Vous êtes sur la page 1sur 24

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2014; 97:683706


Published online 17 December 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/nme.4611

Contact between rolling beams and flat surfaces


Alfredo Gay Neto1, * , , Paulo M. Pimenta1 and Peter Wriggers2
1 Polytechnic

School at University of So Paulo, So Paulo, Brazil


Universitt Hannover, Hannover, Germany

2 Leibniz

SUMMARY
This work presents a new approach to model the contact between a circular cross section beam and a flat
surface. In a finite element environment, when working with beam elements in contact with surfaces, it is
common to consider node or line to surface approaches for describing contact. An offset can be included in
normal gap function due to beam cross section dimensions. Such a procedure can give good results in frictionless scenarios, but the friction effects are not usually properly treated. When friction plays a role (e.g.,
rolling problems or alternating rolling/sliding) more elaboration is necessary. It is proposed here a method
that considers an offset not only in normal gap. The basic idea is to modify the classical definition of tangential gap function in order to include the effect of rigid body rotation that occurs in a rolling scenario and,
furthermore, consider the moment of friction force. This paper presents the new gap function definition and
also its consistent linearization for a direct implementation in a Newton-Raphson method to solve nonlinear
structural problems modeled using beam elements. The methodology can be generalized to any interaction
involving elements with rotational degrees of freedom. Copyright 2013 John Wiley & Sons, Ltd.
Received 21 July 2013; Revised 5 November 2013; Accepted 6 November 2013
KEY WORDS:

contact; rotation; rolling; sliding; beam

1. INTRODUCTION
Many engineering applications present components that are naturally modeled using beam or shell
elements, instead of solids. Contact interaction between these components in such a modeling has
to be considered. As an example, which is the focus of the present paper, one can think of an interaction between beam-like structures and surfaces. Some practical examples are present in offshore
engineering, as long pipelines interaction with the seabed or a rope rolling and sliding in the seabed
as a part of an anchoring system of a ship. The fact is that many practical problems present rolling or
sliding or both situations occurring in different regions of circular cross section beam-like structures.
The rolling/sliding status may even change along time in some transient dynamics problems.
Aiming to model such a kind of problem using a FEM environment with beam elements, leads
to an issue to consider properly the rolling effect with the classical contact formulations. One could
think of modeling everything using solid elements. Then, not only a huge number of degrees of freedom (DOFs) can be achieved, but also very difficult to find convergence nonlinear models have to be
solved, once the nature of rolling includes successive changes of contact points, which is usually a
difficult task when solving a nonlinear contact model. This motivates the improvement of a contact
formulation to be used together with beam elements or, more generally, with structural elements
containing rotation DOFs. For that, it is necessary to include rotations DOFs in a proper way into
a contact formulation. Once this is done, structural elements can be used to represent the structure

*Correspondence to: Alfredo Gay Neto, Polytechnic School at University of So Paulo, So Paulo, Brazil.
E-mail: alfredo.neto@gmail.com
Copyright 2013 John Wiley & Sons, Ltd.

684

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

Figure 1. Example of representation of a cylindrical cross section beam with four nodes. For contact detection and description purposes, each node can be faced as a rigid sphere, which movement is described by the
node degrees of freedom.

involving such a kind of more complex contact kinematics, preserving the usual small number of
DOFs when compared to solid models.
In this work the focus will be given to the contact problem between a beam with circular cross
section and a flat rigid surface, which is a quite simple geometry. The here developed idea can naturally be expanded to more complex geometries and contact pairs, such as beam to beam, shell to
shell and beam to shell interactions.
The contact between a beam and a surface is a classical problem and can be treated using the
node to surface approach. Furthermore, it is possible to consider a line to surface description for
a list of references in classical contact formulations one can refer to Wriggers (2002) [1]. When the
nodes or lines used in the contact formulation to represent the slave points are from a beam or shell
element, it is possible to modify the gap function including the size of the element as an offset. Such
a procedure can be seen in Benson and Hallquist (1990) [2] for shell contact and was also addressed
by Wriggers and Zavarise (1997) [3] when treating the beam to beam contact frictionless modeling.
Zavarise and Wriggers (2000) [4] visited the same problem again, but assuming a frictional contact.
However, usually the moments of friction forces are not considered, assuming implicitly that radii
of contact beams are small. The inclusion of such effect in this kind of interaction is still an open
subject, yet to be incorporated in computational contact mechanics formulations.
The present work introduces a new technique to modify the definition of the gap function in a
node to surface approach. The main idea is to include in the gap function the information of the displacements and rotations of the underlying body. The aim is to represent phenomena that physically
are described by changes in contact points (such as a rolling contact) only with the displacements
and rotations of each node of the modeled bodies in contact.
Each node in a structural element can be geometrically represented as a small body. In cylindrical cross section beam elements, for example, each node can be seen as a small cylinder or, in an
approximated description, as a small radius sphere (Figure 1), equal to beam radius. For contact
description purposes it is possible to assume that the geometry of each body related to each node
of a structural element can be described as a rigid body . With that, it is possible to write for each
node an equation to describe the movement of any generic point of the assumed rigid body, only

Here, the rigid body assumption implicitly is consistent with a rigid cross section of a beam or constant thickness of a
shell assumption. To generalize this idea to more elaborated structural element theories considering cross section radius
or shell thickness variations, the same assumption of rigid body could be addressed in the first moment but, after, the
current radius of the sphere would have to be updated during the solution of the model, according to the modifications
occurred. This contribution would change the weak form of the contact theory as well as some extra contribution would
appear in tangent operator of the model.

Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

685

containing information of displacements and rotations DOFs of the node. With this equation, it is
possible to describe the movement of exterior points of each sphere, which can be assumed to be the
actual points of contact, instead of the center of the sphere (the slave point itself). Such a procedure
can be done with any shape of rigid body but, definitely the simplest choice is a sphere. In this work,
the sphere approximation will be assumed from now on.
In order to include the sphere movement in the gap function to describe the contact kinematics, it is necessary to study, first, the kinematics of each sphere in 3D space. This is done in the
next section.
2. KINEMATICS OF SPHERE
Once each node will be assumed to represent a rigid sphere, it is necessary to adopt a rotation parameterization to describe the rigid body movement. This is a classical subject and will not be discussed
deeply here. Information about the topic can be found in works from Spring (1986) [5] and Altmann
(2005) [6], regarding many possible options to describe rotations.
The present formulation assumes description of movement of the sphere using an updated
Lagrangian framework, there is, the movement will be divided into steps and, from one step to
the other, it is necessary to describe a rotation and displacement of the sphere in order to map its
movement. Figure 2 shows a scheme of this description for a given example of sphere. The idea is
to deal with small (but finite) rotations in each step and, because on that the updated Lagrangian
technique was a natural choice.
On the basis of Figure 2, Rodrigues rotation parameters were adopted for writing a rotation tensor
between the current and the next configuration (Q / (Refer to Pimenta & Campello (2001) [7] for
applications of Rodrigues parameters with beams and Pimenta, Campello, & Wriggers (2011) [8]
for shells dynamics and Gay Neto, Martins, & Pimenta (2013) [9] for models with beams involving
contact).
Rodrigues parameters are related to the Eulers parameters, which can be used to define the rotation tensor. One has to calculate a rotation pseudo-vector to evaluate Q . Eulers parameters will
be here briefly described and also their relation with Rodrigues parameters, for completeness of the
present formulation. Let   D e be the Eulers rotation pseudo-vector, which magnitude is the
0
rotation angle . The mapping of a generic vector v suffering a rotation and leading to the vector v
0
can be done using the rotation tensor Q , such that v D Q v.
The Rodrigues rotation parameter is defined by D 2 tan . =2/ and the Rodrigues rotation
pseudo-vector ( / is given by:  D e. It is possible to obtain the following expression
for the rotation tensor using Rodrigues parameters, see, e.g., Campello, Pimenta, and Wriggers

Figure 2. Updated Lagrangian description for a sphere. A framework accompanying the rigid body motion
of a sphere can suffer displacements and rotations. This scheme shows different configurations: reference
configuration, current configuration .i/, and next configuration .i C 1/. The transformation between the
current and the next configuration is referred as .
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

686

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

(2003) [10]:


1
Q D I C g./ A C A 2
2

(1)

 
4
with A D skew  and g./ D 4C
2.
It is also possible to define the angular velocity skew-symmetric tensor given by:  D
P  .t /QT .t /. The relation between P  and ! D axial ./ is given by:
Q
! D  P 

(2)

Using some algebra, it is possible to obtain the following expression to the operator  :


1
 D g./ I C A
2
Furthermore, it is useful to calculate the time derivative of  , given by:

i
h
1
P
P D  g./   P    A
2

(3)

(4)

Going back to sphere description for each node, the only important geometric parameter that
affects the model is the external radius of the cross section of the beam, which defines the size of
the sphere that will be adopted to represent the movement of each node in central axis of the beam.
Thus, the present formulation is valid for tubular cross section beams. Then, the external radius has
to be considered to define external surface of the beam from each node.
If one considers an offset radius in each point of the beam for correcting the classical gap functions of a node to surface or line to surface element, this will not be enough to represent the correct
kinematics of the problem. The only exception is when the problem is frictionless, when only that
offset is enough. Assuming that contact point is not exactly located at the centroid of the cross section leads to friction force non-null moments related to that point. This always occurs and can be
very important in rolling problems when the friction plays a role not permitting sliding. The action
line of normal force crosses the center of the considered cross section and, naturally, results in null
moment related to that pole. The friction, however, does not have the same property. Neglecting the
friction force moment does not permit a good physical description of rolling phenomenon.
To explain the basic idea of the present formulation kinematics one can refer first to a classical
(and simple) 2D problem, illustrated in Figure 3. A circular rigid body with pure sliding in a flat
surface is described in Figure 3a. In this problem, one can see that the point of contact of the body
C .i/ does not change with the movement but, once the body translates, that point moves to another
position. In the second example in Figure 3b) when there is sliding and rolling together, the situation
is completely different: the contact point in the beginning C .i/ changes into C .i C 1/, which is a
different point of the circular cross section body, once it presented a rotation given by '. Then,
aiming to describe these different contact interactions using a tangential gap function, the change
in the contact point presented in the second scenario has to be considered to describe it as different
from the first scenario.

Figure 3. Example of movement of a circular rigid body representing the cross section of a beam. (a) Pure
sliding and (b) sliding and rolling.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

687

For Figure 3a, it would be correct to say simply that tangential gap is given by the vector defined
by successive positions of C .i/ and C .i C 1/ in the plane of contact. However, for the scenario
described in Figure 3b, it is possible to adapt the tangential gap definition including the rotation
. Describing for simplicity the tangential gap as a scalar variable for the 2D problem, gap function could be calculated simply by s  r, already considering in this definition the amount of
rolling given by the radius r multiplied by the angle  and the change in the contact point total
distance, given by s.
If one has pure rolling situation, s D r, this new definition of tangential gap would give a
null result, which is correct, once the tangential gap function has to describe the amount of sliding
between two points according to its classical definition (see, e.g., Wriggers (2002) [1]).
The formulation of contact described next generalizes this basic idea for 3D rotations of a sphere
in contact with a flat surface in 3D problems.
3. CONTACT FORMULATION
First of all, it is useful to define some nomenclature used from now on in the text:










C is the point of contact between the flat surface and the sphere in configuration (i C 1).
O is the centroid of the sphere in configuration (i C 1).
f is the friction force in configuration (i C 1).
M f is the friction force moment in configuration (i C 1) related to the point O, calculated by
M f D .C  O /  f .
f is the friction force pseudo-moment in configuration (i C 1) related to the point O (energetically conjugated with the rotation parameter time derivative see Pimenta, Campello, and
Wriggers (2008) [11] for more details about pseudo-moment and its implications to the beam
formulation).
vc is the velocity of point C .
vo is the velocity of point O.
! is the angular velocity of a sphere representing a node movement as described previously.

Let P f be the power of the friction force acting in the contact between the flat surface and the
sphere. It is possible to write
P f D f  vc D f  vo C !  .C  O/ D f  vo C f  !  .C  O/ D f  vo C M f  ! (5)
T

with f D   M f . Equation 5 leads to the following weak form:


T

Wf D f  x o C   M f  

(6)

This weak form is convenient to work together with structural models. Once x o and  are
respectively the displacements and rotations of the nodes of the structure, these can be elected as
slave points from now on. Then, one can deal directly with the DOFs of the beam (or if wanted,
for a shell) when developing this weak form and its tangential operator. From now on, the pseudovector  will be represented simply by , only for a cleaner notation purpose. It remains with the
physical interpretation of rotation between configurations .i/ and .i C 1/.
The major issue to describe correctly the rolling behavior using only that beam element information is related to the kinematics of the cross section and the subsequent changes that rolling causes
in the contact point. For representing that, special defined gap functions can be constructed using
only information of the DOFs of each node of the beam to describe the contact situation. Let r be
the radius of the cross section of the beam and then defining the radius of the sphere associated
to the node of interest. Figure 4 shows the master surface (rigid and flat) with the defined normal
(n) and tangential (t1 / and (t2 / directions. The sphere oriented by the node DOFs changes is represented in configuration .i C 1/. The previous contact point of configuration .i/ C .i/ turns into
C 0 .i C1/ because on rigid body rotation of the sphere. The new point of contact is C .i C1/, defined
in configuration .i C 1/, as shown in Figure 4.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

688

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

Figure 4. Geometric description to define the tangential gap function.

The amount or rolling between configurations .i/ and .i C 1/ is given by the arc length s,
also described in the Figure 4, assuming implicitly the simplest movement history connecting
configuration .i/ to .i C 1/.
Assuming that the magnitude of rotation is small from one configuration .i/ to the next .i C 1/ it
0
is possible to write an approximation to represent a vector amount of rolling,
given
by: r  r.


0
0
Once r D rn and r D rQ n, one can write: r  .r/ D r Q  I n. Using this
expression, it is possible to quantify the amount of rolling related to the displacement of the center of the sphere from O.i/ to O.i C 1/. If one projects this vector describing the amount of
rolling in the master
possible to define
hsurface plane,
 i it would beh
i
the tangential gap simply by:


O(i+1)O.i/ C r
Q  I n  t1 t1 C r
Q  I n  t 2 t 2.
However, this information has to be propagated to the next step properly. For that, it is convenient
to define a fictitious point of initial contact between the sphere and the flat plane x f .i C1/ , which has
to be updated after each step of the updated Lagrangian scheme:
h
 i
h
 i

x f .iC1/ D x f .i / C r
Q  I n  t 1 t 1 C r
Q  I n  t 2 t 2
(7)
This fictitious point of initial contact is physically associated to the information of the total
amount of accumulated rolling that occurred until the configuration .i C 1/. The idea is to subtract this amount from the difference between the centers of the spheres in order to define tangential
gap function. Finally, with this information, one can define the so called elastic gaps for each sphere
related to the configuration .i C 1/ of an updated Lagrangian description:
g t .i C1/ D gt1.i C1/ t 1 C gt 2.i C1/ t 2

(8)



gt1.iC1/ D x o  x f .i C1/  t 1  g st.i /  t 1

(9)



gt 2.iC1/ D x o  x f .i C1/  t 2  g st.i /  t 2

(10)



gn.iC1/ D x o  x f .i C1/  n  r

(11)

The vector g st is the accumulated sliding tangential gap, which is null in the beginning g st.0/ D 0
and has to be updated when sliding occur. The name elastic gaps is related to the classical and
commonly mentioned analogy between the Coulomb friction model with sticking/slipping and the
structural material models of elasticity/plasticity. This analogy will be assumed in the present model
together with a penalty method formulation, which is also used from now on (the reader can refer
to Wriggers (2002) [1] for more information about different methodologies of imposing contact
constraints, as Penalty method, Lagrange multiplier methods, Augmented Lagrange, etc.).
All this procedure for defining the gap function is done for creating a fictitious initial point of contact, considering the possibly rolling on its composition. The only geometric approximation made
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

689

is related to the actual arc of rolling, considered as the cord of the same arc, projected in the master
surface. This is valid for small rotations in each step of the updated Lagrangian formulation. The
asymptotic tendency is to represent the actual arc of rolling, as the step between two configurations
tends to zero. The expression (7) actually does not make a special update for the axial or transversal
directions of the cylindrical beam element and, everything occurs as if the rolling was of a sphere
with a flat surface in each node of the beam. This simplifies a lot the formulation. Rotations of nodes
causing penetration due to bending could make the sphere assumption a bad choice. However, these
rotations naturally do not occur due to contact restriction in the problem, which constrains the rotations of the beam in order to keep the structure in the contact plane or above it, obeying the inequality
constraint. Thus, the sphere rigid body can be used as a good approximation for representing the
rolling of a cylinder rigid body in the contact situation.
As usual, in any Coulomb friction model, one has to check the condition of sliding or sticking.
This can be made using a tangential penalty approach. Let t be the tangential penalty stiffness and
n the normal penalty stiffness. Then, one can define the trying friction force:
f try D t .gt1 t 1 C gt 2 t 2 /

(12)

If this trying


 friction force norm is less than the maximum allowed friction magnitude, given
 slide 
by f
 D jn gn j, then one has the sticking situation. Otherwise, there is slipping and it is
necessary to update the slide accumulated tangential gap by:
g st.i C1/ D g st.i / C nT

(13)

With
 D


  slide 
1 

f try   
f

t

(14)

and the direction of sliding nT can be calculated by:


g et
nT D  .i C1/ 
 e

g t .i C1/ 

(15)

with the elastic tangential gap used to calculate the friction try force given by:
g et.i C1/ D g t .i C1/  g st.i /

(16)

Weak form and tangential operator sticking status


In sticking situation, the friction force is given by Equation (12). Then, one can obtain the following
weak form:
T

Wfst i ck D t .gt1 t 1 C gt 2 t 2 /  x o C rt   .gt 2 t 1  gt1 t 2 / 

(17)

Including the normal contribution one obtains


T

W st i ck D t .gt1 t 1 C gt 2 t 2 / C n gn n  x o C rt   .gt 2 t 1  gt1 t 2 / 

(18)

The tangent operator is given by the linearization of the expression (18):


T

W st i ck D t .gt1 t 1 C gt 2 t 2 / C n gn n  x o C rt   .gt 2 t 1  gt1 t 2 / 


T

C rt   .gt 2 t 1  gt1 t 2 / 


Each one of the contributions can be given by:

h

 i
i
h
T
Ct 1
T
T
 T r  skew Q n t 1
gt1 D ..x o / / x f .i /
t 1
Copyright 2013 John Wiley & Sons, Ltd.

(19)

(20)

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

690

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

gt 2 D

x Tf .i /

x To

gn D

Ct 2
t 2

x To


 i
h
T
 T r  skew Q n t 2

x f .i /T

Cn
n

(21)

(22)

To calculate the contribution given by   , one can use the result from Equation (4):
iT
i
h
h
T
1
1
T
P
P
P  A
P  CA
D  g./ .  /
D  g./ .  /
P
2
2
Operating in a generic vector v, one has
i
h
h
i
T
1
T
P D  1 g./  T vT P  v  P
P   v C Av
v D  g./ .  /
P
2
2
Analogously, it is possible to write a linearization that defines the operator O1 .v/.

i
h

T
T
1
   v D  g./   vT  V  D O1 .v/ 
2

(23)

(24)

(25)

It is possible to write the tangent operator in a ready to implement matrix form, separating each
contribution for each term. Once rotations of the cross section of the beam were considered, some
contribution appears related to the rotation parameters and also some cross terms between and
x o DOFs of the cross section of the beam. Once the flat surface is supposed to have no meshed
bodies, but it is defined only as a rigid body surface, its contribution will not be considered from this
point on in the tangent operator. Then, only the slave point DOFs should be altered by this approach.




x o
W st i ck D x To T K 1 C K 2 C K 3 C K 4 C K 5
(26)

With:

"
K1 D
"
K2 D

t t 1 t T1
0

#


t rt 1 t T1 skew Q n  
0

(27)

t t 2 t T2
0

#


t rt 2 t T2 skew Q n  
0

(28)


K3 D

K4 D
"
K5D

rt  

0
t 1 t T2  t 2 t T1

n nnT
0

0
0

(29)

0
0
0 rt O1 .gt 2 t 1  gt1 t 2 /

r 2 t  

h
T

(30)

#

0


i
t 1 t T2 skew Q n    t 2 t T1 skew Q n  
(31)

Weak form and tangential operator sliding status


Noting that when one has contact, the normal gap function gn becomes negative; in the sliding
situation, the friction force can be written as:
f slide D n gn nT
Copyright 2013 John Wiley & Sons, Ltd.

(32)
Int. J. Numer. Meth. Engng 2014; 97:683706
DOI: 10.1002/nme

691

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

and also the friction force moment related to the circular cross section centroid is given by:
M slide
D rn gn .n  nT /
f

(33)

Using the weak form general expression (6) with the friction force (32) and the friction moment
(33), one obtains the weak form for the status when sliding occurs, which has the contributions of
forces and moments of friction contact interaction for the slave point x o :
T

Wfslide D n gn nT  x o C rn gn   .n  nT / 

(34)

Including the normal direction, one obtains


T

W slide D n gn nT C n gn n  x o C rn gn   .n  nT / 

(35)

The tangent operator is given by the linearization of the weak form (35) and can be written as:
W slide D n gn nT  n gn nT C n gn n  x o C


T
T
T
C rn gn    .n  nT / C   gn .n  nT / C   gn N nT 
(36)


T
The linearizations necessary to calculate the tangent operator are gn ,    , and nT . The
first two are already calculated and used in the sticking status tangent operator. The last one has to
be evaluated. Such a procedure for that is showed next:
!
gt1.i C1/ t 1 C gt 2.i C1/ t 2

(37)
nT D  
gt1
t 1 C gt 2
t 2
.i C1/

.i C1/

which can be developed by (see, e.g., Zavarise and Wriggers (2000) [4] for algebraic details):
nT D 
gt1





1
 .I  nT nT /  gt1.i C1/ t 1 C  gt 2.i C1/ t 2

t C gt 2.iC1/ t 2
.iC1/ 1

with



 gt1. i C1/ t 1 D gt1 t 1


 gt 2. i C1/ t 2 D gt 2 t 2

(38)

(39)

By using the results from Equations (20) and (21) together with (38), analogously to that done to
evaluate the sticking status tangent operator, one can write for the sliding situation the matrix form
ready to implement tangent operator:




x o
slide
T
T
W
D x o
K 3 C K 6 C K 7 C K 8 C K 9 C K 10 C K 11 C K 12

(40)
with K 3 already calculated in Equation (29) and the other contributions given by:


n nT nT 0
(41)
K6 D
0
0
"
K7 D

 .n gn / 



gt1.i C1/ t 1 Cgt 2.i C1/ t 2 



#
.I  nT nT / t 1 t T1 C t 2 t T2
0

(42)

3
2




1
 .I  nT nT / t 1 t T C t 2 t T skew Q n  
0 r .n gn / 

1
2
5
gt1.i C1/ t 1 Cgt 2.i C1/ t 2 
K 8 D4
0
0
(43)
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

692

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS


K9 D

0
0
0 rn gn O 1 .n  nT /

"
K 10 D
2
6
K 11 D 4

rn  

.rn gn /   N

0
.n  nT / nT

(44)
0
0

#
(45)

3
0
7


 .I  nT nT / t 1 t T C t 2 t T
0 5

1
2

1


gt1.i C1/ t 1 Cgt 2.i C1/ t 2 

2
0

6 

T
K 12 D 4
0 r 2 n gn   N 

(46)
3
0

7




 .InT nT / t 1 t T Ct 2 t T skew Q n   5

1
2

gt1.iC1/ t 1Cgt 2.iC1/ t 2 

(47)
Equations (18), (26), (35), and (40) are ready to implement in a FEM code. Once the beam nodes
that have contact with the flat surface are set-up as slave points, these equations represent direct
contributions to force vector and tangent operator of the problem.
As usual, in contact problems numerical implementation, it is important to monitor the contact
status along the evolution of the simulation. For that, one can use the evolution of the sign of normal
gap function. If a slave point loses contact and then starts having contact again, it is important to
set up the accumulated tangential gap and fictitious initial point of contact as null, as if the contact
would have beginning at that time of contact recovering. Otherwise, a false time history of contact
force interaction could be registered.
4. NUMERICAL EXAMPLES
This section shows numerical examples with applications using the proposed contact formulation.
The first is related to transient dynamics of a very simple problem of a disc sliding and rolling in a
flat surface; the second example is a static nonlinear problem of a beam under rolling contact conditions suffering large displacements and large rotations, and the third and fourth examples are related
to transient dynamics involving impact problems. All examples were chosen to identify situations in
which the rolling or alternating rolling/sliding plays a role to demonstrate the usage of the proposed
methodology.
The beam model used in all examples is the presented by Gay Neto, Martins, and Pimenta (2013)
[9] for statics and complemented by Gay Neto, Pimenta, and Martins (2013) [12] for dynamics. For
dynamics, one could naturally apply the proposed contact formulation to the Pimenta, Campello,
and Wriggers (2008) [11] beam element. All these referenced works assume a geometrically-exact
beam formulation able to consider large rotations in 3D problems. For more details about the beam
formulation, the reader can refer to the mentioned works. Once the focus here is to discuss the
contact problem and its details, no further comments are made about the beam formulation.
It is important to emphasize that the proposed here contact modeling can be used together with
any beam theory to model the structure. The only requirement is the correct treatment of the large
rotations in a geometrically nonlinear approach.
4.1. Dynamics of a cylinder initially sliding and turning into rolling
This numerical example presents a dynamic problem of classical mechanics illustrated in Figure 5.
A rigid disc of unit length has initial angular velocity given by ! D !0 k and its center has an
initial velocity given by V 0 D V0 i . There is friction between the disc and a flat surface defined by
a friction coefficient .
A specific initial condition can be imposed such that, in the beginning, there is sliding, which
changes the angular velocity of disc and the velocity of point O along time. A sticking condition is
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

693

Figure 5. Numerical example 1: description of the problem.

then achieved and the disc starts rolling with no sliding. The analytical solution of this problem is
straightforward and can be found using the Newton Law applied for both translation and rotation of
the rigid disc. If r is the disc radius and g is the gravitational field magnitude (defined in negative
y direction), the results are that during the sliding, one can write the following expressions for !.t /
and V 0 .t /:


2g
!.t / D !0 C
t k
(48)
r
V 0 .t / D .V0 C gt / i

(49)

The kinematic condition of rolling is given by kV 0 .t /k D k!.t /k r, which is achieved in time


given by trol l :
trol l D

!0 r  V 0
3g

(50)

After start rolling, the angular velocity of the disc and the center of disc velocity remain for any
time value. With the numerical values given by !0 D 127.03, V0 D 2.355, r D 0.2,  D 0.25, and
g D 9.81, one can calculate trol l D 3.13.
A simulation was performed with the proposed contact method applied to the beam model utilized
in [9]. Only one element discretization and the appropriate boundary conditions to simulate the disc
problem were considered. For dynamics, the Newmark trapezoidal rule was utilized, as mentioned
in [12]. The time step considered is constant and equal to 1E-3. The penalty parameters considered
are n D 1E8 and t D 1E7. The amount of penetration registered is of order 1E-4. The total mass
of the considered disc in the simulation is 1E3.
Figures 69 show the time evolution of the displacement of the center of the disc, total rotation
of the disc  , magnitude of velocity of the center of the disc, and magnitude of angular velocity of
the disc. It is possible to see clearly that the proposed rolling model could represent very well the
large amount of sliding and the transition from sliding to rolling as well. It is of interest to mention
that the chosen value of tangential penalty stiffness causes very small displacements in the sticking
situation, once it is a very high value. It is almost impossible to see such an oscillation in the presented plots during the sticking steady state (Figures 8 and 9). If one diminishes this penalty value,
the condition of sticking would be relaxed and it would be possible to visualize an oscillation in
velocity of the center of the disc and in the angular velocity of the disc along time during the steady
state rolling behavior, with mean equal to the obtained by the analytical solution. Obviously, this
oscillation goes to zero if one increases the value of t , as in the results here shown.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

694

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

Figure 6. Numerical example 1: x displacement of point O as a function of time.

Figure 7. Numerical example 1: angular coordinate measuring rotation of cylinder as a function of time.

Figure 8. Numerical example 1: magnitude of velocity of the point O of the disc as a function of time.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

695

Figure 9. Numerical example 1: magnitude of the angular velocity of the disc as a function of time.

Figure 10. Numerical example 2: geometric description.


Table I. Time series for the y direction imposed
displacement at point B.
Time

y direction imposed displacement at point B

0
1
2
3
4

0
0
+5
0
+5

4.2. Cantilever beam with rolling and sliding


In this example, it was considered a quasi-static nonlinear model of a scenario of an initially straight
beam connecting the points A and B (Figure 10). The cross section of the beam was assumed to
be circular with diameter D D 0.3, and the length of the beam is L D 10. The material properties
are Young Modulus E = 1E7, Poisson ratio  D 0, and density D 1E5. The gravitational field
considered has magnitude 9.81 in negative direction of . The coefficient of friction in the interface
was assumed to be  D 0.5.
The displacements and rotations are imposed to be zero in point A (fixed point) and, in point B,
the y direction movement is imposed using the time series described in Table I. The other DOFs in B
are free, including all the rotations. Imposing this transversal (y direction) displacement in point B
causes the beam to roll and slide or alternate it, depending on the analyzed cross section and on the
amount of imposed displacement. Naturally, to make normal contact forces non-null, in a first step
of solution, the structures own weight was considered and equilibrated by normal contact forces.
This first step is considered in a time history evolution of loading from time D 0 to time D 1. Then,
in time D 1, the transversal displacements start being imposed progressively, according to Table I.
Once working with penalty method, a study of influence of penalty parameters was done in the
model. To get good results with small values of penetration in normal direction (of maximum order
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

696

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

of 1E-4), the chosen values of parameters were n D 1E9 and t D 1E7. Quadratic (three nodes)
beam elements were used in a 100 elements uniform mesh, which showed to present good results.
For a comparison, the same conditions were simulated using ANSYSTM 14 solver, however, using
solid elements to represent the beam. Then, the external surface of the whole beam was defined as
slave surface, and a defined plane flat region just below the cylinder was defined as the master
surface. The same boundary conditions and imposed displacements were considered as in the beam
model. Second order (SOLID186) elements were used in a wedge shaped elements sweep mesh (see
ANSYSTM (2012) [13] for details about the formulation). For achieving convergence, it was necessary to use the Augmented Lagrange formulation, with manual stiffness factor D 1.0 and permitting
the stiffness updating automatically in each iteration, according to the evolution of the simulation.
It is important to emphasize that this setup was not straightforward and many other parameters and
combinations with other algorithms were performed, with many difficulties to achieve convergence,
due to the extremely nonlinear behavior that this problem presents, once the contact points in the
solid model change every time when there is rolling. A good mesh quality in the external surface of
the beam was required in order to capture smoothly the changes in contact points (or at least to try to
do that). Here, the results presented that 200 divisions were considered along the length of the beam
and 32 divisions along the circumferential direction of the cross section. With this discretization, the
final mesh, which results are here shown, has 45,363 nodes and 13,825 elements. To input the own
weight and the sequence of imposed displacements in the beam tip 3385 iterations were necessary,
of course, due to the large number of sub-steps to perform the rolling contact properly. This took a
time of many hours of simulation in a 4 core XeonTM processor computer with 8Gb RAM memory.
The tendency of the presented formulation when the beam radius (and consequently the sphere
radius associated with each node) tends to zero is to go to the classical node to surface treatment,
with no rolling properly treated. A third simulation considering this assumption was performed for
comparisons with the proposed formulation in this paper, just to check the influence of the rolling
in the results. Results related to this check (r D 0) simulation are named beam model (no rolling).
The results considered using the here presented and proposed model with the rolling treated properly are simply named beam model. Both simulations solved using beam models can give results
in some seconds or no more than few minutes using the same computer 4 core XeonTM processor
computer with 8Gb RAM memory.
Figure 11 shows the deformed beam structure at time D 2.0, there is, after the first lateral displacement imposition. It is possible to see that the proposed model captured well the rolling behavior that
occurred in a large amount of the beam, giving very good agreement with ANSYSTM solution. The
beam model with no rolling could not capture correctly the scenario, once rolling plays an important
role. This is clearly shown in Figure 11.

Figure 11. Numerical example 1: deformed shape of the beam structure at time D 2.0 (for the solid model
(ANSYSTM / the central line passing through the centroids of cross sections were plot).
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

697

Figure 12. Numerical example 2: force reaction to impose the y direction displacement at point B of the
beam (fy / from time D 1.0 to time D 2.0.

Figure 13. Numerical example 2: moment of torsion in the beam structure at time D 2.0 as a function of the
reference configuration x coordinate (undeformed structure length coordinate).

Furthermore, one can check in Figure 12 the force reaction at point B in y direction because of the
imposed displacement. This can be faced as a global friction action in the problem. Again, the proposed contact model agrees very well with ANSYSTM solid model results. The non-smoothness
behavior of ANSYSTM results is due to the mesh discretization, which can never capture smooth
changes in contact points during rolling. For the proposed method, this smoothness is straightforward once the contact point changing is automatically taken into account in the definition of
tangential gap. The beam model with no rolling gives no proper results, once considered larger friction actions due to a false sliding identified instead of rolling. Actually, the rolling diminishes the
amount of friction work when compared with the no rolling assumption, even with more movement
of the whole beam.
Looking now to the comparison between the torsion moment along length of the beam between
the beam model and the beam model (no rolling), one can notice in Figure 13 that all the torsion
present comes only from the moment of friction forces, and simply neglecting these effects leads to
incorrect null torsion predictions.
Looking forward in time history results, after a cycle of movement, in time D 4.0, one can look
at the deformed structure comparison between the models in Figure 14. Again, the beam model
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

698

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

Figure 14. Numerical example 2: deformed shape of the beam structure at time = 4.0 (for the solid model
(ANSYSTM / the central line passing through the centroids of cross sections were plot).

Figure 15. Numerical example 2: x coordinate of point B of the beam (xB / as a function of time.

represented very well the rolling behavior when compared with the solid model (ANSYSTM /,
and the beam model (no rolling) could not capture the important contribution of rolling in the
cyclic behavior, giving improperly results. The Figure 15 gives the additional information of the
x coordinate of the point B as a function of time. It is possible to see clearly that the proposed
model can represent very well the solid behavior, but the beam model (no rolling) could not deal
with that.
A last plot in Figure 16 shows the hysteresis of the friction problem behavior in cyclic loading
situation. One can see the force reaction at point B in y direction due the imposed displacement
from time D 2.0 to time D 4.0. The comparison of beam model and the solid model (ANSYSTM /
is very good. The beam model (no rolling), however, could not represent well again the friction
forces, giving improperly results. Note the interesting fact that the work of friction force is larger
in this cycle than in the rolling situation, for the same imposed displacement in the tip of the
beam. This could naturally be expected once rolling is a natural way of movement that does not
dissipate energy.
Figure 17 shows some post-processing images to illustrate the movement of the beam. It is possible to see the solid model and a post-processed as solid result regarding the beam model. This can
be useful to see the regions of rolling or sliding in the model.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

699

Figure 16. Numerical example 2: force reaction to impose the y direction displacement at point B of the
beam (fy / from time D 2.0 to time D 4.0.

Figure 17. Numerical example 2: top view of the deformed structure for different time values. On left column side are the present contact algorithm solved using beam model and rendered (post-processed) as a
solid structure. On right side column are the solid model solutions made using ANSYSTM solver.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

700

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

4.3. Puck sliding on a flat surface with successive lateral impacts


This example presents a transient dynamics problem involving impacts. The idea is to represent the
movement of a cylindrical disc, named puck, on a rigid and flat surface. The surface is a squared
region with edge s. The puck radius is given by r and its movement is described using the coordinate system shown in Figure 18. Puck is initially located in the centroid of the squared region and
oriented according to Figure 18, with null velocity and null spin. It cannot roll over the surface, but
only slide, once it is laid laterally (see Figure 18 showing the top view of the proposed scenario).
The puck is initially subjected to two simultaneous time history loadings, described in Figure 19,
being one force and one moment, injecting kinetic energy in the system causing translation and
spinning to the puck. The square region is limited by some lateral surfaces, named walls, which
limit the puck movement range. Puck was modeled by one only finite element beam subjected to
boundary conditions to keep its central point in the plane x. The only free rotation is in y direction.
Interaction between puck and the flat surface was modeled just imposing the y direction movement
equal to zero in the nodes of the beam representing the puck, which represents a frictionless behavior with no possibility of status changing, there is, the puck never leaves the flat surface. The walls
were considered to be rigid and the puck interaction with them was addressed considering the here
proposed methodology for contact interaction.

Figure 18. Numerical example 3: geometric description.

Fz

My

1E4

5E3

0.1

time

0.1

time

Figure 19. Numerical example 3: loading time history in each node of the beam representing the puck. The
beam has three nodes, being the total load given by 3 times the shown in these plots.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

701

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

The numerical parameters used in the example were chosen such that both loadings have a very
short duration, going to zero prior to the first puck impact in one of the walls. Then, it is possible
to analyze the system behavior and kinetic energy variation only due to impacts occurrences. All
the results were generated using the following parameters: axial stiffness of puck EA D 1E8; bending stiffness of puck EI D 1E8; torsion stiffness of puck GJ D 1E8; mass of the puck 1E3; puck
radius r D 0.25; puck height is unitary; total simulation time 10; time-step 1E-4; normal stiffness
penalty factor n D 1E8; and tangential stiffness penalty factor t D 1E7. The maximum registered
penetration was less than 7E-3. The solver used in this example was the same used in Example 1.
Some structural parameters here described are not important for the contact method, such as the
stiffness of the puck, modeled as a beam. Actually, the beam behaves as a rigid body in this example;
however, these parameters were here provided for completeness. Once dealing with impact problems, such stiffness parameters can be important due to numeric difficulties, once high frequency
vibration modes are exited during impacts.
Two different values of friction coefficient between puck and the walls were considered:  D 0.0
and  D 0.4. The idea is to see a comparison and show that friction plays a role as a way for the
system transform kinetic energy from rotation into translation and also the opposite, and that the
proposed contact modeling can capture that.
Figure 20a shows the centroid of puck trajectory for  D 0.0 case. The disc simply stays in
a steady state straight line movement, going and returning in the same path, suffering successive
impacts in only two walls. The coordinates of puck centroid along time can be seen in Figure 21a,

1.0
0.8
0.8
0.6
0.6
0.4

0.4

0.2

0.2

0.0

0.0

-0.2

-0.2

-0.4

-0.4

-0.6

-0.6

-0.8

-0.8

-1.0
-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

(a)

-1.0
-1.0

1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

(b)

Figure 20. Numerical example 3: trajectory of puck center in the surface (a) frictionless interaction with the
walls (b) frictional interaction with the walls.
3

1
30

0.8

0.8
0.6

0.4
20

0.2
0
0

1015

-0.2
10

-0.4
-0.6

x and z

x and z

0.4

0.2
0

1.5
0

10

-0.2
1

-0.4
-0.6

0.5

-0.8

-0.8
-1

2.5

0.6

25

Time
x

-1

(a)

(b)

Time
x

Figure 21. Numerical example 3: translational and angular coordinates describing puck center and its
orientation along time (a) frictionless interaction with the walls (b) frictional interaction with the walls.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

702

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

showing that angular coordinate only increases uniformly due to the constant angular speed after
instant 0.1, when the applied loading moment My becomes zero (Figure 19). The time evolution of
velocity components of puck, such as the angular speed can be seen in Figure 22a, with a typical
alternating pattern due to the successive impacts. Kinetic energy due to puck spin remains during the
successive impacts for the whole analysis period, as one can see in Figure 23a, which also shows
2.5
2

30

1.5

25

1.5

25

20

1
20

0.5
0

15
0

10

-0.5

Vx and Vz

Vx and Vz

30

15

0.5

10

0
2

10

-0.5

10

-1
0

-1

-1.5

-1.5

-5

-2

-2

-10

-2.5

Time
Vx

(a)

Time

Vz

Vx

(b)

Vz

Figure 22. Numerical example 3: translational speeds of puck center and angular speed of puck along time
(a) frictionless interaction with the walls (b) frictional interaction with the walls.
25000
25000
20000

Kinetic Energy

Kinetic Energy

20000

15000

10000

15000

10000

5000

5000

0
0

10

Time
Total

Translation

10

Time
Total

Rotation

(a)

Translation

Rotation

(b)

Figure 23. Numerical example 3: kinetic energy evolution along time (a) frictionless interaction with the
walls (b) frictional interaction with the walls.

Figure 24. Numerical example 4: top view of the initial beam configuration. The black line represents the
trace of the vertical plane in which the beam impacts during its movement.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

703

that the total kinetic energy of the system (composed by the rotation and translation movements)
remains constant, only achieving minimum values during the impacts, when for some very small
period the normal to the contact surface component of velocity goes to zero.
Looking now at results considering  D 0.4 for the interaction between puck and the walls one
has a completely different behavior. The same initial kinetic energy is supplied to the system, but just
after the first impact, some spin turns into translational movement in x direction, because of friction
impulse during impact. Then, in other impacts, one has again the friction playing a role as a mechanism of changing the kinetic energy from spin into translation, and also the opposite. However,
always some sliding occurs during contact and a fraction of the kinetic energy is dissipated.
Figure 20b shows the trajectory in plane x, and Figure 21b shows the evolution of the coordinates
of the centroid of puck. Furthermore, Figure 22b shows the evolution of the velocity components,
clearly totally different from the frictionless example. Now, part of the spin turns into translational
speed and vice-versa during each impact. The most interesting is to analyze the kinetic energy evolution, shown in Figure 23b. Clearly there is a loss of energy in each impact occurrence, and in
the long term, it is possible to say that the system spin energy goes to zero and it stays only with
translational kinetic energy. All the dissipation is due to sliding friction occurred during successive
impacts once no other damping sources were considered in the model.
4.4. The flying beam
This example aims to show a mechanical system subjected to impacts with more than one surface,
involving essentially 3D behavior. The structure simulated is a straight beam with circular cross
Fy
1E3

1.0

time

Figure 25. Numerical example 4: loading time history applied in the center of the beam.

Figure 26. Numerical example 4: top view of the beam trajectory. In red bullets and blue crosses one can
see both beam tips centers successive positions and, at some equally spaced instants, a black bar linking the
points showing instantaneous configurations of the beam.
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

704

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

section, shown in Figure 24 by a top view. The centroid of the beam is located at coordinates
(0,0,r/, with r being the external radius of the cross section. The beam is laying on a horizontal
plane, located in D 0. There is another contact surface constructed in vertical direction and shown
in Figure 24, with which there is no initial contact.
The beam numerical data is given by: axial stiffness EA D 3.6E8; bending stiffness EI D 9.2E9;
torsional stiffness GJ D 9.2E9; shear stiffness GA D 1.8E8; mass per unit length 1.3E-3; r D 25;
and length L D 1000. Environment data is given by the following: gravitational field g D 9.81k;
the vertical contact plane contains the point (0,1000,0) and its normal direction is (-1,-1,0); friction
coefficient in all contact regions is  D 0.8; normal stiffness penalty factor n D 1E3; and tangential stiffness penalty factor t D 1E3. Simulation total time is 20 units and time step 1E-3. The
beam is uniformly meshed using 10 elements, with quadratic shape functions for displacements and
rotations DOFs.
Initially, the beam is statically subjected to the gravitational field, and the contact with the horizontal surface plays a role, only equilibrating the weight of the structure. After, a transient dynamics

Figure 27. Numerical example 4: isometric view of the beam trajectory. In red bullets and blue crosses one
can see both beam tips centers successive positions and, at some equally spaced instants, a black bar linking
the points showing instantaneous configurations of the beam.
60

1500
1250

55

1000
50

750

45

500
250

40

x and y

0
-250 0

10

12

14

16

18

20 35
30

-500
-750

25

-1000
20

-1250

15

-1500
-1750

10

-2000
5

-2250
-2500

Time
x

Figure 28. Displacements time history of beam point initially located at position (0,-L=2, r).
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

705

CONTACT BETWEEN ROLLING BEAMS AND FLAT SURFACES

1500
1250
1000
750
500
250
0
-250 0
-500
-750
-1000
-1250
-1500
-1750
-2000
-2250
-2500
-2750
-3000

60
55
50
45
40
2

10

12

14

16

18

20

35
30

x and y

simulation starts being performed, because of the application of a loading in the centroid of the
beam, according to Figure 25 time history. Because of friction effect, some angular speed appears
in the axial direction of the beam (x/. When the beam suffers the first impact with the vertical plane,
only its tip touches that plane, and the angular speed makes the beam suffer a vertical force in upward
direction, due to sliding friction in the vertical wall. This tip leaves the horizontal plane, loosing that
contact condition. The movement continues until the opposite beam tip touches the vertical plane,
and the same phenomenon is observed. After, beam starts shaking a lot due to excited modes in successive impacts. It stays alternating touching in the horizontal plane in both tips, sometimes flying
for a while, but always returning to the horizontal plane, due to gravitational field action. Figures 26
and 27 show the trajectories of both beam tip points, marked in bullets and crosses in both plots.
At some instants, a bar connecting both tips is shown to identify the orientation of the beam, for
clarifying the history of movement. Displacements time histories from both tip points and middle
point of the beam are shown in Figures 2830.

25
20
15
10
5
0

Time
x

Figure 29. Displacements time history of beam point initially located at position (0,+L=2, r).
1000

60

750

55

500

50

250
45

x and y

-250

10

12

14

16

18

2040

-500

35

-750

30

-1000

25

-1250

20

-1500

15

-1750

10

-2000

-2250
-2500

Time
x

Figure 30. Displacements time history of beam point initially located at position (0,0,r).
Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

706

A. G. NETO, P. M. PIMENTA AND P. WRIGGERS

5. CONCLUSION
A new methodology was presented to include the rigid body motion of rotation in structural elements contact scenarios. The formulation here developed applies for a beam element contact with
a flat surface; however, the same concepts could be used in more complex interactions involving
beam to beam or shell interactions with structural or solid elements. It was shown that the proper
definition of the tangential gap function could take into account correctly the rotations occurred
in the numerical examples shown, even in situations of large amount of sliding, alternating sliding/rolling, and impact problems. The model can naturally be used in static nonlinear or transient
nonlinear dynamic models. The moment of friction forces and kinematics can be addressed when
there is rolling situation. This is not possible with previous contact formulations that consider only
an offset in the normal gap due to the thickness effect of structural elements. Then, the ideas and
concepts here discussed open a field of development to be continued in many situations in order to
consider properly the rigid body rotation of structural elements in contact interactions when they
are being considered in the finite element model.
ACKNOWLEDGEMENTS

The authors acknowledge FAPESP (Fundao de Amparo Pesquisa do Estado de So Paulo) for the support
under the grants 2012/09912-0 and 2012/21167-8 (PostDoc). The second author acknowledges the support
by CNPq (Conselho Nacional de Desenvolvimento Cientfico e Tecnolgico) under the grant 301279/2009-8.
REFERENCES
1. Wriggers P. Computational Contact Mechanics. John Wiley and Sons Ltd.: England, 2002. 440 p.
2. Benson DJ, Hallquist JO. A single surface contact algorithm for the post-buckling analysis of shell structures.
Computer Methods in Applied Mechanics and Engineering 1990; 78:141163.
3. Wriggers P, Zavarise G. On contact between three-dimensional beams undergoing large deflections. Communications
in Numerical Methods in Engineering 1997; 13:429438.
4. Zavarise G, Wriggers P. Contact with friction between beams in 3-D space. International Journal for Numerical
Methods in Engineering 2000; 49:9771006.
5. Spring KW. Euler parameters and the use of quaternion algebra in the manipulations of finite rotations: a review.
Mechanism and Machine Theory 1986; 21:365373.
6. Altmann SL. Rotations, Quaternions, and Double Groups. Dover Books on Mathematics: New York, 2005.
7. Pimenta PM, Campello EMB. Geometrically nonlinear analysis of thin-walled space frames. Proc. of the Second
European Conference on Computational Mechanics, II ECCM, Krakow, 2001.
8. Pimenta PM, Campello EMB, Wriggers P. An exact conserving algorithm for nonlinear dynamics with rotational
dofs and general hyperelasticity. Part 2: shells. Computational Mechanics 2011; 48:195211.
9. Gay Neto A, Martins CA, Pimenta PM. Static analysis of offshore risers with a geometrically-exact 3D beam model
subjected to unilateral contact. Published online in Computational Mechanics July 2013. DOI: 10.1007/s00466-0130897-9.
10. Campello EMB, Pimenta PM, Wriggers P. A triangular finite shell element based on a fully nonlinear shell
formulation. Computational Mechanics 2003; 31:505518.
11. Pimenta PM, Campello EMB, Wriggers P. An exact conserving algorithm for nonlinear dynamics with rotational
dofs and general hyperelasticity. Part 1: rods. Computational Mechanics 2008; 42:715732.
12. Gay Neto A, Pimenta PM, Martins CA. Loop formation in catenary risers on installation conditions: a comparison of
statics and dynamics. Proceedings of the 32nd International Conference on Ocean, Offshore and Arctic Engineering
OMAE 2013, Nantes, 2013.
13. Ansys Inc. Mechanical APDL Theory Reference 2012. version 14.

Copyright 2013 John Wiley & Sons, Ltd.

Int. J. Numer. Meth. Engng 2014; 97:683706


DOI: 10.1002/nme

Vous aimerez peut-être aussi