting b ˆ R=r, then
V
º 8ðbK w PN (18) ¢
r3
ˆ (1 ¡ cos â0 ) ¡ b(1 ¡ cos á0 ) (19)
r
Figures 5 and 6 show the variations in ¢=r and V =r 3

respectively with the parameter Kw PN both computed from where sin á0 ˆ (sin â0 )=b, and
equations (2), (12) and (14) (â0 increasing in 58 steps from
58 to 908) and from the small-angle asymptotes [equations V
ˆ (â0 ¡ sin â0 cos â0 ) ¡ b2 (á0 ¡ sin á0 cos á0 )
(17) and (18)]. Values of b have been chosen as 1, 2 and 5. Br 2
(20)

where B is now the axial length of the bearing sleeve. If the

local Archard wear coefficient is again Kw , the nominal
4 JOURNAL BEARING pressure P is

The radial wear rate for journal or sleeve bearings under W

Pˆ (21)
unlubricated conditions is especially important as it often 2Br
determines their effective useful life. Halling [8, p. 140]
has derived an equation for the rate of radial wear, starting and N is the number of revolutions made by the journal,
with the assumption that the volume wear rate is propor- then
tional to the rate of frictional energy dissipation at the
… â0
rubbing interface. With such an assumption, he showed that 1
K w PN ˆ [â0 ‡ 0:5 sin(2â0 )]
the rate of radial wear, i.e. the rate at which the dimension 4ð 0
¢ grows (see Fig. 15) is proportional to the product of  ´
radial load and linear or rotational velocity. However, the sin â0 cos â0
3 sin â0 ¡ d â0 (22)
influence of the bearing clearance on either ¢ or V, the b2 ¡ 1 ‡ cos2 â0
corresponding volumetric wear, was not considered.
In the derivation in Appendix 3 it is supposed that wear If â0 is small, then equations (19), (20) and (22) reduce to

Fig. 5 Dimensional wear ¢=r as a function of the load parameter Kw PN for a crossed cylinder wear test: n , s , d ,
computed from equations (2) and (16) for values of â0 from 58 to 908 at 58 intervals; - - -, small-angle
asymptotes [equation (17)] for indicated values of the diameter ratio b

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 393

Fig. 6 Volumetric wear V =r 3 as a function of the load parameter Kw PN for a crossed-cylinder wear test: , , h , j ,
computed from equations (14) and (16) for values of â0 from 58 to 908 at 58 intervals; - - -, asymptotes
[equation (18)] for indicated values of the diameter ratio b

 ´1=3  ´1=3
¢ ð2 1 Figures 7 and 8 show the variations in ¢=r and V =(Br 2 )
º 9 1¡ (K w PN )2=3 (23) respectively with the parameter Kw PN. Both of these are
r 2 b
computed from equations (19) to (21) (in steps 58 from 58
to 908) and from the small-angle asymptotes (23) and (24)
and and the large-angle asymptotes (25) and (26). Values of b
equal to 1.003, 1.1 and 2 have been chosen. The gradual
V change in the slope of the log–log plot of ¢=r versus
º 4ð K w PN (24)
r3 Kw PN from 2=3 to unity is clear.
These results will be applicable to journal bearings
On the other hand, when the angle â0 is large and, in the generally in which the journal is much harder and wear
limit, approaches ð=2, a similar treatment leads to the resistant than the surrounding sleeve; this is the usual
corresponding asymptotes preferred arrangement. Clearly, in the case of plastic
bearings, or other situations in which the bearing first
¢ forms a transfer film on the journal, the results will be
º 8K w PN (25)
r restricted to steady state conditions after this film has been
established.
and Data from tests of this sort have recently been published
by Trevüz [9] and are reproduced in Fig. 9 for nominal
V bearing pressures of 125 and 500 kPa. The journal was of
º 16K w PN (26)
Br 2 440 stainless steel with diameter 40 mm and the bearing
sleeve a polytetrafluoroethylene (PTFE)–carbon composite
Both of these are independent of the radius ratio b and material. The axial length B was equal to 50 mm and the
correspond to a simple ‘tunnelling’ relation as the journal initial clearance such that b ˆ 1:003. Initially, in both
moves into the bearing: cases, there was a region of relatively high wear, associated
with the formation of a stable transfer film as part of the
V ¢ wearing-in or running-in process, after which the rate of
ˆ2 (27) mass loss became linear with sliding distance in accord
Br 2 r
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
394 J J KAUZLARICH AND J A WILLIAMS

Fig. 7 Dimensional wear ¢=r as a function of the load parameter Kw PN for journal bearing geometry: d , m , h ,
computed from equations (19) and (22) for values of â0 from 58 to 908 at 58 intervals; - - -, asymptotes
[equations (23) and (25)] referred to in the text for the values of b indicated

Fig. 8 Volumetric wear V =(Br 2 ) as a function of the load parameter Kw PN for journal bearing geometry: j , m , d ,
h , computed from equations (20) and (22) for values of â0 from 58 to 908 at 58 intervals; - - -, asymptotes

[equations (24) and (26)] referred to in the text

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 395

V ð
ˆ [(2 ¡ 3 cos â0 ‡ cos3 â0 )
r3 3
¡ b3 (2 ¡ 3 cos á0 ‡ cos3 á0 )] (29)

and
… â0
2
K w PN ˆ I(sin â0 ¡ tan á0 cos â0 ) dâ0 (30)
ð2 0

„â „ç
where I ˆ 0 0 0 cos2 å cos3 â då dâ and cos ç ˆ (cos â0 )=
(cos â). When the angle â0 is small, these reduce to
s
 ´
¢ 1 p
º ð 1¡ K w PN (31)
r b

Fig. 9 Wear, measured as mass loss, versus sliding distance for

carbon-filled PTFE bearing sleeves (data from Trevüz and
[9]). The indicated angles are calculated values of â0
V
º ð2 K w PN (32)
r3

On the other hand, when â0 approaches 908, the situation is

with the Archard relation. Taking the density of the sleeve one in which the sphere ‘tunnels’ into the seat. The
material as 2200 kg=m3 , the slopes of these experimentally corresponding large-angle asymptotes are
determined curves can be used to evaluate an effective
value of Kw (equal to 7 3 10¡7 mm3 =N m) and then ¢
º 3ð K w PN (33)
equation (22) utilized to investigate the way in which the r
angle â0 changes with sliding distance; these values are
shown on the two experimental curves. As the bearing and
wears, so the intervals in sliding distance (or time) for
equal intervals of â0 increase; for this angle to become V
º 3ð2 K w PN (34)
equal to 908, sliding distances of 3318 and 13 275 km are r3
required for the two loading cases respectively.
Under some circumstances and with some materials, as Figures 10 and 11 illustrate the match between these
the bearing wears under repeated loading, the values of ¢ approximate solutions and the exact solutions represented
or V may suddenly increase as fatigue or delamination wear by equations (28), (29) and (30). Values of b equal to
is initiated. The equations developed here are then no 1.003, 1.02, 1.1, 2, 5 and 20 have been chosen. Once again
longer immediately applicable to this change in wear in the plot of ¢=r versus the group Kw PN there is a clear
regime. transition from the initial region, where in this case the
slope is one-half, to a linear region at greater wearing
distances. It can be noted that eliminating the group Kw PN
between equations (33) and (34) (where â0 is large)
5 SPHERICAL BEARING recovers the conventional tunnelling equation

As a final example, consider the case of a loaded ball, of V ¢

3
ˆð (35)
radius r, supported in a spherical housing or cup which has r r
a radius R equal to br. The normal load is again W and N is
now the number of revolutions of the sphere. Again, on the Dowson et al. [10] have presented both volumetric and
basis of local Archard wear conditions (Appendix 4) and dimensional wear data from tests on a hip prosthesis in
referring to Fig. 16, it is shown that which an alumina femoral head was reciprocated in an ultra-
high molecular weight polythene (UHMWPE) acetabular
¢ cup. Figure 12 shows a plot of the penetration of their heads
ˆ (1 ¡ cos â0 ) ¡ b(1 ¡ cos á0 ) (28)
r number 1 and 2 into the cup, i.e. the extent of dimensional
wear as a function of the number of loading cycles during
the volumetric wear loss V is given by each of which the head moved through an angle of 558. Six
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
396 J J KAUZLARICH AND J A WILLIAMS

Fig. 10 Dimensional wear ¢=r as a function of the load parameter Kw PN for spherical bearing geometry: n , s ,
, , ¤, e , computed from equations (28) and (30) for values of â0 from 58 to 908 at 58 intervals; - - -,
asymptotes [equations (31) and (33)] referred to in the text

Fig. 11 Volumetric wear V =r 3 as a function of the load parameter Kw PN for spherical bearing geometry: n , s , , ,
¤, e , computed from equations (29) and (30) for values of â0 from 58 to 908 at 58 intervals; - - -,
asymptotes [equations (32) and (34)] referred to in the text

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 397

Fig. 12 Penetration of an alumina femoral head into an UHMWPE acetabular cup: m , d , j , from reference [10];
——, computed from equations (28) and (30)

data points are shown. The difference between the radii of will be related to the total applied load or global sliding
the head and cup corresponded to a value of b ˆ 1:02 and history. This is because the interfacial pressure is not
the mean pressure P to a value of 3.62 MPa. The solid curve known a priori but will rather adjust itself as wear proceeds
is a plot of equation (31) using these values together with a so that both the demands of force equilibrium are satisfied
value of Kw equal to 4:58 3 10¡7 mm3 =N m. The experi- and the overall wear geometry is consistent with the
mentally determined mean value is very close, being quoted maintenance of geometric compatibility between both
as 4:49 3 10¡7 mm3 =N m. Values of the angle â0 , estab- elements of the tribological pair. In practice, it may also be
lished from equation (28), are indicated by the full triangles that the dimensional wear rate, i.e. the rate at which the
on the computed curve in Fig. 12. clearance in the device increases, is of greater practical or
economic significance than the overall loss of volume,
although, of course, these must be geometrically related.
In this paper it is demonstrated for all the cases
6 DISCUSSION AND CONCLUSIONS examined, i.e. a spherically ended pin rubbing on a flat
surface, a pair of crossed cylinders, and both journal and
The resistance of a material to wear is often assessed using spherical bearings, that, at any stage from initial contact to
a test geometry which is different from those of real final steady state conditions, the short-time volumetric
components that might be manufactured from it. In wear loss increases linearly with both increasing load and
particular, wear tests are often carried out using either a increasing incremental sliding distance. However, in the
pin-on-disc arrangement, in which the test material is case of crossed cylinders, journals and spherical bearings
the sacrificial pin, or a crossed-cylinders set-up where the the leading coefficient in this relation is not constant so
stationary test piece is worn by a rotating cylinder of the that
harder counterface material. In order to translate empirical
data from such tests to field conditions involving sliding Total volume loss ˆ f (sliding distance)
contacts, such as occur in journal or spherical bearings, the
resistance of the material to dry wear is conventionally 3 (K w 3 load 3 sliding distance)
expressed in units of volume lost per unit sliding distance
per unit normal load. This is often known as the If the evolution of this factor (which can be as great as 3)
dimensional wear resistance Kw . This can be normalized by with changing geometry is not taken into account, then
dividing by the value of the material hardness to give a experimental and field data may appear to have a much
non-dimensional wear resistance. greater scatter than, in reality, they possess.
Even if every element at the loaded and wearing In addition, when wear is expressed in terms of
interface in the field application behaves locally in dimensional loss, then in all these cases the initial rate of
accordance with the Archard wear law, i.e. a linear relation dimensional wear growth is dependent on the group
between incremental wear volumes and local loads and (K w 3 load 3 sliding distance) raised to some power
sliding distances, it is not immediately obvious how the which is less than unity. The assumption of linearity
overall or global volumetric wear loss from the component between wear and either load or sliding distance during this
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
398 J J KAUZLARICH AND J A WILLIAMS

‘running-in’ stage can lead to very significant errors in APPENDIX 1

estimates of values of Kw .
The relationships developed are substantially verified by
Spherically ended pin on a flat surface
appeal to a number of wear studies in the literature. The
frictional torques or sliding resistances have not been Figure 13 shows the geometry. Assume initially that all the
evaluated although these can be readily determined from a wear takes place on the sphere which is of radius r and that
knowledge of the local interfacial stress distribution and there is no loss of material from the rotating disc. If ¢ is
geometry which form intrinsic elements of the analysis the wear dimension, then
presented following the methods suggested by Kauzlarich
et al. [7]. Finally, it should be noted that this paper does not ¢
ˆ 1 ¡ cos â0 (36a)
take into account any fundamental changes in wear r
mechanism that may occur during the life of the component
due, for example, to the generation of transfer films on the and the volumetric wear V is given by
counterface or the loss of material through some form of
fatigue or delamination mechanism which requires the V ð
3
ˆ (2 ¡ 3 cos â0 ‡ cos3 â0 ) (36b)
accumulation, with time, of subsurface damage. r 3

If the Archard wear coefficient Kw is the same at all points

across the interface, then this must remain flat and parallel
to the harder disc surface; it follows that the pressure p
across the worn flat must be uniform:
ACKNOWLEDGEMENTS
W
The authors would like to express their appreciation for the pˆ sin2 â0
ð r2
comments and suggestions offered by Dr J. A. Greenwood
and Professor K. L. Johnson. and, in accordance with the Archard relation (1), that

d¢ ð r 2 sin2 â0 ˆ K w W dL (37)

where L is the sliding distance. Thus, differentiating

REFERENCES equation (36a) to obtain a second expression for d¢ and
equating this to equation (37) gives
1 Preston, F. W. Theory and design of glass polishing … â0
machines. J. Soc. Glass Technol., 1927, 11, 214–256.
K w WL ˆ ð r 3
sin3 â0 dâ0
2 Archard, J. F. and Hirst, W. Wear of metals under 0
unlubricated conditions. Proc. R. Soc. Lond. A, 1956, 236,
3–55.
3 Archard, J. F. Contact and rubbing of flat surfaces. J. Appl. If P is defined as W =(ð r 2 ) and N as L=r, then
Physics, 1953, 24, 981–988.
4 Rabinowicz, E. Friction and Wear of Materials, 1965 (John
Wiley, New York).
5 Williams, J. A. Wear modelling: analytical, computational
and mapping: a continuum mechanics approach. Wear, 1999,
225–229, 1–17.
6 Pödra, P. and Andersson, S. Simulating sliding wear with
finite element method. Tribology Int. 1999, 32, 71–81.
7 Kauzlarich, J. J., Bhatia, K. G. and Streitman, H. W. Effect
of wear on pivot thrust bearings. Trans. Am. Soc. Lubric.
Engrs, 1966, 9, 257–263. Correction: Fig. 4 is incorrect;
replot equation (13) and correct data for run-in after
5:9 3 105 rev by 4:4 in3.
8 Halling, J. Principles of Tribology, 1978 (Macmillan, Lon-
don).
9 Trevüz, T. Tribological behaviour of carbon filled polytetra-
fluoroethylene (PTFE) dry journal bearings. Wear, 1998, 221,
61–68.
10 Dowson, D., Jobbins, B. and Seyed-Harraf, A. An evaluation Fig. 13 A spherically (or cylindrically) ended pin of radius r
of the penetration of ceramic femoral heads into polythene carrying load W wearing against a flat counterface. The
acetabular cups. Wear, 1993, 162–164, 880–889. shaded area of depth ¢ represents the worn volume V

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 399

… â0
appropriate expressions for dimensional and volumetric
K w PN ˆ sin3 â0 dâ0
0
wear become respectively

ˆ 13(2 ¡ 3 cos â0 ‡ cos3 â0 ) (38) ¢

º 0:5 3 91=3 (K w PN )2=3 (45)
r
and thus
and
V
ˆ ð K w PN (39)
r3 V
ˆ 2K w PN (46)
Br 2
When the angle â0 is small, equations (36a) and (38)
reduce to where P ˆ W =(2Br) and, again, N ˆ L=r.
¢
º (K w PN )1=2 (40)
r
APPENDIX 2

In this particular case, equations (36a) and (38) can be

Crossed cylinders
combined to give the exact relation between the dimen-
sional wear and the group Kw PN: Figure 14 shows the geometry of the situation under
 ´ ´ consideration. It will be assumed that all the wear takes
¢=r ¢ 2 place on the stationary lower cylinder of radius R and that
1¡ ˆ K w PN (41)
3 r there is no loss of material from the upper rotating cylinder
which has radius r. Applying the Archard wear equation (1)
locally to a different element at the rubbing interface yields
Spherical pivot
dV ˆ r dâ B cos â d¢
If, instead of sliding, the wearing pin rotates about its own
axis, then, while there will be no change either in the ˆ K w ó r dâ Brö dt (47)
relation between the worn volume V and the angle â0 or in
equation (36a), the distance slid by any element depends where B and â are the dimension and the angle respectively
on r, its radial distance from the centre-line. It follows that indicated in Fig. 14; ó is the local interfacial normal stress.
equation (39) is replaced by The dimension B is given by the equation
… â0
B
K w PN ˆ sin2 â0 dâ0 ˆ [b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2 (48)
0 2r

ˆ 12[â0 ¡ 0:5 sin(2â0 )] (42) in which b is the radius ratio R=r and â0 is the maximum
value of angle â.
where N is now the total number of revolutions. If the angle Equation (47) can be rearranged as
â0 is small, then these equations reduce to
r
¢ d¢ ˆ K w ó ö dt (49)
º 0:5(3ð)2=3 (K w PN )2=3 (43) cos â
r
However, in addition, because the vertical displacement d¢
and must be independent of the chosen angle â, the local
normal stress ó must be proportional to cos â:
V
º 0:25(34 ð7 )1=3 (K w PN )4=3 (44)
r3 ó ˆ ì cos â

These are identical with the corresponding equations It thus follows from equation (49) that d¢ ˆ K w ì rö dt
derived by Kauzlarich et al. [7] for this geometry and given which, if N represents the number of revolutions that
as equations (10) and (13) of reference [7]. cylinder r completes, can be written as

d¢ ˆ K wì r2ð dN (50)
Sliding cylindrical contacts
If Fig. 13 is interpreted as representing a softer cylinder, of The value of the factor ì can be evaluated „from considera-
â
length B, sliding in a direction normal to its axis, then the tions of overall equilibrium, since W ˆ 2r 0 0 ó B cos â dâ,
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
400 J J KAUZLARICH AND J A WILLIAMS

Fig. 14 Geometry of the crossed-cylinder wear test. The upper harder cylinder of radius r rotates against the
stationary lower cylinder of radius R ˆ br, so generating a wear scar of depth ¢

„ â0
i.e. 2rì 0
B cos2 â dâ, where B comes from equation (48), f (â) ˆ arcsin(1=b)[b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2
so that
¡ (1=b)2 (b ‡ cos â0 ¡ cos â)
… â0
ìˆ
W
B cos2 â dâ 3 [b2 ¡ (b ‡ cos â0 ¡ cos â)2 ]1=2
(51)
2r 0

Equating the two equatons (50) and (52) for d¢=r provides
the relation
Now the relative displacement of the two cylinders, ¢, is
… â0 … â0
related to the angle â0 simply by 1 B
K w PN ˆ cos2 â sin â0 dâ dâ0 (54)
ð
2 b 0 0 2r
¢
ˆ 1 ¡ cos â0 (36a)
r
where P is the nominal pressure on the contact:

so that W W
Pˆ ˆ (55)
4rR 4br 2
d¢
ˆ sin â0 dâ0 (52) For a given value of the radius ratio b, equations (36a), (53)
r
and (54) can be evaluated for a range of the angle â0 and
plots drawn of the dimensional wear ¢=r and volumetric
and the volumetric wear loss V is given by wear V =r 3 versus the group Kw PN.
Asymptotic solutions to these equations can be readily
… â0 obtained for small values of the angles â and â0 so that
V
ˆ 2b2 [á ¡ 0:5 sin(2á)]cos â dâ dimensional and volumetric wear rates can be expressed in
r3 0 terms of the group Kw PN directly; they are

¢ pp
º 2 2b1=2 K w PN (56)
where the angle á ˆ arcsin[B=(2R)] so that r

… â0 and
V
ˆ 2b2 f ( â)cos â dâ (53)
r3 0 V
º 8ð bK w PN (57)
r3

where respectively.
Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 401

APPENDIX 3 where N represents the number of revolutions that the

cylinder of radius r completes.
The value of the factor ì can be evaluated from
Journal bearing geometry
considerations of overall equilibrium; since
Figure 15 shows the geometry of the situation under
… â0
consideration. It will be assumed that all the wear takes
W ˆ 2Br ó cos â dâ
place on the stationary bearing sleeve which has radius R 0
and that there is no loss of material from the rotating
… â0
journal which has radius r. Applying the Archard wear
ˆ 2Brì cos2 â dâ
equation (1) locally to a differential element at the rubbing 0
interface yields
so that
dV ˆ r dâ B cos â d¢
W
ìˆ [â0 ‡ 0:5 sin(2â0 )]
ˆ K w ó r dâ Brö dt (58) Br

where B is the axial length of the rubbing contact and â is it follows that
the angle indicated in Fig. 15; ó is the local interfacial
d¢ W
normal stress. ˆ K w2ð dN [â0 ‡ 0:5 sin(2â0 )] (61)
Equation (58) can be rearranged to give r Br

r Now the relative displacement of the two cylinders, ¢, is

d¢ ˆ K w ó ö dt (59) related both to the angle â0 and to b, the ratio of the radius
cos â
of the bearing to that of the journal, i.e. b ˆ R=r, by the
equation
However, because the vertical displacement d¢ must be
independent of the chosen angle â, the local normal stress ¢
ó must be proportional to cos â: ˆ (1 ¡ cos â0 ) ¡ b(1 ¡ cos á0 ) (62)
r

ó ˆ ì cos â where sin á0 ˆ (sin â0 )=b so that

It thus follows from equation (59) that d¢ ˆ K w ì rö dt, d¢

ˆ sin â0 dâ0 ¡ b sin á0 dá0 (63)
which can be written as r

d¢ ˆ K w ì r2ð dN (60) Similarly, the volumetric wear loss V is given by

V
ˆ (â0 ¡ sin â0 cos â0 ) ¡ b2 (á0 ¡ sin á0 cos á0 )
Br 2
(64)

Equating the two expressions (61) and (63) for d¢=r

provides the relation
…â
1 0
K w PN ˆ [â0 ‡ 0:5 sin(2â0 )]
4ð 0
 ´
sin â0 cos â0
3 sin â0 ¡ 2 dâ 0 (65)
b ¡ 1 ‡ cos2 â0

where P is the nominal pressure on the contact:

W
Fig. 15 Geometry of a wearing journal bearing arrangement. Pˆ (66)
The shaft of radius r carries a uniform and stationary
2rB
load W while rotating at speed ö. The dimension ¢
represents the depth by which the journal has penetrated For a given value of the radius ratio b, equations (62) to
the softer bearing sleeve which has radius R ˆ br (66) can be evaluated for a range of the angle â0 and plots
J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J
402 J J KAUZLARICH AND J A WILLIAMS

drawn of the dimensional wear ¢=r and volumetric wear that there is no loss of material from the rotating sphere
V =(Br 2 ) versus the group Kw PN. which has radius r. If R=r ˆ b, then the subtended angles
Asymptotic solutions to these equations can be readily á0 and â0 are related by
obtained for small values of the angle â0 so that
dimensional and volumetric wear rates can be expressed in sin â0 ˆ b sin á0
terms of the group Kw PN directly; they are
 ´1=3  ´1=3 The relative displacement ¢=r of the sphere into its seat is
¢ 9ð2 1
º 1¡ (K w PN ) 2=3
(67) related to the angles â0 and á0 and the radius ratio b simply
r 2 b by

¢
and ˆ (1 ¡ cos â0 ) ¡ b(1 ¡ cos á0 ) (71)
r
V
º 4ð K w PN (68)
Br 2 and the volumetric wear loss V is given by

respectively. V ð
ˆ [(2 ¡ 3 cos â0 ‡ cos3 â0 )
When the angle â0 is large and in the limit approaches r 3 3
ð=2 a similar treatment leads to the corresponding
¡ b3 (2 ¡ 3 cos á0 ‡ cos3 á0 )] (72)
asymptotes

¢
º 8K w PN (69) Applying the Archard wear equation (1) locally to a
r differential element at the rubbing interface yields

and ó
d¢ ˆ K w rö dt (73)
cos cos â
å
V
º 16K w PN (70)
Br 2
where å and â are the angles indicated in Fig. 16; ó is the
local interfacial normal stress and ¢ is the penetration of
which corresponds to a simple tunnelling relation.
the sphere.
However, because the displacement d¢ must be
independent of the particular values of angles â and å, the
local normal stress ó must be proportional to cos å cos â:
APPENDIX 4
ó ˆ ì cos å cos â
Sphere in a spherical seat
Figure 16 shows the geometry. It will be assumed that all It thus follows that d¢ ˆ K w ì r2ð dN , where N is the
the wear takes place on the stationary cup of radius R and number of revolutions that the sphere of radius r completes.

Fig. 16 Geometry of a wearing spherical bearing. The sphere of radius r carries a uniform and stationary load W
while rotating at speed ö. The dimension ¢ represents the depth by which the sphere has penetrated the
softer surrounding bearing cup which is of radius R ˆ br

Proc Instn Mech Engrs Vol 215 Part J Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 J04100 # IMechE 2001
 ARCHARD WEAR AND COMPONENT GEOMETRY 403

The value of the factor ì can be evaluated from considera- where P is the nominal pressure on the contact:
tions of overall equilibrium; since
W
… â0 … ç Pˆ (77)
ó cos å cos 2â ð r2
W ˆ 2r 2
då dâ
âˆ0 刡ç

… â0 … ç For a given value of the radius b, equations (71), (72) and

ˆ 4 r2ì cos2 å cos3 â då dâ (74) (76) can be evaluated for a range of the angle â0 and plots
0 0 drawn of the dimensional wear ¢=r and volumetric wear
V =r versus the group Kw PN.
If the subtended angles are small, then asymptotic
so that
expressions are readily obtained from the equations above:
W s
ìˆ  ´
4r 2 I ¢ 1 p
º ð 1¡ K w PN (78)
r b
where
… â0 … ç
and
Iˆ cos2 å cos3 â då dâ
0 0
V
cos â0 º ð2 K w PN (79)
cos ç ˆ r3
cos â
(75)
On the other hand, when â0 approaches 908 the situation is
one in which the sphere ‘tunnels’ into the seat. The
and thus
corresponding large-angle asymptotes are
d¢ ð2
ˆ K w P dN ¢
r 2I º 3ð K w PN (80)
r
Equating the two expressions for d¢=r and integrating
gives and

… â0 V
2 º 3ð2 K w PN (81)
K w PN ˆ I(sin â0 ¡ tan á0 cos â0 ) dâ0 (76) r3
ð2 0

J04100 # IMechE 2001 Downloaded from pij.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 Proc Instn Mech Engrs Vol 215 Part J