Professional Documents
Culture Documents
D. M. Bailey
Research Engineer,
Friction on Contact Stresses
Westland Helicopters Ltd.,
Yeovil, Somerset, England The stress distributions associated with smooth surfaces in contact are rarely ex-
perienced in practice. Factors such as surface roughness, lubricant films, and third
body particulates are known to influence the state of stress and the resulting rolling
R. S. Sayles contact fatigue life. This paper describes a numerical technique for evaluating the
Lecturer, complete subsurface field of stress resulting from the elastic contact of noncon-
Department ot Mechanical Engineering, forming rough bodies, based on measurements of their profile. The effect of sliding
Imperial College of Science, friction is included. The presence of asperities within the contact region gives rise
Technology and Medicine, to high shear stresses near the surface. Realistic coefficients offriction for lubricated
London SW7 England sliding contacts (i.e., ix~0.1) causes the "smooth body" shear stresses to interact
with the asperity stresses to produce a large, highly stressed region exposed to the
surface. The significance of these near-surface stresses is discussed in relation to
modes of surface distress which lead to eventual failure of the contacting surfaces.
Introduction
Metal fatigue is known to depend upon the internal stress However, the stress distributions associated with idealised
field. Nonconforming surfaces such as those of rolling elements smooth surfaces in contact are rarely experienced in practice.
and gear teeth are subject to loads which cause high subsurface Factors such as surface roughness (Berthe et al., 1977), lu-
stresses. During rolling, the cyclic nature of these stresses often bricant films (Kweh et al., 1989), surface furrows and debris
initiates fatigue cracks and subsequent failure by pitting or dents (Webster et al., 1985a and Sayles and Macpherson, 1982),
spalling. are known to influence the state of stress and the subsequent
The classic study of the complete stress field in smooth bodies fatigue life. Failure by rolling contact fatigue is generally con-
in line contact is by Smith and Liu (1953). They included sidered to be a process involving competition between the var-
frictional effects as a tangential load proportional to the Her- ious types of localized stress concentrations, both on the contact
tzian normal pressure distribution. The results showed, in com- surface and at defects within the material (Chiu et al., 1971;
bination with the theory for rolling contact fatigue by Lundberg Tallian and McCool, 1971; and Tallian, 1971).
and Palmgren (1947), that the friction-induced shear stresses This paper examines the state of subsurface stress associated
were of special importance for fatigue. For example, using a with "real" roughness within Hertzian contacts subject to both
coefficient of friction ft = 0.3, Smith and Liu calculated the normal and tangential loads. The analysis assumes elastic be-
maximum reversing shear stress to be at the surface, a feature havior (i.e., no plastically deforming asperities) and the contact
which could greatly reduce fatigue life compared to the fric- is modelled as two-dimensional. Although roughness is gen-
tionless case. erally three-dimensional, it is idealized as two-dimensional,
Hamilton and Goodman (1966) developed equations for the with ridges oriented in the third dimension.
complete stress field beneath a circular sliding contact. They
used the von Mises yield criterion as a measure of the risk of
mechanical failure. The results showed that increasing the coef- Rough Surface Contact Model
ficient of friction from 0 to 0.5 moved the point of yield from The analysis of rough surfaces in contact has, over the past
a depth of 0.5a (where a is the Hertzian half width), to the 30 years or so, progressed from the simple asperity-on-asperity
surface, and that the magnitude of the yield parameter in- approach of Archard (1957) to the complex random process
creased with increasing friction coefficient. Explicit equations models of Bush et al. (1975 and 1979). These models yield
for calculating the stresses within smooth, circular sliding con- important results about the average properties of the contact
tacts have since been given by Hamilton (1983) and Sackfield of rough surfaces. However, due to the statistical nature of
and Hills (1983a). The analysis has been extended to the general the approach, interactions between adjacent asperities are ne-
case of elliptical contacts by Bryant and Keer (1982) and by glected and little information is gained on the real pressure
Sackfield and Hills (1983b), who show that the contact pressure distribution or deformed shape of individual asperities. Recent
for first yield is almost independent of the shape of the contact developments in numerical methods has enabled the contact
ellipse. of rough surfaces to be simulated without the need for statis-
tical modelling.
A numerical contact technique developed by Webster and
Sayles (1986), based on the force-displacement relationship for
Contributed by the Tribology Division for publication in the JOURNAL OF
TRIBOIOOY. Manuscript received by the Tribology Division July 25, 1990; revised a half space, simulates the dry, frictionless, elastic contact of
manuscript received October 1990. Associate Editor: F. E. Kennedy, Jr. a smooth cylindrical body normally loaded against digitized
- 2.0
.. 1.0
25.0 Microns
(0
Q.
Q
Fig. 3 Idealisation of the two-dimensional contact of nonconforming
bodies showing plane of action and notation for subsurface stresses
2TT
(cos 20!-cos 262) Solution Procedure ( lc )
Figure 3 shows the two-dimensional contact of two non-
where tan 8l2 = z : conforming smooth bodies idealized as an elastic half-space
* (x±a) and a cylinder. A computer program has been developed to
The stresses due to a uniform distribution of tangential trac- evaluate ax, az, and TXZ at each subsurface point for any given
tion acting on the strip (b<x<a) are given by: distribution of surface pressure and coefficient of friction.
The main features of the subsurface stress program are rep-
f*Hs - (cos 20i - cos 202) (2a) resented in the form of a flow chart in Fig. 4. The pressure
distribution is read from the results data file produced by the
rough surface numerical contact program previously described.
(cos 20!-cos 202) (2b) Header information giving details of the contact analysis is
2TT
written to the screen. Next the operator specifies the size and
{2(01-02) + (sin20 1 -sin20 2 )J (2c) resolution of the subsurface mesh in the horizontal and vertical
2TT directions. The mesh is automatically positioned centrally be-
where rh2= {(x±a)2 + z?}ln neath the pressure distribution.
Therefore, given a surface pressure distribution consisting The solution proceeds by selecting each subsurface point in
of uniform elements (as produced by the Webster and Sayles turn, on a layer by layer basis and performing the following
contact model) and a frictional relationship to give the tan- calculations:
gential tractions, the principle of superposition can be used to (i) Using equation (1) the contribution to ax, az, and TXZ from
yield the complete subsurface stress field. each of the normal pressure elements at the surface is evaluated.
In this analysis the imposition of tangential traction is as- (ii) If the friction coefficient is greater than zero then using
sumed to have no effect upon the normal pressure. This is true equations (2) the contributions to ax, az, and TXZ from each of
only when the two bodies are smooth or have the same elastic the tangential traction elements at the surface is evaluated.
constants. In the case of a rough-on-rough contact, asperity (iii) Using the summation of the values from (i) and (ii) the
principal shear stress T\ and the principal direct stresses <xi and Ox+Oz
(3c)
a2 can be calculated using equations (3) ov
2-\ 1/2
When the required stresses for a complete layer of the sub-
TI= W + (3a) surface mesh have been calculated the results are saved to the
output datafile. When the complete stress field has been cal-
ffv+ff.
= +T, (3b) culated and saved a comprehensive "header" is appended to
°v the end of the file. This gives details of the original measured
profile, the simulated contact analysis, and the subsurface The complete distribution of orthogonal and principal shear
stress analysis. stresses beneath a smooth elastic Hertzian contact as predicted
The data file are designed to be compatible with the three- by the computer model is shown in Fig. 6. The stresses are
dimensional isometric and contour plotting routines developed displayed as isometric views as well as the more usual contour
at Imperial College (Webster et al., 1985b) for displaying dig- plots. The complex stress distributions from rough surfaces
itised rough surface data; the stress, normalised byp0, replacing are often easier to visualize when displayed as isometric views.
the surface heights in micrometers. Figures 6(d and g) illustrate how sliding friction (/x = 0.1) in-
fluences the distribution of subsurface shear stresses.
Validation of Numerical Results
To assess the accuracy of the subsurface stress model a series Analysis of Rough Surfaces
of test cases using "smooth body" Hertzian pressure distri- The measured surface profiles considered in the subsequent
butions were computed and compared with classical analytical analyses were recorded from disc machine specimens during a
solutions. programme of tests to study the influence of surface roughness
The Hertzian elastic contact stress formulas for the two- on rolling contact fatigue. The specimen material, lubricant
dimensional contact of nonconforming bodies idealized as an and slide to roll ratios were selected to be representative of
elastic half-space and a cylinder, as shown in Fig. 3, are as gear tooth contacts within helicopter transmission systems.
follows; Figure 7 shows two roughness profiles recorded from a ground
/APR specimen before and after testing, and one from an unrun
Semi-contact width a- (4) ground and polished specimen. Profiles (a) and (b) were taken
WE
from the same position on the specimen using precise relocation
where P is the load per unit length, E* is the reduced Young's techniques. This permits detailed analyses of roughness changes
modulus, and R the cylinder radius. to be conducted. The original topography is clearly evident,
PE* although plastic deformation of asperity peaks has occurred.
Maximum contact pressure p0 = { (5) Figures 8 and 9 show the distribution of contact pressure
wR
and subsurface orthogonal and principal shear stress for the
The subsurface stresses along the axis of symmetry (z-axis) simulated elastic contact of profiles (a) and (b) shown in Fig.
are given by: 7, with a smooth cylindrical counterface. The results are pres-
A> ent in the same format and to the same scale as those of Fig.
{(a1 + 2z2)(a2 + z2)'y2-2z} (6a) 6, for the contact of smooth surfaces, thus permitting direct
comparison. The analyses were conducted using conditions as
oz= -Poatf + z2)- (6b) near as possible to those of the disk machine test programme.
These are principal stresses so that the principal shear stress The indentation hardness of the specimen material is ap-
is given by: proximately 7 x 109 N/m2 (700 Hv), which corresponds to 1.7/?0.
2 2 l/2 The elastic contact analysis for the unrun surface (Fig. 8(a)),
Tl=p0a{z-z (a -z?r } (6c) shows that many of the asperity pressures exceed this value,
For all examples of smooth body contact used to evaluate thus suggesting that localized plastic flow may occur. The
the subsurface stress distributions, either with or without fric- pressure at which asperities deform plastically is influenced by
tion, the results agreed well with those given by Johnson (1985). local strain hardening and interactions between neighbouring
The variations of ax, az, and TXZ with depth below the surface asperities. Studies by Childs (1977), of rough surfaces in static
(along the axis of symmetry) as calculated by the computer contact, show that the supportable pressure increases with the
model are plotted in Fig. 5(a). A closer examination of the ratio of real to apparent contact area for ratios above 0.5. The
near surface stresses is shown in Fig. 5(b). Accuracy is main- present analyses show that the contact ratio of the ground
tained even when the depth below the surface is less than the surface increased from 0.69 to 0.87 during testing, and that
sampling interval of the digitized profile. the asperities deformed plastically to give a peak contact pres-
sure slightly less than the material indentation hardness of 1.7p0 The surface pressures associated with asperity contacts pro-
(see Fig. 9(a)). This suggests that in a rolling and sliding lu- duce highly stressed zones which reach a maximum at, or very
bricated contact the local pressure is limited by the independent near, the surface. In the case of the unrun surface many of
plastic deformation of the surface asperities. the subsurface "asperity" shear stresses are twice the mag-
The computer model predictions of subsurface principal and nitude of the maximum "smooth body" values. However,
orthogonal shear stress for each contact analysis are shown as these stresses are unrealistically high in that the contact analysis
isometric and contour plots. At depths below the surface in was purely elastic and many of the asperities would be expected
excess of approximately 0.5a (where a is the Hertzian half- to, and in practice do, deform plastically. However, this anal-
width) the stress distributions are as predicted by smooth body ysis represents the worst case of stress, as experienced during
theory. However, consideration of surface roughness and fric- the first contact cycle. Indeed cyclic plastic damage done to
tion influences significantly the distribution of stresses in the asperities during the initial "running-in" period is a principal
near surface region. source of surface distress micro-pitting.
Notice how the asperity stresses interact with the highly A similar set of results for the ground and polished surface
stressed subsurface region dominated by Hertzian effects, even shown in Fig. 7(c) are given in Fig. 10. In this case the mag-
under conditions of zero friction, to distort the stress distri- nitude of the asperity stresses is less than the maximum
bution toward the surface. This is particularly apparent with "smooth" body values, and are much shallower than for the
the unrun surface. The combined effect of roughness and fric- ground surface previously analysed.
tion is shown in Figs. 8(rf and g) and 9(d and g) for the contact In Fig. 11 the effect of roughness and friction on the sub-
conditions previously described. A frictional coefficient of surface principal shear stress is illustrated in the form of se-
/x = 0.1, considered realistic for lubricated contacts, has been lected cross sections from the stress distributions previously
assumed for these analyses. The well-known effect of the described. Figure 11(a) compares cross sections of principal
"smooth-body" stresses rising toward the surface is clearly shear stress through the peak asperity values from Figs. 8(e)
evident. and 10(e) with a similarly located section through the smooth
(a)
body stress distribution (Fig. 6(e)). These distributions rep- subsurface stress model represents a useful technique for ana-
resent normal loading conditions (/* = 0) and clearly demon- lyzing the effect of roughness and friction on the distribution
strate the interaction of the asperity stresses with the Hertzian of stresses likely to cause fatigue failures.
stresses. Figure 11(b) shows similar cross sections of principal Stress predictions for polished surfaces show that smoothing
shear stress taken through the right-hand edge of the Hertzian of the asperity peak significantly reduces the magnitude of the
contact of Figs. 6(f), 6(g), and 8(g). The combined influence near-surface stresses. This is in agreement with the results of
of roughness and friction produces a large highly stressed re- disk machine tests conducted in conjunction with this theo-
gion exposed to the surface. retical study which show that surface polishing eliminates as-
perity deformation and micro-cracking.
The prediction of high shear stresses immediately beneath
Discussion the asperities of ground surfaces is consistent with experimental
Notwithstanding the limitations of the contact model, the results which show the propensity of ground surfaces to suffer
Conclusions
A theoretical technique has been developed to determine the
subsurface stress field within real rough surfaces subjected to
both normal and tangential loads. The accuracy and resolution
of this technique is sufficient to permit an analysis of individual
asperity contacts within a normal Hertzian contact of realistic
dimensions.
It is shown that for a range of real rough engineering surfaces
the maximum orthogonal and principal shear stresses occur at
Depth below surface (z/a) or very close to the surface, even if frictional effects are ig-
Fig. 11(6) Cross sections through edge of Hertzian contact for smooth nored. For a polished surface the magnitude and position of
and unrun ground surface the maximum shear stress values are as predicted by smooth
Fig. 11 Cross-sections of principal shear stress taken from subsurface
body theory.
distributions previously described to illustrate effect of roughness and Plastic deformation of asperities during running-in reduces
friction the peak contact pressures and the resulting subsurface stresses.
However, for a well run ground surface the maximum or-
thogonal and principal shear stresses still occur at or very near
from micro-pitting and transverse asperity cracking (Olver, the surface.
1984 and 1986). The formation of micro-cracks in ground Using realistic coefficients of friction (//. ~ 0.1) the smooth
surfaces, usually during the early stages of running, is tempered body stresses interact with the asperity stresses to produce a
by their reluctance to propagate readily into the highly stressed large, highly stressed region exposed to the surface.
subsurface region dominated by Hertzian effects. Generally, It is postulated that the stress distributions associated with
most of the cracks grow very slowly and remain small, with certain asperity sizes could bridge the quiescent stress zone and
the exception of one or two which in the latter stages grow provide a mechanism for pit formation.
rapidly to form a pit (Clarke et al., 1984).
The foregoing stress predictions clearly identify a region of
relatively low stress separating the highly stressed near-surface Acknowledgments
region due to asperity interactions and the subsurface region The authors gratefully acknowledge the assistance of Dr.
dominated by hertzian effects. This low stress region or ' 'quies- Andy Olver, Westland Helicopters Ltd for his encouragement
cent" zone, discussed by Tallian et al. (1987) and identified and suggestions throughout this study. The authors are also
experimentally by Clarke et al. (1984), is the most likely ex- indebted to Dr. Martin Salzer, Westland Helicopters Ltd for
planation for this type of crack growth behavior. Clearly, the his continued support and for permission to publish.
formation of a pit requires some means whereby the surface This paper is based on research work funded by Westland
micro-cracks can propagate through this low stress region. Helicopters Ltd and the Department of Trade and Industry.
Motivated by the examination of failed test specimens, Cheng
et al. (1984), analyzed the possibility of subsurface defects
providing the necessary bridge. Their results indicated that even References
small inclusions, if strategically located, could provide a mech- A.F.B.M.A., 1960, "Method of Evaluating Load Ratings for Ball Bearings,"
anism for pit formation. I.S.O. Standard, Section 9, Revision 4.
Archard, J. F., 1957, "Elastic Deformation and the Laws of Friction," Pro-
However, this study, utilizing a contact model based on ceedings, Royal Society, Vol. A, No. 243, pp. 190-205.
measured roughness profiles, indicates that the stress distri- Berthe, D., Michau, D., Flamand, L., and Godet, M., 1977, "SurfaceRough-
bution associated with certain asperity sizes bridges the quies- ness Effects in Lubrication-Effect of Roughness Ratio and Hertzian Pressure
cent zone and provides an additional mechanism for pit on Micropits and Spalls in Concentrated Contacts," Proceedings of the Leeds-
Lyon Symposium.
formation. Bryant, M. D., and Keer, L. M., 1982, "Rough Contact Between Elastically
The main hypotheses concerning the critical stresses in roll- and Geometrically Similar Curved Bodies," ASME Journal of Applied Me-
ing contact fatigue are reviewed by Popinceanu et al. (1981). chanics, Vol. 49, pp. 345-352.
Although the hypotheses of maximum orthogonal shear stress, Bush, A. W., Gibson, R. D., and Thomas, T. R., 1975, "The Elastic Contact
of a Rough Surface," Wear, Vol. 35, pp. 87-111.
proposed by Lundburg and Palmgren (1947), was accepted by Bush, A. W., Gibson, R. D., and Keogh, G. P., 1979, "Strongly Anisotropic
the International Organsiation for Standardisation Rough Surfaces," ASME JOURNAL OF LUBRICATION TECHNOLOGY, Vol. 101, pp.
(A.F.B.M.A., 1960) for calculating the dynamic capacities of 15-20.
rolling bearings, there is no general consensus as to the truly Cheng, H. S., Keer, L. M., and Mura, T., 1984, "Analytical Modelling of
Surface Pitting in Simulated Gear-Teeth Contacts," SAE Technical Paper Series,
critical stresses in rolling contact fatigue. No. 841086,
In addition, contacting surfaces subjected to cyclic loading Childs, T. H. C , 1977, "The Persistence of Roughness between Surfaces in
may strain harden and develop residual stresses due to local Static Contact," Proceedings, Royal Society London, Vol. A353, pp. 35-53.
It is significant that the major tribology research developments of the past 75 years were driven
by the needs of the automotive, aerospace and nuclear industries. This book serves as a compre-
hensive review of each of the major disciplines in tribology over the past 75 years. An overview is
given of the technology then and now, as well as future needs and expectations, including the
economic and technical drivers of tribology research.
To order, write ASME Order Department, 22 Law Drive, Box 2300, Fairfield, NJ 07007-2300
or call 1-800-THE-ASME (843-2763) or FAX 1-201-882-1717.