Professional Documents
Culture Documents
28 (1974) I-13 1
Q Elsevier Sequoia S.A., Lausanne - Printed in The Netherlands
SUMMARY
INTRODUCTION
0
0
CONTACT PRESSURE
Fig. 1. Ideal static friction curve. b vs. contact pressure for metal surfaces (qualitative)
FRICTION OF MECHANICAL AND STRUCTURAL JOINTS 3
formulated and may consist of the four regions shown qualitatively in Fig. 1.
This portrays the friction properties of metal surfaces which are reasonably
smooth, e.g. machined surfaces, and greaselfree.
Referring to Fig. 1, there are two regions in which M is constant as
predicted by theory and commonly tlbserved in practice. In one of these, region
I, both surfaces are covered by contaminant films, e.g. oxides, and the friction
observed is mainly that between these films. Contaminant films are less reactive
than clean metals hence friction forces are small in this region.
As contact pressure is increased surface films are progressively broken
down, some metal-to-metal contact occurs, and friction forces rise sharply (region
II). This effect has been observed by Whitehead’ in sliding friction of copper
oxide and also by Ernst and Merchant3 in iron although there it was not closely
investigated.
In region III substantial metal-to-metal contact takes place and surface
films have no effect. k is again independent of pressure but is much larger
than in region I.
At extreme contact pressures involving extensive plastic surface deformation,
region IV, the coefficient of friction decreases with increasing pressure and
eventually becomes zero when the material fails in compression.
Contact pressure as used above is the apparent contact pressure, defined
as the ratio of normal force to nominal contact area. This pressure is quite
different from the true contact pressure existing at real areas of contact which
is, according to the adhesion theory, equal to the flow stress of the softer
material.
For comparison with the above theory Fig. 2 shows friction data taken from
the literature. Static friction coefficients for clean dry steel surfaces are plotted
against nominal contact pressure. Many of these results were obtained with
“point-contact” devices in which the contacting surfaces are spheres sliding on a
plane or which use crossed cylinders. In this type of apparatus the normal pressure
varies over the area of contact, so for comparison with other data an average
pressure has been calculated from Hertzian stress theory4.
The graph shows low friction coefficients, approximately 0.3, at contact
pressures less than 50 p.s.i. At 10,000 p.s.i. and above several workers have found
that ,r+ is constant over a small range of pressure. However there is little
similarity between the respective numerical values, these ranging from 0.4 to
1.0. A steady decrease in b with contact pressure has also been observed.
Over an entire decade of pressure between 70 and 700 p.s.i. no data were found;
a wide gap in knowledge of this important engineering parameter. All of the
features of Fig. 1 are evident in Fig. 2 although, in places, some of the evidence
is fragmental or displays a marked scatter.
Much of the variation in the results is due to the different surface prepara-
tion techniques used by different people. Schmidt and Weiter’ used flat ground
surfaces washed in carbon tetrachloride. Surface oxides still present after this
cleaning account for the low values of ~1 recorded. Gaylord and Shuh used
flat surfaces polished on emery paper and washed in alcohol, producing similar
results. Paslay and Plunkett7 used cylindrical shrink-fit joints to obtain friction
data. The surface condition of these joints is unknown but it is likely that a
considerable amount of oxidation occurred during assembly. Campbell’, Hardy’,
Tomlinson lo, Bowden and Tabor”, and Wiid and Beezhold12 all used sphere-on-
plane point contact devices. Spurri3 and Ernst and Merchant3 used crossed
cylinders to obtain point contact. These devices all produced very large contact
pressures; different results again being due to different material properties and
cleaning procedures. The effect of surface treatment on friction between structural
surfaces has been reported by Vasarhelyi rt ~1.‘~. They give values of ~4 in a
bolted structural joint ranging from 0.065 for surfaces painted with red lead to 0.51
for rusted surfaces cleaned with a wire brush.
Wide variation in friction data is also evident in engineering handbooks,
which differ over the value of ,LLfor clean dry steel surfaces. Marks15 suggests
0.39 for grease-free surfaces in air; Eshbachi6 gives 0.15, while Machinery’s
Handbook” quotes 0.8. The latter does however mention that this value may
change considerably with changes in test conditions.
It is clear that the available data are inadequate for predicting, with any
accuracy, the static friction between a pair of steel surfaces. It also appears that
the transition between regions I and III of Fig. 1, in which K changes, may
take place at pressures within the range applicable to a variety of mechanical
joints and bearing pads. This could further complicate the problem of predicting
friction behaviour in real situations.
SCOPE OF INVESTIGATION
important in the study of frictional energy loss during vibration. In this work
the static friction coefficient is determined under dynamic load conditions since
previous work 6 ’ ~3. ’ 9 has suggested that this can be different from the result obtained
under purely static loading. It is important to distinguish between the static
coefficient of friction, the static coefficient under dynamic load, and, thirdly, the
kinetic coefficient. The latter is the one that determines the energy loss in surfaces
in relative motion, but it is the second that controls the breakaway point or
incipient movement under increasing load. In this sense the static coefficient is
also related to the energy dissipation characteristics of an interface.
EXPERIMENTAL EQUIPMENT
SECTION X-X
out the apparatus loads are transferred by point contact wherever possible so that
unwanted forces, e.g. due to misalignment, are not transmitted to the friction
surfaces. It has been shown ‘O that extra forces or moments in the slip plane
have a significant effect on the friction and adhesion of metals. A sectioned
view of the setup is shown in Fig. 3.
One specimen (A) is rigidly fixed to a vertical mounting surface. The
other (B) is attached to a holder (C) supported by a self-aligning ball bearing
(D). Nominal contact area is 1.29 sq. in. with a mean diameter of 3.25 in.
Normal loads are applied to the specimen holder by a pneumatic cylinder
piston (E) through a ball-ended rod (F). The load is measured by strain gauges
attached to this rod. Static tangential forces are applied by a similar piston
(G) and rod (H), to disc (I) which is radially supported by roller bearings (J).
Torque is transmitted to the specimen holder (D) from disc (I) via the ball
joints (K). Strain gauges on rod (H) measure the tangential force. During the
static tests both force signals were displayed on a pen chart recorder.
With this apparatus contact pressures between 80 and 2000 p.s.i. can be
produced. Narrower specimens (0.20 sq. in.) allowed this to be extended to
10,000 p.s.i. In each test the moving specimen rotated through an angle correspond-
ing to a sliding distance of 0.093 in.
In the dynamic tests a sinusoidal tangential force is applied by an electro-
dynamic shaker (L) through a rod with spherical end bearings (M). A piezo-
electric force transducer is used to measure the dynamic load and an accelerometer
mounted on the specimen holder is used to determine the onset of slip. Force
and acceleration signals were recorded on an ultra-violet recorder. The normal
load is applied as before and both static and dynamic tangential forces can be
applied simultaneously.
TEST PROCEDURE
In some tests with small normal loads the slip load was dependent on the
rate at which the tangential force was applied. When this occurred the loading
rate was reduced until it ceased to have any effect on the point of slip. A similar
procedure was adopted by Campbell’.
1” n
NO. OF TESTS
Fig. 4. H OS. no. of tests, static loading. Surfaces ground, contact pressure 250 p.s.i.
8 H. NOLLE. R. S. H. RICHARDSON
although only in spherical sliders. Following this bedding-in period, the region
in which k increases slowly with sliding distance appears to correspond to the
process generally referred to as mild wear ” . This process, which commonly occurs
between oxidised metals, produces little subsurface damage. Fine black wear debris,
apparently iron oxide, often observed during the tests is an expected consequence
of mild wear of steel surfaces. The final section of Fig. 4, in which K increases
very quickly, corresponds to the severe wear process. Large scale surface damage
occurred and the specimens became so badly scored that mechanical interference
outweighed surface friction. The results then cease to be useful.
The variation of ,u, with sliding distance makes it difficult to associate a
Fig. 5. 1~ US. no. of tests, static loading. Rough surfaces (mill scale); contact pressure 300 pxi.
01 / , I
IO 50 100 5cn to*0 5000 1, 00
CONl;PCT PRESSURE -pm
single numerical value with a particular set of conditions. The results given below
were obtained after the surfaces had been run-in but before significant wear had
taken place.
The results for rough surfaces, Fig. 5, show some of the same trends as the
machined surfaces. K is initially small and increases slowly but then remains almost
constant, although considerable scatter is evident. The test was concluded before
the severe wear stage was reached. Figure 6 shows one of the rough surfaces
after testing. Small contact spots (white areas) are evident around most of the
circumference of the specimen; there are no heavily scored regions. The mill
scale surface is worn away as a fine powder and seems to act as a lubricant
to prevent subsurface damage to the base metal.
Results showing M plotted against contact pressure are shown in Fig. 7.
All of the surfaces studied yielded constant values of ,~k over a wide range of
contact pressure. For turned and ground surfaces the numerical values of K are
almost identical. It appears that for this type of machined surface friction is
insensitive to surface roughness. Ernst and Merchant3 recorded similar results for
a crossed cylinders device but Schmidt and Weite? found that surface roughness
did have some influence on friction in more finely finished surfaces.
The results show that the coefficient of static friction is constant for contact
pressures between 100 and 10,000 p.s.i. The value of h,, 0.19 for machined
surfaces, appears to correspond to region I of Fig. 1; i.e. the contaminant film
region. This is to be expected since the cleaning methods used were not intended
to produce chemically clean surfaces. For the material and surface preparation
used the transition to metal-to-metal contact, and large values of k, did not occur
below 10,000 p.s.i. This places region III outside the range of pressures used in
mechanical joints. It appears that friction data used in joint design should be ob-
tained from the contaminant film regime /rather than from contact between
chemically clean metals.
Many of the values of h for steel given in the literature, Fig. 2, are larger
than those found above. This is ascribed to the widespread use of point contact
10 H. NOLLE, R. S. H. RICHARDSON
devices for friction experiments. These produce very high localised contact pressures,
even under small loads, and may lead to misleading results if the data obtained
are applied to flat surfaces. The fact that small localised areas under very high
contact pressures lead to metal-to-metal contact and to a “ploughing” effect when
surfaces are in motion, indicates that the use of calculated mean pressures for
point contact devices cannot be justified. The importance of maintaining geometric
similarity between friction experiments and applications has been stressed previ-
0uslyS.
The results for rough surfaces show fairly large values of k, caused by
mechanical interlocking and ploughing between the surfaces. However Amontons’
Law, that b is constant, is still obeyed over the range of pressures used in
the tests. The value of k for rusted surfaces, 0.49, compares favourably with that
of 0.48 given by Vasarhelyi et a/.14.
TABLE I
5 0.10
10 O.b4
20 0.025
50 0.005
.6 -
.5 -
.L - %
lb
.3 - d
.2- k0
0 0-lP--4 Q
., -
Fig. 8. Dynamic load tests; k us. no. of slip cycles. Turned surfaces, frequency 20 Hz, contact
pressure 100 p.s.i.
Ps
.2 -
.I5 -
-0 0
0
.I
.05
t
0 I / I
2 5 10 20 so
FREQUENCY -Hz
Fig. 9. Dynamic load tests; K vs. frequency. Turned surfaces, contact pressure 100 p.s.i.
.‘
.3 - a-,-,C-3-o
lo
&
.2 - ,,._.-9--
.I -
0 / I
10 100 1000 IDWO 100000
NO. OF CYCLES
Fig. 10. Dynamic load tests; k us. no. of slip cycles. Rough, rusted, surfaces, frequency 20 Hz,
contact pressure 100 p.s.i.
12 H. NOLLE, R. S. H. RICHARDSON
CONCLUSIONS
(1) Available experimental data for static friction of dry steel are inadequate
and often misleading.
(2) Experimental equipment has been developed for measuring static friction
of metals over a wide range of contact pressures with either static or dynamic
tangential loading.
(3) For machined steel surfaces Amontons’ Law, that the coefficient of
static friction is constant, was obeyed for contact pressures between 100 and
10,000 p.s.i.
(4) Static friction coefficients obtained under sinusoidal tangential loads were
numerically less than those found for static conditions.
(5) Friction data for use in vibratory environments should be obtained from
tests conducted with the same type of dynamic loads as expected in practice.
REFERENCES
1 F. P. Bowden and D. Tabor, The Friction and Lubrication of Solids, Part II, Clarendon Press,
Oxford, 1964, pp. 98-104.
FRICTION OF MECHANICAL AND STRUCTURAL JOINTS 13
2 J. R. Whitehead, Surface deformation and friction of metals at light loads, Proc. Roy. Sot.
(London), A201 (1950) 1099124.
3 H. Ernst and M. E. Merchant, Surface friction of clean metals-A basic factor in the metal
cutting process, Proc. Spec. Conf. on Friction and Surface Finish, M.I.T., 1940, pp. 76-101.
4 S. Timoshenko and J. N. Goodier, Theory of Elasticity, McGraw-Hill, New York, 1951, p. 372.
5 A. 0. Schmidt and E. J. Weiter, Coefficients of flat-surface friction, Mech. Eng., 79 (12) (1957)
1130-l 136.
6 E. W. Gaylord and H. Shu, Coefficients of static friction under statically and dynamically applied
loads, Wear, 4 (1961) 401412.
7 P. R. Paslay and R. Plunkett, Design of shrink-fits, Trans. ASME., 75 (1953) 119991202.
8 W. E. Campbell, Studies in boundary lubrication, Trans. ASME, 61 (1939) 6333641.
9 W. B. Hardy, Collected Papers, Cambridge Univ. Press, Cambridge, 1936, p. 648.
10 G. A. Tomlinson, A molecular theory of friction, Phil. Mag., 7 (7) (1929) 905-939.
11 F. P. Bowden and D. Tabor, The Friction and Lubrication of Solids, Part II, Clarendon Press,
Oxford, 1964, p. 323.
12 D. H. Wiid and W. F. Beezhold, The influence of wear on the coefficient of static friction in the
case of hemispherical sliders, Wear, 6 (1963) 383-390.
13 R. T. Spurr, Frictional behaviour of metals on a crossed cylinder apparatus, Wear, 5 (1962) 55-59.
14 D. D. Vasarhelyi, S. Y. Beano, R. B. Madison,’ Zung-An Lu and U. C. Vasishth, Effects of
fabrication techniques on bolted joints, A.S.C.E. J. St. Div., 85 (3) (1959) 71-116.
15 L. S. Marks, Standard Handbook for Mechanical Engineers, McGraw-Hill, New York, 1967, p. 3.35.
16 0. W. Eshbach, Handbook of Engineering Fundamentals, Wiley, New York, 1961, p. 4.52.
17 MachinerJq.7 Handbook, 18th edn., The Machinery Publishing Co. Ltd., New York, 1968, p. 522.
18 A. Seireg and E. J. Weiter, Frictional interface behaviour under dynamic excitation, Wear, 6 (1963)
66-77.
19 A. Seireg and E. J. Weiter, Behaviour of frictional Hertzian contacts under impulsive loading,
Wear, 8 (1965) 208-219.
20 W. P. Mason, Adhesion between metals and its effect on sliding contacts, ASLE Trans., 2 (1959)
4749.
21 F. P. Bowden and D. Tabor, The Friction and Lubrication of Solids, Part II, Clarendon Press,
Oxford, 1964, pp. 350-364.