Professional Documents
Culture Documents
The scope of the series covers the entire spectrum of solid mechanics. Thus it includes
the foundation of mechanics; variational formulations; computational mechanics;
statics, kinematics and dynamics of rigid and elastic bodies: vibrations of solids and
structures; dynamical systems and chaos; the theories of elasticity, plasticity and
viscoelasticity; composite materials; rods, beams, shells and membranes; structural
control and stability; soils, rocks and geomechanics; fracture; tribology; experimental
mechanics; biomechanics and machine design.
The median level of presentation is the first year graduate student. Some texts are mono-
graphs defining the current state of the field; others are accessible to final year under-
graduates; but essentially the emphasis is on readability and clarity.
Edited by
l.A.C. MARTINS
lnstituto Superior Tecnico,
Departamento de Engenharia Civil, Lisbon, Portugal
and
Preface........................................................................................... ix
Measurements of impacts with friction... ............. ....... .... ........ .......... ............. 17
Friedrich Pfeiffer
Collisions in systems made of rigid bodies... .... ........... ...... .... ............ ......... ..... 33
Eric Dimnet
A controllability criterion for linear juggling mechanical systems........ ....... ......... .... 75
Bernard Brogliato, Mongi Mabrouk, Arturo Zavala Rio
Longitudinal waves in elastic rods with discontinuous cross sections.. .. .. .. .. ...... .. .. .. . 117
Werner Schiehlen, Bin Hu, Peter Eberhard
Friction coefficient evolution during wear tests related to infrared emission............... 203
Nadine Stalin, Jean Christophe Eytard
The real area of contact - a combination of experimental and numerical approaches.... 219
Frank Bucher, Rob S. Dwyer-Joyce
vii
Prediction of wheel wear for rail vehicles - methodology and verification.. . . . . . . . . . . . . .. 229
Tomas Jendel. Mats Berg
Multi-body modeling of paper calendering unit by contact dynamics fonnulation.... .... 237
Erno Keskinen. Sirpa Launis. Juha-Matti Kivinen
4. Mathematical analysis
Existence and uniqueness for quasi static contact problems with friction................... 245
Lars-Erik Andersson, Anders Klarbring
Augmented lagrangian methods for a class of non convex contact problems in struc-
tural mechanics...................................................................................... 261
W.R. Bielski, A. Galka, J.J. Telega
Convergence for a time discretization of dynamic contact problems with friction........ 317
E. Pratt, J.-M. Ricaud
5. Numerical methods
Unilateral contact, friction and adhesion: 3D cracks in composite materials............... 333
Michel Raous, Yann Monerie
The unilateral frictional contact of a piezoelectric body with a rigid support.............. 347
Paolo Bisegna, Frederic Lebon, Franco Maceri
viii
Convergence studies for 3D smooth frictional contact elements based on the quartic
Bezier surfaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... . . 371
Lovre Krstulovic-Opara, Peter Wriggers
Multicontact structures and parallel computing........ ................ ......... ....... ......... 379
P. Alart, M. Barboteu
We thank all the authors for their valuable contributions to this volume.
We are indebted to the members of the Scientific Committee for their help in
refereeing the submitted abstracts and manuscripts. We also thank the Series
editor, Prof. Graham Gladwell, for his assistance in the revision process.
We gratefully acknowledge the support of our research centers, the
lCIST (lnstituto Superior Tecnico) and the CMAF (Universidade de Lisboa),
and the support of the institutions listed in page viii. In particular, we thank
ix
x
the Funda9ao Calouste Gulbenkian for partially funding the publication of
this volume.
We address special heartfelt thanks to our colleagues of the Organizing
Committee, Eduardo Borges Pires, Fernando Simoes, Ant6nio Pinto da
Costa and Anca-Maria Toader, for their great effort, efficiency and
enthusiasm in the organization of the Symposium.
CMIS 2001
June 17-21, 2001
Praia da Consola~iio, Peniche, PORTUGAL
1. Organizing Committee
J.A.C. Martins, Dep. Eng. Civil and ICIST, Instituto Superior Tecnico, Lisboa
M.D.P. Monteiro Marques, Fac. Ciencias da Universidade de Lisboa and CMAF, Lisboa
E.B. Pires, Dep. Eng. Civil and ICIST, Instituto Superior Tecnico, Lisboa
F.M.F. Simoes, Dep. Eng. Civil and ICIST, Instituto Superior Tecnico, Lisboa
A. Pinto da Costa, Dep. Eng. Civil and ICIST, Instituto Superior Tecnico, Lisboa
A.-M. Toader, Faculdade de Ciencias da Universidade de Lisboa and CMAF, Lisboa
2. Scientific Committee
J.R. Barber, Michigan, USA A. Klarbring, Linkoping, Sweden
S. Cescotto, Liege, Belgium F. Maceri, Rome, Italy
A. Curnier, Lausanne, Switzerland J.J. Moreau, Montpellier, France
G. Del Piero, Ferrara, Italy F. Pfeiffer, Miinchen, Germany
M. Fremond, Champs sur Marne, France M. Raous, Marseille, France
C. Glocker, Zurich, Switzerland J.R. Rice, Cambridge, USA
J. Haslinger, Praha, Czech Republic D. Stewart, Iowa, USA
M. Jean, Marseille, France W.J. Stronge, Cambridge, United Kingdom
K.L. Johnson, Cambridge, United Kingdom P. Wriggers, Hannover, Germany
J.J. Kalker, Delft, The Netherlands
xi
xii
The Organizing Committee of CMIS 2001 thanks Prof. Joao Bento and
Prof. Eduardo Pereira of the ICIST, for having kindly provided some
database software, and Dr. Cristian Barbarosie of the CMAF, for his help
during the meeting.
The Organizing Committee of CMIS 2001 also thanks the staffs of
Camara Municipal de Peniche, Hotel Atlantico Golfe and ICIST, for their
efficient collaboration.
NUMERICAL DYNAMICS
OF GRANULAR MATERIALS
1. Introduction
1.1. Addressed problems
In diverse domains, computational methods are needed for the statics
or the dynamics of collections of rigid bodies subject to the constraints
of non-interpenetrability, with friction taken into account in the event of
contact. This includes the dynamics of machines, in particular robots,
the dynamics of masonry works submitted to transient actions (earth-
quakes, gusts of wind or impacts), animated computer graphics and
numerical simulation in granular mechanics. Possibly, some variables
are added in order to also account for a certain deformability of the
bodies without essentially changing the computational strategies; see
Jean (1999). Of course, for the handling of non-interpenetrability and
friction, much may be learned from the rich literature devoted to the
treatment of contact between deformable media, discretized through fi-
nite elements; see e.g. Alart and Curnier (1991), Chabrand et al. (1998),
Cristensen et al. (1998), Curnier (1984), Vola et al. (1998).
The main part of the lecture at the Symposium consisted of projected
animations, intended to show what sort of knowledge may be gained
J.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 1-16.
© 2002 Kluwer Academic Publishers.
2
from the numerical simulation of granular processes. The present paper
begins with an overview of the 'Contact Dynamics' time-stepping tech-
nique used in these simulations. More detailed accounts of the method,
originated in Moreau (1988b), may be found in Moreau (2000), Moreau
(2001). Finally, some drawings created in the course of numerical exper-
iments are shown and commented.
where the functions ga(t, q) are measures of the gaps between bodies,
counted as negative if the configuration q at time t involves overlap.
The event of equality in (1) for some value of a, corresponds to contact
between a member B of the system and a body B' which may be another
member of the system or an external obstacle.
Non-interpenetrability is a unilateral constraint; as always in Mechan-
ics, its description cannot reduce to merely specifying the corresponding
set of feasible states (such, in contrast, is the meaning of the word 'con-
straint' in Optimization and similar contexts). Some information must
be added about the mechanical process through which the geometrical
restriction is enforced. For instance the use of servomechanisms to se-
cure (1) should result in other motions than those investigated in what
follows.
Generically, let us conceive a contact law as a relationship involving
the contact force n a experienced by B from B' at the contact point Ma
and the local velocity Ua of B relative to B', relationship which a priori
depends on time and on the configuration attained by the system, say
(2)
For numerical and analytical purposes, the definitions of the elements
n
Ua and a of E3 are extended, at least locally, to configurations with
nonzero ga(t, q). Saying that (2) models a contact phenomenon involves
that this law yields n a = 0 whenever ga(t, q) > o. Of course, more
complex situations may be addressed, for instance with contact actions
not reducing to single forces.
3
1.3. Approximation
Relationships (1) to (5) convey all the retained information about the
mechanical system. For the numerical treatment of evolution problems,
it is easy to conceive time-stepping procedures of approximation to the
differential equation (3), under the kinematical relationship connecting
q and u. Through (4) and (5), the contact laws (2) are entered into the
management of each time-step. The difficulty comes from the necessity
of taking the non-interpenetration inequalities (1) also into account.
In the majority of the techniques described in literature and applied
in commercial software packages, this is achieved through the classical
trick of penalizing inequalities. Mechanically, this means that the strict
non-interpenetrability of a pair of bodies is approximately replaced by
elastic repulsion forces which become effective when the bodies come
close to each other. Such a procedure amounts to approximate the joint
conditions (1) and (2) by relationships connecting the same variables,
but smoother. In principle, the resulting system of differential equations
should be sufficiently regular for standard time-stepping procedures to
apply. Similar interaction laws are used in the computer simulation of
molecular motions, hence the name of Molecular Dynamics (abbr.: MD)
commonly given to this approach.
The drawback is that the need of precision requires of the artificial
repulsion laws to be very stiff. Numerical stability in integration then
calls for the use of very short time-steps and frequently also for the
introduction of some artificial damping or artificial increase of inertia.
When treating dynamical applications, the effect of such alterations of
the mechanical data may blur the picture. Significant simulations of
loose, collisional, flows of granular materials have been obtained in that
5
way, but when dense collections of bodies are concerned (masonry works
or compact granulates) the method is mainly applied to quasi-static
evolutions in which only a succession of equilibrium states is looked for.
For these reasons, other numerical strategies may be preferred in uni-
lateral multibody dynamics.
In the line of the traditional stereodynamic practice are the approaches
qualified as event-driven (abbr: ED). Starting from a state in which
certain contacts are in effect, one attempts to calculate the subsequent
motion under the provisional assumption that these contacts persist and
that their status - sticking or sliding - in regard to Coulomb law is in-
variant. Calculation is then the same as with bilateral constraints, but
requires to watch the evolution of some indicators. If, for instance, the
normal components of some contact forces take directions incompatible
with unilaterality, one concludes that the subsequent motion has to be
calculated otherwise. Finding the further status of contacts without hav-
ing to try all combinations is a nontrivial matter. It is usually reduced
to Linear Complementarity Problems; see Moreau (1966), Pfeiffer and
Glocker (1996), Vola et al. (1998), Abadie (2000).
The ED strategy becomes unpractical if the number of contacts in
presence is too large. Some time-stepping integration procedures with
preselected time-intervals have been developed instead; see Pang and
Stewart (1999), Stewart (1998). The following one belongs to this class.
iterations with zero initial velocity; if the step-end velocity is found zero,
the values obtained for the contact forces agree with equilibrium.
contact and Coulomb friction, without any adhesive effect, makes a suf-
ficiently rich concept for investigating some fundamental questions.
For Civil Engineers, dry sand , to which the above model is relevant,
is only a special instance of soil.
In the numerical treatment of most engineering problems, the soil
material s viewed as a continuous medium, for which some empirical
constitutive law should be available. Finite Element software packages
are then applied to the corresponding boundary value problem. The
difficulty lies in the determination of the constitutive law which above all
has to be tested against the experimental information available about the
concerned soil. Engineering needs may be met through this approach,
but no insight into the intimate granular behaviour is gained.
While the abbreviation FEM (for Finite Element Method) is used to
refer to the above strategy, the representation of a granular sample as
a collection of solids is called a Distinct Element Method (abbr. DEM,
sometimes also read as Discrete Element Method). Numerical simulation
may be conducted thanks to the methods referred to in the foregoing.
Clearly the boundary problems of Civil Engineering cannot directly be
handled that way since the number of grains involved in real soils exceeds
the possibility of computers. The availability of scaling rules, which
would allow one to approximate these boundary problems through DEM
with oversized grains and possibly with more elaborate interaction laws
than Coulomb contact, is still an object of investigation; cf. Nouguier
et al. (2000).
-
.;§~:::::. .-
~~~~m~·:g~~l~~,~;,. The deposited layer nearly at rest
with conjugate
directions
References
Abadie, M. (2000) Dynamic simulation ofrigid bodies: modelling of frictional contact,
in Impacts in Mechanical Systems. Analysis and Modelling, edited by B. Brogliato,
Springer-Verlag, Berlin Heidelberg, 61-144.
Alart, P., Curnier A. (1991) A mixed formulation for frictional contact problems prone
to Newton like methods, Comput. Meth. in Appl. Mech. Engng. 92, 353-375.
Brogliato, B. (1999) Nonsmooth Mechanics, 2d. edition. Springer-Verlag, London.
Brogliato, B., ten Dam, A. A., Paoli, L., Genot, F. and Abadie, M. (2001) Numerical
simulation of finite dimensional multibody nonsmooth mechanical systems, ASME
Applied Mechanics Reviews, to appear.
Chabrand, P., Dubois, F., and Raous, M. (1998) Various numerical methods for solv-
ing unilateral contact problems with friction, Mathl. Comput. Modelling 28, 97-
108.
Christensen, P. W., Klarbring, A., Pang J. S., and Stromberg N. (1998) Formulation
and comparison of algorithms for frictional contact problems, Int. J. Num. Meth.
Engng. 42,145-173.
Curnier, A. (1984) A theory of friction, Int. J. Solids Struct. 20, 637-647.
Daudon, D., Lanier, J., and Jean, M. (1997) A micromechanical comparison between
experimental results and numerical simulation of a biaxial 2D granular material,
in Powders and Grains 97, edited by R. P. Behringer and J. T. Jenkins, Balkema,
Rotterdam, 219-222.
Fremond, M. (1995) Rigid body collisions, Physics Letters A 204, 33-41.
Genot, F., and Brogliato, B. (1999) New results on Painleve paradoxes, European
Journal of Mechanics, A/Solids, 18, 653-677.
Goddard, J. D. (1998) Continuum modelling of granular assemblies, in Physics of Dry
Granular Materials, edited by H. J. Herrmann et al., Kluwer, Dordrecht Boston
London, 1-24.
Ivanov, A. P. (1997) The problem of constrained impact, J. Appl. Math. Mech. 61,
341-353.
Jean, M. (1999) The Non Smooth Contact Dynamics method, in Computational Mod-
eling of Contact and Friction, edited by J. A. C. Martins and A. Klarbring, special
issue of Computer Meth. in Appl. Mech. and Engng. 177, 235-257.
Jean, M. (2001) Simulation numerique discrete de materiaux granulaires, in Mi-
cromecanique des materiaux granulaires, edited by B. Cambou and M. Jean, Her-
mes, Paris.
16
Johansson, L., and Klarbring, A. (2000) Study of frictional impact using a nonsmooth
equation solver, ASME J. Appl. Mech. 67, 267-273.
Jourdan, F., Alart, P., and Jean, M. (1998) A Gauss-Seidel-like algorithm to solve
frictional contact problems. Computer Meth. Appl. Mech. Engng. 155, 31-47.
Kunze, M., and Monteiro Marques, M. D. P. (2000) An introduction to Moreau's
sweeping process, in Impacts in Mechanical Systems. Analysis and Modelling, edited
by B. Brogliato, Springer-Verlag, Berlin Heidelberg, 1-60.
Moreau, J. J. (1966) Quadratic programming in mechanics: dynamics of one-sided
constraints, SIAM J. Control 4, 153-158.
Moreau, J. J. (1988a) Bounded variation in time, in Topics in Nonsmooth Mechanics,
edited by J. J. Moreau, P. D. Panagiotopoulos, and G. Strang, Birkhiiuser, Basel
Boston Berlin, 1-74.
Moreau, J. J. (1988b) Unilateral contact and dry friction in finite freedom dynam-
ics, in Nonsmooth Mechanics and Applications, edited by J. J. Moreau and P. D.
Panagiotopoulos, CISM Courses and Lectures, Vol. 302. Springer-Verlag, Wien
New York, 1-82.
Moreau, J. J. (1989) An expression of classical dynamics, Ann. Inst. H. Poincare Anal.
Non Lineaire, 6 (suppl.), 1-48. Volume also available as Analyse Non Lineaire,
edited by H. Attouch, J.-P. Aubin, F. Clarke, and I. Ekeland, Gauthier-Villars,
Paris.
Moreau J. J., (1997) Numerical investigation of shear zones in granular materials,
in Proc. HLRZ- Workshop on Friction, Arching, Contact Dynamics, edited by P.
Grassberger and D. Wolf, World Scientific, Singapore, 233-247.
Moreau, J. J. (1999) Some basics of unilateral dynamics, in Unilateral Multibody Con-
tacts, edited by F. Pfeiffer and Ch. Glocker, Kluwer, Dordrecht/Boston/London,
1-14.
Moreau J. J. (2000) Contact et frottement en dynamique des systemes de corps rigides,
Rev. Europ. des Elements Finis 9, 9-28.
Moreau J. J. (2001) An introduction to unilateral dynamics, in Novel approaches
in Civil Engineering, edited by M. Fremond and F. Maceri, Springer-Verlag, to
appear.
Nouguier, C., Bohatier, C., Moreau, J. J., and Radjai, F. (2000) Force fluctuations in
a pushed granular material, Granular matter 2, 171-178.
Pang, J. S., and Stewart, D. E. (1999) A unified approach to discrete frictional contact
problems, Int. J. Engng. Sci., 37, 1747-1768.
Pfeiffer F., and Glocker Ch. (1996) Multibody Dynamics with Unilateral Contacts,
John Wiley and Sons, New York.
Stewart, D. E. (1998) Convergence of a time-stepping scheme for rigid body dynamics
and resolution of Painleve's problem, Arch. Rational Mech. Anal. 145, 215-260.
Stoianovici, S. P., and Hurmuzlu, Y. (1996) A critical study of the concepts of rigid
body collision theory, J. Appl. Mech. 63, 307-316.
Vola, D., Pratt, E., Jean, M., and Raous, M. (1998) Consistent time discretization
for a dynamical frictional contact problem and complementarity techniques, Rev.
Europ. des Elements Finis 7, 149-162.
MEASUREMENTS OF IMPACTS
WITH FRICTION
Friedrich Pfeiffer
Lehrstuhl fUr Angewandte Mechanik, Technische Universitiit Miinchen
D-85747 Garching, Germany
Abstract Existing models of impacts with friction consider compression and ex-
pansion phases, where during compression impulses are stored, and
where during expansion they are released with frictional losses, both
in normal and tangential contact directions. Measurements performed
with a specially designed ballistic machine verify these theories and give
an indication for improvements.
1. Introduction
Impacts with friction play an important role in machines and mecha-
nisms. Therefore many efforts have been made to establish good models,
where the most realistic ones are based on Moreau's ideas (1986, 1988),
continued and extended by Glocker (1995). The theory takes into ac-
count compression and expansion phases during the impact and con-
siders storing and restoring energy in normal and tangential directions.
Friction is included by the corresponding complementarities, restoring is
governed by Poisson's friction law. The theory has been applied success-
fully to many industrial problems, but in the meantime it has been also
systematically verified by laboratory tests. For this purpose a machine
was designed, where a rotating mechanism on a rotating arm releases a
disc within milliseconds, at the same time controlling nearly exactly the
translational and the rotational speeds of the disc. These speeds can be
pre-programmed. The ballistic flight of the disc, the contact with the
ground and the disc-ground behavior are tracked stroboscopically and
by a high-speed camera. Special marks assure a precise evaluation of
the experiments. More than 600 experiments have been carried through
with various material pairings. The outcome confirms the theory excel-
lently, and, moreover, it has induced a small correction of the corner law
during expansion.
17
l.A. C. Martins and M.D.P. Monteiro Marques (eds.). Contact Mechanics. 17-24.
© 2002 Kluwer Academic Publishers.
18
First ideas for impact modeling came from Newton (1687) and Poisson
(1833). The interest in problems of that kind arose anew in the last fif-
teen years when Moreau (1986, 1988) and Panagiotopoulos (1985, 1993)
started to consider mechanical systems with unilateral properties first
more in a statical sense, then more and more also dynamically. Both
scientists have the merit to have created and established this area with
great impact to modern mechanical sciences. A very good review of
impact dynamics is presented by Brogliato (1996). Murty (1988) gives
a collection of practical rules to deal with the numerical problems of
complementarity and, more generally, of nonlinear optimization.
The author's own institute develops research in that area for sixteen
years, very much related though to applications of practical relevance. A
survey of these activities is given by Pfeiffer, Glocker (1996). In partic-
ular an impact theory including frictional effects has been developed by
Glocker (1995) and experimentally confirmed by Beitelschmidt (1998).
The paper at hand is mainly based on these findings.
momentum drive
release unit
release unit
As a result, the machine of Fig. 1 was designed and built, which met
all requirements. A release unit containing the disc is mounted at the
end of a rotating arm with mass balance. The unit itself drives the disc
giving it a prescribed rotational velocity. Main drive and momentum
drive are decoupled allowing to control the two speeds independently.
The rotation of the arm mainly generates a translation, the rotation
of the release unit a rotation of the disc. The flight of the body is
photographed under stroboscopic exposure in a dark room before and
after hitting his target. From the evaluation of the photographs one can
20
calculate the velocities and the position of the body immediately before
and after the impact.
Figure 2 depicts the structure of the test set-up. A computer performs
all control calculations, processes sensor data, evaluates control torques,
releases stroboscope and camera and records all measured data. Within
this overall structure we find for each drive an individual control con-
cept, which has thoroughly been optimized with regard to the above
requirements (Beitelschmidt (1998)).
flash sensor
camera
AT-BUS card
Data Input
Throw Planning
Initialization Hardware
Trajectory Planning
6 Sensor Acquisition
'"'" Drive Control
8
e
p.., Release Disc
Release Flash, Camera
Storage of Results
Post-processing
All computer codes have been realized in C++, which was feasible
due to the fact that the PC-Mode activities are not critical with respect
to time.
21
3. Results
The evaluation of the measurements as recorded by the camera and
the processor was straightforward. Figure 3 illustrates the method and
shows additionally two photographs of experiments. Especially the rub-
ber disc experiment shows nicely a reversal of the trajectory due to the
disc's rotation. The experimental process provided thus a very precise
and well reproducible basis for determining the properties of impacts
with friction.
a)
b)
c)
I§I .. ~/~"
0.5
.
J"]ZI'-"
#,.,r
~'
Ai...."
o • .- -"
-0.5 ,,.'N.
-"til'
,,~
...., ....
~~
-1
/~ pvc
4.
I--+--i
Theory ---------.
-1.5 -1 -0.5 o 0.5
y
Figure 4. Dimensionless tangential relative velocity
after vs. before the impact, PVC-body
The experiments are marked by crosses, the dotted line shows the the-
oretical result. For small tangential relative velocities before the impact,
sticking occurs, and the rolling constraint between disc and ground is
fulfilled after the impact. If the relative velocity is big enough, the body
slides throughout the impact and has a redcuced tangential relative ve-
locity at the end of the impact. At this impact no tangential reversion
occurs.
23
2.5
1.5
w
;t::
•
0.5
0
/o-f--o
-0.5
-8 -6 -4 -2 o 2
y
Figure 5. Dimensionless tangential relative velocity
after vs. before the impact, rubber-body
References
Beitelschmidt, M. (1998) Reibstope in Mehrkorpersystemen, Dissertation TU-Miinchen.
Brogliato, B. (1996) Nonsmooth Impact Dynamics, Springer, London.
Glocker, Ch. (1995) Dynamik von Starrkorpersystemen mit Reibung und StofJen.
Fortschr.-Ber. VDI. Reihe 18, Nr. 182, VDI-Verlag, Diisseldorf,
Moreau, J.J. (1986) Une formulation du contact a frottement sec; application au calcul
numerique, Technical Report 13, C.R. Acad. Sci. Paris, Serie II.
Moreau, J.J. (1988) Unilateral Contact and Dry Jilriction in Finite Freedom Dynam-
ics, Non-Smooth Mechanics and Applications, CISM Courses and Lectures, 302,
Springer Verlag, Wien.
Murty, K.G. (1988) Linear Complementarity, Linear and Nonlinear Programming,
Sigma Series in Applied Mechanics (ed. White, D.J.), Heldermann Verlag, Berlin.
Newton, I. (1687) Principia, Corol.
Panagiotopoulos, P.D. (1985) Inequality Problems in Mechanics and Applications,
Birkhiiuser, Boston, Basel, Stuttgart.
Panagiotopoulos, P.D. (1993) Hemivariational Inequalities, Springer Verlag, Berlin,
Heidelberg.
Pfeiffer, F., and Glocker, Ch. (1996) Multibody Dynamics with Unilateral Contacts,
Wiley & Sons, New York.
Pfeiffer, F. (1996) Assembly processes with robotic systems, Robotics and Autonomous
Systems 19, 151-166.
Pfeiffer, F., Fritz, P. and Srnik, J. (1997) Nonlinear Vibrations of Chains, J. of Vi-
bration and Control 3, 397-410.
Pfeiffer, F., and Stiegelmeyr, A. (1997) Damping Towerlike Structures by Dry Jilriction,
Proc. of DETC '97, ASME Design Eng. Techn. Conf.
Poisson, S.D. (1833) 1hliU de mechanique, Bachelier, Paris.
Srnik, J., and Pfeiffer, F. (1997) Dynamics of CVT Chain Drives: Mechanical Model
and Verification, Proc. of DETC '97, AS ME Design Eng. Techn. Conf.
FORMULATION AND WELL-POSEDNESS
OF UNILATERAL MULTIBODY
DYNAMICS
Patrick Ballard
Labomtoire de Mecanique des Solides - Ecole Polytechnique
91128 Palaiseau Cedex - Jihlnce
Abstract The classical theory of rigid bodies systems undergoing perfect bilateral
constraints has received firm mathematical foundations for a long time.
We extend it to the case where there are perfect unilateral constraints
also. The formulation follows the line initiated by Schatzman (1978) and
Moreau (1983). We give a parametrization-free formulation and try to
identify the most general form of impact constitutive equation which is
compatible with weU-posedness. Then, weU-posedness is proved under
the assumption of analyticity of the data, since it is known, thanks to
Bressan (1960) and Schatzman (1978), that uniqueness does not hold
in the Coo setting.
where D/dt denotes the covariant derivative along the curve q(t) associ-
ated with the Levi-Civita connection, and ~ the isomorphism from TQ
onto T*Q canonically associated with the Riemannian metric. For any
virtual velocity (q, v) ofthe system, the real number (~Dq/dt, v}q is nat-
urally called the virtual power of inertial forces in the virtual velocity
(q, v). Given any local parametrization of the system (local chart), it
admits the well-known representation:
Next, we are given an initial state (qO, vol E TQ. Then, the evolution
problem associated with the dynamics of rigid bodies system with perfect
bilateral constraints is the Cauchy problem:
Problem I. Find T > 0 and q E 0 2 {[O, T[; Q) such that:
• (q(O), q(O)) = (qO, vol,
• \:ItE[O,T[, b~q(t)=f(q(t),q(t),t).
Well-posedness is ensured by the following.
Theorem 2 (Cauchy) The Riemannian configuration manifold is as-
sumed to be of class 0 2 and the mapping f : TQ x R+ -+ T*Q is of class
0 1 . Then, there exists a unique maximal solution for problem 1.
More precisely, theorem 2 states that there exists Tm > 0 (Tm E
R+ U {+oo}) and qm E 0 2 ([0, Tm [, Q) being a solution of problem I such
that any other solution of problem I is a restriction of qm. Of course, we
expect that Tm = +00, in which case the dynamics is said to be eternal.
This situation cannot be taken for granted, in general. In the usual cases
encountered in mechanics, eternal dynamics is ensured by the following
general sufficient condition.
Theorem 3 The configuration manifold Q is assumed to be a complete
Riemannian manifold. The effort mapping f is supposed to admit the
following estimate:
for all (q, v) E TQ and almost all t E [to, +00[, where d{·,·) is the
Riemannian distance and l{t), a (necessarily nonnegative) function of
L}oc{lR; R).
Then, the dynamics is eternal: Tm = +00.
With this convention, the final form of the equation of motion is:
R = bDq+ - f(q(t), q+(t)j t) dt E -N*(q(t)) (5)
It is classical that this equation of motion has to be completed with
some constitutive impact equation describing the outcome of any impact
occurrence. We shall assume an impact constitutive equation of general
form:
'v't, q+ (t) = F (q(t), q- (t» . (6)
To ensure compatibility with the equation of motion (5), the mapping
F should satisfy:
F(q,v-) E V(q),
F (q, v-) - v- E -N(q). (7)
30
Moreover, we add the assumption that the kinetic energy of the system
cannot increase during an impact:
• (q(O),q+(O)) = (qO,vo),
• 'v't E [0, T[, q(t) E A,
• R ~ ~Dq+ - f(q(t), q+(t); t) dt E -N*(q(t)),
• 'v'tE]O,T[, q+(t)=F(q(t),q-(t)).
Corollary 5 There exists an analytic solution (Ta, qa) for problem II.
ql[O,To[ = qal[O,To[·
In other terms, there is local uniqueness for problem II.
It was noticed above that the analytical solution for problem II fur-
nished by corollary 5 fails to exist at the first instant of impact. To
overcome this fact, we have proved that local uniqueness still holds in
the wider class of motion MMA which allows impacts. However, the
maximal solution for problem II may stop to exist at finite time for un-
physical reasons. In other terms, we still do not know if the class MMA
is wide enough. Actually, it is wide enough as shown by the following
theorem which should be brought aside theorem 3.
32
for all (q, v) E TQ and almost all t E [0, +00[, where d(·,·) is the
Riemannian distance and l(t), a (necessarily nonnegative) function of
L}oc{1R; JR).
Then, the dynamics is eternal, that is, the maximal solution for prob-
lem II is defined on [0, +00[.
References
Ballard, P. (2000) The dynamics of discrete mechanical systems with perfect unilateral
constraints, Archive for Rational Mechanics and Analysis 154, 199 - 274.
Ballard, P. (2001) Formulation and well-posedness of the dynamics of rigid bodies
systems with perfect unilateral constraints, Philosophical 7ransactions of the Royal
Society Serie A (to apppear).
Bressan, A. (1960), Incompatibilita. dei teoremi di esistenza e di unicita del moto per
un tipo molto comune e regolare di sistemi meccanici, Annali della Scuola Normale
Superiore di Pisa Serie III, Vol. XIV, 333 - 348.
Lotstedt, P. (1982), Mechanical systems of rigid bodies subject to unilateral con-
straints SIAM J. Appl. Math. 42, no 2, pp 281 - 296.
Moreau, J.J. (1983) Standard inelastic shocks and the dynamics of unilateral con-
straints, in Unilateral problems in structural analysis (G. Del Piero and F. Macari
Eds), Springer-Verlag, Wien, New-York, 173 - 221.
Percivale, D. (1985) Uniqueness in the elastic bounce problem, I, Journal of Differ-
ential Equations 56, 206 - 215.
Schatzman, M. (1978) A class of nonlinear differential equations of second order in
time, Nonlinear Analysis, Theory, Methods (1 Applications 2, No 2, 355-373.
Schatzman, M. (1998) Uniqueness and continuous dependence on data for one dimen-
sional impact problems, Mathematical and Oomputational Modelling 28, No. 4-8,
1-18.
COLLISIONS IN SYSTEMS MADE OF
RIGID BODIES
Eric Dimnet
Laboratoire Lagrange, Laboratoire Central des Ponts et Chaussees
58, boulevard Lefebvre, 75732 Paris cedex 15.
Abstract We describe the collision of a point with a fixed plane and the simul-
taneous collisions of a collection of rigid bodies following the theory of
collisions by M. Fremond. We then give some existence and uniqueness
results for simultaneous collisions of a collection of rigid bodies, and
we show some numerical simulations of the evolutions of collections of
solids during which such collisions happen.
1. Introduction.
In this paper, we discuss instantaneous collisions involving rigid and
deformable solids by describing the interior percussions of the system
made of the colliding bodies. The system made of all the colliding el-
ements is a deformable system: its form changes, even if it is made of
rigid elements! If the duration of a collision is small compared to the
duration of the evolution, we assume that the collision is instantaneous;
thus the velocities are discontinuous.
33
JAC. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 33-43.
@ 2002 Kluwer Academic Publishers.
34
piD~x,t)
x(t)
Figure 1. Collision of a point with a rigid fixed plan and sliding after the collision.
J/
wint(tl,t2' V) = - 12 Rint(T). V(T)dT
(1)
_pint-(t) . V-(t) - pint+(t) . r(t)
35
where Rint (r) is the contact force between the point and the plane. Both
quantities, pint-(t) and pint+(t) are percussions.
The virtual work of the acceleration forces is
+m(U+(t) -
~ -+
U-(t)). V (t)!V (t).
(2)
Let us note that the virtual work of the acceleration forces is such that
the actual work is the variation of the kinetic energy between the times
tl and t2. We suppose that there are no exterior forces applied to the
system. The equations of motion result from the principle of virtual
work
\::Itl, \::ft2, \::IV, wacc(tl, t2, V) = wint(tl, t2, V) (3)
They are
m dU
dr
=_Rint ' (4)
almost everywhere and
W,\::Iw,
m(U+ - U-). y++y- + I(n+ - n-). w++w-
_ 2_ _ _ 2 (7)
= _pint. D(V+,w+,A)+D(V-,w-,A»),
.... .... .... 2 r;-t
D(V,w,A) = V +w X GA.
36
Figure 2. Collision of a rigid body with a rigid fixed plan. The contact point is A.
(17)
for a point. In the case of the collision of a point with a plane (figure 1),
one can easily show, cf Dimnet (2001), that the linear percussion leads
38
-c
..
.u X10-a
.0
:+
:u ..
,.
I .~ : .:~ \- •. :
::.,t+.: ' •~
.~--------~--~--~---~
3 ....
0.1
":1
.:
.. ..
. !t.:
...
0.1
o. . :
:
0.5
00 •
.
.. ";41 ...
:
0.1
,
.• ----.,----.J
O·~~---oC_U'---_~'------c....'=-.--;,----"""'U-~--,,7- •.•
This constitutive law, called the Coulomb law, does not derive from
a pseudo-potential.
When more than two solids collide together at the same time, the
relative velocities of all the couples of colliding solids are discontinuous, it
is a simultaneous collision. This situation happens very often in nature or
40
in numerical simulations of the evolution of collections of solids. Figures
5 and 6 show examples of such simulations.
which take into account the impenetrability condition, the principle (20)
becomes
In order to use law (16), we assume that there exist a normal vector
directed from solid j toward solid i with i < j.
This implies that one of the solids has to be regular at the contact
point but. not both. The more general situation where irregular solids
collide afangular vertices is studied in Dimnet and Fremond (1999a),
Fremond(2000) and Fremond (2001).
The function
(23)
(24)
or by letting
0++0-
X= 2 ' (28)
(30)
where the vector PN(X) contains the actual (but unknown!) values of
the quantities /-LPN which intervenes in the tangential constitutive law
(18). One can try to solve this problem by an iterative method
+ ~(PN(Xn), Y) -
X n +1 = Arglnf {
y2
Iy E R 6N
(20-, Y) }
.
(32)
References
C. Cholet, (1998a), Chocs de solides rigides, these de I'Universite Pierre et Marie
Curie, Paris.
E. Dimnet, M. Fremond, (1999a), Chocs de solides rigides, 4e Colloque national de
calcul des structures, Giens.
E. Dimnet, M. Fremond, (1999b),Choc instantane d'un solide deformable sur un sup-
port rigide, 14e congres franc;ais de mecanique, Toulouse.
43
Christoph Glocker
IMES - Center of Mechanics, ETH Zentrum
CH-8092 Zurich, Switzerland
1. Introduction
Newton's classical kinematic impact law provides a way to calculate the post-
impact velocities of two colliding particles. It reverses the sign of the relative
velocity 'Y at the impact and takes into account dissipation by a coefficient of
restitution € such that 'Y+ = -€ 'Y-. When this impact law is applied to multi
contact problems, a formulation in terms of inequalities is required, such as done
by Pfeiffer and Glocker (1996). Here we assume in addition that"the coefficients
of restitution of the different contacts are all equal to each other. Such an event
is then called a Newtonian impact with global dissipation index, and it agrees
with Moreau's non-smooth dynamical equations, see e.g. Moreau (1988), when
the latter are evaluated at a single instant of time. We present a geometric
interpretation of this class of impacts which is based on the decomposition of
the pre-impact velocity with respect to a pair of orthogonal convex cones. Most
of these results are already found in Moreau (1988). We have put them together
to find a natural generalization to the non-convex case of re-entrant comers
which is addressed at the end of the paper. The resulting formulation goes far
beyond simple unilaterally constrained motion and includes, for example, the
situation of two rectangular blorks hitting each other at their comers.
45
J.A.c. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 45-52.
© 2002 Kluwer Academic Publishers.
46
2. Cones and Convexification
In this section some basic definitions, notations and properties of cones used
in non-smooth analysis are put together. These cones characterize sets in the
neighborhood of a chosen point and indicate whether convexification, as the
next step available after linearization, is possible or not. As a standard tool for
inequality systems, these cones will be used throughout the paper. The material
is mainly taken from the books by Aubin and Ekeland (1984) and Rockafellar
(1972), but applied to finite-dimensional manifolds.
Let M be a proper n-dimensional Riemannian manifold. We denote the
tangent space to M at q E M by TqM, the cotangent space by r: M, and the
duality pairing between r:M and TqMby (', '}q. The inner product and the
norm on TqM induced by the metric on M is denoted by (', ')q and 1I·lI q , and
the corresponding expressions on r: M by (', .); and 11·11;, respectively. We
will also need the natural vector bundle isomorphisms ~ : T M -+ T* M and
U= ~-1.
Let V denote the set of admissible points on M which is assumed to be a
closed subset of M with piecewise smooth boundary avo
In order to charac-
terize the behavior of av in some neighbourhood U of a point q E V we adapt
the definitions of the contingent and the tangent cone in Aubin and Ekeland
(1984) to manifolds. We denote the contingent cone to the set Vat the point
q E V by KqV, and the tangent cone by CqV, respectively. Then
Finally let us set up a pair of closed convex cones (CqV, ctV) orthogonal
to each other via the variational inequality
CfV := {v E TqM I (v, u}q :::; 0 \lu E CqV}. (2)
For convenience, we denote the elements of ctv by vl. when v are the corre-
sponding elements of CqV. One observes that CqV = TqM {:} C;V = {O}
and vice versa. If CqV is a half-space then CiV degenerates to a ray orthogo-
nal to the boundary of Vat q. If CqV is a subspace of TqM, then ctv is the
classical othogonal vector space complement in TqM. According to (2), the
elements of (CqV and ctV) are characterized by the inequality
(3)
Figure I shows the cones Kq V, CqV and ctv for the situations of a tangentially
regular set and a re-entrant comer. More examples on how these cones look
like may be found in (Glocker, 2001)
3. Orthogonal Decompositions
In this section we review briefly the decomposition of a vector with respect to
a pair of orthogonal convex cones and discuss some other related formulations.
The following statement is contained as a special case in Moreau's Theorem,
see e.g. Rockafellar (1972) for the full version and the proof, and generalizes
the classical orthogonal vector decomposition.
Theorem 3.1. Let U be a finite-dimensional real inner product space, and
(R, Rl. ) an orthogonal pair of closed convex cones in U. Any u E U can then
be decomposed uniquely into a sum u = v + vl. such that
(4)
48
Figure 2. Orthogonal cones R, RJ. and the orthogonal tangent cone of Rat (.), Cf)R.
This decomposition is depicted in the left part of Figure 2. For u ¢ R URi. one
obtains v =f. 0, vi. =f. 0, whereas vi. = 0 or v = 0 as soon as u E R or u E Ri..
Apparently is v the nearest point to u in the set R, which is usually denoted by
v = ProxR(u). Thecorrespondingmapu -t proxR(u) is called aproximation
which is, in fact, a projection because proxMu) = proxR(u). In addition, this
projection is orthogonal since (proxR(u) , (u - ProxR(u))) = O. Of course,
the same properties apply for the second term vi. in the decomposition due to
symmetry, i.e. vi. = proxRJ.(u) .
Proposition 3.2. Let U, Rand Ri. be defined as in Theorem 3.1. Then (4) is
equivalent to any of the two conditions
(5a)
(5b)
where C~ Rand C;J.Ri. denote the cones orthogonal to the tangent cones to
R at v and Ri. at vi., respectively.
The proof may be found in (Glocker, 2(01). The cone C~ R is depicted in the
right part of Figure 2 for some elements v E R.
Here, ~u(t) E Tq(t)M denotes the covariant derivative ofu along the curve
q(t) with velocity q(t) = u(t) E Tq(t)M , and f E T;(t)M are the applied
49
forces. Note that Tq(t)M = Cq(t) V holds in this particular situation, because
q{t) E int V by assumption.
Suppose now that q{t) from (6) reaches the boundary of V. In order to
ensure that q{t) does not leave V, a discontinuity in the velocity u{t) has to be
permitted. To include such events in the formulation of the dynamics, equation
(6) is rewritten as an equality of measures (Moreau, 1988)
with u+(t) and u-{t) the right and the left limit of u{t), and R{t) the impact
percussion. In accordance with perfect constraints, the most natural cone of
possible reactions and percussions is ~q(t) C~t)V, hence we claim
(9)
which incorporates already one part of the impact law: For q{ t) E int V one has
C~t)V = {O},henceR(t) = Oandu+(t) = u-(t) by (8), i.e. no impact. Note,
however, that (9) together with (8) is still insufficient to determine uniquely
a post-impact velocity u+{t) or, in other words, to single out one particular
element -~q(t) R(t) from C~t)V. The missing information will be added in
the next section by means of constitutive equations, called the impact law.
(10)
(11)
50
This decomposition is unique. The tenu v plays the role of the tangential com-
ponent of u- which remains unchanged by the impact. The normal component
vol is "inverted" by the impact rule V X := -EVol, where E is the global coeffi-
cient of restitution, a magnitude expressing the global dissipation behaviour of
the impact when chosen between 0 and 1. The post-impact velocity is then set
to be u+ := v + V X which yields
according to the construction shown in Figure 3, see e.g. also Brogliato (1999).
By taking the difference of (12) and (10) the impact percussion R in (8) is
identified as
(13)
From Figure 4 one recognizes that the impact law is always energetically con-
sistent: The kinetic energy T satisfies 2 T+ = lIu+ II~ ~ lIu-ll~ = 2 T-,
where equality holds for E = 1 and maximal dissipation is achieved for E = O.
In tenus of a minimization problem v and vol are the nearest points to u-
in the sets CqD and CiD, respectively. The corresponding maps are called
proximations and are denoted by
(14)
For example, the impact law (10)-(12) might equivalently be stated in tenus of
proximations as
(15)
51
when the first equation in (14) is used. Further, we recognize that the proxima-
tion in (15) becomes the identity whenever u- E Cq'D. In this case u+ == u-,
thus no impact occurs.
7. Conclusion
In this paper a geometric interpretation of two extensions of Newton's kine-
matic impact law were presented: Moreau's multi-contact impact law for fric-
tionless unilateral constraints, and a version which is capable of handling even
the reentrant corner problem. Although clearly originating in the collision prob-
lem of rigid bodies under simple unilateral constraints, these methods are useful
for many impact problems in general finite-degree-of-freedom dynamics.
References
Aubin, J., and Ekeland, I. (1984) Applied Nonlinear Analysis, John Wiley & Sons, New York.
Ballard, P. (2000) The dynamics of discrete mechanical systems with perfect unilateral con-
straints, Arch. Rational Mech. Anal. 154, 199-274.
Brogliato, B. (1999) Nonsmooth Mechanics, 2nd ed., Communications and Control Engineering
Series, Springer, London.
FIimond, M. (1995) Rigid bodies collisions, Physics Letters A 204, 33-41.
Glocker, Ch. (2001) A geometric interpretation of Newtonian impacts with global dissipation
index, Cahiers Stephanois de Mathematiques Appliquees Publication de l'Equipe d'Analyse
Numerique 2, UPRES EA 3058, Universite Jean Monnet, Saint-Etienne, online publication
at http://wwwean.univ-st-etienne.fr. 1-14.
Kane, c., Repetto, E.A., Ortiz, M., and Marsden, J .E. (1999) Finite element analysis of nonsmooth
contact, Compo Methods Appl. Meck Eng. 180, 1-26.
Moreau, J. J. (1988) Unilateral contact and dry friction in finite freedom dynamics, in Non-
Smooth Mechanics and Applications, edited by J. J. Moreau and P. D. Panagiotopoulos,
CISM Courses and Lectures 302, Springer, Wien, 1-82.
Pfeiffer, F., and Glocker, Ch. (1996) Multibody Dynamics with Unilateral Contacts, John Wiley
& Sons, New York.
Rockafellar, R. T. (1972) Convex Analysis, Princeton Univ. Press, Princeton, New Jersey.
UNDERSTANDING IMPACT THROUGH
CONTINUOUS MEDIUM VIBRATIONS
Lretitia Paoli
UMR CNRS MAPLY and Equipe d'Analyse Numerique (EA9058) Faculte des Sciences,
Universite Jean Monnet, 42029 St-Etienne Cedex 2 France
Michelle Schatzman
UMR CNRS MAPLY
Universite Claude Bernard - Lyon 1, 69622 Villeuroonne Cedex France
Abstract We write a model for the impact of a slender bar making an angle
(J with a rigid foundation, starting from first principles. We obtain
some explicit solutions, with the help of asymptotics and simplifying
assumptions. We compare the corresponding results to the experiments
of Hurmuzlu and Stoianovici, and we obtain good qualitative agreement
in all the cases and good quantitative agreement for 5 out of 8 of the
bars in their experiments. We rule out the role of inharmonicity due
to heterogeneity as a cause for loss of macroscopic energy when the bar
makes the angle 7l" /2 with the rigid foundation, concluding thus that
this loss is mainly due to plastic deformation.
1. Introduction
In a seminal series of experiments Stoianovici and Hurmuzlu, 1996
dropped a slender bar onto a massive rigid foundation while the initial
position of the bar made an angle (J with the horizontal. They found
that the apparent coefficient of restitution, defined aB the ratio of the
post-collision to pre-collision vertical velocities, measured with the help
of a high-speed camera system, depended strongly on (J.
In numerical simulations presented in Paoli and Schatzman, 1999, we
have obtained results with an excellent qualitative agreement and a good
quantitative agreement with those of Stoianovici and Hurmuzlu, 1996;
in our discrete model we considered a frictionless rigid contact on the
foundation, while Stoianovici and Hurmuzlu, 1996 included a nonlinear
damped model of the contact with the foundation.
53
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 53-64.
© 2002 Kluwer Academic Publishers.
54
The analysis of Hurmuzlu, 1998 offers a way of understanding the
relation between continuous medium modes and restitution: it is based
on energy arguments, giving an ansatz for the form of the elastic en-
ergy stored into the continuous media modes, and then identifying the
relevant parameters for one set of geometrical data; these values are
subsequently validated by comparing experiment to theoretical results
in other experiments.
This explanation is quite satisfactory, since it accounts well for the
experimental observations; however, it contains an argument which we
do not understand completely, namely the choice of the ansatz. In this
article, we give a different and more theoretical argument, relying on
the analysis of the dynamics of the system in infinite dimension, i.e.
continuous medium modelling. This analysis does not answer all the
questions, but we hope that it sheds a bit more light on a problem
which would deserve to become classical.
The article is organized as follows: in section 2, we establish a La-
grangian formulation for the impact of the bar; this formulation in-
cludes the continuous medium modes; there are two different kinds of
constraints: one type of constraints is the non penetration constraint;
we have to impose three equality constraints to remove any indetermi-
nacy in the coordinates. All these constraints will be treated through
the use of Lagrange multipliers. We obtain thus a complicated system
of ordinary and partial differential equations with unilateral constraints.
In order to get to practical results, we make approximations and sim-
plifying assumptions in section 3; we estimate the physical quantities in
terms of the small parameter which is the ratio c of the vertical velocity
of the bar before impact to the velocity c of compressive waves in the
beam. We also assume that the reaction force due to the non penetration
constraint is constant throughout the first microcontact time interval;
in order to prove that this is close to truth, much mathematics might be
needed. Nevertheless, this is a simple enough assumption, which leads
to a number of definite conclusions: first, we obtain a kinetic coefficient
of restitution when there is only one contact interval; this coefficient is
qualitatively correct for the eight steel bars which have been tested by
Stoianovici and Hurmuzlu, and quantitatively correct for the five longer
ones.
In section 4, we go beyond the first microcontact time interval and
we produce an approximation of the solution, which enables us to find
the smallest initial angle for which there is a second micro contact time
interval soon after the first. We also obtain qualitatively correct results
for all bars, and they are quantitatively correct for the five longer ones.
55
2L/e
L/e
x -L/2 o L/2 x
Figure 1. Choice of axes and coordi- Figure 2. The regions involved in
nates for the problem. the definition of the function it in (17)
L/2 jL/2
j u(x, t) dx = v(x, t) dx = o. (1)
-L/2 -L/2
56
The bar makes an angle 8 with the horizontal axis and the vanishing
moment condition
L/2
/ xv (X, t) dx = 0, (2)
-L/2
C = pAL
x2 +2 y2 + "211L/2 pA[(u - . ..
vO)2 + (u + x)202 + 2uvO
-L/2
Then, the motion of the bar can be, at least theoretically, described
as an extremum of the action A = f C+ dt, under the constraint (4);
in fact, when the constraint is satisfied, this extremum of the action
is a maximum; in order to understand why this must be so, it suffices
to consider a linear case, where the sense of the variational inequality
corresponds to a maximum of the action.
When the constraint (4) is not active, the coefficients of each varia-
tion must vanish; when the constraint is active, the variations of Y, 8,
u( -L/2, t) and v( -L/2, t) must satisfy the following inequality
pALl' = A, (5)
d
pA-
dt
l L/ 2
-L/2
•
[v(vO - u)
•
+ O(u + x)2 + uti] dx (6)
l
1-'1 = --L-' 1-'2 = --L-'
(13)
L3 L L/2 •..
1-'3 12 = 2"ACOSO + pA x(20ut + uO) dx.
-L/2
o if (x, t) E R 1 ,
u{x, t) = { ct - x - L/2 if (x, t) E R2, (17)
2{ct - L) if (x, t) E R31
with the regions Rl, R2 and R3 of the plane being depicted in Fig. 2.
It is important to observe that
¢i
4) = et¢k, ¢Z{O) = ¢Z{I) = 0, ¢Z'{O) = ¢Z'{I) = 0,
(19)
111¢k12dx = 1, ¢k{O) = 2.
59
1
tPk(X) = -¢k
..fL
(x-L + -21) (20)
(23)
We use now relations (5) and (6) to find Y and 0, where we neglect
the quadratic terms in c; we find
.. c2
Y = c( L' (24)
.. ~
8 = -6 L2 c( cos 8, (25)
e-
Y(r) - O(r)L(cosO(O))j2 .
- IY(O)I '
we obtain finally
1 + (3 - J-t) cos2 0(0)
e= -~-..:......:..--:::--~ (28)
1 + (3 + J-t) cos2 0(0)'
(29)
61
Table 1. The critical angles and times for the second microcontact time interval
with free end boundary conditions, and initial data at time T deduced
from (23). Therefore, v( -L/2, t) - v( -L/2, T) is given explicitly by
v{ -L/2, t) - v( -L/2, T)
_ r 4L3 O~ COS Wk(t - 1 + COSWkT
- C.. ",2 cos L....J 4 e
T) - COS(Wkt) -
•
k~l k
(30)
We sought numerically the value Oc of 0 and t' for which (30) vanished
for the first time; the results are given in table 1 for Stoianovici and
Hurmuzlu bars, which are made of steel with E = 2.11011Pa and p =
7.86kg x m- 3 .
In order to compare our theoretical results with the experimental
measurements of Stoianovici and Hurmuzlu, 1996 and Hurmuzlu, 1998,
we have represented at figure 3 the plots of the coefficient of restitution
e given by (28) on the angle interval [Oc, 71'/2]; on the plot, the angle
is measured in degrees. This figure is to be compared to figure 12 of
Hurmuzlu, 1998, and we see that for bars 4 to 8, our results seem closer
to the experimental results than the numerical results of Stoianovici and
62
. e /'
e
/
r- ~
'r- bo" 8.2
/
/
(J (J
e e
/
= 80"
u-
IL7 _ Bill: 4
(J
e e
r-
I- a..,S
r- ... 6
I-
e e
ur-
u r-
r-
Bar 7
r- B.B
(J
Figure 3. The restitution coefficient for bars 1 to 8 (left to right and then top to
bottom).
Hurmuzlu, 1996. However, our results are much worse for the first three
bars. It should be emphasized that our analysis ignores friction, and
this may account for the discrepancies.
Another experimental comparison is the time duration between the
first and the second impact; in the case of bar 3, we have performed ex-
tensive numerical experiments, reported in Paoli and Schatzman, 1999;
the observed time which can be read on Fig. 5 of that paper is 0.410- 3 S
which is completely compatible with the theoretical value found and
reported in table 1.
63
References
Brahim-Otsmane, S., Francfort, G. A., and Murat, F. (1992). Correctors for the ho-
mogenization of the wave and heat equations. J. Math. Pures Appl. (9),71(3}:197-
23l.
Francfort, G. A. and Murat, F. (1992). Oscillations and energy densities in the wave
equation. Comm. Partial Differential Equations, 17(11-12}:1785-1865.
Gerard, P. (1991). Mesures semi-classiques et ondes de Bloch. In Seminaire sur les
Equations aux Derivees Partie lies, 1990-1991, pages Exp. No. XVI, 19. Ecole Poly-
tech., Palaiseau.
Hurmuzlu, Y. {1998}. An energy based coefficient of restitution for planar impacts of
slender bars with massive external surfaces. ASME J. Appl. Meek., 65:952-962.
Paoli, L. and Schatzman, M. (1999). Dynamics of an impacting bar. In Wunderlich,
W., editor, European Congress on Computational Mechanics, Munich. CD-ROM.
Stoianovici, D. and Hurmuzlu, Y. {1996}. A critical study of the applicability of rigid
body collision theory. ASME J. Appl. Meek., 63:307-316.
SOLVING ROCKING BLOCK PROBLEMS
WITH MULTIPLE IMPACTS
Cuneyt Yilmaz
Mechanical Engineering Department, Southern Methodist University
Dallas, TX-75275-0337 USA
Yildirim Hurmuzlu
Mechanical Engineering Department, Southern Methodist University
Dallas, TX-75275-0337 USA
1. Introduction
Multiple impact problems still have many difficulties and unanswered
questions (see Marghitu and Hurmuzlu 1995, and Brogliato, 1996). The
rocking block problem, where multiple impacts are likely to occur, is one
of the simplest rigid body impact problems that may involve multiple
collisions.
Housner (1956) introduced the first study to derive the mathematical
equations of a free-standing rigid block under base excitation. Aslam
et al (1980) investigated the earthquake rocking response of rigid bod-
ies, and defined angular coefficient of restitution by using experimental
data. Lipscombe and Pellegrino (1993) presented a solution for free rock-
ing prismatic blocks, and discussed the angular coefficient of restitution
for it. Moreau (1994) used a complementarity dynamical formulation
to solve the multi-body impact problems. This research was followed
by Cholet (1998) and Johansson (1999). Johansson (1999) investigated
the rigid body impact dynamics by using a linear complementarity al-
gorithm considering Coulomb's friction effect. Cholet (1998) used an
adaptation of Moreau's sweeping process (see also Fremond 1995). This
65
J.A.C. Martins and M.D.P. Monteiro Marques (eds.). Contact Mechanics, 65-73.
© 2002 Kluwer Academic Publishers.
66
work produces unique and energetically consistent solutions. The only
drawback of the approach is that the problems are formulated in terms
of parameters that do not have obvious physical meanings.
Ceanga and Hurmuzlu (2000) introduced a method that produces
unique and energetically consistent solutions in multiple impact prob-
lems. They developed a new approach that used the energetic coefficient
of restitution, and proposed a new constant, that they called the "Im-
pulse Transmission Ratio". This parameter is physically meaningful and
more effective in dealing with multiple impact problems. They applied
the method to the multiple impacts that took place in an linear N-ball
chain.
2. Problem Description
In this article we consider the system given in Fig. (1). The impact
problem takes place as a result of the block striking the left surface (at
Od while resting on the right surface (at O 2 ). The block is symmetrical
with a width of 2w, a height of 2h, mass m, and a centroidal moment
of inertia of fern. To simplify the calculations, we choose a coordinate
system whose origin is attached to the center of the block and its x-axis
parallel to the horizontal edge of the block at the impact instance. At
the instant immediately before impact, the block is undergoing a non-
centroidal rotation about 02 with an angular velocity of w-. In addition,
we consider frictionless contacts only and choose surface inclinations of
fh and (}2 (see Fig. (1)) at the the respective contact points 0 1 and O2 •
In this paper we apply the Impulse Transmission Ratio (ITR) concept
that was developed in Ceanga and Hurmuzlu (2000) to resolve the diffi-
culty that is encountered in the present problem. In the next section we
use compliance contacts to derive the ITR for the rocking block problem.
Y
m,Icm
h
x
r1 r2
h
Vern
W
.1. w
(2)
where P1, (12, ).1, ).2, a1, a2, a3, (31, (32, (33, and f are functions of m, h,
w, I cm , (h, (h, k, and 'Y. The impulses acting at the two contact points
can be computed as follows:
(4)
4. Velocity-Impulse Relationships
Using the laws of conservation for the impulse and momentmn, we
obtain the following equations:
(7)
(8)
where, m is the mass, Icm is the centroidal moment of inertia of the
block, l'cm and w are the linear and angular velocities of the mass center
of the block respectively, 71 and 72 are the impulses, and r1 and r2 are
the vectors from the mass center to the contact points. The velocities
at the contact points can be written as follows:
(9)
(10)
69
where V1 and V2 are the velocities of the block at 0 1 and 02 respectively
(see Fig. (1)).
We use Eqs. (7), (8), (9), and (10) to derive the velocity expressions
in terms of the collision impulses.
5. Bouncing patterns
We consider the case where the block strikes the external surface at
0 1 while resting at 02 (vr- i= 0 and v~- = 0, or more specifically
W- = Wo i= 0). There are two possible bouncing patterns that result
from the collision at 01, they can be enumerated as follows:
-dv~ = --sm(h
dv~. dv~
+ -COS(}2 =
dT1 dT1 dT1
(h 2 + w 2) COS((}l - (}2) + [-2 ~ + (w 2 - h 2)] COS((}l + (}2)
2~ + (h 2 + w 2) + (w 2 -
h 2) cos 2{}2 - 2hw sin2(}2
2hw sin((}l + (}2)
> 0 (11)
2 ~ + (h + w ) + (w 2 - h2) cos 2{}2 - 2hwsin2(}2
2 2
restitution to compute the final impulse for the first collision at 01 as:
f 2(1 + edr1WOCos81
71 =
rg + r10
and the final impulse for the first O2 collision as:
72f = r1 r 1S
r17(r 2 cos 82 + r4 sin2 82 - r3 sin(282))
2
[V
r 14+ e2 -r16a+r 214]
where
r 1S = (1 + edwo cos 81
(12)
Simplifying the expression given in Eq. (12) yields the identical upper
bound for the compliant case, which was presented in Eq. (5).
I.SO
1.25
1.00
'0'
g 0.7S
",- Single Impact
D 01 - Tan-'[(2r'.l)l(3rl] aJ-
9 1 -Tan" [(3r)l(r'+4l]
0.25
r< .Jf
0.00 .....,...",...-:-:---:-:--'-:-:-:......,...,,.,......:.;~~
l .s<l 1.25 1.00 0.7S 050
92 (Rad)
(a)
blocks (large r values) multiple impacts are more likely occur, while for
a narrow block (r is small) we are more likely to have a single impact.
1. Conc1 us ion
In this paper we developed a new approach to solve the multiple im-
pact problem of a rocking block. The methodology is based on the use of
impulse momentum methods. The approach uses the Impulse Transmis-
sion Ratio that was developed previously to solve the multiple impact
problems in a linear chain of balls (Ceanga and Hurmuzlu, 20(0). The
method also utilizes the energetic coefficient of restitution and yields
energetically consistent solutions (although this issue is not valid in the
present paper since friction is neglected).
The problem considered here is simplified by neglecting friction at the
contact points. This was a necessary simplification in this initial stage
of the development. Yet, including friction will be a good natural step
for future research efforts.
References
Aslam, M., Godden, W.G 1956, Scalise, D. T., 1980, "Earthquake Rocking Response
of Rigid Bodies" , Journal of Structural Division, Feb. 1980, pp. 377-392.
Brogliato B., 1996 Nonsmooth Impact Mechanics: Models, Dynamics and Control,
Springer Verlag, LNCIS 220.
Ceanga V., Hurmuzlu Y. 2000, "A New Look At an Old Problem: Newton's Cradle" ,
to appear in the ASME Journal of Applied Mechanics.
Cholet C. 1998 Chocs de Solides Rigides, Ph. D. Thesis, University of Paris 6 and
LCPC-CNRS, March, 1998)
Fremond M. 1995, "Rigid bodies collisions", Physics Letters A, 204, pp 33-41.
73
Bernard Brogliato
INRIA RhOne-Alpes, ZIRST Montbonnot, 655 avenue de l'Europe
38334 Saint lsmier cedex, France
Mongi Mabrouk
Laboratoire de Mecanique Appliquee Rene Chaleat
24 chemin de l'Epitaphe, 25030 Besan(,!on cedex, France
1 Introd uction
Manipulating objects by pushing and hitting (also called nonprehen-
sile manipulation) is an important robotic task, see Akella et al (2000)
and references therein. It is easily recast in the setting of so-called jug-
gling systems (a class that encompasses systems with dynamic backlash,
75
I.A.C. Martins and M.D.P. Monteiro Marques (eels.), Contact Mechanics, 75-82.
© 2002 Kluwer Academic Publishers.
76
2 System's dynamics
Let us consider the following class of complementarity dynamical sys-
tems that we may call linear jugglers
Mlih = AT>.
(1)
O~Aq+B..l>'~O
The "proxM" denotes the proximation in the kinetic metric and the
times tk generically denote impact times. The tangent cone to the do-
main Aq + B ~ 0 is defined as V(q(t» = {x E lR 2 X lR 2 I Ax =
~ ~
3 A controllability criterion
In view of this let us investigate another path to characterize the con-
trollability properties of linear juggling systems as in (l).We make the
following
Assumption 1. Let Al be made of the first ~ columns of A. The
matrix Al satisfies rank(At) ~ 2.
77
The control problem is decoupled into two parts: the control of the
zl-dynamics using A as the input (i.e. using impacts), then the control
of the overall system with u. Here we focus on the first part only.
Let us formulate the following controllability problem which we name
Controllability Through the Impacts (CTI)
Problem 1 (CTI). Given (q},vt) and (qf,v~) with vi =F 0, find
{ql(k)h$k$n, {lh(k)}2$k$n, {lJ2(k)h$k$n (2)
such that (qt{tr), ql(ti» = (q}, vt) and (ql(t n), ql(tt» = (qf, v~).
Notice that n ~ 2 by construction. Controllability is here to be
understood as a general assertion and obviously can be replaced by ac-
cessibility. The constraint on vi will be made clear later. The reachable
subspaces are defined as 'R,n [(qt,vt)] = {(qf,v~) I problem 1 possesses
at least one solution}, and denote tin [(qt,vi)] = U3$k$n'R,k [(q},vi)].
Following the standard terminology Nijmeijer and van der Schaft (1990),
the dynamics of the object may be called accessible when tin [(q}, vt)]
contains an open set for any n ~ 3, and accessible in N impacts if
tin [(q},vt)] contains an open set for all n ~ N. Let us denote xf =
xr
(qf(2), ... ,qf(n-1», = (qf(l), ... ,qf(n», xf = (qf(2), ... ,qf(n-
1», xT = (qf(2), ... ,qf(n».
The main goal of this note is to prove the following
Lemma 1. Let assumption 1 hold. Problem 1 has a solution if and
only if the constrained equation H(Xl' X2, X4) = 0, G(XI' X2, X4) ~ 0 has
a solution, where H(·) and G(·) are some nonlinear functions. Further-
more H(xI, X2, X4) = HX4 (Xl, X2) and G(XI, X2, X4) = GX4 (Xl, X2), where
H X4 (·) and GX4 (-) are linear.
The functions H (.), G(·), HX4 (.), GX4 (.) can be constructed as follows
AIj(vi,A,MI,M2,X4)XI + BIj(vi,A,q},qf,MI ,M2,X4) = 0
J. E {1 , ... ,m(n-l)}
(3)
78
(7)
1l(X4).X1 + I(x4) ~0
fo--------.
. . ._ . . . ._ _ _. . ._ ..... M2
.. ,
,,
'
.. ,
, L+L =L
1 2
v=( o) (10)
(He)m2(eml-m2) (He)ml
(ml+ma)2 - ml+m2
1 -vi
m1+ema v +
:F= (He)ma
ml+ma -1 0 g= ml+m2 1
+ (He)am1m2 +
1 0 0 eV1 - (ml +ma)2 v1
(11)
-1
0
-ev+ +
I
(1-te)2ml m2 v+ +K:x4
(ml +m2)2 I
-qf1l12
1l=
-1 1
I= 0
~+
ml+m2 VI + c. X4 ml+m2 VI + C. X4
~+
1
vt
0 ::#
VI
(12)
(13)
(14)
where X4,i is the ith component of X4. The inequalities in (14) define
a domain Dl C IR3 for X4 E F that clearly has a nonempty interior:
int{Dl) -=J. 0. Actually Dl is the intersection of three half-spaces whose
boundaries are three planes. From the second line in (7) and from (10)
it follows that X4 has to belong to a plane PI that is orthogonal to
the plane (O, x, y) where 0 is the origin and x, yare the first two
coordinates. The intersection of PI with (O,x,y) is a line L 1, which in
turn intersects (O, y) at the point {O, -£ (fi~)~~' 0). Clearly the signum
of £ can be varied arbitrarily by modifying vt, which corresponds to
translating PI along its normal 'D in (10). Since int{Dd -=J. 0, it follows
that there exists a set 81 for vt such that PI n Dl =1= 0 for all vt E 81,
and int{8d -=J. 0.
Let us now examine the denominators of 11. and I. The values of X4
which yield singularities belong to two planes P2 and P3, which are both
orthogonal to (0, X, y) in the 3-dimensional X4 space. Therefore even in
the worst case where P2 n PI n DI -=J. 0 and P 3 n H n DI =1= 0, the set
of admissible X4 remains of nonempty interior and so does 8 1 3 vt. It
remains now to examine the last inequality in (7), which provides the
existence of Xl E IR2. The first and third inequalities yield XI,2 ~ qt
or XI,2 ~ qt and XI,1 ~ qi or XI,I ~qi. The second inequality defines
a half plane whose boundary is a line containing the origin O. There
exists a set 8 2 with nonempty interior such that for every qt E 82 the
intersection of these three half planes has a nonempty interior. Finally X2
can be calculated from (8). Since 81 and 82 have non-empty interiors,
the reachable set 'R,4[(qi, vi)] contains an open set. Since the above
reasoning can be redone with new initial data chosen as the quantities
at time t2, one concludes that 'kn [(qi,vi)] contains an open set for all
n ~ 4.
82
References
S. Akella, W.H. Huang, K.M. Lynch, M.T. Mason (2000) Parts feeding
on a conveyor with a one joint robot, Algorithmica, vo1.26, pp.313-344.
B. Brogliato (1999) Nonsmooth Mechanics, Second Edition, Springer
London CCES.
B. Brogliato and A. Zavala-Rio (2000) On the control of complementary
slackness juggling mechanical systems, IEEE Transactions on Auto-
matic Control, vo1.45, no 2, pp.235-246, February.
M. Mabrouk (1998) A unified variational model for the dynamics of per-
fect unilateral constraints, European Journal of Mechanics A/Solids,
vo1.l7, no 5, pp.819-842.
H. Nijmeijer and A. J. van der Schaft (1990) Nonlinear Dynamical Con-
trol Systems, Springer Verlag.
ON INTEGRATING
STIFF MULTIBODY DYNAMICS
WITH CONTACT AND FRICTION
Mihai Anitescu
Department of Mathematics University of Pittsburyh,
Pittsburyh, PA 15213, U.S.A.
Florian A. Potra
Department of Mathematics, University of Maryland, Baltimore Oounty,
Baltimore, MD 21250, U.S.A.
1. Introduction
In Anitescu and Potra (2001) we proposed a time-stepping method
for stiff multibody dynamics, subject to joint, contact and frictional
constraints. The method is set up in a velocity-impulse framework, and
is shown to be consistent for a wider class of configurations than the
ones covered by classical mechanics, similar to previous such algorithms
cf. Anitescu and Potra (1997), Anitescu and al. (1999), Stewart and
Trinkle (1995) and Stewart (2000).
such vectors d1 (q) and d2 (q) are needed in this formulation. We de-
note by D(q) = [d 1(q),d2 (q), ... ,dmc(q)]. A tangential impulse will be
:E~1 (1ic4(q), where {1i ~ 0, i = 1,2, ... ,me. The friction model, which
ensures maximum dissipation for given normal impulse en and velocity
v, as well as that the total contact force is inside the discretized cone, is
D(q)Tv + .Ae ~ 0 ..L {3 ~ 0, Jl.Cn - eT{3 ~ 0 ..L .A ~ o.
Here e is a vector of ones of dimension me, e = (1,1, ... , l)T, Jl. is
the friction parameter and {1 is the vector of tangential forces {3 =
({31, f32, ... , (3mc ). The additional variable .A is approximately equal to
the norm of the tangential velocity at the contact, if there is relative
motion at the contact, or IID(q)TvII =I- 0 cf. Stewart and Trinkle (1995)
and Anitescu and Potra (1997).
Dynamical data of the simulation. We denote by M(q) the sym-
metric, positive definite, mass matrix of the system and by k(t, q, v) the
external force.
0 0 t-t -ET 0 .A 0 (
(5)
[l n~ 1
86
= 0,
c- =
Here 1/- -- [1/(1) , 1/(2) , ... , I/(m)] ,II
[ II ~ 0, [ ~ 1~ O. (6)
11,11,···,11,
[C(1) C(2) C(m)]T n- = [n(i!) "n(jI)
Using these approximations when computing ij{i) (7) and k(i) (8), it
was shown, cf. Anitescu and Potra (2001), that
1 The matrix M(l) is positive definite for any value of h and the
scheme (5-6) is thus unconditionally consistent.
2 An energy inequality holds from one step to the next, which ensures
that a system with constant mass and linear mappings defining the
stiff forces is stable.
3 If damping is the dominant stiffness effect and the friction cone
is pointed, then we obtain that, for high stiffness, the solution set
of the time-stepping scheme (5-6) approaches the solution set of
the problem with the springs and dampers replaced by rigid joint
constraints.
,
ri.T 0 0 0 0 + 0 15
jjT 0 0 0 E {3 0 u
0 0 Ji. -EfT 0 A 0
(13)
nn~
88
[~ 0, [ ~ 1~ 0, [ ~ 1~ 0. (14)
We now make the assumption that the friction cone of the limit system
(13-14) is pointed. This is a key assumption for proving that, in the
nonstiff case, as h -7 0, the sequence created by this algorithm converges
in a measure differential inclusion sense, cf. Stewart (2000). In our case,
this is equivalent to assuming that any solution of (13-14) for a zero free
term (v(l) = On, rv = Om+n.s-y and k~') = On) must satisfy v(l+l) = 0,
en
~/I = 0, = 0 and ,B = O.
We now focus on the LCP (5-6), where M (7) and k (8) are com-
puted using the approximations (10) and (11). We assume that we have
a sequence of stiffness parameters, -r:,6~, k = 1,2, ... ,ny6, P = 1,2, ...
such that all elements of the sequence increase to infinity, but the dom-
inant stiffness or damping characteristic of the individual component is
6P 7f
preserved, or :t -7 0, k = 1,2, ... , n'Y and ~ -7 0, k = n'Y + 1, n'Y +
'Yk k
2, ... , n'Y6 . Denote by v(i+l),p, ~, ~, ~P, .xp a solution of (5-6) thus ob-
tained at any p. Although in some perfectly behaved cases there may
be an unbounded choice of .xp, we can always choose .xp to be the rel-
ative velocity at the contact. Therefore we can assume without loss of
generality that we can choose .xp such that II.xPII ~ CcIlV(l+l),PII, where
Cc is a constant that depends only on the friction cone, but not on p.
For k = 1, 2, ... , n'Y6, define dt = h( 6: !h;Y:) , ~ = 6t J;Y: and
tions and using ~ = [~, ~,m+1' ~,m+2' ... ,~,m+no-y] in the framework
of the limit problem (13-14) we obtain that v(l+1),p, ~, ct:., fjP, .>:p must
satisfy
M -n -D 0 v(l+1),p
-/I - M v(l) - hkl') 0
-;::p
-:::;T
/I DPI 0 0 0 Cv 77-
lTv 0
fiT 0 0 0 0 cP
~n + 0 P
jjT 0 0 0 E fJP 0 U
0 0 J.t -E
~T
0 ),.P 0 (
(15)
l ~;~]Tlli]{= 0, lct:.]
~; ~ 0, lli]
{~ O. (16)
-;::p
Let AP00 = Ilv(l+l),p , cv' cPn' JJ,
RP ),.PII 00· Assume that a subsequence of
~-
References
Anitescu, M., and Potra, F .A., " A time-stepping method for stiff multibody dynamics
with contact and friction", Reports on Computational Mathematics MCS-P884-
0501, Mathematics and Computer Science Division, Argonne National Laboratory,
200l.
Anitescu, M., and Potra, F.A., "Formulating rigid multi-body-dynamics with contact
and friction as solvable linear complementarity problems", Nonlinear Dynamics
14, 231-247, 1997.
Anitescu, M., Stewart, D. and Potra,F. A., "Time-stepping for three-dimensional rigid
body dynamics", Compo Methods Appl. Mech. Eng. 177(3-4), 183-197, 1999.
Anitescu, M., Cremer, J., and Potra, F. A., "Formulating 3D contact dynamics prob-
lems", Mechanics of Structures and Machines 24(4), 405-437, 1996.
Cremer, J., and Vanecek G., "Building simulations for virtual Environments", Pro-
ceedings of the IFIP International Workshop on Virtual Environments, October
1994, Coimbra, Portugal.
Dirkse, S. P., and Ferris, M. C., "The PATH solver: a non-monotone stabilization
scheme for mixed complementarity problems", Optimization Methods and Software
5, 123-156, 1995.
Haug, E. J., Computer Aided Kinematics and Dynamics of Mechanical Systems, Allyn
and Bacon, Boston, 1989.
Munson, T. S., "Algorithms and Environments for Complementarity", Ph.D Thesis,
Department of Computer Science, University of Wisconsin-Madison, 2000.
Stewart, D. E., and Trinkle, J. C., "An implicit time-stepping scheme for rigid-body
dynamics with inelastic collisions and Coulomb friction", International Journal for
Numerical Methods in Engineering 39, 2673-2691, 1996.
Stewart, D., "Rigid-body dynamics with friction and impact", SIAM Review 42 (1),
3-29,2000.
ANALYSIS OF SYSTEMS WITH MULTIPLE
FRICTIONAL CONTACTS
A. P. Ivanov
Moscow State Textile University named after A.N.Kosygin
M.Kaluzhskaya 1, Moscow 119991, Russia
Abstract Different types of paradoxical situations in systems with two or more frictional
constraints are discussed. Along with well-known Painleve paradoxes of non-
uniqueness and non-existence, a new kind of singularity is detected: co-
existence of several stable solutions. Besides, it is shown that in a regular case
where a unique solution exists, this solution might be unstable.
1. Introduction
with Q being the generalized force and R(I) being the constraint force. For
a fixed instant to and given values of q and q, system (1) is to be solved
with respect to q and R(l). We suppose that at t =to the normal velocity at
the point of contact vanish (i.e. ql = 0) while the tangential velocity does
not. In this case, the tangential components of reaction force ~I) , • •• ,R~I) can
be expressed as linear functions of the normal component R?) by means of
Coulomb's law. Substituting these expressions in equation (l) and then
excluding q2'" .qn' we arrive at the single equation for ql and R?)
.. -AR(l)
ql - 1 +
B (2)
.. > 0, R(l)
ql- I
>
-
0,ql
.. R(I)
I
=0 . (3)
93
Obviously, in the case A> 0 equation (2) has a unique solution which
agrees with (3): if B > 0 (B < 0), then ij} = B, R?) = 0 (ij} = 0, R?) = -B / A).
Paradoxes arise in the case A < 0: if B > 0, both the above solutions are
suitable, while for B < 0 no solution exists. All these possibilities can be
represented in the (B,x) - plane (see Fig.1) with
if ij} > 0
(4)
if Ril) ~ 0
x x
B B
.....
....
A>O
A<O
We discuss the stability of these solutions (in a sense). For such purpose,
we apply a quasi-static approach, usually adopted in bifurcation theory. The
variables qj' ifj (j * 1) are "frozen", and A and Bin (2) are fixed. Further,
small violations of the constraint q} ~ 0 are possible, so that q} = -E, E << I .
Moreover, the constraint force R?) is a function of s and i, R?) = F(s,i) .
From physical premises, this function should be monotone increasing with s
and dissipative, i.e. F; > O,Fl > O. It is easy to show that an equilibrium
with x < 0 is stable ¢:;> A > o. Concerning a solution with x> 0 , it is stable
since x = 0 for B fixed. Thus, in the left diagram in Fig. I, both branches are
stable while in the right diagram the upper branch is stable and the lower one
is unstable.
It is worth noting that in the regular case we have a unique "proper"
solution for small initial perturbations. In paradoxical case, however, in the
absence of stable solution (B < 0) the system experiences a velocity jump
after which the quantity A becomes positive.
94
3. Systems with two frictional constraints
Equation (1) should include two constraint forces R(I) and R(2) . Excluding
the unknowns Q3" .. ,qn' we get the following equation similar to (2):
c c: a
l. All > 0, ~ > 0, ~ = All ~ - A12~1 > °- regular case where unique
2.
3.
All > 0, ~ > 0, ~ < 0, AI2 > °-
solution exists for any B» B2 (cf. Cottle et al. (1992»;
cusp;
~I > 0, ~ >0, ~ <0, AI2 <0 - fold;
4. AII~ <0, ~ <0 - fold;
5. AIIA22 <O,~>O,AII~I >O-fold;
6.
7. All < 0, ~ < 0, ~ > °-
~I~ <O,~>O'~I~I <O-cusp;
double fold. To imagine this singularity, one
should bend a sheet of paper in halves and then bend it once more
along a line which is non-orthogonal to the line of the first bending;
8. ~I <0, ~ <O'~<O'~I <O-fold;
9. All <O,~<O,~<O'~I>O-CUSp.
Now we argue the stability of solutions of equation (5) in the same sense
as previously in Section 2. Thus, we "freeze" variables qj' (t (j > 2) and
present reaction forces as some functions of small deformations:
(6)
Generally speaking, there exist three types of solutions: (i) those where
°
Xi and X 2 are positive (i.e. both deformations vanish); (ii) those where
Xix2 < (i.e. one of the deformations is non-zero); (iii) solutions those where
Xi and X 2 are negative (both deformations are non-zero). Solutions of the
first type are stable since for zero reactions and BI and B2 being constant,
we obtain from (5) ~ = x2 = 0. Furthermore, a solution of the second type is
stable ~ the corresponding element of matrix A is positive (i.e. An > 0 for
a solution where R?) =0, RJ2) > 0 etc.).
The third case is most complicated: in general, it is impossible to make
conclusion on stability without specifying the strain-stress functions (6). A
special subcase where matrix A =IIAy II is symmetric can be studied with the
help of Kelvin theorem: the solution is stable ~ A is positive definite. Due
to the presence of dry friction, A is non-symmetric in general. An example
was considered by Popp and Rudolph (2001) who considered a system with
linear functions (6) and detected regions of stability and of instability as
well. Some other conditions of stability and instability which are invariant to
the choice of functions (6) were obtained by the author (2002).
We apply these arguments to the list of singularities.
96
1. The unique solution may be stable or unstable depending on strain-
stress functions (5) (we denote such kind of stability (D»;
2. One stable solution, or two stable (S) + one unstable (U);
3-4. No solution or S+U;
5. No solution, S+U, or D+U;
6. S, D, or S+U+D;
7. No solution, U+U, S+U, or S+U+U+U;
8. No solution, or S+N;
9. U, or S+U+U.
These results allow to understand the dynamics of system (1). In such
cases where either no solution, or no stable solution exists, the system
experiences an instantaneous velocity jump - impact by friction. If there
exist one stable and one or more unstable solutions, the stable one is realised
for most initial conditions. Finally, if we have two stable solutions at once,
each of them can be realised for certain initial conditions. This situation is
similar to the famous example of bifurcation - the Euler pinned column.
3.3. Example
The equations of motion are (we assume that the wheel has unit mass):
97
In the simplest case where the wheel is homogeneous and flt =/12 =/1 ,
we have a = b = 0 and AlJ > 0, A22 > 0,!1 > o. Therefore, a unique solution
of system (5) exists for any external forces. However, the stability problem
is still open. To solve it, let reaction forces be similar linear functions of
deformations: RI = gEl + / &pRz = gE2 + /&2· Then equation (5) is
(10)
Acknowledgments
References
Cottle, R.W., Pang, lS., and Stone, R.E. (1992) The linear complementarity problem.
Academic Press, Chichester.
Ivanov AP. (2002) On dynamical paradoxes in systems with dry friction. To appear
Klein, F. (1910) Zu Painleves Kritik des Coulombschen Reibungsgesetze. Zeitsch.Math.Phys.
58, 186-191.
Lecomu, L. (1905) Sur la loi de Coulomb. C.R.Acad.Sci.Paris. 140,847-848.
Lotstedt, P. (1981) Coulomb friction in two-dimensional rigid body systems. ZAlvIM. 61,605-
615.
Lotstedt, P. (1982) Time-dependent contact problems in rigid body mechanics. Math.Progr.
Study. 17, 103-110.
Painleve, P. (1905) Sur 1es lois du frottement de glissement. C.R.Acad.Sci.Paris. 121, 112-
115; 141,401-405,546-552.
Pang, l-S., Trinkle, lC. (1996) Complementarity formulations and existence of solutions of
dynamic multi-rigid-body contact problems with Coulomb friction. Math.Progr.73, 199-
226.
Popp, K., Rudolph M. (2001) Brake squeal. Detection, utilization and avoidance ofnonlinear
dynamical effects in engineering applications. Shaker Verlag. Aachen. 197-225.
Prandtl, L. (1910) Bemerkungen zu den Aufsatzen der Herren F.Klein, R.v.Mises und
G.Hamel. Zeitsch.Math.Phys. 58, 196-197.
Trinkle, lC., Pang, l-S.,Sudarsky, S., Lo, G. (1997) On dynamic multi-rigid-body contact
problems with Coulomb friction. ZAlvIM. 77,267-279.
Whitney, H. (1955) On Singularities of Mappings of Euclidean Spaces I. Mappings of the
Plane into the plane, Ann.Math. 62, 374-410.
INITIATION OF FRICTION INSTABILITY
ON A PLANE FAULT SYSTEM
loan R. IONESCU
Labomtoire de Mathematiques, Universite de Savoie, 73376 Le Bourget du Lac, France
Michel CAMPILLO
Labomtoire de Geophysique Interne et Tectonophysique, BP53, 38041 Grenoble, France
Cristian DASCALU
Institute of Applied Mathematics, P.O. Box 1-22, 70700 Bucharest, Romania
Pascal FAVREAU
Institute of Crustal Studies, University of California, Santa Barbam, CA 93106, USA
Christophe VOISIN
Dept. of Geological Sciences, San Diego State University, San Diego, CA 92182, USA
Abstract We study an unstable elastodynamic process during the initiation phase (i.e. the
period between a perturbation of a unstable state and the onset of rupture propa-
gation associated with the seismic wave radiation). We consider the elastic anti-
plane problem for a system of finite faults under a slip-weakening friction law.
A spectral analysis is used to determine the existence, or not, of a catastrophic
evolution ofthe slip. We find that long initiation durations are expected. We also
investigate the possibility of defining an effective friction law for a finite fault
with a small scale heterogeneity. The "spectral equivalence" between an het-
erogeneous fault system and an homogeneous fault is pointed out. Surprisingly
good agreements are found between the heterogeneous fault model and the ho-
mogeneous fault with an effective friction law. Finally we analyze the initiation
pattern as a possible signature of instability and we show how the weakening
rate is transmitted in the elastic medium through a "domain of confidence".
99
J.A. C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 99-108.
© 2002 Kluwer Academic Publishers.
100
1. Introduction
Laboratory experiments on friction (Dieterich (1979), Ohnaka et al. (1987),
Ohnaka and Kuwahara (1990» have pointed out the existence of a phase of
slow but accelerating motion, called the initiation or nucleation phase. This
initiation stage precedes the propagation of the dynamic instability along the
fault surface. The pre seismic slip associated with the nucleation process should
be recognized as a manifestation of some phenomena precursor to the rupture.
Following these experiments, we shall consider here a slip-weakening friction
law.
The elastic quasi-static problem with slip dependent friction was studied
in Ionescu and Paumier (1996) where results concerning non homogeneous
bifurcation of the static equilibrium positions were obtained. Having in mind
the multiplicity of the equilibrium positions they concluded that it is difficult
to predict the new equilibrium position with a quasi-static analysis and that a
dynamic analysis is required.
We shall concentrate here on the elastodynamic analysis of the friction in the
anti-plane case. More precisely, we focus on the initiation of the shear process
during the weakening stage to point out simple mathematical properties of its
unstable evolution as a result of a slip dependent friction law. Our aim is to
describe the growth of the instability in a form very simple to evaluate and to
interpret. The present paper does not discuss the long term evolution of the
system as done in Cochard and Madariaga (1994), Geubelle and Rice (1995).
(1)
101
°
for t > and y > 0. The boundary conditions on y = °are:
W(t,x,O) ° if x ¢ r/, (2)
aoyw(t,x,O) +r~ = ~(x,w(t,x,O)))S, if Otw(t,x,O) > 0, (3)
aoyw(t,xO) +r~ < ~(x,w(t,x,O))S if Otw(t,x,O) = 0, (4)
where ~ (x, s) is the coefficient offriction which is a function of the slip sand
may be non-homogeneous. The initial conditions are denoted by Wo and WI,
that is,
w(O,x,y) = wo(x,y), Otw(O,x,y) = WI(X,y). (5)
Since our intention is to study the evolution of the elastic system near an un-
°
stable equilibrium position, we shall suppose that r;:O = S~(x, 0). We remark
that taking was a constant satisfies (1)-(4); hence w == is an (unstable) equi-
librium position, and wo, WI may be considered as small perturbations of the
equilibrium.
In order to give a non dimensional formulation we introduce a, the charac-
teristic length, and we put Xl = Xf a, X2 = Y f a.
°
Suppose that the initial perturbation is small and the nonlinear function ~
may be approached in a neighborhood of s = by its linear approximation i.e.
(6)
and we introduce the non-dimensional function:
where (7)
We can state now the following linearized evolution problem:
oltw(t, Xl, X2) (cfa)2\1 2w(t, XI, X2) (8)
w(t, Xl, 0) 0, for Xl ¢ r I, (9)
OX2 W(t,XI,0) -,B(XI)W(t,XI, 0), for Xl E r/,(lO)
W(0,XI,X2) = WO(XI,X2), Otw(O, XI, X2) = WI (Xl, X2). (11)
,B(XI)<P(XI,O) = -~FP
7r
r <p(s,O)KI~As I-s - ~l I) ds,
ir, Xl
(15)
where A 2:: 0, K I is the modified Bessel function of the second kind and the
integral is taken in the finite-part sense. In the case of homogeneous single
fault system (i.e. r / = [-a, a], and ,B(XI) = canst) this integral equation
has been solved by Dascalu et al. (2000) for small values of A using a semi-
analytical method. Recently, this method have been improved by Dascalu and
Ionescu (2001) to work in the case of a system of multiple and homogeneous
faults (i.e. ,B(XI) = ,Bk on the fault k) and for arbitrary A.
The second technique, developed in Voisin et al (2001) uses a finite element
approach. To do this the finite fault zone is embedded in a bounded elastic
domain n =] - L, L[ x ]0, L[. The infinite elastic half space is limited by a
fictitious boundary all over which the displacement is negligible, i.e. a null
displacement all along r d, the part of the boundary of n which is not on the
fault r /' is imposed. The variational formulation of (12)-(14) is
~l sinh(cIAnlt/a)
f;:o [cosh(cIAnlt/a)Wn + a
0 I
CiAni Wn]<Pn(Xl, x2),(16)
~ 0 sin(cIAnlt/a) I
n~[COS(CIAnlt/a)Wn + a CiAni Wn]<Pn(XI, X2). (17)
where wg, W~ are the projections of the initial data on the eigenfunctions.
We remark that the part of the solution associated with positive eigenvalues
103
-+_.... n. I
dstO'lCe (m)
OQO """ """
A2 will have an exponential growth with time. Hence, after a while this part
will completely dominate the other part which has a wave-type evolution. This
behavior is the expression of the instability caused by the slip weakening fric-
tion law. The use of the expression of the dominant part leads to a solution in
which the perturbation has been severely smoothed by the finite wavenumber
integration. The propagative terms are rapidly negligible and the shape of the
slip distribution is perfectly described by the dominant part.
104
5. Stability analysis
One can easily remark that w == 0 is a stable position if >'5 < 0 (i.e. N = 0).
In this case the dominant part w d vanishes and the system has a stable behavior.
Hence it is important to obtain a simple condition on the distribution /3(X1)
which determines the positiveness of the eigenvalue >'5.
Let us suppose in the following that we deal with a homogeneous fault sys-
tem (i.e. /3(xd = canst). In this case the spectrum (>';(/3))n~O is a function
of the non-dimensional parameter /3. Let 0 < /30 ::; /31 ::; ... be the intersection
points of the curves /3 ----7 >'~(/3) with the axis >.2 = 0, i.e. >'~(/3k) = O. The
constant /30 depends only on the geometry of the anti-plane problem (i.e. the
distribution of the faults) and it is independent on all physical entities involved
in our problem. This non dimensional parameter gives quantitatively the limit
between the stable (/3 < (30) and unstable (/3 > (30) behaviors of the fault.
In the case of a homogeneous single fault system (i.e. r f = [-a, aD the
value of /30 was computed by Dascalu et al. (2000) to be /30 = 1.15777388 .....
105
For each representative physical quantity which is included in the non di-
mensional parameter f3 (the friction weakening slope SJ.L'(O) = (J.Ls -J.Ld)S/ L e,
the interface stiffness G/ a, the fault half length a, the elastic bulk modulus G,
etc... ) we can define a "critical" value.
.,go.s
5
(e) 0
0.6
iil 0.4
>
~ 0.2
en O~~~~~~~~~
-4000 -3000 -2000 -1000 0 1000 2000 3000 4000
POSITION (ml
Figure 2. Comparison between two slip rate distributions Otw(t, x, 0) on the fault computed
with a FD method at the same times during the initiation phase. The slip rate corresponding to
the heterogeneous model of the local friction (continuous line) vs the slip rate corresponding to
the homogeneous model with the effective friction law (dashed lines).
(19)
In Figure 2 we have plotted the spatial distribution of the slip rate at different
times during the initiation phase for a heterogeneous fault (solid lines) with
a weakening parameter f31oc. With dashed lines we have plotted the spatial
distribution of the slip rate at the same times on a homogeneous fault with a
weakening parameter f3equiv computed from (19). Note that the homogeous
fault gives a good description of the heterogenous fault at a macroscopic scale.
This ratio represents the information about the weakening rate of the fault,
when it is defined. Let us analyze now the parameter , in three different
cases. If we use the expression of the dominant part (see Campillo and Ionescu
(1997» in the case of the homogeneous and infinite fault we deduce that the
parameter, is constant in all the elastic space, i.e. we have
__ 8yW d(t,x,y) - f3
, (X, y ) - a d( ) - , everywhere in R x R+, (21)
w t,x,y
In the case of a single finite fault the function 'Y has its support in a narrow
band of the size of the fault length. The general shape of'Y defines an initiation
pattern that qualitatively characterizes the unstable behavior of the fault. The
most interesting point is the existence of a domain, including r f' over which
,(x, y) = f3 (see Figure 3). We now define the "domain of confidence" as
107
X103
4. 2
4.25
4. 3
4.35
4. 4
4 .45
Figure 3. Spatial distribution of the derivative of strain with respect to displacement, param-
eter ,(x, y) computed with a FD method, during the initiation on a single finite homogeneous
fault. Note the constant value of, in abroad region around the fault Dc(fJ) called the "domain
of confidence".
As it follows from Campillo et al. (2001), Voisin et al. (2001) in the case of
a heterogeneous fault, described in previous section, close to each individual
fault segment, an initiation pattern develops in the elastic medium, associated
with a local domain of confidence over which ,(XI,X2) = {3loc' But now,
the striking feature is the existence of a wide domain over which ,(Xl, X2)
is nearly constant, independent of the individual fault segments but closely
related to the whole fault system (see Figure 4). All over this domain, we have
,(Xl, X2) = (3equiv. That is, over this wide domain, it is possible to measure
the collective behavior of all the fault segments, similar to the behavior of a
homogeneous fault.
References
Ampuero J.-P., l-P' Villote and FJ. Sanchez-Sesma (2001) Nucleation of rupture under slip
dependent friction law: simple models of fault zone, submitted to J. Geophys. Res..
CampilIo, M. and Ionescu, I. (1997) Initiation of anti-plane shear instability under slip-dependent
friction, J. Geophys. Res.• 102, 20,363-20,371.
Campillo, M., Favreau, P., Ionescu I.R. and Voisin, C. (2001) On the effective friction law of an
heterogeneous fault, J. Geophys. Res .• 106. 88, 16,307-16322.
Cochard, A and R. Madariaga (1994) Dynamic faulting under rate dependent friction, Pure
Appl. Geophys. 142,419-445.
Dascalu, C., Ionescu, R.I., and CampilIo, M. (2000) Fault finiteness and initiation of dynamic
shear instability, Earth Planet. Sci. Lett.. 177, 163-176.
Dascalu C. and I.R. Ionescu (2001) Spectral analysis of instabilities for the wave equation with
application to eathquake initiation, submitted.
Dieterich, J.H. (1979) Modeling of rock friction 1. Experimental results and constitutive equa-
tions, J. Geophys. Res .. 84, 2161-2175.
108
x 10'
1
1. 1
1. 2
1. 3
1 •
1. 5
1. 6
1. 7
1. 8
1. 9
Figure 4. Spatial distribution of the derivative of strain with respect to displacement, param-
eter ,(x, y) computed with a FD method, during the initiation on a heterogeneous fault. Note
the constant value of, in a "collective domain of confidence" Dc ({3equiv) .
Favreau, P., M. Campillo and I.R. Ionescu (2000) The initiation of inplane shear instability
under slip dependent friction. Bull. Seism. Soc. Am. 89, 1280-1295.
Favreau, P., M. Campillo and I.R. Ionescu (2001) The initiation of shear instability in three
dimensions: theoretical and numerical studies, to appear in 1. Geophys. Res..
Geubelle, Ph. and J.R Rice (1995) A spectral method for 3D elastodynamic fracture problem, 1.
Mech. and Phys. Solids, 43, 1791-1824.
Ionescu I.R. and Campillo M. (1999) The influence of the shape of the friction law and fault
finiteness on the duration of initiation, 1. Geophys. Res. , 104, 3013-3024.
Ionescu, I.R., and J.-C. Paumier (1996) On the contact problem with slip dependent friction in
elastostatics, Int. 1. Eng. Sci. 34(4),471-491.
Knopoff L., J.A. Landoni and M.S. Abinant (2000) The causality constraint for fractures with
linear slip-weakening, 1. Geophys. Res., 105, 28035-28044.
Ohnaka, M., Kuwahara, Y. and Yamamoto, K. (1987) Constitutive relations between dynamic
physical parameters near a tip of the propagating slip zone during stick slip shear failure,
Tectonophysics, 144, 109-125.
Ohnaka, M. and Kuwahara Y. (1990) Characteristic features of local breakdown near a crack-
tip in the transition zone from nucleation to unstable rupture during stick-slip shear failure,
Tectonophysics, 175, 197-220.
Voisin C, I. Ionescu, Michel Campillo, Riad Hassani, and Q.L. Nguyen (2001) Initiation process
on a finite segmented fault: a spectral approach, to appear in Geophys. 1. Int ..
AN EXAMPLE OF STICK-SLIP WAVES
Franck Moirot
PSA, Centre Technique La Garenne-Colombe
Paris, France
Quoc-Son Nguyen
Laboratoire de Mecanique des Solides, CNRS-umr7649
Ecole Polytechnique, Paris, France
1. Introduction
It is well known that in frictional contact of solids, a contact point
may have a slip or stick or separation regime according to Coulomb's law
of dry friction. For dynamical problems, the study of the propagation
of these zones on the contact surface may be interesting in different ap-
plications : ultrasonic motors (cf. Zharii, 1996), seismic predictions and
simulations (cf. Cochard and Madariaga, 1995), mechanical vibrations
and instabilities induced by friction (cf. Martins and Simoes, 1995, Vola
et al., 1999, Oancea and Laursen, 1997, Oestreich et al., 1996). In the
particular case of a solid in contact with a moving obstacle, some recent
results of the literature on the flutter instability of the steady sliding
solution (cf. for example Adams, 1995 or Martins and Simoes, 1995 or
Vola et al., 1999) have suggested that we investigate dynamic periodic
responses. A possible transition of this response to a cyclic response in
the spirit of Poincare-Andronov-Hopf bifurcation, cf. (Nguyen, 2000),
is the long term goal. As an example, the study of the phenomenon of
brake squeal has been discussed in this context, Moirot, 1998. A simple
modelling of a drum brake Ie? ds to the study of an academic example of
109
1.A. C. Martins and M.D.P. Monteiro Marques (eds.). Contact Mechanics. 109-116.
© 2002 Kluwer Academic Publishers.
110
encased cylinders, and has enabled us to obtain an interesting example
of stick-slip waves. Such an example is new in the literature.
€= (\7u)s,
Diva = ,iL,
a = (1+ V )(1-2v)Tr(€) 1+ l~v€'
u(~, (), t) = v(~, (), t) = 0, (1)
a rr (1 , (), t) = -p((),t), a r o(l,(),t) = -q((),t),
u ~ 8, p ~ 0, p( u - 8) = 0,
Iql ::; jp, q(l - v) - jpl1 - vi = 0.
Since r = a ± iw, the general expression of Ll. is Ll. = eCtCP(M cos w¢> +
Nsinw¢». There are 4 unknowns (e, W,M,N). The continuity of the
displacement gives V(O) = V(T). The stick region leads to two supple-
mentary equations V'(O) = V'(WT) = -lie. Thus, the following system
of equations must be satisfied :
e
2
> a+ ~ 4h '
Ifg8 + hLl.1 < f( ~ + g8)
{
(14)
g8 - DLl.' ~ 0, 1 + ell.' ~°V¢> E (WT, T),
V¢> E (0, WT).
A family of stick-slip waves is considered for which the continuity of Q
at ¢> = T is ensured :
(15)
This condition implies that hV(T_) = fg8 + 1..f!- and leads to the follo-
wing system of governing equations and inequalities after the elimination
of M and N:
T(l - w) + ~ealJ!T sinw'ItT = 0,
{ 1 + ~ealJ!T sinwWT - ealJ!T cos wwT = 0,
(3- (16)
a -- ~,
-fD
v - a2,
- c2h-a' W -_.1(3
Ifg8 + hLl.1J!1 ::; 1..f!- + fg8, 1 + ell.' ~ 0.
The equations can be solved numerically by Mathematica for different
values of k and f. It is found that e must have the sign of D; these
waves propagate in the sense of the previous exploding perturbed mo-
tions. The value of 8 must be strong enough to avoid possible separations.
114
The results (phase diagram, displacement and reactions) are given res-
pectively in Figure 2 and Figure 3 for the particular case of a common
metal with f = 1, k = 8 and D < O. In this case, the physical velocity
is c = 1255m/s and the wave propagates in the opposite sense to the
implied rotation.
References
Adams, G. (1995). Self-excited oscillations of two elastic half-spaces sliding with a
constant coefficient of friction. J. Appl. Mech., 62 :867-872.
Co chard, A. and Madariaga, R. (1995). Dynamic faulting under rate-dependent fric-
tion. Pure Appl. Geophys., 142 :419--445.
Martins, J. and Simoes, F. (1995). On some sources of instable/illposedness in elasti-
city with Coulomb's friction. In Contact mechanics, New York. Plenum Press.
Moirot, F. (1998). Etude de la stabilite d'un equilibre en presence du frottement de
Coulomb. Application au crissement des freins a disque. These, Ecole Polytech-
nique, Paris.
Nguyen, Q. (2000). Stability and Nonlinear Solid Mechanics. John Wiley and Sons,
Chichester.
Oancea, V. and Laursen, T. (1997). Stability analysis of state-dependent dynamic
frictional sliding. Int. J. Nonlinear Mech., 32 :837-853.
Oestreich, M., Hinrichs, N., and Popp, K. (1996). Bifurcation and stability analysis
for a non-smooth friction oscillator. Arch. Appl. Mech., 66 :301-314.
Renard, Y. (1998). Modelisation des instabilites liees au frottement sec des solides
elastiques, aspects tMoriques et numeriques. These, Univ. Grenoble.
Vola, D., Raous, M., and Martins, J. (1999). Friction and instability of steady sliding
squeal of a rubber/glass contact. Int.J. Num. Meth. Engng., 46 :1699-1720.
Zharii, O. (1996). Frictional contact between the surface wave and a rigid strip. J.
Appl. Mech., 63 :15-20.
115
0.02
0.01
I
I
0.00
0.0
I ~~~_~.J
0.2 0.4 0.6 0.8
~ (radian)
36
p
..
Figure 2 - The phase diagram 32
__ _ _
of the periodic solution associated
with a stick-slip wave is given for
k = 8, f = 1 and 8 = 0.005. :l~ ~_. ~_~~
M U
~
U
(radian)
U U
l-~-/-I
1.5 ---.--.-.-~. ----.- ---~ -- ------,
-I
0.5
t ) .•
Q I I
--0.5
r -.
- 0.5
-1
O.
II "'<.
• dl
l
...
""" ,'
.'I~,
I" I'"~
I'
~.
)6
)5
'~.,
• \1 .. ! . I" H
))
)2
II
10 20 )0 40
22
20
-I
LONGITUDINAL WAVES IN ELASTIC
RODS WITH DISCONTINUOUS CROSS
SECTIONS
Peter Eberhard
Institute of Applied Mechanics, University of Erlangen
Egerlandstr. 5, D-9I058 Erlangen, Germany
Abstract In this paper longitudinal impact waves in an elastic rod with discontin-
uous cross section struck by a rigid sphere are studied numerically and
experimentally. In a rod with finite length, waves move back and forth
along the rod; therefore, it may happen that two incident waves cross
a discontinuity interface simultaneously. Thus, the reflected and trans-
mitted waves induced at the interface will influence the incident waves
due to the wave reflection at both ends of the rod. For these general
cases, to the authors' knowledge, there are no computational methods
for the transmitted and reflected waves. This problem will be addressed
in this paper in some detail. In the numerical simulations, the classical
Hertzian contact law and the wave equation are combined to compute
impact responses. In the experimental setup, advanced instruments like
Laser-Doppler-Vibrometers for measuring the high frequency impact re-
sponses are used. It is shown that numerical results computed by this
approach agree very well with experimental results.
1. Introduction
During the last decades, scientists and engineers have developed an
increasing interest in the solution of problems involving the impact of
solid bodies. Among solid bodies, rigid bodies and linearly elastic rods
are the simplest engineering structures. However, even wave propagation
phenomena in elastic rods resulting from longitudinal impacts by rigid
bodies may be complex due to the dynamical boundary conditions and
117
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 117-124.
© 2002 Kluwer Academic Publishers.
118
wave reflections. Considerable efforts have been put in understanding
the mechanism of wave reflection and transmission in rods. Many results
for longitudinal waves in rods produced by impacts were reviewed by
Goldsmith (1960), Johnson (1985) and Stronge (2000). Experimental
results were reviewed by Al-Mousawi (1986).
When an incident wave crosses an interface between two different cross
sections or materials, a part of the wave is reflected and another part
transmitted. Among the few investigations dealing with longitudinal
impact waves in discontinuous rods, the important references by Rip-
perger and Abramson (1957), Plunkett and Doolen (1963), Habberstad
and Hoge (1971), Barez, Goldsmith and Sackman (1980) are mentioned
here. However, in a rod with finite length, two waves move along the rod
in opposite directions. Therefore, it may happen that two incident waves
cross an interface simultaneously. Further, the reflected wave and the
transmitted wave induced at the interface will influence the two incident
waves crossing the interface later. For these general cases, to the authors'
knowledge, there are no computational methods with one-dimensional
wave theories for the wave propagation in the rod. This problem will
be addressed in this contribution in some detail. Longitudinal impact
waves in an elastic rod with a discontinuous cross section struck by a
rigid sphere are investigated analytically and experimentally.
~q
Vo..
C3 L..---L----''---'--''''''-'-~ m2, E2, v2' A2,P2
~I
(1)
(2)
At the interface of the two homogeneous rods, i.e, at Xl = L1, X2 = 0,
there are constraints on the displacements
(3)
and on the interface forces
(4)
The boundary condition for the inhomogeneous rod at the free end X2 =
L2 reads
oU2{L2, t) = O. (8)
OX2
For the contact end Xl = 0, different impact theories give different
boundary conditions. A comparison of different theories applied to lon-
gitudinal impacts of a rigid sphere striking an elastic rod can be found
in Hu and Eberhard (1999). According to St. Venant's impact theory,
the boundary condition for the contact end reads
However, this theory is not suitable here since the surfaces of the col-
liding bodies are not planar. Instead, a boundary approach combining
both, St. Venant's theory and the Hertzian contact law, is used in the
following. Details about the boundary approach are given in Hu and
Eberhard (1999). According to this approach, the boundary condition
for the contact turns into
-E A 8Ul(O, t) - F( ) (10)
1 1 8 - t.
Xl
The contact force F(t) is given by the static Hertzian contact law,
2
F(t) = ko~(t) (11)
where o+(t) is the indentation between the striking sphere and the struck
end, i.e.,
(12)
The coefficient k is a constant depending on the geometry and elastic
properties of the sphere and the struck end,
4VR (13)
where R, E and 1/ are the radius, Young's modulus and Poisson's ratio
of the striking sphere, respectively, see e.g. Johnson (1985). Now, the
impact problem of the sphere striking the inhomogeneous rod is mathe-
matically described by equations (1-8) and (10-13). In the following this
mathematical model will be further simplified.
The general solutions of the partial differential equations (1) due to
d' Alembert, is
The integral constants C1 and C2 can be set to any values since they have
no influence on the displacements U1(Xl, t) and U2(X2, t). For simplicity,
these constants are set to zero resulting in
91 (e) = 0 for 0 ~ e~ L1 ,
92 (e) = 0 for 0 ~ e~ L2 . (16)
(19)
(20)
Combining the constraints (17) and (18), the boundary conditions (19)
and (20) with the equation of motion of the striking sphere (5), one
obtains a set of nonlinear ordinary differential equations with time delay.
They have to be integrated with very small time steps since the impact
response during contact includes very high frequency components. The
unknown functions are h, 91, h, 92 and q. The initial conditions are
given by equations (6) and (7).
section. The striking sphere was suspended by a thin kevlar wire forming
a 'V'. In the experiments it is released at a vertical height h above
the axis of the rod leading to the initial impact velocity of the sphere
Vo = J2gh. The rod has cylindrical sections with radii r1 = 10 mm,
r2 = 20 mm and lengths L1 = 1 m, L2 = 0.5 m, respectively. The rod is
suspended by two sets of thin kevlar wires at two locations and carefully
leveled horizontally.
The strains were measured with strain gauges and high quality DC-
amplifiers. The strain gauges are of type 3j120XY13, made by Hot-
tinger Baldwin Messtechnik GmbH, and the glue is Z70. The two DC
signal conditioners are of type DMS805B made by Rohrer GmbH with
a frequency range up to 1 MHz. For the measurement of displacements
and velocities, Laser-Doppler-Vibrometers of type OFV-3000jOFV-302
made by Polytee GmbH were used. The measured locations for the
strain signals are marked with letters A, B, C and D in Figure 3 with
the distances 0.2m, 0.8m, 1.2m and l.4m from the struck end of the
rod, respectively. The measured locations for the velocity and displace-
ment signals are marked with x in Figure 3 and the directions of the
velocity mJ
2 2
0.2
2
S
Vl
S 0 0
-
~ q)
E .5
-I 0 .5 -0. 1
0.5
-2 -0.2
0.5 1 1.5 0.5 1 1.5
position [m] position [mJ
~
~ 0.2 ... ... .. .. ... : ..............•.. .. ....... ....•. ........ ... ..
····J\WJiV\\l{
f.1
~ 0 ..... .• .... ... .•.. .. .. .. . .....
.
~-0.2~
., ~~'
.
:
[E~fl~ o 0.5 1
Time [ms]
1.5 2
I:~m
is 0 0.5
Time [ms]
1 1.5 2
measured and the simulated results for the strains at the locations A
and C and the velocity and the displacement at the free end of the rod,
where the simulated results are denoted by black solid lines and the
measured results by gray dashed lines. This also emphasizes that the
computational method using a set of ordinary differential equations with
time delay can be used to predict the longitudinal impact waves with
high accuracy in inhomogeneous rods.
4. Conclusions
In this paper, longitudinal wave propagation in a finite elastic rod
with discontinuous cross sections struck by a rigid sphere is studied nu-
merically and experimentally. A set of ordinary differential equations
with time delay is derived describing longitudinal wave propagation in
the finite rod. Not only is the wave reflection and transmission at the
interface with discontinuous cross sections considered, but also the wave
reflection at both ends of the rod. Experimental results show that the
boundary approach which combines the Hertzian contact law and the
wave equation with boundary conditions satisfactorily predicts the im-
pact wave in the rod with discontinuous cross sections. Moreover, the
computational method developed in this paper may also be used for
rods with different materials, or with both cross sectional and material
discontinuity.
References
AI-Mousawi, M.M. (1986) On experimental studies of longitudinal and flexural wave
propagations: An annotated bibliography, Appl. Mech. Rev. 39, 853 - 864.
Barez, F., Goldsmith, W. and Sackman, J.L. (1980) Longitudinal wave propagation
in axisymmetric structures with material and/or areal discontinuity, Exp. Mech.
20, 325 - 333.
Goldsmith, W. (1960) Impact: The Theory and Physical Behaviour of Colliding Solids,
Edward Arnold Ltd, London.
Habberstad, J.L. and Hoge, K.G. (1971) Effects of discontinuity in a cross section on
an elastic pulse, J. Appl. Mech. 38, 280 - 282.
Hu, B. and Eberhard, P. (1999) Experimental and theoretical investigation of a rigid
body striking an elastic rod, Report IB-32, Institute B of Mechanics, University of
Stuttgart, Stuttgart.
Johnson, K.L. (1985) Contact Mechanics, Cambridge University Press, New York.
Plunkett, R. and Doolen, R.B. (1963) Reflection and transmission coefficients for
stress waves in bars, J. Acoust. Soc. Am. 35, 1457 - 1458.
Ripperger, E.A. and Abramson, H.N. (1957) Reflection and transmission of elastic
pulses in a bar at discontinuity, Proc. 3rd Midwestern Conference on Solid Me-
chanics, 135 - 145.
Stronge, W.J. (2000) Impact Mechanics, Cambridge University Press, Cambridge,
U.K.
THE INFLUENCE OF CONTACT
PROPERTIES ON FRICTION-INDUCED
BRAKE VIBRATIONS
1. Introduction
In recent years comfort features of cars have become more impor-
tant to car customers. Therefore, brake suppliers have to avoid friction-
induced vibrations during the braking action. Vibrations showing fre-
quencies above 1 kHz and a narrow bandwidth are very perceptible by
the human ear and are therefore most annoying. These are commonly
called 'brake squeal'.
As there is no general concept of designing a silent brake from the very
beginning, expensive and time consuming tests have to be conducted to
investigate the noise generated by a particular brake design. Detected
noise is then fought by modifications of the existing design and repeated
tests. This procedure is supported by FE-calculations, which are only
valid for the brake under investigation.
125
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 125-132.
© 2002 Kluwer Academic Publishers.
126
2. Brake model
For the brake a 14-DOF multibody model has been developed. This
approach cannot mirror each detail of the brake design but allows an easy
variation of design features to check their influence on squeal generation.
Furthermore, the results from a rather rough model are more general
than those from a detailed FE model for a specific brake design.
As Figure 1 shows, the model consists of 6 rigid bodies connected
by coupling elements. As rather high frequencies are involved, elastic
modes of certain components have to be considered. This is done at
the carrier by splitting it into two rigid bodies connected by a coupling
element. The stiffness and inertia properties of the disc in the model
are chosen so that the representation in the model has eigenfrequencies
which are close to frequencies of elastic modes of the disc, known from
FEM analysis. The coupling elements connect the bodies by moments
or forces which are dependent on relative displacements and velocities.
These dependencies can be linear or nonlinear; here, only linear charac-
teristics are considered. Frictional contacts are introduced between disc
and pads.
Parameter values have been determined by measurements conducted
at brake components, or by calculations using a validated FE-model of
the brake under investigation.
127
+[
--------
M
+ C3
--------
D
,
Cl
(J.Lh - S}C3 (s - J.L~)~:s + C2 ] [ : ]
,
- 0, (1)
...
c
• • • ::• • iI. • • ~l
Im(>')
rtr~:
o -.~ ..... -:-,,"~ ....
: ,~IJ :
•••••• t····'lt •••••• t•••••••
o Re(>')
a) b)
Figure 2. 2-DOF model of nonconservative restoring forces (a) and related root
locus plot (b).
....
:10 -------J-II-+-------tr-----+------+--------
-------- - --:--------l\------r--------:---------
~ -1~ n ----lt~~~-l---~~~=t===::
! : : : _ _::Jf::::::::t:::::::- :~--l~i--~--~-:::::
i i 0-+-+
: ::::: : ::::::F::::::F:::::-l~---=--t---:::::
~L-~~ __ ~ __ ~ __ ~ __ ~~
o 5 10 15 20 25 30
Eigenvalue index i
Figure 3. Change of eigenvalue real parts due to variations of the parameter group
'piston' (c=stiffness).
kHz, see Treyde (2001). Bearing in mind that the model has a rather
simple structure, and, therefore is valid mainly for the lower frequencies
of squeal, these results are a good confirmation for the approach chosen
in this work.
The main results from the investigations of the parameter influence can
be summarized as follows:
Parameters close to the friction contact play an important role,
increase in damping does not always stabilize the system,
increase of inertia parameters lowers the stability,
increase of the distance between two connecting points has a
stabilizing effect.
(2)
Ignoring viscous damping, the equation of motion for the sliding fric-
tion oscillator reads
(3)
D = - Fn k < o. (5)
2y'ciii
It is clear that, due to the friction characteristic, negative damping
occurs. This damping is dependent on the dynamic parameters m, c, as
well as on the slope k and on the normal force Fn.
16r-~~~~'--'-'~~1
14
12
.... --------t-------- --------1--+--+-1"'--+-
,....... I I I
.~.
I
I., 10
t,
C. 8
........
- 4 :::::::l::::::: _::::::::~::~:::::I::::::::
.:( 6 ,--------t-------- --------;---------1-------
]' ..... --------t--+--+- .--+--~--+-----j--------
4 -----rnn-tn-l"n-n:m--1--nn: : I.T.t
--------T------- --------T--------r-------
-m-r--- ni\2140HZ t-~--
2
2 +------.:--------+-----+-+-- -\--
o . .. .. 1 + + .. ..
-7 -6 -5 -4 -3 -2 -1 5 10 15 20 25 30
ReP'i) [10 3 s- 1] Brake pressure [bar]
a) b)
average value of k = 0.0576 sm- 1 has been taken from the measure-
ments. A stability analysis with changing brake pressure from 0 bar to
25 bar was conducted and it was found that instability occurs at a brake
pressure of about 15 bar and a mode frequency of 2140 Hz.
Figure 4 a) shows the calculation results in a section of the complex
plane.
Acknowledgments
This study has been funded by the 'Bundesministerium fur BiIdung
und Forschung'. The authors are grateful for this support. Furthermore
the authors would like to thank Dr. T. Treyde, J. Korte and A. Stache
of TRW Automotive company for their fruitful discussions.
References
Allgaier, R., L. Gaul, W.Keiper and K Willner (1999) Mode lock-in and friction
modelling, In GAUL, L. and C. A. BREBBIA (editors): Computational Methods in
Contact Mechanics IV, 35-47, WIT Press, Southampton, Boston.
Magnus, K, Popp, K (1997) Schwingungen, B. G. Teubner, Stuttgart.
Mills, H. R. (1938,1939) Brake Squeak, Technical Report 9000 B, 9162 B, Inst. of
Automobile Engineers.
North, M. R. (1972) Disc Brake Squeal - A Theoretical Model, Technical Report,
MIRA. No. 1972/5.
Popp, K, Rudolph, M. (2001) Brake Squeal, In Popp, K. (editor): Detection, Utiliza-
tion and Avoidance of Nonlinear Dynamical Effects in Engineering Applications,
197-225, Shaker Verlag, Aachen.
Treyde, T. (2001) Bremsenquietschen, In Popp, K. (editor): Detection, Utilization
and Avoidance of Nonlinear Dynamical Effects in Engineering Applications, 227-
244, Shaker Verlag, Aachen.
ANALYSIS OF EIGENVALUE PROBLEMS
MODELLING FRICTION: SUFFICIENT
CONDITIONS OF NON-UNIQUENESS FOR
THE ELASTIC EQUILIBRIUM
Riad Hassani
Laboratoire de Geophysique Interne et Tectonophysique, Universite de Savoie
73376 Le Bourget du Lac, France.
Abstract This study is concerned with the Coulomb frictional contact problem in
elastostatics. Introducing a convenient eigenvalue problem, it becomes
possible to establish sufficient conditions of non-uniqueness for the con-
tinuous model. It can be also proven that these sufficient conditions are
fulfilled under appropriate hypotheses.
1. Introduction
In structural mechanics, the Coulomb friction model is currently used
when studying the contact problems. The variational formulation of the
continuous problem in the case of Coulomb's simplified law in elasto-
statics was given by Duvaut and Lions (1972). In this case, the first
existence results obtained by Necas, Jarusek and Haslinger (1980) were
followed by Jarusek (1983), Kato (1987) and more recently by Eck and
Jarusek (1998). In these references, existence is proved if the friction
coefficient is lower than a critical value.
A regularizing procedure of the normal stresses, the so-called "nonlo-
cal Coulomb friction model" introduced by Duvaut (1980) and studied
by Demkovicz and Oden (1982), Cocu (1984), Kikuchi and Oden (1988)
permits to prove existence for all friction coefficients and uniqueness
when the friction coefficient is sufficiently small.
133
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 133-140.
© 2002 Kluwer Academic Publishers.
134
The "normal compliance model" introduced by Oden and Martins
(1985) (see also Martins and Oden (1987), Klarbring, Mikelic and Shillor
(1988, 1989)) leads to similar existence and uniqueness results.
The finite element problem, corresponding to the continuous static
Coulomb friction problem, admits always a solution according to the
works of Haslinger (1983, 1984) (see also Haslinger, Hlavacek and Necas
(1996)). Furthermore, the discrete solution is unique if the friction co-
efficient is small enough (the denomination "small" depends also on the
discretization parameter). Besides, some convergence results of the fi-
nite element solutions to a solution satisfying the continuous model are
established in Haslinger (1983).
Again in the finite dimensional context, several studies using truss
elements have led to examples of non-uniqueness in the static and quasi-
static context (Janovsky (1982), Klarbring (1990), Alart (1993) and Bal-
lard (1999)).
In this study, we propose sufficient conditions of non-uniqueness for
the continuous model by using an appropriate eigenvalue problem. If
the friction coefficient is an eigenvalue of this problem, then the couple
(geometry, material) is candidate to non-uniqueness. In such a case,
we give sufficient conditions for non-uniqueness to be satisfied by the
solution of the continuous model. Since there are to our knowledge no
available analytical solutions, the fulfillment of these sufficient conditions
can be illustrated only numerically by using finite elements.
where u stands for the stress tensor field in 0, c(u) = (Vu + VT u)/2
denotes the linearized strain tensor field, C is a fourth order symmetric
135
and elliptic tensor of linear elasticity and div represents the divergence
operator of tensor valued functions. On r c, we adopt the following
notation: U = unn + Utt and cr(u)n = crn(u)n + crt(u)t. The equations
modelling unilateral contact with Coulomb friction can be written:
Un ~ 0, (3)
Ut = 0 ===}
{ (4)
Ut =1= 0 ===}
Remark: Let us mention that the true Coulomb friction law involves
the tangential contact velocities and not the tangential displacements.
However, a physically meaningful problem analogous to the discussed
here can be obtained by a time discretization of the quasi-static frictional
contact evolution problem.
The variational formulation of problem (1)-(4) consists of finding u
such that
Figure 1. The elastic body n in frictional contact with the rigid foundation.
136
(8)
Moreover, suppose that one of the following two conditions (i) or (ii)
holds:
(i) Ii is an eigenvalue of (5)-(7) with the corresponding smooth eigen-
function <I? and
where
(11)
model do not exist, the numerical computations cannot stand for a rigor-
ous mathematical proof of the sufficient conditions for non-uniqueness.
Our aim here is only to illustrate the methodology given in the contin-
uous context.
As geometry, we choose an unit square n (in Figure 2) with Pois-
son ratio and Young modulus 1/ = 0.4 and E = 10 GPa respectively.
The body lies on a rigid inclined plane and it is loaded by a density of
gravity forces 1 and by an imposed displacement field U. The following
values are chosen: 0 = Arctan 2, <p = Arctan 2.5, 111 = 223.6 N.m- 3 ,
lUI = 0.005385. The discrete eigenvalue problem (11) is solved with an
10 20 30 40 &0
Figure 2. An elastic square lying on a rigid inclined plane, the deformed configura-
tion u~ and the Von-Mises stress field
0.87
0.86
0.85
:'.t
0.84
0.83
0.82
20 40 60 80 100 120 140 160 180 200
llh
o 10
References
Alart, P. (1993) Criteres d'injectivite et de surjectivite pour certaines applications de
Rn dans lui meme : application ala mecanique du contact, Math. Model. Numer.
Anal. 27, 203 - 222.
Babuska, I., and Osborn, J . (1991) Eigenvalue problems, in Handbook of Numerical
Analysis, Volume II, Part 1, P.G. Ciarlet and J.L. Lions, eds., North Holland,
Amsterdam, 641 - 787.
Ballard, P. (1999) A counter-example to uniqueness in quasi-static elastic contact
problems with small friction, Int. J. Engng. Sci. 37, 163 - 178.
140
K.L. Johnson
Department ofEngineering, Cambridge University
Trumington Street, Cambridge CB2 1PZ, U.K.
J.A.Greenwood
Department ofEngineering, Cambridge University
Trumington Street, Cambridge CB2 1PZ, U.K.
1. Introduction
The incentive for this research arose from the development of microprobe
instruments to measure the surface energy and other surface properties of
polymeric materials. Such probes commonly comprise an effectively
141
J.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 141-160.
© 2002 Kluwer Academic Publishers.
142
spherical tip pressed against a flat surface or, as in the case of the Surface
Force Apparatus, two crossed cylinders of equal radii, which are
geometrically and mechanically equivalent to a sphere in contact with a
plane. A typical experiment consists of loading and subsequently unloading
the contact with a controlled force, while measuring the displacement and/or
contact area, until the surfaces separate. If the surfaces adhere during the
compressive part of this cycle, a tensile force is required to pull them apart,
which is referred to as the 'pull-off' force. Microprobe instruments are
particularly advantageous for adhesion measurements for two reasons: (i)
adhesion forces become increasingly significant with a decrease in the size
of the contact; (ii) adhesion is strongly affected by surface roughness whose
effect is minimised by using a very small area of contact. This paper up-
dates an earlier 'progress report' [Johnson (2000)].
For perfectly elastic solids and reversible adhesive forces, the contact
mechanics is well developed through the elastic fracture mechanics principle
of matching the release of strain energy to the work done to separate unit
area of the surface (surface energy). See Johnson, Kendall and Roberts
(JKR) (1971), Derjaguin, Muller and Toporov (DMT) (1975), Maugis
(1992). But the adhesion of viscoelastic solids, notably rubber, is well
known not to be reversible. More work is required to separate two surfaces
than is returned when they come together, an effect known as 'adhesion
hysteresis'. This irreversibility arises from viscoelastic dissipation within the
contacting solids. The mechanism can easily appreciated in a qualitative
way. During separation the work done by the external force has to overcome
the dissipation in addition to doing work against the surface forces (surface
energy). When the surfaces are coming together the surface energy is
providing the driving force, which has to provide for the dissipation as well
as overcoming the external load. Hence the external force on separation
exceeds that on coming together by virtue of the internal viscoelastic
dissipation. In common with any non-conservative system, the response is
dependent upon the loading path.
The results of the JKR and Maugis elastic theories will be briefly
reviewed, since they form the basis of the viscoelastic analysis.
spherical surfaces and (ii) an adhesive tension Pa(a), which gives rise to the
adhesive force Pa. The net contact force P can be expressed by
8 = (a 2 / R) - .J27r Ay a / E * , (2)
_I=_ v12 + __
1- _ 1- v 22
E* E1 E
2 '
8 - - + -_ _ 8
(a)
(b) c c
·1
Figure 1. Adhesive contact of elastic spheres. Surface traction comprises two terms: Hertz
pressure PI and adhesive tension Pa. (a) JKR model; (b) Maugis-Dugdale model.
144
the relative radius of the two spheres lIR = lIRJ + lIR2 and the combined
surface energy of the two surfaces 11 r = rJ + r 2 - rl2· The surfaces snap
apart at a 'pull-off force Pc given by
(3)
(4)
(5)
and
Pc =-112. (6)
The infinite tension at the edge of the contact in the JKR model [Figure
l(a)] leads to difficulties in a viscoelastic analysis since, with a moving
contact boundary, it corresponds to an infinite strain rate. The singularity
has been removed in the contact problem by Maugis (1992), by analogy with
a Dugdale crack in an elastic-plastic material. As shown in Figure l(b), the
tensile adhesive traction is taken to have a maximum finite value 0"0 which is
assumed to be constant in a 'process zone' a ::s; r ::s; c in which the separation
of the surfaces h( r) ::s; ho' where ho is related to the surface energy by
I1r = (joh o· The width d of the process or Dugdale zone is given by
E* h 2
7r
d=c-a::;;.- Q (7)
4 Ar '
provided that d/a « 1, in which case the JKR equations (1) to (6) become
effectively asymptotic solutions to the Maugis analysis.
rate U =Po / to' where Po is the maximum compressive load. Such a ramp
cycle is shown in Figure 2.
To be mathematically tractable, analytical continuum mechanics
modelling of rate-<lependent materials is effectively confined to linear
viscoelasticity. We shall use the simplest model which encapsulates the
essential features of small strain polymer deformation: the 3-element solid. It
is represented in Figure 3 by two springs and a dashpot. At very low rates of
strain the material deforms elastically with a modulus EOC) represented by the
two springs in series. At high rates of strain the deformation is again elastic,
with a modulus Eo = EOC) Ik represented by the single spring. At intermediate
rates, the strain at time t after a step change in stress is given by the creep
compliance function <I>(t) = (1 / EOC» fjJ(t) , where
Figure 2. Ramp cycle in which the load is increased and then reduced at a unifonn rate U.
Separation occurs at the pull-off force Pc.
-
1
r;;
.i
~
~
e-
o(,)
~
~
1 I--~~=-- ------- --------
To
o
10-2
Time tIT
Figure 3. Creep compliance function of a three-element model of a linear viscoelastic
material: equation (8).
146
and T = (1- k) 1] I Eoo is the relaxation time of the material. <D(t) is plotted
in Figure 3. It can perhaps be thought of as the reciprocal of a time
dependent modulus E(t).
If we refer to the surface traction distributions in Figures la or Ib, it is
apparent that there are two different locations of high elastic strain having
very different scales of size. There is the bulk deformation of the contact
arising principally from the Hertz contact pressure, whose size is
characterised by the radius a of the contact. There is also an intense
concentration of strain at the periphery of the contact produced by the
adhesive forces, whose size is characterised by the length d of the Dugdale
zone. During the cycle shown in Figure 2 the representative time for bulk
deformation is to but, for the strain concentration at the periphery, the
representative time is that for the edge of the contact to pass through the
Dugdale zone, i.e. t d = d I a, where the average value of a=ao I to and ao is
the contact size under Po' Hence the ratio of characteristic times governing
the strain rate in the two locations is
~=!!..
to a0
Putting P = Uf and using the creep compliance function of equation (8) gives
1.
C
IU
~O.B
fIl
::s
i3 0.6
f
o ~--~--~--~----~--~--~--~~--~--~--~
o 0.2 0.4 0.6 O.B 1 1.2 1.4 1.6 1.B 2
As we have seen in the JKR elastic theory, the adhesion term can be
obtained by elastic fracture mechanics; in the present problem we tum to
viscoelastic fracture mechanics. The situation is illustrated in Figure 5,
which shows an interface crack with a Dugdale cohesion zone of width d.
The crack may be envisaged as being propagated at a steady speed V by a
frictionless pin in the crack mouth. First consider elastic solids. If the pin
moves to the left, since the stored elastic energy does not change, the work
done by the pin exactly balances the work done against the surface forces l1y.
So we can write:
For elastic solids this result is perfectly reversible and applies equally if the
pin moves to the right and the crack closes. A viscoelastic material, on the
other hand, involves energy dissipation D(V) as the strain field moves
through the solid. If we postulate that D(V) is a fraction a of the input work
G, for an opening crack we have:
or
(12)
In a closing crack the input work comes from the surface energy so that the
equivalent relationship becomes:
G
fJcl == ~
fly
= 1- a cl (V):::; 1 (13)
V(closing)
L(»d)
V(opening)
Figure 5. Representation of a steadily moving (opening and closing) interface crack with a
Maugis-Dugdale adhesive traction
Figure 6. Dugdale process zone for an opening crack; material points move from left to right
with velocity V.
150
With a Dugdale crack 1* equals the Dugdale length d which, in the
visoelastic case, is given by equation (7) with E* replaced by E: I ¢(t) , i.e.
I* 7r E* h2 7r 1
(t * / l).~(t * / T) =- ~(t*) =_....£..Q.... == - - , (15)
cVT 4c /).yVT 4c v
where c is the constant (1/3) or (2In). Schapery then showed that for an
opening crack the apparent work of adhesion is given approximately by:
(16)
The closing crack was examined by Schapery (1989). He showed that the
representative length and time were approximately the same as for the
opening crack (equation (14», but the apparent work of adhesion was given
by:
which gives the neat result: Pop x Pcl ~ 1. Note that this corresponds to the
qualitative equations (12) and (13) if the dissipation coefficients in opening
and closing, a op and acl, are equal.
For a 3-element viscoelastic solid with the creep function of equation (8), the
impulse response function rf>i(t) is given by:
'f',
d {l-k}
"'. (t) == -dt ~(t) =
-T exp( -t / 1) . (19)
If the crack opens with a steady speed V, the point now at a distance x ahead
of the current crack tip would have been a distance x, = x - Vt from the tip a
h(x)=~
20" d
TfE*
f
d/
0'
v
~(t)
,x-
time t ago. It follows that the relaxed (viscoelastic) shape is given by:
-
d
Vt) 20" d
dt+~~(O)F(x/d),
TfE*
(20)
where the second term is the instantaneous elastic response, for which
(J(O) = Eoo* / Eo* == k.
Now the crack opening displacement at the end of the Dugdale zone is
related to the surface energy by: h( d) = ho = dr /0"0' Inserting this
condition, together with the impulse response (19), into equation (20) and
changing the integration variable to u = 1 - Vtld gives an implicit equation
for the true length d of the cohesive zone: i.e.
VTdr ~l
v== E*h 2 =[r~(r,k)] , (21)
co 0
where
11 (r,k) [r
= ~ J~ (1- k) exp[- r(l-u)]F(u)du + kF(1) ]
and r = dlVT. This integral has been evaluated numerically and equation (21)
inverted to obtain d = VTr( v, k).
152
Returning to the contact stress distribution shown in Figure 6, it was first
shown by Barenblatt (1962) that, to maintain finite stresses at the tip of a
crack, the remote stress intensity KJ must be balanced by the stress o(x) in
the process zone according to:
K = f2 roo CJ(x) dx
J 1; Jo..rx '
which, for an opening Dugdale crack, gives
(22)
Since the stresses remote from the tip are relaxed, G is related to KJ by
G = K} / 2E:, whereupon
(23)
Now consider the crack in Figure 6 to be closing, i.e. moving from left to
right. Superposing instantaneous stress distributions to find the relaxed crack
shapes does not work in this case, since the need for contact of the surfaces
behind the crack front would be violated. Instead it is necessary to superpose
instantaneous crack shapes and to calculate the resulting relaxed stresses.
The required shape need not be known ab initio, but will be found in course
of the analysis. Following Schapery (1989) we shall refer to the stress
distribution which would produce this shape in an elastic solid of modulus
E: as the 'reference stress' and denote it by (TR(X]) = (T "g(x] / d), where d
remains to be found.
A point now at x would have been at x, = x + Vt a time t ago, so that the
relaxed stress at x is given by
(24)
153
where If/i(t) is the impulse relaxation function and 1fJ(0) = 11k. Making use of
the reciprocal nature of the relaxation and compliance functions, equation
(24) can be inverted to obtain the reference stress:
(:l(x/d)-=()og(x/d) = Jo
r(d-X)IV
¢i(t)()
(xd+ Vt) dt +¢(0)CJ(x/d).(25)
For a Dugdale crack relevant values of O"{ (x+ Vt)/d} are equal to 0"0 which,
when substituted into equation (25) and integrated, gives
~
hex) = 2(} fdg(X' /d){2
Jr E: 0
Ix -In ..rx -HU
1t..rx +H11dx' . (27)
At the end of the Dugdale zone h(x) = h(d)= ho = LirlO"o which, when
substituted in equation (27) gives an equation for the Dugdale length d.
Values of r -= d / VT as a function of v and k have been found numerically
by Greenwood (2001). It may be shown that the remote 'effective' stress
intensity factor is determined by the reference stress if rather than by the
true stress at the crack tip. Thus:
(28)
(29)
decreases with speed of closing. Plots of the Schapery equations (16) and
(17) are added to Figure 8 with the coefficients cop = 4.5 and Cel = 6 taken to
give a best fit with the exact results of equations (23) and (29). For the case
of a closing crack, computed values of the reference stress cI(x) compared
with the true stress a(x) in the vicinity of the crack tip for r = 10 and k =
0.01 are shown in Figure 9, from which it is seen that the reference stress
approaches the true stress away from the tip. Similar results have been found
by Baney and Hui (1999).
10 2 r----r----.----.----~--~----~--~----=
'tI
:S
....fi
CD
10 1
~
eq
=
CD
,
U
0
, ,
t ,
, ,
, ,
,,
10 1 10 2 10 3 10 4 10 5
relationship between load and contact size is given by equation (4). In the
viscoelastic case the surface energy Liy is scaled by the factor Pel during
loading and by POP during unloading, for which Schapery's approximate
equations (16) and (17) will be used. In dimensionless variables equation (3)
becomes
- -3 r::=3
p= a -12/Ja (30)
or
* IT = In~ .
-k
t (31)
1- ¢J(t * I T)
-g.
'-'
ctl
tlo
.E!
10 1
~
g.
~
ci- 10 0
II
IQ.
A1VT u~ VT
Ivl = --.--z = - . -
EC1J ho Eo A1
Figure 8. Ratio Pof apparent / true surface energy for a moving crack.
k =0.01; --Exact, eq. (23) & (29); -.-.-.-.- Schapery, eq. (16) and (17)
with empirical shift factors cop = 4.5 and Cel = 6.0.
156
Substituting in equation (15) gives
where daldt is the sQ..eed of the edge of the contact, which corresponds to the
crack speed Vand d P / dt = p" / to = U, whereupon
(33)
1 "",'
/". \
I · . . .
. .t . .... :.........\ ....... :........ :...... .
I~~ ... ~........ ~.'~.~>,.~:
0.8
0.6
i :
.......~ .... _.. .
(1 : ..................... :
bO i· . .,....-.,;"..
"-
b 0.4 i ...... . :................ :........ : ..... -
;
gj ; .
Q)
; I ;
-0.4
-1 -0.5 o 0.5 1 1.5
x
Figure 9. Closing crack, k = 0.01, VTld = 10. Reference stress d(x) [eq. (25)] compared with
true Maugis-Dugdale stress o(x).
157
3
lea
2.5
...
~
fIl
"(j
ea
....""CJ 2
....ea . . .
. fl-1- . .
§ 1.5 '" "f.~ ..: .... : .... : ..
CJ
"(j
Q)
fIl
. . .. . . . . . '.. . . . . . . . . . ... . .
~ 1 '.' '.' ~
e
~ ... ... ... ,' .... : .... ..
0.5 ,', ,', ',
o
-6 -4 -2 o 2 4 6 8 10 12
Normalised load P
(a)
6
'()
5 .. ·.
.... , .. · .
~ 4 .. .'.
~
CJ 3 ... .. .. ·.
lIS
...i
"(j
2 ..,. ..
-6 -4 -2 o 2 4 6 8 10 12
Normalised load P
(b)
Figure 10. Simulation of ramp loading I unloading cycle. k =0.01; A =JO, 500.
(a) Variation in contact radius a; (b) variation in displacement (5.
158
The simultaneous equations (30), (31) and (33), can be integrated step-
by-step to give the variation of P with a and hence the variation of contact
size with load. During loading (crack closure) P= Pel given by equation (16);
during unloading (crack opening) P = Pop given by equation (17). In an
experiment, it is commonly the displacement 0 rather than the contact size
which is measured. It is given by equation (5) with the surface energy scaled
by the factor Pwhich, in dimensionless variables becomes
(34)
Computations have been carried out using the creep compliance function
in equation (8), with k = 0.01 and C = Cel = cop = 1/5, for various values of the
loading rate parameter A, where
(35)
The variation of contact size with load is plotted in Figure lOa. The elastic
(JKR) reversible relationship is shown by the chain line. At finite loading
rates adhesion hysteresis is clearly visible. When the surfaces first touch they
snap into contact at an initial size given by the elastic theory with E* = E; .
With increasing compressive load the contact grows; its size is less than the
JKR value since Pel < 1. When the load is first reduced, the contact size
continues to grow slightly until the surfaces begin to separate at the higher
apparent surface energy associated with an opening crack (Pop> 1). The
contact size then decreases until pull-off occurs at an enhanced pull-off
force. The contact size approaches its high speed values at the beginning and
end of the cycle. This is on account of the cube-root relationship between a
and P which, for dP/dt = constant, gives high values of dP/da when a is
small.
Having determined the variation of contact size a with load, the variation
of displacement ois given by equation (34), see Figure lOb.
8. Conclusion
more work is required to separate the surfaces than is extracted when they
come together and the process is rate dependent.
In the absence of adhesion, the contact mechanics of linear viscoelastic
solids is well established through the work of Ting (1966) and others. As in
the elastic case, the presence of adhesive forces at the interface introduces a
fracture mechanics problem at the edge of the contact area, where a Iprocess
zonel of length d can be identified. Provided d « a, where a is the contact
radius, the strain rates in the process zone are much greater than in the bulk.
This enables viscoelastic effects on the adhesive force to be separated from
those in bulk, as in elastic fracture mechanics. This condition is likely to be
satisfied with probes giving a contact size measured in ~, but may not be
with an Atomic Force Microscope in which a may be of order nm. The
analysis in this paper is simplified further by assuming that the bulk
deformation is carried out sufficiently slowly to show an approximately
elastic response.
Linear viscoelastic fracture mechanics developed from the work of
Schapery (1975) and (1989) who introduced the concept of a 'representative'
time t* and 'crack length' 1* (=Vt*), comparable in magnitude with the
process zone length d. He then showed how t* and the work of adhesion
G(V) for both opening and closing cracks could be found approximately
from the creep compliance function ¢i..t) of the material [equations (14), (16)
and (17)].
To investigate the accuracy of Schapery's results, Greenwood has
conducted a rigorous viscoelastic analysis of opening and closing 'Dugdale'
cracks, i.e. where the adhesive stress distribution has a constant value aD
over the process zone d, in which the separation of the surfaces is less than
h o • Good working agreement is found with Schapery's equations if 1* is
taken to have the value 4.5d when opening and 6d when closing, or even 5d
for both.
Finally, Schapery's equations have been used to simulate an adhesion test
in which a spherically tipped probe is brought into contact with a
viscoelastic solid, loaded at a slow uniform rate to a maximum compression,
and then unloaded at the same rate until the surfaces separate.
References
Johnson, K.L. (2000) in Microstructure and microtribology ofpolymer surfaces, Am. Chern.
Soc. Symposium Ser.741, 24-41.
Maugis, D.l (1992)J.Coll.Interface Sci. 150,243-269.
Schapery, R.A. (1975) Int.J.Fracture, 11,141-159,369-388.
Schapery, RA (1989) Int.J.Fracture, 39, 163-189.
Ting, I.C.I. (1966)ASMEJ.Appl.Mech. 33, 845-854.
A MODEL OF ADHESION
ADDED TO CONTACT WITH FRICTION
1. Introduction
A simple and realistic law of normal and tangential adhesion (in-
cluding decohesion) is added to the classical law of unilateral contact
with threshold friction. The motivation for this extension is the study
of debonding and damage in fibrous composites by means of computa-
tional and experimental methods. The adopted framework is the con-
stitutive theory of standard generalised materials of Moreau (1973) and
Halphen and Nguyen (1975). More specifically, the model is inspired by
the pioneering work of Fremond (1987) and subsequent improvements
of Raous, Cangemi and Cocu (1997) on surface adhesion, as well as the
Lemaitre and Chaboche (1985) precedent on the analogous phenomenon
of damage in the bulk.
The proposed model differs from existing ones by the use of an ir-
reversible adhesion gap !f (similar to a plastic strain) as the damage
internal variable instead of the usual dimensionless fraction {3 of intact
adhesive links. By energetic duality, there corresponds a dual stress pa
to ga representing the adhesion intensity, which simultaneously derives
from an adhesion interface energy density q, and a dissipation power
161
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 161-168.
© 2002 Kluwer Academic Publishers.
162
2. Contact mechanics
2.1. Kinematic and static variables
Given two bodies 0 C R3 and 0' C R 3 , the gap vector g between a
striker and a target particles in regards, within the hypothesis of small
slips and strains, is simply defined together with its rate by:
g =y-y'
. .
g=y-y ,
., (1)
p=Pnn+pt, (3)
2.2. Tribology
A basic tribologicallaw, such as unilateral contact or dry friction, is a
relationship p between the kinematic variables g, g and the static stress
vector p, which typically is set-valued:
P E p[g,gl . (4)
Addition of secondary irreversible processes such as running-in, wear or
decohesion requires the introduction of additional internal state vari-
ables. Let gi be such a generic kinematic variable and pi its dual flux
(in the sense that pi . gi is a power). The tribological state law p and
the complementary law pi governing the evolution of gi take the form:
(5)
Adopting the framework of standard generalised materials, tribological
laws can be equivalently defined by a free energy potential w[g, gil, and a
dissipation power function <1>[g, gi; g, gil. For a broad class of materials,
the state and complementary laws are derived from these two potentials:
pi E -OgiW[g, gil
{ . . . . (6)
pl E Ogi<1>[g, gl; g, gl]
PM
{ P:Pt E -Og:~:g~'sgt,:a'cgt)
E Og;<I>(g ,gti9 ,Pn)
(17)
Therefore contact and friction are mounted in series as usual, whereas
adhesion is added in parallel.
p{ = pC(p;,)bt(gt - gio)) ,
(21)
pa = pad = p[_oo,s(ga)]baPR+(llgll - go))d
p{
slip
adherence It f
gr
I
gi'
- - ~I I-lP~
PM~
Finally, the expression of the contact force results from summing the
contributions of each component:
p = p~ n + p{ + pa = I~PR - (gn) n
(22)
167
5. Benchmark problems
After deriving the appropriate Jacobian matrices for the generalised
Newton method of resolution (Curnier and Alart, 1988), the unilateral
contact with friction and adhesion law is implemented in the finite-
element code TACT. Numerical simulations of simple test problems of
unilateral contact and tangential friction coupled with adhesion are run.
The results displayed in figures 3 and 4 illustrate and validate the law.
30
Pn Pn
4
20
10
0.2
-10
-0.2 -0.1 0.4
-20
-2
-30
-40
-4
-50
Figure 3. Numerical simulations of penalised contact and penalised contact with adhesion.
60
Pt 60
Pt
40 40
20 20
-20 gt -20 gt
-40 -40
-60 -60
Pt
60
t~
40
20
-20
-40
-60
Ti 300
350
250
I weak load
) no debonding
~
200
150
100
a perfect adhesion
50 friction
~z
10 15 20 25 30
I
~
Ti 300
350 medium load
250
200 b) partial debonding
150
100 adhesion-decohesion
50 friction
10
~
15 20 25
----z
30
~ L_________
--_ ___._
strong load
c) complete debonding
Ti friction
5 10 15 20 25 30 Z
Figure 5. Pull-out of a fibre from a matrix. Analytical (plain lines) and FE simulation
(dots) of interfacial shear stress Ti = Pt versus z.
References
Curnier, A. (1994). Computational Methods in Solid Mechanics. Kluwer.
Curnier, A. and Alart, P. (1988). A generalised newton method for contact problems
with friction . J. Theor. Appl. Mech., 7(1):67- 82.
Fremond, M. (1987). Adhesion of solids. J. Theo. Appl. Meeh., 6(3) :383- 407.
Halphen, B. and Son, N. Q. (1975) . Sur les materiaux standards generalises. J. Meca,
14:39- 63.
Kim, J.-K. and Mai, Y-W. (1996). Modelling of stress transfer across the fibre-matrix
interface. In Numer. anal. and model. of compo mat., pages 287-326. Chap. & Hall.
Lemaitre, J. and Chaboche, J. (1985). Mecanique des materiaux solides. Dunod.
Moreau, J. (1973). On unilateral constraints, friction and plasticity. In Capriz and
Stampacchia, editors, New var. teeh. in math. phys., II-73, pages 175-322. Crem.
Raous, M., Cangemi, L., and Cocu, M. (1997). Un modele couplant adherence et
frottement pour Ie contact entre deux solides deformables. Cptes Rend. A cad. Sc.
Paris, 325:503- 509. Serie IIb.
ANALYSIS OF A 'WALKING' PUNCH
D. A. Hills
Department of Engineering Science, University of Oxford, Parks Road,
Oxford, OX1 3PJ.
A. Sackfield
Department of Mathematics, Nottingham Trent University,
Burton Street, Nottingham, NG1 4BU.
1. Introduction
The problem to be studied is shown schematically in Figure 1. A
rigid punch rests on an incompressible half plane. A force, P, is applied
to the centre of the punch, and a cord, attached to the base of the punch,
is used to develop a constant shearing force, Q (Q < f P, where f is the
coefficient of friction). We assume that the position of the normal force,
s, then varies harmonically with time, so as to cause the punch to rock.
If the punch is of half-width, c, we will assume that the maximum extent
of the load offset, So, is less than c/2, so that complete contact is made
throughout. The object is to track out the stick/slip regime, shearing
traction distribution and slip displacement as a function of time. The
conditions under which 'walking' proceeds will then be found, and an
estimate of the size of the 'steps' found.
169
f.A. C. Manins and M.D.P. Monteiro Marques (eds.). Contact Mechanics. 169-178.
© 2002 Kluwer Academic Publishers.
170
s p
-c x
2. Formulation
2.1 Starting Conditions
First, push the punch normally onto the surface with s = 0, so that
the punch face is parallel with the half-plane, and the contact pressure
is given by
p
p( x) = ---;=~~ (1)
7rv'c2 - x2
Now apply a shearing force, Q. Providing there is complete adhesion,
this will lead to a shearing traction, Hills et. al. (1993)
Q
7rv'C2 - X 2
q(x) = (2)
given by
.!. avo = _!
A ax 7r
l c
-c t -
pet) dt
x
(5)
where
A=1-v. (6)
J-L
p(x;s)=.;)2P 2
2SX]
[ 1 +-
2 . (8)
7r C -x C
q(x)
p(x) =
(Q)
P c2
c2
+ 2sx' (9)
q(x) ~
p(x)
(.9...)
IP
[1 + 2SX]
c
.
2
(10)
3,0
2,5
"
,.:1
"
2,0
i\:
l,.,. 1.5
d'
sIc =-0,25,0,0,25"k:\ 1.0
\... .
"', = O~~C~-=='OC,,~ 0.5
-1.0 -0.5 0
x!c
Figure 2, Plot of the dimensionless normal pressure, cp( x) I P, along the contact for
various positions of the normal load, sic
adhesion while
!L
fP
< 1- 21s1
c . (11)
If this inequality does not hold, and assuming, from now on, that s <
0, we expect stick to occupy the region -c S x S b. The equation
connecting the slip displacement, g(x), to the shear traction distribution
is
(12)
~ dg _ 1+
A dx -
.!.l
7r
b
-c X - ~ ,
q'(~)d~ (13)
where
I-
-
.!.l
7r
c
-c 7rVc2 -
fP
~2
[1 + 2S~]
c
.-!!:L _ -2sf P
x- ~ -
2 7rC 2 .
(14)
q'(x)=2s f P
7rC 2
Jb-X,
c+ X
(15)
173
Q/fP
1.0
0.6
0.4
0.2
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
sIc
Figure 3. Plot of the initial extent of the slip zone (b/c) as a function of the position
of the normal load, s / c, and Q/ f P
Also, the connection between the shear force, Q, and the stick zone
extent, b, is given by the tangential equilibrium equation
The integral here is trivial, and provides the connection between the
dimensionless stick zone size, blc, and the load offset, 81c:
8 QlfP-l
(17)
c l+blc'
Note that, for complete contact I 8 I Ic < 1/2. If slip is incipient blc -+ 1
and equation (11) is recovered. As the value of QI f P is increased the
value of bI c decreases linearly, until, as Q I f P -+ 1, bI c -+ -1, which is
as required, for -1 I 2 < 81 c < O. The full stick regime and extent of the
initial slip zone are plotted out in Figure 3.
It may be verified that the shear traction in the stick region is every-
where less than the limiting value.
The next phase in the development of the solution is to establish the
relative slip displacement, g(x), in the slip region, b < x < c. Now,
equation (13) holds everywhere throughout the contact patch, although
the integral no longer assumes a Cauchy Principal Value. When evalu-
ated this gives
1 dg _ 28fP ~
A dx - - ---;;? ;+c V b< x < c. (18)
174
ltglx)/2fPA
2.5
blc
-0.75
2.0
-0.5
2sfPA [ -..jx2+(c-b)x-bc-!(b+c)ln(b+c)+ 1
g(x)= 7rC2 !(b+c)ln(2..jx2+(c-b)x-bc+2x+c-b) '
(20)
and is plotted out in Figure 4. This, then, is the value of the relative
slip displacement in the slip zone, at any point during the first quarter
cycle of the loading history. To measure the 'step' length it is useful to
keep a record of the displacement at some particular point on the punch
surface. The most obvious point is the origin, so that, providing b < 0
() ~fflf#ffl#ffl#fflffffflfr 1
~ c
sdl
-so< s "I
-s.+ £ bfflfffflfffflffffflfff#fff'1
~ c
'---Ib,
~(X)l /
-~ .. -bo C
~
Figure 5_ Evolution of the stick/slip regimes
extreme position of the load, the stick region occupies (-C, bo ) and the
slip region (b o , c)_
Now rock the punch back, i.e. tilt the punch in a clockwise direction.
An infinitesimal rotation will cause the value of the contact pressure
to increase within the slip region, hereby causing stick, but adhesion
will still be maintained within the stick region, and hence the entire
interface is adhered. A further, finite, rotation will leave the problem
substantially unchanged until, at some new positive value of 8, slip will
start from the left hand end of the punch. Suppose that the value of
8 is increased still further, to a value 82, (0 < 82 < 8 0 ) and determine
the position of the new stick-slip transition point, b2 . The slip region
will now occupy the region (-c, b2) and stick will occupy the region
(b 2 , c). Within this new stick zone there will be a 'locked in' effect on
the relative displacements due to it previously being part of a slip zone.
Equation (8) continues to give the pressure distribution, p(x; 81), which
is uncoupled from the shearing traction, equation (12) continues to apply,
and therefore equation (13) becomes
.!. Je q'(~)d~ =
'Ir x- ~
2fP
'lrC
(81_Jx -+ b), b <
C X C
2 X < c. (23)
b2
} ]
(24)
Hence
(25)
where
(26)
and
(27)
Note that we must have b2 > bo in order for k2 < 1. This function
does not appear to be capable of analytical integration, but numerical
integration, to give the slip displacement is possible.
177
4. Walking Problem
The procedure described above gives the state of slip, shearing trac-
tions and slip displacement during the first three quarter cycles of the
rocking punch problem. Clearly, the solution is becoming increasingly
complex, because, with each rock, the expression for the 'locked in' slip
displacement is becoming more complex. It does not, therefore, seem
feasible to continue this solution towards a steady state. It is, how-
ever, possible to infer certain features from the solution obtained so far.
These are, first, the condition for the punch to remain stationary (which
is weaker than the condition for the punch to adhere). Secondly, an ap-
proximate measure of the 'step length' when the punch is walking may
also be found.
A good first approximation to the point representing the transition
from remaining stationary to walking is to determine the set of conditions
which cause the slip zones emerging from each corner of the punch just
to meet at the centre. Clearly, if the slip zones never touch, central
particles will remain adhered, and there can be no rigid body motion. It
will not be easy to determine the exact conditions for meeting, because
the problem will, inherently, contain a certain element of asymmetry,
but a sensible simple approximation would appear to be to determine
the condition where the slip zones move just as far as the centreline
(x = 0), from either side, i.e. Ib/cl - t O. The condition for this is that
walking cannot occur providing that
(28)
5. Conclusion
An analytical solution has been developed for the transient problem
of a punch, initially at rest and with a central normal load, subject to
an oscillatory moment of constant amplitude. It has been possible to
produce an analytical description of the behaviour of the punch for the
first three quarters of a cycle of loading. From this, the condition for
the punch to remain stationary, and an estimate of the initial step length
when slip occurs has been found. It is expected that these results will
form the basis of a fuller, numerical solution for further steps, against
which it may be calibrated.
References
Muskhelishvili, N.!. (1953) Some Basic problems of the Mathematical Theory of Elas-
ticity, Noordhoff, Gronigen.
178
Gradshteyn, 1. S. and Ryzhik, 1. M. (1980) Tables of Integrals, Series and Products,
Academic Press, New York.
Hills, D. A., Nowell, D., and Sackfield, A. (1993) Mechanics of Elastic Contact,
Butterworth-Heinemann, Oxford.
Sackfield, A., Truman, C. E., and Hills, D. A. The tilted punch under normal and
shear load (with application to fretting tests), Int. J. Mech. Sci., in press.
PHENOMENOLOGICAL MODEL
OF FRICTION ACCOUNTING FOR
SUBSURFACE PLASTIC DEFORMATION
IN METAL FORMING
Stanislaw Stupkiewicz
Institute of Fundamental Technological Research
Swi~tokrzyska 21, Warsaw, Poland
Zenon Mr6z
Institute of Fundamental Technological Research
Swi~tokrzyska 21, Warsaw, Poland
Abstract The paper deals with the effect of subsurface plastic deformation on
the evolution of the real contact area and on friction in metal forming
processes. A model of evolution of real contact area is derived within a
phenomenological framework by considering a thin homogeneous surface
layer representing the asperities and the layer of inhomogeneous defor-
mation induced by deforming asperities. The friction model is then
derived by assuming asperity flattening and ploughing mechanisms to
occur at different scales.
1. Introduction
Accurate modelling of contact interactions is crucial for reliable simu-
lations of metal forming processes. The real contact area fraction, which
is an important state variable governing the phenomena of friction, lu-
brication and heat transfer, is usually high in these processes due to
high contact pressures. An additional increase of the real area of con-
tact is caused by plastic deformation of the underlying bulk material
which significantly accelerates asperity flattening process. This effect is
observed experimentally and also predicted by micrcrmechanical models.
The corresponding models have been developed on the basis of the slip
line method (Sutcliffe, 1988), the upper bound approach (Wilson and
Sheu, 1988; Kimura and Childs, 1999) and finite element solutions {Ko-
179
l.A.C. Martins and M.D.P. Monteiro Marques (eds.J, Contact Mechanics, 179-186.
© 2002 Kluwer Academic Publishers.
180
v'
workpiece PN~
/
tool tool
(a) (b)
,,,,,,,,,,,,,,,
PN
- v
Figure 3. Dual asperity model: (a) asperity interaction, (b) average local stresses
at real workpiece asperity contacts.
iv) elastic strains are neglected and rigid-plastic, rate independent ma-
terial model is assumed for the bulk and the surface layer.
bulk material
92 [1 + 92 (1 _ R2) ( Eo ) 2] R2 + 91 + 92 R2 _ 1 = 0, (5)
39~ 91 T e~ax N 9~ T
The yield condition (5), also called the asperity flattenin9 condition,
can be resolved for Eo, so that Eo = Eo(PN,Fr,a). In view of the
definition (6) of Eo the following evolution law of the real contact area
fraction is obtained
(7)
This evolution law has successfully been verified against existing micro-
mechanical models and experimental data, cf. Stupkiewicz and Mroz
(2001). In Fig. 5 the predicted evolution of a as a function of plastic
strain in the bulk ell is compared to the results of asperity crushing
experiment of Sutcliffe (1988). In this experiment, wedge-like asperities
were compressed with a constant normal pressure and lateral tension
was applied, as depicted in Fig. 5.
4. Friction model
Let us now apply the dual asperity concept, Fig. 3(a), to derive a
friction model combining the two basic asperity deformation mechanisms
of flattening and ploughing. The average dimensionless local contact
stresses at workpiece asperity contacts, cf. Fig. 3(b), are given by
p~ = PN = PN , ~ = Fr = PT . (8)
a ak a ak
The effect of bulk plastic deformation on asperity flattening and the
related growth of real contact area a is discussed in the previous section.
185
Figure 5. Real contact area fraction 0: as a function of lateral strain ell: model
predictions ("1 = 1.20) and experimental data (Sutcliffe, 1988).
where J.t is the friction coefficient and m is the friction factor. Combining
eqns. (9) and (8) the following limit friction condition is obtained
(a) (b)
Acknowledgments
This work has been partially supported by the Polish State Committee
for Scientific Research (KBN), Grant No.8 T07A 022 20.
References
Avitzur, B. and Nakamura, Y. (1986). Analytical determination of friction resistance
as a function of normal load and geometry ofsurface irregularities. Wear, 107:367-
383.
Bay, N. (1987). Friction stress and normal stress in bulk metal forming processes. J.
Mech. Working Technol., 14:203-224.
Ike, H. (1995). Plastic deformation of surface asperities associated with bulk defor-
mation of metal workpiece in contact with rigid tool. In Raous, M., Jean, M., and
Moreau, J., editors, Contact Mechanics, pages 275-286. Plenum Press, New York.
Kimura, Y. and Childs, T. (1999). Surface asperity deformation under bulk plastic
straining conditions. Int. J. Mech. Sci., 41:283-307.
Korzekwa, D., Dawson, P., and Wilson, W. (1992). Surface asperity deformation dur-
ing sheet forming. Int. J. Mech. Sci., 34(7):521-539.
Mr6z, Z. and Stupkiewicz, S. (1998). Constitutive model of adhesive and ploughing
friction in metal forming processes. Int. J. Mech. Sci., 40:281-303.
Stupkiewicz, S. (2001). Extension of the node-to-segment contact element for surface-
expansion-dependent contact laws. Int. J. Num. Meth. Engng., 50:739-759.
Stupkiewicz, S. and Mr6z, Z. (1999). A model of third body abrasive friction and wear
in hot metal forming. Wear, 231:124-138.
Stupkiewicz, S. and Mr6z, Z. (2001). Phenomenological model of real contact area
evolution with account for bulk plastic deformation in metal forming. Int. J. Plast.
(to appear).
Sutcliffe, M. (1988). Surface asperity deformation in metal forming processes. Int. J.
Mech. Sci., 30(11):847-868.
Wanheim, T., Bay, N., and Petersen, A. (1974). A theoretically determined model for
friction in metal working processes. Wear, 28:251-258.
Wilson, W. and Sheu, S. (1988). Real area of contact and boundary friction in metal
forming. Int. J. Mech. Sci., 30(7):475-489.
THERMOELASTIC INSTABILITIES IN
AUTOMOTIVE DISC BRAKES -
FINITE ELEMENT ANALYSIS AND
EXPERIMENTAL VERIFICATION
D.L.Hartsock
Ford Scientific Research Laboratories
Dearborn, MI48121-2053, USA
1. Introduction
Automotive disc brakes are susceptible to a low frequency vibration
known as 'judder' or 'roughness', particularly during high speed engage-
ment (Kubota et al., 1998). Major improvements in noise reduction in
other areas of automotive design have reduced customer tolerance of
brake noise, resulting in a significant volume of warranty claims asso-
ciated with brake judder (Steffen and Bruns, 1998, Lee and Dinwiddie,
1998).
Various mechanisms have been proposed to account for brake jud-
der, including rotor imbalance, non-uniform accumulation of frictional
transfer films, core shift and disc thickness variation, but considerable
evidence has now accumulated to show that it is associated with an un-
stable interaction between thermoelastic distortion and frictional heat-
ing known as Thermoelastic Instability or TEl (Jacobsson, 1999, Yi et
al., 2000). Temperature measurements of noisy brakes typically show
a non-uniform circumferential temperature variation (Kreitlow et al.,
1985; Thoms, 1988, Little et al., 1998) and examination of brake discs
after engagement reveals evidence of focal hot spots distributed approx-
imately evenly around the circumference (Anderson and Knapp, 1989).
The feedback mechanism responsible for TEl is illustrated by the flow
diagram of Figure 1. Frictional heating during braking causes thermoe-
lastic distortion, which in turn modifies the contact pressure distribution
p(x, y, t) and hence the distribution of frictional heating q(x, y, t). Cou-
pling between the mechanical and thermal problems is introduced by the
energy balance relation
q(x,y,t) = jVp(x,y,t) , (1)
where j is the coefficient of friction and V is the sliding speed. It is
clear from Figure 1 that the product jV functions as the gain in the
feedback process and it follows that there will generally be a critical
speed Vcr above which any given sliding system will be unstable (Dow
and Burton, 1972). Above the critical speed, non-uniform perturbations
in the temperature field will grow, leading to a characteristic pattern of
hot spots or bands on the brake disc.
Ideally, we would like to be able to design the brake system so that
the operating conditions always lie below Vcr. However, Vcr is signifi-
cantly affected by the geometry of the system (Lee and Barber, 1993)
and hence it is generally necessary to use numerical methods (typically
the finite element method) to obtain appropriate design estimates. Two
broad categories of numerical solution have been attempted. In numer-
ical simulations, the system is discretized by the finite element method
and the nodal temperatures are evaluated after each of a succession of
189
/
Solution of contact
problem
~ Frictional heat generation
q(x,y,t) = jVp(x,y,t)
v = sliding speed
f = friction coefficient
pad
hat
section
....-
rotor
where Kf3, Pf3, cf3 are the thermal conductivity, density and specific heat
respectively of material f' and wf3 is the angular velocity of body f' in
the given frame of reference. If we denote the rotor as f' = 1 and the
pad/ caliper assembly as f' = 2, then WI will be the rotational speed of
the rotor and W2 will be zero.
Substituting (2) into (3), canceling the common exponential factor
and discretizing the resulting equation by the finite element method
leads to the matrix equation
(K + C + bH)9 + Q = 0, (4)
where 9, Q are the vectors of nodal temperatures and nodal heat sources
respectively,
and N(r, </J, z), W(r, </J, z) are vectors of shape and weight functions re-
spectively, defined in the domain Of3.
192
The nodal heat sources Q are non-zero only at the contact interface,
where they result from frictional heating. For the disc brake geometry,
the sliding speed V varies with radius r, being given by V = wr. The
discrete form of equation (1) is therefore
Q = jVipP, (6)
where P is the nodal contact force normal to the contact interface which
is defined only at the Nc contact nodes, V is the diagonal Nc x Nc matrix
defined by
(7)
ip is an Nc x N matrix defined by
(8)
(10)
e
where E (-1,1) and'fJ E (-1,1) are dimensionless local coordinates in
and perpendicular to the corivective direction respectively. The resulting
shape and weighting functions are compared in Figure 3 for a = 1.
w,
Figure 3. Comparison of the shape function and the upwind weighting function for
a bilinear quadrilateral element.
4. Experimental investigation
The evolution of hot spots in a caliper disc brake during drag braking
was monitored using an array of fiber optic cables connected to single-
point, two-colour, infrared (IR) detectors. The detectors were housed in
liquid nitrogen cooled dewars to maintain a constant temperature. The
IR system is described in detail by Hartsock et al. (2000).
Five sensors were arranged along a radial line on each of the inboard
and outboard sides of the rotor. During braking, each sensor sweeps
the entire circumferential distribution of temperature at a given radius
with every revolution, permitting a time history of the complete surface
temperature distribution to be reconstituted from the recorded data.
The IR system was set to take temperature readings at a frequency of
2000 Hz, which corresponds to one reading every 4.2 degrees of rotation
at the highest rotational speed used (1400 rpm). This was considered
to give adequate resolution of the spatial distribution of temperature for
typical hot spot patterns.
The use of two-colour detectors reduces the sensitivity of the system
to emissivity, but does not completely eliminate it. The system was
therefore calibrated against a set of thermocouples embedded 0.5 mm
below each surface of the disc. The rotor was subjected to 200 burnish
stops and several long drags to condition the surfaces. It was then heated
to 600°C by dragging at low speed and calibration readings were taken
during subsequent cooldown at 30 rpm with the brakes released and no
external cooling air. During this period, the surface of the rotor was
195
o test data
- 9th order polyfit
500
~ 400
~
::J 300
~
Q)
~ 200
.Sl
-200L-----~-------L-------L------~----~
o 2 4 6 8 10
voltage (V)
10'
\
o
Qj)
Figure 6. Logarithmic plot of average amplitude against time for the mode n = 7.
The slope of the straight line fit corresponds to a growth rate R(b) = 0.0368 S-1.
The dimensions of the basic brake design tested are shown in Figure 7.
The rotor contains 37 vents (not shown in Figure 7) of thickness 12 mm.
The vent:vane ratio is 4:1 and hence the vents occupy 80% of the total
circumference of the rotor at the mid-plane. For the standard brake, the
pad arc length <pp = 66°, but tests were also conducted on brakes with
reduced arc lengths of 53° and 40°. Other design modifications tested
included the reduction of the total rotor thickness dr (by machining a
standard rotor) from 28 mm to 21 mm and of the pad friction material
thickness dp from 10 mm to 5 mm.
Theoretical predictions of the critical speed were obtained using the
finite element method of section 3. In order to model the vents, 74
elements were used around the circumference of the disc, resulting in
the model of Figure 8, which contains about 3000 eight node solid el-
ements. Some approximation is involved in this representation of the
vents, since in reality the vent boundaries move in the given frame of
199
E v a K k
CPa x10- 6 deg K- 1 Wlm deg K mm 2 /s
Rotor 112.4 0.25 13.25 57.0 17.2
Pad 2.03 0.35 30.0 0.93 0.52
Backing plate 200 0.3 12.0 42.0 11.9
Experimental result
3-D FEA model
600
•
_500
-
E
c.
....
- 0400 • 0 .
0
CI)
• 0 0 ~
~
!!!.co 300
(.)
• • •
0
:0=
.~
(.) 200
100
2 468 10 12
hot spot number
element method also predicts 7 hot spots, with a critical speed of Yc~ =
316 rpm.
Similar comparisons were performed for brakes with other values of
the parameters <pp, dp, dr and the results are summarised in Table 3.
In all cases the predicted number of hot spots n* is within ±1 of the
experimentally observed value and the maximum error in the predicted
critical speed is 27%, which represents an exceptional level of agreement
for such a complex problem.
The experimental results show that reducing the pad arc angle <pp
from 66° to 53° has comparatively little effect on the critical speed, but
that a further reduction to 40° increases it by a factor of 2.4. This
increase is associated with an increase in the number of hot spots from 7
to 10, which is close to the minimum number that ensures that at least
one hot spot is located in the contact zone at all times (Lee and Barber,
1993).
Reducing the pad thickness dp reduces the critical speed significantly.
This is attributable to the increased stiffness associated with the thinner
layer of compliant friction material. It suggests that brakes are more
likely to exhibit judder when the brake pads are significantly worn.
These effects are clearly identified by the finite element analysis. Other
investigations, not reported here, have shown that the FEA also cor-
201
rectly identifies the effect of changes in the geometry of the hat section
on susceptibility to judder.
6. Conclusions
We have shown that a finite element implementation of the eigenvalue
method gives generally excellent predictions of the critical speed for TEl
for a caliper disc brake and also successfully identifies the effect of minor
design changes on the critical speed. The analysis of one brake design
by this method takes about 5 hours CPU time on a workstation. By
contrast, no results have been reported for direct numerical simulation
of a brake system with anything like the present level of geometric detail.
We conclude that the eigenvalue method is the best numerical tool for
TEl brake design presently available.
Acknow ledgments
The authors are pleased to acknowledge support from the Ford Mo-
tor Company and from the National Science Foundation under contract
number CMS-9619527.
202
References
Anderson, A.E. and Knapp, R.A. (1990) Hot spotting in automotive friction systems,
Wear 135, 319-337.
Christie, I., Griffiths, D.F., Mitchell, A.R. and Zienkiewicz, O.C. (1976) Finite element
methods for second order differential equations with significant first derivatives,
International Journal for Numerical Methods in Engineering, 10, 1389-1396.
Dow, T.A. and Burton, R.A. (1972) Thermoelastic instability of sliding contact in
the absence of wear, Wear 19, 315-328.
Du, S., Zagrodzki, P., Barber, J.R. and Hulbert, G.M. (1997) Finite element analysis
of frictionally-excited thermoelastic instability, J. Thermal Stresses 20, 185-201.
Hartsock, D.L., Fash, J.W., Dalka, T., Smith, G., Yi, Y-B, Hecht, R. and Dinwiddie,
R.B. (2000) Development of a High-Speed IR System for Temperature Mapping of
a Rotating Target, 14th Annual International Symposium on Aerospace/Defense
Sensing, Simulation and Controls. Thermosense, Orlando, FL.
J.C.Heinrich, J.C., P.S.Huyakorn, P.S. and O.C.Zienkiewicz, O.C. (1977) An 'upwind'
finite element scheme for two dimensional convective transport equation, Interna-
tional Journal for Numerical Methods in Engineering, 11, 131-143.
Jacobssen, H. (1999) Analysis of brake judder by use of amplitude functions, SAE
99-01-1779.
Kreitlow, W. Schrodter, F. and Matthiii, H. (1985) Vibration and hum of disc brakes
under load, SAE 850079.
Kubota, M., Suenaga, T. and Doi, K (1998) A study of the mechanism causing high
speed brake judder, SAE 980594.
Lee, K and Dinwiddie, R.B. (1998) Conditions of frictional contact in disk brakes
and their effects on brake judder, SAE 980598.
Lee, K and Barber, J.R. (1993) Frictionally-excited thermoelastic instability in au-
tomotive disk brakes, ASME J. Tribology 115, 607-614.
Little, E., Kao, T-K, Ferdani, P. and Hodges, T. (1998) A dynamometer investigation
of thermal judder, SAE 982252.
Steffen, T., and Bruns, R. (1998) Hot spot formation on passenger car disks, ATZ
Automobiltechnische ZeitschriJt 100, 408-413.
Thoms, E. (1988) Disc brakes for heavy vehicles, Proc.Int. Conf. Disc Brakes for Com-
mercial Vehicles, Inst.Mech.Eng. (London), C464/88, 133-137.
Yi, Y-B., Barber, J.R. and Zagrodzki, P. (2000) Eigenvalue solution of thermoelas-
tic instability problems using Fourier reduction, Proc.Roy.Soc. (London) A 456,
2799-2821.
Zagrodzki, P. (1990) Analysis of thermomechanical phenomena in multidisc clutches
and brakes, Wear 140, 291-308.
Zagrodzki, P., Lam, KB., Al-Bahkali, E.and Barber, J.R. (2001) Nonlinear transient
behaviour of a sliding system with frictionally excited thermoelastic instability,
ASME J. Tribology, in press.
FRICTION COEFFICIENT EVOLUTION DURING
WEAR TESTS RELATED TO INFRARED
EMISSION
Nadine Stalin
Laboratoire de Mecanique des Solides, CNRS-UMR.7649, Ecole Polytechnique,
91128 Palaiseau cedex, France.
Presently at Direction des Constructions Navales, Etablissement S.c., Site du Mourillon,
BP 30, 83800 Toulon Naval, France.
1. Introduction
2. Experimental methodology
LOADING
t
BALL
INf'RARED THERMOGRAPHY
15:06:18 «15:06:22 .441 5:06: 26 4C'I 5:06:30.4115:06:34.4:1 5:06:38.4415 06:42 «15.06:46 4Sl5:06:5CI.«15.06:54 41
.!l .J -!..l
Figure 2. Sequence of thennal maps
~
~
0,7
0,6
1;'i 05
e '
- - - ..:
- - - - - - - - - - - - - - i1
---- -_1-
:
!:.
~: ~~
-*------~-
--------
0,7
0,6
0,5 ....
~ 04 +-___ d __ :I
I:
.
.. : ····1
!!!'(*i rl_+-~'-II--\
l:'
, ,:il'! r
0,4
0,1 -
o
o 20 40 60 80 100 120 140 160 180 200 220 240
time(s)
The loading dropped because there were some debris in the contact
interface and this probably modified the control of the loading. When the
sliding velocities are larger, the loading is almost constant: see Figures 6 and
7, where v=17 mm1s and v=120 mm1s respectively.
4,5 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0,9
0,8
3,5 - 0,7
0,6
0,5 ...
0,4
1,5 ---------------"-----. ------- 0,3
0,5 0,1
°° 20 40 60 80 100 120
time(s)
140 160 180 200 °
220
28
26
0,9
24
22 0,8
20 0,7
18
E 16
0,6
....
~ 14 0,5
f-<12
<I 0,4
10
8 0,3
6 0,2
4
0,1
2
0 0
0 10 20 30 40 50 60 70 80 90 100 110
time (s)
We can observe a very short transient phase and a large stabilised phase
for both evolution curves. The friction coefficient f increases in the
beginning then it becomes stabilised (average value is 0.6) very quickly
(only 20 seconds after the beginning of the test). The curve corresponding to
!1Tmax has the same evolution although with some delay. The similarity
between the evolution of the friction coefficient and that of the temperature
is quite remarkable.
4. Conclusion
Numerical friction studies require a friction coefficient which is generally
assumed to be constant. The objective of this paper is to check this
hypothesis and it is shown experimentally that instead the friction coefficient
varies continuously in the range of 0.1 to 0.6.
Technological progress in infrared thermography has changed the way
one studies thermal effects in mechanics and, in particular, in frictional
contact problems where direct measurements are difficult.
A new approach based on indirect infrared measurements of temperature
fields and on direct measurements of the friction coefficient has been
210
proposed. The thennal emissions, observed in real time on a surface parallel
to the contact zone, pennit an understanding of the evolution of the friction
coefficient under the contact zone. These results show that the friction
coefficient varies continuously during the tests. Transient and stabilised
phases are observed and we have measured large variations of the friction
coefficient at the beginning of the tests. It varies in the range of 0.2 to 0.5 in
Figure 6 and 0.3 to 0.6 in Figure 7. The nonnalised pressure probably
exceeds the elastic limit which is approximately equal to 3, as shown by
Johnson (1992). In the wear tests the changes in the contact area arise from
indentation, wear or plastic defonnations of the layers near the surface
during repeated sliding contact. The proposed method opens up new
horizons for evaluating the local friction coefficient and for interpreting wear
phenomena.
References
Barghava V., Hahn G. T. , and Rubin C. A. (1985), A.S.ME. 1. Appl. Mech., 52,67-82.
Block H. (1937), Theoretical Study of Temperature Rise at Surfaces of Actual Contact under
Oiliness Conditions, Proc. Inst. of Mechanical Engineers General Discussion of
Lubrication, 2, 22-235.
Dang Van K., and Maitoumam H. (1993), Steady-state flow in classical elastoplasticity :
Applications to repeated rolling and sliding contact, 1. Mech. Phys. Solids, 41,1691-1710.
Dang Van K., and Maitoumam H. (1994), Elasto-plastic calculations of the mechanical state
in reciprocating moving contacts : Application to fretting fatigue, Mechanical
Engineering Publications, London, 161-168.
Dang Van K., Maitoumam H. and Prasil B. (1995), Elastoplastic analysis of repeated moving
contact, application to railways damage phenomena, Wear, 196, 77- 81.
Harish G., Szolwinski M. P. , Farris T. N. and Sakagami T. (2000), Evaluation of fretting
stresses through full-field temperature measurements, Fretting Fatigue: Current
Technology and Practices, edited by Hoeppner D. W. , Chandrasekaran V. and Elliott C.
B. for Testing and Materials, West Conshohocken, PA, ASTM STP, 1367, 423-435.
Jaeger J. C. (1943), Moving Sources of Heat and the Temperature at Sliding Contacts, Proc.
R. Soc. N. S.w., 76, 203-224.
Johansson L. (1992), Elastic and Thermoelastic Contact Problems with Friction and Wear,
Linkoping Studies in Science and Technology, 266.
Johnson K. L. (1992), The application of shakedown principles in rolling and sliding contact,
J. Mech. A. Solids, 11,155-172.
Luong M. P. (1999), Infrared thermography of macrostructural aspects of thermoplasticity,
IUTAM Symposium on Micro and Macrostructural Aspects of Thermoplasticity, 0. T.
Bruhns and E. Stein, Kluwer Academic Publishers, 20, 437-446.
Sakagami T., Ogura K., Kubo S., Lesniak J. R., Boyce B. R. et B. I. Sandor B. I. (1997),
Visualization of Contact Stress Distribution Using Infrared Stress Measurement System,
SPIE, 3056, 250-259.
Stalin N., and Dang Van K. (1997), Numerical simulation of the sliding wear test in relation
to material properties, Wear, 203-204, 180-186.
A CONTACT FORMULATION FOR ELECTRICAL
AND MECHANICAL RESISTANCE
G. Zavarise
Department o/Structural Engineering and Geotechnics
The Turin Polytechnic
Abstract In this paper a formulation to deal with the Finite Element modelling of
coupled electric-mechanical contact resistance is presented. The electrical
contact resistance is based on a micro-macro approach, i.e. the contacting
surfaces are not considered as perfectly smooth, but are characterised at the
microscopic level by the presence of valleys and peaks. Such roughness
reduces the real contact area to a very small percentage of the apparent one.
The microscopic geometry thus plays a very important role in perturbing any
type of field (mechanical, thermal, electrical etc.) in the contact zone. Some
macroscopic models, based on microscopic characterizations, have been
proposed to describe the normal and tangential contact stiffness and the
thermal contact resistance. Within such models, the surface microrugosity of
the asperities in contact is taken into account by replacing each surface by a
statistically equivalent one. The load-flattening mechanical behaviour of the
contacting asperities is modelled by choosing a perfectly plastic constitutive
law as well as an elastic one. In this way the mean plane approach is obtained
as a function of the applied load variation.
The smallness of the real contact area causes a perturbation also of the electric
field. The flux lines must concentrate where the asperities are in contact. In
analogy with the thermal contact resistance this electrical constriction
resistance is studied, supposing a flux tube around each asperity, and choosing
a suitable geometry for its narrowing at the contact zone. The contributions,
due to each flux tube, are added in parallel to determine the electric resistance
of the apparent contact area. The contact constitutive law so obtained, is then
consistently linearized, in order to keep the quadratic convergence of the
Newton-Raphson iterative scheme.
211
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 211-218.
© 2002 Kluwer Academic Publishers.
212
1. Introduction
During the last few decades the family of contact problems has been
growing in complexity more and more, due to the progresses made in
advanced fields such as fusion technology, astronautics, metal forming
processes etc ..
In several applications, contact conditions have to be satisfied very
accurately, i.e. global parameters like the Coulomb friction coefficient have
to be replaced with more accurate relationships. This means that the global
contact behaviour, expressed through contact constitutive laws, has to be
based on the real physical behaviour of the contacting surfaces.
In this study attention is focused on the development of a constitutive law
suitable for contact problems with coupled electric and mechanical fields.
2. Model set up
_ Nse _ 1 m 2 exp - d 2 / a 2
n,e - A - 16 a 2 eric d/,[ia '
(1)
where d is the mean plane distance, m is the mean asperity slope and
a is the surface roughness.
The mean radius of the circular contact zone is
-a =V;
[8 m
a exp[ 2a2
d 2} dJ
ric ( ,[ia . (2)
(3)
where p is the pressure and H the hardness. From eqs. (2) and (3) the
ratio between real and apparent contact area can also be written as
(4)
Comparing eqs. (3) and (4) we obtain a relationship between the applied
pressure and the mean plane approach
(6)
where d v is the diagonal of the indenter and C/' C2 are two parameters
depending on the material. Assuming that the contact micro hardness of the
surface being penetrated by the asperities of the harder surface is the same as
the Vickers micro hardness (cOrre.10nding to the equivalent Vickers
indentation diagonal) we have d v = 2Jra, which combined with eqs. (2)
and (6) gives the asperity hardness
(7)
The electrical constriction resistance can be studied like the thermal one,
by supposing a flux tube around each asperity in contact, and choosing a
suitable geometry for its narrowing at the contact zone. The contribution of
each tube flux is then added in parallel to determine the electric resistance of
the apparent contact area. The most suitable narrowing is the flat circular
one, and for a disk and a half-space the constriction resistance can be
expressed as (see Holm (1981))
1
R=-- (9)
I 4k*a'
c =_2_k_*a_; = 2k*a i
c, 'If; (I -a;/bi ys . (10)
Applying this result to each ideal tube, we obtain the conductance of the
joint:
(11)
and by recalling eq. (1), (2) and (3) we obtain, after some algebra
h.
(
= k' m
2(5
(1- VnrP)-]'S 2
exp(- d /(2(52 )) .
Jii (12)
(13)
product of the apparent contact area and the apparent pressure. Recalling eq.
(8) we have
(15)
Substituting equation (5) in eq. (12) we obtain the exact form for the
electric current
=A k*m (1-
a 2a
where gv is the contact voltage jump. Note that the relationship depends
on the apparent contact area, the mechanical and the electrical gap, i.e. it is a
function which may be easily expressed in terms of primary variables.
217
Master segment
Acknowledgments
The authors wish to thank the support of MURST Res. Contr. "40% -
Contact mechanics: constitutive laws for interfacial phenomena and discre-
tization techniques", and EURATOM/ENEA Res. Contr. "EFDAlOO-521".
218
References
FHBucher
Institutfor Luft- und Raumfahrttechnik. Technische Universitiit Berlin.
Marchstr. 12. 10587 Berlin
R S Dwyer-Joyce
Department of Mechanical Engineering. University of Sheffield.
Sheffield. S1 3JD
Abstract In this work the real area of contact between two rough surfaces has been in-
vestigated by means of ultrasonic experiments and numerical modelling. A
wave of ultrasound is partially reflected from a rough surface interface. The re-
flection can be related to the stiffness of that interface. However, the stiffness
does not uniquely define the real area of contact within the interface. The first
loading of the rough pair was shown to occur plastically. A numerical contact
model has been adapted to approximate plasticity at the asperity contacts. The
contact pressure at the asperity peaks is truncated and the plastically flowed
material distributed into nearby valleys. Predictions of the interface stiffness
from this model compare favoumbly with those measured in the ultrasonic ex-
periments. This combined approach provides a way of either validating rough
contact models or to help in the interpretation of measured stiffness results.
1. Introduction
K=I~· (2)
We use the term nominal pressure,Pnom (i.e. load divided by nominal area
of contact) to distinguish from the real contact pressure, Pr (the load divided
by the real area of contact) which develops at the contact regions.
The measurement of the reflection coefficient thus gives a method for de-
termining the stiffness of an interface. In parallel we use a numerical contact
model to determine the relationship between the applied load on the surfaces
and their approach and hence the stiffness of the interface according to equa-
tion (2).
2.2. Apparatus
The signal from the loaded interface is divided by that from an unloaded
interface. This gives the reflection coefficient as it varies with frequency.
The spring model (equation I) is then used to determine the stiffness of the
interface, K (which is independent of frequency). Further details of this ex-
perimental procedure along with the development and validation of the ap-
proach is given in Dwyer-Joyce et al. (2001).
applied load
In this work the lower specimen is grit blasted aluminium (H30TF). The
upper specimen is hardened steel (EN24) which has been ground. The di-
ameter of the nominal contact patch is 9 mm. Here, the idea is to load the
rough aluminium against the smooth hard steel such that the majority of the
deformation will occur on the aluminium surface.
uz+h-u z
I:: {= 0 point is in contact ~ p >0
(3)
> 0 point is not in contact ~ p =0
with Uz as the normal displacement, h as the surface geometry and Oz as the
rigid body displacement. The system of equations in the plain strain contact
model for the contact points is given by:
When using the half space assumption for modelling the contact of rough
bodies, one is restricted to pure elastic behaviour of the material.
Considering higher spatial frequencies of the rough surfaces, leads to unrea-
sonably high surface stresses and lower real area of contact (see figure 3).
Even at low loads the material close beneath the surface will plastically de-
fonn due to the very high local pressures. Assuming that every local contact
spot is a small Hertzian contact, one can introduce a maximum value for the
local nonnal pressure (i.e. a shakedown limit), above which the material in
the vicinity of the asperity is assumed to plastically defonn.
Bucher et. al (2000) introduced an iterative algorithm to keep the nonnal
pressures below that maximum pressure value and still maintain equilibrium.
However this does not take into account the flow of material from the plasti-
cally defonned asperity tips into the neighbouring valleys (as observed in the
experiments of Pullen and Williamson 1972).
.....\-----.-..--------.-..
PI--------~~ ..~---
Pm..
/: pressure
undeformed
\ surface profile
For this work, a new procedure for approximate consideration of the mi-
cro plasticity has been developed. The changes in the local contact spots that
had elastic pressure values larger than Proax were considered in the following
way: a very small change in the asperity height was applied; the curvature
radius was increased and the "defonned mass" was distributed amongst the
neighbouring nodes (see Figure 2). This was done for every load considered
until the solution contained no local pressures larger than Pmax.
I 1.2
I
I
1
i
0'
~
Fhi;;;;.;-l
0.' -+-FIrltUnIoad
0.4
..•.. '."""'lDod
··.··SecondU""*'
- .. ·ThinlLoId
.j
0.2
-0-. Third UnIoacI
~ ~ w ~ ~ M ~ U
NomIMI Contact P.....un, GP.
Fig. 3: Interface stiffness for an aluminium steel contact as it undergoes three cycles ofload-
ing and unloading.
Surface profIles were recorded from the specimen before loading took
place. The numerical contact procedure described above was perfonned.
Figure 4 shows the pressure distribution predicted for two elastic cases (with
differing threshold roughness wavelengths) and a plastic case. Figure 5
shows the original and defonned surface profIle after the plastic loading.
The stiffness determined from the numerical model is shown in figure 6
compared with the first loading cycle extracted from figure 3.
The agreement between the experimental stiffness and the modelled re-
sult for Pmax = 1.2 GPa and considered maximum spatial frequency of
30 microns is good. The selection of maximum spatial frequency is not par-
ticularly significant. The stiffness of the interface is dominated by the con-
tact of the taller peaks; the small wavelengths (high frequencies) have little
contribution to stiffness. It is clear that significant plasticity is taking place,
and that this is modelled reasonably well by the approximate procedure
adopted.
226
li-fU,-ll,lk,---;fl-;--~-~-;---~~i~ -
8 Pnom = 0.24 GPa - - At = 8 ~m. elastic case
-1--!J-l-.
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
8
- - A t = 30 ~m, elastic case
~6
8:.
4
Q.2
~ 0~-4~~-.~~-.~~~~~-.~~~~~~L-~
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
- - At = 30 ~m, Pma, = 0.9 GPa
+0.02 , - - - - - - - - - - - , - - - - - - - - - - - - - ,
undeformed
+0.01 deformed after full loading with Pnom =0.333 GPa
E
oS +0.00
N
-0.01
E:1. 1.75
1.50
CO
a.
Q, 1.25
''c:""
Q)
1.00
0.75
~
Q) 0.50
.g" 0.25
~
.5 0.00
0.0 0.4
0.1 nominal prgJsure [GPa] 0.3
Fig. 6: Stiffness detennined from the numerical model - different maximum pressures and
maximum spatial frequencies.
The ultrasonic method gives only interface stiffness, which does not
uniquely determine the real area of contact. This interpretation of the results
using a numerical model provides a way of calibrating the experimental ap-
227
proach. Figure 7 shows the predicted real area of contact and local pressure
for the model with plasticity parameters chosen to best-fit experimental
stiffness measurements.
1.0 1.0
0.9
A,= 30 ~m, p.... = 1.2 GPa 0.9
: : 0.7 0.7~
III
~ 0.6 0.8£1.
E
c:
0.5 0.5 ~
8 0.4 0.4 I!!~
a.
'00 0.3 0.37ij
I!!
""~ 0.2 0.2
0.1 0.1
M M
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
nominal pressure [GPa)
Fig. 7: Real area of contact and real pressure as determined from the numerical model.
5. Conclusions
(i) A numerical model has been developed to predict the real area of contact
between rough surfaces undergoing elastic-plastic deformation. Plastic-
ity is modelled approximately using a limited pressure and surface flow
algorithm.
(ii) Experiments have been performed to investigate some rough surface
contacts using ultrasonic reflection. The method can be used to deter-
mine the stiffness of the rough surface contact.
(iii) The model and experiments agree well for appropriate selection of plas-
ticity parameters. This suggests that this simple approach to plastic flow
gives a realistic approximation.
(iv) The interpretation of stiffness measurements with a numerical model
provides a method for determining real area of contact.
6. Acknowledgements
The authors would like to thank: Dr Bruce Drinkwater and Mrs Amy
Quinn of the University of Bristol who carried out the experimental work.
Thanks also to Professor Knothe of the Technical University of Berlin who
organised the exchange programme which facilitated this joint research
work.
228
References
Bucher, F. H., Theiler, A., Knothe, K., 2000, ''Nonnal and tangential contact problem of sur-
faces with measured roughness", Contact Mechanics and Wear of Rai/lWheel Systems, 5th
International Conference Tokyo, Japan, 25-28th July 2000, to appear in: WEAR
Dwyer-Joyce, R. S., Drinkwater, B. W., and Quinn, AM., (2001), "The Use of Ultrasound in
the Investigation ofRough Surface Interfaces", ASME Journal ofTribology, Vol. 123, pp.
8-16.
Drinkwater, B. W., Dwyer-Joyce, R. S., and Cawley, P., 1996, "A Study of the Interaction
between Ultrasound and a Partially Contacting Solid-Solid Interface", Proc. Royal Soc. A
London, Vol. 452, pp. 2613-2628.
Greenwood, 1. A. and Williamson, 1. B. P., 1966, "Contact of Nominally Flat Surfaces",
Proc. Royal Soc. A London, Vol. 295, pp. 300-319.
Kalker, J. 1., "On the rolling contact of two elastic bodies in the presence of dry friction", PhD
thesis, Delft.
Kendall, K. and Tabor, D., 1971, "An Ultrasonic Study of the Area of Contact between Sta-
tionary and Sliding Surfaces", Proc. Royal Soc. A, Vol. 323, pp. 321-340.
Pullen, 1. and Williamson, J.B.P., 1972, "On the Plastic Contact of Rough Surfaces", Proc.
Royal Soc. A, London, Vol. 327, pp.l59-173.
Tattersall, H.G., 1973, ''The Ultrasonic Pulse-Echo Technique as Applied to Adhesion Test-
ing",1. Appl. Phys. D6, pp. 819-832.
Venner, C. H., Lubrecht, A A, 2000, "Multilevel Methods in lubrication", Elsevier
Wang, G., 1991, "Rollkontakt zweier viskoelastischer Walzen mit Coulombscher Reibung",
PhD thesis, TU Berlin
Webster, M. N. and Sayles, R. S., 1986, "A Numerical Model for the Elastic Frictionless Con-
tact of Real Rough Surfaces", Trans. ASME, 1. Trib., Vol. 108, No.3, pp. 314-320.
PREDICTION OF WHEEL WEAR FOR RAIL
VEHICLES - METHODOLOGY AND
VERIFICATION
Tomas lendel
Div. ofRailway Technology, KTH
S£-10044 Stockholm, Sweden
Mats Berg
Div. ofRailway Technology, KTH
S£-/0044 Stockholm, Sweden
Abstract This paper presents a simulation tool for systematic prediction of consecutive
wheel wear of rail vehicles. The methodology includes steps like load
coJlective design, vehicle-track interaction simulation, wear modelling, wheel
profile updating and in-field verification.
1. Introduction
In the railway field the vehicle wheels wear when they run on the rails.
The degree of wear depends on horizontal curve radii, type of vehicle
bogies, lubrication etc. Wheel (and rail) wear is not only a maintenance
problem but also significantly affects the vehicle dynamic performance in
terms of safety against derailment, dynamic stability, ride comfort etc.
It is therefore of interest to investigate how wheel rim shapes (profiles)
change due to wear over the time for various railway applications. In
particular, simulation of such consecutive wear has a great potential. Such
simulation tools are rare today, but see for instance Pearce and Sherratt
(1991) and Li and Kalker (1998), partly due to rather long simulation times
but also due to a lack of methodology and verification. In this paper both a
methodology and in-field verification are presented, see also lendel (2000).
229
J.A.C. Martins and M.D.P. Monteiro Marques (eels.), Contact Mechanics, 229-236.
© 2002 Kluwer Academic Publishers.
230
2. Methodology
design geometry
Track data: { rail profiles
track irregular.
Coefficient of friction
Braking and acceleration
Vehicle traffic conditions
Figure 1. Flow chart of the proposed wheel profile wear prediction tool.
'"?' I
~
:"
I
~ I
dy006. :-' ,d1 ·6.
dyo·2.
dyoo.
I
I Wheel tread
I I
I
I
I
i I t£
i\
~
11"< , -
~-
i I ' I
II
I I
I I
~
;-
Rail gauge
II
comer "-
I
I I
I
I
f'--
I
~ I
i
I
I I
I
I
I
I
I
I
I
I ~ I
Figure 2. Example of wheel-rail match: nominal VIC S I 002 wheel profile and nominal
UIC60 rail profile, inclined 1:30. dy =£\y. Units in mm. Results from the MBS software
GENSYS, DEsolver (2000).
232
Then we need to model the normal and tangential contact problems. Here
Hertzian theory respectively Kalker's simplified theory, Kalker (1990) and
the software FASTSIM, are used. This choice implies, among other things,
that the contact area is elliptical and small compared to the overall size of the
bodies and their radii in the contact vicinity. For wheel tread - rail top
contacts, the chosen contact model is good. However, for wheel flange - rail
gauge comer contacts the choice may be more questionable. But since we
will include thousands of contact problems for each wear step, cf. Sections 2
and 4, an averaging effect will appear. Methods like BEM (software
CONTACT, Kalker (1990)) and FEM are not realistic in the present context
with today's computers.
The chosen wheel-rail contact model as well as multibody models of the
vehicle and track at hand, define a dynamic model for vehicle-track
interaction simulations. Note that due to the "stiff' wheel-rail contact, time
steps are usually only about I ms. This means that even a single simulation,
e.g. describing the vehicle negotiating a curve of a certain radius, is not a
simple computational task. The present vehicle-track simulations are
performed with the multibody dynamics software GENSYS, DEsolver
(2000). See also, for instance, Andersson et al. (1999) for general
information on rail vehicle dynamics.
Figure 3. Definition of quantities in the contact ellipse for wear calculations. The vehicle
travels in the positive ~-direction.
~s = kp(ru)/H (1)
where k is the wear coefficient, p the contact pressure (here given by the
Hertzian solution) and H the hardness of the softer material. ru is the sliding
distance experienced by a material point on the wheel profile when passing
through an element. This sliding distance is given by
(2)
where VsJip is the magnitude of the slip vector VsJip, M the time it takes to pass
through the element, and Vvehic/e is the vehicle speed.
The wheel wear due to one wheel revolution is found by adding the
contributions from the elements along each longitudinal strip as implied in
Figure 3.
The wear coefficient k in (1) is generally a function of slip velocity,
pressure, temperature, contact environment etc. Laboratory tests have been
made on wheel and rail steels to determine the wear coefficient, Telliskivi
and Olofsson (2001). The tests were performed under dry, room temperature
conditions for various slip velocities and pressures. Both disk-on-disk and
pin-on-disk machines were used in order to cover a wide range of velocities
and pressures. Results from these tests, and measurements on similar steels,
gave the wear chart according to Figure 4. As can be seen from Figure 4 the
wear chart has been divided into four different regions, with the most severe
wear occurring in the upper region withp>0.8H.
234
k] = 300-400' 10-4
0.8H
k2 = k4 =
k3 = 30-40 .10-4
1-10,10-4 1-10,10-4
Vslip (m/s)
o 0.2 0.7
Figure 4. Wear chart for wheel and rail steels. Wear coefficient k for different regions of slip
velocity and pressure. Dry, room temperature conditions. Telliskivi and Olofsson (2001).
The span for each wear coefficient, k, to k4' implies scatter in the
measurements. Still an average k-value has been chosen for each region.
Note that wheel tread contacts usually occur in the lower left comer, whereas
wheel flange contacts can be found in all four regions.
The effect on the wear coefficient due to natural lubrication (weather
etc.) and man-made lubrication, on the high rail in curves with radii less than
600 m in the application of Section 5, has been estimated based on different
field measurements. In this way the coefficients of Figure 4 have been
reduced. Laboratory tests for lubricated conditions will soon be carried out.
When the wear in a wear step - corresponding to, say, 1500 km of travel
or a maximum wear depth of 0.1 mm - is determined, the wear distribution
along a wheel profile must be smoothed somewhat to avoid difficulties in the
following wear steps. The smoothed wear distribution is subtracted from the
wheel profile giving an updated wheel profile, which in tum needs some
additional smoothing. A cubic spline interpolation algorithm has been used
for both smoothings and tested out carefully.
5. In-field verification
The present wheel wear simulation tool has been verified by comparing
simulated and measured wheel wear on a commuter vehicle trafficking the
commuter railway network in Stockholm. A vehicle XlOB, with two non-
powered and rather track-friendly two-axled bogies, has been selected for the
main comparisons. The network comprises about 200 km of track and has
many curves with radii in the span of 300-2000 m. A load collective have
235
been designed with seven curve radii and three rail profiles (small, medium
and large wear). The measurements have been carried out by Nilsson (2000).
The present vehicle and track have been properly modelled and the MBS
software GENSYS, DEsolver (2000), has been used for the simulation. The
well-known software MATLAB has been used to administrate the entire
simulations, cf. Figure 1.
Figure 5 shows an example of the simulation-measurement comparison.
Here we can see how a start wheel profile, with a straight tread, is worn after
200000 km of travel. Simulated and measured wheel profiles agree well.
Scalar wheel wear measures such as flange thickness, flange height, flange
inclination and decrease in cross-sectional area can also be shown to give
good agreement.
-10
-20
-30
~------~------~------~----~~----~~----~
-60 -40 -20 o 20 40 60
10.-------,-------,-------,-------,-------.-------~
8
6
4
2
o
_2L-------L-------~------~------~------~------~
-60 -40 -20 o 20 40 60
Figure 5. Simulated and measured wheel profile (top) and wheel wear (radius decrease,
bottom) after 200000 km running distance for Stockholm commuter vehicle XIOB-3149.
Simulation (solid line), measurement (dashed line), start profile (dash-dotted line, in top
figure only). Simulated results are for wheel set I and 4. Measured results are for the left
wheel of wheel set I. Units in mm.
236
6. Concluding remarks
Acknowledgments
References
Andersson, E., Berg, M., and Stichel, S. (1999) Rail vehicle dynamics - fundamentals and
guidelines, Compendium, Div. of Railway Technology, KTH, Stockholm.
Archard, 1. F. (1953) Contact and rubbing of flat surfaces, Journal of Applied Physics 24,
981-988.
DEsolver (2000) GENSYS User's Manual, Release 0003, Ostersund.
Jendel, T. (2000) Prediction of wheel profile wear - methodology and verification, Licentiate
Thesis, Report TRITA-FKT 2000:49, Div. of Railway Technology, KTH, Stockholm.
Kalker, 1. 1. (1990) Three-dimensional elastic bodies in rolling contact, Kluwer Academic
Publisher, Dordrecht.
Li, Z. L., and Kalker, J. J. (1998) Simulation of severe wheel-rail wear, Computers in
Railways, 393-402.
Nilsson, R. (2000) Wheel and rail wear - measured profile and hardness changes during 2.5
years for Stockholm commuter traffic, Railway Engineering 2000, London.
Pearce, T. G., and Sherratt, N. D. (1991) Prediction of wheel profile wear, Wear 144, 343-
351.
Telliskivi, T., and Olofsson, U. (2001) Wear and plastic deformation of two rail steels - full
scale test and laboratory study, 2nd World Tribology Congress, Vienna. To appear.
MULTI-BODY MODELING OF PAPER
CALENDERING UNIT BY CONTACT DYNAMICS
FORMULATION
Emo Keskinen
Tampere University of Technology, Laboratory ofMachine Dynamics
P.O. Box 589, FIN-33JOJ Tampere, Finland
1. Introduction
shells are in rolling contact for calendering, see Smook (1992), the axially
moving web material. In calendering the paper web is manipulated between
two compressed rolls in the contact zone called nip. In order to extend the
nip manipulation time the rolls are covered by polymer layer providing a
wide contact zone.
2. Roll dynamics
a) b)
Figure 2. Angular degrees of freedom related to roll spin motion (a) and orientation (b).
239
(1)
where s=(u'vl is the rigid body position of the roll c.o.g. Angles Q x and
Q y are small roll orientation inclinations around global coordinate axis, as
illustrated in figure 2b. Vector s=(u, v/ is the bending displacement vector of
roll cross-section and vector er=(e(e!»,s(e!»Yhas elements
e(e!» = eDse!> , s(e!» = sine!> . Matrix
a) b)
Figure 3. Roll loaded by line loads (a) and roll interaction forces in calendering (b).
240
Suppose now the roll is loaded on its outer surface by a set of distributed line
loads fi.(z) = pJz)er(Vi)+qJz)e'(Vi) and line moments t;{z) together
with concentrated bearing reaction forces Fj acting on the roll ends.
A relatively long but straightforward equation derivation, see Keskinen
(200 1), produces the rigid body equations of motion in form
e/2
MU= L f [C(Vi)Pi(Z)-S(Vi)qi(Z)}iZ+ LC(Vj)Fj
i_~ j
e/2
MV= L f ~(Vi)P;(Z)+C(Vi)qi(Z)}iZ+ LS(Vj)Fj -Mg
i -e/2 j
e/2
IQ x =-L f [S(Vi)Pi(Z) + c(Vi)q;(z)]zdz- LS(Vj)Fjz (4)
i_~ j
e/2
IQ y =L f [c(V;)p;(Z)-S(Vi)qi(Z)]zdz+ LC(Vj)Fjz j
i -e/2 j
e/2
I i = L J[q/z)R + Ii (z)}tz + T
i -e/2
where they are dynamically decoupled from the bending equations of motion
e/2
mii-2mev+(k-e 2m)u= L J[C('IIi -9)Pi(Z)-S('IIi -9)qi(Z)g(z)dz
i -e/2
+ LC('II j )Fjg(z j)
j
(5)
e/2
mv + 2meu+(k-e 2m)v = L J[S('IIi -9 )Pi(Z)+C('IIi -9 )q;(z)g(z)dz
i -e/2
+ LA'll j )Fjg(z j)
j
e/2 e/2
with m=pAfg(z)g(z)Tdz and k=El x f g"(z)g"(zldz . Eqs. (4)
-e/2 -e/2
and (5) have similar uncoupled structure than the ones obtained for a
vibrating unrestrained airplane structure in Bisplinghoff et al. (1955), p. 114.
This unrestrained body formulation is new in the dynamics of elastic rotors.
241
3. Nip dynamics
(6)
where h(z,t) is the thickness variation of paper web entering the nip, s is the
thickness of roll cover and Il is the out-of-roundness profile of permanent
cover deformation. The rate of penetration is now
If one of the rolls, say roll A, has been covered by soft polymer of thickness
s and characterized by its elastic and viscous moduli E and 11 corresponding
242
to the classical Kelvin-Voigt material model, the theory of elastic
foundation, Johnson (1985), is available giving contact parameters
i
a = E ~ ; b =!.- ; c =
3 S v~ 2 S v~
2rlf!o
with 2~ =(RA + S 'j1 + RB-1.
Tangential traction force may be evaluated by equation
4. Case study
b)
Figure 4. a) Main elements of test calender: Frame I, loading ann 2, lower roIl 3, upper roIl 4,
locking ann 5, loading cylinder 6, damping cylinder 7, locking cylinder 8, loading linkage 9,
load ceIl 10 and. b) System diagram of multi-body simulation model.
e [mm] 0.1.----,----,-----,--r===:===ll
o :~~~
··············j················T················[·····...
-0.1 . \...j..... ..+..........;. ........... ;
=
-0.2
\:
····\·j""'· ......·..·t:. . ·. ·. . ·!"". ..·...... :·. ..· .. ··
~: -=~m---r--
-0.6 ............. -!--.......... + .........: ....... j
o 2 4 6 8 10
t (s]
a)
e[mm]-0.5r-----,----:---,-r===~
10 12 14 16 18 20
t (s]
b)
Figure 5. Relative roll approach during nip closing maneuver (a) and under periodic paper
thickness variation (b).
The calculated responses show that the new loading mechanism is able to
limit the roll vibrations in the excitation amplitude level. Numerical time
integration of equation system gave satisfactory results for time step 1([4 S.
244
5. Conclusions
A two roll paper finishing unit has been modeled using multi-body
contact dynamics approach. This work contains the development of a roll
element that can be flexibly connected to mechanism links, bearings, drives
and hydraulic loading actuators. The complete system model consists of
dynamic equations of motion for the multi-body system. Algebraic contact
equations represent the coupling of neighbouring rolls in terms of contact
force components depending explicitly on the relative positions of the roll
surfaces. Dynamic excitation arising from periodic paper thickness variation
makes the system to vibrate the effect of which has been investigated by
means of a new kind of nip loading mechanism. The results show that the
new modeling approach is able to describe the contact dynamical
phenomena of cylinders in rolling contact. The system model has been
utilized in the design of the new calendering mechanism.
References
Bisplinghoff RL, Ashley H. (1962) Principles ofAeroelasticity. New York: Dover.
Bisplinghoff RL, Ashley H., Halfman RL, (1955) Aeroelasticity. New York: Dover.
Haug EJ. (1989) Computer-aided kinematics and dynamics of mechanical systems. Volume I:
Basic methods. Boston: Allyn and Bacon.
Johnson KL. (1985) Contact mechanics. Cambridge: Cambridge university press.
Keskinen EK. (1994) A contact dynamics formulation for distributed simulation of flexible
hydromechanical systems. Ph.D. Thesis. Tampere: Tampere university of technology
publications 134.
Keskinen EK. (2001) Multi-body contact dynamics modelling of roll mechanisms in paper
manufacturing systems. Proc. of ASME International Mechanical Engineering Congress
and Exposition, Symposium on the Vibration and Control of Mechanical Systems, Nov
11-16,2001, New York.
Kivinen J-M. (2001) A variable parameter facility for dynamic testing of polymer covered
paper machine rolls. Ph.D. Thesis. Tampere: Tampere university of technology
publications 347.
Lubkin JL. (1962) Contact problems. In: Fluegge W, editor. Handbook of engineering
mechanics. New York: McGraw-Hill.
Meirovitch L. (1967) Analytical methods in vibrations. New York: Macmillan.
Shabana A.A. (1998) Dynamics of multibody systems. Sec. ed. Cambridge: Cambridge
university press.
Smook GA. (1992) Handbook for pulp and paper technologists. Second ed. Vancouver:
Angus Wilde Publications.
EXISTENCE AND UNIQUENESS FOR
QUASISTATIC CONTACT PROBLEMS
WITH FRICTION
Lars-Erik Andersson
Department of Mathematics
Linkoping University, SE-581 83 LINKOPING, Sweden
Anders Klarbring
Department of Mechanical Engineering
Linkoping University, SE-581 83 LINKOPING, Sweden
Abstract This paper deals with mathematical results of existence and uniqueness
of solutions to problems of linear elasticity involving friction, in partic-
ular for the quasistatic case when inertia forces are neglected. Both the
case of a continuum body and that of a space discrete one are treated.
The intention is to describe the state of the art for these problems.
1. Introduction
The present paper reviews results, some of them recent, concerning
existence and uniqueness of solutions for the problem of a linearly elastic
body in frictional contact with a rigid obstacle.
The review is limited to quasistatic problems where inertia terms are
assumed negligible. Both the case of a continuum body and that of a
space discrete one are treated. We also restrict our attention to so called
coercive problems, where some part of the boundary of the elastic body
has a prescribed motion, excluding rigid body motions.
The mathematical analysis of frictional contact problems is very dif-
ferent when the coefficient of friction, fJ, is small or large, respectively.
It is well known that for large fJ one cannot expect any general existence
or uniqueness results for quasistatic problems, not even for the simplest
possible case with a single particle system with two degrees of freedom.
There is a well known counterexample, due to Klarbring (1987, 1990),
which is presented in Section 2. Otherwise we focus on existence and
245
J.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics. 245-260.
© 2002 Kluwer Academic Publishers.
246
uniqueness results which can be obtained under the assumption that fJ
is sufficiently small. Furthermore, we consider only the case of small
deformations and assume that the equations of linearized elasticity can
be used.
The rest of the article is organized as follows. In Section 2 we give
the above mentioned counterexample. In Section 3 we first give classical
formulations of the quasistatic problem for a continuous elastic system,
including the s~called normal compliance approach. By classical we
mean partial differential equations and boundary contact conditions,
assuming sufficient regularity of displacement and stress fields. Next
we reformulate the problems as a variational time-dependent problem in
appropriate function spaces and present some recent results, in particular
for the case with Signorini contact conditions.
In Section 4 we specialize to the space-discretized problems, for which
the mathematical analysis is quite different. Contrary to the case with
a continuous elastic system, it is here possible to obtain some rather
general uniqueness results, provided that one imposes some condition of
time-regularity for the exterior force-field.
The previously introduced time derivatives PT' PN and WT' and the time
derivative of the normal displacement, WN' are linearly coupled through
247
a stiffness equation:
where kNN' etc. are real numbers such that the system matrix is positive
definite and j N and iT are time derivatives of the external forces. Since
PN < 0 it holds that WN = 0 so we may conclude from (3), by multiplying
the first equation with /-t and subtracting, that
(4)
From (1) and (4) it follows that ifthe node goes into a stick state, then
/-tiN ~ iT.
From (2) and (4) one finds that if the node goes into a slip state, the
condition on derivatives of external forces depends on the sign of the
"effective stiffness" in (4):
if -/-tkNT + krT > 0 then /-tiN ~ iT
if -/-tkNT + kTT = 0 then /-tiN = iT
if -/-tkNT + kTT < 0 then /-tfN ~ fT.
We conclude the following for the problem of determining the right hand
time derivative of contact forces and displacements when given a rate of
change of the external forces:
• if -/-tk NT + kTT > 0, then there exists a unique solution;
• if -/-tk NT + kTT :::; 0, then, depending on the direction of change
of external loading, there does not exist any solution or there are
multiple solutions.
Note that /-t > 0 and kTT > 0 while kNT may have any sign. This later
constant represent a stiffness coupling between normal and tangential
degrees of freedom, so, clearly, for large such couplings we may experi-
ence non-uniqueness or non-existence. The non-existence of right hand
time derivatives means that the quasistatic problem cannot have any
solutions that are continuous in time.
u(t)lsu = ii.
Su ~"""""'rl
-divO"(u(t)) = f(t) in n
O"(u(t))ft = t(t)
-+
a(Jij f i= 0 in 0, (5)
ax;
(Jij = aijkl a
aUk
Xl
in 0, (6)
(Jijnj = ti on St· (7)
u=u, on SUo (8)
Here u = (Ul' U2, U3) is the displacement vector, U = {(Jij} is the stress
tensor and nj are the components of the outward unit normal vector.
249
The elasticity coefficients aijkl satisfy the usual symmetry and ellipticity
conditions. Furthermore, i, j, k, I = 1,2,3, the summation convention is
used and (0, Xl, X2, X3) is the cartesian reference frame.
To state the laws of contact and friction we decompose the displace-
ment and traction vectors on Se into normal and tangential components:
(11)
where Or and mT are new physical parameters. One retrieves (10) from
(11) and (12) by putting Cr = I-'CN and mT = mN'
Relations (5)-(8), (11) and (12) constitute the quasistatic frictional
contact problem with normal compliance.
°
for all t and all w E K. Here 'f/;c E Cr(Rn ) is a cut-off function which
equals 1 on a neighbourhood of Sc and on a neighbourhood of Sa U Su.
For the initial state u(O) = Uo we have the following necessary com-
patibility condition
a(Uo,l1 - uo) - ('f/;cl-'UN(U), I11T - uOTI)
~ (/,11 - uo) + (t,l1 - uo) (15)
for all 11 E K. Formally, the latter condition expresses that the initial
state is in equilibrium, satisfies the contact law (9) and that IUT(UO)I ~
-I-'UN(UO) on Sc·
Cocou et al. (1996) gave an existence result for a regularized version
of this problem, when UN is replaced by u~, where * denotes a bounded
linear mapping
a(u,v - ti) + CPnc(u,v- ti) + inc ('1£, v) - inc(u,ti) ~ (F,v- ti) (16)
for all v E V.
Here we have introduced the notation CPnc( '1£, v) = Isc CN (UN -g )~N UN dS
and inc ('1£, v) = Isc CT(UN - g)~ IVTI dS, and (F, '1£) = In J . v dx +
fS t t . v dS.
The problem VPQNC and variants thereof was investigated by Klar-
bring et al. (1988, 1989, 1991).
In Andersson (1999a) the problem VPQNC was treated with mT =
mNand
and
inc ('1£, v) = A r Il( UN -
lsc
g)~N IVTI dS
fixed nodes
external forces
[f II]
+r = [Kll
(KJ)t (19)
where
Kll KJ]
[ (KJ)t K '
is a symmetric stiffness matrix which, if the structure is sufficiently an-
chored, is positive definite and a superscript t denotes the transpose of
254
cp(L, n) = max
ufO
{m~n
l~t:Sl
rpi(L, n; u)} ,
Ui.lni
°
real analytic (RPRA) if it is continuous and if for every to E [0, T) there
exists f > such that
L Fk{t -
00
Right piecewise real analyticity implies that v{t) = MF{t) has at most
countably many discontinuities.
We first formulate an existence and uniqueness result for the case that
we have two degrees of freedom for each node.
Theorem 3 Assume that J.ti < fl., that F : [0, T) -+ R21 is RPRA and
AG and that the initial values r{O) and u{O) satisfy the natural compat-
ibility conditions. Then there exists a unique RPRA mapping
solving our quasistatic friction problem (for all t). Further there exists
a constant C = C{J.t) such that
Sketch of proof:
1. For a given to construct a formal power series solution r{t)
00
E rk{t - to)k. The coefficients rk are uniquely determined by a se-
k=O
quence (tree) of NLCP(k), for k ~ 1. These NLCP:s are similar to
that for the Rate Problem, although a little more complicated.
2. Prove that the power series converges in some interval [to, to + f),
giving a local solution.
3. Repeat with new initial point.
257
The proof does not go through completely for the case with 3 DOF
at each node. However we can still construct unique formal power series
solutions
00
Acknowledgments
This work was supported by the Swedish Research Council for Engi-
neering Sciences under contract no. 223-99-676.
258
Appendix: Notation and definitions
For the subsets Sc, Sa and Su c an
introduced earlier we make the further as-
sumption that they have mutually disjoint closures.
The (Hilbert space) Sobolev norm of order p over 0. or its boundary is denoted by
II· 11",0 or II· 11",110 respectively. The same notation is also used when the functions
are vector-valued. So we write, e.g., II· II(L 2 (0»ft =
11,110,0 and II· II(H1(or=II ·Ih.o.
Similarly, the dual pairing, e.g., between the spaces H- 1/ 2 (an) and H1/ (an)
is writ-
ten (', ')-1/2,1/2(80). When no confusion is likely to appear, the sub-indices denoting
the particular spaces are omitted.
To denote the dual pairing between a Hilbert space H and its dual H' we use
(-,.) H' ,H' The inner product and the norm in H will be denoted by (',')H and II· IIH'
We also introduce a gap function 9 E H 1 / 2 an defining the initial gap between
the elastic body in the reference configuration. Note that we do not assume that
9 ~ 0, which means that the body may be pre-loaded by the obstacle in the absence
of external forces f and t.
The trace operator is denoted by
(A.i)
Further we use the same notation for the trace operator tr : (HI (0.» n -+ (HI/2 (an»n
and denote its norm by IItrll. We also use that there exists a linear bounded extension
operator
=
with norm denoted by IICII, such that troC id(H1/2(IIO»ft. It is clear that IIClilltrll ~ 1
and that an optimal value of this product depends only on the geometry of n. For
the case that 0. = R+ = {(x, Xn) : Xn > O} these mappings are denoted by tro and
Co.
Next, let us introduce the affine subspace
= =
stress tensor is Uij Uij(1£) aijkl ~:~. For the coefficients aijkl ofthe elasticity ten-
sor we require, besides the usual symmetry conditions, that aijkl E Loo(n). Further
we assume that there exists a constant 00 > 0 such that
1
mapping
(1£, v) t-+ a(1£, v) = o
auk aVi
aijkl(X)~~ dx
UXI uXj
then defines a bounded, symmetric bilinear form on V x V. By Korn's inequality and
the boundedness of aijkl there exist constants Co = co(o, 0., Su) and Co such that for
alluEV
Coll1£ll~,o < a(1£,1£) =a(1£) < Coll1£ll~,o. (A.2)
259
For convenience we may let the non-negative coefficient of friction I' be defined on
the whole of the boundary and we assume that I' E L oo with sup-norm 111'1100'
For the coefficient of friction I' we assume further that it is a multiplier on H 1 / 2 (8n),
i.e., that the mapping
Hl/2(8n) 3 v ~ I'v E Hl/2(OO)
A sufficient condition for this is for example that 8S... and 8Sc are Lipschitz curves.
We finally assume that 9 E Hl1fc2+Q(Se) n Hl/2(OO).
The parameters 0: and /3 should satisfy the inequalities, 0 < 0: < /3 ~ 1 and
0: ~ 1/2.
References
Andersson, L.-E. (1991) A quasistatic frictional problem with normal compliance,
Nonlinear Analysis, Theory Methods & Applications 16,347-369.
Andersson, L.-E. (1999a) A quasistatic frictional problem with a normal compliance
penalization term, Nonlinear Analysis, Theory Methods & Applications 37,689-
705.
Andersson, L.-E. (1999b) Quasistatic frictional contact problems with finitely many
degrees of freedom. LiTH-MAT-R-1999-22.
Andersson, L.-E. (2000) Existence results for quasistatic contact problems with Coulomb
friction, Applied mathematics and Optimization 42 (2000), 169-202.
260
W. R. Bielski
Institute of Geophysics, Pol. Acad. Sci.,
01-452 Warsaw, ul. Ksi~cia Janusza 64, Poland
A. Galka
Institute of Fundamental Technological Research, Pol. Acad. Sci.,
Swi~tokrzyska 21, 00-049 Warsaw, Poland
J. J. Telega
Institute of Fundamental Technological Research, Pol. Acad. Sci.,
Swi~tokrzY8ka 21, 00-049 Warsaw, Poland
Introduction
In Bielski and Telega (1998) the study of unilateral contact problems
for elastic plates was started. Such a study was next extended to the
linear Koiter shell model in Telega et al. (2001). The aim of the present
contribution is to develop augmented Lagrangian methods applicable to
nonconvex unilateral contact problems arising in structural mechanics.
To this end we have extended the approach proposed by Ito and Kunisch
(1990, 1995) for convex problems. An illustrative example pertains to a
geometrically nonlinear elastic beam.
261
I.A. C. Martins and M.D.P. Monteiro Marques (eds.). Contact Mechanics. 261-268.
© 2002 Kluwer Academic Publishers.
262
where Ea,B(U) = u(a,,B) = H~+ ~:~). We note that only the first strain
measure is nonlinear.
The constitutive equations have the form
U = 0 on ro , w
aw
= -an = 0 on r .
An appropriate space for displacements is
min{Lc(u, A) I u E K},
where
Lc(u, A) = J(u) + CPc(Au, A), (9)
CPc(v, A) = inf{cp(v - u) + (A, u) H + ~lIull~}. (10)
Here (c, A) E lR+ X H. We observe that cp(., A) is (Lipschitz) continuously
Fh~chet differentiable.
Ito and Kunisch (1995) developed the following augmented Lagrangian
method involving a sequential minimization:
A ugmented Lagrangian Algorithm
1: Choose a starting value Al E H, a positive number c and set k = 1.
Step 2: Given Ak E H find Uk E K by
Lc(Uk, Ak) = min{Lc(u, Ak) I u E K}.
Step 3: Update Ak by Ak+! = CP~(AUk' Ak), where cp' denotes the Frechet
derivative of the functional cp(., A).
Step 4: If the convergence criterion is not satisfied then set k = k + 1
and go to Step 2.
Under suitable, physically plausible assumptions, the augmented La-
grangian algorithm just sketched converges.
Obviously, this algorithm is not directly applicable to nonconvex con-
tact problems of say finitely deformed elastic bodies and geometrically
nonlinear structures. There are three basic sources of nonconvexity: (i)
266
a nonconvex functional J, (ii) a nonconvex functional cp, (iii) a nonlin-
ear operator appearing in the functional cpo Such an operator is denoted
by N. Obviously, in practice various combinations of cases (i)-(iii) are
important.
For geometrically nonlinear problems the functional cp is usually an
indicator function of a (weakly) closed and nonconvex set, cf. Example
2.1 and He et al. (1996). We already know how to generate a sequence
of convex sets of constraints.
A large class of geometrically nonlinear problems leads to the functional
J of the form,
J(u) = G(Au) + F(u),
where G represents the functional of the total internal energy whilst
F is a loading functional, usually a linear one. The functional G is
nonconvex. For nonlinear structures it may often be written as follows
(11)
where
J(u,w) = 0 10.5 (
-0.5
1
u,x + 2"W'XW 2H
,x )2 dx + -3-0
3 1°·5 (w,xx)2dx
-0.5
267
{0.5
+ i-O.5 bw{x)dx,
Ku = {u I U E W 1,4{-O.5,O.5), u{-O.5) = u{O.5) = O},
Kw = {w IwE W 1,2(O.5, 0.5), w( -0.5) = w(0.5) = w,x( -0.5) =
w,x(0.5) = 0, g(w) ~ O}.
and C > 0, Particular forms of the function 9 are given below. Anyway,
we assume that the set Kw is convex. To solve this problem we introduce
the sequence of functionals, cf. the previous section,
Jm{u, w) = C 10.5
-0.5
{u,x
1
+ -2W,xW~-1)2 dx + -3-C
2 H3 1°·5 (w,xx)2 dx
-0.5
+ 10.5
-0.5 bw(x) dx, m= 1,2, ...
Lm,c,>.(u, w) = Jm(u, w) + 21
C
1°·5 [(sup{O,). + cg(w)})2 ]dx.
-0.5
_).2
0.002
0.0015
0.001
0.0005
Acknowledgments
The first author was partially supported by the State Committee for
Scientific Research (KBN, Poland) through the grant No 6 P04D 039
15.
References
Bielski W. R., and Telega, J. J. (1998) Existence of solutions to obstacle problems for
linear and nonlinear elastic plates, Math. Compo Modelling, 28, 55-66.
Telega, J.J., Galka, A., and Bielski, W. R. (2001) Augmented Lagrangian methods
for a class of convex and nonconvex problems, J. Theor. Appl. Mech., 3, 741-768.
Ito, K., and Kunisch, K. (1990) An augmented Lagrangian technique for variational
inequalities, Appl. Math. Optim., 21, 223-241.
Ito, K., and Kunisch, K. (1995) Augmented Lagrangian methods for nonsmooth,
convex optimization in Hilbert spaces, in: Control of Partial Differential Equations
and Applications, ed. by E. Casas, pp. 107-117, Marcel Dekker.
Lewinski, T and Telega, J. J. (2000) Plates, Laminates and Shells: Asymptotic Analy-
sis and Homogenization, Series on Advances in Mathematics for Applied Sciences,
vo1.52, World Scientific, Singapore.
He, Q.-C., Telega, J. J., and Curnier, A. (1996) Unilateral contact of two solids subject
to large deformation and existence results, Proc. R. Soc., London, A452, 2691-
2717.
APPROXIMATION OF QUASISTATIC
SIGNORINI PROBLEMS WITH LOCAL
FRICTION BY A MIXED METHOD
Marius Cocou
Laboratoire de Mecanique et d'Acoustique - C.N.R.S., 91 chemin Joseph Aiguier,
19402 Marseille Cedex 20 and Universite de Provence, Marseille, France
Remi Rocca
Laboratoire de Mecanique et d'Acoustique - C.N.R.S., 91, chemin J.Aiguier,
19402 Marseille Cedex 20, France
1. Introduction
This paper deals with the numerical analysis of the quasistatic prob-
lem of unilateral contact with local Coulomb friction between two elas-
tic bodies. The approach we use here follows that of Rocca and Cocou
(2001b), where the approximation of the contact problem for an elastic
body by a mixed finite element method is studied.
Contact problems with local friction law were first considered in the
static case by Necas, Jarusek and Haslinger (1980). They proved an
existence result in the case of a strip by a shifting technique when the
friction coefficient is sufficiently small. This result was extended by
Jaruaek (1983) for a domain with smooth contact zone and for two elastic
bodies.
269
lAC. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 269-276.
© 2002 Kluwer Academic Publishers.
270
3. Discrete formulations
In order to prove the existence of discrete solutions converging towards
u, we consider the following finite element approximations.
Let (r'/)h be a set of regular triangulations of Of3, {3 = 1,2, see
Ciarlet (1978). Each triangulation is a collection of elements (Or) such
that Of3 = UiEIf O~ with O~ n Or = 0 for all k,1 E If, k =f. I. We
assume that each triangulation is compatible with the partition of r f3 ,
that is, each point where the boundary condition changes is a node of
a set O~ and we require also that the nodes lying on r3 belong to both
triangulations. We shall use the notation Th = {T,/, 1h2}. Let (r~)H be
a set of triangulations of r3 = UiE.1H r 3 ,;. We suppose that the elements
r 3I i, i E .JH, are segments for d = 2 .and rectangles
..
for d = .
3. We
.
set
HI = length (r3,i) for d = 2, and HI = (HI ; H~), where HI, H~ are
the lengths of the edges of r 3 ,i for d = 3. We assume that (r~)H are
regular that is there exists a constant '1 > 0 such that ~in(Hi)/H ~ '1,
IE.1H
where H = ~ax(Hi) for d = 2, and ~in(HLm)/H ~ '1, where
IE.1H IE.1H
H = ~ax(HL H~) for d = 3.
IE.1H
We introduce the following sets:
5. Convergence results
In this section we prove that there exists a sequence of discrete solu-
tions of problems (P~H)i which converges towards (Uh, AH), solution
of problem PhH' If IIJ.tIIM(Hl/2(r)) < ji., then there exist two constants
C3 > 0, C4 > 0 such that
Theorem 5 Let (1hj )jEN and (r~j)jEN be such that for all j EN we
have 71 ~ hj/Hj ~ 72. Let I' E M(Hl/2(rl)) be such that IIJ.tllv"'(r1) <
1'* and IIJ.tIlM(Hl/2(r1)) < ji.. Then there exists a subsequence (jk)kEN such
that (Uh" Ie , AH" Ie )kEN, the solutions of problems (Ph'11e H,' Ie ) , converge
weakly in W 1,2(0, Tj V)XW 1,2(0, Tj H-l/2(r 1 )) towards (u, A = 8UN(U))
solution of problem P. Moreover, we have
276
The result follows by using the previous estimates and by passing to the
limit in (10,11).
6. Conclusions
In this paper we have considered the unilateral contact with local fric-
tion between two elastic bodies. An implicit Euler scheme with respect
to time and a mixed finite element method for the space discretization
have been adopted. The fully discrete problem was solved by a fixed
point approach. This result and some estimates independent of the dis-
cretization parameters enable us to pass to the limit with respect to the
mesh size and to time. It would be interesting to consider the numeri-
cal analysis of unilateral contact problems with local friction for curved
contact zones. The problem of obtaining rates of convergence for the
quasistatic contact problem is open.
References
Andersson, L. E. (2000) Existence results for quasistatic contact problems with Coulomb
friction, Appl. Math. Opt. 42, 169 - 202.
Ciarlet, P. G. (1978) The finite element method for elliptic problems, North-Holland,
Amsterdam.
Cocou, M., Pratt, E., and Raous, M. (1996) Formulation and approximation of qua-
sistatic frictional contact, Int. J. Engrg. Sci. 34, 783-798.
Cocou, M., and Rocca, R. (2000) Existence results for unilateral quasistatic contact
problems with friction and adhesion, Math. Modelling and Num. Analysis 34,981-
100!.
Ekeland, 1, and Temam, R. (1974) Analyse convexe et problemes variationnels, Dunod,
Paris.
Haslinger, J. (1983) Approximation of the Signorini problem with friction, obeying
the Coulomb law, Math. Meth. Appl. Sci., 5, 422-437.
Hlavacek, J., Haslinger, J., Neeas, J., and LoviSek, J. (1988) Solution of variational
inequalities in mechanics, Springer-Verlag, New-York, Berlin, Heidelberg.
JaruSek, J. (1983) Contact problems with bounded friction - coercive case, Czechoslo-
vak Math. J. 33, 237-26!.
Netas, J., JaruSek, J., and Haslinger, J. (1980) On the solution of the variational
inequality to the Signorini problem with small friction, Bolletino U.M.l. (5) 17-
B,796-81!.
Rocca, R. (1999) Existence of a solution for a quasi static contact problem with local
friction, C. R. Acad. Sci. Paris 328 Serie I, 1253-1258.
Rocca, R., and Cocu, M. (2001) Existence and approximation of a solution to qua-
sistatic Signorini problem with local friction, Int. J. Engrg. Sci. 39, 1233-1255.
Rocca, R., and Cocou, M. (2001) Numerical analysis of quasistatic unilateral contact
problems with local friction, SIAM Journal of Numerical Analysis, to appear.
Zeidler, E. (1993) Nonlinear functional analysis and its applications, Vol I, Springer-
Verlag, New-York.
A BEAM IN ADHESIVE CONTACT
W.Han
Department of Mathematics,
University of Iowa, Iowa City, IA 52242, USA
K. L. Kuttler
Department of Mathematics
Brigham Young University, Provo, UT 84602, USA
M. Shillor
Department of Mathematics and Statistics,
Oakland University, Rochester, MI 48309, USA
M. Sofonea
Laboratoire de Theorie des Systemes,
Universite de Perpignan, 66 860 Perpignan, France
1. Introduction
Processes of adhesion are very important in industry, especially when
composite materials are involved. There exists extensive engineering
literature on various aspects of the subject, however, general mathemat-
ically sound models are very recent. A novel approach to the modeling
277
lAC. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 277-284.
© 2002 Kluwer Academic Publishers.
278
2. The model
We consider an elastic beam of length L that is clamped at its left
end while the right end is free. The beam is being acted upon by an
applied force of (linear) density f, and it may come in adhesive contact
with a rigid foundation below it. The setting is depicted in Figure 1.
We denote OT = (0, L) x (0, T), for T > 0, and let u = u(x, t) represent
the vertical displacement of the beam, at (x, t) E flT. We assume that
the obstacle is described by the function y = 4>(x), for 0 ~ x ~ L. We
denote A = El, where I is the beam's moment of inertia and E the
Young modulus. We assume that the acting forces vary slowly in time
and the process is quasistatic. Then, the equation of motion for the
279
beam is
in Or, (1)
e
where = e(x, t) denotes the reaction force of the foundation and the
adhesion force. The displacements of the beam are restricted by the
rigid foundation to be above it, thus
(2)
which represents a nonpenetration condition. When contact takes place
e
the foundation's reaction force is directed upward,
in 0T. (3)
• !
f
!
0r---~----------~----~_
L x
y=~
Next, following Raous et al. (1999), the evolution of the adhesion field
is given by
in flT, (6)
where 'Y is the adhesion rate constant, and r + = max{r,O} denotes the
positive part of r. We note that re-bonding is excluded in this model,
however, our methods can be used in the case when it is included. In (6),
and everywhere in the sequel, a prime represents the time derivative.
To complete the model we prescribe appropriate initial and boundary
conditions. The initial condition takes the form
We denote by H the space L2 (0, L) and by h ·)H, I·IH its inner product
and the associated norm, respectively. Let K denote the convex subset
of V defined by
K = { v E V I v ~ ¢ on [0, L] }.
(19)
282
Here, p = p(.) ~ 0 is a given function which vanishes for nonnegative
values of its argument. We note that when u > 0 then = -1i:(u+)/32e
e
represents the adhesion force and when u ~ 0 then = p(u) represents
the reaction force. Next, the evolution equation for the adhesion field
(6) is modified as follows:
/3' + 'Y1i:((u)+)2({3)+ = 0 in !IT. (20)
We use (u)+ to ensure that only tension contributes to debonding, since
compression does not affect it.
The classical formulation of the contact problem between a beam and
a flexible foundation with adhesion is:
Problem PNC. Find a displacement function u : !IT ~ R and an
adhesion function /3 : !IT ~ R such that (1), (19), (20), (7)-(8) hold.
We have the following weak formulation of problem PNC.
Problem PNC-V. Find a displacement function u: [0, T] ~ V and
an adhesion function /3 : [0, T] ~ LOO(O, L) such that, for all t E [0, T],
a(u(t),v) + (1i:/32(t)u(t),V)H = (f(t) +p(u(t)),v)H \::Iv E V, (21)
/3'(t) + 'Y1i: (u+)2(t)(/3(t))+ = 0, (22)
/3(0) = /30. (23)
5. Fully-discrete approximation
We now turn to numerical approximations of the quasistatic elastic
problem with normal compliance studied in the previous section. We
introduce a partition of the spatial domain [0, L] : 0 = Xo < Xl < ... <
XM = L. Denote Ii = [Xi-I, xd and hi = Xi - Xi-l for i = 1, ... , M, and
h = maxl~i~M hi the meshsize. We define the finite element spaces
Vh = {v h E V I vhlIi is cubic, 1 ~ i ~ M},
Qh = {qh E D)O(O,L) I qhl Ii is constant, 1 ~ i ~ M}.
Thus, Vh consists of piecewise cubics, and Qh of piecewise constant
functions. We define a piecewise averaging operator ph : Ll (0, L) -+ Qh
by
h
p UIIi
1 r
= IIil iIi udx, (24)
(28)
284
We note that if we take ,8e = ph,Bo, then conditions (28) and (29) are
easily satisfied.
Acknowledgments
The work of the first author was supported by NSF under Grant
DMS-9874015.
References
Fremond, M. (1982) Equilibre des structures qui adherent a. leur support, C.R. Acad.
Sci. Paris 295, Serle II, 913-916.
Fremond, M. (1987) Adherence des solides, Journal de Mecanique Theorique et Ap-
pliquee 6, 383-407.
Han, W., and Reddy, B. D. (1999) Plasticity: Mathematical Theory and Numerical
Analysis, Springer-Verlag, New York.
Han, W., Kuttler, K. L., Shillor, M., and Sofonea, M. (2001) Elastic beam in adhesive
contact, preprint.
Han, W., and Sofonea, M. (2000) Evolutionary variational inequalities arising in vis-
coelastic contact problems, SIAM J. Numer. Anal., 38, 556-579.
Kikuchi, N., and Oden, J. T. (1988) Contact Problema in Elasticity, SIAM, Philadel-
phia, 1988.
Klarbring, A., Mikelic A., and Shillor, M. (1988) Frictional contact problems with
normal compliance, Int. J. Engng. Sci., 26, 811-832.
Raous, M., Cangemi, L., and Cocu, M. (1999) A consistent model coupling adhesion,
friction, and unilateral contact, Computer Meth. Appl. Mechan. Engng., 177,383-
399.
Rojek, J., and Telega, J. J. (2001) Contact problems with friction, adhesion and
wear in orthopaedic biomechanics. Part I: General development, J. Theoretical
and Applied Mechanics, 39, to appear.
Rojek, J., Telega, J. J., and Stupkiewicz, S. (2001) Contact problems with friction,
adhesion and wear in orthopaedic biomechanics. Part II: Numerical implementation
and application to implanted knee joints, J. Theoretical and Applied Mechanics,
39, to appear.
COULOMB FLUID-SOLID
INTERFACE LAW IN LUBRICATION
Guy Bayada
CNRS-UMR 5585 and 5514 LMC-INSA
401 Math, Villeurbanne 69621 Prance.
Mahdi Boukrouche
CNRS-UMR 5585 and UPRES 3058
E.A.N. Saint-Etienne 42023, Prance.
Abstract The asymptotic behaviour of a Stokes flow with Coulomb free boundary
friction condition when one dimension of the fluid domain tends to zero
is studied. The specific Reynolds equation associated with variational
inequalities is obtained and uniqueness is proved.
1. Introduction
An accurate choice of the boundary conditions on the fluid-solid inter-
face is of particular interest in lubrication area, which is concerned with
thin film flow behavior. In that case, the difference of velocities between
the surrounding surfaces is the governing phenomenon that allows the
pressure in the fluid to build up and prevent the solid surfaces to collapse
which is the main objective of the lubrication. In operating conditions,
no slip condition is induced by chemical bounds (cf. Pit (1999)) between
the lubricant and the surrounding surfaces and by the action of the nor-
mal stresses, which are linked to the pressure inside the flow. On the
contrary, tangential stresses are so high that they tend to destroy the
chemical bounds and induce slip phenomenon. This induces a transposi-
tion of the well-known Coulomb law between two solids (Duvaut (1980))
to the fluid solid interface. Although being implicitly used in numeri-
cal procedures for lubrication problems, a Reynolds thin film equation
taking account of such slip phenomenon seems not to have been stud-
ied in a somewhat mathematical aspect until recently for the simplified
Tresca interface condition in Bayada-Boukrouche (2001). The goal of
285
1.A. C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 285-292.
© 2002 Kluwer Academic Publishers.
286
the present paper is not only to give existence and uniqueness for this
problem but also to obtain rigorously the equation describing such a
phenomenon in a thin film flow by way of an asymptotic analysis in
which a small parameter is the width of the gap. The departure point
is the Stokes equation with the Coulomb boundary conditions and fall
within the scope of the work of Consiglieri (1992). The related formula-
tion is given in section 2. Following the same idea as in Assemien et al.
(1994) Bayada-Boukrouche (2001), Bresh et al. (1997), we introduce a
new scaling taking into account a small parameter. It is to be noticed
that we are led in a natural way to express the Coulomb coefficient in
terms of this small parameter in a way that induces the uniqueness of the
related Stokes problem. Then in section 3 we get estimates for velocity
and then pressure. Passing to the limit in the initial inequalities using
the results of Bayada-Lhalouani (2001), we are able to obtain the limit
problem, which can be expressed in terms of the (unknown) velocity and
tangential stresses. In section 4 a specific decomposition based upon
a new div-curl decomposition allows us to study the uniqueness of the
limit problem.
Due to (3) it is well known cf. Girault-Raviart (1979) (lemma 2.2 p.24)
that there exists a function Gc such that Gc E (Hl(Oc))3 with
(5)
On ri, the velocity is known and is parallel to the w- plane :
(6)
On w, there is a no flux condition across w so that
U3 = g~ = 0 on w. (7)
The tangential velocity is unknown and satisfies the Coulomb friction
law with k c the ratio of the tangential stress and the normal stress
where 1.1 here is the JR2 Euclidean norm, n = (nl' n2, n3) is the unit
outward normal to f c , and
are, respectively, the normal and the tangential velocity, the components
of the normal and the tangential stress tensor :
8uEC 8uj
(Ji,j = _pc 6i,j +v(8x; + 8Xi) (1~i,j~3),
for all x in w, where <p is a given positive function of class Coo with
support in w. We will denote in the following s := gc on w. To get a
weak formulation, we introduce
288
L6(ft) = {q E L2 (nc) { qdxdx3 = 0, where dx = dX1dx2}.
Joe
A formal application of Green's formula, using (1)-(3) and (5)-(8) leads
to the weak formulation: For Gc as in (4), find u C = (uj, u~, u~) in
K(nc) and pC in L5(n c), such that for all (q, cp) E L5(Oc) x K(Oc):
Let us assume the following dependence of the data (with respect to £):
(11)
The second assumption means that the friction coefficient is the ratio
of the tangential stress inside the film (which has the same order of
magnitude as the ratio between the tangential velocity and the gap: c~)
and of the normal stress, which is close to the actual pressure which has
order of magnitude of £-2.
289
Let us define the c-independent vector G(x,y) = (G 1,G2,G3)(X,y) :
2
L ro-i,j 88,x) (rpi -
i,j=l in
uDdxdy + 1 w
kS(o-;)(lrp - sl-Iuc - sl)dx+
(13)
(15)
8u§ -" 0
c;- (16)
8y ,
ail
(i=I,2), II-a I
y H-l(fl)
~ E.C2. (19)
02U~ op*
_ v _l + - = 0 (i = 1,2) in L2(0), (20)
oy2 OXi
where
rh(x)
jJi(x) = Jo fJi(X, y)dy.
(25)
291
where U = (C, D) is the solution of the following "non linear" variational
problem: Find U E Hl(w) X H6(w) such that
=j r 6vg.ncp.
J(¢)
wkS( -p*)lh Vcp + h- curl(8) - sldx, £¢ =
2 1
kw
Proof. It follows the general line of the one of theorem 5.1 in Bayada-
Boukrouche (2001). The first step is to prove that any function in
(L2(w))2 can be written as (25) so that (C, D) satisfies (26). To prove
uniqueness of s* ( and then of p* and T*) we have to prove uniqueness
of C and D. Let U1 , U2 be two solutions to (26), then:
Combining (27), (30), and (29), we deduce that if Ilkllux,(w) ~ k*, where
(31 )
292
then
therefore s~ - si = O.
Remark 2 For IlkllL'xl(w) ::;
k*, under the same hypothesis of theorem
4, the uniqueness of p* follows from
1(W
h3 \lp* - '3:. s*) \ltpdx = [ tpg.n. \;ftp
1211 2 Jaw
E Hl(w), (32)
which can be obtained from (23) and (24). And then 7* is also unique,
from the uniqueness of p* and s*, by using (23).
References
Assemien, A., Bayada, G., and Chambat, M., Inertial effects in the asymptotic be-
haviour of a thin film flow. Asymptotic Analysis, 9, (1994), 117-208.
Bayada, G., Boukrouche, M., On a free boundary problem for Reynolds equation
derived from the Stokes system with Tresca boundary conditions. 325, (2001), on
http://numerix.univ-Iyon1.fr / publis / publi v/ publis.html
Bayada, G., Lhalouani, K. Asymptotic and numerical analysis for unilateral contact
problem with Coulomb's friction between an elastic body and a thin elastic soft
layer. Asymptotic Analysis, 25 (2001) 329-362.
Bresh, D., Lemoine, J., and Simon, J., Ecoulement engendre par Ie vent et la force de
Coriolis dans un domaine mince: I Cas stationnaire. C.R.A.S. Paris, t.325, Serie
I, (1997), 807-812.
Consiglieri, L., Stationary Solutions for a Bingham Flow with Nonlocal Friction. Pit-
man Research Notes in Mathematics Series 274, (1992), 237-243.
Duvaut, G., Equilibre d'un solide elastique avec contact unilateral et frottement de
Coulomb. C.R.Acad.Sc.Paris, t 290, (1980), 263-265.
Ekeland, I., Temam, R., Analyse convexe et problemes variationnels. Dunod et
Gauthier-Villars, Paris, {1974}.
Girault, V., Raviart, P.A., Finite element Approximation of the Navier-Stokes Equa-
tions. Springer-Verlag, {1979}.
Pit, R., Mesure locale de la vitesse a l'interface solide-liquide simple: Glissement et
rle des interactions. These Physique, Univ. Paris XI (1999).
Strozzi, A., Formulation of three lubrication problems in term of complementar-
ity. Wear, 104, (1985), 103-119.
ELASTODYNAMIC FRICTION PROBLEM
WITH A "SURFACE INERTIA"
PERTURBATION
Jean-Claude Paumier
LMC-IMAG, BP 53X
38041 Grenoble, France
Yves Renard
MIP, INSA de Toulouse
31077 Toulouse, France
1. Introduction
Here, we present an approach of the elastodynamic friction problem
(Le. not in the visco-elastic or quasi-static framework) which considers
a regularization in a layer of small thickness localized on the contact
zone. We called this regularization a surface inertia perturbation. A
simple geometry is considered and also regularized friction and unilateral
contact conditions.
The main motivation is to try to generalize theoretical and qualitative
results established in Ionescu and Paumier (1994) and in Renard (2000)
for the one-dimensional problem to the multi-dimensional case. In the
one-dimensional case, the analysis of the problem showed that the use
293
JA.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 293-300.
© 2002 Kluwer Academic Publishers.
294
of a non-monotone slip dependent friction coefficient in the purely elas-
to dynamic problem introduces a multiplicity of the solutions and shocks
in velocity. This non-uniqueness is still present if the friction and unilat-
eral contact condition are regularized. The one-dimensional case has the
interest to show precisely the regularity which can be expected for the
solutions. In this case, it has been proved that the surface inertia allows
to recover the uniqueness of the solution and, when the perturbation
parameter goes to zero, the solution tends to a particular solution to the
non-perturbed problem which is related to the perfect delay criterion in-
troduced for this problem in Ionescu and Paumier (1994). Unfortunately,
this criterion has not a clear extension to the multi-dimensional case, and
the discussion is still open to know whether or not non-uniqueness can be
observed with a non-decreasing friction coefficient (this is not the case in
the one-dimensional case) and whether or not stick-slip instabilities can
be observed with a constant friction coefficient (positive elements are
presented in Renard (1998), Moirot and Q.-S. Nguyen(2001), Simoes
and Martins (1998) and in Martins et al. (1999)).
I
lisboabstract.tex using the following multi-valued functions:
{O} if ~ < 0,
{I~:I}' if vT # 0,
e
[0, +00[, if = 0, and DirAv) =
{Wi Iwi ~ l'~N = O},
¢, if ~ > 0, If vT = 0,
295
and with p the density, u(u) the stress tensor, e(u) the linearized strain
tensor, A the elasticity tensor which satisfies usual conditions of sym-
metry and coercivity, n the outward unit normal to 0 on 80, u(u)n =
uN(u)n+uT(u) and u = uNn+uT on ro the usual decompositions in
normal and tangential part on r 0' u D, g, f are given, TT the tangential
velocity ofthe rigid foundation, and J.t(.) the friction coefficient which is
allowed to depend on the sliding velocity (itT - T+). It is assumed that
this dependence is such that J.t(') lies in W1,OO(lR ), but no assumption
is made for the friction coefficient to be small enough.
The following regularizations of JN and DirT correspond to classical
regularizations of unilateral contact and friction conditions:
0 if ~ < 0,
{
~(~) = ~
- if ~ ~ o.
."
UN -~(UN)'
UT -~(UN)J.tl1(itT - TT),
3. Simplified problem
For the sake of simplicity, and as a first approach to avoid difficulties
coming from the geometry, it is assumed that 0 is a cylinder (see fig. 1 ),
i.e. 0 = wx]O, D[ where w is a domain of lRn - 1 and D > 0 is the height.
We assume that r D = W x {D}, r N = 8wx]0, D[ and r a = w x {O}. We
denote x = (x',x n ) where x' E wand Xn E ]O,D[.
It is also assumed for simplicity that p == 1, and even if it means a
translation of the solution that TT(t) = 0, UD = 0, u(O) = it(O) = o.
296
Xn =
Xn = 0
Ec = wx]O,c[ CO,
Ic1 : TT
Vc ----->..
---r Hl (rc.' lIl>n)
~ 2
Ic: -----'- L2(rc·,lIl>n)
H c ---r ~
and
V 1---7 V 1---7
(F;'T/(t),v)V;,Ve = r -jZ(u~'T/(t))J.t'T/(u~'T/(t)).vTd(T,
ire
ue'T/(O) = ue'T/(O) = 0,
The term "surface inertia" is motivated by the fact that this pertur-
bation can be viewed as the addition on the contact surface of a surface
operator (Submitted paper).
°
Theorem 1 Under assumptions J.t(') E W1,00(lR+) and (2) Problem (3)
°
has a unique solution ue'T/(t, x) for all c > and'fJ > which belongs to
Loo(O, Tj Ve) n W1,00(0, Tj He).
The principle of the proof comes from the classical existence and
uniqueness result for initial value problem of differential equations
(Cauchy-Lipschitz theorem). We show, thanks to an a priori estimate
and a reformulation of the problem in terms of fixed point, that a p-
iterated operator is a contraction in a certain closed ball of Loo(O, Tj Ve) n
W1,00(0, Tj He). The reinforced traces on replay an important role.
Detailed proofs an be found in Paumier and Renard (1999) and in the
submitted paper.
6. A priori estimate
Proposition 1 Under the same assumptions as Theorem 1, ue'T/(t) be-
ing the unique solution to Problem (3) belonging to Loo(O, Tj Ve) n
W1,00(0, Tj He), there exists a constant C > °
independent of c and
299
1} such that for almost all t in (0, T) one has
+ [t [ J~ (u~l1(t))hl1(lu~l1(t)l)dO"ds ~ C,
10 1re
where hl1 : R+ ~ R+ is defined by hl1 (e) = J.t(e) e
1}
if e< 1}, and hl1 (e) =
e
J.t(e)e if ~ 1},
In particular the {double} sequence u el1 is bounded in Loo(O, Tj Vo) and
the sequence u el1 is bounded in Loo(O, Tj Ho).
The proof is based on energy conservation. Due to the lack of regu-
larity, a finite dimensional approximated problem is first considered.
8. Conclusion
The proposed perturbation in an arbitrary thin layer near the contact
boundary, allows to regain the existence and uniqueness of the solution of
300
the dynamic problem of friction, at least for the simplified geometry we
considered, and for an arbitrary large friction coefficient. It is known, for
such a problem, that uniqueness is lost for large coefficients of friction.
So it should be interesting to know if, as it is the case in dimension one
(see Renard (2000)), the solution of the perturbed problem tends to a
particular solution of the non-perturbed problem related with a certain
criterion.
In any case, this perturbation seems to be a useful tool to deal with
dynamic contact and friction problems.
References
G. DUVAUT, J.L. LIONS (1972), Les inequations en mecanique et en physique, Dunod
Paris.
I.R. IONESCU, J.-C. PAUMIER (1994), On the contact problem with slip rate dependent
friction in elastodynamic, Eur. J. Mech., A/Solids, N° 4, pp 555-568.
J.A.C. MARTINS, S. BARBARIN, M. RAous, A. PINTO DA COSTA (1999), Dynamic
stability of finite dimensional linearly elastic systems with unilateral contact and
Coulomb friction, Compo Meth. Appl. Mech. Engng., vol. 177, pp 289-328.
F. MOIROT, Q.-S. NGUYEN (2000), An example of stick-slip waves, C.R. Acad. Sci.
Paris, II B-Mec., 328 (9), pp 663-669.
J.-C. PAUMIER, Y. RENARD (1999), F'rottement Elastodynamique avec Inertie de Sur-
face. Rapport IMAG RT 179, Universite Joseph Fourier, Grenoble, France.
J.-C. PAUMIER, Y. RENARD (2001), Existence and uniqueness result for a perturbed
purely elasto-dynamic problem with friction. Submitted to the European Journal
of Applied Mathematics.
y. RENARD (1998), Modelisation des instabilites liees au frottement sec des solides
elastiques, aspects thioriques et numeriques., These de doctorat, LMC-IMAG Greno-
ble.
y. RENARD (2000), Singular perturbation approach to an elastic dry friction problem
with a non-monotone friction coefficient, Quarterly of Applied Mathematics, vol.
LVIII, W 2, pp 303-324.
F.M.F. SIMOES, J.A.C. MARTINS (1998), Instability and ill-posedness in some friction
problems, Int. J. Engng. Sci., 36, pp 1265-1293.
SOLVABILITY OF THERMO-VISCO-
ELASTIC CONTACT PROBLEMS
WITH COULOMB FRICTION AND
NONLINEAR HEAT CONDUCTIVITY
Christof Eck
Institute of Applied Mathematics,
University Erlangen-Niimberg
Jifi J arusek
Mathematical Institute,
Academy of Sciences of the Ozech Republic
1. Introduction
In models for contact problems with friction it is often necessary to
include the aspect of heat transport, because friction may generate a
non-negligible amount of heat. Since the solvability of unilateral dy-
namic problems with Coulomb friction seems to be up-to-now estab-
lished only for material having some viscous behaviour (cf. Jaruaek and
Eck (1999)), a suitable model for such a problem consists of a viscoelastic
contact problem and a heat equation. As in the cited paper, the approx-
imate contact condition in velocities is employed here. This limits the
applicability of the presented result in the manner mentioned there.
Let us formulate the problem to be considered. We assume a bounded
domain n whose Lipschitz boundary r is composed of three measurable
mutually disjoint parts ru, rT and re. Let Irr = (O,!) be a bounded
time interval. Let Qrr == Irr )I n be the appropriate time-space domain
301
I.A. C. Martins and M.D.P. Monteiro Marques (eels.), Contact Mechanics, 301-308.
© 2002 Kluwer Academic Publishers.
302
and let S'! denote its lateral boundary consisting of the parts SX,'! ==
I,! x rx for X = U, T, C. For TEl,! an analogous notation In Q-r, S-r
will be used. We shall employ a usual summation convention, dots for
time derivatives, ',i for the derivatives with respect to i-th variable, u for
a displacement, 6 for temperature, n for the unit outer normal vector
and subscripts n and t for normal and tangential components of vectors.
We assume the linear strain-stress relation of the Kelvin-Voight type
with eij(u) == ~(UiJ + Uj,i)' With this notation the problem has the
following classical formulation:
Find a couple [u,8] such that the following relations are satisfied:
Uj - (1ijJ(U, 6) = Ii, i= 1, ... ,N, in Q'!, (2)
u=U on SU,,!, (3)
T(u) = h on Sr,'!, (4)
Un ~O, Tn ~O, Tnun = 0,
Ut = 0
Ut7~O
=}
=}
ITtl ~ ~ITnl,
Tt = -~ITnllutl
Ut } on Sc,'!, (5)
Lipschitz with respect to the space variable and globally bounded. The
tensor of thermal conductivity Cij is assumed to be symmetric and locally
Lipschitz-continuously solution dependent. Two different models of this
dependence are studied:
Case A (cf. Eck (2000)) Cij == Cij(9) such that
eo(l + lel'Y)eiei ~ Cij(e)eiej ~ Co (1 + lel'Y)eiei (12)
for all vectors {ei} E]RN and all x E n with 0 < eo ~ Co < +00.
Case B (cf. JaruSek (2000)) Cij == Cij(Ve) satisfies the growth condition
Cl (1 + IV91'Y) eiei ~ Cij(Ve)eiej ~ C2 (1 + IV91'Y) eiei, e E JRN, (13)
the strong monotonicity
(cij(V9)9,j - Cij(VS)S,j, e,i - S,i}Q$
~ c311V(9 - S)III~!2(Q$) + c411V(9 - S)IIL(Q$) (14)
for each e, S E L'Y+2 (i'I'j W~+2(n)), and the continuity relation
for 9(k) --+ 9 strongly in L'Y+2 (I'I'j W~+2(n)) and '1* = (-y+2)/(-y+
1).
In both cases '1 is a positive number. The aim is to minimize '1 with
respect to the dimension N.
3. Result
For the generalized Lame system we assume:
Assumption 1 In addition to the above mentioned conditions on the
domain 0, its parts of boundary rx, X = U, F, C and the coefficient
functions a~j~l' bij and Cij, we assume re E c2+{3 for some f3 > 0 ar-
bitmrily small, f E L 2(Q't) UHI(I't;HI(O)*), hE HI (h; H I / 2(rT)*) ,
UO,UI E H3/2(O) and U E H2(Q't) such that U E L2(I't; HI(O)). More-
over, U satisfies the compatibility conditions U = 0 on Se,'t, U(O,') = Uo
and U(O,.) = Ul on O. The coefficient of friction J =
J(x,u) is a
nonnegative function satisfying the Camtheodory conditions, supp J c
=
re,w x ~N with re,w {x Ere; dist(x, are) ~ w} for some w > 0 and
IIJIIL oo (rc) < C'J, where the constant C'J (always dependent on the viscos-
ity of the material only) is given in Jarusek and Eck (1999), Proposition
4 and formula (4.23) for an anisotropic material and in Eck and Jarusek
(1998) for an isotropic material in two dimensions.
For the heat equation we assume:
Assumption 2 The input data satisfy eo
~ 0, 0:$ T E L2(S't), Ko :$
K E Loo(r) with a constant Ko > O. The mdiation R R(x, e) satisfies =
the Camtheodory conditions, R(·,O) = 0, R =
0 or R(x,') is increasing
on [0, +(0) and satisfies the growth condition
cller R :$ R(e) :$ c21erR + C3 for e~0
305
(with p = 2N~i~~~+1) in the second relation) for the case with and
without radiation, respectively, is obtained with a constant indepen-
dent of the approximation parameters {j, 'fJ and~. These and the ap-
propriate dual estimate derived with the help of test functions from
If
Loo (h; if" (0)) with '" > 1 + finally leads to the a priori estimate
+ lIeIlLrR+l_-y(8'X) ~c
with 0 < s < 4~N and a constant independent of {j, 'fJ and~. This
estimate enables to perform successively the necessary limit procedures
for all approximating parameters and Theorem 1 is proved in this case
for, E (To, 1). The case, ? 1 is not so difficult.
In Case B the scheme of estimation is similar. For N = 2 the most
difficult term to be estimated is the frictional heat. For R = 0 we put
308
Acknowledgments
The second author gratefully acknowledges support by the Grant
Agency of the Acad. Sci. of the Czech Republic, grant No. 107 5005.
References
Eck, c., and JaruSek, J. (1998), Existence of solutions for the dynamic frictional con-
tact problem of isotropic viscoelastic bodies. Preprint Nr. 242, Inst. Appl. Math.,
Univ. Erlangen-Niirnberg.
Eck, C. (2000), Existence of solutions to a thermo-viscoelastic contact problem with
Coulomb friction and heat radiation. Preprint Nr. 271, Inst. Appl. Math., Univ.
Erlangen-Niirnberg.
JaruSek, J., and Eck, C. (1999), Dynamic contact problems with small Coulomb fric-
tion for viscoelastic bodies. Existence of solutions. Math. Models Meth. Appl. Sci. 9,
11- 34.
Jarusek, J. (2000), Solvability of nonlinear thermo-viscoelastic contact problem with
small friction and general growth of the heat energy. Preprint Nr. 273, Inst. Appl.
Math., Univ. Erlangen-Niirnberg.
A SIMPLIFIED MODEL OF IMPACT
Jeongho Ahn
University of Iowa
Department of Mathematics
David E. Stewart
University of Iowa
Department of Mathematics
1. Introduction
In this paper we consider a simplified model for impact where the
standard pointwise Signorini contact condition is replaced by an inte-
grated version which corresponds to an a priori normal contact force
distribution. Comparison is also made to some recent work of the au-
thors for the standard Signorini contact conditions. Proofs are omitted
for conciseness.
The simplified contact conditions have the form
where the contact force (or traction) on the boundary is - N (t) 'IjJ (x) n( x).
There is no friction in this model. The reason for this kind of contact con-
dition is that it results in a scalar (dynamic) complementarity problem
while keeping all the infinitely many elastic modes of vibration. Since
simulations appear to indicate that these elastic vibrations are a major
309
J.A.C. Martins and M.D.P. Monteiro Marques (eds.). Contact Mechanics. 309-316.
© 2002 Kluwer Academic Publishers.
310
8 2u
{)t2
V . O"[u] + f(x) inn (2)
O"[u]· n N·n on8n (3)
O?u·n-g ...L N 5, 0 on 80. (4)
faan N(u·n-g)ds 0 for all t (5)
a[u](x, t) . n(x) 'I/J(x) n(x) N(t) for all x E an, and t, (6)
02: N(t) 1- ( 'I/J(x) (u(x, t) . n(x) - g(x)) dS(x) sO. (7)
Jan
If we use the mid-point rule for the elasticity operator and the implicit
Euler rule for the simplified contact conditions using ul(x) ~ u(x, tl) and
vl(x) = u(x, tl) for tl = to + 1h, we get
vl+ 1 _ vl l l
h
= \1·a [u +12+ u ] +1 in 0., (8)
ul+1_ ul 1
= _(vl+ 1 + vl) in 0., (9)
h 2
a[ul+ 1]. n 'l/JnNl on an, (10)
02: N l 1- 1an
'I/J(u1+1. n - g)dSSO. (11)
Uj = r
Jan
N(t)'l/J(x)n(x)· <Pj dS - Aj Uj + fj(t) for all j. (13)
Let wJ
= Aj and Wj ~ o.
The simplified contact law (1) can be written in terms of the modal
coefficients Uj as
o~ L
00
3. Numerical approximation
The system that we want to solve is
(15)
L {3jUj(t) ~ 0,
00
o ~ N(t) 1. (16)
j=l
for all j = 1,2,3, .... Numerical approximations are needed to prove ex-
istence of solutions to this system of differential equations and comple-
mentarity conditions. Let v = U and Vj = Uj. Let h > 0 be the step-size
and tl = l h. Given u; ~ Uj(tl) and v; ~ Vj(tl) we wish to compute u;+1
and v;+l. To explicitly show the dependence on h > 0 we write these as
ul+l;h
j , Vjl+l;h ,etc. L et UI -_ (UI'
I I
u2' I
u3'··· )
an d vI -_ (VI'
I I I
V2' V3' ... ) ,etc.
As usual, £2 is the Hilbert space of sequences x = (Xl, X2, X3, •.• ) where
Ilxll = J2:: j IXjl2 is finite.
Now all we need to check is that the complementarity condition (11)
can be made to hold, where (using (12)):
00
L{3jU;+1
j=l
313
Let us assume that the energy in (u l , vl) is finite, and that fl is in £2.
Then as long as the sequence ((3jl(1 +Wj) I j = 1,2, ... ) is in £2, then all
sums above are finite. Also assuming f3 =1= 0 (that is, f3 = ((31, (32, (33,· .. )
contains a non-zero entry) makes l:: j (3J I (1 + h2 Aj 14) a finite, positive
number for any h > o. The complementarity condition (11) can then be
seen to have the form
o ~ N l .1 cl(h) + c2(h)Nl ~ 0
where c2(h) is positive. This is a simple one-dimensional complemen-
tarity problem and has a finite solution N l ~ O. Substituting this value
for N l shows that there are solutions to (8-11) provided f3 =1= 0 and
((3jl(1 + wj)li ~ 1) is in £2.
The energy E(u,v) =! l::~1(V] + AjU] + 2/juj) is dissipated in the
scheme (8-11); that is, E(u l+1,vl+1) ~ E(ul,v l ), provided L-j(3jU; +
'flo ~ o. Thus we can obtain numerical trajectories uh(t) by piecewise
linear interpolation of Uh(tl) = ul;h and vh(t) by piecewise linear inter-
polation of Vh(tl) = vl;h. The function Nh(t) = Nl for tl ~ t < tl+1 is
the numerical force function for this step size: Nl = (1Ih) Jt~l+1 Nh(t) dt.
4. Convergence theory
Taking (Stewart, 1998) as a guide, it would be desirable and appro-
priate to build on these results to give a complete convergence theory
for this simplified impact problem. Because we are dealing with an infi-
nite dimensional problem, some convergence issues become considerably
trickier than in the finite-dimensional case in (Stewart, 1998).
One issue that remains for the finite-dimensional case is that the func-
tions Nh do not converge to an ordinary integrable function, but rather
to a measure, such as a Dirac-8 function. It can be proven that, pro-
vided E(uO,vO) < 00, limsuPhloJoT Nh(t)dt <00. Since Nh is a scalar-
valued function, we can use the Helly selection theorem (Natanson, 1961,
Ch. VIII, §7, pp. 222-223) or Alaoglu's theorem (Lang, 1993, p. 71) to
conclude that there is a subsequence Nhk which converges weak* in the
space of measures to a measure N on [0, T] with scalar non-negative val-
ues. In the finite-dimensional case, this is sufficient to obtain convergence
of u hk and v hk • However, it is not generally so in infinite-dimensions.
Convergence of subsequences of (u h , v h ) in the following weak senses can
be shown. Since Ilvh(t)11 is uniformly bounded by the energy bound, it
follows that v h is uniformly bounded in V(O, T; £2) for 1 ~ p ~ 00,
the space of functions 'l/J : [0, T] ~ £2 where J[ I 'l/J (t) liP dt < +00.
Then by Alaoglu's theorem, there are weakly convergent subsequences
in V(O, T; £2) for 1 ~ p < 00, but only weakly* convergent subsequences
314
in LOO(0,T;£2) = L1 (0,T;£2)*. In the search for strong convergence,
there are two cases to consider: (a) the "easy" case where f3 E £2, and
(b) the "hard" but physically appropriate case where f3 ¢ £2.
This result can be extended to the full elasticity operator with more
work. These are now used to bound Nl;h.
First we note that l:~1 f3jU~+1 must be finite. Writel:~1 f3jU~+1 =
l:f=l f3jU~+1 + l:~J+1(f3j/Wj)(WjU~+l). Since wou l+1 E £2 and (f3j/Wj I
j > J) E £2 (0 is the Hadamard or componentwise product of vec-
tors in £2/' we see that the sum on the right is finite, as desired. If
l:~l f3ju/ 1 < 0 then N l = o. So for N l < 0 we have l:j f3jU~+1 = 0
but 'E j f3jU~ :::; OJ thus INII (h 2/2) 'Ej f3J/(1 +h2>'j/4) is bounded by the
absolute value of l:j( -h2>.jf3juV2 + hf3jv~ + (h 2/2)f3jfj )/(1 + h2>'j/4).
For the term 'E j -h2wjf3jwju;/(2(1 + h2>'j/4)) we bound the £2 norm
of the vector h 2diag(2 + h2>../2)-I(w 0 (3). Using (17) this can be es-
10
timated by the square root of 00 h 4 >.>.-1/2/(2 + h2>'/2)2d>' = 8(h).
Thus Ilh 2diag(2 + h 2>../2)-1(w 0 (3)11 = 8(hl/2). Similarly, Ildiag(1 +
h2>"/4)-1 (h{3) II = 8(hl/2), and II(h2/2)diag(1 + h2>"/4)-1{311 = 8(h 3 / 2).
Combining these asymptotic results, we obtain,
Nl;h = O(h-l/2).
This result is indicative of the fact that the contact force functions Nh(t)
are not only uniformly bounded as measures, but are bounded under
stronger norms.
References
Anitescu, M., Potra, F. A., and Stewart, D. E. (1999). Time-stepping for three-
dimensional rigid body dynamics. Compo Methods Appl. Meeh. Engineering,
177:183-197.
Lang, S. (1993). Real and Functional Analysis, volume 142 of Graduate Texts in Math-
ematics. Springer, Berlin, Heidelberg, New York, 2nd edition.
Natanson, I. P. (1961). Theory of Functions of a Real Variable. Frederick Ungar Publ.
Co., New York. Revised edition. Original translated from the Russian by L.F.
Boron and annotated by E. Hewitt.
Oleinik, O. A., Shamaev, A. S., and Yosifian, G. A. (1992). Mathematical Problems
in Elasticity and Homogenization, volume 26 of Studies in Mathematics and ist
Applications. North Holland, Amsterdam, London, New York, Tokyo.
Renardy, M. and Rogers, R. C. (1993). An Introduction to Partial Differential Equa-
tions, volume 13 of Texts in Applied Mathematics. Springer Verlag, New York,
Berlin, Heidelberg.
Schatzman, M. (1973). Sur une classe de problemes hyperboliques non liw3aires. C.
R. Acad. Sci. Paris Ser. A-B, 277:A671-A674.
Stewart, D. E. (1998). Convergence of a time-stepping scheme for rigid body dy-
namics and resolution of Painleve's problems. Archive for Rational Mechanics and
Analysis, 145(3):215-260.
Stewart, D. E. (2000). Rigid-body dynamics with friction and impact. SIAM Review,
42(1):3-39.
CONVERGENCE FOR A TIME DISCRET-
IZATION OF DYNAMIC CONTACT PRO-
BLEMS WITH FRICTION
E. Pratt
Universite de Provence and
Laboratoire de Mecanique et d'Acoustique, C.N.R.S.,
31 chemin J. Aiguier, 13402 Marseille cedex 20.
J.-M. Ricaud
Laboratoire de Mecanique et d'Acoustique, C.N.R.S.,
31 chemin J. Aiguier, 13402 Marseille cedex 20.
1. Introduction
The models adopted to describe the physical phenomena of frictional
contact induce numerous difficulties, such as strong non-linearities or the
implicit status of the constraint convex. Most of the results obtained are
issued from the study of various operators coupled with an unilateral
constraint (see Brezis (1971), Moreau (1983)). From a frictional contact
mechanics point of view, Duvaut and Lions (1972) elaborated variational
formulations through which they obtained some first results. Confronted
with the difficult task of formulating dynamical contact problems, Mar-
tins and Oden introduced the so called normal compliance model for
which they obtained an existence and uniqueness result in Martins and
Oden (1987). Recently, Jarusek and Eck (1999) obtained some results
concerning dynamical unilateral contact problems with local Coulomb
friction, using a velocity formulation for the contact.
We begin in section 2 by setting the generic implicit variational in-
317
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 317-324.
© 2002 Kluwer Academic Publishers.
318
2. Variational formulation
We consider a viscoelastic body that initially occupies a domain 0 in
IRd , d = 2,3. The boundary of 0 denoted by r is sufficiently regular.
This boundary is divided into three disjoint and open parts: r =
r u u r fur c. The body is subjected to volume forces of density F I ,
prescribed zero displacements and tractions F2 on the parts r u and r f
respectively. On r c, the body may come into contact with a foundation.
We shall denote by u = (Ui) the displacement field, by (J=((Jij) the stress
tensor, by € = (€ij) the strain tensor, for i,j E {1, .. ,d}. We denote by
p the mass density of the material, and respectively by A = (aijkl) and
B = (bijkl) the elasticity and viscosity tensors with the usual properties
of ellipticity and symmetry. The strong formulation of the problem is as
follows.
Find u = u(t,x) such that u(O,x) = ua,u(O,x) = ua and for all t E]O,T[,
div (J + FI = P ii. in 0, (J = A €(u) + B €(u),
U = ° on r u and (J.n = F2 on r f.
As frictional contact can occur on the part r c of the boundary, some
extra conditions depending on the models of contact and friction adopted
should be added to this formulation. In the following generic weak for-
mulation of the problem, these additional boundary conditions are taken
into account by the mapping ¢.
Find u E K such that for almost all t E]O, T[ and for all v E V:
According to the model one adopts for contact and friction, the map-
ping ¢ is given by:
ior
liminf T ¢(t,Uk(t), Uk (t))dt 2:: r T ¢(t,u(t),u(t))dt,
k--+oo io
for all u E V, ¢l (u, .) is linear,
¢2(0, uo,.) = 0, ¢2(t, u,.) is a semi norm,
3 'fJ > 0, 'v'Vb V2, Wb W2 E V, such that 'v'tb t2 E [0, T],
I¢(tb VI, wt} - ¢(tb vb W2) + ¢(t2,v2,W2) - ¢(t2,v2,WI)1
~ 'fJ {IIVI - v211 + It I - t21} IIWI - w211·
Lemma 2
There exists no E 1N* and C E R+ such that:
'Vn ~ no, Ildlllv ~ c.
Lemma 3
There exists no E 1N* and Cll C2, C3 E R+ such that 'Vn ~ no:
'Vi E {1,··· ,n -1}, Iluiliv ~ Cll IlfJillv ~ C2, l'YilH ~ C3, Ilunllv ~ Cl.
The proof of these lemmas rests uniquely on the assumptions of sec-
tion 2 and the constants exhibited depend only on the initial data. Let
us point out that the presence of viscosity is essential in these proofs
nevertheless these lemmas hold however little viscosity is present and
also however stiff the contact and friction conditions are.
321
Lemma 4
There exists a subsequence (nk)kElN such that:
°
The convergence result is now obtained in the following way. We be-
gin by integrating (4) between and T and going to the limit in the
resulting inequality. The limit u obtained in lemma 4 is thus shown to
satisfy the following inequality:
322
Vv E L2(O, Tj V) :
Theorem 1
The sequences Un, Un and Un converge weakly (in the sense of lemma 4)
towards the unique solution of problem (1).
4. Numerical results
In this section we give the numerical results we have obtained by com-
puting a Q2 finite element approximation of the incremental solution
given by (3). Contact is modelized by a normal compliance law, friction
by a Coulomb law. The discretized problem thus obtained becomes a
finite dimensional implicit inequality that can be set as a minimization
problem by using a fixed point procedure.
The mechanical benchmark consists of a two dimensional rectangu-
lar solid, 20mm by 160mm which may come into contact with a plane
obstacle on one of its long sides and has some prescribed displacement
on its two sides. The prescribed force is sinusoidal in time. Such a me-
chanical test was studied in Martins and Oden (1985) but they used a
different normal compliance law and a regularization of the friction law,
instead of the Coulomb law.
The geometry of the problem with the data concerning the loading,
the prescribed displacement and the contact zone are described in Fig-
ure 1, because of the symmetry we consider only the left hand side of
the solid.
323
.n'" .n
~
Scm
The Young modulus equals 210000 mPa, the Poisson ratio equals
0.25 and the mass density equals 7900kg/m 3 •
A Rayleigh damping method is used to take the viscosity into account.
The evolution of the normal and tangential displacements of a node at
the centre of the contact zone are plotted in Figure 2. The evolution of
the ratio UT / UN and the evolution of the tangential velocity of the first
node of the contact zone are plotted in Figure 3.
0.3
l l
0.2
0.1
-0.1 l, r
0.0
0.0 , 0.00 , 0.00 , 0.0 01
-0.2
-0.3
References
BR Brezis, H., Problemes unilateraux. Master Thesis, Universite de Paris VI, fevrier
1971.
Cocu, M., Existence of solutions of Signorini problems with friction. Int. J. Engng
Sci., Vol. 22, p. 567-575, 1984.
Duvaut, G., and Lions, J.-L., Les inequations en mecanique et en physique. Dunod,
Paris, 1972
JaruSek, J., and Eck, C., Dynamic contact problems with small Coulomb friction for
viscoelastic bodies. Existence of solutions. Math. Models and Meth. Appl. Sci.,
Vol. 9, p.11-34, 1999.
Martins, J.A.C., and Oden, J. T., Existence and uniqueness results for dynamic contact
problems with non-linear normal friction and interface laws. Nonlinear Analysis,
Theory Meth. Applic., Vol. 11, p. 407-428, 1987.
Moreau, J.-J., Liaisons unilaterales sans frottement et chocs inelastiques. C. R. Acad.
Sci. Paris, Vol. 286, p. 1473-1476, 1983.
Oden, J.T., and Martins, J.A.C., Models and computational methods for dynamic
friction phenomena. Comput. Meth. Appl. and Engng., Vol. 52, p. 527-634, 1985.
Ricaud, J.-M., and Pratt, E., Analysis of a time discretization for an implicit varia-
tional inequality modelling dynamic contact problems with friction. Mathematical
Methods in the Applied Sciences, Vol. 24, p. 491-511, 2001.
Ricaud, J.-M., Etude d'une classe d'inequations d'evolution implicites et application
a des problemes dynamiques de contact avec frottement. Ph. D. thesis, Universite
de Provence, novembre 1999.
SOLUTION METHODS
FOR STRUCTURAL OPTIMIZATION
IN CONTACT ROD PROBLEMS
Isabel N. Figueiredo
Departamento de Matematica, Universidade de Coimbra
Apartado 3008, 3001-454 Coimbra, Portugal
Joaquim J. Judice
Departamento de Matematica, Universidade de Coimbra
Apartado 3008, 3001-454 Coimbra, Portugal
Silverio S. Rosa
Departamento de Matematica, Universidade da Beira Interior
6201-001 Covilha, Portugal
1. Introduction
We address a structural optimization problem which involves a unidi-
rectional fiber reinforced composite rod in contact with a rigid founda-
tion. The purpose of the model is to find the fiber volume fraction and
the value of some geometric parameters of the rod in order to minimize
the compliance of the structure. The variables are continuous and vary
in an admissible set.
325
l.A.C. Martins and M.D.P. Monteiro Marques (eels.), Contact Mechanics, 325-332.
© 2002 Kluwer Academic Publishers.
326
ph~
12
- Ph~l
12 . (3)
328
minO(x,u) = minluTA(x)u
subject to:
u E RIHI, W E RIHI
MPEC (8)
xEX and { A(x)u - F(x) = W
uJ~O, wJ~O, WK=O
U}WJ = O.
For each x, the vector u(x) is the solution of the complementarity prob-
lem and 'V xF and 'V xA are the gradients of F and A defined by
4. A Numerical Example
We have tested the previous algorithms in an example. The material is
an unidirectional fiber reinforced composite material, whose longitudinal
modulus is E = EIVI + Em(l- VI) with EI the modulus of the fiber,
Em the modulus of the matrix and VI the fiber volume fraction which
belongs to [0,1]. The optimization variable is VI and the data of the
problem are displayed in the table below.
Parameter Value
Em (OPa) - modulus of the matrix 3.45
E, (OPa) - modulus of the fiber 86
Vr Vrin , ax -lower and upper bounds for V, 0.01,0.99
L (m) - length of the rod 10
q (N) - distributed force in the direction of the rod axis -18000
p (N) - distributed force in the direction perpendicular to the rod axis -200
9 (m) - obstacle (constant) -0.001
Iwl (m 2 ) - area of the cross section 0.004
J(m4) - moment of inertia 2.1e- 6
The symbols (GPa), (N) and (m) denote the units Giga Pascal, New-
ton and meter, respectively. The interval [0,10] has been discretized
with 50 finite elements, whose length hi is constant and equal to 0.2.
The projected-gradient algorithm has successfully found an optimal
solution VI = 0.3518 in 7 iterations. We have employed the block pivot-
ing algorithm described in Fernandes et al. (2002) to process the LCP's
required by the projected-gradient method. The block pivoting algo-
rithm has required a total of 150 iterations to process all the 44 LCP's
needed by the projected-gradient algorithm. In these tests the subrou-
tine beam2e of the CALFEM toolbox of MATLAB has been used to
evaluate the stiffness matrix A and the force vector F. The comple-
mentarity algorithm has been implemented in MATLAB. The figure 1
shows the plot of the equilibrium bending displacement of the rod, at
the optimum value VI = 0.3518.
As stated in this section, the projected gradient algorithm has per-
formed well for solving this example of MPEC problem discussed in this
paper. Further computational investigation is required to evaluate the
performance of the algorithm in practice. A description of this experi-
ence will be reported in the near future.
332
X 10-3
2 4 5
References
Fernandes, L.M. ; Figueiredo, I.N. and Judice, J.J. (2002) On the solution of a finite
element approximation of a linear obstacle plate problem, International Journal
of Applied Mathematics and Computer Science (to appear).
Harker, P.T. and Pang, J.S. (1990) Finite-dimensional variational inequality and non-
linear complementarity problems: a survey of theory, algorithms and applications,
Mathematical Programming 48,2, 161-220.
Haslinger, J. and Neittaanmi:i.ki, P. (1997) Finite Element Approximation for optimal
Shape Design, Theory and Applications, Wiley, Chichester.
Nocedal, J. and Wright, S. (1999) Numerical Optimization, Springer, New York.
Outrata, J. ; Koevara, M. and Zowe, J. (1998) Nonsmooth Approach to Optimization
Problems with Equilibrium Constraints, Kluwer Academic Publishers, Dordrecht.
Petersson, J. (1995) Optimization of Structures in Unilateral Contact, Linkoping Stud-
ies in Science and Technology, Dissertations n.397, Linkoping University, Sweden.
Trabucho, L. and Viano, J.M. (1996) Mathematical Modeling of Rods, in P.G. Ciarlet
and J.L Lions (edts), Handbook of Numerical Analysis, VoI.IV, 487-974, North-
Holland, Amsterdam.
UNILATERAL CONTACT, FRICTION AND
ADHESION: 3D CRACKS IN COMPOSITE
MATERIALS
Michel Raous
Laboratoire de Mecanique et d'Acoustique - CNRS
Marseille - France
Yann Monerie
Laboratoire de Mecanique et d'Acoustique - CNRS
Marseille - France
Abstract The model coupling adhesion, unilateral contact and friction, devel-
oped by M. Raous, L. Cangemi and M. Cocou, is extended here for
studying propagating cracks in fiber-reinforced composites. The brittle
fracture of the bulk materials (fiber and matrix) is described by a time-
independent version of the model. In this case, jumps in the evolution of
the solution may occur and a dynamic formulation has to be used. This
is performed by adapting a dynamic formulation due to J.J. Moreau
and M. Jean to this problem. Numerical simulations of 3D fracture of
composite materials are then presented.
1. Introduction
The present work deals with interface problems where an initial ad-
hesion due to chemical or thermal treatments can disappear during the
loading and be replaced by frictional sliding. A model taking into ac-
count adhesion, friction and unilateral contact was proposed by
M. Raous, L. Cangemi and M. Cocou (the RCC model) with a quasi-
static formulation in Raous et al. (1997, 1999). It was used to simulate
the micro-indentation of a fiber in a composite material. A comparison
between this model and those developed by Tvergaard (1990), Allix et
al. (1995, 1998), Ladeveze (1995), Michel and Suquet (1994), Michel et
al. (1994) and Chaboche et al. (1997, 2001) was presented in Monerie
et al. (1998). In the RCC model, using the adhesion intensity vari-
333
I.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 333-346.
© 2002 Kluwer Academic Publishers.
334
able introduced by F'remond (1987, 1988) and Trong Dinh Tien (1990),
adhesion, friction and unilateral contact are described via an unilateral
elastic behaviour of the interface which fulfills the non penetration con-
ditions and depends on the intensity of adhesion. Viscosity was taken
into account in the evolution of the adhesion intensity.
In the present work, the model is extended and a new formulation is
proposed in order to describe the propagation of a crack in a composite
material and to study the interactions occurring between the crack and
the fiber-matrix interfaces. To study the crack propagation, a model
without interface viscosity has to be used (and friction will be neglected).
In this case, rapid changes in the evolution of the solution may occur
and a dynamic formulation has to be used. We present here :
- a new form of the RCC model introducing a new term which gives a
better transition from adhesion to friction,
- a dynamic formulation, which is an extension to the RCC model of the
Non Smooth Contact Dynamics method (NSCD method) developed by
Jean (1999) and Moreau (1988, 1999),
- the simulation of 3D fracture in a composite material.
(2)
with
if IIRT - R~II < 1-'(1- (3) IRN - (32CNUNI =} uT = 0,
if II~ - R~II = 1-'(1- (3) I~ - (32CNUN =} 3'\ ~ 0, u
T = ,\(~ - ~),
(3)
begins to decrease, the apparent stifi'nesses (32CN and (32CT also begin
to decrease (softening) and the friction begins to act. When (3 tends to-
wards zero, classical Coulomb friction behaviour is obtained (the limit is
asymptotic). This behaviour can be compared with that of the standard
RCC model given in Figure 3, where the dotted line gives the evolution
with infinitely small velocities. Figure 3 gives a clearer picture than
Figure 2 of the energy exchanges. The interface energy is dissipated by
viscosity, friction or adhesion (when the intensity of adhesion begins to
decrease). A reversible elastic part is stored at the interface.
i
I
l -------------+~
i
1
·2
-5
----
. 1+ - - - - - --11
~ cydo •• p.-!b«I
-4
~ty
-f-.--r-'..:...
-3
,.-.-....""
...-.-
-2 -I
...
~r__+_._+_~~r___.,....._+
0
Utr/UON
2 3 -4
At the point P :
• Energy dissipated
by viscosity
1+ ~ I RN I Energy consu med
Ro by the damage
• Energy stored In the
interfa.ce
$ Stored complementary
energy (viscosity effect)
~ Energy diBBipated
by friction
1 2 3 4
At the point P ,
• Energy dissipated
1 + J'IRNI by viscosity
Ro
Energy consumed
by the damage
• Energy stored in the
interfacial bond
J'IRNI
Ro
1 2 3 4
where:
-u = (uI, u2) where u1 and u2 define the displacements in 0 1 and 0 2,
- V = (VI, V 2), va = {va E [H 1(oa)]3;va = 0 a.e. on a = 1,2 ru},
(ru are the parts of the boundaries where the displacements are pre-
scribed),
- H = LOO(re ),
- K = {v = (vI,v 2) E VI X V2;VN ~ Oa.e. onre}, where re is the
contact boundary between the two solids 0 1 and O2,
- a(.,.) is the bilinear form classically associated to the elasticity map-
ping,
- j(,8,UN,VT) =! J.t(1- ,8)lu(RN) - ,82CN uN lllvT llds , where U(RN) is
rc
a linear and compact mapping, u(.) : H-l/2(r e) ~ L2(re) (non local
friction),
- F = (PI, p 2 ) are the given force densities applied respectively to solid
1 and to solid 2.
In Raous et al. (1999), an incremental formulation was given and the
problem was solved by associating minimization techniques and a fixed
point method on the sliding threshold for the contact, and a O-method
for the differential equation. Details on the computational methods are
presented extensively in Raous (1999). This quasi-static formulation was
used to simulate the micro-indentation of a fiber in a SiC/SiC composite
and identify the constitutive parameters of the fiber-matrix interface.
This formulation has been extended to hyper-elastic bodies in Bretelle
et al. (2001).
339
Theorem
There exists a fixed point, k, for the application s :
s : k -+ 1/ ([u(k)]) = R([u(k)] , h([u(k)]))
where u is the solution of :
a(u, v - u) - (1', v - u) + r k. ([v] -
ire
[U]) ds ~0 'Iv E K. (11)
and
where:
- CO is the Lipschitz constant for the application
R(.,.) : L2(rc)3 -+ L2(rc)3,
- I is the constant of coercivity of a(., .),
- >. is a constant related to the continuity of the injection from Hl/2 (r c)3
to L2(rc)3 and to the one of the trace mapping from Hl(OOi)3 to
Hl/2(rc)3 where a = 1,2.
rt2
it!
dq = q(tt) - q(tl) 'it2 > tl , (15)
341
and then:
l
~ ~ l~,bl
t2
q(t2) = q(tt} + q(r)dr. (17)
tl
The formulation due to Jean (1999) and Moreau (1988, 1999) is ex-
tended here to the adhesive frictional model (Monerie (2000)). The non
linear problem is then solved by using a generalization of a Newton-
Raphson algorithm to non-smooth functions due to Alart and Curnier
(1991). The implementation is performed in the 3D finite element code
LMGC (see Jean (1999) and Monerie and Acary (2001)).
'-
P=
r
INITlAL
_ MFrH!..<. Cn~ rt>-U. GFaiu~ CRACK
1 '111 ... 4 .~'t:~ .
Figure 5. Mesh of the fiber and the surrounding matrix (12246 ddl).
z
::::;0.08
1(i/~f = 0.35 1
...-
Q)
u
Q)
-50.06
s:::
o
s:::
o
·ZO.04
~
~
0.02
Figure 6. Propagation of the crack and of the breakdown area of the interface
(strong interface).
,---..
Z
"--'0 . 08
:::l
Q)
u
Q)
-50.06
s:::
o
s:::
o
'ZO . 04
~
~
0.02
Figure 7. Propagation of the crack and of the breakdown area of the interface (weak
interface) .
345
References
Alart, P., and Curnier, A. (1991) A generalised Newton method for contact problems
with friction, J. Mech. TMor. Appliq. 7, 67 - 82.
Allix, 0., Ladeveze, P., and Corigliano, A. (1995) Damage analysis of interlaminar
fracture specimen, Composite Structures 31(1), 61 - 74.
Allix, 0., Leveque, D., and Perret, L. (1998) Identification and forecast of delam-
ination in composite laminates by an interlaminar interface model, Composites
Science and Technologie, 58,671 - 678.
Bretelle, A.-S., Cocou, M., and Monerie, Y. (2001) Unilateral contact with adhesion
and friction between two hyperelastic bodies, Int. J. Engrg. Sci. 39, 2015 - 2032.
Cangemi, L. (1997) Frottement et adherence: modele, traitement numerique et appli-
cation al'interface fibre/matrice, Thesis, Universite de la Mediterranee, Marseille.
Chaboche, J.-L., Girard, R., and Levasseur, P. (1997) On the interface debonding
model, Int. J. Damage Mechanics 6, 220 - 257.
Chaboche, J.-L., Feyel, F., and Monerie, Y. (2001) Interface debonding model: a
viscous regularization with a limited rate dependency, Int. J. Solids Structures
38(18), 3127 - 3160.
Cocou, M., Pratt, E., and Raous, M. (1996) Formulation and approximation of qua-
sistatic frictional contact, Int. J. Eng. Sci. 34(7), 783 - 798.
Cocou, M., Cangemi, L., and Raous, M. (1997) Approximation results for a class
of quasistatic contact problem including adhesion and friction, in Variation of
domains and free-boundary problems in solids mechanics, edited by A. Argoul,
M. Fremond and Q.S. Nguyen, Solid Mechanics and its Applications 66, Kluwer,
211 - 218.
Cocou, M., and Rocca, R. (2000) Existence results for unilateral quasistatic contact
problems with friction and adhesion, Math. Model. Num. Anal. 34(5),981 -1001.
Fremond, M. (1987) Adherence des solides, J. Mec. The. Appl. 6(3), 383 - 407.
Fremond, M. (1988) Contact with adhesion, in Nonsmooth Mechanics and Appli-
cations, edited by J.J. Moreau and P.D. Panagiotopoulos, CISM - Courses and
Lectures 302, Springer Verlag, 177 - 221.
He, M.Y., and Hutchinson, J.W. (1989) Crack deflection at an interface between
dissimilar elastic materials, Int. J. Solids Structures 25, 1053 - 1067.
Jean, M. (1999) The Non Smooth Contact Dynamics method, Computer Meth. Appl.
Mech. and Engrg. 177(3-4), 235 - 257.
Ladeveze, P. (1995), A damage computational approach for composites: basic aspects
and mircomechanical relations, Computational Mechanics 17, 142 -150.
Leguillon, D., Lacroix, C., and Martine, E. (2000) Interface debonding ahead of a
primary crack, J. Mech. Phys. of Solids 48, 2137-2161.
Mathur, K.K., Needleman, A., and Tvergaard, V. (1996) Three dimensional analysis
of dynamic ductile crack growth in a thin plate, J. Mech. Phys. Solids 44(3), 439
- 464.
Michel, J.-C., and Suquet, P. (1994) An analytical and numerical study ofthe overall
behaviour of metal-matrix composites, Model. Simul. Mater. Sci. 2,637 - 658.
Michel, J.-C., Suquet, P., and Thebaud, F. (1994) Une modelisation du rOle des inter-
faces dans Ie comportement des composites a matrice metallique, Rev. Eur. Elem.
Finis 3(4), 573 - 595.
Monerie, Y. (2000) Fissuration des materiaux composites: role de l'interface fibre-
matrice, Thesis, Universite de la Mediterranee, Marseille.
346
Monerie, Y., Raous, M., Leroy, F.-H., Sudre, 0., Feyel, F., and Chaboche, J.-L. (1998)
Comparaison de lois d'interface fibre/matrice sur la base d'un modele uni-axial
d'experience de micro-indentation, in Comptes Rendus des Onziemes Jounees Na-
tionales sur les Composites, edited by J. Lamon and D. Baptiste, AMAC publisher,
565 - 574.
Monerie, Y., and Raous, M. (2000) A model coupling adhesion to friction for the
interaction between a crack and a fibre-matrix interface, Zeitschrift fUr Angewandte
Mathematik und Mechanik 80, 205 - 209.
Monerie, Y., and Acary, V. (2001) Formulation dynamique d'un modele de zone
cohesive tridimensionnel couplant endommagement et frottement, Rev. Eur. Elem.
Finis 10, 489 - 504.
Moreau, J.-J. (1988) Unilateral contact and dry friction in finite freedom dynamics, in
Non Smooth Mechanics and Applications, edited by J.J. Moreau and
P.D. Panagiotopoulos, CISM Courses and lectures 302, 1 - 82.
Moreau, J.-J. (1999) Some basics of unilateral dynamics, in Unilateral Multibody Con-
tacts, edited by F. Pfeiffer and Ph. Glocker, Solid Mechanics and its applications
72, Kluwer, 1 - 14.
Needleman, A. (1990) An analysis of tensile decohesion along an interface, J. Mech.
Phys. Solids 38(3), 289-324.
Needleman, A., and Tvergaard, V. (1998) Dynamic crack growth in a nonlocal pro-
gressively cavitating solid, Eur. J. Mech. A/Solids 17(3), 421 - 438.
Raous, M. (1999) Quasistatic Signorini problem with Coulomb friction and cou-
pling to adhesion, in New developments in contact problems, edited by P.Wriggers-
P.Panagiotopoulos, CISM Courses and Lectures 384, Springer Verlag, 101 - 178.
Raous, M., Cangemi, L., and Cocou, M. (1999) Consistent model coupling adhesion,
friction and unilateral contact, Compo Meth. in Appl. Mech. and Engrg. 177(3-4),
383 - 399.
Raous, M., Cangemi, L., and Cocou, M. (1997) Un modele couplant adherence et
frottement pour Ie contact unilateral entre deux soli des deformables, C. R. Acad.
Sci. Paris, Serie II b 329, 503 - 509.
Siegmund, T., Fleckand, N.A., and Needleman, A. (1997) Dynamic crack growth
across an interface, Int. J. of Fracture 85, 381 - 402.
'Truong Dinh Tien, J.-M. (1990) Contact avec adherence, Thesis, Universite de Paris
VI, Paris.
Tvergaard, V. (1990) Effect of fibre debonding in a whisker-reinforced metal, Mat.
Sci. Engrg. A125, 203 - 213.
Tvergaard, V., and Needleman, A. (1995) Effects of nonlocal damage in porous plastic
solids, Int. J. Solids Structures 32, 1063 - 1077.
Tvergaard, V., and Needleman, A. (1997) Nonlocal effects on localization in a void-
sheet, Int. J. Solids Structures 34(18), 2221 - 2238.
Vola, D., Pratt, E., Jean, M., and Raous, M. (1998) Consistent time discretization for
dynamical frictional contact problems and complementarity techniques, Rev. Eur.
Elements Finis 7, 149 - 162.
Xu, X.-P., Needleman, A., and Abraham, F.F. (1997) Effect of inhomogeneities on
dynamic crack growth in an elastic solid, Model. Simul. Mater. Sci. Engrg. 5, 489
- 516.
Xu, G., Bower, A.F., and Ortiz, M. (1998) The influence of crach trapping on the
toughness of fiber-reinforced composites, J. Mech. Phys. Solids 46(10), 1815 -
1833.
THE UNILATERAL FRICTIONAL
CONTACT OF A PIEZOELECTRIC BODY
WITH A RIGID SUPPORT
Dedicated to the memory of Professor P. D. Panagiotopoulos.
Paolo Bisegna
Dipartimento di Ingegneria Oivile, Universitd di Roma "Tor Vergata", Rome, Italy
Frederic Lebon
Labomtoire mecanique materiaux, Universite Lyon 1, Villeurbanne, I+ance
Franco Maceri
Dipartimento di Ingegneria Oivile, Universitd di Roma "Tor Vergata ", Rome, Italy
Abstract This paper is devoted to study the frictional contact of a linearly piezo-
electric body with a rigid foundation, under a quasi-static monotone
loading. Contact and friction are modelled by the laws of Signorini
and Coulomb. Primal and dual formulations of the contact problem
are presented and condensed on the boundary. A new numerical al-
gorithm, which is the dual version of the celebrated PANA algorithm
by P. D. Panagiotopoulos, is introduced to solve the dual discretized
formulation by a block-relaxation technique. Numerical examples are
presented, showing the effectiveness of the proposed algorithm.
1. Introduction
The problem of a piezoelectric body in frictional contact with a rigid
support, under a quasi-static monotone loading, is considered. This
is an extension of the classical Signorini-Coulomb problem. It has a
perspective technological importance: as an example, the relative motion
of two bodies may be detected by a piezoelectric sensor in frictional
contact with them.
In the first part of the paper, the governing equations are presented
in the framework of linear piezoelectricity (Tiersten, (1969». Two weak
347
J.A.C. Martins and M.D.P. Monteiro Marques (eels.), Contact Mechanics, 347-354.
© 2002 Kluwer Academic Publishers.
348
formulations are established taking into account the non-associated char-
acter of the friction. The first one, extending to piezoelectric materials
the classical primal approach (see, e.g., Lebon and Raous, (1992)), is
written in terms of a variational equation coupled with an implicit varia-
tional inequality, and involves the displacement and electric-displacement
fields as unknowns. The second one, generalizing the dual formulation (
Telega, (1988); Klarbring et al., (1989)), is written in terms of a varia-
tional equation coupled with a quasi-variational inequality, and involves
the stress and electric potential fields as unknowns.
The second part of the paper is concerned with the numerical so-
lution of the nonlinear problem associated with the dual formulation
presented above. The nonlinear discretized problem is solved by using
an iterative algorithm based on a condensation procedure coupled with
a block-relaxation technique. The idea of this technique is to solve, at
the k-th step, two sub-problems: the first is a problem of friction with
given normal forces, and the second is a problem of unilateral contact
with prescribed tangential forces. This is the dual version of the PANA
algorithm due to P. D. Panagiotopoulos (Panagiotopoulos, (1975)).
In the last part of the paper a numerical application is presented,
in order to show the effectiveness and robustness of the proposed algo-
rithm. The frictional contact of a linear piezoelectric plane body with
a rigid support is analyzed. Different values of the friction coefficient
are considered and the influence of an actuation electric potential on
the detachment zone and on the distribution of the contact forces is
investigated.
2. Formulations
2.1. Notations
Let R be the set of real numbers. Let (.) and (x) denote the inner
and vector products, respectively. A body, comprised of linear piezoelec-
tric material, occupies an open bounded region 0 of R3 with sufficiently
smooth boundary 00. The body, subject to monotone slowly increasing
loads, is in unilateral frictional receding contact with a rigid obstacle
along the part aco of 00. This problem is quasi-static both in the
sense of electromagnetism (essentially, electromagnetic waves uncouple
from elastic waves), and in the sense of mechanics (Le., inertia forces
are negligible). The following fields are introduced: (i) the electric po-
tential cp, the electric field E, the electric-displacement field D, (ii) the
displacement field u, the strain field e and the stress field u.
The body 0 is subject to: (i) a volume charge p, a surface charge w
on awo, a prescribed electric potential CPo on al"0' where {awO, al"0}
349
with
bl(u,T) f S U· T dx,
0
Ml(T) - f uo· Tn dl,
Bd O
~(rp,X) f f grad rp. grad X dx,
0 (9)
M2(X) - f p X dx - f w· X dl,
0 Bw O
h(rp, T) - f d T . grad rp dx,
0
l2(U, X) - f d U· grad X dx.
0
351
3. Numerical procedures
3.1. Introduction
Classically, the primal formulation is equivalent to a fixed point prob-
lem on the sliding limit (Duvaut and Lions, (1976». In order to compute
the solution, the problem is discretized by finite elements and usually
condensed on the contact zone, on the degrees of freedom corresponding
to the normal and tangential displacements.
Here we focus on the dual formulation. The discretized problem is
condensed on the contact zone on the degrees of freedom corresponding
to the normal and tangential reactions, and solved by a block-relaxation
method.
(10)
(11)
352
x
3
o v
l:S::,ssS::Ss:S;ssiSss~s;s:sssss:S;ssiSs~ss;s:sssss:S;ss~sss----",x
2
where bCT (resp. bTl corresponds to the normal (resp. tangential) degrees
of freedom.
At the k-th step of the proposed algorithm, the first su~problem
allows to find Tk, solution of:
mi~nTtaTT l' - Tt(bT - aTCT uk-d, 11'1 ~ -I-'Uk-I}, (12)
where Uk-I ~ 0 comes out from the step k - 1. Since aTT is positive
definite and the minimization set is nonempty closed convex, this sub-
problem has a unique solution.
The second sub-problem allows to compute Uk, solution of:
minqHutaCTCT u - Ut(bCT - (aTCT)tTk)' u ~ o}. (13)
As before, since aCTCT is positive definite and the minimization set is
nonempty closed convex, this su~problem has a unique solution.
In Bisegna et al., (2001) it is shown that, for a sufficiently small friction
coefficient 1-', the proposed D-PANA algorithm is convergent.
4. Numerical results
In this section, a numerical application is presented. A Cartesian
coordinate frame (O,XI,X2,X3) and a piezoelectric body extending in-
definitely in the Xl direction are introduced. The body has a rectangular
cross section, with dimensions 0.1 m and 0.05 m along the X2 and X3
axes, respectively (figure 1). It is initially in contact with a rigid plane
support along the side X3 = O. It is acted upon, on the opposite side,
by a uniform normal pressure of 100 N/mm2 distributed over the in-
terval 0.0625 m ~ X2 ~ 0.1 m, and by a difference of electric potential
V between its vertical sides. The electric and mechanical loadings are
constant along Xl.
353
(a) (b)
2OOr.====:::::;-~~-~---"I 100 r;=:====:::;-~-~-""7---'
-&- 11=0 -&- 11=0
150 -+- 11=0.2 -+- 11=0.2
~
..
~ --- 11=1 50 ---11=1
(c) (d)
200 ,
1'_' '02: 1 ,,
~
100
- OV
~ 150 - - 2V ~ 50 --2V I
'"
!!:.
.. 100
.. ../... I
~ 0 ,.' I
~
"., "
iii I----~c\,
,
iii 50 ~ \'
E ~ (
0
t:
0
S -50 \
,, \
, \
-50 -100'---~~~-~""""'~....L......I
0 0.02 0.04 0.06 0.08 0.1 o 0.02 0.04 0.06 0.08 0.1
contact zone (m) contact zone (m)
Figure 2. Influence of the friction coefficient J! and of the applied voltage V on the
contact forces: (a) normal and (b) tangential forces for three values of the friction
coefficient: J! = 0, J! = 0.2, J! = 1, the sliding limit is presented in the case J! = 0.2.
(c) normal and (d) tangential forces for three values of the applied voltage: V = -2
V, V = 0 V, V = 2 V, the sliding limit is presented in the case V = 2 V
Acknowledgments
The financial supports of CNR and AS! are gratefully acknowledged.
This research was developed within the framework of Lagrange Labora-
tory, an European research group between CNRS, CNR, University of
Rome "Tor Vergata", University of Montpellier II, ENPC and LCPC.
References
Bisegna, P., Lebon, F., and Maceri, F. (2001) D-PANA: a convergent block-relaxation
solution method for the discretized dual formulation of the Signorini-Coulomb
contact problem, C. R. Acad. Sci. Paris Ser. I Math., to appear.
Capuzzo-Dolcetta, I., and Matzeu, M. (1980) Duality for implicit variational problems
and numerical applications, Numer. Punct. Anal. Optim., 2, 231 - 265.
Duvaut, G., and Lions, J. L. (1976) Inequalities in mechanics and physics, Springer-
Verlag, Berlin-New York.
Tiersten, H. F. (1969) Linear piezoelectric plate vibrations, Plenum Press, New York.
Klarbring, A., Mikelic, A., and Shillor, M. (1989) On friction problems with normal
compliance, Nonlinear Anal., 13, 935 - 955.
Lebon, F., and Raous, M. (1992) Friction modelling of a bolted junction under internal
pressure loading, Comput. & Structures, 43, 925 - 933.
Maceri, F., and Bisegna, P. (1998) The unilateral frictionless contact of a piezoelectric
body with a rigid support, Math. Comput. Modelling, 28, 19 - 28.
Panagiotopoulos, P. D. (1975) A nonlinear programming approach to the unilateral
contact and friction boundary value problem in the theory of elasticity, Ing.-Arch.,
44, 421 - 432.
Panagiotopoulos, P. D. (1985) Inequality problems in mechanics and applications.
Convex and nonconvex energy junctions, Birkhuser, Boston.
Panagiotopoulos, P. D., Haslinger, J. (1992) On the dual reciprocal variational ap-
proach to the Signorini-Fichera problem. Convex and nonconvex generalization,
Z. Angew. Math. Mech., 72, 497 - 506.
Telega, J. J. (1988) Topics on unilateral contact problems of elasticity and inelastic-
ity, in Nonsmooth mechanics and applications, edited by J.-J. Moreau and P. D.
Panagiotopoulos, CISM Courses and Lectures 302, Springer-Verlag, Vienna, 341 -
462.
Dostal, Z., Haslinger, J., and Kucera, R. (2001) Implementation of the fixed point
method in contact problems with Coulomb friction based on a dual splitting type
technique, Preprint.
Note. The D-PANA algorithm has been presented for the first time during the
CMIS 2001 Conference. Prof. J. Haslinger, attending the Conference, drew to the
authors' attention a preprint (Dostal et al., (2001)) presenting the same algorithm
named FPMII.
CONTACT BETWEEN 3D BEAMS WITH
RECTANGULAR CROSS-SECTIONS
Przemyslaw Litewka
Institute/or Structural and Computational Mechanics, University Hanover
Appelstr. 9A, 30167 Hanover, Germany
Peter Wriggers
Institute/or Structural and Computational Mechanics, University Hanover
Appelstr. 9A, 30167 Hanover, Germany
1. Introduction
2. Contact search
l'~l~
1 edge to 1 edge I edge to 2 edges
2 edges to 2 edges
not allowed
The first step of contact search is to find the pair of closest elements. The
pair of elements with the shortest distance between fictitious middle points
defmed by the current positions of real element nodes is to be located.
Further on pairs edge-edge or edge-axis are considered. They have initial
position vectors X, and X2 , displacements u, and U2 and current position
vectors: Xj =X j + Uj, i =1, 2. Orthogonality conditions (1) are used to fmd a
pair of closest points on edges, C, and C2, given by co-ordinates rCl and rC2:
x
where =X2 - XI and O,i =dO,i / dri'
Generally (1) represents a set of non-linear equations. To solve it
iteratively by Newton's scheme its linearisation is required:
Vcc .---------
VCA2
With the closest points for each pair of edges located, a penetration check
is done. To this end the following three vectors are defined (see Fig. 2):
• vector between closest points C I and C2: VCC =XCI - XC2,
• vector between C I and its counterpart on beam axis AI: VCAI =XCI - XAh
• vector between C2 and its counterpart on beam axis A 2: VCA2 =XC2 - XA2.
358
Further on the values of angles u, =L(vee , VeAl) and U2 =L(-Vee, VeA2)
are considered. If there is penetration both angles are acute, in the opposite
case they are obtuse, see Fig. 2. Hence, the condition for penetration reads:
(3)
(4)
Expressions for L\gN and 3gNwere derived in Wriggers and Zavarise (1997):
Yn,
L\g N =(L\U C2 - L\uCl
(6)
3g N = (3u C2 - 3UCl Yn,
(,
+ \OU C2 + XC2 0f 2 oU CI - XCI1 0fCl
C2 -
Y(l-n(8)n) . (7)
, , gN
Here the contact contributions to (4) and (5) are expressed in terms of
nodal displacements for a pair of contacting elements UM) and UM2. This is
done by using the representation of edge functions X;, evaluated at the closest
points C) and C2 • In this way the extra terms to be included in the tangent
stiffness matrix and the residual vector for the Newton iterations scheme are
obtained.
Values of L\uc;, oUc;, L\uc;,; and oUc;,; present in expressions (6) and (7)
are determined by the partial differentiation of displacement approximations
Uj with respect to nodal displacements and local co-ordinates rio This yields
the following matrix relations:
(9)
(10)
into three vectors: G;=(G;J, G a , G;3) and the perturbation method yields six
(12xI2)-matrices Gdij,j= 1, 2, 3. They are used to get:
(11)
With matrix relations (8), (9), (10) and (11), variation, linearisation and
linearisation of variation of gap function (6) and (7) can be expressed by
nodal displacements. Their substitution into (4) and (5) yields the contact
contributions to variation and linearisation of variation of strain energy:
Finally the contact contribution to the residual vector has the form:
(13)
o
3
- LnjG djJ
j=J
Three beams, shown in Fig. 3 are analysed. Two of them, crossed at the
angle 15.40 in the plane XY, are of cantilever type with displacements
d3 = At = 10 along Z-axis moving their tips against each other. The third
beam, the "free" one, is moved as a rigid body towards two former ones by
displacements dl = d2 = 100 along X-axis applied at its tips. Displacements
are applied in 80 increments, all beams are divided into 10 elements and
have the same data: E=20000, v =0.3, dimensions 5x5xlOO. Penalty
parameters are: for contact between the crossed beams E= 22500, for contact
between the "free" beam and two other ones E= 1200.
Results of the convergence study for four selected load increments are
presented in Table 1. The energy tolerance is taken as (etol)= 10-8• The
corresponding stages of a deformation process are also shown in Fig. 3.
A very fast convergence is observed throughout the whole process of
deformation. Only 3 to 4 iterations per increment suffice to get eqUilibrium.
The maximum penetration value is about 1% of the size of beam cross-
section. Reaching this level was the objective of penalty parameter tuning.
3D VIEW
T= 1.00
6. Concluding remarks
Acknowledgment
References
1. Introduction
The present paper concerns the numerical treatment of the ther-
moelastic model of contact friction and wear developed in Stromberg
et al. (1996). Here two different approaches are presented and com-
pared. Both are based on the Newton method for non-smooth equations
suggested by Pang (1990). The first idea is to apply the Newton method
directly to the system of coupled thermomechanical equations. This ex-
tends the approach developed in Stromberg (1998) for one thermoelastic
body in frictional wearing contact with a rigid support to the case of
two-body contact. The second idea is to use a strategy where essen-
tially mechanical and thermal subproblems, respectively, are identified
and solved in an iterative process, which from an algorithmic point of
view is a Gauss-Seidel scheme. Also in this case the non-smooth New-
ton method is utilized as solver for the mechanical subproblem, while
the thermal subproblem is simply a system of linear equations.
= 0, (1)
where Yc = (dl,d 2,T 1 ,T2,Pn,Pt). The first two equations are the
discrete equilibrium equations for each body, respectively, where, letting
index l = 1, 2 refer to the two bodies, Klare the stiffness matrices,
KI are the dilatation matrices, d l are vectors of nodal displacements,
TI are vectors of nodal temperatures, FI are external nodal forces, P n
is a vector of the nodal normal contact forces and P t is a vector of
nodal tangential contact forces. The matrices en and C t are kinematic
transformation matrices. The next two equations are the discrete bal-
ance of energy for each body respectively, where MI are the capacity
matrices, 0 1 the conductivity matrices and QI = MITI(tk). Assuming
that the nodal values are ordered such that indices corresponding to con-
tact nodes are put first, the vectors LI, representing the heat generated
by friction and wear and the heat exchange between the two bodies, have
the form LI = (Lf, 0). The entries of L} and L~;' respectively, are given
by
365
(2)
and
(3)
where (x)+ = max(O, x), {}l are the thermal contact conductances, p, is
the coefficient of friction, k is the wear coefficient, {wf} = C t (d1 - d2)
and Wt = wt(tk+d - Wt(tk). Here and in the following, index M refers
to contact nodes. One might also note that the assumption that the
overall contact conductances depend linearly on the contact pressure is
included explicitly in (2) and (3), {}l are treated as constants.
Finally, lIn is the projection equivalent to Signorini's contact con-
ditions with Archard's law of wear inserted and lIt is the projection
equivalent to Coulomb's law of friction. These projections are defined
by
and
3. Algorithms
In this section we present two different strategies for solving H(y) = 0
where H(y) is B-differentiable.
366
(4)
Algorithm GS: Repeat the following steps for each time increment
k + 1:
o Set y~=o = Y2(tk)
1 Solve if I (yi+ l ) = HI (yi+l, y~) = 0 to get yi+l.
2 Solve if2(y~+I) = H2(Yi+l,y~+I) = 0 to get y~+I.
3 Terminate if !H(yi+I)H(yi+l) ~ f. Otherwise replace i by i +1
and return to step 1.
When this strategy is applied, we identify HI and H 2 with the sub-
problems defined by
} =0,
and
4. Numerical example
Consider the two thermoelastic bodies, denoted punch and founda-
tion, shown in figure 1. The potential contact surface is defined by the
lower end of the punch.
368
u(t)
Punch
Foundation
N ~ 7-J-
1.0
0.12 0.12
S 0.1
,, S 0.1
..§, 0.08 /
/
/
S
~ 0.08
3 0.06
3 0.06
0.04 0.04
-10 -5 0 5 10 -10 -5 0 5 10
x [mm] x [mm]
0.14
alum. on steel 0.14
alum. on alum.
0.12 0.12
-----
S 0.1
S 0.1
..§, 0.08 /
/ "-
S
~ 0.08
/
3 0.06
3 0.06 ...... ------
0.04 0.04
-10 -5 0 5 10 -10 -5 0 5 10
X [mm] x [mm]
Figure 3. Wear gap after 100 cycles when thermal effects are included (solid line)
and excluded (dashed line)
References
Ireman, P., Klarbring, A. and Stromberg, N. (2001) Finite element algorithms for
thermoelastic wear problems. Submitted for publication.
Johansson, L. and Klarbring, A. (1993) Thermoelastic frictional contact problems:
modelling, FE-approximation and numerical realization. Compo Meths. in Appl.
Mech. Engng. 105, 181 - 210.
Klarbring, A. (1992) Mathematical programming and augmented Lagrangian methods
for frictional contact problems. Proc. Contact Mechanics Int. Symp., 409 - 422.
Pang, J.-S. (1990) Newton's method for B-differentiable equations. Math. Operations
Research 15, 311 - 34l.
Stromberg, N., Johansson, L. and Klarbring, A. (1996) Derivation and analysis of a
generalized standard model for contact, friction and wear. Int. J. Solids Structures
33, 1817 - 1836.
Stromberg, N. (1997) An augmented Lagrangian method for fretting problems. Eur.
J. Mech., A/Solids 16, 573 - 593.
Stromberg, N. (1998) Finite element treatment of two-dimensional thermoelastic wear
problems. Compo Meths. in Appl. Mech. Engng. 177, 441 - 455.
CONVERGENCE STUDIES FOR 3D SMOOTH
FRICTIONAL CONTACT ELEMENTS BASED ON
THE QUARTIC BEZIER SURFACES
1. Introduction
which tangential slip is split into stick ("elastic") and slip ("plastic") part
(Wriggers (1987}), thus enabling quadratic rate of convergence within the
Newton-Raphson iteration loop. The weak fonnulation and the penalty
method are used to describe the contact contribution to the slave nodes.
The main benefit of smoothing the master surface is that even a relatively
coarse mesh can provide results that are satisfactorily close to the physical
reality. To obtain similar results with the simple node-to-surface elements,
the discretized mesh must have a very high density. Hence, if the discretized
mesh density is increased, a point where there is no difference between the
simple and the smooth node-to-segment element should be reached.
According to this, the simple node-to-surface and the smooth elements are
compared in Chapter 3.
Figure 1. The tetrahedral element and the smooth contact surface consisting of the six Bezier
surfaces
The symbol!!. denotes the summation over all i,j,k ~ 0 for i + j +k =4.
Due to this particular definition of quartic Bezier surfaces (Figure 1,,), the
C1-continuity between the adjacent surfaces is always satisfied.
1 "
O,
Figure 2. Definition of the B6zier points needed for the definition of one quartic B6zier
surface
(2)
Furthermore the contact check, i.e. the detection if the slave node is
inside or outside the master body, is performed. In the case of contact, the
normal gap (penetration) is calculated according to the relation:
The direction of sliding is defined by the vector connecting the previous, "x,
and the current, ,,+l x{'l+la ;+Ij3), solution points, projected into the tangential
375
plane of the current solution, as defmed in equation (61). The position ofthe
second and the third Bezier point is chosen to be third of Ildll multiplied with
the unit tangent vectors. Due to the explicit form and the convex hull
property (for more details see Farin (1993», a cubic Bezier curve is chosen
for the definition of the path. The path length s is evaluated as the arc length
of the Bezier curve from the integral (611).
n+le
T ~T
(6)
As the integral in (611) does not have an explicit analytical solution, Gauss
integration is used for the calculation of the arc length s.
The increase in the total tangential slip vector is defmed by equation
(6111). With equations (6) all relations needed for the formulation of the
residual vector and the tangent matrix are defmed. More about the
formulation of residual vector and tangent matrix can be found in Krstulovic
et al. (2000) or Krstulovic (2001).
I O.
I :: Smp_~_1 I
i Smooth frlet_
t 0.3
.. 02
L. _______ .~...
__fd,IP-I"'--nt)------'--------T-
.... -f-__r--_ _ _ _ ~
...
0.8
17970DOF
1
100llepo
0
•
_ 0.5
1
I
0• - - Smp_fr_nal
.. ..... S..-h frletlOnal
! 03
~ 0.2
"'~'8J
Figure 6. Total tangential reactions for 800 load steps and various mesh densities
For the tangential reaction, the L z norms of the difference between the
reactions of simple (FTRU) and smooth (FSTR) elements are displayed in
Figure 7. The L z norm (Le., the square root error) of the difference in
reactions for simple and smooth elements is calculated from the relation (7,).
From Figure 7 it can be observed that, by increasing the mesh density, the
error of approximately 2.4% corresponds to the mesh in Figure 5". That is a
discretization for which the smooth and the simple approaches give
approximately the same results. The difference in total reaction for various
mesh densities appears to be due to the different overlapping caused by FE
discretization. Therefore, results on the discretization error for systems with
a smaller number of degrees of freedom show the advantage of the smooth
formulation in comparison with the simple one.
From the presented error diagram it can be concluded that the smooth
mesh of 17970 DOF (Figure 5,,) gives the best result. Comparison with other
meshes should show the convergence of the solution when increasing the
number of DOF. For the case of 500 load steps, Figure 8 depicts the Lz norm
of the difference in reactions according to the Lz norm of reaction for the
smooth 17970 DOF mesh. This error is calculated from relation (7,,).
378
1=1 1=1
1 03
0.275
0.25
--8----
----<7-~-
--B---
552 OOF
289200F
17970 OOF
l ------
2.5
~~~·---·~~----l,
I
I
I
1";-
r0.2
1=0.175
~0.225
~--------;>,--------"
I
d 0.15
e: 0.125
11f~
I ~ 01
It-0075 ,;r--+------~-----0
1=000:: ~ G-----B------e--~
___
L_~ ~~_~O:_::ioa~~p~OO 600 700 800
References
Agelet de Saracibar, C. (1997) A New Frictional Time Integration Algorithm for Large Slip
Multi-Body Frictional Contact Problems, Computer Methods In Applied Mechanics And
Engineering, 142, 303-334.
Bohm, W., Farin, G., Kahmann, 1. (1984) A survey of curve and surface methods in CAGD,
Computer Aided Geometric Design, 1, 1-60.
Farin, G. (1993) Curves and Surfaces for Computer Aided Geometric Design. A Practical
Guide, third edition, Academic Press, Boston.
Krstulovic-Opara, L., Wriggers, P., and KoreJc, 1. (2000) Symbolically generated 3D smooth
polynomial frictional contact element based on the quartic B6zier surfaces, In Proceedings
of ECCOMAS 2000, Barcelona, Spain.
Krstulovi6-Opara, L. (2001) A C1-continuous formulation for finite deformation contact,
Ph.D. Thesis, FOIlI, IBNM, Hannover.
Wriggers, P. (1987) On Consistent Tangent Matrices for Frictional Problems, In Proceedings
ofNUMETA 87 Conference, Eds. 1. Middleton, G. N. Pande, Nijhoff, Drobrecht.
Wriggers, P., Krstulovi6-0para, L., and KoreJc, 1. (1999) Development of 2D Smooth
Polynomial Frictional Contact Element based on a Symbolic Approach, In Proceedings of
ECCM, Wunderlich (ed.), Miinchen.
MULTICONTACT STRUCTURES AND
PARALLEL COMPUTING
P. Alart
Laboratoire de Mecanique et Genie Civil UMR 5508, CNRS-UniversiU Montpellier 2
Pl. E. Bataillon, 34 095 Montpellier Cedex 5, France
M. Barboteu
Laboratoire de Theorie des Systemes, UniversiU de Perpignan
52, avenue de Villeneuve, 66860 Perpignan Cedex, France
Abstract In this paper we present a numerical strategy to solve large scale fric-
tional contact problems by using balancing domain decomposition meth-
ods which are well adapted to parallel computing (scalability property).
The motivation is given by the study of the mechanical behaviour of
three multicontact structures (rolling shutters, collections of deformable
grains and wood micro cellular media).
1. Introduction
The aim of this paper is to present an efficient numerical scheme for
a multibody system with frictional contact between the bodies or sub-
structures, we call these multicontact structures. We focus our attention
on three applications:
- rolling shutters composed of many slats jointed by a hinge with
clearance and eventually rotative friction;
- collections of deformable grains with contact interfaces between
them;
- wood microstructure composed of a cell wall network with self
contact between the different walls of the cells.
To handle these multicontact problems it is necessary to develop rigorous
modelling and new numerical tools adapted to the strong non-linearity
and non symmetry due to the large proportion of degrees of freedom
related to the contact conditions. To overcome these difficulties, we pro-
pose in Section 3 a domain decomposition method generalized to non
symmetric systems. The balancing preconditioners have to be improved
379
J.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 379-386.
@ 2002 Kluwer Academic Publishers.
380
{
Fint - Fext+ F(u, A) = 0 (1 )
-~(A - F(u, A)) = 0
We let u denote the kinematic variables (displacements or rotations) and
A the static variables (contact forces or torques). Fint and Fext denote,
respectively, the internal and the external discretized forces. F(u, A)
defines the discretized contact operator, and r is the penalty coefficient.
We have chosen to treat both variables u and A simultaneously through
Newton's method. The system of equations is split into two parts involv-
ing the pair x = (u, A): a differentiable part G and a non differentiable
frictional contact F,
G(x) + F(x) = O. (2)
To overcome the non differentiability in equation (2), Newton's method
may be extended to the following iterative form (G.N .M.):
xi+! = xi _ (Ki + A~) -1 (G(xi) + F(x i )) , (3)
Ki = 8G(xi), A~ E 8F(x i ).
The matrix Ki is the usual tangent matrix and A~ represents the gener-
alized Jacobian of F at xi. Usually, xi is in a region of linearity, and A~
reduces to a single classical Jacobian matrix. Each region of differentia-
bility of the operator corresponds to a status of each contact node of the
discretization. The G.N .M. leads us to solve at each iteration i a non
symmetric linear system. The non symmetry of the matrix A;:" is due to
the friction terms. The contact interface is discretized by contact finite
elements which yield elementary non symmetric tangent matrices when
the friction appears (slip status). To solve the non symmetric linearized
systems, we choose the GMRes method as an iterative non symmetric
domain decomposition method (Schur complement form).
Here, the space Va contains the elements of the kernel of S; (and not
Si as for the symmetric case !). The preconditioning step consists in
multiplying the preconditioning matrix M- 1 by the residual vector r.
This product, ¢, can be decomposed into the sum of ¢f E Vf = L~l ~.l
382
The terms 4>t and 4>'G are given by the solutions of local Neumann prob-
lems and a global coarse problem which take the following variational
forms:
4. M ulticontact applications
The efficiency of this domain decomposition strategy will be assessed
in three examples of multicontact structures. Some characteristics of
these examples are summarized in the following table.
383
"floating" subdomain
with a hinge (contact interface)
~~e+bodYI
body 2
F
~
~
[ Non symmetric 2- le.el Newnann- Neumann
I- level Newnann-Neumann
I ""f)~ •~ 1::
ifau ~ 200
~ ~~. °1 00
50'-:'-:-~-""""''''''''''''--'- o~~~~~~~~
0.0 0.5 1.0 1.5 2.0 10 20 30 40 50 60 70
friclion coefficient Number of subdomains (II = 0.2)
• elastic node
one subdomain
Collection of defonnable grains
of the walls and self contact inside the cells between the walls. The simu-
lation of such a process requires accounting for the large deformations of
the walls and eventual large slips. The connectivity of the medium leads
to identifying a subdomain as a cell, with half walls, or a set of cells.
In the first approach the microstructure is fully periodic. Using only a
5. Conclusions
The numerical tool developed above may be used for cellular media,
such as wood, in order to understand the densification process of wood,
to determine the region with micro-buckling in an indentation test of the
386
0.8
'"....
Q)
<) 0.6
<E
s:: 0.4
.2
t)
..s
0
..... 0.2 linear behaviour
0.0
0 5 10 15 20 25 30 35
compression ratio
References
Alart P. and Curnier A. (1991) A mixed formulation for frictional contact problems
prone to Newton like solution methods, Comp o Meth . Appl. Mech . Eng, 92, 353-
375.
Alart P. , Barboteu M., Le Tallec P. and Vidrascu M. (2001) Additive Schwarz method
for non symmetric problems : application to frictional multicontact problems, to
appear in the proceeding 's of the 13 t h International Conference on Domain De-
composition Method (plenary lecture).
Barboteu M., Alart P. and Vidrascu M. (2001) A domain decomposition strategy
for non classical frictional multicontact problems, Compo Meth. Appl. Mech. Eng,
190/37-38, 4785- 4803.
Jourdan F., Alart P. and Jean M. (1998) A Gauss-Seidel-like algorithm to solve fric-
tional contact problems, Compo Meth . Appl. Mech. Eng, 155, 31-47.
Le Tallec P. (1994) Domain decomposition methods in computational mechanics,
Compo Mech. Adv., 1, 21-220.
Moreau M. and Jean M. (1996) Numerical treatment of contact and friction: the
contact dynamics method, Eng. Sys. Des. Anal., 76, 201-207.
A NEW APPROACH TO THE
2D TRANSIENT ROLLING
CONTACT PROBLEM
Abstract This work presents a new approach to the transient rolling contact of
two-dimensional elastic bodies. Solutions will be obtained by mini-
mizing a general B-differentiable function representing the equilibrium
equations and the contact conditions at each time step. Inertial effects
are not taken into account and the Boundary Element Method is used to
compute the elastic influence coefficients of the surface points involved
in contact.
1. Introduction
The mechanical study of rolling contact problems started in 1875 with
Reynold's investigation of the creep phenomena, observed when a rubber
cylinder rolling on a steel plate moves forward more than its undeformed
rolled perimeter. He attributed this effect to the elastic deformations in
the contact area and introduced the idea that the contact region would
be composed of stick and slip zones. It was about fifty years later in 1926,
when Carter published his work presenting the steady-state analytical
solution for the two dimensional case of similar material cylinders, and
the non-linear relation between longitudinal creep force and creepage.
Historically, the first approximation to the transient rolling contact
problem was reported by Kalker in 1970 for the particular no-slip case;
the formulation was extended in 1971 to the complete adhesion-slip case
using an original minimum principle. In both works, a constant Hertz
normal pressure distribution over time was assumed, and uncoupled
normal-tangential problems were solved.
387
J.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 387-394.
© 2002 Kluwer Academic Publishers.
388
Driven
Roller 2a SOLIDB
Pn ~ 0 (1)
(4)
where V(.) jVr is the Total time derivative, sometimes called Substantial
or Material derivative (following a particle of fixed identity relative to
fixed axes) and 8t represents the tangential separation, given by
8t (x, r) = {xA(r) - xB(r)} + {uf(x, r) + uf(x, r)} . (5)
We assume quasi-flat contact surfaces, and that the mean rolling ve-
locity V(r) governs the rotation of the two cylinders. Under these hy-
potheses the individual velocity Va(r) of each cylinder can be substi-
tuted in the equations by the mean velocity V (r).
The slip velocity is obtained from equations (5) and (4), based on the
Total time derivative
(6)
and presents the advantage that the formulation for each time step only
involves tangential displacements at the current and the previous time
step and the drawback that we have to interpolate them between nodes
with the shape functions, if tlTJ = V(r)tlr is not equal to the nodal
separation.
3. Problem formulation
The step to step time integration scheme is adopted, forcing the ful-
fillment of contact constraints at all nodes for every discrete time step.
390
The suggested algorithm is based on Mathematical Programming tools
and solves the problem in a combined manner, without separating it into
normal and tangential parts, and with a fully automated iteration con-
trol.
where nts is the total number of steps, and p~,t = p~,{ (TJk) are the
surface nodal tractions for discrete k-th step.
Similarly, the tangential displacement difference can be obtained in
condensed form as
u tk = gtk + StnP nk + S ttPtk (8)
After the resolution of the elastic problem using the BEM and later
condensation, we have obtained the linear relation between gap-tangential
displacements vector (o~, u~) and normal-tangential tractions vector in
the contact area (p~, p~), expressed as
The equilibrium equation (9) together with the contact conditions ex-
pressed by equations (1) and (2), for all contact point pairs, completes
the mathematical description of the problem. To calculate the slip ve-
locities at the contact nodes, the Material formulation based on (6) is
adopted.
(11)
Introducing (8) into the slip velocity main expression (11) and con-
sidering (9), we finally obtain the grouped equation for the Material
approach
1 [_<lk-l
k t:k
St = '" + tl17 'U + gtk + S tnPnk + S ttPtk] (12)
= 0,
(13)
which is a B-differentiable function, because of the non F-differentiable
contact conditions, that involve non-linear directional derivatives.
To solve the problem the Generalised Newton's Method with line
search (GNMIs) has been used. GNMls is an extension of Newton's
Method for B-differentiable functions formulated by Pang in 1990.
4. Results
The example problems presented are for cylinders with the same ma-
terial properties. The BE mesh is represented in Figure 2. The first
transient rolling contact problem consists of two rotating cylinders of
identical material. This problem does not present normal-tangential
coupling under the half-space hypothesis, and Hertz normal pressure
solution is exact for normal tractions.
The friction coefficient between surfaces is J1 = 0.1 and two cases
are considered, both starting from a static contact situation but with
different applied forces.
i"
~
a.- 0.4
...... ..... T_
..... -O-Slip
5. Conclusions
The transient two-dimensional rolling contact problem between cylin-
ders can be solved using a new numerical technology based on the use
of Boundary Element Method and Mathematical Programming Tech-
niques.
In this new technology BEM has been used to compute the constant
influence coefficients based on true cylinder geometry, not on the half-
space approximation, while MPT was used to model the contact condi-
tions and to solve the highly non-linear problem associated with tran-
sient rolling at each time step. The algorithm and proposed formulations
prove to be very strong and efficient for 2D, similar transient problems.
References
F. W. Carter. On the action of a locomotive driving wheel. Proc. Roy. Soc. Ser. A.,
112:151157, 1926.
J. A. Gonzalez and R. Abascal. An algorithm to solve coupled 2d rolling contact
problems. Int. J. Num. Meth. Engng, 49:1143-1167, 2000.
J. J. Kalker. Transient phenomena in two elastic cylinders rolling over each other with
dry friction. Jour. Appl. Mech., 37:677-688, 1970.
J. J. Kalker. A minimun principle for the law of dry friction, with applications to
elastic cylinders in rolling contact. part 2: Application to nonsteadily rolling elastic
cylinders. Jour. Appl. Mech., 38:881-887, 1971.
J. J. Kalker. Three-dimensional elastic bodies in rolling contact. Kluwer Academic
Press, Dordrecht, 1990.
J. S. Pang. Newton's method for b-differentiable equations. Math. Oper. Research,
15(2):311-341, 1990.
O. Reynolds. On rolling friction. Philosophical Transactions of the Royal Society of
London, 1:155 174,1876.
NUMERICAL APPROXIMATION OF THE
ELASTIC-VISCOPLASTIC CONTACT
PROBLEM USING NONCOINCIDING
FINITE ELEMENT MESHES
Patrick Hild
Laboratoire de Mathimatiques. University of Savoie
73376 Le Bourget du Lac, France
1. Introduction
The object of this paper is to study the convergence of finite element me-
thods for elastic-viscoplastic models submitted to frictionless unilateral con-
tact conditions. Considering the contact problem between two bodies, we are
particularly ~nterested in finite element approximations involving nonmatching
meshes on the contact part in which the nodes inherited from the triangulations
of the bodies may not coincide.
By adapting and extending the techniques developed for the linear elastic
case with nonmatching meshes in Ben Belgacem-Hild-Laborde (1999) and for
the elastic-viscoplastic case with matching meshes in Han and Sofonea (2000),
we are able to study the convergence of different finite element approaches
handling nonmatching meshes: a first method where the discrete contact con-
ditions are of local type and another one where global contact conditions hold.
395
l.A. C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 395-402.
© 2002 Kluwer Academic Publishers.
396
The Hilbert spaces V(Oi), Q(Oi) and Q1 (Oi) are endowed with their canoni-
cal inner products denoted (., ·)V(nl), (., ·)Q(nl) and (., ·)Ql(nl). Setting V =
V(01) x V(02), Q = Q(01) x Q(02) and Q1 = Q1(01) x Q1(02), these
product spaces are equipped with the inner products (., ·)v, (., .)Q and (., ·)Ql
and the associated norms are denoted I· Iv , I·IQ and 1·IQl respectively. Let 82
be the space of second order symmetric tensors on ~2 and let· and I . I denote
the inner product and the norm on this space.
Next, we specify the properties satisfied by the functions t;i and C i (£ =
1, 2) which will be incorporated in the viscoplastic constitutive law.
The function t;i : Oi x 82 -+ 82 is a bounded symmetric positive definite
fourth order tensor
{
t;fjkl E Loo(Oi), 1 ~ i,j, k, 1 ~ 2,
£iui . Ti = ui· £iTi vui, Ti E 8 2 ,
°
(1)
3 a i > s.t. t;iTi . Ti ~ a i lTil2 VT i E 82 .
397
l _ 1 (8vf 8v1)
e(v ) - 2" 8xj + 8Xi '
e(v) = (e(v 1),e(v 2)) T/v = (V 1,V2) E V,
ce = (c 1e 1 , c 2e2) T/E: = (e 1, e2) E Q,
G(O",e) = (G 1(0"1,e 1),G2(0"2,e 2)) T/e, 0" E Q.
For all t E [0, T], let F(t) represent the element of Vi given by
2 2
(F(t),v) = L:(f"(t),v")[L2(nt )]2 + L:(gl(t),vl)[L2(r~)J2 T/v E V.
~1 ~1
Suppose that
I" E W 1,OO(0, Tj [£2(0")]2), y! E W 1,OO(0, Tj [£2(r1v )]2)j (3)
Assume that
1.£0 = (uA, u~) E U, (0"0, e(uo))Q = (F(O),uo). (5)
The variational formulation of the elastic-viscoplastic unilateral contact pro-
blem (see Rochdi and Sofonea (1997» consists then of finding the displace-
ment fields 1.£ : [0, T] -+ U and the stress fields 0" : [0, T] -+ Q1 satisfying:
1.£(0) = 1.£0, 0"(0) = 0"0,
u(t)=ce(u(t))+G(O"(t),e(u(t))) a.e. tE(O,T), (6)
(O"(t),e(v) - e(u(t)))Q ~ (F(t),v- u(t)) T/v E U, t E [O,T].
The existence and uniqueness statement for this problem has been proved
by Rochdi and Sofonea (1997). We recall this result in the following theorem.
Theorem 1 Let the assumptions (1), (2), (3) and (5) hold. Then there exists a
unique solution ojproblem (6) such that
(1.£,0") E W1,OO(0,TjU x Q1).
398
Note that both approximations are different and nonconforming so that gene-
rally U'h i= ul, U'h rt rt
U and Ul u. This means that (slight) interpenetra-
tions are allowed in these approximations.
Now, we consider a fully discrete approximation of problem (6). Given a
partition of the time interval [0, T]: 0= to < tl < ... < tN = T, we denote
the step size by k n = tn - tn-I, for n = 1,2, ... , N. Let k = maxn k n be
the time discretization parameter. For a sequence {w n };;=o, we define 6w n =
(w n - wn- 1)jkn. Finally, we denote Fn = F(t n ), un = u(tn),u1,n =
u1(tn), u 2,n = u 2(t n ), un = u(tn ), u~ = uv(tn) for n = 0, 1, ... ,N.
The discretized problem issued from (6) uses the above-mentioned noncon-
forming finite element approaches and a backward Euler scheme. It consists
then of finding the displacement fields Uhk = {Uhk};;=O c Uh and the stress
fields Uhk = {uhd;;=o c Qh such that ug k E Uh, ugk E Qh and for
n = 1,2, ... ,N,
Using fixed point arguments, the existence and uniqueness of solution to pro-
blem (7) is proved for a time parameter k small enough. In order to obtain
error estimates for the discrete solutions, we first establish the following result.
max {Iu n -
l::;n::;N
u~klQ + lu n - u~klv} :::; C(luo - ugklQ + 11.1.0 - u~klv
+ max {
l::;n::;N
inf (IVh - unlv
'lJhEUh
+\(
Jre
U~[(Vh - un). V]dI'\t)
+ inf I (
Jre u~[(V - Uhk)' V]dI'/t
t
'lJEU
1ire
( u~[(v - u~k) . v] drl t ~ (C hflun - u~klv ) t + C(hl + h2),
where constant C depend,; linearly on lu1,nl[H2(f21))2 and lu 2,nl[H2(f22)]2 and
p, = 1, r = 1 ijUh = UK and p, = ~, r = ~ ijUh = Uf
We finally obtain the convergence result of the fully discrete approximation.
where Ai is a viscosity constant and PBl is the projection operator (with respect
to the norm l'T i l2 = ('Ti, £i'Ti)Q(f2l)) on the plasticity convex set Bi C 8 2
401
defined by:
where O"~ represents the uniaxial yield stress. Moreover, both plane stress
elasticity tensors [l (f = 1,2) are given by:
where El and y;,l denote Young's modulus and Poisson's ratio of the material
that occupies the domain nl.
In order to compare results obtained via (local) approximation set Ul and
(global) approximation set U'h, we implement both methods for the simple
problem shown in Figure 1 (see Fernandez Garcia-Hild-Viafio (2000» invol-
ving two viscoplastic bodies (El = 3 X 105 M Pa, y;,l = 0.3, )..l = 100 N .
sec/m2 , O"~ = yTON/m 2, f = 1,2). A density of forces gl(Xl,x2,t) =
(0, -lOt) N/m linearly increasing in time is applied on the upper boundary
of n1 whereas body forces are absent. Also, we consider T = 1 sec, 0'0 =
ON/m 2 anduo = Om.
:r.-'
N
:r::Z
o
Q-------Q------Q------~------Q--------------
Figure 1. Uk: Initial and deformed boundaries at final time T = 1 sec. and contact area.
Figure 1 shows also the deformed boundaries and the deformed meshes
when U'h is assumed. We remark that no interpenetration has been produced
and the obtained stress field is constant (0"11 = 0"12 = 0, 0"22 (Xl, X2, t) = lOt).
In the case where the local approach is chosen (Uh = Ul), the deformed
meshes near the contact area are shown in Figure 2. We now discover a
non negligible and non realistic penetration of n2 into n1 as well as artificial
stresses, particularly near the contact zone.
402
:aD _ 23.05
i. _ 21.15
18 _ 20.51
17 _ 11.80
l' _ 19.03
15 _ 11.25
1& _ 17.&'
13 _ lli.l1
12 _ 15.9'
11 _ 15.16
10 _ 1.... "
,_ 13.11
• _ 12 ....
7 _ 12.0'7
5 _ 11.30
5 _ 10.52
1.750
8.176
•• 203
1 _ 7.&3D
Figure 2. U[: Contact area and stress field at final time T = 1 sec.
References
Ben Belgacem E (1999) The mortar finite element method with Lagrange multipliers, Numer.
Math. 84,173-197.
Ben Belgacem E, Hild P. and Laborde P. (1999) Extension of the mortar finite element method
to a variational inequality modeling unilateral contact, in Math. Mod. Meth. Appl. Sci. 9 (2),
287-303.
Fernandez Garcia, 1.R., Hild, P. and Viano, 1.M. (2000), Resolution numenque d'un probleme
de contact entre corps elasto-viscoplastiques et maillages elements finis incompatibles, C. R.
Acad. Sci. Paris, Serie 1331, 833-838.
Fernandez Garcia, 1.R., Hild, P. and Viano, lM. (2001), Numerical approximation ofthe elastic-
viscoplastic contact problem with non-matching meshes, Prepublication LAMA, Ol-Ola
Submitted.
Han W. and Sofonea M. (2000) Numerical analysis of a frictionless contact problem for elastic-
viscoplastic materials, Comput. Methods Appl. Mech. Engrg. 190, 179-191.
Ionescu, I. R. and Sofonea, M. (1993) Functional and Numerical Methods in Viscoplasticity,
Oxford University Press, Oxford.
Rochdi M. and Sofonea M. (1997) On frictionless contact between two elastic-viscoplastic bod-
ies, Quart. J. Mech. Appl. Math. SO, 481-496.
CONTACT IN THE ARLEQUIN FRAME-
WORK
Malek Zarroug
Ecole Centrale de Paris
Laboratoire MSS-Mat, CNRS UMR 8579, lirance
1. Introduction
There is an important range of mechanical contact problems for which
the choice of appropriate refined numerical discrete schemes is an im-
portant issue. Among other applications, let us mention the case of
contact edges and more generally problems for which the curvature of
the candidate contact surface is sufficiently large, at least locally. A
soil/foundation interaction problem is a typical example_ Another one
is given by the classical contact test used to measure a material tough-
ness. The multiscale character of contact problems and impact are other
sources of motivation. There is clearly for such kind of problems a need
of a (local at least) numerical refinement in order to capture mechanical
quantities of interest with sufficient accuracy. Classical h-adaptive ap-
proaches (cf. ego Johnson (1991) and Wriggers et al (1998)) can of course
be used to address these contact problems. To ensure the improvement
of the accuracy most of these methods modify the approximation space
by adding independent vectors to enrich the coarse base where needed.
403
l.A.C. Martins and M.D.P. Monteiro Marques (eds.), Contact Mechanics, 403-410.
© 2002 Kluwer Academic Publishers.
404
But let us quote here that this independence requirement constitutes a
lack of "flexibility" of these classical adaptive methods.
{ -ct:.u = f in 0
(1)
u = 0 onaO
where c is a positive real coefficient and where 0 is a regular bounded
domain of 'R2. The weak "mono-model" formulation of this problem is
the following (well-posed one):
with
Let G denote the function describing the obstacle surface and let A
be the contact pressure on the potential contact surface Oc (figure 1).
The local system taking into account the Signorini unilateral contact
5. Solution strategy
The problem (9) is a nonlinear one. This is due to the projection
application. A numerical strategy is needed to solve it. The one we use
consists of introducing the following status function:
X(g) = 1 if 9 ~ 0
{ = 0 otherwise (10)
9 = A - pd(U C ) in !lc
Find (Uhl UX, Fhl Ah) in W h such that for all (Vhl vhl Fit, Ah) in
W hl
[ o
o K2 +0 Ak (Bkf
K1 0 (O )T
(02)T
l
1(U~) = ( L1 )
U~ L~
Bk Fk 0 Ak Lk (15)
h 3
0 1 0 2 0 0 Fk 0
where the matrices and the right hand side terms of (15) can be evalu-
ated in a classical manner and where k refers to the iteration.
(16)
where r~, Ah, v~ and Al stand for the "slave" surface, the contact pres-
sure, the virtual displacement-like field, defined on r~ (the "master"
surface) and the classical target application, respectively. A promising
409
idea consists of distributing this integral as follows:
t,.,.,.v ...• u,
, ,,
,
, '
,'
: :
: :
, ,
: :
, ,
, ,
~~,wh~~),;}/.i'##~/.i'.d7ld
Contact pressure
- - 6 - Coarse model ,
i
if
JI
Ii
-------.-- -------......
__ -------:-+-->-.+':::<.!!.~(+
__ __ __ x
Contact pressures obtained by both the coarse model (figure 2-a) and
the Arlequin one (figure 2-b) are plotted in figure 3 where one can notice
the improvement of the accuracy of the result by Arlequin approach.
References
Ben Dhia, H. (1998) Multiscale mechanical problems: The Arlequin method, C. R.
Acad. Sci. Paris, Ser. II, 326, 899 - 904.
Ben Dhia, H. (1999) Numerical modelling of multiscale mechanical problems: the Ar-
lequin method, ECCM'99 CD-ROM, Miinchen, Germany, August 31 - September
3.
Ben Dhia, H., Vautier, L, and Zarroug, M. (2000) Problemes de contact frottant en
grandes transformations: du continu au discret, Revue Europeenne des Elements
Finis, 9, 243 - 26l.
Ben Dhia, H., Rateau, G., (2001) Mathematical analysis of the Arlequin method,
C.R. A cad. Sci. Paris, Ser. I, 332, 649 - 654.
Ben Dhia, H., Zarroug, M., (2001) Mixed frictional contact elements and local "volu-
mic" contact interfaces, ECCM'01 CD-ROM, Cracow, Poland, June 26- 29.
Brezzi, F., Hager, W., and Raviart, P.A., (1978) Error estimates for the finite element
solution of variational inequalities, Numer. Math., 31, 1 -16.
Hallquist, J.O. et al (1985) Sliding interfaces with contact-impact in large-scale La-
grangian computations, Comput. Meth. Appl. Mech. Engng., 51, 107 -137.
Haslinger, J., Hlavacek, 1., and Necas, J., (1991) Numerical methods for unilateral
problems in solid mechanics, In Handbook of Numerical Analysis, IV Part 2, Eds.
P.G. Ciarlet and J.-L. Lions, North Holland, 313 - 485.
Johnson, C., (1991) Adaptative finite element methods for the obstacle problem,
Technical Report, Chalmers University of Technology.
Wriggers, P., Scherif, 0., (1998), Different a posteriori error estimators and indicators
for contact problems, Mathl. Comput. Modelling, 28,4-8, 437 - 447.
Appendix A
CMIS 2001
June 17 - 21,2001
Praia da Consolaryoo, Peniche, PORTUGAL
411
413
~onda~June18,2001
CMIS 2001
June 17-21, 2001
Praia da Conso/ac;:iio, Peniche, PORTUGAL
419
421
1. R.T. Haftka, Z. Giirdal and M.P. Kamat: Elements ofStructural Optimization. 2nd rev.ed., 1990
ISBN 0-7923-0608-2
2. U. Kalker: Three-Dimensional Elastic Bodies in Rolling Contact. 1990 ISBN 0-7923-0712-7
3. P. Karasudhi: Foundations of Solid Mechanics. 1991 ISBN 0-7923-0772-0
4. Not published
5. Not published.
6. IF. Doyle: Static and Dynamic Analysis of Structures. With an Emphasis on Mechanics and
Computer Matrix Methods. 1991 ISBN 0-7923-1124-8; Pb 0-7923-1208-2
7. 0.0. Ochoa and J.N. Reddy: Finite Element Analysis of Composite Laminates.
ISBN 0-7923-1125-6
8. M.H. Aliabadi and D.P. Rooke: Numerical Fracture Mechanics. ISBN 0-7923-1175-2
9. 1 Angeles and e.S. Lopez-Cajun: Optimization of Cam Mechanisms. 1991
ISBN 0-7923': 1355-0
10. D.E. Grierson, A. Franchi and P. Riva (eds.): Progress in Structural Engineering. 1991
ISBN 0-7923-1396-8
11. R.T. Haftka and Z. Giirdal: Elements of Structural Optimization. 3rd rev. and expo ed. 1992
ISBN 0-7923-1504-9; Pb 0-7923-1505-7
12. J.R. Barber: Elasticity. 1992 ISBN 0-7923-1609-6; Pb 0-7923-161O-X
13. H.S. Tzou and G.L. Anderson (eds.): Intelligent Structural Systems. 1992
ISBN 0-7923-1920-6
14. E.E. Gdoutos: Fracture Mechanics. An Introduction. 1993 ISBN 0-7923-1932-X
15. J.P. Ward: Solid Mechanics. An Introduction. 1992 ISBN 0-7923-1949-4
16. M. Farshad: Design and Analysis of Shell Structures. 1992 ISBN 0-7923-1950-8
17. H.S. Tzou and T. Fukuda (eds.): Precision Sensors, Actuators and Systems. 1992
ISBN 0-7923-2015-8
18. lR. Vinson: The Behavior of Shells Composed of Isotropic and Composite Materials. 1993
ISBN 0-7923-2113-8
19. H.S. Tzou: Piezoelectric Shells. Distributed Sensing and Control of Continua. 1993
ISBN 0-7923-2186-3
20. W. Schiehlen (ed.): Advanced Multibody System Dynamics. Simulation and Software Tools.
1993 ISBN 0-7923-2192-8
21. C.-W. Lee: Vibration Analysis of Rotors. 1993 ISBN 0-7923-2300-9
22. D.R. Smith: An Introduction to Continuum Mechanics. 1993 ISBN 0-7923-2454-4
23. G.M.L. Gladwell: Inverse Problems in Scattering. An Introduction. 1993 ISBN 0-7923-2478-1
Mechanics
SOLID MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
24. G. Prathap: The Finite Element Method in Structural Mechanics. 1993 ISBN 0-7923-2492-7
25. 1 Herskovits (ed.): Advances in Structural Optimization. 1995 ISBN 0-7923-2510-9
26. M.A. Gonzalez-Palacios and J. Angeles: Cam Synthesis. 1993 ISBN 0-7923-2536-2
27. W.S. Hall: The Boundary Element Method. 1993 ISBN 0-7923-2580-X
28. 1 Angeles, G. Hommel and P. Kovacs (eds.): Computational Kinematics. 1993
ISBN 0-7923-2585-0
29. A. Curnier: Computational Methods in Solid Mechanics. 1994 ISBN 0-7923-2761-6
30. D.A. Hills and D. Nowell: Mechanics of Fretting Fatigue. 1994 ISBN 0-7923-2866-3
31. B. Tabarrok and EP.I. Rimrott: Variational Methods and Complementary Formulations in
Dynamics. 1994 ISBN 0-7923-2923-6
32. E.H. Dowell (ed.), E.E Crawley, H.C. Curtiss Jr., D.A. Peters, R. H. Scanlan and E Sisto: A
Modem Course in Aeroelasticity. Third Revised and Enlarged Edition. 1995
ISBN 0-7923-2788-8; Pb: 0-7923-2789-6
33. A. Preumont: Random Vibration and Spectral Analysis. 1994 ISBN 0-7923-3036-6
34. J.N. Reddy (ed.): Mechanics of Composite Materials. Selected works of Nicholas 1 Pagano.
1994 ISBN 0-7923-3041-2
35. A.P.S. Selvadurai (ed.): Mechanics of Poroelastic Media. 1996 ISBN 0-7923-3329-2
36. Z. Mr6z, D. Weichert, S. Dorosz (eds.): Inelastic Behaviour of Structures under Variable
Loads. 1995 ISBN 0-7923-3397-7
37. R. Pyrz (ed.): IUTAM Symposium on Microstructure-Property Interactions in Composite
Materials. Proceedings of the IUTAM Symposium held in Aalborg, Denmark. 1995
ISBN 0-7923-3427-2
38. M.I. Friswell and lE. Mottershead: Finite Element Model Updating in Structural Dynamics.
1995 ISBN 0-7923-3431-0
39. D.E Parker and A.H. England (eds.): IUTAM Symposium on Anisotropy, Inhomogeneity and
Nonlinearity in Solid Mechanics. Proceedings of the IUTAM Symposium held in Nottingham,
U.K. 1995 ISBN 0-7923-3594-5
40. J.-P. MerIet and B. Ravani (eds.): Computational Kinematics '95. 1995 ISBN 0-7923-3673-9
41. L.P. Lebedev, l.l. Vorovich and G.M.L. Gladwell: Functional Analysis. Applications in Mech-
anics and Inverse Problems. 1996 ISBN 0-7923-3849-9
42. J. Mencik: Mechanics of Components with Treated or Coated Surfaces. 1996
ISBN 0-7923-3700-X
43. D. Bestle and W. SchieWen (eds.): IUTAM Symposium on Optimization ofMechanical Systems.
Proceedings of the IUTAM Symposium held in Stuttgart, Germany. 1996
ISBN 0-7923-3830-8
44. D.A. Hills, P.A. Kelly, D.N. Dai and A.M. Korsunsky: Solution of Crack Problems. The
Distributed Dislocation Technique. 1996 ISBN 0-7923-3848-0
45. V.A. Squire, R.I. Hosking, A.D. Kerr and P.I. Langhorne: Moving Loads on Ice Plates. 1996
ISBN 0-7923-3953-3
46. A. Pineau and A. Zaoui (eds.): IUTAM Symposium on Micromechanics of Plasticity and
Damage of Multiphase Materials. Proceedings of the IUTAM Symposium held in Sevres,
Paris, France. 1996 ISBN 0-7923-4188-0
47. A. Naess and S. Krenk (eds.): IUTAM Symposium on Advances in Nonlinear Stochastic
Mechanics. Proceedings of the IUTAM Symposium held in Trondheim, Norway. 1996
ISBN 0-7923-4193-7
48. D. Ie~an and A. Scalia: Thermoelastic Deformations. 1996 ISBN 0-7923-4230-5
Mechanics
SOUD MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
49. J.R. Willis (ed.): IUTAM Symposium on Nonlinear Analysis of Fracture. Proceedings of the
IUTAM Symposium held in Cambridge, U.K. 1997 ISBN 0-7923-4378-6
50.· A. Preumont: Vibration Control ofActive Structures. An Introduction. 1997
ISBN 0-7923-4392-1
51. O.P. Cherepanov: Methods of Fracture Mechanics: Solid Matter Physics. 1997
ISBN 0-7923-4408-1
52. D.H. van Campen (ed.): IUTAM Symposium on Interaction betwten Dynamics and Control in
Advanced Mechanical Systems. Proceedings of the IUTAM Symposium held in Eindhoven,
The Netherlands. 1997 ISBN 0-7923-4429-4
53. N.A. FleckandA.C.F. Cocks (eds.): IUTAM Symposium on Mechanics ofGranular and Porous
Materials. Proceedings of the IUTAM Symposium held in Cambridge, U.K. 1997
ISBN 0-7923-4553-3
54. J. Roorda and N.K. Srivastava (eds.): Trends in Structural Mechanics. Theory, Practice, Edu-
cation. 1997 ISBN 0-7923-4603-3
55. Yu.A. Mitropolskii and N. Van Dao: Applied Asymptotic Methods in Nonlinear Oscillations.
1997 ISBN 0-7923-4605-X
56. C. Ouedes Soares (ed.): Probabilistic Methods for Structural Design. 1997
ISBN 0-7923-4670-X
57. D. Fran\rois, A. Pineau and A. Zaoui: Mechanical Behaviour ofMaterials. Volume I: Elasticity
and Plasticity. 1998 ISBN 0-7923-4894-X
58. D. Fran\rois, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume II: Visco-
plasticity, Damage, Fracture and Contact Mechanics. 1998 ISBN 0-7923-4895-8
59. L.T. Tenek and J. Argyris: Finite Element Analysisfor Composite Structures. 1998
ISBN 0-7923-4899-0
60. Y.A. Bahei-EI-Din and OJ. Dvorak (eds.): IUTAM Symposium on Transformation Problems
in Composite and Active Materials. Proceedings of the IUTAM Symposium held in Cairo,
Egypt. 1998 ISBN 0-7923-5122-3
61. 1.0. Ooryacheva: Contact Mechanics in Tribology. 1998 ISBN 0-7923-5257-2
62. O.T. Bruhns and E. Stein (eds.): IUTAM Symposium on Micro- and Macrostructural Aspects
of Thermoplasticity. Proceedings of the IUTAM Symposium held in Bochum, Germany. 1999
ISBN 0-7923-5265-3
63. F.C. Moon: IUTAM Symposium on New Applications of Nonlinear and Chaotic Dynamics in
Mechanics. Proceedings of the IUTAM Symposium held in Ithaca, NY, USA. 1998
ISBN 0-7923-5276-9
64. R. Wang: IUTAM Symposium on Rheology of Bodies with Defects. Proceedings of the IUTAM
Symposium held in Beijing, China. 1999 ISBN 0-7923-5297-1
65. Yu.I. Dimitrienko: Thermomechanics of Composites under High Temperatures. 1999
ISBN 0-7923-4899-0
66. P. Argoul, M. Fremond and Q.S. Nguyen (eds.): IUTAM Symposium on Variations ofDomains
and Free-Boundary Problems in Solid Mechanics. Proceedings of the IUTAM Symposium
held in Paris, France. 1999 ISBN 0-7923-5450-8
67. FJ. Fahy and W.O. Price (eds.): IUTAM Symposium on Statistical Energy Analysis. Proceedings
of the IUTAM Symposium held in Southampton, U.K. 1999 ISBN 0-7923-5457-5
68. H.A. Mang and F.O. Rammerstorfer (eds.): IUTAM Symposium on Discretization Methods in
Structural Mechanics. Proceedings of the IUTAM Symposium held in Vienna, Austria. 1999
ISBN 0-7923-5591-1
Mechanics
SOLID MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
69. P. Pedersen and M.P. Bendsj/le (eds.): IUTAM Symposium on Synthesis in Bio Solid Mechanics.
Proceedings of the IUTAM Symposium held in Copenhagen, Denmark. 1999
ISBN 0-7923-5615-2
70. S.K. Agrawal and B.C. Fabien: Optimization of Dynamic Systems. 1999
ISBN 0-7923-5681-0
71. A. Carpinteri: Nonlinear Crack Models for Nonmetallic Materials. 1999
ISBN 0-7923-5750-7
72. F. Pfeifer (ed.): IUTAM Symposium on Unilateral Multibody Contacts. Proceedings of the
IUTAM Symposium held in Munich, Germany. 1999 ISBN 0-7923-6030-3
73. E. Lavendelis and M. Zakrzhevsky (eds.): IUTAMlIFToMM Symposium on Synthesis ofNon-
linear Dynamical Systems. Proceedings of the IUTAMlIFfoMM Symposium held in Riga,
Latvia. 2000 ISBN 0-7923-6106-7
74. J.-P. Merlet: Parallel Robots. 2000 ISBN 0-7923-6308-6
75. J.T. Pindera: Techniques of Tomographic Isodyne Stress Analysis. 2000 ISBN 0-7923-6388-4
76. G.A Maugin, R. Drouot and F. Sidoroff (eds.): Continuum Thermomechanics. The Art and
Science of Modelling Material Behaviour. 2000 ISBN 0-7923-6407-4
77. N. Van Dao and EJ. Kreuzer (eds.): IUTAM Symposium on Recent Developments in Non-linear
Oscillations of Mechanical Systems. 2000 ISBN 0-7923-6470-8
78. S.D. Akbarov and AN. Guz: Mechanics of Curved Composites. 2000 ISBN 0-7923-6477-5
79. M.B. Rubin: Cosserat Theories: Shells, Rods and Points. 2000 ISBN 0-7923-6489-9
80. S. Pellegrino and S.D. Guest (eds.): IUTAM-IASS Symposium on Deployable Structures: Theory
and Applications. Proceedings ofthe IUTAM-IASS Symposium held in Cambridge, U.K., 6-9
September 1998. 2000 ISBN 0-7923-6516-X
81. A.D. Rosato and D.L. Blackmore (eds.): IUTAM Symposium on Segregation in Granular
Flows. Proceedings of the IUTAM Symposium held in Cape May, NJ, U.S.A, June 5-10,
1999.2000 ISBN 0-7923-6547-X
82. A. Lagarde (ed.): IUTAM Symposium on Advanced Optical Methods and Applications in Solid
Mechanics. Proceedings of the IUTAM Symposium held in Futuroscope, Poitiers, France,
August 3 I-September 4, 1998. 2000 ISBN 0-7923-6604-2
83. D. Weichert and G. Maier (eds.): Inelastic Analysis ofStructures under Variable Loads. Theory
and Engineering Applications. 2000 ISBN 0-7923-6645-X
84. T.-J. Chuang and J.w. Rudnicki (eds.): Multiscale Deformation and Fracture in Materials and
Structures. The James R. Rice 60th Anniversary Volume. 2001 ISBN 0-7923-6718-9
85. S. Narayanan and R.N. Iyengar (eds.): IUTAM Symposium on Nonlinearity and Stochastic
Structural Dynamics. Proceedings of the IUTAM Symposium held in Madras, Chennai, India,
4-8 January 1999 ISBN 0-7923-6733-2
86. S. Murakami and N. Ohno (eds.): IUTAM Symposium on Creep in Structures. Proceedings of
the IUTAM Symposium held in Nagoya, Japan, 3-7 April 2000. 2001 ISBN 0-7923-6737-5
87. W. Ehlers (ed.): IUTAM Symposium on Theoretical and Numerical Methods in Continuum
Mechanics of Porous Materials. Proceedings of the IUTAM Symposium held at the University
of Stuttgart, Germany, September 5-10, 1999. 2001 ISBN 0-7923-6766-9
88. D. Durban, D. Givoli and J.G. Simmonds (eds.): Advances in the Mechanis o/Plates and Shells
The Avinoam Libai Anniversary Volume. 2001 ISBN 0-7923-6785-5
89. U. Gabbert and H.-S. Tzou (eds.): IUTAM Symposium on Smart Structures and Structonic Sys-
tems. Proceedings of the IUTAM Symposium held in Magdeburg, Germany, 26-29 September
2000.2001 ISBN 0-7923-6968-8
Mechanics
SOLID MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
90. Y. Ivanov, V. Cheshkov and M. Natova: Polymer Composite Materials - Interface Phenomena
& Processes. 2001 ISBN 0-7923-7008-2
91. R.C. McPhedran, L.C. Botten and N.A. Nicorovici (eds.): IUTAM Symposium on Mechanical
and Electromagnetic Waves in Structured Media. Proceedings of the IUTAM Symposium held
in Sydney, NSW, Australia, 18-22 Januari 1999.2001 ISBN 0-7923-7038-4
92. D.A. Sotiropoulos (ed.): IUTAM Symposium on Mechanical Waves for Composite Structures
Characterization. Proceedings of the IUTAM Symposium held in Chania, Crete, Greece, June
14-17,2000.2001 ISBN 0-7923-7164-X
93. V.M. Alexandrov and D.A. Pozharskii: Three-Dimensional Contact Problems. 2001
ISBN 0-7923-7165-8
94. J.P. Dempsey and H.H. Shen (eds.): IUTAM Symposium on Scaling Laws in Ice Mechanics
and Ice Dynamics. Proceedings of the IUTAM Symposium held in Fairbanks, Alaska, U.S.A.,
13-16 June 2000. 2001
ISBN 1-4020-0171-1
95. U. Kirsch: Design-Oriented Analysis of Structures. A Unified Approach. 2002
ISBN 1-4020-0443-5
96. A. Preumont: Vibration Control ofActive Structures. An Introduction (2 nd Edition). 2002
ISBN 1-4020-0496-6
97. B.L. Karihaloo (ed.): IUTAM Symposium on Analytical and Computational Fracture Mechan-
ics of Non-Homogeneous Materials. Proceedings of the IUTAM Symposium held in Cardiff,
U.K., 18-22 June 2001. 2002 ISBN 1-4020-0510-5
98. S.M. Han and H. Benaroya: Nonlinear and Stochastic Dynamics of Compliant Offshore Struc-
tures.2oo2 ISBN 1-4020-0573-3
99. A.M. Linkov: Boundary Integral Equations in Elasticity Theory. 2002
ISBN 1-4020-0574-1
100. L.P. Lebedev, 1.1. Vorovich and G.M.L. Gladwell: Functional Analysis. Applications in Me-
chanics and Inverse Problems (2 nd Edition). 2002 ISBN 1-4020-0667-5; Pb 1-4020-0756-6