Vous êtes sur la page 1sur 17

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/245369380

A Dynamic Friction Model for Unlubricated


Rough Planar Surfaces

Article in Journal of Tribology October 2003


DOI: 10.1115/1.1573229

CITATIONS READS

9 60

2 authors:

Xi Shi Andreas A. Polycarpou


Shanghai Jiao Tong University Texas A&M University
19 PUBLICATIONS 201 CITATIONS 246 PUBLICATIONS 2,999 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Andreas A. Polycarpou on 29 November 2014.

The user has requested enhancement of the downloaded file.


A Dynamic Friction Model for
Unlubricated Rough Planar
Surfaces
Xi Shi Modeling dynamic or kinetic friction for realistic engineering surfaces continues to be a
challenge, partly due to the coupling between system dynamics and interfacial forces. In
Andreas A. Polycarpou this paper, a dynamic friction coefficient model for realistic rough surfaces under external
normal vibrations is developed. From the system dynamic model, the instantaneous time
Department of Mechanical varying normal separation at the interface is obtained under normal harmonic excitation.
and Industrial Engineering, Subsequently, the instantaneous dynamic contact and tangential (friction) forces are cal-
University of Illinois culated as a function of the instantaneous normal separation. The dynamic friction coef-
at Urbana-Champaign, ficient defined as the ratio of the time varying friction to the interfacial normal forces that
Urbana, IL 61801 explicitly includes interfacial damping, is also calculated. The results show that a mean
increase in the instantaneous normal separation may or may not lead to a decrease of the
mean friction force and the mean friction coefficient, which is supported by published
data. For unlubricated elastic sliding contact conditions considered in this paper, the
effect of damping on the dynamic friction coefficient is found to be negligible, whereas
loss of contact causes significant apparent dynamic friction force and dynamic friction
coefficient reductions. Several different interpretations of the time varying dynamic fric-
tion coefficient are presented and the implications of using a simple constant value to
represent the time varying dynamic friction coefficient are discussed.
DOI: 10.1115/1.1573229

1 Introduction mass block with surface roughness in contact with a rough sur-
face. F is the external force; P is the contact force due to the
Many important mechanical devices and machines include slid-
asperity deformation and F s is the adhesion or intermolecular
ing contacts or interfaces, which must operate under transient,
force. Notice that the direction of P and F s are opposite as the
oscillatory or other dynamic conditions. Despite great advances in
adhesion force is an attractive force. The mean normal separation
modeling and measuring friction since Leonardo da Vinci 1,2,
between the two surfaces d is equal to the static equilibrium sepa-
Amontons 3 and Coulomb 4, the majority of the research work
ration d o plus the mass displacement y measured from d o , Static
that is reported is empirical in nature and specific to the conditions
contacts, shown in Fig. 1b involve only the contacting interface
studied. When it comes down to an engineer needing to know the
with fixed relations between the external force F o and the fric-
friction or the friction coefficient for a specific contacting inter-
tion force Q o , e.g., a friction coefficient ( s Q o /F o ). In the
face such as a control engineer in a robotics application 5, then
static case, the forces are only a function of the static normal
the usual source of information are tabulated handbook values,
separation, d o , which is determined once the external force F is
e.g., 6. Blau 7,8 recently compiled various sources of friction
known as in typical contact applications. Dynamic friction in-
coefficient values and he reported that identical interfaces could
have different friction coefficient values. For example, based on 3 volves both the interface and the dynamics of the sliding system,
different references, the kinetic friction coefficient for unlubri- as shown schematically in Fig. 1c. F(t) and Q(t) vary with time
cated steel-on-steel ranges from 0.120.6 8. Clearly, the design as a function of the instantaneous absolute and relative displace-
engineer is faced with a predicament on choosing the appropri- ments, y(t), velocities, and accelerations of the sliding bodies.
ate friction coefficient value for a specific application. Ap- Also, under dynamic conditions damping is present at the inter-
proaches that can account for system dynamics including damping face, designated by a linear damping coefficient c. Under dynamic
and the contact behavior, i.e., all time varying contributions to loading conditions, dividing the time dependent friction force
normal and tangential contact loads are needed to address current Q(t) with a nominal constant normal load e.g., mass of the slid-
and future technological needs in terms of basic understanding ing system to obtain the kinetic friction coefficient is not correct,
and simulation needs. since the normal load is usually not constant unsteady under
Static friction, defined as the force just sufficient to prevent sliding dynamic conditions.
relative motion between two bodies, implies that static friction is A further complication of the interpretation of dynamic friction
a constant value and can easily be divided by the external load arises from the uncertainty of loss of contact during sliding, which
to give the static friction coefficient. On the other hand, kinetic will result in an apparent reduction in friction 9. Also, an uncer-
or dynamic friction force is the minimum force needed to main- tainty arises when converting time varying dynamic friction quan-
tain relative motion between two bodies. This definition implies tities to constant numbers that are typically reported in handbooks.
that the dynamic friction and dynamic friction coefficient are Numerous researchers have studied dynamic friction and asso-
time varying, even though they are typically reported as constant ciated dynamic motions and forces. For unlubricated or dry con-
values 6 8. ditions, see Refs. 918 and for boundary/mixed lubricated con-
An illustration of the differences between static and dynamic ditions see for example Ref. 19. The majority of these research
friction is shown schematically in Fig. 1. Figure 1a shows a works are experimental in nature and propose models that are
based on specific experiments. Inherently such models, which in
Contributed by the Tribology Division for publication in the ASME JOURNAL OF
some cases are complex indeed, are empirical or semi-empirical in
TRIBOLOGY. Manuscript received by the Tribology Division March 6, 2002 revised nature and their range of applicability limited. Tolstoi 10 and
manuscript received December 11, 2002. Associate Editor: M. D. Bryant. Budanov et al. 16 measured the slider motions in the normal to

788 Vol. 125, OCTOBER 2003 Copyright 2003 by ASME Transactions of the ASME
Fig. 1 Contacting pair with friction: a mass block with roughness contacting a surface, insert shows the
Greenwood-Williamson roughness surface model; b model of a static friction pair showing interfacial forces
with adhesion; c model of dynamic friction pair for a sliding mass block with adhesion force; and d model of
dynamic friction pair showing interfacial forces without adhesion

the sliding direction, which depend on the system dynamics, and 2 Dynamic Contact and Friction Modeling
postulated that these normal motions change the normal contact
Even though the concept of friction coefficient, defined as the
strain and thus the magnitude of the kinetic friction force. Also,
friction force divided by the normal force is dating back to daVin-
Godfrey 9 measured friction and friction coefficient reductions
chi, there could be some slightly different definitions and interpre-
with normal vibrations that were attributed to loss of contact. To
tations, especially in relation to dynamic friction. According to
the contrary, based on analytical considerations and using argu-
Tabor 2325 there are three basic elements that are involved in
ments of nonlinear normal contact stiffness e.g., using planar
the friction of unlubricated solids, and any generalized static or
contacts and surface roughness theory 20 it was shown that
kinetic friction model should include all three elements. These
with sliding there is a mean increase in the normal separation
elements are: a The area of true contact between the sliding
between the sliding surfaces compared to the static mean separa-
surfaces; b The type of strength of bond that is formed in the
tion, without any significant changes in the friction coefficient
interface where contact occurs; and c The way in which the
14,21. Such analyses were based on the classic adhesion
material in and around the contacting regions is sheared and rup-
theory of friction, which assumes that the friction force is propor-
tured during sliding.
tional to the real area of contact with planar contacts the load is
It is important to note that the above three elements that govern
proportional to the real area of contact. These findings were also
friction behavior do not differentiate between static and dynamic
supported by the experiments of Soom and Chopra 17. On the
situations.
other hand, based on the same analysis for concentrated Hertzian
contacts, a significant reduction in dynamic friction with an in- 2.1 Static Friction Coefficient. In the simpler static case
crease in the mean separation during sliding 13 is predicted shown in Fig. 1b, the friction coefficient is usually defined as the
with Hertzian contacts the load is nonlinearly related to the real friction force, Q o divided by the external normal load, F o . Such a
area of contact. definition was adopted by Chang et al. 26 28 in developing the
Finally, in dry friction phenomena in general, the friction coef- CEB static friction coefficient model for unlubricated planar sur-
ficient is observed to reduce with increased speed, see for example faces. The CEB model contains the three basic elements men-
Ref. 22. There are several possible explanations for this phe- tioned above and is given by:
nomenon: At sufficiently high sliding speeds and rougher sur-
faces, the sliding system can be considered as a base excitation Qo Qo
system. In such cases, with increasing speed, the excitation am- s (1)
plitude and perhaps frequency also increase causing an apparent F o P o F so
friction coefficient reduction. Another reason is that with increas-
ing speed interfacial temperature effects become more significant, Under static conditions, the external normal force F o is equal to
causing an apparent friction coefficient reduction. the actual contact load in the true area of contact P o minus the
In this paper, a dynamic friction coefficient model for realistic adhesion force, F so . As also shown in Figs. 1a1b the contact
rough planar surfaces under unlubricated sliding conditions is de- asperity force is represented by a non linear stiffening spring,
veloped. The model is based on a static friction coefficient model whereas the adhesion force is represented by a nonlinear attrac-
that incorporates key interfacial forces and rough surfaces and is tive spring acting in opposite directions. Q o is the tangential
combined with a system dynamic model to represent the sliding friction force needed to shear the junctions between the contact-
tribosystem. From the system dynamic model, the instantaneous ing surfaces, and is indirectly related to the contact and adhesion
time varying normal separation at the interface is obtained under forces via the normal separation d o . The contact load is related to
normal harmonic excitation. Subsequently, the instantaneous dy- the true area of contact through the general problem of contacting
namic contact load, interfacial damping, friction force and dy- rough surfaces 20,26. The adhesion relates to the strength of the
namic friction coefficient are calculated as a function of the in- bond formed at the interface 27,29, and the tangential force to
stantaneous time dependent normal separation. the shearing of the contacting asperities and, hence to the friction

Journal of Tribology OCTOBER 2003, Vol. 125 789


force 28,30. A key element of the CEB static friction model is elastic/plastic 26. In the presence of significant intermolecular/
the inclusion of the surface roughness, which is modeled using the adhesion forces, an additional non-linear attractive force is present
GW statistical contact model 20. at the interface, as shown in Fig. 1c. Under sliding conditions
Pollock 31 suggested a different definition for the static fric- the dynamic excitation can be:
tion coefficient: the friction force divided by an apparent normal
force, i.e., s Q o /F app . The apparent normal force F app is the a. Internal excitation due to the surface roughness, which is
equivalent external normal force and is different than the normal usually referred as random roughness induced vibration 11.
external force due to the presence of adhesion. This definition This will be the case when no significant external loading is
poses difficulties in calculating the apparent normal external force applied to the sliding mass.
for general rough surfaces. Note that both static friction coeffi- b. External loading applied to the mass. This is the case when
cient definitions entail the coupling between friction, contact and significant external forces are exerted on the mass during
adhesion forces via the normal separation between the contacting sliding, and the roughness excitation is negligible.
surfaces. The latter case is considered in this paper. Furthermore, to mi-
2.2 Dynamic Friction Coefficient. Referring to the CEB nimize the effects of internal excitation, the simulations in
static friction model Eq. 1, a question that arises is whether the this work are restricted to relatively low sliding speeds, where
process of sliding kinetic friction is different than that of static roughness induced excitations and thermal effects on friction are
friction. As a first approximation, under low sliding velocities, the negligible.
thermal effects can be neglected. Therefore, shearing of the as- Consider the model shown in Fig. 1c with a harmonic dy-
perities to either initiate or sustain sliding as well as the adhesive namic load applied to the mass block in the normal direction as
forces are the same for both static and sliding conditions. Even FF o (1 cos t) and negligible surface roughness excitation.
though the contact force P has been developed for static contacts, The normal contact force P is modeled with a nonlinear spring
this force is also valid under sliding conditions, since typical as- and the dynamic effect of the system is captured by a linear non-
perities have very small slopes and Hertzian calculations are valid viscous damping element. Under dry sliding conditions, the inter-
32. Thus, the basic interface models in the CEB static friction facial damping is usually very small and it increases slightly with
coefficient model are used as the basis for developing the dynamic wear debris 17. Also, in this work the interfacial adhesion force
contact and friction model. An additional complexity of the dy- is neglected refer to Fig. 1d, which is justified for rougher
namic friction is the time dependency of the interfacial forces interfaces, as detailed in section 3.2.
involved, in addition to their dependency on the normal separa- The surface roughness is modeled using the GW 20 contact
tion. The time dependency will be coupled/derived from the sys- model where it assumes that the asperities are modeled as spheres
tem dynamic model. with the same radius of curvature and their height from the mean
As with the static case, two different definitions for the dynamic plane follows a certain probability density function. It is experi-
friction coefficient are discussed. The first definition is the same as mentally found that many engineering surfaces follow Gaussian
the CEB static friction coefficient case, i.e., the instantaneous time normal or Gaussian type distributions 20,33, which precludes
dependent friction force divided by the instantaneous normal ex- closed form solution of the contact equations. Polycarpou and
ternal force, i.e., k (t)Q(t)/F(t). A second definition that more Etsion 34 were able to obtain closed form solution of the contact
correctly captures the true interfacial friction properties of the load, real contact pressure and number of contacting asperities
contacting surfaces is adopted in this paper. In this case, the dy- using an exponential fitting to the original half Gaussian distribu-
namic friction coefficient is defined as the ratio of the instanta- tion, i.e.,
neous friction force to the instantaneous dynamic normal interfa- * c 1 e d 0 y / (3)
cial force P d :
where * is the normalized probability density function of the
Q t Q t Q t asperity heights; c 1 17 and 3 are curve-fitted constant param-
k t (2)
P d t F t my t P t F s t cy t eters that give results comparable to the Gaussian distribution re-
sults 34; is the standard deviation of asperity heights; d o is the
where my is the inertia term associated with the time dependent static equilibrium distance between two rough surfaces in contact,
normal acceleration; cy is the interfacial damping; P d is the sum measured from the mean of the asperity heights; y is the normal
of the dynamic interfacial forces, obtained from a dynamic force displacement measured from the static equilibrium position. Using
balance of the forces in Fig. 1c, and is equal to the external force this distribution of asperity heights, the resulting nonlinear contact
F plus the inertia term, as well as to the contact force minus the or spring force assuming all contacts are elastic is given by 34:
adhesion force and the damping force, i.e., P d (t)F(t)my(t)
P(t)F s (t)cy(t). Notice that the contact force is acting in
the opposite direction than the attractive adhesion force as also
shown in Fig. 1a. The advantage of this definition is that it
P
c 1 EA n
5/2
d y /
R
e o (4)

couples the system dynamics through the inertia and damping


where A n is the nominal contact area; E is the composite elastic
terms. Note that a similar dynamic friction coefficient definition
modulus for the two contacting surfaces; R is the radius of cur-
was also adopted for planar rough contacting surfaces in Refs.
vature of asperity summits, and R is a roughness param-
14, 17, but without damping and without adhesion forces.
eter. The static equilibrium position is obtain from Eq. 4 and is
2.3 Dynamic System Model. In order to calculate the dy- given by
namic friction and dynamic friction coefficient given by Eq. 2,
one needs to consider the dynamic system model. When consid-
ering a mass sliding on a rough surface with friction Fig. 1a,
several variations of simple 1 degree-of-freedom dof lumped
d 0


lnmgF o
k
, where k
c 1 EA n
5/2

R
(5)

parameter models can be used to model its normal dynamics ig- For the sliding mass system considered in Fig. 1d, two differen-
noring the additional dofs. Such a simple model is depicted in tial equations are necessary to fully describe the system, one dur-
Fig. 1d showing a mass block sliding on a rough surface with a ing contact and one when contact is lost. For a rough surface, a
constant velocity under nominally unlubricated conditions. Typi- reasonable criterion for loss of contact is when the mean separa-
cally, the mass block is supported by a nonlinear spring that rep- tion is greater than 3, a criterion that is also adopted in this
resents the contacting asperities and the contact force is calculated research. Therefore, contact is maintained when d o y3 , and
from a contact asperity model, which can be purely elastic 20, or the equation of motion is

790 Vol. 125, OCTOBER 2003 Transactions of the ASME


Table 1 Simulation parameters

E H An m
nm R ( m) m2 (GPa) (GPa) (m2 ) (Kg)
100 110 0.004 0.044 110 7 1.6104 0.5

mycy PmgF o 1 cos t 0 In Contact (6a)


where P is given by Eq. 4. When d o y3 the mass block
looses contact with the flat surface and the dynamic equation be-
comes:
mymgF o 1 cos t 0 Out of Contact (6b)
Solution of the system dynamic equations gives the instantaneous
normal displacement y, velocity y and acceleration y of the mass
block, at each time step. In the absence of adhesion, the dynamic
normal interfacial force P d is obtained as the sum of the spring Fig. 2 Interfacial forces versus external load under static con-
restoring force given by Eq. 4, and the dynamic damping force ditions. Simulations parameters listed in Table 1.
cy, i.e., P d Pcy. Having obtained the dynamic normal inter-
facial force, it is then used in the dynamic friction coefficient Eq.
2 to obtain the kinetic friction coefficient, k . The friction force
Q in Eq. 2 is calculated quasi-statically at each time step by: which is near the linearized resonance ( o 5.74104 rad/s) of


the dynamic system model depicted in Fig. 1d. Note that the
Q d An d
d c 4
3
ER 1/2 ud 3/2 f
ud
c
, u du (7)
linearized damping ratio is linearized with respect to the linear-
ized resonance, about the static equilibrium position, i.e.,
c/2m o where o is the linearized resonance. The simulation
where the local interference is given by ud, c is the criti- procedure consists of first solving for the instantaneous displace-
cal interference at the interception of plastic deformation, and f is ment of the mass, y, using the dynamic system equations, Eqs. 6.
a dimensionless function, which depends on the failure inception Then at each y, the dynamic contact force P d neglecting adhe-
location 27. sion is calculated as P d (t) P(t)cy(t) and using Eq. 4 for
the contact load P. Also, the friction force is calculated using Eq.
7 and together with the dynamic contact friction force are sub-
3 Simulations and Discussion stituted in Eq. 2 to obtain the time dependent dynamic friction
coefficient. Using this procedure, nine different dynamic simula-
3.1 Simulation Parameters. Table 1 lists the simulation pa- tion cases were run using the same roughness, material, excitation
rameters used in this study. The roughness and material param- and external loading conditions as in the static case. Three differ-
eters represent typical macro engineering applications such as a ent values of the linearized with respect to the linearized reso-
worn compressor surface 35: The roughness and material param- nance damping ratio, 0, 1, and 5 percent, and three different
eters in Table 1 correspond to purely elastic contact conditions as values of the amplitude of the external force excitation, 1, 5,
evident from the value of the plasticity index which is 0.47. Val- and 10 percent were considered as shown in Table 2. Also, sum-
ues less than 0.6 indicate purely elastic contact conditions, and marized in Table 2 are the mean values of the time-varying normal
values greater than 1 indicate purely plastic contact conditions separation, contact force, friction force, and dynamic friction co-
20,26. efficient. Contact of the sliding mass block with the flat surface is
maintained at all times except for case 7 0; 10 percent and
3.2 Static Contact. For a specific contact situation, the case 8 1 percent; 10 percent.
static case has a constant friction coefficient value and is consid- Figures 3 and 4 show the simulation results for two of the cases
ered first. Referring to Fig. 1b, the contact and friction forces are listed in Table 2. Both cases have the same damping ratio of 1
calculated from Eqs. 4 and 7, respectively using Eq. 3 for the percent, but different excitation amplitude: case 5 with 5 per-
exponential distribution of asperity heights. Similarly the adhesion cent and case 8 with 10 percent. Figures 3a and 4a show
force is calculated using Eq. 25 in Ref. 27. The procedure for the time varying steady-state instantaneous displacements of the
calculating the static friction coefficient is implicit: One solves the sliding mass. Also, in the cases shown in Figs. 3 and 4, as well all
contact equations for all relevant mean normal separations, which the cases depicted in Table 2, the nonlinearity of the contact stiff-
determine the individual forces and thus the static friction coeffi- ness causes a positive average deflection. With lower input exci-
cient at all relevant external loads 28. Knowledge of the external tations as in case 5, the amplitude of the normal vibration is small
load in a specific contact application determines the constant ranging from 14.28 nm to 16.38 nm with an increase of the
value for the static friction coefficient. Figure 2 shows the contact mean separation of only 0.6 nm, which corresponds to a 0.2 per-
P o , friction Q o and adhesion F so forces versus the external force cent increase from the static equilibrium position d o see Table 2.
F o for the contact conditions listed in Table 1. Clearly over the Under the same conditions but higher input amplitude excitation
whole range of F o , the adhesion force is about 2 orders of mag- case 8 the normal vibration is much larger, ranging from 57.7
nitude lower than the friction and contact forces, and thus negli- nm to 130.4 nm and the corresponding increase in mean value is
gible. The friction and normal forces for an external load of F o 29.6 nm, which corresponds to a 9.85 percent increase from the
50 N in addition to the mass load are Q o 47.823 N and P o static equilibrium separation.
54.905 N, respectively. The corresponding static friction coeffi- Figures 3be and 4be show the dynamic contact force,
cient is constant s 0.871. Note also that for the loading condi- instantaneous friction force and instantaneous friction coefficient,
tions considered, the static equilibrium separation calculated from for 5 percent and 10 percent, respectively. For the cases shown
Eq. 5 is d o 300.5 nm. in Figs. 3 and 4 as well as all other cases listed in Table 2, the
mean values for both the contact and friction forces are unchanged
3.3 Dynamic Sliding Contact. Simulations under different and are about the same as the static values. This unexpected result
linearized damping ratios and excitation amplitude coefficients occurs irrespective of a the large force oscillations, especially
were run using an excitation frequency of 5.2104 rad/s, with 10 percent, and b the loss of contact. Even though dur-

Journal of Tribology OCTOBER 2003, Vol. 125 791


Table 2 Dynamic simulation results

y(t) nm P d (t) N Q(t) N k (t)


Simulation Loss of
Case % % Mean mean mean mean Std Contact
1 0 1 0.042 54.91 47.82 0.871 4.7108 No
2 1 1 0.023 54.90 47.82 0.871 5.3104 No
3 5 1 0.018 54.91 47.82 0.871 0.0024 No
4 0 5 1.252 54.90 47.82 0.871 6.8108 No
5 1 5 0.594 54.90 47.82 0.871 0.0028 No
6 5 5 0.460 54.91 47.83 0.871 0.012 No
7 0 10 23.23 55.00 47.91 0.731 0.32 Yes
8 1 10 29.58 54.99 47.89 0.632 0.40 Yes
9 5 10 2.102 54.91 47.83 0.872 0.028 No
Static Values: Q47.823 N; P54.905 N; s 0.871

ing loss of contact the friction force is zero, the large oscillations 16. Soom and Chopra 17 measured this phase lag for unlu-
that occur, maintain the same mean values. Clearly a mean value bricated sliding extended contact conditions and reported that
of the time varying friction force is inadequate to explain the 0, 7, and 20 degrees for clean virgin surfaces, surfaces with
observed dynamic friction phenomena. wear debris and surfaces lubricated with graphite, respectively.
The simulation results in this paper also support these findings.
4 Interfacial Damping Effects Figure 5 shows the amplitude and phase plots between P(t) and
Q(t) for simulation cases 1, 2, and 3, respectively. These plots
The effect of interfacial damping on the system response is were obtained using Transfer Function Estimate TFE methods
summarized in Table 3. Increasing the damping ratio will suppress 37. For case 1 shown in Fig. 5a similarly for all cases with
nonlinear behavior and decrease the response amplitude of the 0, the amplitude of the TFE is constant over the overall fre-
system. For example, a 78 percent reduction in the peak-to-peak quency range from DC to 50 KHz, and equals the static friction
amplitude is observed when increasing from 1 percent to 5 per- coefficient value of 0.871. Also, as expected with zero damping
cent for the case of 10 percent see Table 3. This effect will the phase shift is zero. When 1 percent, depicted in Fig. 5b,
delay loss of contact, as clearly seen in case 9, which agrees with the phase shift is about 1 degree at the frequency range from 5
Soom and Chen 36. By suppressing the nonlinear oscillations,
the magnitude of the contact and friction forces also decrease, a
phenomenon that is not captured by the mean values for the time
varying forces. For example, for 10 percent, Q varies from 0 to
311 N, from 0 to 270 N, and from 25 N to 83 N for 0, 1
percent, and 5 percent, respectively, whereas the mean value re-
mains unchanged see Table 2.
An important interfacial damping effect is the phase lag be-
tween the instantaneous friction force and the instantaneous nor-
mal contact force, with the friction force lagging the contact force

Fig. 4 Dynamic contact and friction simulation for Case 8


F o 50 N, 8.3 KHz, 0.01, 0.1: a instantaneous nor-
mal displacement; b dynamic contact force; c friction force;
and d dynamic friction coefficient

Table 3 Amplitude response under different damping ratios

Maximum and peak- 0 1% 5%


to-peak amplitude
response y(t) nm max p-p max p-p max p-p
Fig. 3 Dynamic contact and friction simulation for Case 5 1%: Cases 1,2,3 3.5 6.8 3.0 5.9 1.9 3.8
F o 50 N, 8.3 KHz, 0.01, 0.05: a instantaneous nor- 5%: Cases 4,5,6 19.6 36.2 16.4 30.7 10.1 19.4
mal displacement; b dynamic contact force; c friction force; 10%: Cases 7,8,9 157.4 219.8 130.4 188.1 22.3 40.8
and d dynamic friction coefficient

792 Vol. 125, OCTOBER 2003 Transactions of the ASME


Fig. 6 Dynamic contact and friction simulation for F o 5 N,
3.82 KHz, 0.1, 0.05: a normal displacement; b dy-
namic contact force; c friction force; and d dynamic friction
coefficient

tude and higher damping: F o 10 N, 5 percent, 10 percent,


3.82 KHz. The simulation results are shown in Fig. 6. The
static values are Q o 8.63 N, P o 9.90 ( s 0.871 same as be-
fore since adhesion is ignored. The mean increase of the normal
separation is y 0.13 nm, and contact is maintained at all times.
The average values for the dynamic contact and friction forces are
9.91 N, 8.63 N, respectively. Upon examination of the time-
varying dynamic forces shown in Fig. 6, one can observe the
phase lag between the friction and normal forces. Figure 7 depicts
the amplitude and phase lag of the TFE between the friction and
normal dynamic forces. The phase lag is about 11 deg in the
frequency range around the contact resonance and the associated
dynamic friction reduction is about 2 percent. Clearly such an
interpretation correctly captures dynamic friction coefficient be-
havior. Knowing the excitation frequency, one can readily obtain a
constant value for the dynamic friction coefficient from the am-

Fig. 5 Transfer function estimate between interfacial normal


and friction forces: a Case 1 F o 50 N, 8.3 KHz, 0.01,
0; b Case 2 F o 50 N, 8.3 KHz, 0.01, 0.01; and
c Case 3 F o 50 N, 8.3 KHz, 0.01, 0.05

KHz to 15 KHz, which is negligible. When the damping ratio


increases to 5 percent see Fig. 5c, the phase shift also in-
creases to 5 degrees from 5.8 KHz to 14 KHz. Also, in this fre-
quency range, the amplitude of the TFE decreases slightly by 0.2
percent.
To emphasize the effect of damping on the dynamic friction
coefficient, consider a case with the same roughness and material Fig. 7 Transfer function estimate between interfacial normal
parameters as before, but with a lower normal excitation ampli and friction forces for F o 5 N, 3.82 KHz, 0.05, 0.1

Journal of Tribology OCTOBER 2003, Vol. 125 793


plitude of the TFE. This method/interpretation of dynamic friction ring to k (t) and the standard deviation of the kinetic friction
coefficient is valid with continuous contacting interfaces, i.e., no coefficient, listed in Table 2, in all cases with no loss of contact
loss of contact. k (t) s .
Based on this interpretation of the average k (t) , the simu-
lation results in this paper agree with the work of Refs. 15, 17
5 Average Nature of Dynamic Friction Coefficient and that in the absence of damping there is no significant kinetic fric-
Interpretations tion coefficient reduction with vibrations for unlubricated planar
surfaces, and with Ref. 9 that there is significant friction coeffi-
As discussed in sections 3 and 4 for a sliding block, all dynamic
cient reduction associated with loss of contact.
interfacial forces and consequently the instantaneous dynamic
friction coefficient are varying with time. Due to this dynamic 5.2 Interpretation B. Another interpretation of the average
friction nature, several different interpretations of the dynamic nature of the time varying dynamic friction coefficient is to con-
friction coefficient as it relates to constant fixed or average values sider the ratio of the mean value of the friction force to the mean
that are usually reported in the literature are possible, as discussed value of the normal contact force, denoted as a v as also dis-
below. cussed by 13:
5.1 Interpretation A. Perhaps the most obvious average in- Q Q
terpretation of dynamic friction coefficient is to take the time av- av (8)
erage of k (t), denoted by k (t) , as it is also discussed in Ref. Pcy P cy
13. For the simulation cases considered in this paper, these av-
For the simulations cases considered in this paper, cy is very
erage values are tabulated in Table 2, and again in Table 4. For
small and can be ignored. Therefore, the average coefficient of
simulation case 5 shown in Fig. 3 lower excitation despite the
friction reduces to a v Q / P . Since the average values of the
small fluctuations in the normal displacement and forces, the re-
friction and contact forces are always constant and equal to the
sulting kinetic friction coefficient is relatively constant. The mean
static cases see Table 2, under this interpretation the average
value of the friction coefficient k (t) 0.871, exactly the same
kinetic friction coefficient is always constant, irrespective of loss
as the static case, with an insignificant variation of one standard
of contact, and is equal to the static friction coefficient value, i.e.,
deviation equal to 0.0028 see Table 2. To the contrary when the
a v s . These values are also shown in Table 4, and such a
excitation increases to 10 percent as shown in Fig. 4, the large
dynamic friction coefficient interpretation contradicts the experi-
vibrations of the slider result in large changes in the contact and
mental data of Ref. 9 that there is significant friction coefficient
friction forces. Furthermore, when the oscillations are large, loss
reduction associated with loss of contact. Therefore such an inter-
of contact occurs, with the corresponding friction force and dy-
pretation is questionable.
namic friction coefficient fall to zero. In this case, the average
value of the friction coefficient is k (t) 0.632, which corre- 5.3 Interpretation C. In previous research studies, e.g.,
sponds to a 27.4 percent decrease from the static friction coeffi- Ref. 16, an increase in the mean normal separation was used to
cient value. Based on such a dynamic friction coefficient interpre- explain the reduction in the dynamic friction and dynamic friction
tation, it is predicted that compared to the static case, simulation 5 coefficient. To the contrary, Hess and Soom 15 reported that a
will not lead to a change in the kinetic friction coefficient, positive deflection will sometimes but not always be accompanied
whereas case 8 will lead to a significant decrease in the friction with a friction force reduction. Based on the first two dynamic
coefficient due to loss of contact. friction interpretations, the simulation results in this work do not
Based on the simulation results, there are two possible causes support the claim in Ref. 16. However, if one considers the
for reducing k (t) . The first is due to the phase lag between the increase in the normal separation with sliding in a quasi-dynamic
friction force and the contact force see section 4, and the second sense and calculates the friction force and friction coefficient at
due to the loss of contact of the sliding mass. Assuming that under the static equilibrium position plus the mean separation increase,
normal excitation there is no contact loss cases 1 6, 9 then the i.e., d o y , then a decrease in the friction force, but not the
dynamic friction coefficient depends only on the contact damping. friction coefficient is predicted, as depicted in Table 4 damping is
However, under unlubricated sliding conditions the damping is ignored. For example, considering simulation case 8 with a mean
small and can be ignored. This is justified since for the simula- separation increase of 29.58 nm and statically calculating the fric-
tions in this work, the velocity of the mass block is small, typi- tion force and friction coefficient at this new separation of 329.58
cally of the order of 104 103 m/s, and thus the damping force nm, then a lower friction force of 19.68 N 59 percent reduction
term cy is also very small compared with the spring force P(t) from the static friction force is predicted. Despite this friction
and friction force Q(t). Under these conditions, which are typical force reduction, which appears to be unrealistically high, the con-
for unlubricated sliding under low to moderate harmonic excita- tact force also decreases by the same percentage, and thus the
tions, the kinetic friction coefficient in the absence of adhesion dynamic friction coefficient is unchanged. The authors cautioned
reduces to k Q(y)/ P(y) , which is about constant, and against such an interpretation since in a true dynamic sense, an
equal to the static friction coefficient, as shown in Table 2. Refer- increase in the mean separation does not necessarily result in a

Table 4 Average nature of dynamic friction coefficient and its interpretations

Simulation Case:
Interpretation 1 2 3 4 5 6 7 8 9
A: k (t) 0.871 0.871 0.871 0.871 0.871 0.871 0.731 0.632 0.872
Q
B: a v 0.871 0.871 0.871 0.871 0.871 0.871 0.871 0.871 0.871
Pcy
*C: @d 0 y 0.871 0.871 0.871 0.871 0.871 0.871 0.871 0.871 0.871
C: d 0 y (nm) 300.54 300.52 300.52 301.75 301.09 300.96 323.73 330.08 302.60
C: Q @d 0 y (N) 47.73 47.76 47.77 46.03 46.95 47.14 23.81 19.68 44.87
C: P @d 0 y (N) 54.77 54.80 54.81 52.82 53.87 54.09 27.32 22.58 51.49

*Ignore damping

794 Vol. 125, OCTOBER 2003 Transactions of the ASME


decrease of the average value of the time varying friction force y normal displacement of the mass block, measured
Q(t) and contact force, P(t) due to the nonlinear relation between from the static equilibrium position d o
the forces and the separation. y normal velocity of the mass block
y normal acceleration of the mass block
6 Conclusions external normal force amplitude coefficient
roughness parameter, R
A dynamic friction coefficient model for realistic unlubricated
linearized damping ratio
rough planar surfaces is proposed. The dynamic friction coeffi-
cient is defined as the ratio of the time varying friction to the
areal density of asperities
a v ratio of average friction force to the average normal
interfacial normal forces that explicitly include interfacial damp-
force
ing. The model couples interfacial forces and damping with the
s static friction coefficient
dynamic system of the sliding interface. Based on the simulation
k dynamic or kinetic friction coefficient
results it is suggested that when dealing with dynamic friction
phenomena, it is preferable to consider the individual true time- k time average of the kinetic friction coefficient
constant coefficient in Eq. 3
varying dynamic interfacial forces. When coupling interfacial
phase lag between the friction force and the normal
forces to obtain a dynamic friction coefficient, one needs to exer-
interfacial force
cise extra caution since it is more difficult to interpret, compared
distribution function of asperity heights, */
to static friction. Furthermore, the complications of interpreting
standard deviation of asperity heights
and converting time-varying dynamic friction coefficient to con-
interference
stant kinetic friction coefficient values are discussed. The pro-
c critical interference at the inception of plastic defor-
posed dynamic friction coefficient model predicts that for unlubri-
mation
cated rough planar contacts, the mean value of the dynamic
o linearized resonance about the static equilibrium posi-
friction coefficient remains approximately constant when loss of
tion
contact is avoided and interfacial damping is negligible. Also, the
external excitation frequency
proposed dynamic friction coefficient model predicts a reduction
in the average value of the dynamic coefficient with loss of con-
tact and with significant damping presence at the interface. An References
accurate method to interpret dynamic friction coefficient under 1 da Vinci, L., Notebooks and Manuscripts, presented in the late 1400s, as de-
continuous contact conditions is with the use of the transfer func- scribed in 2.
2 Dowson, D., 1998, History of Tribology, 2nd ed., Professional Engineering
tion between the true time-varying normal and friction forces. Publishers, London.
Such a method captures the friction coefficient reduction associ- 3 Amontons, G., 1699, On the Resistance Originating in Machines, Proceed-
ated with the phase lag between dynamic normal and friction ings of the French Royal Academy of Sciences, pp. 206 222.
forces. When loss of contact is encountered dynamic friction in- 4 Coulomb, C. A., 1785, Theorie des machines simples, en ayant egard au
frottement deleurs parties, et a la roideur dews cordages, Mem. Math Phys.,
terpretation is more ambiguous, and average values may or may pp. 161342.
not correctly capture the true nature of time-varying dynamic fric- 5 Armstrong-Helouvry, B., 1991, Control of Machines with Friction, Kluwer
tion behavior. Academic Publishers, Boston.
6 Booser, E. R., 1997, Tribology Data Handbook, CRC Press, Boca Raton.
7 Blau, P. J., 2001, The Significance and Use of the Friction Coefficient,
Acknowledgments Tribol. Int., 34, pp. 585591.
8 Blau, P. J., 2001, Experimental Aspects of Friction Research on the Macro-
This research was supported by the Department of Mechanical scale, in Fundamentals of and Bringing the Gap Between Macro- and Micro-/
and Industrial Engineering of the University of Illinois at Urbana- Nanoscale Tribology, B. Bhushan, ed., Dordrecht, Kluwer, pp. 261278.
Champaign. The authors gratefully acknowledge this support. 9 Godfrey, D., 1967, Vibration Reduces Metal to Metal Contact and Causes an
Apparent Reduction in Friction, ASLE Trans., 10, pp. 183192.
10 Tolstoi, D. M., 1967, Significance of the Normal Degree of Freedom and
Nomenclature Natural Normal Vibrations in Contact Friction, Wear, 103, pp. 199213.
11 Soom, A., and Kim, C., 1983, Roughness-Induced Dynamic Loading at Dry
An nominal contact area and Boundary-Lubricated Sliding Contacts, ASME J. Lubr. Technol., 105, pp.
c1 constant coefficient in Eq. 3 514 517.
c damping coefficient 12 Martins, J. A. C., Oden, J. T., and Simoes, F. M. F., 1990, A Study of Static
and Kinetic Friction, Int. J. Eng. Sci., 28, pp. 2992.
d mean normal separation based on asperity heights 13 Hess, D. P., and Soom, A., 1991, Normal Vibrations and Friction Under
do static equilibrium normal separation, based on asper- Harmonic Loads: Part I: Hertzian Contacts, ASME J. Tribol., 113, pp. 80 86.
ity heights 14 Hess, D. P., and Soom, A., 1991, Normal Vibrations and Friction Under
E composite elastic modulus for the two contacting sur- Harmonic Loads: Part II: Rough Planar Contacts, ASME J. Tribol., 113, pp.
8792.
faces 15 Hess, D. P., and Soom, A., 1992, Unsteady Friction in the Presence of Vibra-
F external normal force tions, in Fundamentals of Friction: Microscopic and Microscopic Processes,
Fo external normal force under static conditions I. Singer, and H. Pollock, eds., Kluwer Academic Press, pp. 535552.
F app apparent normal force in Ref. 31 16 Budanov, B. V., Kudinov, V. A., and Tolstoi, D. M., 1980, Interaction of
Friction and Vibration, Trenie Iznos, 11, pp. 79 89.
Fs adhesion force 17 Soom, A., and Chopra, A., 2001, In Search of Dynamic Effects in Dry Slid-
F so adhesion force under static conditions ing Friction, in Tribology Research: From Model Experiment to Industrial
g acceleration due to gravity Problem, G. Dalmaz, A. A. Lubrecht, D. Dowson, and M. Priest, Eds.,
H hardness of softer material Elsevier, pp. 5559.
18 Tariku, F. A., and Rogers, R. J., 2001, Improved Dynamic Friction Models
k constant coefficient in Eq. 5 for Simulation of One-Dimensional and Two-Dimensional Stick-Slip Motion,
m mass of the sliding block ASME J. Tribol., 123, pp. 661 669.
P normal contact force 19 Polycarpou, A. A., and Soom, A., 1996, A Two-Component Mixed Friction
Po normal contact force under static conditions Model for a Lubricated Line Contact, ASME J. Tribol., 118, pp. 183189.
20 Greenwood, J. A., and Williamson, J. B. P., 1966, Contact of Nominally Flat
Pd dynamic normal interfacial force Surfaces, Proc. R. Soc. London, Ser. A, A295, pp. 300319.
Q tangential friction force 21 Johnson, K. L., 2001, Dynamic Friction, in Tribology Research: From
Qo tangential friction force under static conditions Model Experiment to Industrial Problem, G. Dalmaz, A. A. Lubrecht, D. Dow-
R average radius of curvature of asperities son, and M. Priest, eds., Elsevier, pp. 37 45.
22 Sakamoto, T., 1987, Normal Displacement and Dynamic Friction Character-
t time istics in a Stick-Slip Process, Tribol. Int., 20, pp. 2531.
u height of asperity measured from the mean of asper- 23 Tabor, D., 1981, FrictionThe Present State of Our Understanding, ASME
ity heights J. Lubr. Technol., 103, pp. 169179.

Journal of Tribology OCTOBER 2003, Vol. 125 795


24 Tabor, D., 1991, Friction, Surface Science and TribologyA Personal View, Friction: Microscopic and Microscopic Processes, I. Singer, and H. Pollock,
Proc. Inst. Mech. Eng., Part C: J. Mech. Eng. Sci., 205, pp. 365378. eds., Kluwer Academic press, pp. 7794.
25 Tabor, D., 1995, TribologyThe Last 25 Years: A Personal View, Tribol. 32 Huang, W., Bogy, D. B., and Honchi, M., 2000, An Asperity Contact Model
Int., 281, pp. 710. for the Slider Air Bearing, ASME J. Tribol., 122, pp. 436 443.
26 Chang, W. R., Etsion, I., and Bogy, D. B., 1987, An Elastic-Plastic Model for 33 Yu, N., and Polycarpou, A. A., 2002, Contact of Rough Surfaces With Asym-
the Contact of Rough Surfaces, ASME J. Tribol., 109, pp. 257263. metric Distribution of Asperity Heights, ASME J. Tribol., 124, pp. 367376.
27 Chang, W. R., Etsion, I., and Bogy, D. B., 1988, Adhesion Model for Metallic 34 Polycarpou, A. A., and Polycarpou, A. A., 1999, Analytical Approximations
Rough Surfaces, ASME J. Tribol., 110, pp. 5056.
in Modeling Contacting Rough Surfaces, ASME J. Tribol., 121, pp. 234 239.
28 Chang, W. R., Etsion, I., and Bogy, D. B., 1988, Static Friction Coefficient
35 Patel, J. J., 2001, Investigation of the Scuffing Mechanism Under Starved
Model for Metallic Rough Surfaces, ASME J. Tribol., 110, pp. 57 63.
29 Derjaguin, B. V., Muller, V. M., and Toporov, Y. P., 1975, Effect of Contact Lubrication Conditions Using Macro, Meso, Micro and Nano Analytical Tech-
Deformations on the Adhesion of Particles, J. Colloid Interface Sci., 53, pp. niques, M.S. thesis, University of Illinois at Urbana-Champaign.
314 326. 36 Soom, A., and Chen, Jern-Wen, 1986, Simulation of Random Surface
30 Hamilton, G. M., 1983, Explicit Equations for the Stresses Beneath a Sliding Roughness-Induced Contact Vibrations at Hertzian Contacts During Steady
Contact, Proc. Inst. Mech. Eng., Part C: J. Mech. Eng. Sci., 197C, pp. Sliding, ASME J. Tribol., 108, pp. 123127.
5359. 37 Proakis, J. G., and Manolakis, D. G., 1996, Digital Signal Processing: Prin-
31 Pollock, H. M., 1992, Surface Forces and Adhesion, in Fundamentals of ciples, Algorithms, and Applications, Englewood Cliffs, NJ, Prentice Hall.

796 Vol. 125, OCTOBER 2003 Transactions of the ASME

View publication stats


Tribology International 34 (2001) 585591
www.elsevier.com/locate/triboint

The significance and use of the friction coefficient


*
Peter J. Blau
Oak Ridge National Laboratory, P.O. Box 2008 Mail Stop 6063, Oak Ridge, TN 37831-6063, USA

Abstract

The quantity known as the friction coefficient (or coefficient of friction) has long been used in science and engineering. It is
easy to define, but not easy to understand on a fundamental level. Conceptually defined as the ratio of two forces acting, respectively,
perpendicular and parallel to an interface between two bodies under relative motion or impending relative motion, this dimensionless
quantity turns out to be convenient for depicting the relative ease with which materials slide over one another under particular
circumstances. Despite the fact that both static and kinetic friction coefficients can be measured with little difficulty under laboratory
conditions, the time- and condition-dependent characteristics of friction coefficients associated with both clean and lubricated sur-
faces have proven exceedingly difficult to predict a priori from first principles.
The shaky nature of frictions fundamental underpinnings, has not prevented investigators from compiling lists of friction coef-
ficients and publishing them for general use. Problems often arise, however, when engineers attempt to use tabulated friction
coefficients to solve specific problems in mechanical design or failure analysis. The systems-dependence of frictional behavior is
sometimes ignored, leading to misapplication of published data. This is particularly true for applications in nano-technology and
others that differ from typical laboratory size scales. This paper will review the measurement and use of static and kinetic friction
coefficients, discuss their usefulness, and describe the sources of frictional resistances in terms of shear localization. 2001 Elsevier
Science Ltd. All rights reserved.

Keywords: Friction; Friction coefficient; Kinetic friction; Static friction; Tribology terminology

1. Historical underpinnings and definitions of the stant over a range of conditions. Amontons, for example,
friction coefficient is remembered for his two laws of friction:

The dimensionless quantity known as the friction I. The force of friction is directly proportional to the
coefficient, or coefficient of friction as it is sometimes applied load
called, evolved from the work of many philosophers, II. The force of friction is independent of the apparent
scientists and engineers; in particular, da Vinci [1], area of contact.
Amontons [2], and Coulomb [3]. These thinkers
attempted to rationalize the sliding resistance between Dowson [4] calls our attention to Semen Kirilovich Kot-
solid bodies with a universal law that explained obser- elnikov (17231806), a former student of the famous
vations of their day. In early work with simple machines mathematician Euler, who is credited for the use of the
and macro-scale tribometers, it was observed that the Greek mu (m) to represent the friction coefficient. Kot-
proportionality of the force opposing relative motion to elnikov conducted some of the earliest Russian studies
the force holding the bodies together seemed to be con- of friction in the late 1700s. In his book on mechanics
[5], he wrote

If we denote the friction content F and the applied



Research sponsored by the US Department of Energy, Assistant force P as unknowns, in the ratio :1, then friction
Secretary for Energy Efficiency and Renewable Energy, Office of F=P.
Transportation Technologies, as part of the High Temperature
Materials Laboratory User Program, under contract DE-AC05-
96OR22464 with Lockheed Martin Energy Research Corp. Thus, despite the credit given to Coulomb for defining
* Tel.: +1-423-574-5377; fax: +1-423-574-6918. the friction coefficient, the Russian academician Kot-
E-mail address: blaupj@ornl.gov (P.J. Blau). elnikov also has a claim to that distinction.

0301-679X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 0 1 - 6 7 9 X ( 0 1 ) 0 0 0 5 0 - 0
586 P.J. Blau / Tribology International 34 (2001) 585591

Two types of friction coefficients can be dis- Table 1


tinguished: one that represents the friction opposing the Friction coefficients for wood-on-wood presented in several compi-
lations
onset of relative motion (impending motion), and one
that represents the friction opposing the continuance of Listed material ms mk Reference
relative motion once that motion has started. The former combination
is called the static friction coefficient, and the latter, the
kinetic friction coefficient. In the case of solid-on-solid Wood on wood 0.250.5 0.19 [14]
Wood on wood 0.250.5 0.38a [15]
friction (with or without lubricants), these two types of (dry)
friction coefficients are conventionally defined as fol- Wood on wood 0.300.70 [16]
lows: Wood on wood 0.6 0.32 [17]
Wood on wood 0.6 0.5 [18]
msFs/P (1) Wood on wood 0.4 0.2 [19]
Oak on oak 0.62 [20]
mkFk/P (2) (para. to grain)
Oak on oak 0.54 0.48 [20]
where Fs is the force just sufficient to prevent the relative (perp. to grain)
motion between two bodies, Fk is the force needed to Oak on oak 0.62 0.48 [21]
maintain relative motion between two bodies, and P is (fibers parallel)
the force normal to the interface between the sliding Oak on oak 0.54 0.34 [21]
bodies. Formally-accepted definitions for the friction (fibers crossed)
Oak on oak 0.43 0.19 [21]
coefficient can be found in ASTM Standard G-40 on (fibers
Wear and Erosion Terminology [6]. Eqs. (1) and (2) perpendicular)
became widely accepted as quantitative friction-related
a
parameters because they seemed to be suitable for Authors suggest reducing ms by 25% to obtain mk.
designing machines and structures. Values for ms and mk
could be conveniently tabulated and incorporated into shows, different tabulations of friction data list different
engineering handbooks. Such tabulations have existed values even for (supposedly) the same sliding materials.
for at least 300 yr and continue to be published today. In light of the differences in friction coefficients
In the 1700s and later, disputes arose as to the correct reported in the literature, and the common lack of experi-
value of m. Some argued that it was 1/3 and others mental details accompanying most compilations, the
asserted that it was closer to 0.3. As the diversity of value of such compilations is dubious at best. Therefore,
machines grew and the range of speeds and normal additional friction testing under known conditions is
forces used in friction experiments increased, the laws often required to solve specific, applied problems. Tests
introduced by Amontons, Kotelnikov, and Coulomb can be run on the actual machinery, but commonly that
were found to break down in many instances. Further- approach is impractical, and sub-scale, or laboratory
more, ms and mk were eventually recognized to be both tests are conducted instead. In the latter case, it becomes
material- and system-dependent. The implications of important to establish a quantifiable linkage between the
those findings were sobering to those who believed that data obtained in laboratory tests and field performance.
friction was somehow an intrinsic property of the two To develop simulative tests, the factors that affect fric-
contacting materials. The systems approach has therefore tional behavior must be recognized and placed in a pro-
become a tool for the interpretation and use of friction per priority within the tribosystem.
data in modeling friction, developing friction-mitigating
materials, developing friction test methods, and design-
ing machinery. 3. Factors affecting frictional behavior

The forces that resist sliding occur in the regions near


2. Tabulations of friction coefficients and between solid surfaces. The problem of establishing
exactly which attributes of the contact conditions and the
As young science students, we are given the erroneous materials contribute most to the friction force is a major
impression that all friction problems can be solved either one for developing friction tests and analytical friction
by conducting simple experiments or by looking up models. Models for friction have used geometric argu-
values in published tables of friction coefficients [7]. ments (surface roughness and asperity interlocking),
Engineers and scientists confronted with real friction mechanical properties-based arguments (shear properties
problems in machinery or industrial processes often find of the solids and of the substances between the surfaces),
this simple approach insufficient to explain observations fluid dynamics approaches, considerations of electro-
or to enable them to select from among numerous candi- static forces between surface atoms, and chemical com-
date materials and lubricants. Furthermore, as Table 1 patibility arguments. In light of such diverse approaches,
P.J. Blau / Tribology International 34 (2001) 585591 587

the number of possible variables for use in predictive tems, the force of friction is largely a result of shearing
friction models becomes quite large. (See for example, within the lubricant film or the boundary between that
Table 2.) film and one or both contact surfaces. In other cases,
Since the number of potential friction-affecting factors work is done to deform the materials bordering the inter-
is large, it is necessary to identify the set of key variables face. When particles are present, other factors come into
applicable to each particular case in order to select the play. Understanding the manner in which various struc-
appropriate test methods or simulations. tures in and adjacent to the sliding interface contribute
If friction tests are intended to study fundamental to friction is as essential for designing friction test
mechanisms or to evaluate materials or lubricants under methods as it is for the fundamental modeling of fric-
standard controlled conditions, there is more leeway in tional processes.
terms of isolating the system variables. The added con- Fig. 1 depicts the friction stack concept in which a
straints of simulating a specific operating environment number of parallel layers represent potential locations in
are not present. On the other hand, if friction tests are which frictional energy can be dissipated [7,8]. Several
more applications-oriented, then the tribosystem of inter- Frictional Energy Dissipation Zones are identified. If
est must be analyzed and its friction-related variables interfacial shear is localized within Zone 1, then tests
identified to effect a valid simulation. Assessing the that investigate the shear response of fluids in the con-
degree of interfacial shear localization is one approach fines of the contact are appropriate. Two examples of
to identifying the sources of frictional resistance. Zone 1 tests include lubricated disk-on-disk tests with
partial slip, and flat-on-flat tests at contact pressures low
enough so that a full lubricating film is maintained and
4. Interfacial shear localization no solid contact occurs. In Zone 2, some of the frictional
energy is used in deforming (shearing and/or fracturing)
The non-conservative friction force, be it static or kin- the films that cover one or both bodies. Examples of
etic, arises in response to the work needed to enable rela- such films include films formed from friction-modifiers
tive motion between two bodies. In different tribosys- in lubricants, adsorbed films on surfaces, oxides and tar-
tems, the energy associated with this work is distributed nishes, and deposited films of solid lubricants. In Zone 3,
differently. Some of the energy goes into heat, some of energy is dissipated in deforming the contacting solids,
it into the creation of new surfaces (wear), and some is elastically and possibly plastically, depending on the
used in deforming the materials. In well-lubricated sys- severity of the imposed conditions. Accumulated defor-

Table 2
Factors influencing frictional behavior

Category Factor

Contact geometry Conformity of the components (macro-scale mating of shapes)


Surface roughness (microscale featuresasperity shapes, size distribution)
Surface waviness
Surface lay (directionality) with respect to relative motion
Fluid properties and flow Lubrication regime (boundary, mixed, hydrodynamic, elastohydrodynamicfilm thickness and pressure)
Viscosity characteristics of the fluid as it affects Newtonian or non-Newtonian flow
Temperature and pressure effects on viscosity
Shear thinning effects on viscosity in ultra-thin films
Lubricant chemistry Formation of friction-altering films
Stability of friction-modifiers over time
Oxidation and acidification of lubricants
Relative motion Unidirectional or reciprocating motion
Constancy of motion (accelerations, pauses, startstop)
Magnitude of relative surface velocity
Applied forces Magnitude of the normal force (contact pressure)
Constancy of applied forces
Third-bodies Characteristics of particles entrained in the lubricant
Characteristics of particle assemblages contained within the interface (e.g., wear particles, external
contaminants, lubricating powder layers)
Temperature Thermal effects on material properties (thermoelastic instabilities)
Thermal effects on lubricant properties (viscosity, flow, possibility for cavitation)
Friction induced temperature in combination with the temperature of the surroundings
Stiffness and vibrations Contact compliance (stick slip)
Damping of frictional or external vibrations
Feedback between frictional stimulus and structural response
588 P.J. Blau / Tribology International 34 (2001) 585591

(disc brakes, drum brakes, rotating seals, etc.) friction


coefficients are obtained from torque measurements and
component dimensions. Tension-wrap devices use the
differences in tension resulting between the ends of a
sheet of material or a wire wrapped over a circular body.
Oscillation decrement devices involve measurements of
the decrease in swing amplitude of a pendulum sup-
ported by frictional surfaces. Indirect indications are sys-
tem outputs like motor current or vibrations.
A range of friction testing devices, including those
used in ASTM standard test methods, are described else-
where [711]. Each method used to measure friction has
advantages and disadvantages. These must be weighed
in light of the purpose for which friction coefficient data
are needed, and based on the nature of the surface con-
tact in the friction system. Traditionally, if the measured
friction force is F, then there can be various contri-
butions to F from a number of sources, each providing
Fig. 1. Schematic diagram of the friction stack concept showing four
frictional energy dissipation zones (a) and an electrical analog rep- an additional resisting force f. That is,


resentation of sliding resistances in parallel (b). n

F fi (3)
i1
mation can result in fracture (wear). The presence of
highly-deformed layers in polished cross-sections of Examples of fi include the force required to shear the
material subjected to sliding indicates that frictional interfacial medium, the force needed to deform asperities
energy is being dissipated in Zone 3. Zone 4 can extend that have formed adhesive junctions, the force needed to
well beyond the immediate vicinity of the frictional plow hard asperities through the softer surface, the force
interface. It involves the response of the surrounding needed to shear agglomerated deposits of wear debris,
structure to the frictional situation. Only rarely have fric- etc. Omitting one or more sources of frictional resistance
tional tests attempted to vary the stiffness of the sur- from a test method can significantly affect the results.
rounding structure to study its effects on friction. Fric- For example, if a diesel engine ring and cylinder operate
tion-critical applications that must address Zone 4 issues mainly in the presence of sooty oil, friction tests for ring
notably include braking systems in which and liner materials should use sooty oil, not fresh oil.
friction/vibration interactions affect functional perform- Fig. 2 exemplifies frictional situations that require
ance and customer satisfaction with the product. somewhat different approaches to testing. Friction
involving two-body abrasion usually involves high fric-
tion coefficients because the tips of hard asperities on
5. Selection of test methods one surface can dig into the surface and increase the
plowing contribution to the friction force. Friction during
As described elsewhere [7], six categories can be used smooth sliding tends to be lower unless there is a sig-
to characterize friction testing devices: nificant amount of adhesion (as in very clean contacts
or self-mated materials) or if wear occurs so as to
1. Gravitation-based devices roughen the surfaces. Friction in the presence of third
2. Direct linear force measurement devices bodies is rather complex because there can be friction
3. Torque measurement devices within the particle layer itself and friction at the layer-
4. Tension-wrap devices solid interfaces. Density gradients and pores can also
5. Oscillation-decrement devices form in such layers, and the shear strength of particle
6. Indirect indications layers is related to their density in a non-linear manner
[11]. Friction during rolling contact and lubricated inter-
Gravitation-based devices have been proposed for at faces can be affected by the manner in which the fluid
least 500 years, and some of them are shown in the note- is forced into the interface, the roughness of the surface
book sketches of da Vinci [1]. In some configurations, and consequently, a quantity that is known as the -
like flat-on-flat testing or pin-on-disk testing, the friction ratio. The -ratio is the ratio of the mean lubricant film
force can be measured directly with a load cell, strain thickness to the composite surface roughness s. The lat-
gauge, or similar force sensor mounted in line with the ter is defined in terms of the arithmetic averages of the
contact. In other systems, like swept circular contacts surface roughness of the two mating parts as,
P.J. Blau / Tribology International 34 (2001) 585591 589

ence. In a liquid or solid lubricant where simultaneous


localized shear occurs on parallel layers, the friction
coefficient is quite low. When parallel processes operate
in a frictional situation it is possible to experience
momentary siezure on one layer, at which time the shear
switches to another layer of lower shear strength. Shear
layer switching can be responsible for fluctuations in
friction force.
When conducting friction tests to simulate specific
engineering components, visual or microscopic obser-
vations of friction surfaces helps to ensure that the lab-
oratory test method faithfully replicates the shear pro-
cesses that operate in the application.

6. Ambiguity in friction testing terminology

There are cases in which a friction coefficient is


reported, yet there is some ambiguity as to whether that
term is appropriate. Two cases will illustrate this point.
One case involves the term stick-slip as applied to a
periodic instability in the relative motion between bod-
ies. Fig. 3 typifies a tangential force trace associated with
a sliding system in which there is intermittent motion.
Fig. 2. Examples of four different interfacial conditions that com- The linear, upward sloped portions of the curve, labelled
monly arise in frictional systems. Friction in the presence of hard s, indicate times when there is no relative motion
asperities (upper left), smooth sliding with or without an interfacial between surfaces but when the tangential force is rising.
liquid layer (upper right), friction controlled by a sandwiched layer of
particles (lower left), and friction under rolling with slip (lower right).
Only during the portions between the s segments is
there relative motion. Therefore, from the beginning of
ss21+s22
each stick segment until just before breakaway, when
(4) the tangential force can be considered the static friction
The lower the -ratio, the closer the moving surfaces
are together, and the more likely that a protective lub-
ricating film will not be maintained. A value of 3 or
higher is generally favored for good lubricant perform-
ance.
Eq. (3) above suggests that frictional contributions are
additive, like electrical resistances in series, but a con-
sideration of the schematic anatomy of a frictional sys-
tem, as in Fig. 1, indicates that it is possible for frictional
resistances to occur in parallel. One need only consider
the shearing of a deck of playing cards in order to realize
that frictional interfaces can operate in parallel; parti-
cularly, when there are similar shear strengths among
the various layers in a friction stack. Therefore, to under-
stand friction in tribosystems that allow friction pro-
cesses to operate in parallel, one could represent the net
friction force F in the same manner as for electrical
resistances in parallel: Fig. 3. Illustration of stick-slip behavior. The tangential force rises


periodically as the system stores elastic energy with no relative move-
n
1 ment within the contact (s periods in the figure). When the static
F (5) friction force is exceeded, the specimens slip until they come to rest
i1
fi again and the process repeats. By definition, friction force only exists
under conditions of impending motion (static friction force peak just
Note that the operation of each new process tends to prior to slip), and when relative motion is occurring (kinetic friction
reduce the friction force because it provides additional during the slip). A plot of friction force would exclude the portions
means to accommodate the body-to-body velocity differ- labeled s.
590 P.J. Blau / Tribology International 34 (2001) 585591

force, there is no friction force at all, but rather just 7. The relationship between friction and wear
the tangential force. Likewise, the kinetic friction force
only exists during the period between s segments of The energy that is transformed as a consequence of
the curve. It is therefore incorrect to plot a curve such frictional contact can be stored in the tribosystem or dis-
as that shown in Fig. 3 by dividing the force record by sipated in a number of different ways. If Ef is the energy
the normal force and plotting m as a function of time. resulting from sliding contact, Eout is the energy leaving
Secondly, and more relevant to the emerging subfields the tribosystem, and Est is the energy remaining in the
of micro- and nano-tribology, is the issue of friction dur- tribosystem,
ing scratching. Friction coefficients are now being
EfEoutEst (6)
reported in which a fine-scale stylus is drawn across a
surface. Such stylii can be in the form of diamond For example, mechanical energy from sliding can be
indenters (as for scratch hardness or adhesion tests), or converted to heat, vibrations (like sound), to material
perhaps, as narrow silicon nitride pyramids, as in the deformation, or the creation of new surfaces (by
case of atomic force microscope tips. Consider the case fracture). Likewise, it can be stored in the material as
in Fig. 4 in which a sharp point is progressively pushed the energy associated with microstructural defects
into a surface and drawn along it. If this happened to be (deformation twins, work hardening, etc.). Therefore,
a sharp diamond stylus and the surface happens to be two materials exhibiting the same friction coefficient can
copper, would the reported ratio of the tangential force exhibit quite different wear rates because the energy is
to the normal force truly be the kinetic friction coef- partitioned differently between and within the
ficient for diamond on copper? The sharpness of the materials [12].
point affects its degree of penetration, and hence, the The interrelationships between friction and wear can
magnitude of the plowing contribution to the friction change with time, as indicated by considering the
force. Therefore, the reported friction coefficient dur- changes that occur during running in. Fig. 5 shows
ing scratching, even though calculated by taking the ratio results of a series of block on ring sliding tests using the
of tangential to normal force, becomes a strong function same sliding speed, load, and testing environment
of the tip geometry and will probably not match values (sliding dry in air) [13]. The time to reach a constant
for diamond sliding on copper reported elsewhere in wear rate was determined from monitoring the displace-
the literature. A fundamental question then arises: At ment of the block relative to the ring. The time to reach
what degree of sharpness does the term friction coef- steady state friction coefficient was determined from rec-
ficient no longer apply to the material couple? ordings of the friction force versus time behavior. The
The matter is further complicated when considering data show that the times to achieve frictional break-in
the material flow during the process. On a still finer- and to reach wear-in are not generally equal. Thus, there
scale, friction is created as the material displaced from is an interesting interplay between friction and wear as
ahead of the moving stylus slides either around or under- the tribosystem ages. For example, frictional shearing
neath the slider itself. Perhaps it might be better to avoid can lead to the accumulation of damage to the point of
the controversy entirely by coining new terms for stylus fracture, at which time particles are released. These par-
experiments in scratching and lateral force microscopy. ticles alter the frictional resistance, which in turn affects
Perhaps terms like the stylus drag coefficient or the the energy available to continue to generate and expel
tribomolecular force might be more appropriate alter- particles. In the mean time, the interface may be heating
natives for the friction coefficient and the friction force up, changing the mechanical properties and reactivity of
in atomic-level experiments, respectively. the surfaces.

Fig. 4. The slider geometry can affect the magnitude of the plowing contribution to the tangential force, and hence the value of the friction
coefficient reported for a given material combination (say, diamond on a ductile metal). This raises the question as to whether the ratio of the
tangential to the normal force in scratching is truly reportable as a friction coefficient or should be redefined as a different quantity.
P.J. Blau / Tribology International 34 (2001) 585591 591

for a series of controlled stops. However, the safety and


comfort of the vehicles passengers will be directly
affected by the variations experienced in daily driving,
regardless of the average value obtained in a simple test.
It is therefore important, when testing for friction, to note
not only the average value of the friction force (or fric-
tion coefficient), but the time dependence and stability
of the friction force over a range of contact conditions.

References

[1] da Vinci, L. Notebooks and manuscripts, prepared in the late


1400s, as described in [4].
[2] Amontons, G. De la resistance causee dans les machines, Mem.
lAcad. Roy. A 1699;25782.
[3] Coulomb CA. Theorie des machines simples, en ayant egard au
Fig. 5. Relationship between the friction and wear break-in periods frottement de leurs parties, et la roideur des cordages. Mem Math
for various block materials sliding on 52100 steel rings without lubri- Phys Paris 1785;x:161342.
cation [12]. The processes that are responsible for wear and friction [4] Dowson D. History of tribology. 2nd ed. London: Professional
are not the same, and their complex interplay results in different break- Engineering Publishers, 1998.
in periods. [5] Kotelnikov SK. A book containing instruction on the equilibrium
and movement of bodies. St Petersburg; 1774.
8. Concluding remarks [6] ASTM G-40-99, Standard terminology relating to wear and ero-
sion. Annual book of standards vol. 03.02, 2000;15865.
[7] Blau PJ. Experimental aspects of friction research on the macros-
The friction coefficient is an established, but some- cale. In: Bhushan B, editor. Fundamentals of and bridging the
what misunderstood, quantity in the field of science and gap between macro- and micro-/nanoscale tribology. Dordrecht:
engineering. It is a convenient and useful parameter for Kluwer, 2001.
engineering, but care should be exercised when ascribing [8] Blau PJ. Friction science and technology. New York: Marcel
to it a fundamental significance. For hundreds of years, Dekker, 1986.
[9] Budinski KG. Laboratory testing methods for solid friction. In:
friction coefficients have served many useful purposes, ASM handbook, vol. 18friction, lubrication, and wear tech-
like aiding in the design of machines and buildings, nology. New York: ASM International, 1992:4558.
improving devices for enhanced safety (like brakes, floor [10] Budinski KG, Blau PJ. Friction testing. In: ASM handbook, vol.
waxes, tires, and walkways), and improving industrial 8mechanical testing. New York: ASM International, 2000.
processes. While friction coefficients are relatively easy [11] Brown RL, Richards JC. Principles of powder mechanics.
Oxford: Pergamon Press, 1966.
to determine in laboratory experiments, the fundamental [12] Blau PJ. Four great challenges confronting our understanding and
origins of sliding resistance are not as clear. In fact, modeling of sliding friction. In: Dowson D, Taylor CM, Childs
some of the greatest scientists and philosophers have THC, Dalmaz G, Berthier Y, Flamand L, Georges JM, Lubrecht
contemplated friction without managing to produce a AA, editors. Tribology for energy conservation. Amsterdam:
universal, predictive theory. This striking lack of success Elsevier, 1998:177228.
[13] Blau PJ. Friction science and technology. New York: Marcel
is due to the many potential factors that can influence Dekker, 1986:266.
friction in a wide spectrum of physical situations. [14] Beer FP, Johnston ER Jr. Vector mechanics for engineers. New
Tables of friction coefficients can be useful as long York: McGraw Hill, 1962.
as the conditions used to obtain them are clearly stated [15] Anon. Handbook of chemistry and physics, 48th ed, Table F-14.
and qualified. Static and kinetic friction coefficient data Boca Raton: CRC Press;19678.
[16] Hibbeler RC. Engineering mechanics. 3rd ed. New York: Mac-
listed in tables should not be applied to situations greatly millan, 1983.
different from those used to obtain them. The character- [17] Automotive handbook. Bosch, 1993.
istics of frictional behavior, such as break-in transients [18] Zafiratas CD. Physics. 2nd ed New York: J. Wiley, 1985.
and the stability of the friction force, can be more [19] Giancoli D. Physics. 3rd ed Englewood Cliffs (NJ): Prentice
important for the design and optimization of machines Hall, 1991.
[20] Frautschi SC, Olenick RP, Apostol TM, Goodstein DL. The
than the average value of the friction coefficient alone. mechanical universe. Oxford: Oxford University Press, 1986.
For example, the kinetic friction coefficient for a disc [21] Eschbach OW. Handbook of engineering fundamentals. 2nd ed.
brake system on a certain vehicle might average 0.35 Chichester: John Wiley and Sons, 1952.

Vous aimerez peut-être aussi