Proceedings of the Institution of Mechanical

Engineers
http://pme.sagepub.com/

An Analysis of Plastic Deformation in Rolling Contact

Applied Mechanics Group, J. E. Merwin and K. L. Johnson
Proceedings of the Institution of Mechanical Engineers 1963 177: 676
DOI: 10.1243/PIME_PROC_1963_177_052_02

The online version of this article can be found at:

http://pme.sagepub.com/content/177/1/676

Published by:

http://www.sagepublications.com

On behalf of:

Institution of Mechanical Engineers

Additional services and information for Proceedings of the Institution of Mechanical Engineers can be found at:

Email Alerts: http://pme.sagepub.com/cgi/alerts

Subscriptions: http://pme.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://pme.sagepub.com/content/177/1/676.refs.html

>> Version of Record - Jun 1, 1963

What is This?

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014

676

APPLIED MECHANICS GROUP

AN ANALYSIS OF PLASTIC DEFORMATION IN ROLLING

CONTACT

By J. E. Merwin, M.S., Ph.D.* and

K. L. Johnson, M.Sc.Tech., Ph.D., M.A. (Associate Mernber).f-

When two metal cylinders roll together under a contact pressure sufficient to cause yielding, a surprising
mode of plastic deformation occurs. The surface of each cylinder is progressively displaced in the forward
direction of rotation relative to the core by plastic shearing in a thin subsurface layer. This phenomenon
was first observed by Crook (6) in 1957 and the results of a more complete experimental investigation are
reported by Hamilton (9) in an accompanying paper. In this paper an attempt is made to explain the
mechanism of ‘forward flow’ by an approximate numerical analysis of the elastic-plastic stress cycles to which
the material is subject in repeated rolling contact. On the basis of an idealized material which is elastic-
perfectly plastic and isotropic, it is shown that a forward displacement of the surface would be expected as
a result of the complex cycle of stress and strain encountered in rolling. It is also shown that residual com-
pressive stresses are introduced into the immediately subsurface layers during the first few cycles of the
load. If the load does not exceed the elastic limiting load by more than 66 per cent further plastic deforma-
tion would then cease. At higher loads a steady pattern of plastic deformation is predicted which in its
principal features is consistent with the observed behaviour.

INTRODUCTION mation and, further, why the flow should occur in the
IF TWO METAL CYLINDERS are rolled together under a forward direction. For example, it was suggested that a
sufficiently large pressure to cause yielding of the material, lubricant film between the discs would lead to an asymmetri-
an interesting type of deformation takes place. This defor- cal distribution of contact pressure, but experiments
mation is principally one in which an elastic surface layer reported by Welsh showed that ‘forward flow’ was observed
rotates with respect to the elastic core of the cylinder. The in the absence of a lubricant. Again, the Bauschinger effect
two elastic portions are separated by a narrow band of and also the upper-yield-point phenomenon were cited as
plastically deformed material. In the absence of any tan- possible causes of unsymmetrical behaviour in a loading
gential forces, the direction of rotation of the surface cycle in which the shear stress reverses its direction.
relative to the core is in the direction of rotation of the However, as stated at the time by Johnson (S), it is reason-
cylinder. able to expect an asymmetrical contact stress field on general
These observations, when they were first described by grounds. The occurrence of repeated plastic deformation,
Crook (6)$ and Welsh (7), gave rise to considerable specula- whether cumulative or not, involves continuous energy
tion as to why a purely normal force acting between the dissipation. Thus, for steady rolling motion to proceed
cylinders should produce an asymmetrical pattern of defor- external work must be supplied, which demands that the
reaction force between the discs be displaced from the line
The M S . of this paper was received at the Institution on 1st Ociober of centres towards the entry side. A further consequence of
1962.
* Assistant Professor, Department of Civil Engineering, Rice plastic deformation is that the state of stress at any point
University, Houston¶ Texas. in the field depends upon the entire loading history of an
.f- Lecturer in Engineering, Cambridge University. Associate Member
of the Institution. element of material before it reaches the point in question.
$ References are given in the Appendix. It appeared to us, therefore, that any detailed and convincing
Proc Instn Mech Engrs Vo1177 AT0 25 1963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
 AN ANALYSIS OF PLASTIC DEFORMATION I N ROLLING CONTACT 677

picture of the process must trace the stress and strain history magnitude of elastic and plastic strains will be the same. Up
of material elements as they pass repeatedly through the to now solutions in the mathematical theory of plasticity of
contact stress field. problems of contained plastic deformation have been ob-
In the meantime Hamilton* (9) has conducted an exten- tained only for some simple cases. An exact solution, in
sive experimental investigation of the phenomenon which which conditions of equilibrium, compatibility and elastic-
provides a sound basis for analysing the mechanism of the plastic constitutive relations are simultaneously satisfied
process. Hamilton has shown that the phenomena occurs throughout the solid and on the boundaries, seems to be
with all of a variety of metals and is not restricted to steel out of the question at the present time. A principal difficulty
discs used by Crook and Welsh. He observed that the arises from the unknown shape and location of the elastic-
deformation in the first cycle of load was markedly different plastic interface. However, an approximate solution has
from subsequent loadings. A steady state was then quickly been found which exhibits all the important features of the
reached when the total forward displacement of the surface observed behaviour and, for loads not greatly exceeding
layer became cumulative in the sense that it was directly the shakedown limit, is not likely to differ appreciably from
proportional to the number of revolutions of the cylinder. an exact solution.
If two discs of the same metal were geared together so that The problem of two rolling cylinders will be simplified
they ran at exactly the same surface speed their surfaces by the following idealization:
became corrugated, a subsidiary effect which interfered (1) The system is replaced by a rigid cylinder rolling on
with the process of forward flow. It was found that the
the plane surface y = 0 of a semi-infinite solid y < 0
development of corrugations could be eliminated either by (see Fig. 21 for the co-ordinate system and the notation
permitting a slight amount of ‘slip’ between the discs or by
for stresses).
rolling against a harder disc which did not deform plastic- (2) The material of the solid is elastic-perfectly plastic
ally. When the tendency to corrugate was prevented, the
(i.e. non-work-hardening) and isotropic.
forward displacementper cycle was found to be independent
(3) The deformation is plane, hence the axial strain
of the speed of rolling. Finally, by measuring the rolling
eZ is zero and all other stresses and strains are independent
friction, Hamilton estimated the energy dissipated in plastic
of 2.
deformation. This was found to be appreciably greater than
the work which would be necessary to produce the observed The implications of these idealizations will be discussed
forward displacement by a process of simple shear. later.
Hamilton’s experiments appear to support the hypothesis ELASTIC CONTACT
that, in the absence of corrugations, forward flow arises The elastic solution for the contact of two circular cylinders
from the elastic-plastic stress-strain cycle in rolling contact or a cylinder with a plane is well established and is derived
rather than from ‘special’ material properties. It was felt, from the classical Hertz theory. Since the contact region is
therefore, that a relatively simple idealization of an elastic- smaIl compared with the other dimensions, a cylinder can
plastic material could be employed in an analytical study of be considered as a semi-infinite solid which has a prescribed
the process. normal displacement on the surface of the form w = a-/3xz
Before proceeding to the analysis, certain other features over the interval 1x1 Q a where a and /3 are constants and
of the process may be anticipated in qualitative terms. At a is the semi-contact width. Hertz obtained the pressure
small loads the contact stresses will be everywhere elastic. distribution in the contact region necessary to produce the
With increasing load some point in the stress field will reach
its yield point, which may be determined from the elastic
stresses and a suitable yield criterion. This load will be
referred to as the elastic limit. At higher loads some plastic
flow will occur, thereby leaving residual stresses in the
tY
surface region of the cylinders after the load has passed.
During subsequent passes of the load the material is subject
to the combination of residual and contact stresses which
together may not reach the yield point. In other words the
system ‘shakes down’ to a resultant state of stress which is
entirely elastic. The maximum load at which shakedown
occurs will be termed the shakedown limit. Clearly, loads at
which cumulative plastic flow occurs must exceed the
shakedown limit.
It is well known that in elastic contact of solids the yield
point is first reached beneath the surface. Hence, for loads
not greatly in excess of the elastic limit, plastic deformation
will be fully ‘contained’by elastic material and the order of (Co-ordinate system and notation for stresses and strains.)
* Hamilton’s paper should be read i n conjunction with the present Fig, 21. Rigid cylinder rolling on an elastic-perfectly plastic
work. plane
Proc Znstn Mech Engrs Vol177 No 25 1963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
678 J. E. MERWIN AND K. L. JOHNSON

prescribed displacements. Later Beliaev ( I o), Radzimovsky

(II) and others obtained solutions for the complete internal
stress field by various methods. These solutions are exact
for a rigid cylinder in contact with a semi-infinite solid and
closely applicable to two elastic cylinders in the vicinity of
the contact zone.
In plane strain the elastic stresses may be expressed:
4P
u,+ul, = - Q [ i e - r ]
nu
2P .cosh S-cosh
=~ - z
U , - U , + ~ ~ T ~
Ta sinh 5
and
u* = v(u,T+ulJ
where x+@ = a cosh [, 5 = f + i T and P is the contact
J
force per unit axial length of the cylinder.
In this paper, the value of Poisson’s ratio, v, is taken to
be 0.3.
We can now determine the elastic limit. The yield condi-
tion chosen is that proposed by von Mises which states that
the second invariant of the stress deviation tensor, J2,
cannot exceed some limiting value k2. The second invariant
is given by
J2 -
= +[(. u,)” + ->;.” +(a,-.,T>”l
(u,
+ + +722
.I

T1.U” TYZZ (2)

In simple shear Jz = 2, whilst in simple tension J2 = .+”.
Hence k may be interpreted as either the yield stress in
simple shear or the yield stress in tension divided by 4 3 .
In Fig. 22 are shown contours of constant values of
l / ( J z ) expressed in terms of the maximum contact pressure Maximum value is 0.322 at a depth 0.705a.
po. It can be seen in the figure that yielding will first occur Fig. 22. Contours of constant values of 2/( Jz)/po due to the
on the line of centres at a depth y = 0-705a when elastic stress distribution
po = 3-10K . . . (3) . .
In other words) only direct stresses acting parallel to the
R E S I D U A L S T R E S S E S A N D SHAKEDOWN* surface can remain, but they may vary, independently of
If the load exceeds the elastic limit then local plastic each other, with depth below the surface. S d a r l y , the
deformation will take place in a layer at a depth of approxi- possible residual strains are :
mately 0 . 7 ~and, after a passage of the load, the solid will be
left in a state of residual stress.
We now consider possible distributions of residual stresses
(6y)r

( d r =
}
= fdy), (Yzy)r == f h ) ,
= (~yZ)r= ( Y J r = 0
- (5)

and strains which can remain in a semi-infinite solid after The only non-zero residual strains are the direct strain
the passage of a rolling contact load. First, the assumption which accounts for compression of the solid normal to
( E ~ ) ~

of plane deformation eliminates ( T ~ ~ ( )T ~, ~, ) ~(&

, (yYz),, its surface, and the shear strain ( ~which,
3 if~it occurred
(yzz)rt and makes the remaining stress and strain com- with each repeated passage of the load, would produce a
ponents independent of z. Secondly, if the plastic deforma- cumulative tangential displacement of the surface.
tion is assumed to be steady and continuous, the solid will An insight into the way in which the residual stresses
retain a plane surface after loading, so that must be arise can be obtained from an inspection of the elastic stress
zero and the non-zero stress and strain components will field. To simplify the discussion, let us temporarily con-
also be independent of x. Thus, for the system to be in sider only stresses in the x-y plane$.
equilibrium with a traction free surface, it follows that The principal shear stress in the x-y plane is given by
(u& and ( T ~ , , ) ~cannot exist. The only possible system of T = ~1/(uZ-uy)”+4~zy2 . (6) .
residual stresses, therefore, reduces to : Its maximum value lies on the y-axis where T,, = 0.
(%)r = fi(Y)Y (4,= fdy),
}
( u g ) r = ( ~ z y ) r= ( T v J r = ( ~ z & = 0
(4)
The variation of T along the line of centres is shown
in Fig. 23. The absolute maximum value occurs at a
* This aspect of the problem has been treated in greater detail in a depth 0.786 and is equal to 0-300 p,. It should be noted
separate pager by Johnson (12). $ This simplz3cation amounts to substituting the Tresca maximum
f A subscript ‘r’ denotes residual stress or strain. = k, for the von Mises criterion.
shear stress criterion of yield, rmaX
Proc Instn Mech Engrs Vo2177 No 25 1963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
 AN ANALYSIS OF PLASTIC DEFORMATION I N ROLLING CONTACT 679

y,

-a

- 2a
YZ

Principal shear stress T is modified to T' by the addition of the

residual stress ( u ~ ;taken
)~ from Fig. 23.
Fig. 24. Variation of elastic stresses at a constant depth
y = -a
-36
Note that T may be limited to a maximum value k by the addition it follows that a residual stress distribution can be found
of residual stresses ( u ~in) ~the layer y z < y < yl.
which will eliminate yield according to the Tresca criterion
Fig. 23. Variation of elastic stresses with depth along the provided that the maximum value of T , ~anywhere in the
axis of symmetry field is less than k. Conversely, if ( T ~exceeds ~ ) k,~then~ ~
no residual stress system can prevent subsequent yielding.
that this maximum shearing stress, which occurs on the line We may now appeal to Melan's theorem (see Symonds (13))
of centres, is not the shear associated with the cumulative which states that : if any residual stress can be found which,
strain (y& observed in Crook's experiments. This stress, together with the stress due to the load, constitute a
which occurs on planes at 45" to the surface, leads to a system within the yield limit then, under repeated loading,
compressive displacement normal to the surface. Clearly the system will shake down to some system of purely elastic
the normal displacement must be rather small and reach a deformation. Conversely, if no such system of residual
limiting value, since the dimensions of either a cylinder or stress can be found then the system will not shake down
a plane solid cannot decrease indefinitely. Eventually, and continued plastic deformation will occur at each passage
therefore, residual stresses must exist which eliminate of the load.
further yielding in the normal direction. Now the maximum value of T~~ occurs at y = - 0 . 5 ~ ;
If we indicate some yield stress K in Fig. 23 it will be seen x = f 0 - 8 7 a and has the value 0 . 2 5 0 ~Hence
~ . we havefound
that T exceeds the yield criterion in the layer lying between a shakedown limit, i.e. for shakedown ( T ~ ~k, thus< ) ~ ~ ~

y1 and yz. Now the addition of residual compressive < 4.00K . . . . . (8)
stresses This limit may be compared with the elastic limit given in
(O,), = 2(k-T) . . . . (7) equation (3). For clarity of argument the Tresca criterion
would eliminate any further yielding on the y-axis. How- has been used above, but Johnson (12)shows that the von
ever, a residual stress which is capable of eliminatingfurther Mises yield criterion leads to the same shakedown limit.
yielding due to stresses on the y-axis, will not necessarily At loads in excess of the shakedown limit, the yield
prevent yielding on either side where T~~ # 0. To illustrate criterion will always be exceeded on either side of the axis
this point we examine the variation of elastic stress with x of symmetry in the regions where the elastic value T~~
at a constant depth y = -a in Fig. 24. Initially the principal exceeds K . It is now clear how a continuing plastic shear
shear stress T exceeds K by a maximum amount at x = 0. strain ( y ~arises;
~ ) ~it remains to investigate why the strain
The addition of a residual stress (u& given in equation (7) is cumulative in the forward direction, since the peak values
reduces the value of T to K on the y-axis, but it still exceeds of T~~ are equal in magnitude and opposite in sign.
k on either side, in the regions where T , has ~ its maximum
values. It follows from equation (6) that no possible residual NUMERICAL ANALYSIS OF THE
stress can prevent yield if ( T , ~ ) , , , ~ ~exceeds k, but for ELASTIC-PLASTIC S T R E S S CYCLE
(7,Jrnax <
k, a residual stress can be found which will To investigate cumulative plastic deformation under the
reduce the resultant value of T to less than K at all points action of loads greater than the shakedown limit we have
along y = -a. seen that it is necessary to trace the loading cycle of material
The same reasoning applies to other depths, from which elements as they pass through the stressed region. The
Proc Instn Mech Engrs Vol177 No 25 1963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
680 J. E. MERWIN AND K. L. JOHNSON

stress history of an element in elastic loading is given by where

curves of the form shown in Fig. 24; we require similar
curves to include the influence of plastic flow. w = sxdx+sydy+s,d,+T.ryj,y . (12)
In order to proceed to apply a theory of plasticity to the and a dot over any quantity denotes a time rate of change of
problem, we now assume that the strain cycle remains that quantity. These relations apply during plastic deforma-
identical with the elastic strain cycle which is determined by
equation (1). This simplification is not so unreasonable as
may appear at first sight. The deformable solid must have
boundary displacements within the contact region which
”
tion, in so long as J2 = k2 and w
> 0. For J2 < k2 or
< 0 the elastic equations (9) apply. The quantity t6’
calls for comment; it may be interpreted as the rate at
which the stresses do work in connection with a change of
conform to the circular profile of the rigid rolling cylinder shape. During plastic deformation the von Mises flow con-
whether the material yields or not. Similarly, two identical dition ensures that the elastic strain energy of distortion
rolling discs must deform to produce a flat interface under
both elastic and plastic conditions. The only change which remains constant, therefore corresponds to the rate of
~

plastic action can produce in the boundary displacements irreversible energy dissipation through plastic action.
would be an eccentricity of the contact strip. Now for loads Turning now to the rolling problem, the stress and strain
which do not exceed the elastic limit by a large factor, the rates may be expressed:
plastic zone will be small and contained which leads one d a
to suppose that the symmetricalelastic strain field will not be -
s[, * . 7 x y ; e,, * * .YXYI = U G s[, * * .TxY; ex, * * .Y,YI
appreciably distorted by plastic flow. . . . (13)
The stresses, on the other hand, are a different matter and
cannot exceed some yield limit. For the stress-strain rela- where U is the steady peripheral velocity of rolling. Sub-
tions in the plastic zone, the Prandtl-Reuss incremental stitution in equations (13) replaces the time rates of change
relations* will be employed. It is well to recall the postulates by stress and strain gradients with respect to x. U, and hence
upon which these relations are based; an elastic-perfectly time, do not then appear in the stress-strain relations.
plastic (i.e. non-work-hardening) isotropic solid which obeys Since the strains are assumed to be known (equal to the
the von Mises yield criterion and associated flow rule, and elastic strains) throughout the field, if the state of stress is
is deformed both elastically and plastically within the known at any point in the plastic zone, the rate of change of
confines of the small strain deformation theory. stress deviation with x at that point is given by equations
For the application of these equations it is convenient to (11). A step-by-step numerical analysis may now be per-
introduce the stress deviation and strain deviation. If we formed for the stress cycle of an element at any depthy.
let s = (a,+uy+u,)/3 and e = (c,+cy+~,)/3, then the Starting from the elastic stress state of the element at the
stress deviations are s, = u,--s, etc. and the strain point xo when it first reaches the yield point (J2= k2) on
deviations are e, = E,-e, etc. For plane strain the usual the entry side, the stress rates of change with x are found
Hooke’s law stress-strain relation for elastic deformation from equations (11) and used to predict the values of the
may be written: stress components when the element has moved a small
s, = 2Ge,, etc. distance Ax. Thus, step by step, the stress variations with
x for constant y are obtained, as illustrated in Fig. 2%-d.
At this point we might consider the implications of
and assuming that the strains are the same as in the purely
s=3Ke J elastic problem. Obviously these strains constitute a com-
where G is the elastic shear modulus and K is the bulk patible system; consistent and physically reasonable stress-
2(1+v)G strain relations have been used; the boundary conditions
modulus given by K = (prescribed surface displacements within the contact area,
3(1-2v)‘
During plastic deformation the von Mises flow rule zero surface tractions outside) have been satisfied. The
maintains that the second invariant of the stress deviation solution is inexact, therefore, only to the extent to which the
tensor J2 remains constant and equal to k2; i.e. in terms of stresses do not satisfy conditions of equilibrium. Clearly no
the stress deviations attempt is made to maintain equilibrium during the strain-
ing cycle, since stresses at one level are being modified
Jz = + ( S , ~ + S , ~ + S , ) ~ + T , ~ ~ = k2 * (10) independently of stresses at other levels. However, it is
The Prandtl-Reuss equations may now be quoted: possible to restore equilibrium at the end of each loading
cycle.
s,
[“I 1
= 2G k x + s s, ,etc. If the loading cycle had been entirely elastic the stresses
would approach zero after the element had passed under
the load. As a result of the plastic deformation, however,
residual stresses (a,)’,, ( o ~ ) ’(~a,)’,, and (rZy)’,.
are found as
+
x + co in the above-described computation. Such stresses
* See Prager and Hodge (14)for derivation of the equations and do not satisfy the conditions of equilibrium of a plane
notatzon. surface free from traction. These boundary conditions
Proc Instn Mech Engrr Vo1177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014

 AN ANALYSIS OF PLASTIC DEFORMATION I N ROLLING CONTACT 681

require (a,), = (T,~)? = 0 (see equations (4)). The com-

u , putation gives non-zero values (a,)‘, and ( T J , because
- 20 -a elastic recovery of strains has been assumed at the outset.
In order to satisfy the condition of equilibrium governing
the residual stresses, the strains are permitted to relax
~ 0. In this relaxation process
elastically until (cry)r = ( T , ~ )=
plane strain demands (cJT = 0 and symmetry conditions
demand that ( E , ) ~ = 0 (see equation (5)). The residual
strains (cy)r and (Y,)~ are thus given by:

-lJ

-20 - a
I I

Using the residual stresses (a2),. and (uZ), as initial condi-

tions in a repeated numerical integration, the result of a
second passage of the load can be calculated giving new
values of residual stresses. This procedure is repeated until
b
there is no further change in ( u ~and ) ~ (a&, i.e. the incre-
ment of (cy), per pass approaches zero. However, the
increment of (yXy)?per pass, the residual shear strain which
produces a tangential displacement of the surface, does not
u -
approach zero but reaches some constant value. Therefore,
- 30 -2a -a a X 20 30
the overall shear displacement is cumulative with repeated
traversals of load. The direction of (y,y)r corresponds to a
‘forward’ displacement as observed by Crook. If the tangen-
tial displacement of the surface per pass is denoted by 8, it
may be found from the steady-state values of (y,y)l. by
integration through the depth of plastic deformation,
namely :

Curves of a,, a,, a, and rZyat a depth ofy = - a are shown

in Fig. 25 for several traversals of the load. A steady state
t is approached rapidly.
SECOND AND
XY An unfortunate feature of the method, inherent in its
STEADY STATE basic approximations, is that the theoretical pressure dis-
tribution over the surface of contact cannot be determined.
However, the variation of uy on a representative plane,
y = constant, in the plastic zone (e.g. y = -a, shown in
Fig. 2%) is suggestive. For equilibrium, the integral of ow
from x = -a to x = +co must balance the normal
- -0.2p contact force. I n the steady state, at least, it appears that
the magnitude of the force is not very different from the
I ___I
U
d
Variations of the stress components (a) u ~ ,(b) u,, (c) uz and (d) T~~
I equivalent elastic case. It is reassuring to note that the
centre of pressure has been shifted towards the inlet side
with x at a constant depth, y = -a, and loadp, = 5% (Note:p , which overall energy considerations demand.
is the ‘elastic’ maximum Hertz pressure corresponding to a contact An energy balance may be used, in the steady state, to
width 2a.)
calculate the moment which must be applied to the cylinder
Fig. 25. Typical elastic-plastic stress cycles obtained from to overcome the resistance to rolling due to plastic deforma-
the numerical analysis tion. If this moment per unit axial length is denoted by M,
Proc Instn Mech Engrs Vol 177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014

682 J. E. MERWIN AND K. L. JOHNSON

recalling that the rate of energy dissipation per unit volume computation at each depth was repeated until an approxi-
is given by w in equation (12), external work may be mately steady state was reached (usually about four passes)
giving rise to stress cycles of the form illustrated in Fig. 25.
equated to energy dissipated:
The build-up of residual stresses (u& and (u,), with re-
MU/R = ~ddxdy . . . (17) peated loading is shown in Fig. 26. The whole computation
was performed for three values of applied load factor,
where the integration is taken throughout the plastic zone. namely
From equation (13) ~ is seen to be proportional to U, so Polk = 4.00,4.76 and 5.55
that M is, in fact, independent of rolling speed. The quantity In this expression p , denotes the value of the maximum
MIR may be interpreted as the force of rolling resistance. Hertzian contact pressure assuming elastic behaviour. The
actual maximum pressure may differ slightly from this
RESULTS OF THE COMPUTATIONS value.
AND DISCUSSION For the steady-state condition, the forward tangential
The integration of repeated stress cycles was programmed displacement of the surface per pass given by equation (16)
for the Cambridge University ‘Edsac’ computer. The and the rolling resistance, given by equation (17), are
Runge-Kutta-Gill technique (15) was used in the inte- plotted against the load factor in Figs 27 and 28 respectively.
gration process with an interval d x = 0.01~. The plastic The characteristic features of the steady-state pattern of
zone was covered at depth intervals of y = 0.20~.The plastic deformation are illustrated in Fig. 29 for a load factor

-;--ol

-a

- 2a

-k

-a

b I

(u) p , = 4.0k (shakedown limit); (6) p o = 4%; (c) p , = 5%

Fig. 26. Build-up of residual stress (u& and (u,), with repeated passages of the load
Proc Instn Mech Engrs Vo1177 No 25 I963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
 AN ANALYSIS OF PLASTIC DEFORMATION IN ROLLING CONTACT 683

LOAD FACTOR, po/h

Fig. 27. Forward flow per cycle in the steady state computed -3oJ
I
at the values of po/k shown

0.14
N9
O
2 0.12
\ /

FORWARD SHEARING
I-
u
a Zone of plastic action in the steady state. Note that this cor-
responds to a thinner layer than that which undergoes plastic
deformation during the first passage of the load.
b Steady cycle of plastic shear strain (y&. Although the shear
strain reverses sign the ‘forward’ shear outweighs the ‘backward’
shear to accumulate a residual strain (yz& with each passage
of the load.
LOAD FACTOR, po/k
Fig. 29
-0- First passage of the load.
-0- Steady state.
Fig. 28. Rolling resistance due to plastic energy dissipation which may be computed from Figs 27 and 28, with the
computed from the elastic-plastic stress cycles, at the results :
values of p a / k shown po/k Forward flow work/total work
4-76 0.32
Polk = 5.56. The layer in which cumulative plastic action 5.56 0.28
takes place is thinner than the layer which has experienced
initial plastic flow. Referring to Fig. 25d, the plastic compo- The authors have frequently been asked whether they
nent of shear strain (yly)l)is proportional to the difference can give a simple explanation of the forward displacement
between the actual and the corresponding ‘elastic’ stress. without recourse to the complexities of the complete analy-
The steady-state variation of (ys,), and also (.iz,), for the sis. With this object in view the following greatly simplified
deformation cycle at y = -a are shown in Fig. 296. It will model is offered. In the steady state the only component
be seen that direction of plastic shearing reverses. The of strain which can lead to cumulative deformation is the
residual forward flow arises from the fact that the amount of shear yzu. We shall now follow the loading history of a
forward shear outweighs that of backward shear. However, typical material element in the plastically deformed layer
the plastic work done in the reversal of the direction of and ignore the existence of all other stress and strain com-
shearing, together with the plastic work done by the other ponents. This reduces the cycle from one of complex stress
stress components, ensures that the rate of total plastic to one of simple shear. If, as the complete analysis assumes,
work greatly exceeds that which would be required to the strain cycle is symmetrical the element will be subject
produce the forward displacement in simple, uni-directional to equal and opposite peak values of shear strain yzU*.The
shear. The ratio of work to produce forward flow to the stress-strain cycle is shown in Fig. 30. OA represents
total work is given by: elastic loading at entry up to the yield stress -K, AB is
backward plastic shear, BC is elastic stress reversal to the
yield stress + k and CD is forward plastic shear. Since no
residual T~~ stress can exist, elastic unloading at exit brings
Proc Instn Mech Engrs Vol I77 No 25 1963
2
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
684 J. E. MERWIN AND K. L. JOHNSON

curved profile affects the residual stress system given in

equation (4), since the presence of a circumferential stress
(uZ),. necessarily introduces a radial stress (a,),. However,
provided that the depth of the plastic layer is small com-
pared with the radius of the cylinder, i.e. within the restric-
tions of the Hertz theory, the radial stress will be negligibly
small.
With two cylinders of identical materials, so that yield
occurs simultaneously in each, the boundary displacement
within the contact region will conform to a circular profile
as assumed in the analysis. If one cylinder yields before the
other the geometry of the interface will be distorted, but it
is unlikely that the effect on the strain field wiU be appreci-
able.
The conditions of plane deformation assumed in the
analysis are not likely to be approached near to the ends of
The element acquires a permanent strain OE. The ~ ~ ~
contacting
- y
cylinders.
~ ~
In consequence of the high hydro-
cycle taken from the computations for p , = 5 . 3 3 and y = - a IS static component of stress a fair bulk of material is required
plotted to the same scale for comparison (broken line). to provide lateral support and to prevent sideways plastic
Fig. 30. Simple shear cycle OABCDE in which a material flow.
element of yield strength k is subject to a reversal of strain The remaining limitations of the analysis are restricted
between limits *yTY* and JinaIIy left in a stress-free to the material properties. The experimental evidence that
a variety of metals, whether hardened or not, exhibit the
state (E)
phenomenon of forward flow suggests that the influence of
hardening properties is not of primary importance. The
the element to the state at point E in which it has acquired step-by-step nature of the calculation would enable a law
a forward (+ve) residual strain OE. of work hardening to be incorporated without fundamental
For comparison the -rZ,-ym cycle at y = --a taken from difficulties, but, in view of the present uncertainty in
the complete analysis is plotted to the same scale in Fig. 30. specifying a hardening rule which is physically reasonable
Although the simple shear cycle underestimates the residual in a cycle where the shear stresses reverse, the labour
forward shear strain by about one half, it exhibits the same involved would hardly be justified. To apply the results to a
qualitative features; a reversal of strain in which the for- particular metal which work hardens, the question of an
ward shear outweighs the initial backward shear. It is appropriate value for K arises. Initial yield will obviously
immediately evident that the model predicts a value of 3 be governed by the primitive yield point of the material,
for the ratio of forward %owwork to total work. It may be but both shakedown and cumulative forward flow represent
recollected that in the complete analysis the strain cycle is steady-state conditions after repeated loading when con-
continued to the point F where yZy = 0. The residual stress siderable plastic straining will have taken place. A value of
(TJ,. is subsequently relaxed (equation (14)) to return to K associated with the metal in a strain-hardened state would
state E . then be more appropriate. For example, the yield stress
When comparing the quantitative results of the computa- deduced from a Vickers hardness indentation by the method
tions with experimental observations, the limitations of due to Tabor (16)might provide a reasonable value to use.
the theory should be kept in mind*. The assumption of
elastic strains which leads to stresses which do not satisfy
equilibrium has already been emphasized. It is difficult to CONCLUSION
estimate the error thereby incurred, although Hamilton’s The investigation set out to provide a rational explanation
photoelastic experiments suggest that the actual strain of the puzzling phenomenon of cumulative plastic deforma-
distribution is not perfectly symmetrical. It is likely, how- tion which is observed when metal cylinders are in rolling
ever, that the lack of symmetry and the violation of equili- contact. The analysis which has been presented is not exact
brium in the steady state will be appreciably less than during in the sense demanded by the mathematical theory of
the first passage of the load. The probable influence of plasticity, but nevertheless follows an accepted practice for
asymmetrical strain can be inferred from the simple model the approximate solution of problems in applied mechanics
of Fig. 30. Thus an increase in the peak value of yZyon the by starting from an assumed pattern of deformation which
entry side and a corresponding decrease on the exit side is both kinematically compatible and intuitively reasonable.
would reduce the forward displacement but increase the Its success must be measured by the fact that it predicts
plastic work input. qualitatively, and to some extent quantitatively, the principal
The analysis has been carried out for a semi-infinite solid features of the phenomenon; the build-up of residualstresses
with a plane surface. For the contact of two cylinders the in a layer below the surface, cumulative displacement of
* This comparison is made in the paper by Hamilton (9). the surface in the forward direction of motion and resistance
Prof Instn Mech Engrs Yo1 177 No 25 I963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
 AN ANALYSIS OF PLASTIC DEFORMATION I N ROLLING CONTACT 685

to rolling through plastic dissipation in a well-defined sub- passages of the load to produce a forward displacement of
surface layer. the surface.
Since the analysis has been based upon the stress-strain The material is subject to a cycle of reversed shear whose
relations of an ideal, isotropic, elastic-perfectly plastic range is appreciably greater than the residual strain incre-
material, it may be concluded that the phenomenon of ment at the end of the cycle. The energy dissipated during
forward flow is predominantly due to the complex cycle of this cycle of plastic action is three to four times as great as
stress to which the material is subject in rolling contact. would be required to produce the forward displacement in
Special properties of the material such as strain hardening simple unidirectional shear. This energy is responsible for
or softening, the Bauschinger effect, or an upper yield point, the resistance to rolling; it is manifest by an offset in the
which have at various times been invoked to explain the centre of contact pressure towards the entry side.
unsymmetrical shear deformation, whilst possibly contri-
buting to the details of the behaviour, are not necessary for ACKNOWLEDGEMENTS
its occurrence and are unlikely to play a dominant role. We wish to acknowledge the help given by Mr G . M.
The deformation processes which, according to the analy- Hamilton, Dr W. Hirst and Dr A. W. Crook of the A.E.I.
sis, take place when two cylinders roll freely together under Research Laboratory, Aldermaston. The results of Hamil-
a load not greatly in excess of the elastic limit (certainly for ton’s experiments, reported in his paper (g), were placed at
p , < 6k) may now be described. our disposal as they became available and many valuable
During the first passage of the rolling contact load the discussions were held.
material deforms plastically in a layer whose depth below We are especially indebted to Mrs G . M. Birch for her
the surface roughly corresponds to that in which the Hert- work in programming the numerical computations for the
zian elastic stresses, could they be realized, would exceed Edsac 2 computer.
the yield limit of the material. During this deformation the APPENDIX
surface is compressed only slightly below its original level REFERENCES
and residual elastic compressive stresses acting parallel to (6) CROOK, A. W. ‘Simulated gear-tooth contacts: some experi-
the surface are introduced. This form of deformation ments upon their lubrication and subsurface deforma-
continues, to a decreasing extent, with the next four or five tions’, Proc. Znstn mech. Engrs, Lond. 1957 171, 187.
(7) WELSH,N. C. ‘Structural changes in rubbed steeI surfaces’,
cycles of load, after which a steady state is reached when Instn mech. Engrs, Proc. Conf. Lubr. Wear 1957, 701.
there is no further change in residual stress or surface (8) JOHNSON, I<. L. Discussion on Crook (I), 204.
depression. If the load is within the shakedown limit (9) HAMILTON, G. M. ‘Plastic flow in rollers loaded above the
<
( p o 4k) no further plastic flow takes place. Subsequent yield point’, Proc. Znstn mech. Engrs, L m d . 1963 177 (No.
25).
deformation is then entirely elastic.
(10)BELIAEV,N. M. Engineering construction and structural
At higher loads a steady pattern of plastic action is re- mechanics, 1924,27-108 (published in Leningrad).
peated with each passage of the load. This is restricted to (11) RADZIMOVSKY, E. I. Stress distribution and strength condition
a layer which is thinner and centred closer to the surface of two rolling cylinders pressed together, 1953 University of
than initially. The significant component of plastic strain is Illinois, Engineering Experimental Station, Bulletin
now one of shear parallel to the surface. On entering the No. 408.
(12)JOHNSON, K. L. ‘A shakedown limit in rolling contact’,
plastic zone each material element is sheared backwards in Proc. Fourth U.S. Nut. Congr. appl. Mech., Berkeley,
the sense which the elastic shear stress at entry suggests. California, June 1962.
However, before the axis of symmetry is reached the direc- (13) SYMONDS, P. S. ‘Shakedown in continuous rnedia’,J. appl.
tion of the plastic shear straining reverses. Throughout the Mech., Trans. Amer. SOC. mech. Evgrs 1951 73, 85.
exit side the material is subject to a severe shearing in the (14)PRAGER, W. and HODGE, I?. G. Theory of perfectly plastic
solids 1951 (John Wiley and Sons, New York).
forward direction, which more than eliminates the back- (15) GILL, S. ‘A process for the step-by-step integration of
ward shear obtained at entry and gives rise to the residual differential equations in an automatic digital computing
plastic shear strain which accumulates with repeated machine’, Proc. Camb. Phil. SOC.1951 47, 96.
(16) TABOR, D. Hardness of metals 1951 (Oxford University Press).

Proc Instn Mech Engrs Val 177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
686

Communications
Professor G. Chabert (Paris)-These papers indicate ture leads to unfortunate consequences with phenomena
that, in the case of gear wheels, rolling should cause plastic such as creep.
flow of the underlayer towards the base of the teeth of the Thus it would appear that slip phenomena are the only
driving element, and towards the top of the teeth of the important factors in surface creep phenomena when gear
receiving element. However, according to Crook (6),there wheels are used in a through-hardened state or without
is a superimposed slip effect which causes creep, tending heat treatment. I consider, therefore, that the study could
to move the underlayer away from the pitch circle of the with great advantage be extended to include slip and,
driving element and to concentrate it towards the pitch moreover, case-hardened test discs.
circle of the receiving element. This effect is, of course,
well known and has been observed for many years with
gear wheels. Mr G. J. Moyar (Schenectady, New York)-In their
I n any event, the two effects are cumulative at the analysis of a complex problem, Dr Merwin and Dr
bottom of the tooth spaces of the two elements, and Johnson have demonstrated considerable ingenuity in the
opposed at the top of their teeth. selection of pertinent approximations which still maintain
This observation, in fact, led Crook to state: ‘The type many essential features of the actual phenomenon.
of deformation found with the rolling discs is of interest With regard to the method of determining stresses in
since it suggests a new explanation of the greater propen- the elastic-perfectly plastic material, have the authors
sity to pitting of discs with the lower peripheral speed and considered the possible simplification offered by a piece-
of the dedenda of gear teeth.’ wise linear yield surface and associated flow rule ? With
However, the mathematical analysis by Dr Merwin and such theories (17) it is possible to express the stress com-
Dr Johnson (second paper) produces figures showing that ponents directly in terms of the strains and thus eliminate
this argument cannot be maintained. The phenomenon of the step-by-step machine computation. It would be
continuous creep, demonstrated for the first time with necessary to follow the stress point from one regime (face
rolling discs by Mr Hamilton’s firm and discussed in or intersections of the yield polyhedron) to another during
both papers, should be observable only with such heavy the strain cycle. These theories, for example the Tresca
loading as to be met exclusively in industrial applications yield and associated flow rule, are physically acceptable
where case hardening is necessary. according to Drucker’s stability postulate. I n fact, con-
T o show this, let R be the ultimate tensile stress and u sidering the complex loading cycle, there is little experi-
the yield stress (approximately 0.85R for low-carbon mental evidence to justify the selection of the Prandtl-.
alloyed steels designed for heavy-loaded applications). Reuss theory over such theories.
Continuous creep takes place when I should like to amplify some of the discussion of
material hardening behaviour. Although many metals are
not closely described by perfectly plastic behaviour, they
may attain a saturation hardness, as the authors suggest,
(3*4)2 after which the analysis is appropriate. Since there is
Then - 3.85 some indication that steady-state cyclic strain accumulation
is not possible with many hardening rules, such a sug-
Now let us take the load criterion introduced in S.E.I.E. gestion has practical merit. To illustrate the behaviour
Bulletin No. 35, p. 8: when an idealized hardening, as opposed to a hardened
material, is treated, consider the result of isotropic (18)
and kinematic (19)hardening on the basis of the simple
model proposed by the authors. Three cycles for each
This value corresponds to heavy loading applications case are illustrated in Fig. 31. The isotropic model
such as heavy commercial vehicles or tractors. Case rapidly ‘shakes down’ to purely elastic behaviour. I n the
hardening is of general practice for such gears. case of kinematic hardening a stable cycle of energy
It may well be that the presence of a superficial layer dissipation is established although forward strain accu-
different from the core as regards composition and struc- mulation ceases.
Prod Instn Me& Engrs Vol177 No 25 I963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
 PLASTIC FLOW IN ROLLING CONTACT 687

(18) HILL,R. The mathematical theory ofpZmticity 1950 (Claren-

don Press, Oxford).
(19)PRAGER, W. ‘A new method of analysing stress and strain
in work hardening plastic solids’, J . appl. Mech. (Trans.
Amer. SOL.mech. Engrs) 1956 23, 493.
( 2 0 ) MOYAR, G. J. and SINCLAIR, G. M. ‘Cumulative plastic
deformation in rolling contact’, 3. Basic Engng, Tram.
Amer. SOC.mech. Engrs 1963 85 (series D, no. l), 105
(discussion 114).
(21) MOYAR, G. M. ‘Cyclic strain accumula-
G. J. and SINCLAIR,
tion under complex multiaxial loading’, Joint Int. Conf.
on Creep (A.S.M.E.-A.S.T.M.-1.Mech.E.) 1 963, paper
no. 35.
Mr E. Ollerton, B.Sc. (Associate Member)-Mr
Hamilton is to be congratulated on the clarity of his
photoelastic fringe patterns. The production of clear
fringe patterns in this type of problem can be very dif-
ficult. The disc can warp in its own plane under the radial
load, giving a contact area, which is trapezoidal instead
of rectangular. If this occurs the fringe pattern becomes
a Linear isotropic hardening. b Kinematic hardening. indistinct because the stresses vary through the disc
Fig. 31. Simple shear cycles for strain-hardening models thickness. Mr Hamilton evidently avoided this difficulty.
In the calculation of a pressure distribution which
would cause the observed tilting of the fringe pattern, it has
Despite the shortcomings of the current hardening been assumed that the principal directions of stress in
idealizations it does not necessarily follow that the in- the contact surface are radial and tangential; in other
herent hardening properties are not significant. That words, that no shear stresses were transmitted across the
steady-state strain accumulation is possible, even without contact surface. The validity of this assumption can be
macroscopic stress or strain gradients, has been discussed tested very easily by viewing the disc in plane-polarized
previously (20) and explored experimentally (21). Yet this light, with the polarizer and analyser aligned in the radial
qualification does not detract from the authors’ analysis and tangential directions. The resulting isoclinic lines
of the forward shear phenomenon. Their theory is direct will connect all points at which the principal directions of
and convincing without appeal to mysterious material stress are radial and tangential, and if an isoclinic extends
properties. across the full width of the contact surface, the assumption
REFERENCES
is justified.
(17) GOODIER,J. N. and HODGE, P. G. Elasticity and plasticity In the absence of such a justification, the inferred pres-
1958,75 (John Wiley and Sons, New York). sure distribution should be treated with some caution,

Fig. 32. Fringe pattern obtained by loading an Araldite disc against a mild-steel plane and applying a tangential force
Proc Instn Mech Engrs Vo1177 No 25 1963
Downloaded from pme.sagepub.com at UQ Library on June 14, 2014
688 COMMUNICATIONS

because the tilting of the fringe pattern could well be The third observation was that with repeated traversals
due to the combined effect of pressure redistribution and of the same track the track width generally increased and
surface shear stresses. The application of a driving torque the rolling friction diminished until after about 100 tra-
to the disc, tending to roll it to the right in Fig. 19 would versals a steady state seemed to have been reached. At
itself cause the fringe pattern to tilt in the observed this stage it appeared as if the contact were determined by
direction. Could Mr Hamilton say whether the loading elastic deformation of the ball and the groove. The elliptic
rig was capable of transmitting such a torque ? region of contact agreed well with that given by the
Fig. 32 shows a fringe pattern obtained by myself by Hertzian relation. Again the rolling friction was little
loading an Araldite disc against a mild-steel plane and affected by lubricants. It was natural, therefore, to look
applying a tangential force. The disc did not roll or slide upon this as an extension of the plastic case and to con-
and the loading conditions did not produce any plastic sider the rolling friction as arising essentially from elastic
deformation in the steel. The applied tangential force was deformation and the hysteresis losses associated with it.
equivalent to a torque tending to roll the Araldite disc to This explained all the major observed results. The main
the left, so the fringe pattern should be inverted for defect, however, was that in some cases the hysteresis
comparison with Fig. 19. The ratio between tangential losses were enormous : indeed, with copper at higher loads
and radial force was about 0.1, and the coefficient of one would need to assume that practically the whole of
friction about 0-3. At almost limiting tangential forces the the elastic energy of deformation was lost during rolling.
fringe patterns agreed closely with those predicted by I n order to study the elastic hysteresis mechanism in
Poritsky (22). greater detail we left metals and concentrated our study
on rubber. It soon became clear that the hysteresis
REFERENCE
explanation for rubber-like materials was descriptively
(22) H. ‘Stresses and deflections of cylindrical bodies
PORITSKY,
valid and finally that it was quantitatively valid. Here,
in contact with application to contact of gears and of
locomotive wheels’, J . appl. Mech., Trans. Amer. SOC. however, a new point emerged. Every element passing
mech. Engrs 1950 72, 191. under the rolling member is subjected to a reversal of the
shear direction and because of this the hysteresis losses
Mr N. A. Scarlett (Chester)-Having read the two are about three times higher than simpler theories would
papers I would congratulate the authors on the original suggest. These results also led to the view that the rela-
and interesting work described. tively high rolling frictions previously observed with
My only comment is that similar corrugations to those metals in the near-equilibrium condition were probably
described have been experienced on hypoid gears run in due to reversed plastic shear of the elements under the
the laboratory of my firm’s research centre under slow- rolling member. However, no quantitative calculations of
speed heavily loaded conditions. this were attempted.
I n the paper by Dr Merwin and Dr Johnson we have,
Dr D. Tabor (Cambridge)-I should like to express for the first time, such a quantitative estimate carried
my great pleasure on reading these two papers on rolling- out with great perception and ingenuity. It is interesting to
contact phenomena in the plastic range. They make a see how far their results may indeed explain the relatively
very real and much-needed contribution to our under- high frictions we observed. If we approximate our narrow
standing of the processes involved when one body rolls ellipse of contact to a parallel band of contact we may use
on another. I hope Mr Hamilton will not think me un- their results of Fig. 28. Assuming at this stage that po/K
appreciative of his experimental skill if I devote my remarks was about 5, the observed friction for copper is of the
to the paper by Dr Merwin and Dr Johnson. They have same order as that shown in Fig. 28.
tackled a difficult and puzzling experimental observation This raises two interesting additional points, As I
and have provided an explanation that is lucid and understand it the energy dissipated by plastic deformation
convincing. Further they have kept very clearly before is provided by strain energy in the elastic hinterland. This
the reader the nature and the limitations of the assumptions energy loss is equivalent to throwing the centre of pressure
they have made. ahead of the geometric centre of contact. There must be a
When Dr Eldredge and I began our study of rolling limit to this. Clearly the centre of pressure cannot be
friction about 10 years ago we studied in particular the beyond the outer edge of the region of contact and even
rolling of a hard steel sphere over the flat surface of a this is physically unacceptable. A more reasonable limit
softer metal. There were three clear-cut experimental would be when the centre of pressure is about a12 ahead
observations. First the initial traversal generally produced of the geometric centre of contact. This would give a
a permanent groove in the lower surface. Secondly the resistive couple equal to Pa/2. An analysis we made in our
rolling friction for this traversal was not reduced by earlier paper suggests a smaller upper limit of P3116a.
lubricants. These led us to the view that the major part In terms of Fig. 28 the rolling resistance parameter
of the rolling resistance was not due to interfacial slip but MGjRapo2 in this case has a value of about 0.16. This
to plastic displacement of metal ahead of the ball. There implies that one could not easily dissipate more energy
was, indeed, good quantitative agreement between the than this without so greatly modifying the elastic strains
observed friction and that calculated on this assumption. that the rest of the analysis would be of dubious validity.
Proc Instn Mech Engrs Vol177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014

 PLASTIC FLOW IN ROLLING CONTACT 689

otherwise, been achieved. As they point out, this contri-

I ill
0.16 --_ -----
_ _APPROXIMATE UPPER L I M I T
-- butes a fair part to the total friction. Even more important
is the part it must play in the fatigue of the surface. Here
0.14
I am thinking not so much of fatigue in rolling friction
itself, but rather of fatigue and wear of sliding surfaces in
the presence of effective lubrication. For even if the
interfacial adhesion is trivial the continuous passage of one
set of asperities over another will produce stress patterns
resembling those which occur in rolling friction. It is
possible to regard the normal loading-unloading cycle of
the surface asperities as the cause of fatigue and wear and
this has been done in an interesting way by Rozeanu (23).
But if surface shearing occurs in the way described by
Dr Merwin and Dr Johnson it could be a far more
effective cause.
Finally I should like to add two questions. Do the
authors consider it possible to carry out a similar treat-
ment for a hard sphere rolling over a flat surface ? Do they
think a shakedown condition is theoretically possible ?
Our own early measurements suggested a near-equili-
brium state after about 100 traversals but later work
LOAD F A C T O R , po/k
showed a slow but continuous diminution in friction and
Fig. 33. Rolling resistance parameter growth of track width for another quarter of a million
traversals.
Fortunately this is above the calculated values given by Secondly, the results of Figs 27 and 28 suggest that if
Dr Merwin and Dr Johnson but not very much (Fig. 33). rolling involves initial plastic deformation there is a
Secondly, of course, the analysis does not say anything great advantage in using a material which has a large
about energy losses below the elastic or shakedown limit. work-hardening index so that in approaching the shake-
In this range the rolling friction is due to hysteresis losses down condition the pressure becomes less than 4k where
and as a matter of interest I have drawn the loss to be K is the augmented critical shear stress for the work-
expected for an effective hysteresis loss of 2 per cent. It is hardened material. Do the authors know of any practical
seen how small this is compared with the value obtained by experience (say from ball-bearing studies) which would
Dr Merwin and Dr Johnson in the plastic range. support this ?
I should also like to congratulate the authors on the
elegant explanation they give for cumulative plastic shear REFERENCE
in the surface layers even though elastic equilibrium has, I,. Wear (to be published).
(23) ROZEANU,

Authors’ Replies
Mr G . M. Hamilton-Professor Chabert has raised the to the type described in the paper, which shows that such
important question of the loads at which disc machine loads are realized in practice. From this point of view it
tests should be conducted. This is a contentious matter would be interesting to know whether the corrugations on
but it is well known that to reproduce failures such as the hypoid gear mentioned by Mr Scarlett were accom-
pitting and scuffing the loads applied between the discs panied by subsurface deformation.
have to be higher than the nominal values for the gears Many of the questions raised by Mr Ollerton can be
which are being simulated. The normal interpretation of settled by reference to Fig. 34 showing a sketch of the
this fact is that the load on the gear in the region where loading rig used for the photoelastic tests which was de-
the failure occurred is considerably greater than average, signed especially to reduce the shear stresses transmitted
generally presumed to be due to misalignment or in- across the contact. The specimen was driven under load
accurate manufacture. I n these circumstances the entire across the surface of the glass disc which in turn was free
load can be supported on as little as 10 per cent of the to roll across the surface of a hard steel plate, so that its
available face width. Benson (24) has reported an example rolling resistance was negligible. With this arrangement
of a gear failure with subsurface deformation, very similar such rolling resistance as remains causes the fringes to tilt
Proe Insm Mech Engrs Vol177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014

690 AUTHORS’ REPLIES

(n
A+N
G ;,SE
;IG
LO
,;+, LOADING ARM
shown predicts that forward flow (but not plastic deforma-
tion) would eventually cease, in order to reach this limiting
\ state, it would be necessary for the metal to harden to at

\
least twice its original yield stress. It might be mentioned
DRIVE that two mild steel discs which were tested showed a
different behaviour from copper and duralumin discs.
Instead of the forward flow being directly proportional to
the number of stress cycles (see Fig. lo), it showed a
marked decrease after a large number of cycles. Presumably

1 SPECIMEN
this effect is due to the hardening characteristics of mild
steel.
The authors are grateful to Dr Tabor for his communica-
tion relating his pioneering work on rolling friction to
the present theory. The agreement with Dr Tabor’s
rolling resistance measurements for copper is encouraging
since the comparison with Mr Hamilton’s measurements
in Fig. 17 is rather poor. Very reasonably Dr Tabor uses
the asymmetry of the pressure distribution to set an upper
limit for the validity of the theory at a load corresponding
STEEL SLATE to po/k * 6 . Further support for this figure is provided by
Fig. 34. Photoelastic loading rig the fact that at this load a subsurface residual stress equal
to the yield stress would be induced, so that at higher
in the opposite direction to that shown in Fig. 19 and in loads the pattern of deformation would be likely to show
fact stress patterns very similar to the one shown by Mr a marked change.
Ollerton were obtained by restraining the glass disc in its T o perform a similar analysis for the three-dimensional
passage across the steel plate. problem of a sphere rolling over a flat surface would seem
to be a formidable undertaking at the present time. The
Dr J. E. Merwin and Dr K. L. Johnson-We would residual stress pattern introduced by the formation of a
certainly agree with Professor Chabert that frictional groove will be complex and difficult to analyse. However,
tractions due to sliding play an important part in the pro- the problem is under consideration and some preliminary
cess of plastic deformation of rolling surfaces, The present observations have been made in reference (25).
work has been extended to include sliding friction and is The idea of using a material which is initially soft,
reported by Johnson and Jefferis (25). In that investiga- deforms to produce a conforming groove, and thereby
tion it was shown that the value of p o at the shakedown hardens to prevent further deformation, is an interesting
limit is decreased from 4k by the action of a friction force, one. But it does not seem to fit in with ball-bearing experi-
falling to 3k and 1.8k for coefficients of sliding friction of ence. Here the fatigue life improves with increasing initial
0.25 and 0.5 respectively. hardness. Also the reversed shear beneath the surface
Dr Moyar’s suggestion of using the Tresca criterion has been observed to soften the material in that region
and flow rule has been found useful in investigating surface (see Bush, Grube and Robinson (26)).
yielding produced by friction forces. Here the expressions
for the elastic strains are simple. Beneath the surface REFERENCES
numerical analysis seems unavoidable and the Prandtl- L. E.
(24) BENSON, Contribution to discussion Proc. Instn mech.
Reuss equations, being without discontinuities, are par- Engrs, Lond. 1957 171,196.
(25) JOHNSON, K. L. and JEFFERIS,J. A. ‘Plastic flow and residual
ticularly convenient for numerical integration. We are stresses in rolling and sliding contact’, Proc. Symp.
grateful for Dr Moyar’s neat extension of our simple Fatigue in Rolling Contact Instn mech. Engrs, Lond.
model of ‘forward flow’ to examine the effect of strain 1963, paper 5.
hardening. The ‘kinematic hardening’ model represents, (26) BUSH,B. J., GRUBE, W. L. and ROBINSON, G. H. ‘Micro-
in a very idealized way, the Bauschinger effect displayed structural and residual stress changes in hardened steel
due to rolling contact’, Proc. General Motors Symp.
by most metals undergoing reversed plastic strain. Rolling Contact Phenomena Detroit 1960 (Elsevier,
Although Dr Moyar is correct is stating that the model as Amsterdam).

Prae Instn Mech Engrs Vol177 No 25 1963

Downloaded from pme.sagepub.com at UQ Library on June 14, 2014