Anda di halaman 1dari 13

Computers & Strwcrures Vol. 37, No. 3, pp. 319-331, 1990 0045.7949/90 $3.00 + 0.

00
Printed in Great Britain. Pergamon Ptvss plc

FINITE ELEMENT FORMULATION OF LARGE


DEFORMATION IMPACT-CONTACT PROBLEMS
WITH FRICTION

P. WRIGGWS, T. Vu VAN and E. STBIN


Institut fiir Baumechanik und Numerische Mechanik, Universitat Hannover, Appelstrasse 9A,
D-3009 Hannover 1, F.R.G.

(Receded 5 October 1989)

Abstract-Impact~ontact problems including frictional effects appear in many technical applications. For
these problems a tinite element formulation is presented which is based on a new frictional interface law
and fully implicit algorithmic treatment for the integration of the constitutive relations and the dynamics.
The first is performed via radial return schemes, which are well established in computational plasticity.
As radial return methods are amenable to consistent linear&&ions, a full Newton-type algorithm can be
constructed within a time step. This leads, however, to non-symmetric tangent matrices. Several examples
compare this method with other formation and shows its ethciency.

1. INTRODUCTION new frictional interface law with a non-associated slip


rule which takes into account micromechanical
Frictional phenomena have to be considered when
effects. However, the classical Coulomb law can be
the tangential part of a motion is irn~rt~t in the
recovered as a limiting case.
response of two or more bodies coming into contact
The combination of the frictional interface law
during a static or dynamic process. For frictional
with well-known radial return mapping algorithms
problems different finite element formulations have
(see e.g. [6]) leads to construction of a new algorithm.
recently been proposed, most of which rely on the
The formulation is an extension of the algorithm
classical law of Coulomb. However, different fric-
given in [TJ, and opens the possibility of deriving a
tional laws are available which take into account the
consistent tangent matrix. Thus it ensures quadratic
local, ~crom~hani~al phenomena within the con-
convergence near the solution within a Newton iter-
tact interface. An extensive overview may be found in
ation. A similar approach may be found in [S] for a
[l]. In recent years, frictional phenomena have been
different interface model. As a result of the non-
considered within the framework of the theory of
associated flow rule the resulting tangent matrix is
plasticity. For the physical background see, e.g. [2].
non-symmetric.
Elastic-plastic relations for friGtiona1 problems are
Several examples show the performance and
well known and have been proposed, e.g. in [3-S].
efficiency of the developed finite element formulation.
Until now, most of the finite element implemen-
tations of frictional phenomena have been based on
2. THE FRICTIONAL INTERFACE LAW
the classical Coulomb law. Many of the associated
algorithms are constructed on an ad hoc basis. How- Contact constraints can be incorporated via differ-
ever, elastic-plastic relations are also used to devise ent formulations into the general variational frame-
frictional algorithms. In these cases it is often work of finite element methods. The approach
assumed that the normal contact stresses are known, between two bodies is characterized by the following
which implies an associated flow rule (see e.g. [5]). distance function:
This approach leads to a two-step algorithm when
applied to frictional contact problems with an a priori
gN = (4 ’ - (b2). n2 on r, = Wn8B2. (1)
unknown contact stress distribution, in which, in the
first step, a solution of the normal contact problem
is found, and in the second step the frictional stresses gN denotes the gap between two bodies, cftAis the
due to the known normal stresses are calculated. Such deformation of the body BA and rr’ is the normal
algorithms need a considerable number of iterations associated with the surface dBA.The non-penetrating
for convergence. This is also true for algorithms condition
which change normal and tangential contact stresses
simul~eously. g,&O on r, (2)
In this paper a new finite element formulation for
frictional contact problems is given. The basis for this is a commonly used mathematical approximation for
development is an elastic-plastic formulation of a the kinematics within the contact interface.

319
320 P. Wruoosas et al.

However, the above condition does not represent that a sublinear relation exists in general between
the real physical behaviour in the interface, which is pressure and true contact areas.
considered in the following approach. 2.1.2. Approach of the contact surfaces. Based on
statistical models, Kragelski et al. [9] developed a
2.1. Frictionless contact law contact law which relates the normal stress in the
contact interface to the approach between the two
2.1.1. The true contact area. It can be observed in
contact surfaces:
experiments on the micromechanical level that two
bodies come into contact only at the asperities of the
rough surfaces. From these experiments and their ]aJ = c]gJ’“, m = 2-3.3. (4)
statistical evaluation, Tabor [2] extracted a relation
between the true contact area A, and the contact
force. In (4) g, is the approach normal to both contact areas
For elastic deformations of quasi-spherical contact and eN is the mean interface pressure. The parameters
areas one observes a proportionality between A, and c and m describe the stiffness of the contacting
It,#, with the limiting cases. surfaces. These parameters depend on, e.g. the
material and the micromechanics of the surface.
n = f, which means that the number of contact points Remark 2.1. Often the penalty method is used to
is constant during loading; solve contact problems. In this method the term
n = 1, which means that the number of contact points l/2 jas, cgf d(aB) is added to the function to incor-
increases during loading. porate the contact conditions. Its first variation yields
In the case the fully plastic deformation within the JaDcEgg,6gNd(dB), which is equivalent to the virtual
true contact areas or when the form of the surface work of the contact stresses jaBcua,6gNd(c%r). By
roughness is different from a spherical cap, the comparing the last two expressions we recover a
limiting case for the law above is (n = 1), which is also linear interface law for the normal contact stresses
known as the law of Meyer. crN= cg,. Thus, a linear version (m = 1) of the inter-
A similar relationship may be found in [9]; face law (4) may lead to a choice of the penalty
Kragelski et al. proposed for elastic deformations parameter on physical grounds.
within the true contact area the exponent n = 0.8-0.9
and for plastic deformations n = 1. An overview of different experimental data and con-
The evaluation of different contact experiments tact interface laws may be found in [ 111,in which was
leads to a nonlinear relation between the true contact proposed the law (4) up to a pressure of 5 MPa. For
area and the contact load t, (see [lo]). a higher pressure, an exponential law is favoured
(see [12]): (QJ = a exp 6(gN]). c, a, b and m are vari-
ables which describe the interface law. For most
: metallic materials the parameter cNis proportional to
n=-.
the modulus of elasticity, whereas m is approximately
two.
Here A, is the true and A is the nominal contact area,
and H denotes the hardness of the material.
Figure 1 depicts the logarithm of the dimensionless 2.2. Frictional interface law
pressure It,(/AH vs the ratio of the true contact area Frictional phenomena have to be considered in the
to the nominal area. case of a relative tangential motion within the contact
Equation (3) depends strongly on the material and surface. To derive the associated interface laws we
surface parameters. However, it can be concluded will neglect influences of temperature and tangential
velocity on the interface law. As an alternative to the
classical Coulomb law we will develop a frictional
interface law which includes the decrease of frictional
& ‘- coefficient with increasing contact pressure. Based on
A
this relation a frictional law analogous to the theory
of plasticity will be formulated.
By the classical assumptions of the frictional law a
body rests until the maximum shear force is reached.
In reality, however, bodies show tangential micro-
displacements within the contact area before the
sliding process starts. This behaviour is depicted in
Fig. 2, where the relation between the tangential force
IO-' 4 and the micro-displacement is slightly nonlinear in
IO-' IO-' IO-* 10-l I m M the presliding phase (see e.g. [9, 111).
AH
Fig. 1. Dimensionless plot of the true contact area vs A reasonable approximation of the nonlinear be-
pressure [lo]. haviour in the pre-sliding phase is given by a linear
Impact-contact problems with friction 321

tT Using (6), eqn (8) can be rewritten as

I The slip condition for the frictional force follows


from (9):

-- Afo
f=-alfNl”-~l~Nl+lltTtl~O, a-(AHY. (10)

Fig. 2. Friction force vs tangential displacement [9].Zone 1: This condition is analogous to the yield condition
preliminary displacement. Zone 2: sliding. in plasticity. Condition (10) depicts in the three-
dimensional case a slip paraboloid. Thus (10) in-
cludes the phenomenon of decrease in frictional
relation between the tangential micro-displacements coefficient with increasing normal pressure. The par-
and the shear force: ameters CL,/I and n depend on surface properties such
as roughness or hardness and have to be deduced
tT= cTgT, (9 from experiments. It should be noted that the sliding
rule used in conjunction with slip condition (10) is
which will be used throughout this paper. non-associated. Reasons for this have been pointed
When the bodies start to slide within the contact out for the classical Coulomb law in earlier work,
interface we have to consider a different law for the e.g. [13].
tangential forces. From experiments an equation for Assuming that the true contact area is proportional
the frictional shear stresses can be obtained in depen- to the contact pressure p, = H = ItJ/A, (n = l), one
dence on the normal contact pressure, (see e.g. [9]). recovers the slip cone of Coulomb as a limiting case:
The shear stresses in the contact area can be written
as a linear function of the contact pressures (this is f = -,‘#Nl + Ilt,II & 0. (11)
depicted in Fig. 3)
Here, the frictional coefficient
is given by
(6) p = (to/H) + /I. Other formulations, such as that
given in [14], use a slip ellipsoid as the slip criterion
Thus the frictional coefficient is given by together with an associated slip rule.
The formulation of a frictional law analogous to
the theory of plasticity leads to the split of the
tangential velocity in the contact interface:

It should be noted that the frictional coefficient p in


(7) is constant when the pressure pr reaches the yield
pressure H. Here the first part of the right-hand side is associated
We can now combine eqns (3) and (6) to give a with the incremental micro-displacements before
new formulation of a frictional law. As the yield sliding; see eqn (5). With the postulate of an associ-
pressure will not be reached in the true contact area ated slip rule the slip direction is defined by the
(n < l), the mean pressure p, in (6) can be expressed gradient of the slip condition:
by (3):

(8)
However, (13) leads to uplifting displacements
which produce negative dissipation. Thus, because of
thermodynamical restrictions, a non-associated flow
rule has to be used in combination with (lo), (see
e.g. ]3,4,71).
The possibility of circumventing this problem,
based on the assumption of a known normal contact
stress distribution in the contact area, is followed in

most implementations of friction. In this case the
slip paraboloid (10) decreases to a circle at the
known normal stress tN = constant. Thus (12) can be
Fig. 3. Relation between shear and normal stresses in the neglected, which leads to an associated slip rule for
contact area [9]. fixed tN =constant. (For a variational framework,

CAS 37/3-F
322 P. WRIGGERS et al.

see [15].) However, algorithms based on this assump- the so-called stress divergence term and P is the load
tion need a staggered form: first the normal contact pattern. Because of geometrical or material non-
stresses have to be calculated with fixed tangential linearities the system of equations (15) is nonlinear in
stresses and then, in a next iteration, the frictional v. M is the consistent mass matrix. Here we will use
stresses have to be computed with fixed normal a lumping technique to obtain a mass matrix of
stresses. In many cases the convergence properties of diagonal form. (For a discussion of different ap-
these algorithms are not satisfactory. proaches, see e.g. [16].)
In this paper, no assumption of a constant normal Remark 3.1. The degree of nonlinearity depends on
pressure in the contact area is made. According to the the theory used to describe the problem at hand.
thermodynamical restrictions, the following non-
associated slip rule is used to formulate the frictional (1) In the three-dimensional theory of nonlinear elas-
interface law: ticity using St Venant material R is a cubic poly-
nomial in v.
(2) In shell theories for large rotations v depends on
t~gonomet~c functions (see e.g. [17]).
(3) The incorporation of elasto-plastic materials
The parabolic slip surface associated with the
leads-depending on the plasticity model-to differ-
sliding condition (10) and the slip rule (14) is shown
ent types of nonlinea~ty.
in Fig. 4.
The advantage of (14) is that a two-step algorithm
The methods for the solution of (15) are standard.
can be avoided, However, this leads to a non-
For two-dimensional applications Newton’s method
symmetric tangent matrix. Its good convergent
has been proved to be advantageous.
behaviour in combination with a return mapping
scheme will be depicted by means of examples. 3.2. Finite element formulation of frictional contact
Attention is now focused on the planar and
3. FINlTE ELEMENT FORMULATION
axisymmetric case of a straight element boundary
3.1. Standard FE formulation without constrailtts undergoing finite deformations. This formulation is
Standard finite element approximations yield in valid for a class of elements such as beam, axisym-
case of dynamics a nonlinear system of ordinary metrical shell and two-dimensional isoparametric
differential equations: elements.
The constraint equations are based on nodal
G(v,t)=Mii+R(v)-P=O, VCR”. (15) values. In this context, and with reference to Fig. 5,
the gap or penetration is approximated by the finite-
Here the body B is discretixed by n, finite elements q, element method nodally. Here, the gap or ~netration
which leads to Bh = u ::, 1R,. The vectors R, P and gh: associated with a typical slave node s is indicated
the matrix M are given by by the inequality

gN, = (x, -x,)9l>O, vs:E4,. (17)

n denotes the normal to the master segment 1-2,


x, = X, + v, defines the current position of the slave
node, and xl = X, + v, , x2 = X2 + v2 define the cur-
rent position of the master segment. Inequality (17)
has to be checked for all candidate contact nodes s
(16) which are contained in the finite set YC. In general,
9C contains all boundary nodes of the contacting
bodies. For gN, < 0, the constraint equation for a
where N contains the shape functions and B their
node s becomes active (s E fA); otherwise, the con-
derivatives according to the theory used. R denotes
straint is inactive (s E fr), with #C = fAu9[ and
9,n.Ff=0.
It should be noted that eqn (17) is valid for the
general contact of two or more bodies, (see e.g. [18]).
If u~lateral cons~aints are present, the master
surface becomes rigid, which is associated with
Vl -- v2 2 0.

Fu~he~ore, we define the length of the master


element i-2 and the local-base vectors by

- Xl II, t =; 62 - X*)t n = e3 x t.
1 = 11x2 (181
Fig, 4. Parabolic slip surface.
Impact-contact problems with friction 323

gjv=hR-hnr=u.-gR

Fig. 5. Discrete formulation of the approach in the contact area. g, = h$ - h, = u - g$

With these quantities we are able to define the natural define a displacement vector for the three-node
coordinate a : contact elements (s, 1,2)

v’= (~~~*~*~~~~#~)r, (24)


o! = f (x, - xi) *t, a E [0, 11, (19)
and the vectors

which measures the relative position of the slave node


s within the master segment l-2. We are now able to K=[-(:#I}, N,,={ -3, (25)
define the relative slip of the slave node s on the
master surface l-2:
and

g, =&+(a -E)l:

In this equation the overbar denotes quantities


(20)
T’=( -a, (26)
-(;;ct)t}.
To={
related to the reference configuration. Thus the initial
length of a segment is given by r= l/Xl - Xi 11. With these definitions we obtain the matrix formu-
The virtual work of boundary nodes which are in lation of eqns (22) and (23):
contact is now formulated for one slave node s:
=N;6v’,
&?.VS 6gr*=f I(T,+~N,)).~~. (27)

Thus the virtual work which has to be added to the


Here, rlv,and tr, are the normal and tangential force, standard finite element equation (15) because of
respectively, at the node, SgNSdenotes the virtual gap contact can be expressed with (21) for all nodes which
and agr, is the virtual tangential displacement. are in contact by
The variations of the gap and the tangential dis-
placements in (21) are given by

=@v,- (1 -a)&,
~&zs - afhj)~o, (22) In (28) the contact forces trJ, and tr, are known from
the constitutive interface laws. The directions of these
forces are defined by N$, T, and NO. &cause of the
&r,=f (Sv,-(1 -a)&,-a&)-t
[ known interface laws, a pure displacement formu-

1.
lation of the contact problem is possible, which is in
contrast to the Lagrangian multiplier technique often
+g+ (SV~-SV,)‘D 03) used in eqn (2). Furthermore, the proposed method
can be viewed as a regularixation technique, like the
For finite element implementations we need the penalty method, based on physical grounds (see also
matrix form of eqns (21)-(23). For this purpose we Remark 2.1).
324 P. WRIGGFM et al.

It should be noted that the vector R, reduces for the steps than in explicit schemes. Thus, in this work,
infinitesimal theory to Newmark’s method is used (see [23]). As the update
formulae of this scheme are well known (see e.g. [16]),
we do not specify them here explicitly. Their incor-
poration into the nonlinear algorithm leads to the
so-called effective tangent stiffness and effective
Finally, we obtain the global nonlinear finite element residuals which will be used in the general algorithm
equation extended by contact forces (28) (see Sec. 4.3).
4.2. Algorithmic treatment of the contact
G;,(v, t) = Mii + R(v) + R, - P = 0, v E R”, (29) Within the solution ~go~thm Newton’s method
will be used to achieve equilibrium within one time
which is the basis for the algorithmic treatment step. For this purpose a consistent linearization of the
within the finite element method. contact contributions is necessary. We will now de-
rive the associated finite element formulation. For
4. ALGORITHMS FOR FRICTIONAL IMPACTXONTACT more clarity in the subsequent developments we will
PROBLEMS suppress the index s in the following which denotes
the slave node of a contact element (see See 3.2).
The algorithmic treatment of impact-contact prob-
4.2.1. Frictionless contact. The consistent lineariza-
lems involves several aspects which will be discussed
tion of the frictionless part can be found for the
below in detail. As can be seen from eqns (28) and
contact discretization (17) in [24]. It stems from the
(29) we have to solve a nonlinear ordinary differential
linearization of the term tN&glyusing the directional
equation with inequality constraints as a resuJt of
derivative in (28):
contact. For this pupose, we will use a standard
implicit method, the Newmark scheme. Impact-
contact and release conditions which have to be (30)
addressed in Lagrangian multiplier techniques (see
e.g. [19]) do not enter this method. Here the same
arguments apply as in penalty methods (see e.g. [20]).
F~he~o~, physical nonlinea~ty associated with
the frictional sliding or the material model of the
bodies has to be considered. As an iterative scheme
and
we apply Newton’s method for the solution within a
time step. For this purpose, we have to linearize the
standard finite element formulation as well as the AL\6g,=6vc- - f T, N: + N,T:
contact interface laws for the normal and tangential [ (

>I
part. The frictional interface law will be integrated by
+y N,N,T Av’. (32)
using a radial return scheme, which is well established
in computational plasticity (see e.g. [6,21]). Its lin-
earization leads to a non-symmetric tangent operator. Inserting (31) and (32) in (30) yields the tangent
We do not state the linearization procedures for stiffness for the normal contact:
standard f&&e element techniques which lead to the
tangential stiffness matrix of the finite element model,
K -!&qLtN i T~N~+N~T~~~N~N~ . (33)
explicitly. They can be found in text-books (e.g. [22]). N-agN s 5 ( >
Consistent tangent operators for elastic-plastic prob-
lems may be found in, e.g. [6]. The matrices N,, T, and N, can be found in eqns (25)
and (26). The derivation of the normal contact force
4.1. Time discretizatioPt with Newmark’s method with respect to the displacement g, is given by the
Explicit or implicit methods can be used for the interface law (4). Numerical experiments have shown
integration of eqn (29). Explicit methods are often that a linearized interface law does not change the
applied for the short-duration impact-contact prob- solution and yields a robust behaviour. Such a choice
lems. Their drawbacks are the dependence of the time is related to the penalty method for normal contact
step on the smallest element size. On the other hand, (see Remark 2.1).
the implementation is very simple and efficient. 4.2.2. Tangential contact, elastic part. In the case of
Un~ndi~onally stable implicit methods do not friction we have to consider additional terms.
possess this disadvantage. However, the numerical Analogous to (30) we obtain the consistent lineariza-
effort is substantial larger as a nonlinear system of tion for the elastic part of the tangential motion-the
equations has to be solved for every time step. micro-displacements-in the contact interface from
Our experience is that for impact-contact problems
where no inherent shock waves appear, implicit inte- D(tr agr) bgr -%Ag&gr+
agr t,Adgp (34)
gration can be used with considerable larger time -_
Impact<ontact problems with friction 325

Using eqn (28) the first term in (34) can be written as The unknown L is computed by inserting (43) into
(40):

Ag,dg,=6P+-, (35)

with the symmetric matrix + IG+, - Ic, sign(t, + ,)I = 0. (44)

A=(T3+~No)(Ts+~No) For constant parameters a and j? we can solve (44)


with respect to L explicitly. In the case of hardening
effects in the frictional law, these parameter will
depend on the tangential displacements (see e.g. [5]).
=[r.T:+$Nf+N,T:+~
NON:)]. Then (44) is nonlinear in 1, which leads to a local
Newton iteration for the solution of (44).
When 1 is known, the inelastic slip increment is
It should be noted that eqn (36) has the same
structure as (33). Furthermore, we have to compute given by
A6g,, which leads to the matrix formulation
Agi;! = Iz sign(t,), AgFO= IAgfl = 1. (45)

A6g, = 6ti.i B Avc, (37) We may now derive the cons&e@ linearization of the
algorithm stated above which leads to the tangential
stiffness matrix for the sliding part. The frictional
with force depends, in contrast to the elastic part of the
tangential motion, on the normal forces in the
B=(N,N,T+N,N,T)--(T,T,T+T,,T:) contact area. Thus we obtain from the virtual
work expression fT6gT with g = {gN, gT} the lin-
-2e(N earization
0Nr+T
0 0Nr)
0. (38)
I

We can now combine eqns (35) and (37) with (34) to


find the symmetric tangent stiffness of the elastic part
of the tangential motion in the contact area:
+ 2 AgN6g, + fT ASg,. (46)
N

k,=$;A+l,;B. (39)
T
As the linearizations with respect to g, and g,
have already been derived, only the linearization of
In (39) the derivation of the tangential contact forces
the frictional forces has to be considered in the
with respect to the displacement g, is given by the
following.
interface law (5). Thus it may be replaced by cr. The tangential force tr is known only via the
4.2.3. Tangential contact, sliding part. The normal algorithm described above. Thus, the differentiations
and frictional forces at the element nodes have to with respect to gr and gN yield
fulfil the slip condition (10) which can be rephrased
for two-dimensional problems as follows:
ah.,
-=
ah + I
f(fNV rT) = -alt,l” - BM + lb1G0. (40)

With the kinematic split (12) and the non-associated (47)


slip rule evaluated for the plane case, the return
mapping scheme can be formulated: The differentiation of 1 with respect to tN and tS
follows from the slip condition (40) as no explicit
&! = 0 gl;n= n’ sign(t,). (41) dependence of 1 as a function of the forces rN and t;
is known. The final result of this implicit differen-
With the elastic predictor tiation yields the incremental constitutive parameters

%+l = fTn + dan+l -an)k (42) +&=+(1+.&J,

we obtain for the frictional force


c;=-=atTn+ I
-cNcrsign(t,)
af+l/atNn+l
af,+, ,an . (48)
agNn + I
326 P. Wamosas et al.

The non-symmetric frictional tangent matrix follows Update: g$ = g:,, , compute matrices
with lTn+l from (41) and eqns (46)-(48): k,, k,, go to (e)
ELSE
Sliding
k&;A+fm+,;B+c:C. (49) Elastic predictor: t;, = tT,
Compute 1 from
Matrix b differs from kH [see eqn (39)], by the f(n)= -alfN,In-BlfN,I
different incremental constitutive parameter c$, and
+ It!+,- Ic, sign(tt ] = 0
Compute: tT,, g$ and matrices kN,, k,
the non-symmetric part c:C, which stems from the
coupling between normal and tangential force in the go to (e)
ENDIF
contact area. The matrix C yields
(e) Continue.
(5) Solution of the incremental system of equations
C=T,Nf+gFN,N:. (50) KF AV’ + Ga’ = 0.
(6) Check for convergence:
IF IlG@ll > Tol: Set (O)i = (i)‘+’ ELSEIIG&‘lI
4.3 Overall algorithm <Tel: go to (8).
In this section we will state the overall algorithm (7) Update VifL=Vi+AVi and gi+‘,V’+l, go to
for the solution of impact-contact problems. For (2).
simplicity we do not use an index for quantities at (8) Updates for next time step:
time t + At. Quantities associated with the time t will If t <tend: Update (0) =(e), go to (2).
be denoted by an overbar. The global matrix KF is (9) stop.
assembled from the contributions of tangent stiffness
matrices for the elements and the contact segments as 5. NUMERICAL EXPERIMENTS
well as the mass matrix according to the Newmark
formulae. The same considerations hold for Cc”, In this section we present a selected sample of
which is assembled from the stress divergence terms numerical simulations that demonstrate the applica-
of the elements, the contact forces (29), and the mass bility of the method developed in the previous sec-
terms due to the Newmark formulae. tions to a wide range of frictional contact-impact
problems. To illustrate the efficiency and convergence
(1) Time t = 0. characteristics of the proposed method, we report on
Set initial values, loop over all time steps. explicit rates of convergence obtained in the course of
(2) Equilibrium iteration at time t + At: these simulations. The developed algorithms have
set initial values and i = 0. been implemented in the general purpose finite
(3) Loop over all elements, compute K;B’, GeRi. element code FEAP (see Taylor in [22], Chapter 24).
(4) Loop over all contact segments s.
5.1. Contact of an elastic block with a rigidfoundation
Check for contact and compute contact forces.
(a) Check for contact: IF: gN, = (x, - x,).n > 0: For purposes of comparison, a numerical exper-
go to (e). iment of Oden and Pires [25] is investigated in this
(b) Compute normal force: tN, and matrix k, paper. The problem considered is that of an elastic
IF (frictionless contact): go to (e). block pressed against a rough rigid foundation. The
(c) Compute frictional force: tT, = tTs, + cT elastic block is then pulled by a tangential force
(a - E)L uniformly distributed along one of its lateral sides
(d) Check slip condition: (see Fig. 6).
IFf(t,,rrz)= -aIt,I”-pItlv,I+ltT~I~O To compare with the results of Oden and Pires the
THEN contact area is restricted such that the first and the
Elastic last element in the contact area do not come into

I I I

1 ::“,‘I
Fig. 6. System, FE discretization and deformed configuration.
Impact-contact problems with friction 327

I
d. 2.

x-Coordinate x-Coordinate
Fig. 7. Stresses on contact area (conical slip surface). Con- Fig. 8. Stresses on contact area (parabolic slip surface).
tact normal stress: +, [25]; *, computed result. Contact
frictional stress: 0, [25]; x , computed result.
the conical slip surface. Here the total load has been
applied in one step.
contact with the rigid surface. Young’s modulus E
and Poisson’s ratio v are 1000 per length square and 5.2. Sliding and rolling of a ring on a rigid surface
0.3, respectively. The applied loads are p, = 60 and
p,, = 200 per unit length. To reproduce the data given This textbook example has been considered to test
in [25] the properties of the contact surface have been the frictional interface law and the algorithm. A steel
chosen as follows: m = 1, c, = lo*, cr = 104, n = 1 ring of R = 50 cm with a rectangular cross-section
and a +/I = 0.5, which is equivalent to Coulomb’s of A = lOcm* and an elasticity modulus of
conical slip surface. Here, a nonlinear dependence of 2.1 x 104kN/cm* is pushed with an initial tangential
the frictional stress on the normal stress has been velocity of ur= 300 cm/set along a straight rigid
neglected as the formulation of Oden and Pires does surface. The density has been chosen as
not include this effect. 7.85 x lo-’ kNsec/cm4. One hundred and twenty
The block is discretized using 200 four-node iso- finite elements have been used to discretize the ring.
parameter% elements. It should be noted that with To compare with the exact solution, Coulomb’s law
the developed frictional algorithm the total load can has been applied with a frictional constant of
be applied in only one step. The computed contact v = 0.15. The parameters of the interface law are set
stresses are depicted in Fig. 7, and indicate good as follows: m = 1, c, = 10’ kN/cm and cr = lo* kN/
agreement between the solutions. cm. The time integration was performed by
To represent the nonlinear dependence of the Newmark’s method with the parameters S = 0.7 and
friction stress on the normal stress the parameters of j? = 0.4. Throughout the computation a time step of
the friction law are varied: n = 0.5, u = $2, /? = 0.3. At = O.O1s has been used*
This results in a parabolic slip surface. Figure 9 shows different states of the ring, begin-
Figure 8 shows the contact stress distribution ning with pure sliding and ending with rolling. The
for the parabolic slip surface. We observe only a values computed using the exact solution are stated
small change of the normal contact stress; however, in brackets. Figure 9 also shows the velocity profiles
a considerable reduction of the frictional stress is for the different solution states. The good agreement
detected in the region of higher normal pressure, of the finite element solution with the exact solution
which is in agreement with experimental obser- for a rigid ring is noteworthy.
vations.
Table 1 shows the quadratic rate of convergence of 5.3. Sloping impact of a ring on a rigid foundation
the Newton method. It should be noted that only four A ring with a quadratic cross-section of A = 1 and
iterations are necessary to achieve convergence for a radius of R = 10 impacts a rigid plane surface under

Table 1.
Norm of residuum llGj

Iteration no. Conical slip surface Parabolic slip surface


1 0.1713709 x 10’ 0.1713709 x 103
2 0.1196315 x lo2 0.2446232 x lo2
3 0.3218561 x 10’ 0.1265767 x lo2
4 0.4368441 x lo-” 0.5389606 x 10’
5 0.1306383 x 10’
6 0.1098278 x lo-$
7 0.4359430 x 10-u
328 P. Waroosas et al.

t-23* (229) 1x= 382 (376) , [cm]

1 1

Pro&s of velocity [cm/s]

300 (300) 149 (150)

n @JL(75) 147Yo(5)

300 156 (150) 16 (0) 1 (0)


Fig. 9, Rolling of a ring on a rigid surface, velocity profiles.

an angle of incidence of a, = 45” with the velocity a viscoplastic beam element, which can be found
u,= 2, t+ = 2. A uniform mesh of 16 linear beam in [26]. The plastic values for the stress resultants
elements which include finite deformations was used N, Q and M are 1, 0.5 and 0.25, respectively. The
in the computations. The formulation of this element non-uniform deformations and a lower upward
is found in [26,27]. The material data chosen are movement of the ring indicate the loss of energy
E = 100, v = 0 and p = 0.01. The time integration in the system because of plastic deformations [see
was performed by Newmark’s method with the par- Fig. 1l(b)].
ameters 6 = 0.7 and /I = 0.4 and a time step At = 0.1.
The data for the interface law are 1?1= 1, cN= 50, 5.4. Fall of an elastic ring into a funnel
cr= 30, n = 0.8, a = 0.2 and /I = 0.3. In this example, the same ring as in Sec. 5.3 is
The motion of the elastic ring is shown in Figs 10 considered, which falls with an initial velocity v,, = 4
and 11(a) for the case of frictionless and frictional into a rigid funnel. Because of symmetry only one-
contact, respectively. Two hundred time steps have
been used for the overall impact-contact calculation.
Figures 10 and 11 indicate the difference between
the solutions. For the case of frictional contact the
ring suffers an additional rotation after rebounding
from the foundation. The angles of reflection and the
components of the velocity of the ring after contact
are given in Table 2.
Next, the same problem is considered for the
case of inelastic response of the ring. Here we use

Fig. 10. Motion of the elastic ring for the case of frictionless Fig. 11. Motion of the ring for the case of frictional contact.
contact. (a) Elastic; (b) viscoplastic.
Impact-contact problems with friction 329

Table 2.
Velocity Angle of
components reflection, al
Frictionless contact t&=2 * Uv’ 1.7 50”
Frictional contact u, = 1.2, uv= 1.7 35”

half of the ring was discretized, by 30 beam elements.


For the interface law within the contact surface the
parameters m = 1, c, = 10 and e, = 5 have been
chosen, The frictional law (10) is used with the
following data: n = 0.75, a = 0.3 and fi = 0.1.
In Fig. 12 the motion of the ring is plotted in
intervals of five time steps, the time step being
At = 0.1. For the case of frictionless contact a large
elastic deformation of the ring is observed during the
contact process; moreover, the ring rebounds, as
there is no energy loss in the system [see Fig. 12(b)]. Fig. 14. Finite element model of the cork.
By taking into account frictional phenomena, the ring
sticks to the funnel wall for the given data above [see of the interface law are m = 1, cN= 100 N/mm,
Fig. 13(b)]. Furthermore, the elastic deformation is cr = 10 N/mm.
smaller, as in the frictionless case. Figure 14 depicts the geometry of the cork. The
axisymmetric cord was discretized by 114 four-node
isoparametric elements. As the deformation of the
5.5. Champagne cork under internalpressure bottle is negligible, only two elements were used to
In this example, the motion of a cork of a cham- discretixe the bottle neck.
pagne bottle during opening is discussed. The cork is The initial connation of the cork being pressed
described by an isotropic and hypereiastic material into the bottle neck with internal radius of IOmm,
law with the material data E = 10 MPa, v = 0.25, was achieved by a quasistatic contact calculation. As
p = 3.5 x IO-” N sec2/mm4. The material data the cork has a maximum radius of 12mm, large

(a) (bl
v, = 4.0 -

Fig. 12. Motion of the ring for the case of frictionless contact.

(at (b)

12= 0.75
Q = 0.3
p = 0.1

Fig. 13. Motion of the ring for the case of frictional contact.
330 P. Waioozas et al.

7 n

t = 0.0

Ii
ms 0.8 ms

II!
1.0 Ins

1
1.5 ms

t = 2.0 ms
L’11
ms
3.0 Ins
28mls

i
! II 1

Fig. 15. The configurations of the cork at several time instants (0.0,0.8, 1.0, 1.5,2.0,2.6,2.7 and 3.0 msec).

strains appear during this process. Thus a geometric an explicit procedure, a small time step must be
nonlinear theory with finite deformation had to be adapted to avoid considerable errors in the solution.
applied for this example. For the frictional law (lo), Thus a time step of At = 1 x 10e4 s was used as long
the following data were applied: n = 1 and as the cork moved inside the bottle.
h = a + /l = 0.1, where, for the frictional coefficient, The internal pressure was applied in eight time
a linear hardening law p = k + rg! with 5 = 0.001 steps of 0.1 MPa. It was then reduced according to
was used. the above equation for the next time steps. Because
In the initial configuration a normal force of of the assumption that the distance between the
t,,,x 2500 N acts in the contact zone surrounding the liquid and the cork is 3Omm, a pressure of
cork. With the assumed frictional coefficient the cork pi = (30/54) 0.8 = 0.56 MPa acts on the cork just be-
begins to slide at a tangential force of t, x 250 N. fore it leaves the bottle. After that the pressure drops.
This force is created by an internal pressure of about During a short time period just before and after the
0.8 MPa. Because of the high load in relation to the cork leaves the bottle, the time step was reduced to
small mass of the cork, it is essential to describe the At = 5 x 10e6 s, as the cork suffers great deformation
pressure precisely during the dynamic solution. In at this time. After about 3-4 x 10e4 s the larger time
general, this is a coupled problem, as the pressure is step was used again.
inversely proportional to the volume. Here we use an The motion of the cork is shown in Fig. 15. Based
explicit scheme for the coupling. Hence the pressure on the given data the cork leaves the bottle with a
is computed from the law p,, V, = pi Vi = constant as velocity of appromimately 28 m/set. The graphs in
p, = (V,,/Vi)po for the next time step i + 1. As this is Figs 16 and 17 show the pressure and the displace-

81 ,
6. 1. 1. ’ 3.*
Time / ms Time / ms
Fig. 16. The pressure on the cork vs time. Fig, 17. The displacement of the cork vs time.
Impact-contact problems with friction 331

ment of the cork vs time. It was observed by par- 14, F. P. Bowden and D. Tabor, The Friction and Lubrica-
tion of Solids, Part II. Clarendon Press, Oxford (1964).
ameter studies that the velocity of the cork is affected ls
G. Duvaut and J. L. Lions, Znequolities in Mechanics
primarily by the initial gas volume. and Physics. Springer-Vedag, Berlin (1976).
16. T. J. R. Hu&es. The Fmite Element Method: Linew
Static and D&a& Finite Element Analysis. Presitice-
RJlFEitENCXS
Hall. Enalewood Cliffs. NJ 1198n.
1. J. ‘I’. Oden and J. A. C. Martins, Models and compu- 17. F. dm&ann, Theo& und‘ Nurkerik schubelasticher
tational methods for dynamic friction phenomena. Schalen mit endlichen Drehungen unter Verwendung
Comput. Meth. appl. Mech. Engng 52, 527-634 der Biot-Spannungen. Forschung- und Seminarbe-
(1985). richte aus dem Bereich der Mechanik der Universitiit
2. D, Tabor, Friction-the present state of our under- Hannover, Bericht Nr. F 88/l (1988).
standing. l. &u&r. Technol. 103, 169-179 (1981). 18. P. Wriaaers. W. Wanner and E. Stein. Alaorithms for
3. R. Michaloswki and Z. Mroz, Associated and non- nonline: c&tact &&mints with ap&a~on to stab-
associated sliding rules in contact ftiction problems. ility problems of rods and shells. Comput. Mech. 2,1-16
Archs Mech. 30,-259-276 (1978). (1987).
4. A. Cumier. A &ON of friction. ht. J. Soli& Struct. u), 19. T. J. R. Hughes, R. L. Taylor, L. Sackman, A. Curnier
637-647 (1984). and W. Kanoknukulchai, A finite element method for
5. B. Fredriksson, Finite element solution of surface non- a class of contact-impact problems. Cornput. Meth.
linear&s in structural mechanics with special emphasis appl. Mech. Engng 8, 249-276 (1976).
on contact and fracture mechanics problems. Compur. 20. J. Hallquist and D Nike. An implicit, finite deformation,
Struct. a, 281-290 (1976). finite-element code for analysing the static and dynamic
6, J. C. Simo and R. L. Taylor, Consistent tangent response of two-dimensional solids. Rep. UCRL-52678,
operators for rate-independant elastoplasticity. Comput. University of California-Lawrence Livermore National
Math. URRI.Mech. Engng 48, 101-118 (1985). Laboratory (1979).
7. P. Wri&rs, On c&is&t tangent matrices for fric- 21, M. Ortiz and J. Simo, An analysis of a new class
tional contact nroblems. Proc. NUMETA 87 Confi of integration algorithms for elastoplastic constitutive
(Edited by J. I&ddleton and G. N. Pande) (1987). _ relations. ~artment of Engineering, Brown Univer-
8. A. G. Giannokopoulos, The return mapping method for sity (1984).
the integration of friction constitutive relations. Com- 22. 0. C. Zienkiewicz, The Finite Element Method.
put. Strut. 32, 1.57-168 (1989). McGraw-Hill, London (1977).
9. 1. V. Kragelsky, M. N. Dobychin and V. S. Kombalov, 23. N. M. Newmark, A method of computation for struc-
Friction and Wear-Calculation Methoak (Translated tural dynamics. ASCE, J. Engng Mech. Div. 85,67-94
from Russian by N. Standen). Pergamon P&s, Oxford (1959).
(1982). 24. P. Wriggers and J. C. Simo, A note on tangent stiffness
10. K. L. Woo and T. R. Thomas, Contact of rough for fully nonlinear contact problems. Common. appt.
surfaces: a review of experimental works. Wear 58, Numer. Meth. 1, 199-203 (1985).
331-340 (1980). 25. J. T. Oden and E. B. Pires, Algorithms and numerical
11. N. Back, M. Burdekin and A. Cowley, Review of results for finite element approximations of contact
the research on tied and sliding joints. Proc 13th problems with non-classical friction laws. Cornput.
Znt. MTDR Conf: (Edited by S. A. Tobias and F. Struct. 19, 137-147 (1983).
Koenigsberger), pp. 87-97. Macmillan, London 26. J. C, Simo, K. D. Hjehnstad and R. L. Taylor, Numeri-
(1973). cal formation of el~to-~~opl~tic response of beam
12. R. Connoly and R. H. Thomley, Determining the accounting for the effect of shear. Comput. Meth. appt.
normal stiffness of joint faces. ASME Paper 67, Prod. Mech. Engng 42, 301-330 (1984).
6 (1967). 21. J. C. Simo, P. Wriggers, K. Schweizerhof and R. L.
13. D. C. Drucker, Coulomb friction, plasticity and Taylor, Finite deformation postbuckling analysis
limit loads. J. appl. Mech., Trans. ASME 21, 71-74 involving inelasticity and contact constraints. Int. J.
(1954). Numer. Meth. Engng 23, 779-800 (1986).

Anda mungkin juga menyukai