Você está na página 1de 20

95

LUBRICATION AND WEAR GROUP

CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT


SURFACES SUBJECTED TO RADIAL AND TANGENTIAL
FORCES
By D. J. Haines, B.Sc., Ph.D. (Associate Member)* and
E. Ollerton, B.Sc. (Eng.), Ph.D. (Associate Member)JC

The problem of Hertzian bodies in rolling contact and supporting radial and sheering forces in the rolling
direction is considered.
A modified form of the conventional photoelastic frozen stress technique has been used to study the
particular case of flat elliptical contact surfaces. Existing theories are reviewed and new theories are presented
which permit the analysis of the frozen stress results. The dependence of the measured stresses on the
hysteresis of rolling is studied.

INTRODUCTION tive motion occurred between the surfaces of the two bodies
THEPROBLEM OF CONTACT STRESS has of late received in contact.
considerable attention. One of the questions at present Mindlin (2) published a paper dealing with spheres in
remaining unanswered is that of the distribution of shearing static contact under the action of radial and tangential
stress on a flat contact area which has an elliptical boundary, forces. His distribution of surface shear stress agreed with
as defined by the Hertz theory, and is subjected to a non- that of Cattaneo and this paper also contained surface strain
limiting shearing force while the two mating bodies are in results.
rolling contact. In this paper an approximate theoretical Vermeulen and Johnson (3) recently extended Mindlins
solution is presented and experiments to determine the work and have calculated surface strains for a large range
surface stresses using photoelastic frozen stress and relative of ellipses of contact.
creep techniques are described. MEwen (4) published a paper giving the complete
The problem of static contact between elastic bodies solution for the stresses in cylindrical bodies subjected to
having two principal curvatures at the poifit of contact was radial pressure and a limiting shearing traction. Subse-
studied by Cattaneo (I)$. He considered the stresses in the quently, Poritsky (5) studied this problem as well as the
bodies under the action of radial and tangential forces and non-limiting shearing traction case. In the discussion to
obtained the distribution of surface shear stress. The area Poritskys paper, Cain (6) showed from the laws of simple
of contact was divided into two parts: an area of adhesion friction that the adhesion area adjoins the leading edge of
which had the shape of an ellipse, with the same eccentricity contact.
as the ellipse of contact and placed centrally on it, over Johnson (7) studied the creep of a sphere over a plane on
which there was no relative motion between the bodies; and, which it rolled while sustaining radial and shearing forces.
outside the adhesion area, a region of slip over which rela- He assumed that the adhesion area would be circular and,
after Cain, adjoining the leading edge of contact. His
The M S . of this paper was first received at the Institution on 21st experimental results show reasonable agreement with
September 1961, and in its revisedform, as accepted by the Council expressions developed in his paper.
f o r publication, on 6th December 1961. The content forms part of a
Ph.D. thesis submitted to the University of Nottingham by the first
author, who worked under the supervision of the second. NOTATION
* Lecturer in the University of Bristol.
*
t Lecturer in the University of Nottingham.
References are given in the Appendix.
ai, bi Semi-widths of a contact surface measured in
the x and y directions from a point i.
Proc Instn Mech Engrs Vol I77 No 4 196.7
Downloaded from pme.sagepub.com at Gazi University on January 9, 2015
D. 5. HAINES AND E. OLLERTON

Semi-widths of a contact surface measured of the shearing force does not alter the distribution of the
along the x and y axes respectively. radial pressure over any part of the contact area, i.e. the
Semi-widths which define the adhesion area. radial pressure on the surface resulting from the shearing
A coefficient. traction alone is everywhere zero.
Shear modulus of elasticity. In the region of slip (where relative motion is taking place
= a</ao. between the two bodies) it is assumed that the shearing
Total normal traction. stress at any point must be equal to the coefficient of friction
Total shearing traction. p x the radial pressure at that point. Inside the area of
Hertzian normal pressure on the centre of a adhesion the shearing stress must everywhere be less than
contact area. ,u x the radial pressure.
The 2, shearing stress at the centre of a Over the area of adhesion there is no relative motion
contact area due to the presence of a between the surfaces as rolling proceeds and to satisfl this
limiting shearing traction in the x direction. condition it can be shown (7) that the surface strains
Net shearing traction value of &/ax within av/ax and au/ax associated with the shearing traction must
the adhesion area at y = b;. be constant for each body over this region. (See Fig. 1 for
Net shearing traction value of &/ax within notation,) If longitudinal creep is to occur it is necessary
the adhesion area at y = 0. that the values of au/ax must be different for the two bodies.
Displacements in the x, y, and z directions. The surface shearing stresses applied to one body must be
Direct stresses. equal in magnitude and opposite in sign to the shearing
Orthogonal shearing stresses (see Fig. 1). stresses at corresponding points in the other body.
Cartesian co-ordinates where x is the direction Warping of the contact surface due to the loading
of shearing traction and rolling. z is the together of dissimilar bodies will be neglected as in the
normal to the surface and x = y = z = 0 Hertz theory.
at the centre of the contact surface. (See
Fig. 1.)
Cartesian co-ordinates where x = y = z = 0 Areas of adhesion for rolling contact
at the centre of the adhesion area. Cains proof (6), that the area of adhesion extends to the
Coefficient of friction under rolling contact leading edge of contact in the case of parallel cylinders, also
conditions. applies for elliptical contact areas. I n this section this result
Poissons ratio. is applied to elliptical contact areas which support non-
Creep ratio. limiting shearing forces applied in the rolling direction.
Suffix N refers to a net value. If it is assumed that the bodies can be considered to be
Suffix L refers to a limiting shearing traction made up of a series of strips parallel to the x axis and
value. perpendicular to the surface, and if interaction between these
strips is ignored, each strip can be studied with the aid of
A THEORY OF SHEARING TRACTION Carters (8) and later Poritskys (5) two-dimensional theory.
Assumptions For the purpose of this, it is appropriate to let the shearing
The conditions which must be satisfied when two elastic force be given some value so that the extent of the adhesion
bodies are in rolling contact under a shearing load have been area on a strip at y = 0 is 2ao (see Fig. 2). The total
clearly described by Johnson (7). Some of these conditions distribution of shearing stress on this strip may then be
will be repeated here so that the assumptions necessary in considered to be made up of a positive limiting traction
the development of the theory will be understood. over the whole length of contact ( 2 4 and a negative traction
The first assumption which is made is that the presence distributed over the length of the adhesion area.
When superimposedthese two tractions give the resultant
CONTACT ELLIPSE.
traction on the strip. The limiting traction is given by px
the Hertzian radial pressure, and its distribution can be
represented to a suitable scale by the semicircleshown. This
limiting shearing traction gives rise to surface strains which
are proportional to x inside the length of contact.
The requirement that the tractive aulax shall be constant
over the adhesion area gives the result that the negative
traction must also produce strains which are proportional
to x, the constant of proportionality being the same as for
the limiting traction. The distribution of negative shearing
stress which will give rise to this strain distribution is
represented by a semicircle in Fig. 2, and the ordinate
representing this shearing stress will have the same scale as
F&. I . Notation that for the limiting traction.
Proc Instn A4ech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES 97

force distributed over the entire contact area, plus a negative


shearing force carried by the area of adhesion. T o determine
the extent of the adhesion area for a given shearing force
TRACTION SYSTEM P, in this approximate case, it is necessary to equate the
I
f volume under the negative traction surface to the difference
between the limiting shearing force, pP, and P,.
The negative shearing traction being carried on a strip at
y = bi is

and the limit of the negative traction [from equation (1),


and the equation of an ellipse], is

b,,' = bo [2 $- k)']
Hence, substituting a;' in terms of ao, ao', and a; from
equation (l), the total negative traction (pP-Pt) is given
PRESUMED EXTENT by the equation
OF ADHESION AREA

+(ao'--ao)2-2ao (1- ($)*(aO-a..)]dY


The resulting equation is
x
D
_pt - 3
ELF - l-- 2 x
Fig. 2. A distribution of shearing traction on the x axis for
the rolling condition and graphs of &/ax at y = 0
and bi . . . (2)
where K = ao'/aoand P is the total radial pressure between
Now the positive traction is known all over the contact the bodies. This result which follows from the assumptions
area, hence au/ax for this traction is known and it is only made above is plotted in Fig. 3.
necessary to determine the extent of the negative traction
on each strip to give the constant tractive &/ax required on
the adhesion area.
If a section at y = b , width 2ai, is considered, it will be
seen from Fig. 2 that to give the required net value of
&/ax (i.e. Ri= Ro) the centre of the negative traction
distribution must always lie at the same &stance from the
y axis: that is,
ai-ai' = ao-ad . . . . (1)
defines the adhesion area.
Thus the required boundary of the adhesion area is a
reflection of the loading edge of the contact area about a
line perpendicular to the rolling direction.
This solution is of course approximate because of the
assumptions made, but it will be shown later that it is a
satisfactory method for defining the shape of the adhesion
area.

The relation between area of adhesion and shearing


traction
In the above section the roral shearing force carried on Fig. 3. Graph showing relation between applied traction
the contact area is made up of a limiting positive shearing and adhesion area
Proc Imrn Mech Engrs Vo1177 hTo4 I963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


98 D. J. HAINES AND E. OLLERTON

LONGITUDINAL RELATIVE CREEP 5v 1


and - = - [Yy-v(Z,+XJl
Over the adhesion area there is no relative movement 8y E
between the contacting bodies; thus if one of the bodies is which reduce to
unstrained in the direction of rolling (i.e. the value of
aujax associated with the shearing traction is zero over this
region) and the other has a surface direct strain 5u/&
which is a constant (tensile), then the strained body will
behave as if it were larger and in fact will rotate at a lower and
velocity than would be expected when &/ax = 0, i.e. the Since &/6x and are independent of y these stress
tensile strain effectively increases the circumference of the differences are constant along the y axis.
strained body. If at the same time the other body had a From the symmetry of the adhesion area
compressive strain, the total effective strain would be the
sum of the two strains. In the case considered the surface
strains over the adhesion area due to shearing traction are
equal in magnitude but of opposite sign and thus the creep
ratio f, = 2 1 au/axNl where au/axN(= R ) is the net value of therefore the values of X,--Z, and X x - Y y which
this strain in the adhesion area. I-v
Now Mindlin (2) has calculated the limiting traction may be calculated along they axis of the adhesion area are
value of au/ax for the case of a circular contact area and the mean values of these stress differences within the
Vermeulen and Johnson (3) have performed the same cal- adhesion area.
culation for an elliptical area. If the shearing traction values of au,/ax and av/ay are
Reference to Fig. 2 will show that the creep ratio 6, is known within the adhesion area these stress differences can
related to the limiting traction values of &/ax, i.e. auj8xI-, be determined along the y axis of the adhesion area, i.e.
by the equations from equation (3),
(ao-ao) 8u _
au ----
ao-a; au
R=--
x ax, &N x axL
and and it can also be shown that
av
-=---
ao-a0 %a
%N x SYL
Hence, if the creep ratio is known, the extent of the ad- where Lu/axL and a@yL are the limiting shearing values
hesion area associated with any shearing traction may be of these strains.
estimated from equation (3), provided that the resulting
stress system is a simple summation of a Hertzian system PHOTOELASTIC TECHNIQUES
and a shearing system which does not alter the radial
Hertzian stress on any surface element. Fig. 4 shows the basic apparatus which has been used for
experiments on stress freezing and relative creep. The
SURFACE X , AND Yy STRESSES apparatus is shown positioned in a hot air oven for an
The theoretical values of these stresses may be determined experiment on frozen stress.
from the work of Mindlin (2) and Vermeulen and Johnson The models, which are nominally of 5 in. diameter, were
machined and surface ground from Araldite Casting
(3)-
It has already been stated that a condition for no longi- Resin B and were mounted in self-aligning bearings. The
tudinal slip is that &/ax associatedwith the shearingtraction top model assembly is free to pivot about a fulcrum
must be constant over the adhesion area. A similar require- positioned near the front of the oven.
ment exists for iiv/ax if no transverse slip is to occur. There are two holes in the oven floor: normal load is
But Vermeuleiis and Johnsons results (3) show that for applied through one to the top model, and shearing traction
all ellipses of contact the z, displacementis proportionalto the is applied through the other. This application of load
product of x andy. Therefore in the case of limiting traction produces the required shearing force because the tensions
aujax and av/ay are proportional to x and independent ofy. in the cables attempt to rotate the top and bottom models
Since the adhesion area has symmetry about a y axis in opposite directions.
(y),on this axis the negative traction values of &/ax and Rolling occurs when an out-of-balance load is supplied
&jay are zero. Therefore the net shearing traction values to the top model and overcomes losses due to hysteresis.
Before the strain system can be frozen into the models
of the stress differences x,--
1-v
2, and X,- Yy can the rolling motion must be stopped. If the losses in the
top assembly due to hysteresis equal those in the bottom
be calculated on this axis from the equations
assembly, the rolling motion is stopped and the shearing
traction remains a constant if the out-of-balance load is at
this point shared between the models. This may be achieved
?roc Instn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES YY

Fig. 6. A contact surface obtained under a state of


constant strain during freezing

regions of adhesion and slip within the area. In this experi-


ment the rate of rolling was very low and, because of this,
cold welding between the models was appreciable at the
testing temperature (125C). Undue reliance should not,
therefore, be placed on the shape of this adhesion area.
When the rate of rolling was increased to eliminate cold
welding, the tracks on the surfaces of the models were found
to taper, becoming wider as rolling proceeded. Moreover,
after rolling was stopped the contact area increased con-
siderably beyond its size at the end of rolling. These effects
are due to the visco-elastic behaviour of the Araldite at
125"C, and thc fact that the strained state of the resin does
Fig. 4. Apparatus for experiments on stress freezing and not remain constant during the freezing cycle.
Now at the end of rolling the models are almost in a state
relative creep
of equilibrium, and if the radial and shearing loads are
adjusted after this point is reached, the models may be
held in approximately this state throughout the remainder
of the loading cycle. Subsequent analysis will then r e v 4
the approximate end of rolling conditions.
Fig. 6 shows a contact surface (ao/bo= 1.00) produced
in this way. (The terminal velocity was approximately
0.010 in./min.) The distinction between a region of slip and
one of adhesion can barely be seen when cold welding is
eliminated. It is always lost after exposure to air for one or
two hours.
The adhesion area in this figure is 'lemon'-shaped as
predicted in equation (1).
Results from this and two other models (in which
= 2.0) will be studied in the next section.

CALIBRATIONS FROM RELATIVE C R E E P


A N D FROZEN S T R E S S R E S U L T S
Fig. 5. A contact surface soon after freezing The loads carried by the photoelastic models were not
constant during the freezing cycles and because of this it
by the use of a simple stop, and the strain-freezing cycle was not possible to conduct independent calibration tests
can then proceed. to relate the frozen stresses to the applied loads. Since no
Fig. 5 shows contact surface soon after freezing. When complete theoretical treatment of the problem exists it was
sliced and analysed it was found that the two regions (light also impossible to perform simple direct calibrations of the
and dark) within the contact surface corresponded to the models.
Proc Instn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


100 D. J. HAINES AND E. OLLERTON

-.
I I 1 I II

.
.-
a
m
C
c'20c

.-L
c
IlOC
G

200 I I I 1 I

ru*
0
1 J/l-y\
oo = 0.487 in.
i-
-

oo = 0.474 in.

Fig. 7. Measured and theoretical surface traction stresses. Fig.8. Measured and theoretical surface traction stresses.
Model 3 3 Model4B

These difficulties were overcome, however, with the aid negative shearing traction is distributedover the strip-theory
of equations (1) and (3) and longitudinal relative creep adhesion area, but that its value is everywhere less than its
experiments conducted between rubber models which had predicted value by a factor C. C is independent ofy for any
the same Poisson's ratio (0.50) as the frozen models. given ellipse of contact and shearing traction, but is a
To calibrate the models in this way it was necessary to function of the traction being carried and the shape of the
establish three results: the extent and shape of the adhesion ellipse of contact.
area; the distribution of the assumed negative shearing Equation (3) indicates that if the coefficient of friction is
traction on the adhesion area; the magnitude of the surface known and the creep ratio is carefully measured, the extent
shearing stress, and in particular the maximum value of this of the adhesion area can be determined for any known radial
stress if the models had supported a limiting shearing force. and tangential forces. An experimental value of C can then
Study of the measured surface shearing stress in Figs 7,8, be obtained by equating the volume of the negative shearing
and 9 shows that the distribution of stress was not in exact stress distribution (acting over a known area) to the dif-
agreement with that predicted by the simple strip theory. ference between the limiting tangential force and the
In each case it was necessary to reduce the magnitude of tangential force actually applied. Experimental values of C
the negative shearing stress ordinate below the value pre- were determined for Poisson's ratios of 0.50 (rubber models)
dicted in order to obtain reasonable agreement between the and 0-39 (Araldite models loaded at room temperature).
experimental results and the 'theoretical' curve in the These results are presented in Fig. 10 and indicate that the
planey = 0. distributions of surface traction (defined by the values of C)
The ratio between the measured maximum negative were sensibly independent of Poisson's ratio.
shearing stress and the value according to the strip theory The values of C used in Figs 7-9 are the same as those
was called C. The curves shown in Figs 7, 8, and 9 for shown in Fig. 10 and illustrate the validity of this approxi-
strips other than the ones at y = 0 were constructed with mate approach to the work. To determine the extent of an
the same values of C as for the central strips, and it is seen area of adhesion for a given ellipse of contact and shearing
that there is reasonable agreement between these curves force, it is convenient to read the required coefficient from
and the measured stress values. Fig. 10 and substitute this in the construction shown in
The conclusion drawn from these results is that the Fig. 3.
Proc Instn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES 101

The final requirement in the calibration of the frozen permitted the relevant Hertzian stresses to be calculated.
stress models concerned the relative magnitudes of the This was possible because the Hertzian surface 2, shear
shearing stresses and the Hertzian stresses. Since the ratio stress was zero.
of the shearing load to the radial load was known in each The form of the 2, shearing stress in Figs 7 and 9
case, a summation of the 2, stress across the contact surface suggests that the Hertzian radial pressure was not greatly
altered by the application of the shearing force, although
the nature of the stress-freezing process could well allow
some slight redistribution of stress at the end of rolling and
before freezing.
The deviation of the coefficient C in Fig. 10 from 1-00
indicates a discrepancy between the simple strip theory and
the experimental observations. Vermeulen and Johnson (3)
have evaluated &/ax and &/ax for surface shearing
tractions having ellipsoidal distributions on elliptical contact
areas. A value for C can be obtained by assuming that the
adhesion area is elliptical and supports an ellipsoidal dis-
tribution of shearing stress with a height that ensures a
constant net value of &/ax all over the adhesion area. The
ratio of the height of this negative traction distribution to
that predicted by the strip theory will then give a value of C.
It is found that the two conditions required for no slip in
the adhesion area (attiax and avjax to be both constant) can
only be satisfied simultaneously if the ellipse defining the
adhesion area has the same eccentricity as the ellipse de-
fining the contact boundary, e.g. for a circular contact area
the adhesion area must also be circular. This method of
approach can therefore only lead to approximate answers
because under these conditions part of the region of slip
carries shearingtractions which contradict the laws of simple
friction, as explained by Johnson (7).Values of C have been
calculated for the case of a circular contact area by fitting
ellipses approximating to the adhesion area, as shown in
a, - 0.500 in.
Fig. 11, and choosing the height of the ellipsoid, qo', to
give constant values of first h / a x and then &/ax over this
Fig. 9. Measured and theoretical surface traction stresses. region. These values are shown in Fig. 12 along with the
Model 5B experimental C values. It is evident from Fig. 12 that the

9
Fig. 10. Values of C
Proc Znstn Mech Erigrs Val 177 No 4 I963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


102 D. J. HAINES AND E. OLLERTON

actual shearing stress distribution was considerably dif- The effect of rolling hysteresis, which had not been
ferent from that predicted when an elliptical adhesion area allowed for in the theory. The presence of hysteresis in the
was assumed. material of rolling bodies results in the necessity for the
The disagreement between the experimental and application of a turning moment to maintain a constant
theoretical results is felt to have been due mainly to the velocity of rolling. The turning moment must be opposed
following two factors. in the contact area by the resistance of the material to
The simplifying assumptions necessary in the develop- deformation, and because of this it is expected that the
ment of the theory were not entirely justified, particularly contact area will move forward relative to the line joining
the assumption that the contact region could be divided the centres of the rolling bodies and that the pressure
into strips having no interaction between themselves. distribution will change, becoming greater towards the
I leading edge of the contact area and smaller towards the
trailing edge. This effect has been confirmed very recently
by Hunter (9) in a paper dealing with the rolling contact of
a rigid cylinder with a visco-elastic half-space, It has not
yet been possible to extend the theory to elliptical contact
areas.
The precise effect of the redistribution of radial pressure
is difficultto estimate, because the lack of symmetry of the
new pressure distribution will result in surface strains
which are non-linear with respect to x. In view of these
considerations it is not surprising that the simplified theory
does not exactly describe conditions in the contact area.
Measurements an the frozen stress models showed that the
contact surface was displaced to a position in advance of the
plane joining the axes of rotation of the two similar models.

'-.
/

'ELLIPSE
I
HAVING THE SAME AREA AND THE
SAME VALUE OF do' AS THE STRIP THEORY
ADHESION AREA
F&. 11. The fitting of elliptical adhesion areas to obtain
czpproximate values of C

0.407in.
I I I I I do=

0 0.2 0.4 p.6 0.0 1.0 .f Driving model x Driven model


--
a0
a0
do
-- --
-_.--- Strip theory
Theoretical result computed from (3)
F g . 12. Comparison of experimental and approximate Fig. 13. Measured and calculated surface (X,-Z,, stress.
theoretical values of C Models 3
Proc Imtn Mech Engrs
V d I77 N o 4 1963
Downloaded from pme.sagepub.com at Gazi University on January 9, 2015
CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES 103

-4OOL
a. = 0.474 in.

+ Driving
----- model x Driven model
Strip theory I I
--- Theoretical result computed from (3) do = 0,500in.
Fig. 14. Measured and calculated surface (X,-Z,) stress. + Driving model x Driven model
Models 4 ----- Strip theory
--- - -Theoretical result computed from (3)
DIRECT S T R E S S E S Fig. 15. Measured and calculated surface (X,-Z,) stress.
Measured values of the surface X,-Z, stresses in the ex- Models 5
periments on stress freezing are presented in Figs 13-15
together with theoretical lines which have been derived
with a simple theoretical prediction. For a flat contact
from: equation (4), and references (2) and (3); the strip
surface the adhesion area is lemon-shaped and its rear
theory, the actual (ao-ao) values, and the assumption
boundary is a reflection of its forward boundary, which
that C = 1.00.
It can be shown from the Hertz theory that when forms part of the ellipse of contact.
The distribution of surface shearing stress within the
Poissons ratio is 0.50 the value of (X,-Z,) is, for the case
of radial load, everywhere zero in the contact surface. adhesion area is different from that predicted by a simple
Hence the measured (X,-Z,) values are all associated with strip theory. This difference is sensibly independent of
the applied shearing tractions and hysteresis effects. The Poissons ratio and the experimental values show consistent
difference between the measured and mean predicted trends, thus enabling the approximate distribution of
stresses is thought to be due to the hysteresis of the model shearing stress to be predicted for particular cases.
material. The mean experimental values of the stresses on The magnitudes of the surface direct stresses in the plane
the adhesion area for each pair of models approximate to of rolling are affected by the hysteresis of the materials.
the mean predicted stresses.
In each experiment the maximum tensile stress in the ACKNOWLEDGEMENTS
adhesion area and in the slip region approximates to a value The authors wish to thank the sponsors of the work, the
according to the strip theory. British Transport Commission, for permission to publish
this paper, and the Department of Mechanical Engineering
CONCLUSIONS in the University of Nottingham for the facilities provided.
The shape of the adhesion area obtained under rolling Thanks are also due to Metalastik Ltd for designing the
contact conditions in the presence of shearing forces agrees rubber moulding equipment and casting the rubber models.
Proc Instn Mech Engrs Vo1177 No 4 1963
Downloaded from pme.sagepub.com at Gazi University on January 9, 2015
104 D. J. HAINES AND E. OLLERTON

APPENDIX contact with application to contact of gears and of


locomotive wheels, 3. appl. Mech., Trans. Amer. SOC.
REFERENCES mech. Engrs 1950 72, 191.
(I) CATTANEO, C . Sul contatto di due corpi elastici, R.C. Accad. (6) CNN, B. S. Contribution to discussion on (5), 3. appz. Mech.,
Lincei 1938 Series 6, 27, 342, 434, 474. Trans. Amer. SOC.mech. Engrs 1950 72,465.
(2) MIHDLIN, R. D. Compliance of elastic bodies in contact, (7) JOHNSON,K. L. The effect of a tangential contact force upon
J. appl. Mech., Trans. Amer. SOC.mech. Engrs 1949 71,259. the rolling motion of an elastic sphere on a plane, 3.
( 3 ) VERMEULEN, P. J. and JOHNSON, K. L. Private communication appl. Mech., Trans. Amer. SOC.mech. Engrs 1958 80, 339.
dated 16th March 1960. (8) CARTER, F. W. On the action of a locomotive driving wheel,
(4) MEwEN, E. Stresses in elastic cylinders in contact along a Proc. roy. SOC.A 1926 112, 151.
generatrix (including the effect of tangential friction), (9) HUNTER, S. C. The rolling contact of a rigid cylinder with a
Phil. Mag. 1949 7th Series 7,454. viscoelastic half space, J. appl. Mech., Trans. Amer. SOC.
(5) PORITSKY, H. Stresses and deflections of cylindrical bodies in mech. Engrs 1961 83, 611.

Communications
Professor Dr Ir A. D. de Pater (Delft, Holland)-Various half-space problems, chiefly because I myself am engaged
experimental investigations on the tangential contact prob- in a theoretical investigation of contact problems of that
lem in which two elastic bodies are pressed upon each other sort.
and then shifted statically over a certain distance have been (1) The problem of high-velocity rolling of symmetrical
executed in the past; until now, however, such problems bodies, upon which a force is acting in the direction of
in which the two contacting bodies roll stationarily along rolling, has been studied in the present paper and, in my
each other have only been investigated theoretically. The opinion, nothing needs to be added.
paper by Dr Haines and Dr Ollerton describes the first (2) As far as I know, there are no reports published of
experiments on these rolling contact problems and their photoelastic experiments on the problem of spin and
results are most interesting. They are to be congratulated transverse creep in high-velocity rolling of symmetrical
for the handsome and ingenious way in which they have elastic bodies. Such experiments would interest me very
executed these experiments. much indeed, especially those on the problem of spin. On
In the paper I missed a representation of the total shearing my side, I can offer some numerical results which pertain
traction P, as a function of the creep ratio tX.The relation to infinitesimal creep and spin, and which are valid for an
between these two quantities is of special interest to an elliptical area of contact. The results for a circle agree
engineer; but I assume that this relation can easily be very closely with Johnsons measurements of creep and
deduced from Fig. 3. spin. Later in this communication I will present these
In my opinion the strip theory, which the authors have results, together with a brief summary of the method
mentioned, fails in various respects, and a more exact employed.
theory is very desirable. My co-worker Mr J. J. Kalker has (3) Dr Ollerton has performed photoelastic experiments
done much valuable work in this direction and some of his (10) on the shift problem of Cattaneo and Mindlin (I) (2).
results are given in the following communication. Again, nothing needs to be added, even though the agree-
ment is fairly rough. I shall return to this in point (5).
Mr J. J. Kalker (Delft)-First of all, I should like to (4) In 1956 Hetenyi and McDonald published a paper
express to Dr Haines and Dr Ollerton my admiration for the (11) on the contact stresses that occur when two bodies are
clarity and accuracy of their experimental work, which pressed together and then twisted. A correct theory is given,
required, besides most painstaking effort, a thorough mastery and one photoelastic experimentis described. Unfortunately,
of the difficult frozen-stress technique. in their experiments, these authors treated only the case of
I believe that the picture of the stress distribution over large slip, to which their theory is confined. It is a pity
the contact area is an invaluable help to the theoretical that the authors did not make greater use of their apparatus.
student of elastic contact problems, because he now knows I shall also return to this presently.
exactly what his results should look like upon substitution (5) Dr Ollerton, in his experiments on Cattaneo-Mindlin
of the elastic moduli that obtain for the materials used by shift, found that the stress distribution is very strongly
Dr Haines and Dr Ollerton. He only asks for more experi- influenced by small differences in the history of the motion.
ments of the same type and quality on other contact prob- In the present paper, however, on experiments on rolling,
lems. I will now attempt to list the contact problems that nothing of the sort is reported, and I should be grateful if
interest me, and possibly, other theorists, and for which I the authors could give some more information on this point.
should like to have an experimental picture of the stress Personally, I am inclined to surmise, on theoretical grounds,
distribution. I limit myself to three-dimensional problems, that the elastic field in high-velocity rolling is independent
without elasto-dynamic effects, which can be reduced to of the history of the motion.
Proc Instn Mech Engrs Vol 177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES 105

(6) I think it would be very interesting to have photo- Now I would like to make a few remarks on the theory
elastic evidence concerning the manner in which a Cattaneo- presented to us by Dr Haines and Dr Ollerton.
Mindlin stress pattern becomes a high-velocity rolling On p. 102 of the paper it is stated that The precise effect
pattern. To catch the stress in this early stage, I suggest that of the radial pressure is difficultto estimate. . .. Perhaps I
it should be frozen when the distance rolled is between can help here, since it is easy to verify that the total stress
and 3 times the diameter of the contact area in the T~~ in symmetrical bodies in contact, in which the plane
direction of rolling. z = 0 is the plane of symmetry in which the contact area
(7) It would be interesting to verify photoelastically the lies, can be resolved into a stress 75, due to the Hertz
stress distributions described by Mindlin and Deresiewicz deformation, and a stress T& due to friction. I n Cartesian
(12). They consider the influence of the history of the
co-ordinates (x, y, z) = (xl, x2, xJ, these stresses are:

I)
motion when the bodies are so constrained that they can only
.:+:
shift with respem to each other in one particular direction.
They explicitly solve this problem in the form of a time in-
tegral, in which a loading history term plays a prominentpart.
Tt]

(7;)
=

=
($??),(
72:;

721 7 2 2 723
t

TCI) =
( ~ +
711 712 3:.
721 722 T 23 ;
(8) There is a fundamental difference between symmetri-
772 4 3 731 T32 733
cal contact problems, in which the bodies involved have 7
1
;-

the same elastic properties (e.g. Araldite on Araldite), and where T& y, z) = ~ { T , ( x , y, z ) + ~ , ( x , Y , -4>,
contact problems where this is not so (Araldite-rubber or and Ti(& YYz) = ~ { T z J (YYZ)-TZ~(XY YY
~Y
Araldite-steel). The reason is that, in symmetrical contact,
the tangential quantities (eg. sliding velocity and tangential The n and t stresses each form an equilibrium system; they
load distribution) do not influence the normal quantities, are produced by displacements u: and u: where the totd
such as the contact area and the normal load distribution. displacement u, is u:+u:, with u: = (u;, u ; , u ; ) and u: =
This is not so in asymmetrical contact. (ui, u;, uf).
For a start, one could confine oneself to the problem Speaking in geometrical terms, we can say that the
treated by Dr Haines and Dr Ollerton in the paper under n system is mirror-symmetrical about z = 0, and the t
discussion. The elasticfield here, as I surmise (see point (5)), system is mirror-anti-symmetricalabout the same plane. We
is rather insensible to the detailed history of the motion. can, of course, add a rigid displacement to each of the
Information on the areas of slip and adhesion would be of displacement systems, as long as we do it in such a way that
great interest to the theorist, and in general he would like to the stress is not disturbed.
gain some understanding of the interaction of tangential The above property has been used in symmetrical contact
and normal quantities. problems since the time of Carter who dealt with the two-
To finish the experimental section of this discussion, I dimensional case only (8).
should like to point out that experimenters should not feel I should also like to point out that a strip theory is
they must confine themselves to the verification of existing somewhat unsatisfactory from a purely theoretical point of
theories. Such unexplained experiments are of the greatest view and in the form presented here it has the very serious
value to the theorist, as I pointed out in the beginning of practical drawback that neither spin nor transverse slip
my remarks. One can also argue that a theory may be forth- can be interpreted by it. By the same token, I cannot accept
coming after the photoelastic work has been completed: it the statement on p. 95 of the paper that Cain has shown that
may even be that the theory will be inspired by the photo- the adhesion area adjoins the leading edge of the contact
elastic work. region. Is this true also when spin and transverse creep
Moreover, a theoretical problem of the half-space can be occur, and how does the adhesion area get there from its
considered as solved, when the load distribution on the Mindlin-Cattaneo position ? Many plausible arguments
surface of the half-space is known, for displacements, can be devised, but plausibihty is a tricky thing and should
strains, and stresses can be very easily computed by means not be confused with proof. The only real proof we possess
of the integral representation of Cerrutti (13). When the is the experimental proof of Dr Haines and Dr Ollerton.
surface integral is reduced to a weighted summation over In the above I hope I have made it clear why we simply
100 points I estimate that the time needed by an IBM 650, must have a truly three-dimensional theory, not only for
to find all the necessary quantities at a single point inside certain special cases, but for the complete symmetrical
the half-space, would be 1-3 min. On an IBM 704 the contact problem. Ideally, this theory should be of the type
time would be approximately 50 times shorter. presented by Mindlin and Deresiewicz in their remarkable
For instance, such a programme can be used to determine paper of 1953. If this is impossible at present an acceptable
the elasticfield inside the half-space directly from the surface alternative would be a really efficient computer programme
loads given by Lubkin (14)or Cattaneo (IS). This consider- for the simulation of contact problems.
ably simplifies the task of the experimenter engaged in work At present I cannot put forward such a three-dimensional
on point (4). In 1954 Deresiewicz (16)contended that there theory; the only thing I have is the relation between slip
was no agreement between Cattaneo on the one hand and and spin on the one hand, and the tangential force and the
Lubkin and himself on the other, but I have not looked twisting couple on the other hand, when the two bodies roll
into this. over each other with high velocity and the tangential system
Proc Instn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


106 COMMUNICATIONS

is infinitesimal with respect to the normal system. In fact, which are applied to the upper body. Dynamic effects are
slip is prevented inside the contact area. Moreover, the neglected.
results are restricted to elastic symmetrical half-spaces. It appears that the tangential force and the twisting
(I obtained them in October 1961.) couple can be expressed in the longitudinal slip a, the
I shall not describe the method I used to obtain these transverse slip p, and the spin y in the following manner:
results in detail. Let it suffice that it is a method of con- F, = c2aGA(e, a), F = c2G/3B(eYu)+c3GyC(e, u),
centrated loads on the half-space, and that Johnsons
method of treating spin (17)is a very simple example of it. M , = c3GpD(e, u)+c4GyE(e, u); c = max (a, b)
The method is an adaptation to the tangential case of I have tabulated the functions A, ByC, D, and E of e
Galins and Dovnorovichs method (IS) (19)of finding the and a. The numerical calculation tends to show that
load distribution over an elliptical contact area of a half- C = -D. I have no explanation for this. Where the rela-
space when a frictionless rigid punch is pressed upon it; tion is not satisfied, I suspect that I have not used enough
like Dovnorovich, I avoid Lamks functions. In the con- equations, the number having been kept down to the
tact area the Lam6 functions have the character of eigen absolute minimum of 10 (four Fourier coefficients). The
functions of a finite system of linear equations: they can next larger set, with six Fourier coefficients,has 21 equations.
therefore be avoided by working with the linear equations The results for e = 0 I obtained in 1957 by a method in
themselves, and this is what Dovnorovich and I do. I n the which the displacements and stresses were expressed in
case of rolling, the system of linear equations is actually spheroidal harmonics. In principle, this comes to the same
the truncation of an infinite system because a stress singu- thing as the new method, which can also be used in the
larity which indicates slip must be removed from the whole elliptical case. The difference is computational; for an
of the leading edge. ellipse there are many more equations than for a circle
The removal can be effected by developing the strength when the harmonics are used. As a consequence the
of the singularity into a series of complete independent numerical values for a circle are accurate to the last decimal
functions, and by setting the coefficients of all these func- place given in the Tables. For an ellipse the last decimal
tions equal to zero. This can indeed be done under the place is a little doubtful.
boundary conditions of high-velocity rolling. The choice The results for a circle are marked with an asterisk.
of the set of complete functions has numerical significance We also possess the asymptotic limits, when
only. For the results I present here I used Fourier series. g = (l-e2)1P+O
Starting with the lowest, I set equal to zero as many
Fourier coefficients as the truncation permitted. For the I n the form given here, they are of little numerical import-
total load and the twisting moment (see Definitions) ance, since they are valid when -log(g) % 1, which is a
sufficient accuracy is obtained with 2 x 4 or, better, with pretty tall order. This gives some colour to the experience
2 x 6 Fourier coefficients. of Dr Haines and Dr Ollerton, that the results of the strip
theory cannot be used without some manipulation.
The limits are:

--
Definitions
Rolling takes place in the positive x direction. The co- (a = gb) 22Guba 22Gab13 2416Gua2by log (g).
F,
ordinate system is chosen in such a way that the contact Il+x $)(l-aypY -7675( - 1-a)?;-
area does not move. (&, yk), with k = 1, 2 are the rigid
velocities of the bodies at the centre of the contact area;
-k > 0 is the rolling velocity, d is the angular velocity of
Mz -
2176Gaa%/3 log (g) 136Gab3y
675(1-u)?; +225(1-u)
When u = 0, the term with y in F,, and the term with
the upper body (indicated by k = 2) with respect to the ,B in M, vanish. Instead, we have
lower body (indicated by k = l), about an axis through
thc centre of the contact area and perpendicular to the (F,) : rGaby/3, (M,) : -~G~*bi5/3;
plane z = 0 in which the contact area lies. (b -
= gal 10Ga2u
The motion of the bodies relative to each other can be - F,---------.3 1% (gS
represented by the three parameters a, 13, y which are +x

defined as follows: lOGa2,B ~a3Gy .


Fy -
13 = (Y2-Y1)/(-4, y = Q/(-*);
N
3(1-u) log (g)-3(1-a) log (g)

(-1
a = (22-*1)/(-4,
a>P, Y < 1
The contact area is given by (x/a)2+(y/b)2 = 1. The
rGa3P
+-
184 1 - 2 ~
M z N - 3 (1 -~ ) log (g) 225 1-u Ga2b2y
largest half-axis is denoted by c. The eccentricity of
~_ We observe that, when a = gb, C = -D, only
the ellipse is e = dl-(a/c)2 when c = b 2 a, and e = approximately. This is probably due to the fact, that we
>
- l/l-(b/c)? when c = a b. The ratio of the shortest have not used enough equations. As for an ordinary
to the largest half-axis is denoted by g; apparently g 1. < ellipse we used only 10 equations.
The normal pressure is Hertzian. G is the modulus of In some way or other, our results can be connected with
rigidity, and o is Poissons ratio. Fx, Fy, and M , are the the value of the constant C, introduced by Dr Haines and
components of the tangential force and the twisting couple Dr Ollerton. I suspect that the slope of the C curve can be
Proc Imtn Meeh Engrs VoI I77 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES 107

Table 1. A, B, C, D,and E as functions of e, with u = 0.3 Table Za. A , B, C, D,and E as functions of u, with e = -0.9

e lI(e, A B i(e, 0.3);


l 0.3)
O.3),(ey c (e,

-0.9 I 2.44 1 2.39 I 0.88 1 -0.88 1 0.14 I 0.44 0 1 2.05 2.05 I 0.76 I -0.76 1 0.18

0.1 I 2.17 i 2.15 I 0.80 I -0.80 I 0.17

-0.5
3.27
1-28 -1.29 0.65
-- _ _ _ _ _ ~ _ _ _
3.83 3.43 1.35 -1.36
I
0.80

0.80 0.87
1 0
yfi
' ~ - _ _ _ _ _ _ _ _ - ~
0.16
____-
0.14

0.13
x
-0.4
-__ -
4.00 3.55 1.41 1 -1.42 1 0.93 I 0.92 0.1 1
-0.3 4.12 3.63 1.45 1 -1.46 1 1.03 1 0.95

-0.21 4.21 3.69 I 1.48 I -1.49 ~ 1.11 1 0.98

Table 2b. A, B, C, D,and E as functions of u, with e = O*

0
!1 (051
3.40
~

1
(0%
3.40
Ii &)
1.33
I
~
4
:
(
-1.33
I1 (0%
1.21

+0.7 1 2.85 1 2-38 I 0.71 1 -0.72 1 0.75 I 0.71 0.5 5.20~3.98


-~___________--
1.63 -1.63 1 ! 1 1.16

Johnson's experimental results (steel) (7)(16)


0-3 I
I
4.19 1 3.64 1 1.56
I
- [ 1.011.3

* M y 'circle' results were mentioned by Professor A . D. de Pater in a


paper read by him at the General Motors Symposium on Rolling
Contact Phenomena, held at Detroit in October 1960. ( T o be
published, seep. 114.) Table 2c. A , B, C, D,and E as functions of u, with e = +O-9
found; I have, however, been unable to find it, because I
do not understand the relation (3) between Ifxi = ( C L / and
the quantity (ao-ao'). I shall give here the pertinent values
of F,.
u = 0.39, alb = 1 : F, = 4-63aZGa;
u = 0.50, a/b = 1 : F, = 514azGa;
u = 0-39, a/b = 0-5: F, = 2.09bZGa;
u = 0.50, a/b = 0.5: F, = 2.40bzGa;

,
a = 0.39, aib = 2.0: F, = 2.81a2Ga;

I 1
u = 0.50, a/b = 2.0: F, = 3.04a2Gu;
u = 0.39, a/b = 1.5: F, = 3-41a2Ga; 0.5 2.08 ~ 1.32 0.29 ~ -0.30 0.48
0 = 0.50, a/b = 1.5: F, = 3-73a2Ga;
u = 0.39, a/b = 0.4: F, = 165b2Ga;
u = 0-50, alb = 0.4: F, = 1.91b2Ga.
The calculation of this last list took 40 min on the 'Zebra' (11) HETENYI,M. and MCDONALD, P. H. 'Contact stresses
computer. under combined pressure and twist', J. Appl. Mech.,
Trans. Amer. SOC.mech. Engrs 1958 80, 396.
REFERENCES (12) MINDLIN,R. D. and DERESIEWICZ, H. 'Elastic spheres in
(These do not form a complete bibliography) contact under varying oblique forces', J . Appl. Mech.,
(10)OLLERTON, E. Photoelastic investigation of contact stresses Tram. Amer. SOC.mech. Engrs 1953 75, 327.
between curved surfaces under radial loads 1959 Thesis (13) LOVE, A. E. H. A treatise on the mathematical theory of
presented to the University of Nottingham for the degree elasticity 4th edition 1927, 243 (Cambridge University
of Ph.D. Press).
Proc Instri Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


108 COMMUNICATIONS

(14) LUBKIN,J. L. The torsion of elastic spheres in contact, results obtained confirm the theoretical tests in a satisfac-
J . Appl. Mech., Trans. Amer. SOC. mech. Engrs 1951 tory manner.
73, 183.
(15) CATTANEO, C. Annu. Sci, Norm. Sup., Pisa 1952 6 (Series Though the strip theory employed in these investigations,
3), 1. (In Italian.) and particularly the assumption that the one strip does not
(16)DERESIEWICZ, H. Contact of elastic spheres under an act upon another, i.e. the assumption of two-dimensional
oscillating torsional couple, J . Appl. Mech., Trans. theory as given by Carter and Poritsky, may give rise to
Amer. SOC.mech. Engrs 1954 76, 52. doubts, the results obtained are very satisfactory. These
(17) JOHNSON, K. L. The effect of spin upon the rolling motion
of an elastic sphere on a plane, 3. Appl. Mech., Trans. results also serve to explain more accurately the distortion
Amer. SOC.mech. Engrs 1958 80, 332. of contact area and the loss of energy in rolling.
(18)GALIN,A. S. Conlact problems in the theory of elasticity When there is actual contact between two spherical
1961 (State College Publication, Raleigh, Korth surfaces having anisotropic roughness, the shape of the
Carolina).
(19)DOVNOROVICH, V. I. Three-dimensional contact problems of lemon-like adhesion area may undergo changes, relative to
the theory of elasticity 1959 (Minsk). (In Russian.) the direction of the tangential force, depending on the
(20) JOHNSON, K. L. The influence of elastic deformation upon ratio of the tangential force to the radial force. It is not
the motion of a ball rolling between two surfaces, Proc. easy, however, to determine the degree of this deformation,
Instn mech. Engrs, Lond. 1959 173, 795. e.g. in machine parts.
Dr K. L. Johnson, M.A., M.Sc.Tech. (Associate
Member)-The authors have devoted a large part of their Mr S. Wise (Member)-The work described in this
paper to the reconciliation of an inexact elastic theory with paper is part of a comprehensive investigation into the
measurements upon an inelastic material. In that discussion stresses set up in rolling contact between wheel and rail,
the positive achievements of an exceedingly difficult inves- and the authors are to be congratulated both on the quality
tigation have become somewhat obscured. of their experimental techniques and on the value of the
Perhaps the most striking of these achievements is that resulting paper.
Figs 7, 8, and 9 demonstrate, for the first time in any It is of interest to note that the results obtained concern
direct way, the validity of the assumption that the tangential the fundamentals of railway engineering, for not only do
traction in the slip region is a constant proportion of the they help the reader to understand the critical relation
normal pressure at any point. All theoretical work in the between wheel diameter and axle load that is necessary to
field so far has been based upon that proposition, so it is prevent spalling failure between wheel and tyre, they also
comforting to have its validity confirmed. have direct relevance to the problems of adhesion and wear
The earlier investigation of the problem (7) left the between wheel and rail.
actual shape of the area of adhesion as an unknown quantity. One interesting aspect of this general problem not dealt
That shape has now been established by the authors and it with in the paper, and one which would form a valuable
is a merit of their theory, approximate as it is, that it subject for further research, is the effect of the presence of a
predicted an area of adhesion whose shape agrees with the liquid phase between the two members nominally in contact.
observations. Other work on this subject which is now being done by
The measurements of surface stress in Figs 13, 14, and Dr Ollerton has shown that under conditions of dry contact
15 are remarkable for their agreement with the theory. It the fatigue strength of the contact areas is very much higher
appears that the problem of contact stresses due to rolling than might be expected. Since spalling failures do occur,
with a longitudinal traction force has now been effectively particularly on tyres and occasionally on rails (especially
solved. An exact elastic solution would be an achievement rails on American railways), it seems that some other cause
in applied mathematics, but would add little to our engineer- must be sought than direct Hertzian contact stresses. The
ing knowledge. experiments that Dr Ollerton is now making show that
One question arises. Considerations of strength of metallic fatigue strength is very greatly reduced by the presence
solids in rolling contact focus attention upon sub-surface of a liquid film, whether this be oil or water. Similar results
stresses. Have any measurements been made which would have been observed at the National Engineering Laboratory,
yield the variations in orthogonal shear stress below the where it has been suggested that the effect of the liquid is
surface ? Whilst a knowledge of the surface tractions makes primarily due to its capacity to enter the fine transverse
possible the numerical computation of sub-surface stresses cracks which are sometimes present in the contact areas.
such computations would be fairly elaborate and have not Christiansen* and others, however, have reported a very
yet, so far as I know, been attempted. considerable increase in contact pressure much above
Hertzian values when a liquid film is present between two
Mr Stanislaw Pytko (Krakhw, Poland)-The paper rolling bodies with critical separation, and it would therefore
by Dr Haines and Dr Ollerton constitutes a further stage be of great value eventually to know how the contact areas
in the work that aims at determining the contact area of and contact stresses described in this paper would be modi-
two balls subject to radial and tangential forces. The authors fied in the presence of a liquid film.
have succeeded in defining the adhesion area within the * CHRISTIANSEN, H. The oil film in a closing gap, Proc. roy.
contact surface by the use of photoelastic techniques. The SOC.A . 1962 266,312.
Proc Insrn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


109

Authors R eId dv
D r D. J. Haines and D r E. Ollerton-We should like to contact areas under the same conditions. Cains discussion
begin by thanking the contributors for their kind remarks did not embrace spin or transverse creep. The case of a
about the investigations. A few questions have been flat contact surface which supports radial and transverse
raised which require comment and these are discussed forces is being studied by Dr Haines at Bristol University
below. and it can be shown theoretically, and has been confirmed
Professor de Pater questioned the validity and usefulness experimentally, that the shape and position of the resulting
of the strip theory. We are aware that the strip theory is adhesion area are the same as those observed when traction
approximate. The problem under investigation was the is longitudinal. When traction is transverse the distribu-
stress distribution over contact areas between railway tion of traction on the adhesion area is found to be a
wheels and rails, and in railway practice the tractive forces function of Poissons ratio. In the remaining fundamental
often approach the limiting value and occasionally achieve problem, that of spin during rolling, we would refer
it. The theory allows reasonably accuracte prediction of Mr Kalker to the experimental results obtained by Dr K.
the state of stress over an elliptical contact area throughout L. Johnson (20) in which adhesion area boundaries are
the whole range of tractive forces from zero to the limiting presented for this problem.
value when the tractive force coincides with the direction The results quoted by Mr Kalker in the Table on p. 107
of rolling. have been compared with the strip theory results calculated
We wish to congratulate Mr Kalker on his theory for in two ways.
the case of a vanishingly small slip, which will be discussed
in more detail later. Mr Kalker is mistaken in his belief First method. Carter (8) has shown that the adhesion area
that Dr Ollertons thesis (10)is concerned with the shift surface strain on a strip taken parallel to the rolling
problem of Cattaneo (I) and Mindlin (2). As the title direction when ao/60 = 0 is given by
states, the investigation was restricted to radial loads. _au - 2(1--v)T(a0--a,)
- . . * (5)
Neither of the authors has conducted experiments with a 8X aGuo2
direct bearing on the Cattaneo-Mindlin problem. where T is the limiting tangential force per unit width of
The manner in which the Cattaneo-Mindlin distribu- the contact area.
tion of static shear stress changes to the distribution I n an elliptical contact area the maximum pressure
under steady rolling conditions is certainly interesting. 3 P
No experiments were performed to study this because the occurs on the central slice, having a value of -
2n ~ o 6 o
-
state of stress in the transition period is of limited practical The maximum surface shear stress has a value of
value to the engineer. Our main concern in the tests 3 P
described was that the rolling should proceed far enough at this point, and the shear force per unit width
to ensure that a stable adhesion area was established at the PZ.rra,b,
leading edge of the contact area. The distance rolled was is equal to the area under the shear stress distribution
three to four times the length of the contact area in the curve or
direction of rolling.
The redistribution of pressure referred to on p. 102 of
the paper is that induced by the visco-elastic behaviour Substitution in equations (3) and (5) gives
of the model material. The pressure distribution during 3pP(l - v ) ( u ~ - u ~ )
rolling would not be exactly Hertzian and could not be = 7~Ga~~6,
. . . (6)
determined photoelastically because the models had to be Equation ( 6 ) can be used together with Fig. 3 to relate
cooled slowly after rolling had ceased, which allows ample 6, to Pi.There is no simple relation between f, and
time for some redistribution of pressure to occur.
Pt/pP, but the case which Mr Kalker considers is given
Cain (6) proved to our satisfaction that the adhesion
by the slope at the origin of the curve in Fig. 3. At this
area adjoins the leading edge of the contact area in the
case of parallel cylinders in rolling contact under radial pt 37r
point is equal to -and, when it is substituted
forces and driving torques. This was the subject of 4
p P (1-2)
Poritskys paper (5). The proof is valid for elliptical
Proc Instn Mech Engrs Vo1177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


110 AUTHORS REPLY

in equation (6), In each of these three equations the limiting and negative
2.3567rG~,4of~ traction stresses are given by considering the xz plane of
Pt =
3(1--~)
.
. . (7) the driven models as being independent of the remainder
of the model. In all but one case the Z,, Z,, and X ,
This last equation answers Professor de Paters query.
traction stresses correspond to the actual surface values
Equation (7) has been evaluated for comparison with
of Z,; in the remaining case the coefficient l/C is intro-
Mr Kalkers Table on p. 107. The above method of
duced because it is found that the surface X,-Z, tractive
approach uses the line-contact values of &/ax as obtained stress in the driven model corresponds more nearly to
by Carter (8) for both the limiting positive shear distribu-
the negative traction 2, stress according to the simple
tion and the negative shear distribution over the adhesion
strip theory than it does to the actual negative traction.
area. In the surface these equations correspond to the graph
Second method. A more accurate result may be obtained lines shown in Figs 7-9 and to the strip theory graph
by using &/%x values as given by Mindlin (2)or Vermeulen lines shown in Figs 13-15. Reference (21) contains a
and Johnson (3) for limiting shearing tractions, and strip detailed discussion of the reasoning leading to these
theory for the determination of the extent of the adhesion equations.
area. (aujax must be constant over the adhesion area, so The sub-surface limiting and negative traction stresses
the 2ujax value associated with the negative traction must predicted by equations (8)-(10) may be calculated by
have the same proportionality relation to x as that associa- means of equations obtained by Smith and Liu (22).
ted with the limiting positive traction.) [Equations (8), (9), and (10) contain several approxima-

I
tions, but unless a stress function is discovered that can
V FxlGa be added to the limiting traction surface solutions within
the adhesion area, yielding mathematical results similar
aoiba Kalker Strip Exact limiting traction to Fig. 10, and can take care of hysteresis, the exact bound-
theory theory +strip theory

8% 1
alone adhesion area ary conditions for the system will remain unknown.
___ Further, an inspection of Mindlins paper will show that
0.39 1.0 I 4.63~
5.14~
4.05~
4.93~ certain of the functions which disappear in the contact
0.5
0-39 2.09b 2.02b2 surface, but which are required for sub-surface stresses,
0.5 240b 247b2 have never been defined even in the case of limiting
0-39 2.0 2.812 2.02~~
0-5 2-0 3.042 2.47a2 traction.]
0.39 1.5 I 3.41~ 2.70~
0.5 1.5 3.73~ 3.29~~
0.39 0.4 1 1.65b2 1.62b2
0.5 0.4 1 191b 1.97b2
Measured sub-surface stresses
The comparison shows that as a,/b, + 0 the strip theory Z, systems. Figs 16-18 show the surface and sub-surface
and modified strip theory values are similar to Mr Kalkers frozen Z, systems on the central planes of models 3B,
values. The strip theories are entirely accurate when 4B, and 5B. They also show theoretical lines which have
aolB, = 0. been computed from the calibrations and equation (9).
It is difficult to see how Mr Kalkers linear theory for a Of the three sets of results the one with the smallest
negligibly small traction could be extended to the case negative traction is shown in Fig. 18 and a comparison
where the applied traction is a significant fraction of the between theory and experiment in this case indicates the
limiting value. When slip occurs within a contact surface approximate validity of the limiting traction 2, assumption
the force-displacement relation which exists is non-linear. in equation (9). Conversely, the largest negative traction
We agree with Mr Pytko that in the presence of non- occurs in Fig. 17 and the approximate agreement here
uniform surface roughness the distribution of surface substantiates the negative-traction 2, assumption. (For
traction will differ from the simple system studied in the Fig. 17 C = 0-975 and for Fig. 18 C = 0.847.)
paper. Sternberg and Muki (23) have shown that this 2,
Dr Johnson and Mr Wise draw attention to the im- stress is independent of Poissons ratio and, therefore,
portance of sub-surface stresses. These stresses have been the agreement between theory and experiment in these
measured in the central xz plane of each of the driven figures shows the agreement which probably exists in
frozen-stress models, and compared with predictions practice if the coefficients of friction are 0.94, 0.82, and
obtained from the equations listed below. 0.81 respectively. A more realistic coefficient for a large
number of practical problems is 0.3 or less. The agreement
Z, (total) = Z, (Hertz)+Z, (limiting traction) which then exists can be obtained by the appropriate
+Z, (negative traction) (8)
reduction of shearing traction parts of the total stresses
2,(total) = 2,(Hertz)+Z, (limiting traction) throughout the figures. The results when p = 0-3 have
+Z, (negative traction) (9) been studied and are conveniently summarized in terms
X, (total) = X, (Hertz)+X, (limiting traction) of the ranges of Z, shearing stress which the material
+(l/C)X, (negative traction) (lo) experiences while passing beneath a contact surface.
Proc Instn Mech Engrs Vol177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


CONTACT STRESS DISTRIBUTIONS O N ELLIPTICAL CONTACT SURFACES 111

These results are shown on the figures and indicate view. They are, however, important in the calculation of
that the theoretical and prototype ranges of stress are in the sub-surface maximum ranges of shearing stress,
close agreement in the vicinities of the sub-surface and these sub-surface maximum ranges of shearing stress
maxima. can be expected to influence the failure through spalling
of bodies which are in rolling contact.
X,-Z, systems. The theoretical and experimental values
of this stress difference in each of the three models are Maximum ranges of theoretical and actual shearing stress.
shown in Figs 19-21. For each figure the theoretical values The most convenient way of obtaining these results is by
have been calculated from equations (8) and (10). The
sub-surface agreement between theory and experiment is
not comparable with that for the 2, systems, but, if the
values are adjusted to correspond to a prototype p of 0.3
the discrepancies are appreciably reduced, It might further
be argued that these sub-surface X, and 2, stresses
are not in themselves significant from a fatigue point of

L ! .+
H E R T Z PLUS APPROXIMATE ,

-+ ; +-+-
F

-1001 I I I 1

Summary of results for p = 0.3


Summary of resultsfor p = 0.3
I
I Range of Z, shear stress (fringes/in.)
zlao I Range of 2, shear stress (fringeslin.)
zlao
i theory I experiment 1 discrepancy, per cent
I theory I experiment I discrepancy, per cent
0
zz 64
0
0.105 1 81
116 1 123 -6
0.105
0.194
0.306
105
109 1
92
103
105
1
2
4
0.194
0.306
0.404
138 142
-;+
-3
-

Fig. 16. Measured and calculated Z, shear stress on the Fig. 17. Measured and calculated Z, shear stress on. the
central plane of model 3B. ao/b, = 1.00, p = 0.94, central plane of model 5B. aQ/bQ= 2-00, ;E = 0.82,
PTlpP = 0.67 PTIpP = 0.46
Proc Imsn M a h E w s Vo1177 No 4 1963

Downloaded from pme.sagepub.com at Gazi University on January 9, 2015


112 AUTHORS REPLY

the construction of a series of Mohr stress circles for system of diagrams for the position Z = 0.105 ao,
points along y = 0 lines at given depths below the con- p = 0.3 in model 3B, and these diagrams will be used to
tact surface. The maximum range of shearing stress in a illustrate the method of solution.
given direction (at a given depth) could then be obtained From the axes used in the figure it will be apparent
from these diagrams by trial and error. This process can that each point plotted corresponds to a point on a con-
be simplified if all the stress circles for a given depth are ventional Mohr stress circle. A sufficient number of these
superimposed on one another (24). Fig. 22 shows such a points must be plotted to permit a smooth curve to be
drawn through them. The maximum range in the hori-
zontal direction is then given, as shown in the diagram,
by the distance between the horizontal tangents. (This is
true because the Z, shearing stress at each point is repre-
sented by its vertical component.) Similarly, the maximum
range of shearing stress in the plane is given by the maxi-
mum distance between any two parallel tangents, and the
direction in which this shear acts (0) can be read from the
diagram.

HERTZ PLUS APPROXIMATE THEORY


200, &7-
-7

HERTZ PLUS APPROXIMATE THEORY

Summary of results for p = 0.3 HERTZ PLUS APPROXIMATE THEORY


2 o o r i
Range of Z, shear stress (fringes/in.)
~ theory I experiment
~-
discrepancy. per cent
-._
C

n
100
0

1
I
1 m
0 92 ._
C

0.105 102 I 92
106 -4 k o
122 -4
-64
105 ~

I
-6
-- 100
100
I 1
I

Fig. 18. Measured and calculated Z, shear stress on the Fig. 19. Measured and calculated X,-Z, on the central
central plane of model 5B. a,lb, = 2.00, p = 0.81, plane of model 3B. a&, = I.00, p = 0.94, PT/pP =
r,ipr = 0.812 0-61
Proc Instn Mech Engrs Vol 177 No 4 1963
Downloaded from pme.sagepub.com at Gazi University on January 9, 2015
CONTACT STRESS DlSTRIBUTIONS ON ELLIPTICAL CONTACT SURFACES I13

Repeating this process for all other sub-surface cases The theoretical and experimental maximum ranges of
being studied gives Table 3. 6 is the angle which the plane shearing stress show close agreement and this is attributed
sustaining the maximum range of shearing stress makes to the fact that 6 tends to be small, hence the discrepancies
with the x and z axes. in the X,-Z, values do not greatly influence these final
It will be apparent from Fig. 22 that the value of 8 results.
which is obtained by this method may be subject to error,
but, in each case the theoretical and experimental H
values are similar and this is the reason why only one
value is quoted.

L
F

--
HERTZ PLUS APPROXIMATE THEORY
300 r--+ _.
I
I
I

HERTZ PLUS APPROXIMATE THEORY

- 1

HERTZ PLUS APPROXIMATE THEORY


200

-6 100
._
u)
P)
m
2 0
c

-100 -100'

Fig. 20. Measured and calculated X,-Z, on the central Fig. 21. Measured and calculated Xz-Zz on the centraI
plane of model 4B. a,lb, = 2.00, p = 0-82, PTIpP = plane of model 5B. a,lb, = 2.00, p = 0.81, PT/p,P =
0.46 0.812
Proe Instn Mech Engrs 1701 177 N o 4 1963
Downloaded from pme.sagepub.com at Gazi University on January 9, 2015
114 AUTHORS' REPLY

TabZe 3. Maximum range of shearing stress in fringesiin. [ p = 0.3)


I
51 Model 3B
i Model 4B II Model 5B
a,
1 __-
Theory , Experiment 8 i Theory I Experiment 1 B j Theory , Experiment ~
__--
8
-

0.105 102 101 18"


0.194 109 105 1 7"
0.306 110 108 2"
0.404 - 1
i - l

The tables of maximum sub-surface ranges of shearing


stress and of Z, shearing stress (see Figs 16-18) justify
the calibration techniques and the use of the approximate
equations (8)-(10).
It is, therefore, considered permissible to employ these

c
equations, which are exact for line contact, in the study of
the surface stresses and the ranges of sub-surface shearing
stress that occur in practical problems where the driving
or braking loads are applied to flat smooth Hertzian contact
J: surfaces and the coefficient of sliding friction does not
greatly exceed 0.3.

REFERENCES
(20) JOHNSON,K. L. 'Tangential tractions and micro-slip in
rolling contact', Rolling contact phenomena (Proceedings
of a symposium held at the General Motors Research
Laboratories, Warren, Michigan, in October 1960) 1962
(Elsevier Publishing Company, Amsterdam).
(21) HAINES, D. J. A photoelastic investigation of wntact stresses
betmeen curved surfaces under radial and tangential loads
1961 Thesis presented to the University of Nottingham
for the degree of Ph.D.
(22) SMITH, J. 0. and LIU,C . H. 'Stresses due to tangential and
S C A L E : 25 f r i n g e s / in. = 1 SQUARE
normal loads on an elastic solid with application to some
contact stress problems', 3. appl. Mech., Trans. Amer.
x Experimental values. SOC. mech. Engrs 1953 75, 157.
0 Theoretical values when p = 0.3. (23) STERNBERG, E. and MUKI, R. 'Notes on the expansion in
The points shownare at intervals of O.lao from -l.lao to +l.lao. powers of Poisson's ratio of solids in elastostatics', Arch.
Rational Mechanics and Analysis 1959 3, 229.
Fig. 22. Diagrams for the calculation of shear stress (24) OLLERTON, E. An unpublished discussion at the Institute
ranges for model 3B. z = 0.105ao of Physics on 14th May 1958 of contact stress problems.

Proc Instn Mech Engrs Vol177 No 4 I963


Downloaded from pme.sagepub.com at Gazi University on January 9, 2015

Você também pode gostar