Master’s Degree in Numerical Simulation

in Engineering with ANSYS

TWELFTH EDITION

Chapter 2: Contact Mechanics in 1D

Elastostatics Problems

Module: Contact Non-linearities

Ricardo Perera

Alberto Carnicero
 CHAPTER 2

Contact Mechanics in 1D Elastostatics

Problems

Ricardo Perera
Technical University of Madrid, Spain

Alberto Carnicero
ETSI-ICAI. Comillas Pontifical University, Spain

CONTACT NON-LINEARITIES
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

All rights reserved. No part of this publication may be reproduced, stored

in a database or retrieval system or published in any form or in any way
without the prior written permission of the authors.

Copyright  Ricardo Perera and Alberto Carnicero, 2020

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

CONTENTS

2 CONTACT MECHANICS IN 1D ELASTOSTATICS PROBLEMS 1

2.1. Contact conditions 1
2.1.1. Frictionless contact with a rigid obstacle: Rigid-to-flexible contact 1
2.1.2. Contact between to deformable bars: Flexible-to flexible contact 3
2.1.3. Contact with friction 3
2.2. Rigid-to-flexible contact: Finite element approximation and method of constraint
enforcement 4
2.2.1. Classical Multi-Point Constraint or Master-Slave method 8
2.2.2. Lagrange multiplier method 10
2.2.3. Penalty method 13
2.2.4. Pertubed Lagrangian method 14
2.2.5. Augmented lagrangian method 16
2.3. Flexible-to-flexible contact: Finite element approximation and method of constraint
enforcement 34
2.4. Frictional contact: Finite element approximation and method of constraint enforcement 42
2.4.1. Lagrange multiplier method 43
2.4.2. Penalty method 44
2.4.3. Perturbed Lagrangian method 47
2.4.4. Augmented Lagrangian method 48
2.5. Large deformation contact 49
2.5.1. Overview of non-linear analysis 50
2.5.2. Finite element formulaction in large displacement kinematics 51
2.5.3. Updated lagrangian formultarion 57
2.5.4. Finite element implementation of non-linear contact 58
2.6. Impact 64

Contents I
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

In this chapter, we will introduce the basic methodology and concepts required to solve contact
problems. For it, one-dimensional simple contact problems, such as the one body contact problem
against a rigid obstacle and the two deformable body contact problem, will be used. We will
concentrate initially on kinematically linear problems with or without friction. These problems will
allow establishing the framework in which the methods for enforcement of contact constraints within
a finite element analysis will be presented.
Many practical contact problems involve large deformations of the bodies that are in contact.
Because of it, the one-dimensional framework will be used too to formulate all contact relations for
finite deformations. The large deformation kinematics as well as the non-linear finite element
discretization are explained briefly with the purpose of setting in a simple way the basis on which
non-linear two- and three-dimensional contact will be developed in further chapters.

2.1. CONTACT CONDITIONS

2.1.1. Frictionless contact with a rigid obstacle: Rigid-to-flexible contact

Let us consider a contact problem consisting of a bar of length L, cross-section A and Young’s
modulus E (Figure 2.1a). The axial displacement of the bar is restricted by a rigid obstacle surface
whose distance to the bar is equal to g. This type of contact is considered to be as rigid-to-flexible
since one of the contacting surfaces, in this case the obstacle, has a much higher stiffness relative to
the deformable body it contacts, in this case the bar. The rigid surface is denoted as target surface
while the bar is denoted as contact surface. Due to the existence of the rigid surface a condition of
impenetrability should be satisfied since the motion is restricted unilaterally. According to this
condition, the axial displacement of the free end of the bar should not exceed g, e.g.
g − uN ≥ 0 (2.1)

where we have called uN to the axial displacement of the bar. The subscript N has been used to
denote that the displacement follows the normal direction to the rigid surface.

uN
g=uN
RN

L g

a) b)

Figure 2.1. Rigid-to-flexible contact

If g − u N > 0 there is a gap between the bar and the rigid plane. In this case the problem is solved
as if the obstacle does not exist. However, if g − u N =
0 contact occurs and a new constraint should
be applied to solve the problem.
Once the bar contacts the rigid surface a compressive force is exerted to the bar equal to the
reaction force RN exerted by the contact surface (Figure 2.1b). This reaction force between the

Module: Contact Non-linearities 1

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

surface and the bar is assumed to be negative since the contact pressure is by compression. In that
case we have the restriction
RN < 0 (2.2)
However, if the bar does not contact the surface we have
RN = 0 (2.3)
Conditions (2.1)-(2.3) can be grouped as follows
g − uN ≥ 0 RN ≤ 0 RN ( g − u N ) =
0
if g − u N > 0 then RN =
0 (2.4)
=
if g − u N 0 then RN < 0
The contact conditions (2.4) give only a constraint of impenetrability for the point of the bar
contacting with the obstacle. These conditions are known as Signorini conditions in Contact
Mechanics and are equivalent to the Kuhn-Tucker conditions in optimization problems.
From (2.4), it is clear that in the case of a contact problem two different states of the structural
system are possible. One is related to the open gap; in that case the problem is solved as usually
considering that no contact is present. The other is related to the closed gap; the topology of the
structure changes due to the contact and the problem to be solved is different too.

2.1.2. Contact between two deformable bars: Flexible-to-flexible contact

We will extend now the concepts developed in the previous section to the contact between two
deformable bars. Figure 2.2 shows a system of two bars, A and B, separated by a gap g. When the
bars get in contact the gap will close. If we call uA and uB to the axial displacements at the free end of
the bars located on the left and on the right, respectively, the condition of impenetrability is in this
case
u A − uB ≤ g (2.5)

Contact will occur when u A − u B = g and the contact conditions are obtained, in this case, by
replacing the displacement in (2.4) by the relative displacement. We have
g − ( u A − uB ) ≥ 0 RN ≤ 0 RN ( g − u A + uB ) =
0
if g − ( u A − uB ) > 0 then RN =
0 (2.6)
if g − (=
u A − uB ) 0 then RN < 0
where RN is in this case the compression force exerted on both bars from the moment in which
contact occurs.
x uA uB

LA g LB

Figure 2.2. Contact between two deformable bars

2 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

This type of contact is considered to be as flexible-to-flexible since both contacting bodies, in this
case the two bars, are deformable. Any of the two bars can be considered as target surface or contact
surface.
In Equations (2.4) and (2.6) two types of contact conditions appear. Firstly, the bodies cannot
interpenetrate and, secondly, the normal tractions across the contact interface cannot be tensile since
we do not consider any adhesion between the contact surfaces in the normal direction; furthermore,
the balance across the interface must be satisfied. Requirements on the displacements are refered as
kinematic conditions while requirements on the tractions are kinetic conditions; both are part of the
constraints derived of a contact problem.
Until now, in (2.4) and (2.6) we have considered static contact conditions. In dynamic analysis, the
kinematic interface conditions must be satisfied at all instances of time, requiring displacement,
velocity and acceleration compatibility between the contacting bodies.

2.1.3. Contact with friction

If, once two bars or one bar and a rigid obstacle get in contact (RN<0), a lateral force FT is applied on
a bar, this would try to slide away if none resistance force opposes to it (Figure 2.3). However, the
existence of a friction force RT at the interface, acting in the opposite direction of the applied lateral
force and tangent to the contact surface, will resist to it and the bar will stay sticked to the other bar
or surface. In that case there will not be relative tangential displacement between both bars or
between the bar and the rigid support, e.g. uT=0, and the friction force will balance the applied lateral
force to keep the equilibrium.

z
x g=uN
RN

RT
PT

Figure 2.3. Contact with friction

The friction force will continue to increase as the lateral force increases until it reaches its
maximum value RTmax. When the driving force reaches RTmax, relative sliding motion initiates in an
opposite direction to the friction force, e.g. uT ≠ 0 .
The friction force is usually described by using Coulomb’s law. According to this law, the
maximum friction force is proportional to the normal compression force RN

RT max + µ RN ≤ 0 (2.7)

where µ is the static coefficient of friction which is dependent on the material properties. The
absolute value of the friction force has been taken since the tangential force can be positive or
negative and the compression force RN is assumed negative.
According to (2.7), stick occurs when

RT max + µ RN < 0 (2.8)

while sliding occurs when

Module: Contact Non-linearities 3

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

RT max + µ RN =
0 (2.9)
In this case, by analogy with (2.4) and (2.6), the Kuhn-Tucker conditions might be formulated as
uT ≥ 0 RT max + µ RN ≤ 0 uT ( RT max + µ RN ) =
0
if uT > 0 then RT max + µ RN =
0 (2.10)
=
if uT 0 then RT max + µ RN < 0
The condition that the two bodies are in contact at a point implies that the normal reaction force
RN ≤ 0 . Therefore − µ RN is always positive.

2.2 RIGID-TO-FLEXIBLE CONTACT: FINITE ELEMENT APPROXIMATION AND

METHOD OF CONSTRAINT ENFORCEMENT

As it is well known, the weak formulation of a physical problem, which agrees with the Principle of
Virtual Work in mechanical analysis, represents the starting point to carry out an approximation by
using the finite element method. Its general expression for an axial bar (see Eq.(2.18) of module
“Fundamentals and Applications of Finite Element in Mechanical Analysis” is as follows
L L

∫ w, x E (=
0
x)A(x)u, x dx ∫ w( x)b(x)dx + w( x) P( x)
0
Γ
(2.11)

where w is virtual displacement, u is the real displacement of the bar, b(x) is a distributed loading per
unit length, P is a concentrated load, E is the Young’s modulus and A is the cross-section of the bar.
As shown in the previous section, the contact problem is a constrained problem according to the
Kuhn-Tucker conditions. Therefore, the weak formulation developed in Chapter 2 of the module
“Fundamentals and Applications of Finite Element in Mechanical Analysis” would be identical for
bodies in contact except that the contributions to the virtual work from the contact interface should
be added. For it, the boundary Γ is decomposed into three parts. Parts of the boundary, Γu and Γt, are
subjected to prescribed displacements and tractions, respectively, but, furthermore, a new contact
boundary Γc has been included. This contact interface consists of the two physical surfaces of the
two bodies which are in contact, but since they are theoretically coincident we refer to a single
interface Γc. Γu and Γt will never be part of the contact boundary and, therefore, will never be in
contact with the rigid surface or any other body. However, the contact boundary Γc will be in contact
with the surface or other body and on it no displacement nor force is prescribed. In a dynamic case,
this boundary can either be in contact with the obstacle or not along the time.
If we consider firstly the frictionless contact with a rigid obstacle seen in Section 2.1.1, contact
restrictions are given by (2.4). Therefore, (2.11) is rewritten as follows
L L

∫ wN , x E ( x) A( x)u=
0
N , x dx ∫w
0
N ( x)b( x)dx + wN P Γt
+ wN RN Γc
(2.12)

where RN is the reaction force between the surface and the bar once contact occurs, assumed to be
negative, and wN is the virtual displacement in normal direction to the rigid surface. The last term of
(2.12) gives the contribution of the Principle of Virtual Work to the kinetic conditions on the contact
interface. The remaining terms in (2.12) give the balance of the bar before contact occurs. Therefore,
Equation (2.12) represents the equilibrium of the bar including the reaction force when contacts with

4 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

the rigid surface. As usually, we have chosen the virtual displacement to satisfy the homogeneous
boundary condition in the essential boundary Γu.
As is well known, the approximation of (2.11) using the finite element method gives a system as
follows
[ K ] {d } = { F } (2.13)

where [ K ] is the stiffness matrix, { F } is the load vector and {d } is the nodal displacement vector.
The stiffness matrix [ K ] is obtained from the assembly of the contribution of each particular
element e
[K ] ≡ k
e e
ij = ∫ψ i ,x EAψ j ,x dx i , j = 1,...., nne (2.14)
Ωe

where nne is the number of nodes per element, ψ i are the finite element shape functions and  K e  is
the stiffness matrix for element e. In Equation (2.14) the subscripts i and j are referred to the local
numbering of the nodes of the element whose correspondence at the global level is obtained from the
connectivity table.
Similarly to the stiffness, the load vector {F } is obtained from the element contributions which
are defined as follows
{F e } ≡=
fi e ∫ ψ i bAdx +ψ i P Γ =i 1,..., nne
e
(2.15)
Ωe

where, as usually, with the purpose of reaching a higher systematization in the procedure, we have
assumed initially that there are not essential boundary conditions since they are applied once the
system of equations is formed.
Since bar finite elements have been used, the contact between a bar and a rigid surface can be
considered as a node-to-surface or node-to-segment contact. In case of having a structural system
formed by several bars, [K ] and {F } in (2.13) would be the result of the assembly of all the bars of
the system.
If we start now from the weak formulation of the contact problem (Equation (2.12)), the
equivalent system to (2.13) when contact occurs takes the form
 {d } 
([ K ] [E] )
t

 RN 
 = {F } (2.16)

where [ E ] is a matrix of constants used to introduce the contact constraints and whose structure will
t

depend on each particular problem and the type of contact. In the particular case of contact of a bar
with a rigid obstacle it will have one row and as many columns as degrees of freedom are defined for
the structure. All the columns will be zero except that corresponding to the degree of freedom of the
structure contacting with the surface.
In summary, the governing equations to be solved for the one-dimensional rigid-to-flexible contact
problem are the Principle of Virtual Work including the effect of the contact reaction through an
externally applied unknown force (Equation (2.12)), plus the kinematic constraint equation included
in (2.4).
Various approaches have been proposed for enforcing the kinematic contact constraints in finite
element analysis. Most of these approaches can be considered as Multi-Point Constraint methods
since their main purpose is the enforcement of the contact constraints in all the active contact zones.
The most commonly considered are: a) the classical Multi-Point constraint method or Master-Slave

Module: Contact Non-linearities 5

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

method; b) Lagrange multiplier method; c) the penalty method; d) the augmented Lagrangian method
and e) the perturbed Lagrangian method.

EXAMPLE 2.1 Due to symmetry reasons the nodes A, B, C and D of Figure 2.4 are constrained to have the
same vertical displacement. Compute the matrix [ E ] that defines these constraints.

Figure 2.4

Solution: The equations used to define the constraints of the problem are
d Az − d Bz =
0
d Az − dCz =
0 (a)
d Az − d Dz =
0
where z is the global vertical axis.
We can write (a) in matrix form as follows
 d Az 
1 −1 0 0    0
   d Bz   
1 0 −1 0   d  = 0 (b)
1 0 0 −1  Cz   0 
  d  
 Dz 
Therefore, the matrix of constraints of the problem is
1 −1 0 0 
=[ E ] 1 0 −1 0  (c)
1 0 0 −1
 

EXAMPLE 2.2 The height of a beam is 2h and is modeled with three nodes, A, B and C (Figure 2.5).
Compute the constraint matrix of the problem which tie the horizontal displacements of the three nodes
and the rotation of the neutral axis (node B)

6 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.5

Solution: The relationships among the displacements at the top and at the botton of the section and the
displacement and rotation in the middle point B are:
d Ax − d Bx + hθ B =
0
(a)
dCx − d Bx − hθ B =
0

where x is the horizontal global axis.

Rearranging (a) in matrix form we have
 d Ax 
 
 1 −1 h 0   d Bx   0 
  =  (b)
 0 −1 − h 1   θ B   0 
 
 dCx 
Therefore, the constraint matrix is
 1 −1 h 0 
[E] =  0  (c)
 −1 −h 1 

EXAMPLE 2.3 The column shown in Figure 2.6 is rigidly joined to a 2D solid. Once a finite element
discretization is carried out the column appears to be joined to a plane element through the middle point of
the edge 1-2 of length 2b. Compute the matrix [ E ] that allows introducing the restriction in the finite
element formulation.

Module: Contact Non-linearities 7

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.6

Solution: The column is discretized with beam elements of three degrees of freedom per node. Each node
of a plane element has two degrees of freedom. The relation among the horizontal and vertical
displacements of nodes 1, 2 and 3 can be easily written as
d1x = d 2 x
d 2 x = d3 x
(a)
d=
1z d 2 z − bθ 2
d=
3z d 2 z + bθ 2
Equation (a) can be written in matrix form as
 d1x   0 
   
d1 y   0 
1 0 −1 0 0 0 0 
   d2 x   0 
0 0 1 0 −1 0 0    
 d2 y  =  0  (b)
0 1 0 −1 0 0 b 
  d3 x   0 
0 0 0 −1 0 1 −b     
 d3 y   0 
 θ  0
 2   

Therefore the matrix [ E ] is

1 0 −1 0 0 0 0
 
0 0 1 0 −1 0 0
[ ]  0
E =
1 0 −1 0 0 b 
(c)
 
0 0 0 −1 0 1 −b 

8 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

2.2.1 Classical Multi-Point Constraint or Master-Slave method

With this method, constraint equations are added internally to tie the displacements between
contacting surfaces. Basically, the degrees of freedom of the contact surface are eliminated and set to
be dependent on those of the target surface through constraint equations.
To carry out this, when the Master-Slave method is used, the degrees of freedom involved in the
contact constraints are divided in two groups, masters and slaves. Slave degrees of freedom are then
removed, so the updated equation system does not have slave degrees of freedom. This technique
was used in Chapter 2 of the module “Fundamentals and Applications of Finite Element in
Mechanical Analysis” as a means to introduce boundary conditions in the system derived from the
application of the finite element method when in one structure the supports are rotated with respect
to the global coordinate system.
The implementation of the Master-Slave method requires the classification of the degrees of
freedom of the structure in three types, unconstrained, masters and slaves. The first ones are those
which do not appear in any of the multi-point constraints. We label these three sets as {du } , {d m }
and {d s } , respectively, and the partition of the stiffness matrix according these labels takes the form

 [ K uu ] [ Kum ] [ Kus ]   {du }   {Fu } 


 [ K mm ] [ K ms ]   {d m }  =  {Fm }  (2.17)
 symm.
 [ K ss ]   {d s }   {Fs } 
In general form, a multi-point constraint can be written as follows
[ C m ] {d m } + [ C s ] {d s } =
{C} (2.18)
If we solve (2.18) for the slave degrees of freedom we have

{d s } = [Cs ] {C} − [Cs ] [Cm ]{d m } ≡ [T ]{d m } + { g}

−1 −1
(2.19)

where we have assumed that matrix [Cs ] is square and nonsingular. Replacing (2.19) into Equation
(2.17) we obtain
 [ K uu ] [ Kum ] + [ Kus ][T ]   {d u } 
  =
 symm [ K ] + [T ]T [ K ]T + [ K ][T ] + [T ]T [ K ][T ]   {d m } 
 mm ms ms ss 
(2.20)
 {Fu } 
= T 
 [T ] { F } − [T ] [ K ]{ g} + { F } − [ K ]{ g} 
T
 s ss m ms 
As we can observe this procedure is the same used in the module “Fundamentals and Applications of
Finite Element in Mechanical Analysis” to impose displacements in a structure.
The main advantages of this method are that it is exact and easy to understand, retains the
symmetry and positive definiteness of the system matrix. Their main disadvantage is that it requires
user decisions, although nowadays the user interfaces and internal algorithms have released user
most of these tasks. Furthermore, it is restricted to linear constraint.
This formulation is available in ANSYS Workbench and is especially recommended for linear
contact types such as Bonded (Two faces or edges coupled together both in their tangential and
normal directions) and No Separation (Two faces or edges coupled in their normal direction only;
only a small sliding is allowed in tangential direction) since it is a direct and efficiency formulation.

Module: Contact Non-linearities 9

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

EXAMPLE 2.4 In the one-dimension finite element discretization shown in Figure 2.7 introduce the
constraint d=
3 d=
4 d 5 using a Master-Slave method.

Figure 2.7

Solution: The application of the finite element method will originate a system as
 K111 K121 0 0 0 0   d1   F1 
    

1
K 22 + K 222 K 232 0 0 0   d 2   F2 
 K 332 + K 333 K 343 0 0   d3   F3 
   =   (a)

3
K 44 + K 444 K 454 0   d 4   F4 
 K 554 + K 555 K 565   d 5   F5 
 5 
    
 Symmetry
 K 66   d 6   F6 
where the superscripts are referred to the number of each element and the subscripts denote the number of
the nodes.
If we consider the displacement d3 as a master degree of freedom we can write the constraint equations as
follows
 d1  1 0 0 0
   
 d2  0 1 0 0   d1 
 d3  0  
0   d2 
[T ]{d}
0 1
=    =  ⇒ {d } (b)
 d4  0 0 1 0   d3 
d  0  
0 1 0   d6 
 5   
 d6  0 0 0 1 

Replacing (b) in (a) and premultiplying by [T ] the following system is obtained

T

 1
K11 1
K12 0 0   d1   F1 
    

1
K 22 + K 22
2 2
K 23 0   d2   F2  (c)
=
 K 332 + K 33
3
+ 2 K 34
3
+ K 44
3
+ K 44
4
+ 2 K 45
4
+ K 554 + K 55
5 5 
K 56  d3   F3 + F4 + F5 
 
5    
 Symmetry K 66   d6   F6 

2.2.2 Lagrange multiplier method

The constraint equations in case of rigid-to-flexible contact are given by (2.4). According to this
method, the Lagrange multiplier λN can be interpreted as the reactive force RN due to the contact.
10 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Therefore, its effect has been already included in the weak formulation through the contact surface
Γc (Eq.(2.12)). However, a new equation should be introduced with the purpose of considering also
the kinematic constraint g − u N =0 . For it, we use a virtual Lagrange multiplier λNw such that

λNw ( g − u N ) =
0 (2.21)

As λN is an arbitrary constant, the fulfillment of the kinematical constraint equation is guaranteed.

w

Once the finite element discretizacion is carried out, equation (2.21) can be written in the form

[ E ]{d } = {Q} (2.22)

where {Q} is a vector containing constants needed to enforce the kinematic contact constraints. In
( )
this case it agrees with the gap {Q} = g and [ E ] was defined previously.
If (2.16) and (2.22) are combined, we obtain the modified finite element equations by the
Lagrange multiplier method
[K ]  { F }   [ K ] [ E ]   {d }   { F } 
[ E ]   {d }  =
t t

    ⇒  =  (2.23)


 [E] [0]   RN   {Q}   [ E ] 0   λN   g 
where we have used the notation λN to denote the reactive force RN. As commented above, it is clear
from (2.23) that multiplier λN may be interpreted as a force due to the constraint introduced by the
contact. For this particular case, λN is a scalar since there is only one active contact constraint
because the contact is carried out only on a node of the bar. However, in general, λN will be usually a
vector containing as many Lagrange multipliers as there are active contact constraint equations.
As observed in (2.23) the system matrix changes its size or form with active contact constraint
equations when compared to the case with inactive constraints (Equation (2.13)). The matrix is still
symmetric but it is singular because of the zeros on the diagonal. The system should be solved by
using a solver able to handle zero diagonal elements.

EXAMPLE 2.5 The Figure 2.8 shows a bar with a load at the free end. The bar has section of 15 cm2 and is
made of steel (E=2.1·108 kN/m2). On the right side of the bar, at a distance of 1 mm, there is a rigid wall.
Compute using the Langrange Multiplier method the evolution of the displacement and the reaction force
with the load P.

2000 1

Figure 2.8

Solution: The problem has only one degree of freedom. The finite element equation to solve it is given by
AE
d1 = P (a)
L

Module: Contact Non-linearities 11

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Therefore the displacement is

PL
d1 = (b)
AE
This solution is right only if d1 ≤ g being the reaction equal to zero in this case. However, when d1 = g ,
contact occurs and we have one active contact constraint. In this, case (2.22) might be written as
follows
d1 = g (c)
where
[E] =
−1 {d } =
d1 {Q} =
−g (d)

In the same way, from (a) and (d) Equation (2.16) can be written as

 AE  d 
 −1  1  =P (e)
 L   λN 
where λN ≤ 0 is the compressive normal reaction.
Hence, combining (c) and (e) the system (2.23) results
 AE 
 L −1  d1   P 
=
  λN   − g 
(f)

 −1 0

where, by consistency with the global axis we have assumed that λN is positive, i.e., it has the same sense
that P. It makes that some signs in (2.23) are changed, since there it was assumed to be negative. If, after
solving (f) we obtain λN ≥ 0 , it will confirm the previous assumption.
From (b), by replacing the parameters with their values we get the value of the displacement in case of the
rigid obstacle does not exist
2000 ⋅ P
d1 = [mm] (g)
1500 ⋅ 2.1 ⋅ 102
If contact occurs then d1= g= 1 [mm] and (f) becomes

 g   1 
 d1     
= = (h)
   AE  1500 ⋅ 2.1 ⋅ 10 1 − P 
2
λ
 N  g − P 
 L   2000 
Figure 2.9, shows the evolution of the displacement and reaction force with the load P. As we can see the
reaction is zero and the displacement linear while there is not contact and the reaction increase linearly and
the displacement remains constant once the contact is reached.

12 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.9

2.2.3 Penalty method

In the penalty method, the impenetrability constraint g − u N = 0 is imposed as a penalty normal
traction along the contact surface. Therefore, in the case of the bar of Section 2.1.1, Eq. (2.12) should
be reformulated as follows
L L

∫w
0
N ,x E ( x) A( x)u=
N , x dx ∫w
0
N ( x)b( x)dx + wN P Γt
+ wN β ( g − u N ) Γc
(2.24)

where β is a large arbitrary parameter known as the penalty parameter, which can be interpreted as a
spring stiffness in the contact interface between the bar and the rigid surface. With this weak form
the impenetrability constraint is satisfied only in an approximate way. The level of penetration
depends upon the penalty parameter. Only when β → ∞ this constraint is fulfilled. That means the
penalty spring stiffness is infinitely large and, therefore, no penetration occurs.
The assumption of β → 0 only is valid for inactive contact constraint. In case of contact it would
involved a high penetration.
When this method is used, the modified finite element equations take the form

( [ K ] + [ E ] β [ E ] ) {d } =
t
({F} + [ E ] β {Q})
t
(2.25)

In the limit, when β → ∞ , (2.25) gives the exact solution obtained with the Lagrange multiplier
method.
In this case, the normal reaction force is computed as
R=
N λ=
N β ( g − uN ) (2.26)
In comparison with Lagrange multiplier method, penalty method does not introduce new
variables, is easier to implement and is quite widely used. Its implementation is as easy as assigning
a high value to an element already existing in the structure although the choice of the penalty number
should be within an allowable range; it should be large enough to be effective but not so large as to
provide numerical problems derived of systems poorly conditioned. However, this method may

Module: Contact Non-linearities 13

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

significantly increase the bandwidth of the system of equations and, unlike Lagrange multiplier
method, the solution is approximate, e.g., some interpenetration occurs.

EXAMPLE 2.6 Using the penalty method, compute as a function of the load P, the displacement at the free
end and the reaction force of the bar of Example 2.5. Assume that the penalty parameter, β , takes values
of 104, 105 and 108 kN/m; compare the solutions.
Solution: Under non contact conditions, the displacement is given by the Equation (b) in Example 2.5.
Under contact condition a penalty parameter has to be added. This is equivalent to introduce a spring
between the wall and the end of the bar when the bar ‘goes through’ the wall (Figure 2.10)

β
P
2000 1
d1

Figure 2.10

Therefore, the equilibrium given by Figure 2.10 can be expressed as

AE
P − β ( d1 − g )
d1 = (a)
L
that can be written as
 AE 
 + β  d1 =
P+βg (b)
 L 
That is the equivalent expression of Equation 2.21 where

=[ E ] 1={d } d1 =
{Q} g (c)
Hence, under contact condition the displacement is given from (a) by the following equation:
P+βg P + β ·1
=d1 = (d)
+ β 1500 ⋅ 2.1 ⋅ 10 + β
2
AE
L 2000
The reaction force can be computed following Equation 2.26. Figure 2.11 plots the three cases solved for
the three different values of the penalty parameter. From this figure, we can see how important the choice
of the value of the penalty parameter is and how far the solution might be of the real solution if we do not
choose this parameter appropiately.

14 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.11

2.2.4 Perturbed Lagrangian method

By using this method the zero diagonal elements appearing in the Lagrange multiplier method are
removed. Its formulation can be considered as a combination of both penalty and Lagrange
multiplier methods. To apply it, Eq. (2.21) is reformulated as follows
 λ 
λNw ( g − u N ) + N  =
0 (2.27)
 κ 

where κ is a constraint parameter to prevent penetration. With this formulation the term g − u N is
regularized by λN κ , which can be considered as a perturbation of the Lagrange multiplier. The
parameter κ, analogous to a penalty parameter, is chosen to be high compared with the terms of the
stiffness matrix. Therefore, the perturbation term is small. If κ → ∞ , Eq.(2.27) yields the classical
Lagrange multiplier method and, therefore, the exact solution is obtained. However, for finite values
of κ, we obtain an approximate solution and the impenetrability conditions are only met
approximately.
Once carried out a FE discretization, Equation (2.27) can be written in matrix form as follows
1
[ E ]{d } + {λ} =
{Q} (2.28)
κ
In this particular problem {λ} and {Q} are scalars, e.g., {λ} = λN and {Q} = g .
If we combine now (2.16) and (2.28) we will obtain
Module: Contact Non-linearities 15
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

[K ]  { F }   [ K ] [ E ]   {d }   { F } 
[ E ]   {d }  =
t t

    ⇒  =  (2.29)


 [E] [1 κ ]   RN   {Q}   [ E ] 1 κ   λN   g 
As is observed, the zero in the diagonal has been removed, but, as in the penalty method, the
solution is now only approximate.
In this method, the formulation of contact surface conditions is almost identical to those obtained
from the penalty method. So the perturbed Lagrangian can be considered as a penalty method in
disguise.

EXAMPLE 2.7 Solve the Example 2.5 using a Perturbed Lagrangian method. Solve the problem when the
parameter κ takes values of 105, 106, 107 and 108 m/ kN. Compare the results.
Solution: Under non contact conditions, the displacement is given by equation (b) of Example 2.5. Under
contact conditions the equation system that allows computing the displacement and reaction force is given
by equation (2.29) and can be written, by analogy with Equation (f) of Example 2.5, as

 AE 
 L −1
 d1   P 
   =   (a)
 −1 1   λN   − g 
 
 κ
Replacing the parameters with their particular values we have
 1500 ⋅ 2.1 ⋅ 102 
 −1 d
2000   P
  1  =   (b)
 1   λN   −1
 −1 
 κ
Figure 2.12 shows the plots of the different solutions. Again, as in Example 2.6, the parameter κ has a big
influence in the performance of the solution and its behavior is similar to a penalty parameter. The larger
this parameter is the better is the computed solution. In fact, we can get absolutely wrong solutions if the
parameter is not high enough as it can be observed in the first graphs of Figure 2.12.

Figure 2.12
16 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

2.2.5 Augmented Lagrangian method

This method is a combination of the penalty method with the Lagrange multiplier method. It is stated
as a Lagrangian method of the constrained problem but incorporating an additional penalty term (the
augmentation) whose purpose is improving the Lagrange multiplier problem.
On one hand, the kinematic constraint g − u N = 0 is enforced to be satisfied by using an equation
identical to (2.21) in the Lagrange multipler method. But, on the other hand, a penalty term similar to
the last term used in Eq.(2.24) is also incorporated in the weak formulation of Eq.(2.12). We have
L L

∫ wN , x E ( x) A( x)u=
0
N , x dx ∫w
0
N ( x)b( x)dx + wN P Γt
+ wN RN Γc
+ wN β ( g − u N ) Γc
(2.30)

From (2.30) the finite element equations take the form

( [ K ] + [ E ] β [ E ] ) {d } =
t
({F} + [ E ] β {Q} + [ E ] R )
t t
N (2.31)

Now, if (2.22) and (2.31) are combined we obtain the modified finite element equations by the
augmented Lagrange multiplier method
 [ K ] + [ E ]t β [ E ] [ E ]t   {d }   { F } + [ E ]t β {Q} 
 =   ⇒
 [ E ] [ 0 ]   RN   { Q } 
   
(2.32)
 [ K ] + [ E ] β [ E ] [ E ]   {d }   { F } + [ E ]t β {Q} 
t t

  =
 [ ]   λN   
 E 0   g 
When β → 0 the exact solution is obtained.

EXAMPLE 2.8 Solve Example 2.5 using an Augmented Lagrangian method taking as values of the
paremeter β 105, 106 and 108 kN/m.

Solution: In this case, system (2.32) becomes

 AE 
 L +β −1  d1   P + β g 
=
  λN   − g 
(a)
 
 −1 0

which, when solved, gives

 g   g 
 d1     
=  =  AE    1500 ⋅ 2.1 ⋅ 10 + β 1 − ( P + β ·1) 
2
(b)
 λN    + β  g − ( P + β g )    
 L    2000  
Figure 2.13 shows the plots of displacement and reaction for the three different values of β .

Module: Contact Non-linearities 17

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.13

EXAMPLE 2.9 Solve Example 2.5 using ANSYS.

Solution: One of the main problems to solve this simple example with ANSYS Workbench is the
definition of the rigid contact surface; for it, in this case we will model this area by using a rigid behaviour
block.
Firstly, in SpaceClaim, a sketch is created on a XY plane by clicking Select New Sketch Plane button at the
bottom of the screen (Figure 2.14).

Figure 2.14

18 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Now, the line is built using Line tool, included in the Sketch group of the ribbon bar (Figure 2.15). Sketch
Mode view (Mode group of the ribbon bar) should be selected to work in 2D.

Figure 2.15

Once we have defined the line, the cross section (15 cm2) of the bar should be specified. For it, go to the
Prepare tab and click on Profiles in the Beams group. A profile must be assigned to the line of the model.
The profile is directly specified as a Rectangle. To assign the cross section to the line, click on Create in
the Beams group and select the line of the model.
Now, the dimensions of the rectangular cross section must be defined. For it, in the Structure tree of
SpaceClaim, right-click on Rectangle and select Edit Beam Profile to introduce the width (B=0.03873 m)
and height (H=0.03873 m) of the rectangle (Figure 2.16).

Figure 2.16
Module: Contact Non-linearities 19
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

The next step is model the rigid contact zone. For it, in Sketch Mode view (Mode group of the ribbon bar),
we build a square of unit edge using Rectangle tool in the Sketch group of the ribbon bar (Figure 2.17). It is
important to ensure that a gap of 1 mm exists between the square and the left end of the line.

Figure 2.17

Now, using Extrusion option from Pull tool (Edit group of the ribbon bar) the 3D solid is built.

Figure 2.18

20 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Once both bodies are created, we open Mechanical module and define the mesh, the boundary conditions
and the contact properties. First of all, we change the behaviour of the solid body from flexible to rigid
(Stiffness Behavior = Rigid) in the details window of Geometry > Solid in the Model tree (Figure 2.19).

Figure 2.19

In the next step, we define the boundary conditions on the bar, that is, we constraint the axial displacement
at one end and apply a force of 300 kN force at the other end. Next, we fix the displacement of the block.
As the behaviour of the block is rigid we cannot fix directly the displacement on the block since it is not
allowed. However, we can fix a remote displacement, so we choose one of the faces of the box and apply a
remote displacement equal to zero at each translational or rotational direction (Figure 2.20).

Figure 2.20

Once both bodies are fixed, we define the solid-line contact by right-clicking Connections in the Model
tree and selecting Insert > Manual Contact Region (Figure 2.21).

Module: Contact Non-linearities 21

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.21

By default, the type of contact is bonded, i.e., the bodies are coupled together in tangential and normal
directions. The characteristics of the contact are showed in Figure 2.22. This window allows defining the
contact-target pair. In this case the contact is defined as the end point of the line and the target is the rigid
block (1). Once the pair is defined, the field (2) is automatically filled in, giving information about the kind
of geometric entities involved in the contact. In (3) we define the contact model; in this case we will use a
frictionless model instead of bonded. Finally, and within the Advanced group we choose the numerical
model to be used for solving the contact problem; in this case we have selected a penalty method (Pure
Penalty) (4) with a penalty stiffness parameter (Normal Stiffnesss Factor) equal 100 (5).

Figure 2.22

22 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Finally, before solving the problem, some characteristics of the solution procedure should be set in the
details window of Solution > Analysis Settings (Figure 2.23). First of all, the load should be specified to be
applied in some substeps since the problem is nonlinear and, additionally, all the output controls should be
turned on in order to compute all posible results.

Figure 2.23

Once the problem has been solved we can check the results. For example, the displacement is plotted in
Figure 2.24. As we can see in this figure, the number of substeps used in this example might be reduced
dramatically keeping good results.

Figure 2.24

Module: Contact Non-linearities 23

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

In the same way, the reaction force can be plotted by using the option Solution >Insert > Probe > Force
Reaction in the Model tree (Figure 2.25).

Figure 2.25

As we can check both displacement and reaction forces are in good agreement with the theoretical results
obtained in the previous examples.

PROBLEM 2.1 In the previous example change the characteristics of the contact formulation and compare
the results.

EXAMPLE 2.10 The bars of the truss structure shown in Figure 2.26 have a section of 15 cm2. The
structure is 1 mm over a non frictional rigid foundation. Compute using the Langrange Multiplier method
the value of the reaction force.

24 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.26

Solution: By using the finite element method with two elements, the system (2.13), derived from no
contact problems, becomes once applied the essential boundary conditions
 K11 K12  d1x   P1x 
   =   (a)
 K 21 K 22  d1z   P1z 

Considering the geometry of the structure and the mechanical and geometrical properties of the bars the
displacements at the free node are computed as
−1
 d1x   59803 6883.1  60  −3  1.1638 
 =     =  10 ⋅   (b)
 d1z   6883.1 77443   −100   −1.3947 
Taking into account that loads and displacements are considered negatives if their sense is downwards, the
contact constraint for this problem is given by

( g + d1z ) ≥ 0 ⇒ d1z > −1 mm (c)

Then, solution (b) violates the contact restriction and, therefore, the problem should be rewritten in terms
of Equation 2.23. We have
 59803 6883.1 0  d1x   60 
    
 6883.1 77443 −1 d1z  =
 −100  (d)
 0 0    −3 
 −1  λN  1 ⋅ 10 
where

[E] =
(0 − 1) {d } =
( d1x d1z ) {Q} =
t
g (e)
In this example compressive normal reaction force is positive since it follows the positive sense, i.e.,
upwards.
Solving the equation system the solution is:

 d1x  1.118 ⋅ 10−3 m 

   −3 
 d1z =  −1 ⋅ 10 m  (f)
λ   30.255 kN 
 N  

Module: Contact Non-linearities 25

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

EXAMPLE 2.11 Solve Example 2.10 using a penalty method and compare the solution by using different
values of the penalty parameter.
Solution: To solve this problem using a penalty approach, we have to modify the stiffness matrix and the
load vector by using Equation (2.25) and Equation (e) of the previous example. Therefore, instead of
system (d) of Example 2.10, we have the following system
 59803 6883.1   d1x   60 
   =   (a)
 6883.1 77443 + β   1z  
d −100 − 1 ⋅ 10 −3
⋅ β 
The solution is a function of the penalty parameter. Table 2.1 shows the value of the displacement and the
contact force for differente values of β
As the table shows the larger the value of the penalty parameter is the closer the solution is from the real
one. For the smallest values the difference between both can become considerable.

Table 2.1

β d1x [mm] d1z [mm] Rz [kN]

103 1.1632 -1.3896 0.38963
104 1.1586 -1.3492 3.4916
105 1.1381 -1.1713 17.127
106 1.1216 -1.0281 28.101
107 1.1187 -1.003 30.025
108 1.1184 -1.0003 30.232
1020 1.1184 -1 21.684

EXAMPLE 2.12 Solve Example 2.10 using ANSYS Workbench.

Solution: First of all, we define the geometry. It will require using two different bodies for the two lines
defining the bars and another body for the rigid obstacle. The lines will be defined from their end points.
As in Example 2.9 a block will be used as rigid foundation. The block will be 1 mm far away from the
lower vertex of the structure. Figure 2.27 shows the geometric model

26 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.27

Once the geometric model is completed, ANSYS Mechanical is opened to start working in the definition of
the finite element model.
As in Example 2.9, the solid block is defined with rigid behavior (Stiffness Behavior = Rigid) and the two
lines as Link/Truss elements. As the two lines defining the bars are uncoupled, we join them using a joint
connection (Connection > Insert > Joints at the Model tree) (Figure 2.28).

Figure 2.28

The used joint, defined as Revolute, connects all degrees of freedom of both vertices except z-rotation
(Figure 2.29). Although rotations are not considered as degrees of freedom in link elements, with Revolute
Module: Contact Non-linearities 27
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

joint all the translational degrees of freedom associated to common nodes of link elements are connected.
The alternative of using the option End release to model the joint between both bars cannot be considered
because it is incompatible with the contact definition.

Figure 2.29

Next step is the definition of the contact pair between the common vertex of the lines and the rigid
foundation. We will add a manual contact region (Connections > Insert > Manual Contact Region in the
Model tree) (Figure 2.30) and define the contact between the vertices (we will choose both of them) and
the rigid surface.

Figure 2.30

28 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.31 shows details of the contact definition. In (1) we select both vertices as a contact pair and the
block as a target pair. Automatically the options Contact Bodies and Target Bodies (2) are filled in giving
information about the type of entities selected. In (3) we define the kind of contact. In this case we have
select Frictionless because this kind of contact is used to model normal contact. Finally, in (4) we select
the method to solve the problem, Pure Penalty in this case with its corresponding penalty parameter
specified in (5) as Normal Stiffness Factor.

Figure 2.31

Once the characteristics of the contact have been specified, the boundary conditions of the model must be
defined. The displacement boundary condition (Simply Supported for the truss and zero Remote
Displacement for the rigid block) and the force are defined as usually. Figure 2.32 summarizes the
boundary conditions.

Figure 2.32
Module: Contact Non-linearities 29
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

As we are solving a nonlinear problem, therefore, before solving the problem, we have to define the
number of substeps to be used in the solution process in Static Structural > Analysis Settings. In this case
we adopt the default options (Figure 2.33).

Figure 2.33

After solving the problem, we can initiate the postprocessing of the results. The displacement in the contact
point is plotted as usually. Figure 2.34 show the horizontal and the vertical displacements through the
different substeps. We can see very clearly how once the contact occurs the vertical displacement remain
stable. Figure 2.35 shows the total displacement. The vertical reaction is plotted in Figure 2.36 and the
correlation with the theoretical value is excellent.

Figure 2.34

30 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.35

Figure 2.36

Information about the contact can be obtained in the Contact Tool (Solution > Insert > Contact Tool)
(Figure 2.37). We can get information about, for instance, the contact status (Figure 2.38), which allows
knowing when the contact happens, or the penetration (Figure 2.39).

Module: Contact Non-linearities 31

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.37

Figure 2.38

Figure 2.39

2.3 FLEXIBLE-TO-FLEXIBLE CONTACT: FINITE ELEMENT APPROXIMATION AND

METHOD OF CONSTRAINT ENFORCEMENT

The methodology shown in the previous section can be easily extended to a flexible-to-flexible
contact between two bars. In this case, the kinetic contact constraints shown in Equation (2.6) instead
of (2.4) should be included in the weak formulation of the problem. For it, equation (2.12) would be
extended to introduce the contribution of the two flexible contacting bars, A and B, as follows

32 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

LA LB LA

∫w
0
A, x E A ( x) AA ( x)u A, x dx + ∫ wB , x EB ( x) AB ( x)uB , x dx =
0
∫ w ( x)b ( x)dx + ∫ w ( x)b ( x)dx + (2.33)
0
A A B B

wA PA ΓtA
+ wB PB ΓtB
+ wA RN Γc
− wB RN Γc

where the subscripts A and B are used to refer to each one of the two contacting bars, respectively. In
this case, the contribution to the work made by the force RN, originated once the contact occurs, is
due to the two bars. According to the kinetic constraint in (2.6), RN < 0 when contact occurs, i.e. the
force RN is compressive. Therefore, the work carried out at the contact interface by RN is negative for
bar A and positive for B since in the first case the compression force follows the negative sense of
the global X-axis while for the second bar the force and the global displacement have the same sense.
The system derived from the finite element method takes a similar form to (2.16)
 {d } 
([ K ] [E] )
t

 RN 
 = {F } (2.34)

where [ K ] and { F } are refered to the total structural system comprising the two bars; [ E ] is a
matrix of constants whose structure, in this particular case of flexible-to-flexible contact between two
bars, will have one row and as many columns as degrees of freedom are defined for the two bars
together. All the columns will be zero except two corresponding to the degrees of freedom of the two
bars contacting. One of these columns will take the unit value while the other will be equal to -1.
The incorporation of the kinematic constraints can be performed by using any of the previous
methods shown in Section 2 based on Lagrange multipliers or the introduction of penalty parameters.
0 by the constraint g − ( u A − uB ) =
The same equations would be applied replacing g − u N = 0 and
[ K ] , {F } and [ E ] are computed in this case as commented for Equation (2.34).
The contact between two bars is considered as a node-to-node contact from the numerical simulation
point of view.

EXAMPLE 2.13 The steel bars of the Figure 2.40 have a section of 15 cm2. Compute the curve
displacement versus load increasing P from 0 to 300 kN by using the Lagrange multiplier method.

2000 1 1000

Figure 2.40

Solution: As we can see in Figure 2.42, the problem has two horizontal degrees of freedom.

d1x d2x

Figure 2.41

Module: Contact Non-linearities 33

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Before contact occurs, the finite element formulation of the problem gives
 AE 
 L 0 
  1x  =  P 
d
 1 (a)
 AE   d 2 x   0 
 0 
 L2 

where it is clear that both displacements are uncoupled and d 2 x is always zero. The contact condition is
given by Equation 2.5 that, in this case, can be written as
g ≥ ( d1x − d 2 x ) (b)
Therefore the constraint matrix is
[ E ] = ( −1 1) (c)

and (3.24) becomes

 AE 
 L 0 −1  d1x 
 1   d 2 x  =  P  (d)
  
1   λN   
AE 0
 0
 L2 
Using Lagrange multiplier approach constraint (b) is introduced by using a virtual Lagrange multiplier as
follows
λNw ( g − ( d1x − d 2 x ) ) =
0 (e)

Combining (d) and (e) we obtain

 AE 
 L 0 −1
 1   d1x   P 
 AE    
 0 1   d2 x  =  0  (f)
 L2  λ   −g 
 −1  N  
1 0
 
 
Replacing the parameters by their values we have
157500 0 −1  d1x   P 
    
 0 315000 1   d 2 x  =  0  (g)
 −1 0    −3 
 1  λN   −1 ⋅ 10 
Equation (d) will be valid until the contact between the two bars occurs. From that moment, any new load
increment will follow the system (f).
Figure 2.43a shows the value of both displacements when load is increased, while Figure 2.43b shows the
value of the contact force λN as a function of the load P. We can identify the increment of stiffness at the
instant in which contact occurs. As the length of the bar on the right is half the bar on the left, its stiffness
will be double and now, after the contact occurs, the stiffness of the two joined bars will be three times the
stiffness of bar 1. It is clearly observed in the slope of the graph.

34 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.42

EXAMPLE 2.14 Solve Example 2.13 using a penalty method. Make a sensibility analysis of the solution
considering for the penalty parameter β values of 103, 106 and 108 kN/m.

Solution: When the contact condition is active, the penalty method is equivalent to the introduction of a
spring of contant equal to the penalty parameter between the degrees of freedon d1x and d 2 x . The
formulation of the contact problem is given by the following equation system
 AE 
 L +β −β  d
 1   1x  =  P + β g  (a)
  d −β g 
+ β  2x  
AE
 −β
 L2 
and the contact force is computed as
RN = β  g − ( d1x − d 2 x )  (b)

Replacing the parameters by their values we obtain the results plotted in Figure 2.44.

Module: Contact Non-linearities 35

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.43

EXAMPLE 2.15 Solve Example 2.13 using the perturbed Lagrangian method. Make a sensibility analysis
of the solution when the parameter κ takes values of 105, 106 and 108 m/kN.
Solution: In this case, the system of equations to be solved is given by
 AE 
 L 0 −1
 1   d1x   P 
 AE    
 0 1   d2 x  =  0  (a)
 L2  λ   −g 
  N  
1
 −1 1 
 κ
Replacing the parameters by their values we have
157500 0 −1   d1x   P 
    
 0 315000 1   d 2 x  =  0  (b)
  λ   −1 ⋅ 10−3 
1 κ 
 −1 1 N   
Figure 2.44 plots the solution of the system (b) for the different proposed values of κ.

36 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.44

When κ increases the values are closer to the solution of the Lagrange multiplier method. Low values of
this parameter can turn into wrong results, as can be observed in Figure 2.44.

PROBLEM 2.2 Solve Example 2.13 using the augmented Lagrangian method. Compare the solution with
different values of the penalty term.

EXAMPLE 2.16 Solve with ANSYS Workbench the Example 2.13

Solution: The procedure to solve this example is basically the same used in Example 2.5. In this case the
rigid block is substituted by a beam, and the application of the boundary conditions is simpler. The
geometry of the bars is modeled with two lines of cross section equal to 15 cm2 and separated a distance of
1 mm. After that we open Mechanical module.
First of all, we define the mesh, in this case we use one element per line (Figure 2.45)

Module: Contact Non-linearities 37

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.45

Next we define a connection group (Connections > Insert > Connection Group in the Model tree) (Figure
2.46).

Figure 2.46

Now, we will create a named selection (Model > Insert > Named Selection) including the two nodes
involved in the contact process (Figure 2.47). We rename this pair as Node_Node_Contact and will be used
later in combination with APDL commands to define the node-to-node contact.

38 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.47

The next step is the application of the boundary conditions, fixed supports at the beginning of the first
beam and at the end of the second beam and a force of 300 kN. However, the contact has not been defined
yet. That is because the contact node-to-node is not implemented in ANSYS Mechanical in Workbench
environment. So an APDL command serie, such as shown in Figure 2.48, has to be added.

Figure 2.48

Once we have defined the contact pair using APDL, we set the Analysis Settings. In this case we have used
100 substeps to get a good representation of the load process (Figure 2.49), although we might have solved
with less substeps.

Module: Contact Non-linearities 39

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.49

Finally, we solve the problem and check the results. Figures 2.50, 2.51 and 2.52 show the values of the
displacements, axial forces and force reactions, respectively, in both bars. The change of slope indicates
the instant when the contact takes place.

Figure 2.50. Displacement of the bars

40 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.51. Axial forces of the bars

Figure 2.52. Reaction forces of the bars

2.4 FRICTIONAL CONTACT: FINITE ELEMENT APPROXIMATION AND METHOD OF

CONSTRAINT ENFORCEMENT

Once the contact between two structural bodies or between a structural body and a rigid obstacle
occurs we have seen until now that the two faces or edges in contact can remain bonded or coupled
together both in tangential and normal directions or coupled in their normal direction only (no
separation contact) allowing a small sliding on each other in tangential direction or even, depending
on the level of applied lateral force, a free large tangential sliding (Frictionless contact). However, it
is more usual that when two faces or edges get in contact, they may slide only when the tangential
force between them exceedes a critical value or friction force. The kinetic and kinematic constraints
for this contact type were shown in (2.10) and its formulation by finite elements will be developed
now.
In a node-to–node contact between two bars as that shown in Section 2.3, relative tangential
movement of the nodes is not allowed in the contact area. Therefore, the study of the frictional
contact will be performed between a structural system and a rigid surface as, for example, the truss
that comes into contact with a rigid foundation of Example 2.10. If we consider any bar of the truss
system, the weak formulation shown in Equation (2.12) will be modified as follows

Module: Contact Non-linearities 41

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

L L

∫ wN , x E ( x) A( x)uN=
0
, x dx ∫w
0
N ( x)b( x)dx + wN PN Γt
+ wT PT
Γt
+ wN RN Γc
− wT RT Γc
(2.35)

where RT is the reaction force of the support to prevent slip once contact occurs, assumed to have an
opposite sense to the tangential displacement, and wT is the virtual displacement in tangential
direction to the rigid surface. The concentrated loads have been divided in two, PN and PT, depending
on if they are applied in normal or tangential directions, respectively. The last term of (2.35) gives
the contribution of the Principle of Virtual Work to the tangential kinetic conditions on the contact
interface. As commented in Section 2.1.3, the absolute value of the friction force has been taken
since the tangential force can be positive or negative. If Coulomb friction is assumed then
RT = − µ RN
In this case, the system of equations (2.13) defined for all the structural system once the finite
element method has been applied is transformed into the following system
 {d } 
([ K ] [ EN ] [ ET ] )
t t  
 RN  = { F } (2.36)
R 
 T 
where [ EN ] and [ ET ] are the matrices of constants whose values will depend on the kinematic
constraints of the problem in normal and tangential directions to the obstacle, respectively. Both will
have one row and as many columns as degrees of freedom are defined for the structure. In both
cases, all the columns will be zero except that corresponding to the degree of freedom of the
structure contacting with the surface in normal or tangential direction depending if we are defining
[ EN ] or [ ET ] , respectively.
The kinematic constraints are imposed in a similar way to the no friction case developed in
Sections 2.2.1-2.2.5.

2.4.1 Lagrange multiplier method

If we use the Lagrange multiplier method, considering the kinematic constraint in (2.10), Equation
(2.21) would be modified as follows for the stick condition
λNw ( g − u N ) + λTwuT =
0 (2.37)

where, in this case, we have two arbitrary constants or virtual Lagrange multipliers, λN and λT , one
w w

for each direction. When the finite element method is applied this equation can be written in matrix
form as follows
[E ]
{G}  N  {d } −=
= {Q} 0 (2.38)
 [ ET ] 

where {Q} in this case is

g
{Q} =   (2.39)
0
Combining (2.36) and (2.38), we obtain the modified finite element equations by the Lagrange
multiplier method
42 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 [K ] [ EN ] [ ET ]   {d }   F 
t t

 { }
 [ EN ] [0] [0]   λN  =  Q  (2.40)
{ }
 [E ]
 T [0] [0]   λT   
where, as usually, we have used the notation λN and λT to denote the reactive forces RN and RT ,
respectively. As, in this case, we have two constraint equations, the Lagrange multiplier has two
constraints.
In case of slip, Equation (2.21) would be applied instead of Equation (2.37) and (2.38) should be
also replaced by (2.22) since uT ≠ 0 . Furthermore, RT should be replaced by − µ RN in (2.36).
Therefore, (2.40) would become in this case
 [K ] t
 { F } − [ ET ]t RT   [ K ] [ EN ]t − µ [ ET ]t   {d }   { F } 
[ EN ]   {d }  =
   ⇒  =
   λN   {Q} 
(2.41)

 [ EN ] [0]   λN   {Q}  [E ]
  N [ 0 ]   

2.4.2 Penalty method

In the penalty method, Equation (2.20) is reformulated in case of stick as follows
L L

∫ wN , x E ( x) A( x=
0
)u, x dx ∫w
0
N ( x)b( x)dx + wN PN Γt
+ wT PT Γt
+ wN β ( g − u N ) Γc
+ wT β uT Γc
(2.42)

Although here, for simplicity, the penalty factor β has been chosen to be the same for both
constraints, it might be different.
According to (2.42), Equation (2.36) is replaced by

N
t
N T
t
T ({F} + [ E ] β {Q})
( [ K ] + [ E ] β [ E ] + [ E ] β [ E ] ) {d } = N
t
(2.43)

In this method, the constraints are not applied directly. They will appear in the solution in an
approximate way.
In case of sliding, (2.42) should be rewritten as follows
L L

∫ wN , x E ( x) A( x=
0
)u, x dx ∫w
0
N ( x)b( x)dx + wN PN Γt
+ wT PT Γt
+ wN β ( g − u N ) Γc
− wT RT Γc
(2.44)

and (2.43) would be replaced by

( [ K ] + [ E ] β [ E ] ) {d } =
N
t
N ({F} + [ E ] β {Q} + [ E ] R )
N
t
T
t
T (2.45)

EXAMPLE 2.17 Solve Example 2.10 considering that after the contact the truss and the foundation remain
bonded in such a way that the tangencial sliding is zero. Compare the penalty approach with the Lagrange
multiplier method.
Solution: The approach used in Examples 2.10 and 2.11 allows the slip (without friction) between the bars
and the rigid foundation. ANSYS allows the use of another kind of contact called rough contact in which,
when two bodies contact, they cannot slide in tangential direction due to large friction between them. In
this case, when Lagrange multiplier method is used, two contact conditions have to be considered once
contact is reached

Module: Contact Non-linearities 43

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

( g − d1z ) =
0
(a)
(d1x (d1z =−1 ⋅ 10−3 ) − d1x ) ≤ 0
−3
where d1x (d1z =−1 ⋅ 10 ) is the value of the horizontal displacement, d1x , when the vertical displacement is
equal to -1 mm, i.e. the contact occurs. This value is unknown, so it should be determined during the load
process.
In this case, we have to solve a classic linear problem until the contact is detected and afterwards introduce
the contact in the formulation. The larger the error in the detection of the contact is, the larger the error in
the evaluation of the contact forces and displacements will be. There are different ways to detect when
contact occurs. One of them is based on the application of the load incrementally until the truss contacts
the foundation. At that moment we obtain the value of d1x .
From Equation (b) of Example 2.10 we have
−1
 d1x   59803 6883.1  60 
=    ∆  (b)
 d1z   6883.1 77443   −100 
where ∆ is a parameter used to control the magnitude of the applied load at each step and whose values
are between 0 and 1. After each step the contact status is checked. Figure 2.53 shows the evolution of both
displacements with the parameter ∆ . When the vertical displacement reaches the contact surface the ∆
parameter takes a value of 0.717 and the horizontal displacement is 0.8344 mm

Figure 2.53

From (2.40), once contact occurs, the system which allows solving the contact reaction forces is given by
 59803 6883.1 0 1   d1x   60 
    
 6883.1 77443 −1 0   d1z  
=
−100  (c)
 0 −1 0 0   λN   1 ⋅ 10−3 
    
 1 0 0 0   λT   0.8344 ⋅ 10−3 
By solving (c) we get the solution of the problem

44 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 d1x   0.83446 ⋅ 10−3 

   −3 
 d1z  =  −1 ⋅ 10  (d)
 λN   28.3 
   
 λT   16.98 
In (c) we have assumed that λN follow the positive sense of the global coordinate axis while λT follows
the opposite sense. If, after solving them, their values are positive it will confirm the initial assumption, if
they are negative it means that the real sense is opposite to the initially assumed.
This problem can be also solved using the penalty method. In spite of the values of the penalty parameter
can be chosen different at each direction we will use the same value β for all directions. We have to
apply (2.43) for the following matrices
 0.8344 ⋅ 10−3 
[ EN ] =
( 0 − 1) [ ET ] =
(1 0 ) { } 
Q = −3  (e)
 −1 ⋅ 10 
We have

 59803 + β 6883.1   d1x   60 + 0.8344 ⋅ 10−3 ⋅ β 

   =   (f)
 6883.1 77443 + β  d1z   −100 − 1 ⋅ 10−3 ⋅ β 

Table 2.2 summarizes results considering different values of the penalty parameter. As it can be observed,
the lowest values give rise to important discrepancies between the computed values and the real ones.
However if the value is extremely high important mistakes can be also made because of numerical
problems.

Table 2.2

β d1x [mm] d1z [mm] RN= λN [kN] RT= λT [kN]

3
10 1.1578 -1.3891 -0.32331 0.38915
104 1.1118 -1.3455 -2.7732 3.4547
105 0.94777 -1.1639 -11.331 16.389
106 0.85065 -1.0264 -16.193 26.37
107 0.83615 -1.0028 -16.898 28.095
108 0.83463 -1.0003 -16.972 28.28
1020 0.83446 -1 0 21.684

EXAMPLE 2.18 Solve Example 2.10 using the Lagrange multiplier method considering a frictional rigid
surface with a Coulomb friction coefficient µ = 0.3 .

Solution: From (2.41), the contact formulation between the lower vertex of the truss and the rigid
foundation, once contact has occurred, is given by

 59803 6883.1 − µ   d1x   60 

s

    
 6883.1 77443 −1   d1z  =  −100 
(a)
 0 −1 0   λN   − g 

s
where d1x is the additional horizontal displacement once the contact occurs.
In Equation (a), the sense of λN has been assumed to be upwards, i.e., in agreement with the global Y-axis.
The solution of (a) gives

Module: Contact Non-linearities 45

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 d1sx   0.0013 
   −3 
 d1z  =  −1 ⋅ 10 
(b)
 λN   31.337 
   
Therefore, the total horizontal displacement is 0.8344+0.0013=0.8357 mm, where the value of 0.8344 mm
is the horizontal displacement of the vertex when it contacts with the foundation and it was computed in
Example 2.17.

EXAMPLE 2.19. Using a penalty approach solve the problem of the sliding of the two masses shown in
Figure 2.54 on a rough surface. Particularize the solution for the following values: m1=2 kg; m2=1 kg;
F1=0; F2= 12 N; µ = 0.25 ; k1=k2=103 N/m and the penalty parameter equals to 104 N/m

Figure 2.54

Solution: Penalty approach requires the addition of springs (of stiffness β ) between the masses and the
support (Figure 2.55). The friction forces are given by these spring forces.

Figure 2.55

Considering the horizontal equilibrium of forces over each mass (Figure 2.55) we have
−k1d1 + k2 ( d 2 − d1 ) + F1 − β d1 =0
(a)
−k2 ( d 2 − d1 ) + F2 − β d 2 = 0
Equation (a) can be written in matrix form as follows
 k1 + k2 + β −k2   d1   F1 
   =   (b)
 −k2 k2 + β   d 2   F2 
Now we have to take into account the three possible situations we can find during the load process:

Total stick

This occurs when the applied load is not enough to exceed the static friction given by the Coulomb law
46 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

R ≤ µ mg (c)

and the reaction forces are computed as β d1 and β d 2 for the first and second masses, respectively. In this
case equation (b) solves the problem.

Stick and Sliding

In this case one of the reactions exceeds the friction and starts to slide while the other one remains
stopped. The mass that begins to slide has a tangential force of value µ mg .
If the first mass is assumed to slide, the equation system to be solved is
 k1 + k2 −k2  d1   F1 − µ m1 g 
   =   (d)
 − k2 k2 + β  d 2   F2 
If it is the second mass we have
 k1 + k2 + β −k2  d1   F1 
   =   (e)
 − k2 k2  d 2   F2 − µ m2 g 

Total sliding

In this scenario both masses are sliding on the support and the tangential forces in both cases are given by
the Coulomb`s law. The equation system is
 k1 + k2 −k2  d1   F1 − µ m1 g 
   =   (f)
 − k2 k2  d 2   F2 − µ m2 g 

Considering these three possible situations, the load is applied incrementally and the tangential forces are
checked to test in which of the three situations we are. In our case, we will apply the load of 12 N in 100
steps, hence the load vector= (0 12∆ ) , where ∆ is the load step control. The system to solve
t
is F
initially is
 2 ⋅ 103 + 104 −103   d1   0 
   =   (g)
 −10 3
103 + 104   d 2  12∆ 

This system will be valid while

104 d1 ≤ 0.25 ⋅ 2 ⋅ 9.81
(h)
104 d 2 ≤ 0.25 ⋅ 1 ⋅ 9.81
When the load reaches 2.787 N the second mass starts sliding and the new condition introduces a new
reaction force in the second mass (Figure 2.56).

Figure 2.56

Module: Contact Non-linearities 47

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

The system to be solved becomes now

 2 ⋅ 103 + 104 −103   d1   01 
 3  =  (i)
 −103 10   d 2  12∆ − 0.25 ⋅ 1 ⋅ 9.81

which is valid while

104 d1 ≤ 0.25 ⋅ 2 ⋅ 9.81 (j)
This constraint is exceeded for a force of 7.878 N. At that moment, the first mass also starts to slide.

Figure 2.57

The new situation is represented in Figure 2.57 and the system to be solved is
 2 ⋅ 103 −103   d1   −0.25 ⋅ 2 ⋅ 9.81 
   =   (k)
 −10
3
103   d 2  12∆ − 0.25 ⋅ 1 ⋅ 9.81

Figure 2.58 shows the evolution of the displacements of both masses during the loading procedure. We can
determine by the change in the slope the instant in which each mass begins to slide.

Figure 2.58

48 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

2.4.3 Perturbed Lagrangian method

To apply this method in a frictional contact, Eq. (2.23) is rewritten in case of stick as follows
 λ   λ 
λNw ( g − u N ) + N  + λTw ( uT ) + T  =
0 (2.46)
 κ   κ
As, in the previous method, the parameter κ might be the same or different for both constraints.
After applying the finite element method, Equation (2.46) becomes in matrix form as follows
[E ] 1 λ 
{G}  N  {d } +  N  −=
= {Q} 0 (2.47)
 [ ET ]  κ  λT 
Combining now (2.36) and (2.47) we obtain
 [K ] [ EN ] [ ET ]   {d }   F   [ K ] [ EN ] [ ET ]   {d }   F 
t t t t

 { }  { }
 [ EN ] [1 κ ] [0]   RN  = 

 ⇒  [ EN ] [1 κ ] [ 0]   λN  =
{Q} 
(2.48)
 [E ]     {Q}      
 T [0] [1 κ ]   RT   [ T ] [ ] [ ] T 
E 0 1κ λ

PROBLEM 2.3 Derive the formulation of the perturbed Lagrangian method when sliding occurs.

2.4.4 Augmented Lagrangian method

This method is a combination of the penalty method with the Lagrange multiplier method. For it,
firstly, in case of stick, Equation (2.35) is modified as follows
L L

∫ wN , x E ( x) A( x)uN=
0
, x dx ∫w
0
N ( x)b( x)dx + wN PN Γt
+ wT PT Γt
+ wN RN Γc
− wT RT Γc
(2.49)
+ wN β ( g − u N ) Γc
+ wT β uT Γc

An additional penalty term has been introduced in the weak formulation of the problem whose
matrix form, once the finite element method has been applied, is

( [ K ] + [ E ] β [ E ] + [ E ] β [ E ] ) {d } =
N
t
N T
t
({F} + [ E ] β {Q} + [ E ] β {Q} + [ E ] R
T N
t
T
t
N
t
N
t
)
+ [ ET ] RT (2.50)

Furthermore, the kinematic constraints are enforced to be satisfied by using Equation (2.33) or (2.34)
in matrix form.
If (2.38) and (2.50) are combined we obtain the modified finite element equations by the augmented
Lagrange multiplier method for the frictional problem
 [ K ] + [ EN ]t β [ EN ] + [ ET ]t β [ ET ] [ EN ] [ ET ]   {d }  
t t

  { F } + [ EN ] β {Q} + [ ET ] β {Q} 
t t

 [ EN ] [ 0] [ 0]   = RN λ=
N  



     {Q} 
 [ ET ] [0] [0]   RT = λT 
(2.51)
PROBLEM 2.4 Derive the formulation of the augmented Lagrangian method when sliding occurs.

Module: Contact Non-linearities 49

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

PROBLEM 2.5 Using the augmented Lagrangian method solve the Example 2.10 considering a rough rigid
surface with a Coulomb friction coefficient µ = 0.3 .

EXAMPLE 2.21 Solve with ANSYS Workbench the Example 2.10 considering a rough rigid surface with a
Coulomb friction coefficient of µ = 0.3 .

Solution: The model of this example is exactly the same to the one used in Example 2.12. There is only
one difference in the configuration of the contact model that in this case is frictional (Figure 2.59)

Figure 2.59

Results of reaction forces during the loading procedure are shown in Figure 2.60. We can observe that the
value computed of the normal reaction force in Example 2.18 and the value computed with ANSYS are
almost equal (Figure 2.60).

50 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.60

2.5 LARGE DEFORMATION CONTACT

In the two applications of finite elements to contact problems shown previously (rigid-to flexible and
flexible-to-flexible) only small changes in the geometry were assumed so that the geometrically
linear theory could be applicable. In small deformation contact the mesh topology does not change
during the loading process since the slip is infinitesimally small. When large sliding occurs (either
frictionless or frictional) geometrically nonlinear formulations should be carried out to solve the
problem suitably. In this section the finite element equations for geometrically non-linear contact are
derived for a truss element in 2D, not because of the intrinsic importance of this kind of elements,
but rather because they provide in a simple and easy way an introduction of the concepts which will
be used later for two- and three-dimensional problems.

2.5.1 Overview of non-linear analysis

The general form of the global equilibrium equations of a discretized system by the finite element
method is stated as shown in Equation (2.13). These equations correspond to a linear analysis of the
structural problem since we have assumed infinitesimally small displacements and elastic linear
material. In that case, the displacement response is a linear function of the applied load vector. The
assumption of a linear elastic material involves the use of a constant stress-strain matrix. The fact of
considering small displacements allows computing all the integrals involved in the evaluation of the
matrix [ K ] and the load vector { F } over the original volume of the finite elements and,
furthermore, the strain-displacement matrix of the elements is independent of the displacements.
Equation (2.13) can be written alternatively in residual form as follows
{r} = {FEXT } − {FINT } = {0} (2.52)

Module: Contact Non-linearities 51

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

where {r } is the residual or out-of-balance force vector, {FEXT } = {F } is a vector containing all
externally applied nodal forces and { FINT } represents the vector of nodal forces due to the internal
stresses. In the linear case, it agrees with [ K ]{d } .
If the problem is linear, according to (2.13), the displacements are proportional to the loads and
the structural stiffness is independent on the load level. Therefore, the displacement vector is
computed directly. However, in nonlinear analysis { FINT } is regarded as a non-linear function of
{d } but, in any case, we are also required to compute the displacements associated with the applied
loads. As we cannot solve the system of equations directly, we must proceed applying incrementally
the load, i.e. dividing the load into small increments, whereby displacements are updated after each
solution. Equation (2.52), expressing the equilibrium in the current (deformed) geometry, must be
satisfied for any time or load step. In the nonlinear case, it can be rewritten as follows

{F } − {F ({d })} =
{r ({d })} = EXT {0}
INT (2.53)

which indicates the dependency of the residual vector with the displacements. Consequently, the
solution of a nonlinear mechanical problem consists of finding the displacement vector which
produces an internal force vector balancing external applied loads.
The basic approach to solve the nonlinear equation (2.53) is an incremental step-by-step solution
in which a linear approximation for small increments of external force and corresponding increments
of displacement is performed. Thus, if we have a system with only one displacement variable d and
assume that for a prescribed value of force P + ∆P we manage to find a scalar displacement
d + ∆d satisfying the Equation (2.53), we use the following linear approximation of (2.53)
∂FINT ( d )
+ ∆P FINT ( d ) +
P= ∆d (2.54)
∂d
where d is the displacement due to P.
Therefore, we have
∂F ( d )
KT ∆d = INT ∆d =∆P (2.55)
∂d
where
∆P ∂FINT ( d )
=
K = (2.56)
∆d ∂d
T

is the so-called tangent stiffness. It plays a similar role in nonlinear problems to the stiffness in linear
problems (Figure 2.61). Unlike stiffness which is constant, tangent stiffness changes throughout the
load-displacement equilibrium path. For a multiple degree-of-freedom system, we have a tangent
stiffness matrix

[ KT ] =
{
∂ FINT ({d } )} (2.57)
∂ {d }
The most widely used methods to solve non-linear finite element problems are incremental-
iterative methods based on the Newton-Raphson technique. These methods are based on a
combination of incremental and iterative methods and consist of a predictor step involving forward
Euler load increment and a corrector step in which some iterative procedure is applied to enforce
equilibrium. On one hand, the application of the load incrementally allows tracing the complete load-
52 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

displacement response. On the other hand, the application of an iterative method, such as Newton-
Raphson method, reduces the cumulation of errors for each load increment due to imbalance forces
along the incremental procedure. The procedure is explained in detail in Chapter 8 of the module
“Fundamentals and Applications of Finite Element in Mechanical Analysis”.

Figure 2.61. Tangent stiffness concept

2.5.2 Finite element formulation in large displacement kinematics

As categorized in Chapter 8 of the module “Fundamentals and Applications of Finite Element in
Mechanical Analysis”, non-linearity can be due to the material behavior or to the appearance of large
displacements and rotations and, furthermore, to the change of the boundary conditions during the
motion as observed in contact problems. In any analysis, it will be necessary to decide if a problem
falls into one or other category. The non-linearity of material is addressed in great detail in the
advanced module “Advanced non-linearities”. The treatment of contact in kinematic linear problems
was addressed previously in this chapter. We will introduce now the geometrically-nonlinear-only
analysis and, subsequently, we will consider kinematic nonlinear contact problems. As the
development is restricted to one dimension the main features of the formulation can be shown easily
and their extension to non-linear continua in 2D and 3D will be also easier.
Different formulations can be used for large deformation problems. The suitable one depends on
the characteristics of the problem to be solved. Two approaches are commonly taken:
a) Total Lagrangian formulation: In this case, the variables (stress and strain) and integrals are
expressed in terms of the undeformed initial configuration of the solid defined in material
coordinates X.
b) Updated Lagrangian formulation: The variables and integrals are expressed taking as
reference the deformed current (or updated) configuration of the solid defined in spatial
coordinates x.
The major difference between both formulations for geometrically nonlinear problems is the need to
define the coordinates over which the integrals are evaluated and to specify the choice of stress and
strain measures. Although the underlying mechanics of the two formulations are identical and the
total Lagrangian formulation can be transformed to updated Lagrangian formulation and viceversa,
we will adopt firstly a total Lagrangian formulation. This formulation is the most widely used in
FEM codes.
Figure 2.62 shows a two-node element belonging to a pin-jointed plane truss structure in its
undeformed original configuration. In this configuration the element has cross-section area A0 and
length L0. This configuration is called base configuration and it agrees with the reference

Module: Contact Non-linearities 53

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

configuration in the total Lagrangian formulation. Therefore, the reference configuration remains
fixed throughout the solution procedure. All equations are referred to this reference configuration.
The coordinates used to describe the position of any material point in the base configuration, or
material (Lagrangian) coordinates, are denoted by the global Cartesian axes X-Z. As the truss
experiences deformation, the element experiences a displacement. In the deformed or current
configuration the cross-section area and the length become A and L. This current configuration is
defined by the spatial (Eulerian) coordinates, denoted by the Cartesian axes x-z. In the total
Lagrangian formulation, the two Cartesian frames, X-Z and x-z, agree. We assume that the structure
may undergo large displacements and rotations but the strains are small to assume that the material
stays in the linear elastic range, i.e. only geometric non-linearities are considered.

2 (X2, Z2) dX2

A0, L0
Reference configuration=
Base configuration dZ2
e
α
Z, z A, L 2 (x2, z2)
1 (X1, Z1)
dZ1
dX1 Current configuration
1 (x1, z1)

X, x
Figure 2.62. Motion of a truss element in 2D –Total Lagrangian

In the analysis we follow all the points of the element in their motion, from the original to the
final configuration of the element. The displacement at the four degrees of freedom of the element
can be expressed as the difference between the current position and the original position of a material
point, so

{d=} {x } − { X }
e e e
(2.58)

where
 dX1   x1   X1 
     
{d }  d= { xe } { X e}
 dZ1  =z1   Z1 
= e
(2.59)
  x2   X2 
     
X2

 dZ 2   z2   Z2 
Although shape functions are unnecessary to compute the internal force vector for this truss
element, due to its simplicity, they will be used to calculate the external force vector and,
furthermore, they will be useful with a view to more complex elements developed in further
chapters. By using isoparametric formulation, shape functions are used to relate both the geometry
and the displacements to nodal values. We can write

54 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 x1 
 
 x e (ξ )  ψ 1 (ξ ) ψ 2 (ξ ) 0   z1 
   ψ (ξ )  { x }
0
 e  =
e e
(2.60)
 z (ξ )   0 ψ 1( )
ξ 0 ψ 2 ( )
ξ x
 
2

 z2 
 X1 
 
 X e (ξ )  ψ 1 (ξ ) ψ 2 (ξ ) 0   Z1 
   ψ (ξ )  { X }
0
 e  =
e e
(2.61)
 Z (ξ )   0 ψ 1( )
ξ 0 ψ 2 ( )
ξ X
 
2

 Z2 
 dX1 
 u xhe (ξ )  ψ 1 (ξ )  
ψ 2 (ξ ) 0   dZ1 
{u he } = ψ e (ξ )  {d e }
0
=   =  (2.62)
 u z (ξ )   0
he
ψ 1 (ξ ) 0 ψ 2 (ξ )   d X 2  
 
 dZ 2 
where
1− ξ 1+ ξ
=ψ 1 (ξ ) = ψ 2 (ξ ) (2.63)
2 2
are the shape functions defined on a one-dimensional standard element of coordinates ξ1 = −1 and
ξ 2 = 1 (Figure 2.63). In Equation (2.62) superscript h means, as usually, that we are approximating
the displacement.

e x, z ξ
x1, z1 x2,z2 ξ1=-1 ξ2=+1

MAPPING
(ξ ) z z (ξ )
=x x=
Figure 2.63. Map of a master element onto an arbitrary element e

In a large deformation analysis, the configuration of the solid is changing continuously. This is an
important difference compared with linear analysis in which it is assumed that the displacements are
infinitesimally small so that the original configuration is always used. The continuous change in the
configuration of the solid entails, among other important consequences, that the stresses are not only
due to the straining of the material; they must also take into account the rigid body rotation of the
material. This change in configuration is dealt effectively by defining auxiliary stress and strain
measures. The objective is to express the internal vitual work in terms of an integral over a known
volume and to be able to incrementally decompose the strains and stresses in an effective way.
Without getting into details, to state the Principle of Virtual Work using a total Lagrangian
formulation, the internal work is obtained from the Green strain and the second Piola-Kirchhoff
stress. As we are working with a truss element the strain has only one component, the axial strain ε.
The Green axial strain for the element e is referred to the initial length of the element L0 and is given
by
Module: Contact Non-linearities 55
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

L2 − L20
ε =
e
(2.64)
2 L20
In (2.64), the initial length of the element L0 is

L20 = ( X 2 − X 1 ) + ( Z 2 − Z1 ) = X 212 + Z 212

2 2
(2.65)
where (Figure 2.64)
( X 2 − X1 )
X 21 = ( Z 2 − Z1 )
Z 21 = (2.66)

X21 2 (X2, Z2)

A0, L0 cosα = X21/L0

Z21 e Reference configuration= sinα = Z21/L0
Base configuration
α
Z, z x21
2 (x2, z2)
1 (X1, Z1) z21
Current configuration
1 (x1, z1) A, L

X, x
Figure 2.64. Definition of geometric parameters

In the same way, from (2.62), the current length L is

L2 = ( X 21 + d X 21 ) + ( Z 21 + d Z 21 )
2 2
(2.67)
where
(dX 2 − dX1 )
d X 21 = ( dZ 2 − dZ1 )
d Z 21 = (2.68)
Therefore, from (2.64), (2.65) and (2.67), the Green strain becomes

( X 21 + d X 21 ) + ( Z 21 + d=
Z 21 ) − X 21 − Z 21
2 2
X 21d X 21 + Z 21d Z 21 d X2 21 + d Z2 21
2 2

=ε e
+ (2.69)
2 ( X 212 + Z 212 ) L20 2 L20

Equation (2.69) can be re-expressed in matrix form as

ε e  B1e  {d e } + {d e } [ H e ]{d e }
1 t
= (2.70)
2
where
 X 2 − X1 Z 2 − Z1 X 2 − X1 Z 2 − Z1  1
 B1  =
e
− − = ( − cos α − sin α cos α sin α ) (2.71)
 L20 L20 L20 L20  L0

56 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 1 0 −1 0 
 
1  0 1 0 −1
 H  = 2
e
(2.72)
L0  −1 0 1 0 
 
 0 −1 0 1 
In (2.71) α is the angle between local axis of the element and global axis X.
Equation (2.70) has two terms. The first term depends linearly on the nodal displacements and it
is similar to the strain-displacement matrix obtained for linear problems. The second term, however,
depends quadratically on the nodal displacements and it introduces the non-linearity geometric of the
problem.
By analogy with linear problems, Equation (2.70) might be written as follows

( (
 B1e  +  B2e {d e }  {d e } =
εe =
 ) )
 B e {d e }  {d e }
 ( ) (2.73)

where

 B2e
 ({d }) = 12 {d } [ H
e e t e t
] (2.74)

Now, from (2.11) and (2.73), the Principle of Virtual Work can be used to express the equilibrium of
the structure
 
∑ {ce } ∫  Be  σ Ge A0 dx ∑ {c }  ∫ ψ  {b} A
NE NE
dx + ψ e  { P}
t t e t e t t
= 0  (2.75)
=e 1 =
Le e 1  Le Γte 

where the virtual displacement and the virtual strain have been also approximated using (2.62) and
(2.73), respectively, through some arbitrary nodal coefficients c e for each element e. For the { }
following developments, we will drop the summation sign and will work directly at the element
level. The assembly procedure to the structural level will be identical to that carried out for linear
analysis.
In (2.75), the stress σG is the axial stress that is work conjugate to the Green axial strain and it is
called the second Piola-Kirchhoff stress.
{ }
Equation (2.75) must stand for arbitrary c e and, hence using (2.73) we have

( (
 B1e  +  B2e {d e }  σ Ge A0 L0 =))
t

∫  B  σ G A0 dx = ∫ ψ  {b} A0 dx + ψ  {P} Γte

et e e e t t
(2.76)
 
Le Le

Therefore, the internal forces are computed from  B1e  and  B2e  as defined in (2.71) and (2.74),
respectively

(  B  +  B ({d }) ) σ
t
{F=
INT }
e
1
e
2
e e
G A0 L0 (2.77)

If we assume a linear material response σ G = Eε and (2.77) becomes

e e

(  B  +  B ({d }) ) Eε A L
t
{F=
INT }
t
 B e  Eε e=
A0 L0 e
1
e
2
e e
0 0 (2.78)

Module: Contact Non-linearities 57

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

From (2.57), as the problem is non-linear the tangent stiffness matrix is obtained simply by
differentiating the internal force given in (2.78) with respect to the displacement vector.

∂ { FINT }  ∂  B e  e 
t
e t ∂ε
e
 K  =
= e 
EA0 L0  B  + ε  (2.79)
∂ {d }  ∂ {d } ∂ {d } 
T e e e
 
From (2.73) we have
∂ε e
=
∂ {d e }
(  B  +  B ({d }) )
e
1
e
2
e
(2.80)

and from (2.78) we obtain

∂ B {d } ( )
∂  B1  + B2 {d } (
  Be d e t 
 2 { }   ( )) ( )
t
 
t
 e e   e
 e e ∂
     
= = (2.81)
∂ {d }
e
∂ {d }
e
∂ {d }
e

since the linear term  B1e  does not depend on the displacements.
Replacing (2.80) and (2.81) in (2.79) the tangent stiffness matrix is given by

(
∂  B2e {d e }  
)
t


( (
 KT  = EA0 L0   B1e  +  B2e {d e }  ) ) (  B  +  B ({d }) )   
t
e e e e  + ε 
e

∂ {d }
(2.82)
  1 2
 e
 
 
From (2.82), the first term can be written as

(
EA0 L0   B1e  +  B2e {d e }  ( ) ) (  B  +  B ({d }) ) 
t
 KTM
e
 = e e e
(2.83)
   1 2

This term is called material stiffness matrix since it depends on the material properties, in this case
the elastic modulus E.
In turn,  KTM
e
 can be divided into two terms as follows

=e
KTM   KTM
e
  e 
1  +  KTM 2  (2.84)

where
t
 KTM
e
  e  e
1  = EA0 L0  B1   B1  (2.85)

 EA0 L0   B1e   B2e ({d }) +  B ({d }) ({d }) ({d }) 
t t
 B1e  +  B2e  B2e
t
 K
= e e e e e e
TM 2  (2.86)
  2
 
Equation (2.85) provides the standard linear stiffness matrix while (2.86) gives the initial
displacement matrix.
The second term of (2.82) can be written as follows

58 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

 1 0 −1 0 
∂ B2 {d } ( )
t
 e e   
  1 A0σ Ge  0 1 0 −1
= KTG  A0 L0
e
= σG =
e
A0 L0σ G [ H ]
e e
(2.87)
∂ {d e } 2 2 L0  −1 0 1 0 
 
 0 −1 0 1 
This term is called geometric or initial-stress matrix since it depends only on the Greem’s stress
(second Piola-Kirchhoff stress). No material properties appear.
Once the tangent stiffness matrix is obtained (Equation (2.82)) an incremental-iterative solution
procedure using full or modified Newton-Raphson iterations might be applied to obtain the solution
of the non-linear geometric problem.

2.5.3 Updated Lagrangian formulation

In an updated Lagrangian formulation the undeformed base configuration remains fixed but the
reference configuration is periodically updated, i.e., unlike total Lagrangian formulation the base
configuration and the reference configuration do not coalesce. In this way, once a deformed
configuration is known for a state (for instance, the last converged solution), it is taken as reference
configuration for the following state and this is continually updated as the calculation proceeds
(Figure 2.65). The new reference configuration would become the old configuration and the updated
coordinates x e ′ can be expressed as
{ }
}′ { x } + {d }
{ x= e e e
(2.88)

2 (x’2, z’2)

2 (X2, Z2) Current configuration

dZ2
Base configuration 1 (x’1, z’1)
dX1
e dX2
Z z dZ1 An, Ln
α 2 (x2, z2)

1 (x1, z1) Reference configuration

1 (X1, Z1)
(Updated after each load increment)

X x
Figure 2.65. Motion of a truss element in 2D –Updated Lagrangian

The second Piola-Kirchhoff stress, used in the total Lagrangian formulation and, therefore,
referred to the base configuration, should be replaced by the true stress relating to the new updated
configuration. If we denote Ln and An the length and the cross-section area of the bar element,
respectively, in the new configuration, the true or Cauchy stress σ is obtained for the element e as
follows

Module: Contact Non-linearities 59

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

A0 Ln e
σe = σG (2.89)
An L0
Use of different measures for stress and strain in both formulations, total and updated Lagrangian,
follows condition that virtual work of internal forces must be the same irrespective of the volume
{ }
over which is integration taken. Hence, from (2.76) with d e = 0 , the internal force vector in the
updated configuration is given by
t t
 e  e ′  e  e  e ′  e
={FINT } ∫=  B  { x }   σ An dx  B1  { x }   σ G An Ln (2.90)
Le      
where
 e  e ′    x2′ − x1′ z2′ − z1′ x2′ − x1′ z2′ − z1′ 
 1
B { x } =
  − −
L2n 
 (2.91)
    L2n L2n L2n
In a similar fashion, the tangent stiffness matrix would, from (2.82), (2.87) and (2.89) with
{d } = 0 , be given by
e

 1 0 −1 0 
 
 e e ′   e e ′  σ An  0 1 0 −1
t

= KT  EAn Ln  B1 { x }   B1 { x }  +
e
(2.92)
    2 Ln  −1 0 1 0 
 
 0 −1 0 1 

Similarly to (2.90), the updated Lagrangian formulation avoids the terms in  B2

e
 ({d }) .
e

The first term in (2.92) gives the linear tangent stiffness matrix while the second term gives the
geometric stiffness matrix.
In the updated Lagrangian formulation all measures are related to the current updated configuration.
However, until now, we have not commented anything about the strain and the Green strain measure
(Equation (2.64)) is related to the initial configuration. In this case, the Almansi’s strain is used
which is defined as follows
L2n − L20
ε Ae = (2.93)
2 L2n
By using (2.93), a similar equation to (2.70) is obtained for the Almansi’s strain changing L0 by Ln in
the expressions for  B1e  and [ H ] (Equations (2.71) and (2.72), respectively).
e

Further discussions on the total and updated Lagrangian formulations will be given in a
continuum context in Chapter 3.

2.5.4 Finite element implementation of non-linear contact

Until now, when we have introduced the kinematically linear contact problem, we have considered
two broad classes of contact problems, rigid-to-flexible and rigid-to-rigid. However, both types can
be treated simultaneously since the problem of contact of a deformable body with a rigid obstacle
might be considered as a subcase of the two deformable body contact problem. In fact, in the
methodology applied previously for linear problems, the main difference is about the configuration

60 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

of matrix [E]. In this way, the most key details associated with either problem class can be remarked
in a unified framework and that will be the procedure we will follow for the non-linear contact
problem. Furthermore, to give more generality we will assume contact with friction from the start.
The weak formulation of the frictional contact problem (Equations (2.35), (2.42) and (2.49)) is
characterized by the introduction of a new term representing the virtual work due to the contact
forces. According to that, Equation (2.53) might be written for a contact problem as follows

{r ( { d } , λ N , λT ) =} {
{FEXT } + FC ({d } , λN , λT ) − FINT ({d }) =
{0} } { } (2.94)

{ }
where FC ({d } , λN , λT ) groups all the contributions to the virtual work due to the contact interface
and the Lagrange multipliers, λN and λT, denote the reactive forces RN and RT , respectively.
Depending on the adopted constraint enforcing method, { F ( {d } , λ
C N }
, λT ) will change and will
depend on one or more variables among {d } , λN and λT . In the particular case of the kinematically
linear contact problems presented in Section 2.1, if the Lagrange multiplier method (Equation (2.36))
{
is used, FC ({d } , λN , λT ) will be }
λ 
{F ( λ
C N , λT )} = − [ EN ]( t
[ ET ] )  λN 
t

 T 
(2.95)

For the other three methods, penalty, perturbed Lagrangian and augmented Lagrangian methods, we
will have from (2.43), (2.36) and (2.51), respectively, the following expressions

{F ({d})} = ([ E ] β {Q} + [ E ] β {Q}) − ([ E ] β [ E ] + [ E ] β [ E ]){d}

C N
t
T
t
N
t
N T
t
T (2.96)

λ 
{F ( λ
C N , λT )} = − [ EN ]( t
[ ET ] )  λN 
t

 T 
(2.97)

{ F ( {d } , λ
C N )
, λT = } ([ E ] β {Q} + [ E ] β {Q} + [ E ] λ + [ E ] λ ) −
N
t
T
t
N
t
N T
t
T
(2.98)
([ EN ] β [ EN ] + [ ET ] β [ ET ]){d}
t t

Non-linearity of (2.94) is due mostly to the terms {F ({d })}

INT and { F ( {d } , λ
C N }
, λT ) . The
internal force vector, as seen in previous sections, contains both geometric and material
nonlinearities. The contact force vector contains all the non-linearities evaluated on the contact
surfaces. These non-linearities can be due to the contact constraints or the friction. By similarity with
(2.54), the linearization of (2.94) can be carried out in the following way

−
{
∂ r ({d } , {λ} ) } ∆ {d } − ∂ {r ({d } , {λ})} ∆ {λ} =
−
∂ { F ({d } , {λ} )}
∆ {d } −
C

∂ {d } ∂ {λ} ∂ {d }
(2.99)
{
∂ FC ({d } , {λ} ) } ∆ {λ} + ∂ {F ({d })} ∆ {d } =
∂ {λ} ∂ {d }
{r ({d } , {λ})}
INT

where all the Lagrange multipliers have been grouped in the vector {λ} .
In (2.99), the term

Module: Contact Non-linearities 61

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

[ KT ] =
{
∂ FINT ({d } ) } (2.100)
∂ {d }
corresponds to the tangent stiffness matrix and is due to the material and geometric non-linearities
while the terms
∂ { FC ({d } , {λ} )} ∂ { FC ({d } , {λ} )}
[ K C1 ] =
− [ KC 2 ] =
− (2.101)
∂ {d } ∂ {λ}
are the contact stiffness matrices.
From (2.100) and (2.101), Equation (2.99) might be written as follows

T C1 {r ({d } ,{λ})}
([ K ] + [ K ]) ∆ {d } + [ K ] ∆ {λ} = C2
(2.102)

Once [ KT ] , [ K C1 ] and [ K C 2 ] have been evaluated the Newton-Raphson incremental-iterative

procedure might be applied to obtain the solution of the non-linear problem. However, in case of
using the Lagrange multiplier method or any of its variations, perturbed and augmented Lagrange
methods, the linearization of (2.38) or (2.47) should be included too in the procedure

{ } ∆ {d } − ∂ {G ({d } ,{λ})} ∆ {λ} =

∂ G ({d } , {λ} )
−
∂ {d } ∂ {λ}
{G ({d } ,{λ})} (2.103)

Combining (2.102) and (2.103), the resulting equations corresponding to the linearization of the
problem are
  KT  +  K C1   K C 2  
    
{
  ∆ {d }   r ({d } , {λ} )} 

{
 ∂ G ({d } , {λ} ) } {
∂ G ({d } , {λ} ) } 
 ∆ {λ } =
{ } 
  G ({d } , {λ} )
(2.104)
− ∂ {d }
−
∂ {λ} 
 
The computation of [ K C1 ] and [ K C 2 ] as well as the derivatives of {G ({d } , {λ} )} will depend on the
method used to enforce the constraints. In case of using the penalty method the terms in
{G ({d } , {λ})} will not be present in (2.104).
Of particular concern in the solution of contact problems is the capacity of the algorithm to
converge when complex geometries, deformations and contact conditions are analyzed. It will be
addressed in further chapters.

EXAMPLE 2.22 Example 2.4 of the Module “Fundamentals and Applications of Finite Element in
Mechanical Analysis” shows an inextensible wire of cross section A subjected to a tensile force T which
balances a distributed loading b(x) [F/L] (Figure 2.66) whose strong formulation for small deformation is
given by
−Tu, xx + b( x) =
0 (a)

u( =
x 0)= 0 (b)
u(=
x L=
) 0 (c)

62 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

b(x)

Figure 2.66

We suppose now that there is a flat rigid obstacle at a small distance d below (Figure 2.67), in such a way
that when the wire deflects downward due to the applied loading, it is stopped when contacts the obstacle.
Compute the solution of the problem.

b(x)
T

x1
L

Figure 2.67

Solution: With this example, we will demonstrate the non-linearity of the contact problem even if we
consider geometric and material linearity.
Before contacting with the obstacle, the solution of the linear problem is obtained by direct integration of
Equations (a), (b) and (c) is given by
u=
−bx( L − x) / (2T ) (d)
Once the wire contacts the obstacle, an impenetrability constraint should be added to the governing
equations
−Tu, xx + b=
( x) 0 0 ≤ x ≤ x1 (e)

u =−d x1 ≤ x ≤ L / 2 (f)
u(=
x 0)= 0 (g)
u=
,x (x L=
/ 2) 0 (h)

where we have assumed that contact occurs at an unknown point x1 ≤ L / 2 and that the wire keeps lying
on the obstacle for x1 ≤ x ≤ L − x1 . Equations (e)-(g) have been stated only for the left half region of the
wire ( 0 ≤ x ≤ L ) due to the symmetry conditions.
Solution of (e)-(h) is obtained by direct integration
=u bx 2 / 2T + cx 0 ≤ x ≤ x1 (i)

u =−d x1 ≤ x ≤ L / 2 (j)
In (i) and (j) we have two unknowns, c and x1, to be determined. For it, we assume the continuity of the
displacement and its derivative at x1, which is reasonable for inextensible wires. We have

Module: Contact Non-linearities 63

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

u ( x1− ) =
u ( x1+ ) ⇒ bx12 / 2T + cx1 =
− d (k)

u, x (=
x1− ) u, x ( x1+ ) ⇒ bx1 / T=
+ c 0 (l)

and, therefore
c=
−bx1 / T =
− 2bd / T (m)

x1 = 2Td / b (n)
If we compare now the solution of the constrained (contact) and unconstrained problems the first
conclusion is that the unconstrained system is linear and, therefore, the displacement (Equation (d)) is
proportional to the load. However, the solution (i)-(n) of the constrained problem is not proportional to the
load since the contact point x1 depends on the load nonlinearly. The contact problem is, therefore, non-
linear even if the governing equation does not contain a non-linear term and the displacements are
considerably small. This type of non-linearity is due to the contact.
The second conclusion is that the solution and its first derivative of the contact problem are continuous at
the contact point x1; however, unlike the solution of the unconstrained problem, the second derivative is
discontinuous at x1. Therefore, the contact condition reduces the smoothness of the solution.

2.6 IMPACT

When two bodies with different velocities come into contact an impact occurs. Impact problems are
typical problems of Contact Mechanics. Although these problems will not be treated with detail in
this text, at least, we will introduce them.
If we consider, for example, the impact of a bar moving in axial direction with a velocity v0
against a rigid barrier perpendicularly placed at a distance g ahead of the rod (Figure 2.68), the
solution of the problem can be derived from the following governing equations
∂ 2u N ∂ 2u N
EA = − ρ A (2.105)
∂x 2 ∂t 2
where ρ is the density of the bar.

v0
x

L g

Figure 2.68. Impact of a bar against a rigid obstacle

Additionally, the problem has to fulfill the initial and boundary conditions, the impenetrability
contact condition and, furthermore, the contact force must be a compressive force. We have

64 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

∂u N
EA (t, 0) = 0
∂x
uN ( t , L ) ≤ 0
RN ( t , L ) ≤ 0
RN ( t , L )  g − u N ( t , L )  =
0 (2.106)

u N ( 0, x ) = 0
∂u N
( 0, x ) = v0
∂t
The end x = 0 is free of load while the boundary conditions at the end point x = L are typical for
contact problems. The fourth equation is the Kuhn-Tucker condition derived from the contact. The
last two equations are the initial conditions.
The solution of (2.105) in combination with (2.106) gives the velocity of the bar after the impact
as well as the impact force as a function of time. Numerical methods designed for short time
responses, such as the finite element method, are especially suitable to solve impact problems.

EXAMPLE 2.23 Using ANSYS workbench solve the low speed impact problem of a circular rod moving
in axial direction with a uniform initial velocity of 500 mm/s against a rigid wall placed at a distance of 50
mm. The rod has a length of 200 mm and a diameter of 16 mm.
Solution: Typìcal crash problems involve impact and rebound of two objects. These problems are difficult
and computationally intense. We will demonstrate it with this easy example.
First of all, we define the geometry. It requires to use one line for the rod and one solid to define the rigid
wall. The wall has been modeled with a cube of side length equal to 100 mm. The geometry has been built
in SpaceClaim (Figure 2.68).

Figure 2.68

Once the geometry has been built, we open Mechanical and define the behavior of the block as rigid
(Stiffness Behavior=Rigid in Model > Geometry > Solid ) (Figure 2.69).

Module: Contact Non-linearities 65

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.69

The contact pair (right-click Insert > Manual Contact Region in Model > Connections) between the bar
(Contact body) and the rigid surface (Target body) is defined as shown in Figure 2.70. We define a
frictionless contact (Type = Frictionless) and activate a pinball region (Pinball Region = Radius). Then,
the contact status will be checked only in the pinball region which will be a sphere defined by its radius
(Pinball Radius). We suggest the students to solve the problem using smaller values of the pinball region
in order to compare the contact status at each case.

Figure 2.70

In the next step we apply the boundary conditions as usually (zero remote displacement for the rigid block
and constrained X and Y displacements at the two vertices of the line body). As the problem will be solved
66 Chapter 2: Contact Mechanics in 1D Elastostatics Problems
 Master’s Degree in Numerical Simulation in Engineering with ANSYS

using a transient analysis an initial condition of velocity must be specified (Equation 2.106) along the main
axis of the line (Z direction) (Figure 2.71).

Figure 2.71

The analysis settings (Model > Transient > Analysis Settings) are shown in Figure 2.72. We have used an
automatic time stepping (Auto Time Stepping = On) with a maximum time step of 0.01 s (Maximum Time
Step). Finally, we turn on all the output controls. Now the problem is fully defined and we can solve it.

Figure 2.72

Module: Contact Non-linearities 67

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.73 shows the displacement of the bar. A rebound of the rod against the wall is observed.

Figure 2.73

Figures 2.74 and 2.75 show the contact status and the contact gap respectively.

Figure 2.74

68 Chapter 2: Contact Mechanics in 1D Elastostatics Problems

 Master’s Degree in Numerical Simulation in Engineering with ANSYS

Figure 2.75

Module: Contact Non-linearities 69

www.ansys.com/mastersdegree

All rights reserved. No part of this publication may be

reproduced, stored in a database or retrieval system or
published in any form or in any way without the prior written
permission of the authors.

Copyright  Ricardo Perera and Alberto Carnicero, 2020