Beruflich Dokumente
Kultur Dokumente
Abstract
Friction is a phenomenon involving elastic interactions, plastic deformation and failure processes at different length scales. The friction coefficient
is one of the most relevant parameters for the rolling contact of railway wheels and rails. A model of dry friction is established based on the method
of movable cellular automata (MCA). The influence of pressure and sliding velocity has been investigated by means of a large number of numerical
simulations. A general qualitative conclusion is the fact that dynamic processes of plastic deformation and fracture at the nanolevel are of great
importance. Within this model, the contact temperature can also be considered by its influence on the strength of materials. The new friction law
is applied to the calculation of stresses, deformations and tractive forces in wheel/rail contact with rough surfaces. Thus, the understanding for the
physical background of dry friction in engineering applications can be considerably improved.
2006 Elsevier B.V. All rights reserved.
0043-1648/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.wear.2006.01.046
F. Bucher et al. / Wear 261 (2006) 874884 875
Table 1
Parameters of the model material (for determination of strength parameters see
Fig. 5)
Symbol Value Unit
Fig. 3. Instant pictures of a quasi-liquid layer structure for a hypothetical material of ultimate strength 92 MPa: pressure 1 MPa (a) and 26 MPa (b). The sliding
velocity in both cases is 5 m/s. The zone of the quasi-liquid layer is shown by parentheses.
F. Bucher et al. / Wear 261 (2006) 874884 877
Fig. 4. Friction force vs. pressure p in the grinding-in stage (first 100 ns) and in Fig. 6. Friction coefficient vs. velocity for pressure 15 MPa for a hypothetical
the stationary state. The sliding velocity is 5 m/s. material of an ultimate strength of 92 MPa.
coefficient depends solely on the parameters 1 and 2 . Note approximation parameter values:
that Poissons ratio and breaking strain were kept constant in all
numerical simulations. More generally, the friction coefficient 0 = 0.15, 1 = 0.0442, 2 = 0.3243,
could also be a function of these dimensionless parameters as b = 0.195 and c = 0.00212. (3)
independent variables.
Since we have no phenomenological model defining the form The root mean square difference between calculated points and
of the analytical dependence of the friction coefficient on loading approximating surface is near 9% (Fig. 7).
parameters, we approximated the numerical results given as 34 The numerical results suggest that the friction coefficient
discrete points by the simplest rational function of the form drastically decreases at very low pressures. This dependence
is not physically unexpected: for p 0, the friction coefficient
1 1 must vanish. However, this decrease takes place in a small low-
= 0 + 1 + 2 , (2) pressure range only. This fact was left out in the graphs given in
1 + b1 1 + c2
Fig. 7. The construction of a unified analytical approximation
for the entire pressure range is a complicated mathematical prob-
that qualitatively reflects the main features of the numeri- lem. For this reason, the points belonging to the low-pressure
cal results. The least squares optimization gives the following range were not taken into account in constructing the approxi-
Fig. 7. Friction coefficient vs. dimensionless parameters 1 and 2 : analytical approximation Eq. (2) and calculated points (a) and other view point of the dependence
in the area of small 1 (b). Lines from point centers show location of points: upper direction for points under the surface, bottom direction for points over the surface.
The length of lines correlates with deviation of points (root mean square difference is 9%). The two-dimensional plot (projection) of the surface on the 1 and 2
plane (c).
F. Bucher et al. / Wear 261 (2006) 874884 879
Both the theoretical analysis and direct numerical calcula- 4.4. Temperature dependence of the friction coefcient
tions show that the probability distribution to find a specified
pressure at a contact between two fractal surfaces becomes the The friction coefficient of the system under study does not
more uniform, the lower the scale level [21,23]. Thus, at the depend explicitly on temperature. Temperature can, therefore,
microlevel where the peak pressure achieves hardness of the influence the friction coefficient only by the temperature depen-
F. Bucher et al. / Wear 261 (2006) 874884 881
eters has been kept constant for all simulations and we assumed
an averaged linear relation for the temperature dependence of
all strength parameters. Thus, the ultimate tensile strength as a
function of temperature can be written as
0 () = 1000 MPa 1 (10)
1250
with in K. (For the investigations of this paper, the zero-point
of temperature has been chosen as the ambient temperature.
Therefore, the temperature always represents the temperature
rise due to the heat supply within the contact patch.)
Now, we have to calculate the contact temperature using an
appropriate model of thermal contact. Concerning the surface
temperature in frictional contact, we also can distinguish differ-
ent scale levels between the average temperature occurring in
the whole nominal area of contact and the peak temperatures
in micro-contacts (asperities). The heat conduction in Hertzian
contact with smooth surfaces can be treated one-dimensionally
which makes possible a fast and efficient calculation of contact
temperature [29].
If the roughness of surfaces is taken into account, the inves-
tigation of temperatures becomes a very complicated numerical
problem [3034]. The results depend on the local pressure at
micro-contacts as well as on their size and distribution. On the
other hand, Archard has already stated that the influence of sur-
face roughness can be neglected in a first approximation, since
. . . in general, the largest temperatures are those deduced for
the whole region rather than those deduced for the smaller indi-
vidual contact areas [35]. This has been confirmed recently by
Vick and Furey [36].
Justified by the statement of Archard, we used an approximate
Fig. 12. Distribution function of probability to find a specified pressure at the solution for the average contact temperature based on Hertzian
wheel/rail contact with real (experimentally measured) roughness (data from line contact. It can be written as
[22]). The distribution function depends on the wavelength at which the Fourier
spectrum of the surface profile is cut off in order to solve the elastic contact a
= 0.426p 0 vsliding (11)
problem. Three distributions correspond to the spectrum cutoff at wavelengths cv0
10, 0.1 and 0.025 mm, respectively. The finer the asperities taken into account the
higher the peak pressure values achieved at the contacts, and the more uniform with the average friction coefficient , maximum Hertzian pres-
the probability distribution in the moderate pressure range. The lower picture sure p0 , sliding velocity vsliding , Hertzian semi-length of contact
shows pressure distribution at the submicron level examined in this work. To a
area a, thermal conductivity , density , specific heat capacity
first approximation, the probability distribution is considered to be uniform in
the range from zero to hardness. c and vehicle speed v0 . As shown in [17] and [29], this increase
of the average temperature at the contact surfaces is in the range
of 200 K for usual operating conditions in wheel/rail contact.
dence of mechanical parameters. Let us analyze the averaged It does not depend on the local pressure distribution at micro-
friction coefficient Eq. (9). Due to the averaging, it does not contacts, but only on global quantities. For typical values of
depend on pressure. It is also practically independent of veloc- roughness profile and real area of contact in wheel/rail systems,
ity and depends only on the ratio of strength to Youngs modulus. the local peak temperatures at micro-contacts are only slightly
As long as this ratio depends on temperature, there will also be higher (1030%) than the values calculated with the assumption
an influence of the temperature on the coefficient of friction. The of smooth surfaces [17].
strength parameters of steel decrease significantly at high tem-
peratures (Fig. 11) while temperature dependence of Youngs 4.5. Discussion of results
modulus can be neglected at first glance.
Yield strength and ultimate tensile strength of a wheel steel The tangential contact problem has been solved for the
do not depend equally on temperature (Fig. 11). For the accu- parameters given in Table 2. In these calculations, the sliding
rate determination of the friction coefficient as a function of velocity was prescribed and the tangential force results from
temperature, MCA simulations at different temperatures should the integral of tangential stresses in the area of contact. For the
use the respective values of yield strength and ultimate tensile consideration of contact temperature at high creep values, an iter-
strength. In order to avoid this effort, the ratio of strength param- ation was required due to the feedback between friction law, tan-
882 F. Bucher et al. / Wear 261 (2006) 874884
Table 2
Operating conditions for the wheel/rail contact
Symbol Value Unit
5. Conclusions
level on two dimensionless arguments. These arguments, in turn, cesses are the key to understanding mechanisms of dry friction
depend on the material (density, strength and Youngs modu- in real tribological systems.
lus) and loading (pressure and sliding velocity) parameters. We
have found that the friction coefficient is strongly influenced Acknowledgements
by the dynamic processes in surface nanolayers that are formed
and sustained at the contact region within the entire period of One of the authors (V.L.P.) is grateful to O. Dudko for dis-
relative motion of contacting solids. The fact that the friction cussions and comments. Financial support from the Deutsche
coefficient depends only on two independent arguments 1 and Forschungsgemeinschaft (DFG: German Research Council) and
2 (Eq. (1)) and not on three independent variables, does not the German Academic Exchange Service (DAAD) is acknowl-
follow from the analysis of dimensionality and has come out edged.
as an unexpected empirical fact requiring theoretical interpreta-
tion.
It should be noted that the dependence given in Eq. (2) was References
found for a number of simplifying assumptions. The major
[1] F.P. Bowden, D. Tabor, The Friction and Lubrication of Solids, second ed.,
assumptions have been: Clarendon Press, Oxford, 1954.
[2] E. Rabinowicz, Friction and Wear of Materials, Wiley, New York, London,
Eq. (2) is valid only within 2 2 1800. The lower limit of Sydney, 1965.
this range gives the minimum values of 2 used for deriving [3] K. Kraft, Der Einflu der Fahrgeschwindigkeit auf den Haftwert
zwischen Rad und Schiene, Archiv fur Eisenbahntechnik 22 (1967) 5878.
Eq. (2), and the higher limit determines the studied range
[4] K.L. Johnson, Contact Mechanics, Cambridge University Press, Cam-
of values (note, however, that for 2 = 1800 there was only bridge, 1985.
one calculation; the most completely studied region is for [5] J.J. Kalker, Three-Dimensional Elastic Bodies in Rolling Con-
2 620). tact, Kluwer Academic Publishers, Dordrecht, Boston, London,
Poissons ratio was taken to be 0.3, and breaking strain under 1990.
[6] Z.Y. Shen, J.K. Hedrick, J.A. Elkins, A comparison of alternative creep
active loading 4% (Table 1). These values are typical of many
force models for rail vehicle dynamic analysis, in: J.K. Hedrick (Ed.), The
metals. Dynamics of Vehicles on Roads and Tracks, Proceedings of the eighth
Tribological systems whose linear dimension is much smaller IAVSD Symposium held at MIT, Cambridge, MA, 1519 August 1983,
than the elastic coherence length were examined. This Swets & Zeitlinger, Lisse, 1984, pp. 591605.
assumption is reasonably well fulfilled for contacts of interest [7] C.F. Logston Jr., G.S. Itami, Locomotive friction-creep studies, Trans.
ASME, J. Eng. Ind. 102 (1980) 275281.
with dimension of about 10 mm in the case of metals.
[8] B.R. Meyer, High-adhesion locomotives with controlled creep, IMechE
The contacting materials were assumed to be homogeneous. C81/87 (1987) 209218.
Hence all elements of the medium and their strength and elas- [9] F. Rick, Zur Erfassung der Geschwindigkeitsabhangigkeit des
tic characteristics were identical. Evidently, this assumption Kraftschlubei-wertes eines hochbelasteten Rad-Schiene-Kontaktes,
is not fulfilled for many alloys and steels. We suggest that Ph.D. thesis, Technische Universitat Clausthal, 1998.
[10] V.L. Popov, S.G. Psakhie, Theoretical foundations of simulation of elasto-
due to the presence of a dynamic boundary layer, the effect of
plastic media with the method of movable cellular automata. I. Homoge-
structural heterogeneities is averaged. However, further inves- neous media, Phys. Mesomech. 4 (1) (2001) 1525.
tigation into the subject is required. [11] V.L. Popov, S.G. Psakhie, A. Gerve, B. Kehrwald, E.V. Shilko, A.I.
Dmitriev, Wear in combustion engines: experiment and simulation by the
Another important result is concerned with a macroscopic method of movable cellular automata, Phys. Mesomech. 4 (4) (2001) 7383.
[12] V.L. Popov, S.G. Psakhie, E.V. Shilko, A.I. Dmitriev, K. Knothe, F. Bucher,
friction coefficient observed at the contact of surfaces with a
M. Ertz, Friction coefficient in rail-wheel-contacts as a function of mate-
random (fractal) roughness. We brought forward arguments in rial and loading parameters, Phys. Mesomech. 5 (3) (2002) 1724.
support of the fact that in this case, the probability of finding [13] A. Lunenschlo, F. Bucher, K. Knothe, Normalkontakt zweier Korper
a specified pressure at the contact is (approximately) uniformly mit rauen Oberflachen. VDI Fortschritt-Berichte, Reihe 2, No. 596, VDI-
distributed in the range from zero to hardness of material. Aver- Verlag, Dusseldorf, 2002.
aging the friction coefficient over this distribution gives the fric- [14] A. Lunenschlo, F. Bucher, K. Knothe, Numerische Behandlung des
quasistatischen Tangentialkontakt-Problems zweier Korper mit rauen
tion coefficient depending on the material strength to Youngs Oberflachen. VDI Fortschritt-Berichte, Reihe 2, No. 616, VDI-Verlag,
modulus ratio 0 /E and 1 = v2 E/02 alone. Dusseldorf, 2002.
Due to the influence of material properties on the friction [15] F. Bucher, The contact between micro-rough rails and wheels, Ph.D. thesis,
coefficient, the simulation results are only as good as the knowl- Technische Universitat Berlin, 2002.
[16] M. Ertz, F. Bucher, Improved creep force model for wheel/rail contact
edge of material behaviour. This should be taken into account
considering roughness and temperature, in: H. True (Ed.), The Dynamics
especially in the case of temperature dependence. In any case, of Vehicles on Roads and Tracks, Proceedings of the 17th IAVSD Sympo-
the interaction of a temperature-dependent friction law and the sium held in Lyngby (Copenhagen), Denmark, 2024 August 2001. Vehicle
contact temperature induced by friction is a convincing physical System Dynamics Supplement 37, Swets & Zeitlinger, Lisse, 2002, pp.
explanation for the decrease of creep force curves at high creep 314325.
[17] M. Ertz, Temperatur, Materialbeanspruchung und Kraftschluss im Rad-
values.
Schiene-Kontakt. VDI Fortschritt-Berichte, Reihe 12, No. 549 (also Ph.D.
Undoubtedly, a general qualitative conclusion is the fact that thesis, Technische Universitat Berlin), VDI-Verlag, Dusseldorf, 2003.
dynamic processes of plastic deformation and fracture at the [18] K. Knothe, Non-steady state rolling contact and corrugations, in: B.
nanolevel are of great importance. In our opinion, these pro- Jacobson, J.J. Kalker (Eds.), Rolling Contact Phenomena. CISM Courses
884 F. Bucher et al. / Wear 261 (2006) 874884
and Lectures, No. 411, Chapter 4, Springer Wien, New York, 2000, pp. [28] G. Baumann, Untersuchungen zu Gefugestrukturen und Eigenschaften der,
203276. Weien Schichten auf verriffelten Schienenlaufflachen, Ph.D. thesis,
[19] J.B. Nielsen, Evolution of rail corrugation predicted with a non-linear wear Technische Universitat Berlin, 1998.
model, J. Sound Vib. 227 (5) (1999) 915933. [29] M. Ertz, K. Knothe, A comparison of analytical and numerical methods
[20] V.L. Popov, A.V. Kolubaev, Generation of surface waves during exter- for the calculation of temperatures in wheel/rail contact, Wear 253 (34)
nal friction of elastic solid bodies, Tech. Phys. Lett. 10 (1995) 812 (2002) 498508.
814. [30] J. Gao, S.C. Lee, X. Ai, H. Nixon, An FFT-based transient flash temperature
[21] B.N.J. Persson, F. Bucher, B. Chiaia, Elastic contact between randomly model for general three-dimensional rough surface contacts, Trans. ASME,
rough surfaces: comparison of theory with numerical results, Phys. Rev. B J. Trib. 122 (2000) 519523.
65 (2001) 184106. [31] G. Liu, Q. Wang, S. Liu, A three-dimensional thermalmechanical asperity
[22] F. Bucher, A. Theiler, K. Knothe, Normal and tangential contact prob- contact model for two nominally flat surfaces in contact., Trans. ASME, J.
lem of surfaces with measured roughness, Wear 253 (12) (2002) 204 Trib. 123 (2001) 595602.
218. [32] S. Liu, Q. Wang, A three-dimensional thermomechanical model of contact
[23] B.N.J. Persson, Elastoplastic contact between randomly rough surfaces, between non-conforming rough surfaces, Trans. ASME, J. Trib. 123 (2001)
Phys. Rev. Lett. 87 (2001) 116101. 1726.
[24] B.N.J. Persson, Sliding Friction: Physical Principles and Applications, sec- [33] Y.-C. Liu, H. Wang, W.-Z. Wang, Y.-Z. Hu, D. Zhu, Methods comparison
ond ed., Springer, Berlin, Heidelberg, New York, 2000. in computation of temperature rise on frictional interfaces, Tribol. Int. 35
[25] K. Knothe, A. Theiler, Normal and tangential contact problem with rough (2002) 549560.
surfaces, in: Proceedings of the 2nd Mini Conference on Contact Mechanics [34] X. Tian, F.E. Kennedy Jr., Maximum and average flash temperatures in
and Wear of Rail/Wheel Systems, Budapest, 1996, pp. 3443. sliding contacts, Trans. ASME, J. Trib. 116 (1994) 167174.
[26] K. Knothe, R. Wille, B.W. Zastrau, Advanced contact mechanicsroad [35] J.F. Archard, The temperature of rubbing surfaces, Wear 2 (19581959)
and rail, Vehicle Syst. Dyn. 35 (45) (2001) 361407. 438455.
[27] A.H. Cottrell, The Mechanical Properties of Matter, John Wiley & Sons, [36] B. Vick, M.J. Furey, A basic theoretical study of the temperature rise in
New York, 1964. sliding contact with multiple contacts, Tribol. Int. 34 (2001) 823829.