You are on page 1of 62

Project:

FENET EU Thematic Network


(Contract G1RT-CT-2001-05034)
FENET RTD (Durability & Life Extension)

Report Title:

Advanced Finite Element Contact


Benchmarks

Author:

A.W.A. Konter
Netherlands Institute for Metals Research

Date:

20 July 2005

Report No:

FENET-UNOTT-DLE-09

FENET-Advanced Contact Benchmarks

1/62

FENET Foreward
Contact analysis has been identified as an important area of development of FE
technology in the FENET EU thematic network (2001-2005). A series of four
workshops dedicated to contact analysis were launched under the Durability and Life
Extension RTD, chaired by Professor A.A. Becker (University of Nottingham).
The first FENET contact workshop (27-28 February 2002, Copenhagen) was focussed
on the current issues regarding the FE simulation of contact problems. The workshop
stimulated many discussions regarding difficulties experienced by FE users, current
limitations of commercial FE software, desirable contact features not currently being
offered by FE software, and the need for further research in FE contact analysis.
It was agreed that there is a need for developing advanced contact benchmarks
through the FENET network. Further FENET workshops were launched to devise
new advanced contact benchmarks. An invitation was issued to all FENET members
to suggest new potential contact benchmarks, and a new FENET Contact Working
Group was formed to discuss the merits of the benchmarks.
In the second FENET contact workshop (25 March 2004, Majorca), the requirements
for advanced contact benchmarks were discussed and a list of new advanced contact
benchmarks was proposed. The merits and disadvantages of each of the candidate
contact benchmarks were evaluated, and it was agreed to concentrate on only five
advanced contact benchmarks.
It was recognized that the dimensions and the material properties will play an
important role in highlighting the relevant features of the contact benchmarks.
Therefore, further FE analyses were performed to establish the geometric parameters,
material constants, values of the applied loads and the coefficient of friction. The task
of running the benchmarks was assigned to A.W.A. Konter (Netherlands Institute for
Metals Research).
Two further workshops on the FENET contact benchmarks were held; to discuss
comments and solutions received from various FENET members (7 October 2004,
Glasgow) and to discuss the final FE solutions (25 February 2005, Budapest).
This report is the final FENET report on the advanced contact benchmarks, and will
be subsequently released as a NAFEMS document. We would like to acknowledge
the support of the FENET Contact Working Group and the many individuals who
attended the four contact workshops and provided useful feedback on the contact
benchmarks.

Professor A.A. Becker


University of Nottingham, UK
(Coordinator- FENET Durability and Life Extension)

FENET-Advanced Contact Benchmarks

2/62

Table of contents
1.

Introduction.........................................................................................................4

2.

Benchmark 1: 2D Cylinder Roller Contact ......................................................6


2.1
Summary ...................................................................................................6
2.2
Introduction...............................................................................................7
2.3
Requested solutions ..................................................................................7
2.4
Analytical solution ....................................................................................7
2.5
FEM Solutions ..........................................................................................8
2.6
Modelling tips .........................................................................................13
2.7
Parameter study.......................................................................................14

3.

Benchmark 2: 3D Punch (Rounded Edges) ....................................................15


3.1
Summary .................................................................................................15
3.2
Introduction.............................................................................................16
3.3
Requested solutions ................................................................................16
3.4
FEM solutions.........................................................................................16
3.5
Modelling tips .........................................................................................25
3.6
Parameter study.......................................................................................25

4.

Benchmark 3: 3D Sheet Metal Forming .........................................................26


4.1
Summary .................................................................................................26
4.2
Introduction.............................................................................................28
4.3
Required solutions ..................................................................................28
4.4
Experimental results................................................................................29
4.5
FEM solutions.........................................................................................29
4.6
Modelling tips .........................................................................................39
4.7
Parameter study.......................................................................................40

5.

Benchmark 4: 3D Loaded Pin..........................................................................41


5.1
Summary .................................................................................................41
5.2
Introduction.............................................................................................42
5.3
Required solutions ..................................................................................42
5.4
FEM solutions.........................................................................................42
5.5
Modelling tips .........................................................................................53
5.6
Parameter study.......................................................................................53

6.

Benchmark 5: 3D Steel Roller on Rubber ......................................................54


6.1
Summary .................................................................................................54
6.2
Introduction.............................................................................................55
6.3
Required solutions ..................................................................................55
6.4
FEM solutions.........................................................................................56
6.5
Modelling tips .........................................................................................59
6.6
Parameter study.......................................................................................59

7. Concluding Remarks ............................................................................................60


References....................................................................................................................62

FENET-Advanced Contact Benchmarks

3/62

1. Introduction
NAFEMS has published a survey on FE Analysis of contact and friction [1] and a
booklet on how to undertake contact and friction analysis [2]. A contact benchmarks
report on two-dimensional problems was published as the first step in establishing a
set of FE Contact benchmarks [3]. It was acknowledged that although the current
published NAFEMS benchmarks were limited in scope, they were important as the
first step in establishing contact benchmarks.
A small FENET Working Group on Contact has been assembled with the
collaboration of the NAFEMS Non-linear Working Group. Following discussions on
the development of new advanced benchmarks, it was agreed to concentrate on only 5
contact benchmarks, as follows
2D Contact of cylindrical roller
3D Punch (Rounded edges)
3D Sheet metal forming
3D Loaded pin
3D Steel roller on rubber
The selected contact benchmarks exhibit the following features:
- 3D contact
- Frictional stick-slip in contact area
- 2D/3D Linear versus quadratic elements
- Shell contact
- Large strain contact
- Metal forming
- Mesh dependency
- Compression of rubber
- Rolling contact
Further FE analyses were performed to establish the geometric parameters, material
constants, values of the applied loads and the coefficient of friction.
The current report presents the results of the FE analyses performed on 2D and 3D
approximations of the proposed problems. Since all proposed benchmarks can be
reasonably well approximated with a 2D or an axisymmetric solution, all target results
presented here have been obtained with a 2D or an axisymmetric FE analysis. In
addition, 3D analyses have been performed and the results have been compared with
the initial 2D solutions (with the exception of Contact Benchmark 1).
Frequently, reports on results of benchmark analyses present numerical solutions for
selected problems, generated by different users using different FE codes. Often the
selection of different numerical input parameters by the analyst is not presented in the
report and the analyst presents his best choice of the generated solutions for a
particular problem.
As a result the effect of specific parameters, such as applied mesh density, element
type, contact parameter settings, number of loading steps etc. is difficult to quantify
and it is not clear whether obtained differences are caused by differences in the
applied FE code or differences in user input. The results presented in this report not
FENET-Advanced Contact Benchmarks

4/62

only compare results of different FE codes using as closely as possible identical input
parameters, but also show the effect of variations in these parameters.

FENET-Advanced Contact Benchmarks

5/62

2.

Benchmark 1: 2D Cylinder Roller Contact

2.1

Summary

Ref. No.

Contact Benchmark - 1

Title

2D Cylinder Roller Contact

Contact
features

Advancing contact area


Curved contact surfaces
Deformable-deformable contact
Friction stick-slip along the contact line
Comparison of linear and quadratic elements
2D plane strain
Block height = 200 mm
Block width = 200 mm
Cylinder diameter =100 mm

Geometry

Material
properties

E punch = 210 kN / mm 2
E foundation = 70 kN / mm2

punch = foundation = 0.3


Analysis type Linear elastic material
Geometric non-linearity
Non-linear boundary conditions
Displacement Symmetry displacement constraints (half symmetry)
boundary
Bottom surface of the foundation is fixed (u x = u y = 0)
conditions
Applied
Vertical point load F = 35 kN
loads
Element type 2D plane strain
8 node quadratic elements
4 node linear elements
Contact
2 different cases:
properties
coefficient of friction = 0.0
coefficient of friction = 0.1
FE results
1. Plot of contact pressure against distance from centre of contact
2. Plot of tangential stress against distance from centre of contact
3. Plot of relative tangential slip against distance from centre of
contact

FENET-Advanced Contact Benchmarks

6/62

2.2

Introduction

A steel cylinder is pressed into an aluminium block. It is assumed that the material
behaviour for both materials is linear elastic. The cylinder is loaded by a point load
with magnitude F = 35 kN in the vertical direction. A 2D approximation (plane
strain) of this problem is assumed to be representative for the solution. An analytical
solution for the frictionless is known.

2.3

Requested solutions

Two solutions, one using friction coefficient 0.1 between the cylinders and one
frictionless solution, are requested for:
- length of contact zone
- pressure distribution as function of arc-length along contact surface
- tangential stress distribution as function of arc-length along contact
The solutions presented include:
- Element size, in particular near the contact zone
- Method for normal contact detection (penalty stiffness) or hard/direct contact
- Indication which surface is treated as master (contacting) and slave (contacted)
- Method for stick slip approximation (true stick-slip, penalty based, other)
The analysis results are presented with linear and quadratic elements.

2.4

Analytical solution

An analytical solution for this contact problem can be obtained from the Hertzian
contact formulae [4] for two cylinders (line contact). The maximum contact pressure
is given by:

pmax =

Fn E*
2 BR*

where Fn is the applied normal force, E * the combined elasticity modulus, B the
length of the cylinder and R* the combined radius.
The contact width 2a is given by:

8Fn R*
a=
.
BE*
Using the normalised coordinate = x / a with x the Cartesian x-coordinate, the
pressure distribution is given by:
p = pmax 1 2

The combined elasticity modulus is determined from the modulus of elasticity and
Poissons ratio of the cylinder and block E1 , 1 and E2 , 2 , as follows:
FENET-Advanced Contact Benchmarks

7/62

E* =

2 E1 E2
E2 (1 12 ) + E1 (1 22 )

The combined radius of curvature is evaluated from the radius of curvature of the
cylinder and block R1 and R2 , as follows:
RR
R* = 1 2
R1 + R2
For the target solution, the block is approximated with an infinitely large radius. The
combined radius is then evaluated as:
RR
R* = lim 1 2 = R1
R2 R + R
1
2
Using the numerical parameters for the problems the following results are obtained:
a = 6.21 mm
pmax = 3585.37 N/mm
Note that half the contact length is equal to 6.21 mm which corresponds to
approximately 7.1 degrees of the ring. Hence it is clear that in order to simulate this
problem correctly a very fine mesh near the contact zone is needed.

2.5

FEM Solutions

A set of numerical solutions has been obtained with the FE programs


Abaqus/Standard and MSC.MARC. Typical element meshes applied for the
determination of the target solutions are shown in Figure 1. For the solution generated
with MSC.MARC the smallest element edges near the contact zone have been chosen
as 0.59 mm for the cylinder and 1.5 mm for the block. The block consisted of 1400
elements and the cylinder of 1431 elements. The solution obtained with
Abaqus/Standard used approximately the same mesh density. A slightly different
mesh for the cylinder was generated (1703 elements): the same seed points along the
curved edges were chosen, the applied advancing front mesher in both codes resulted
in slightly different meshes.
Numerical solutions have been obtained with plane strain linear elements using
reduced integration and with fully integrated quadratic plane strain elements (applied
nominal thickness 1 mm) . The applied elements types for the listed FE codes are
listed in Table 1.
Table 1 Applied element types in numerical solutions (Benchmark 1)

linear
quadratic

Abaqus/Standard
Type CPE4R
Type CPE8

MSC.MARC
Type 115
Type 27

In the contact algorithm, hard contact (i.e. based on direct coupling of the
displacements using automatically generated constraint equations) has been used. For
the simulations with friction either a true stick-slip model (MSC.MARC) or the
Lagrangian multiplier method (Abaqus/Standard) has been selected. The slave nodes
FENET-Advanced Contact Benchmarks

8/62

(contacting nodes) have been set to the nodes on the cylinder; the master nodes
(contacted nodes) have been set to the upper edge of the block.
The obtained lengths of the contact zones is listed in Table 2. The exact length of the
contact zone cannot be determined due to the discrete character of contact detection
algorithms (nodes are detected to be in contact with an element edge). It is clear
however that the numerical solution is in good agreement with the analytical one.
Table 2 Length of the contact zone (Benchmark 1)

linear
quadratic

Abaqus/Standard
5.88 < a < 6.42
5.88 < a < 6.18

MSC.MARC
5.89 < a < 6.42
5.88 < a < 6.17

Figure 1 Element mesh applied in target solution with MSC.MARC (Benchmark 1)

FENET-Advanced Contact Benchmarks

9/62

Figure 2 Deformed structure plot at maximum load level (magnification factor = 1)


(Benchmark 1)

The deformed structure plot (magnification factor 1.0) is shown in


Figure 2. A plot of the Hertzian contact solution for the pressure and the solutions
along the nodes of the cylinder obtained with linear and quadratic elements with two
different FE codes is shown in Figure 3.
Benchmark 1 - no friction
5000

Analytical
MSC.MARC - linear
MSC.MARC - quadratic
Abaqus - linear
Abaqus - quadratic

Contact pressure [N/mm2]

4500
4000
3500
3000
2500
2000
1500
1000
500
0
0

arc length [mm]


Figure 3 Comparison of analytical and numerical solutions (Benchmark 1)

The contact pressure is plotted for the slave (contacting) nodes and shows even with
this applied mesh density a rather oscillating type of behaviour. Generating the same
plots along the nodes of the master (contacting) nodes produces a smoother curve. A
comparison of the calculated contact pressure along the cylinder node path and along
the block node path is shown in Figure 4.

FENET-Advanced Contact Benchmarks

10/62

Benchmark 1 - no friction
5000

Analytical
linear - cylinder
quadratic - cylinder
linear - block
quadratic - block

Contact pressure [N/mm2]

4500
4000
3500
3000
2500
2000
1500
1000
500
0
0

arc length [mm]

Figure 4 Calculated contact pressure using different node paths [MSC.MARC] (Benchmark 1)

Numerical solutions have been obtained for a contact analysis with a friction
coefficient 0.1 (true stick-slip modelling). The contact normal stress and the contact
tangential stress along the nodes of the circle obtained with MSC.MARC have been
plotted in Figure 5. The same results obtained with Abaqus/Standard have been
plotted in Figure 6.
Benchmark 1 - friction coefficient 0.1
5000

pressure_linear
pressure_quadratic
tangential_linear
tangential_quadratic

4500

stress [N/mm2]

4000
3500
3000
2500
2000
1500
1000
500
0
0

10

x-coordinate [mm]
Figure 5 Contact normal and contact tangential stress [MSC.MARC] (Benchmark 1)

FENET-Advanced Contact Benchmarks

11/62

Benchmark 1 - friction coefficient 0.1


5000

pressure_linear
pressure_quadratic
tangential_linear
tangential_quadratic

4500

stress [N/mm2]

4000
3500
3000
2500
2000
1500
1000
500
0
0

x-coordinate [mm]
Figure 6 Contact normal stress and contact tangential stress [Abaqus/Standard] (Benchmark 1)

The tangential slip at the nodes of the cylinder obtained with Abaqus/Standard have
been plotted in Figure 7.
Benchmark 1 - friction coefficient 0.1
0.025
linear
quadratic

slip [mm]

0.020
0.015
0.010
0.005
0.000
0

arc length [mm]


Figure 7 Slip long nodes of cylinder [Abaqus/Standard] (Benchmark 1)

All path plot results show an oscillating type of behaviour. This can be improved by
refining the mesh in the contact zone (and the surrounding part assuring connection
with the remaining part of the structures.
FENET-Advanced Contact Benchmarks

12/62

2.6

Modelling tips

Although generally accepted as a good problem for testing various contact algorithms,
the numerical solution of the Hertzian contact problem is a challenging one. This is
caused by the fact that contact is only present in a small zone, thus requiring a fine
mesh density. The following are some guidelines and tips for modelling this
benchmark:
For the given numerical parameter combinations, the length of the contact
zone can be determined analytically and is plotted as function of the applied
normal force in
8
7
6

a [mm]

5
4
3
2
1
0
0.E+00 1.E+04 2.E+04 3.E+04 4.E+04 5.E+04 6.E+04

force [N]

Figure 8.
Figure 8 Half length of the contact zone versus applied normal force (Benchmark 1)

For a reasonable contact description, at least 10 contacting nodes should be


present along the slave (contacting) surface. If stick-slip is to be considered in
the analysis this number has to be increased for an accurate description of the
contact zone. Ensuring correct connectivity with the remaining part of the
structure makes the meshing of this problem more difficult than the actual
contact analysis
It is recommended to select the curved surface as the slave (contacting)
surface and the straight surface as the master (contacted) surface. The slave
(contacting) nodes will be positions on the edge between the master
(contacted) surface and this edge is frequently assumed to be straight (linear
elements) or approximated (quadratic elements).
The numerical solution is strongly dependent upon the absence of penetration.
This is ensured in the hard contact algorithm, while penalty based algorithms
cause a specific amount of penetration depending upon the numerical value of
the applied penalty stiffness.

FENET-Advanced Contact Benchmarks

13/62

2.7

Parameter study

It is interesting to generate additional solutions for:


Effect of changing the contact body detection (master and slave)
Effect of different penalty settings for penalty based contact
Effect of different friction approximations

FENET-Advanced Contact Benchmarks

14/62

3.

Benchmark 2: 3D Punch (Rounded Edges)

3.1

Summary

Ref. No.

Contact Benchmark - 2

Title

3D Punch (Rounded edges)

Contact
features

3D contact
Stick/slip behaviour along the contact plane
Comparison of linear and quadratic elements
(Plasticity may be considered)
3D Continuum elements (can also be modelled as axisymmetric)

Geometry

Material
properties

Punch diameter = 100 mm


Punch height = 100 mm
Foundation diameter = 200 mm
Foundation height = 200 mm
Fillet radius at the edge of the punch contact = 10 mm

E punch = 210 kN / mm 2
E foundation = 70 kN / mm2

punch = foundation = 0.3


Analysis type Quasi-static analysis using
Linear elastic material
Geometric non-linearity
Non-linear boundary conditions
Displacement Symmetry displacement constraints (quarter symmetry)
boundary
Bottom surface of the foundation is fixed (u x = u y = u z = 0)
conditions
Applied
A uniform pressure (distributed load) applied at the top surface of the
loads
punch P = 100 N / mm2
Element type 3D continuum
20-node quadratic elements
27-node quadratic elements
or 8-node linear elements
Contact
2 different cases:
properties
coefficient of friction = 0.0
coefficient of friction = 0.1
FE results
1. Plot of contact pressure against radial distance from the centre of
contact
2. Plot of tangential stress against radial distance from the centre of
contact
3. Plot of relative tangential slip against distance from the centre of
contact

FENET-Advanced Contact Benchmarks

15/62

3.2

Introduction

An axisymmetric steel punch is compressed on an aluminium cylinder. It is assumed


that the material behaviour is linear elastic. The punch is loaded by a uniform pressure
with magnitude P = 100 N / mm2 in axial direction. The effect of friction is taken into
account along the contact zone. In 3D analyses coarser element meshes are applied
than in 2D simulations, in order to keep the computer time within reasonable limits.
Since the problem is completely axisymmetric 2D solutions can be used to serve as
target solution for the subsequent 3D analysis. Initial 2D solutions with meshes,
representative for the subsequent 3D simulations are used to determine the target
solutions. It is sufficient to model one quarter of the assembly for the 3D solutions,
applying symmetry conditions.

3.3

Requested solutions

Both 2D (axisymmetric) and 3D solutions are requested. With the axisymmetric


modelling the effect of stick-slip along the contact surface will be investigated. In
particular the effect of applied smoothing techniques in stick-slip modelling, such as
true-stick slip model or Lagrange multipliers, penalty based stick-slip or other
smoothing mechanisms is explored.
In the 3D modelling the effect of symmetry conditions in combination with contact
detection and the effect of linear and quadratic elements are evaluated.
Both for the axisymmetric and 3D analysis solutions are requested for the
displacement and stress field along the top surface of the punch, obtained with:
- Linear and quadratic elements
- Different methods for friction modelling
The results include the following:
- Element sizes near the contact zone
- Radial and axial displacement of top contact surface of cylinder as function of
coordinate
- Method used for normal contact detection (penalty stiffness) or hard/direct contact
- Indication which surface is treated as master (contacting) and slave (contacted)
- Method for stick slip approximation (true stick-slip, penalty based, other)

3.4

FEM solutions

Numerical solutions for an axisymmetric model have been obtained with identical
relatively coarse meshes with Abaqus/Standard and MSC.MARC. Typical element
meshes are shown in Figure 9. Typical element lengths along the punch are 4 mm and
along the foundation 3.5 mm.

FENET-Advanced Contact Benchmarks

16/62

Figure 9 FE meshes applied in target solution (Benchmark 2)

Fully integrated elements have been applied in the numerical axisymmetric


simulations with MSC.MARC and Abaqus/Standard. The applied element types are
listed in Table 3.
Table 3 Applied element types in numerical solutions (Benchmark 2)

linear
quadratic

Abaqus/Standard
Type CAX4
Type CAX8

MSC.MARC
Type 10
Type 28

The straight edge of the foundation is selected as the master surface (contacted edge),
the nodes on the bottom edge of the punch are selected as the slave surface
(contacting nodes). Using this choice, the nodes always touch a straight segment in
the contact detection mechanism. The direct method for normal contact detection is
chosen (constraints on the displacements for the nodes which are found to be in
contact). The initial solution is obtained with the true stick slip implementation in
MSC.MARC and the Lagrangian multiplier method implementation for tangential
contact in Abaqus/Standard. Numerical solutions are presented for 2 values of the
friction coefficient ( = 0 and = 0.1 ), different friction implementations and using
different element types for 2 different FE codes.
The axial displacement as a function of the radial coordinate along the top surface of
the foundation, obtained with MSC.MARC using linear axisymmetric, fully integrated
elements, is shown in Figure 10. The analysis with frictions shows a slightly stiffer
behaviour.

FENET-Advanced Contact Benchmarks

17/62

0.00
0

10

20

30

40

50

60

70

80

90

100

-0.02

Axial displacement [mm]

-0.04

-0.06

no friction
friction 0.1

-0.08

-0.10

-0.12

-0.14
radius [mm]

Figure 10 Axial displacement as a function of the radial coordinate (friction coefficient 0.0 and
0.1) obtained with linear elements [MSC.MARC] (Benchmark 2)

Radial displacement as a function of the radial coordinate at the top edge of the
foundation, from the same analyses, are shown in Figure 11. In the frictionless
solution all points move inwards, while for the solution with friction the points closest
to the axis move outwards.
0.002

0.000
0

10

20

30

40

50

60

70

80

90

100

-0.002

Radial displacement [mm]

-0.004

-0.006

-0.008

-0.010
no friction
friction 0.1

-0.012

-0.014

-0.016

-0.018
radius [mm]

Figure 11 Radial displacement as function of the radial coordinate (friction coefficient =0.0 and
0.1) obtained with linear elements [MSC.MARC] (Benchmark 2)

Comparisons of both the axial and the radial displacements obtained with different
element types and FE codes are shown in Figure 12 and Figure 13. The differences
FENET-Advanced Contact Benchmarks

18/62

between the results from different FE codes are smaller than the difference due to
different element types.
0.00
0

10

20

30

40

50

60

70

80

90

100

-0.02

Axial displacement [mm]

-0.04

-0.06

-0.08
linear MARC
quadratic MARC
linear Abaqus
quadratic Abaqus

-0.10

-0.12

-0.14
radius [mm]

Figure 12 Axial displacement along top surface of foundation (Benchmark 2)


0.002

0.000
0

10

20

30

40

50

60

70

80

90

100

-0.002

Radial displacement [mm]

-0.004

-0.006

-0.008

-0.010

linear MARC
quadratic MARC
linear Abaqus
quadratic Abaqus

-0.012

-0.014

-0.016

-0.018
radius [mm]

Figure 13 Radial displacement along top surface of foundation (Benchmark 2)

The effect of different methods for handling the strong non-linearity of the stick-slip
condition is evaluated with Abaqus/Standard using linear elements. The closest
approximation of the stick-slip condition is obtained with the Lagrangian multiplier
technique while often in numerical simulation a penalty formulation is used due to the
better convergence characteristics. The resulting radial displacements are shown in
Figure 14.
FENET-Advanced Contact Benchmarks

19/62

Friction approximation effect


0.002
0.000
0

10

20

30

40

50

60

70

80

90

100

-0.002

radial disp [mm]

-0.004
-0.006
-0.008
Lagrange
Penalty
no friction

-0.010
-0.012
-0.014
-0.016
-0.018
radial coordinate [mm]

Figure 14 Effect of different friction coefficient and method of friction handling on the radial
displacement of the foundation edge [Abaqus/Standard linear elements] (Benchmark 2)

The calculated contact stress, obtained with the linear elements (true stick-slip model
with friction coefficient 0.1) with both FE codes are displayed in Figure 15. For the
same analyses the contact tangential stresses are displayed in Figure 16. Relatively
small differences between the results obtained with both codes are present.
In Figure 17 and Figure 18 a comparison is made of the contact stresses for the nodes
along the punch (slave or contacting nodes) and along the foundation (master or
contacted edges/nodes). The quadratic elements show an oscillating type of
behaviour.
The computed contact stresses and contact tangential stresses along the foundation
edge using different 2D element types and linear 3D elements have been compared in
Figure 19 and Figure 20. For the 3D solution a 90o expanded segment of the mesh
using linear elements, as shown in Figure 21 is used. Figure 22 shows that the
obtained solution for the contact stress is completely axisymmetric.
The computed tangential slip is plotted for the axisymmetric mesh with linear and
quadratic elements in Figure 23.

FENET-Advanced Contact Benchmarks

20/62

axi-symmetric - linear elements


450

contact stress [N/mm2]

400
MSC.MARC
Abaqus

350
300
250
200
150
100
50
0
0

10

20

30

40

50

60

arc length [mm]

Figure 15 Comparison of contact stress along punch using linear elements (Benchmark 2)

axi-symmetric - linear elements


45
40
MSC.MARC
Abaqus

shear stress [N/mm2]

35
30
25
20
15
10
5
0
-5

10

20

30

40

50

60

arc length [mm]

Figure 16 Comparison of contact tangential stress along punch using linear elements
(Benchmark 2)

FENET-Advanced Contact Benchmarks

21/62

450

normal pressure [N/mm2]

400
MARC_linear_foundation

350

MARC_linear_edge

300
250
200
150
100
50
0
0

10

20

30

40

50

position from axis [mm]


Figure 17 Comparison of contact stresses along punch and foundation using linear elements
(Benchmark 2)

700

normal pressure [N/mm2]

600

MARC_quadratic_foundation
MARC_quadratic_punch

500
400
300
200
100
0
0

10

20

30

40

50

position from axis [mm]


Figure 18 Comparison of contact stresses along punch and foundation using quadratic elements
(Benchmark 2)

FENET-Advanced Contact Benchmarks

22/62

400

normal pressure [N/mm2]

350

MARC_linear
MARC_quadratic
MARC_linear3D

300
250
200
150
100
50
0
0

10

20
30
position from axis [mm]

40

50

Figure 19 Comparison of contact stress along foundation using 2D linear and quadratic elements
and 3D linear elements [MSC.MARC] (Benchmark 2)

100
90

shear [N/mm2]

80

MARC_linear
MARC_quadratic
MARC_3D

70
60
50
40
30
20
10
0
0

10

20

30

40

50

position from axis [mm]


Figure 20 Comparison of contact tangential stress along foundation using 2D linear and
quadratic elements and 3D linear elements [MSC.MARC] (Benchmark 2)

FENET-Advanced Contact Benchmarks

23/62

Figure 21 3D expanded segment used in 3D contact simulation (Benchmark 2)

Figure 22 Contour plots of contact normal stress and contact tangential stress in 3D solutions
with linear elements obtained with MSC.MARC (Benchmark 2)

FENET-Advanced Contact Benchmarks

24/62

0.008
0.007
linear
quadratic

relative slip [mm]

0.006
0.005
0.004
0.003
0.002
0.001
0
0

10

20
30
distance [mm]

40

50

Figure 23 Relative slip along edge of interface for linear and quadratic elements plotted along
slave nodes [Abaqus/Standard] (Benchmark 2)

3.5

Modelling tips

The problem to be analysed is relatively simple in terms of the geometric description


and contact can be defined relatively simply. The following are some guidelines and
tips for modelling this benchmark:
To avoid problems with nodes touching a linear approximation of a curved
segment it is advantageous to model the foundation as the master (contacted)
surface.
Based on this contact definition it is better to give the punch the highest
number of nodes (i.e. the smallest elements)
Ensure that if a penalty stiffness for normal contact detection is applied that
the contact stiffness is sufficiently small (detect the amount of penetration and
increase the stiffness if necessary)
Evaluate the effect of the approximation of the true stick-slip model
The contact normal stresses and contact tangential stresses do not directly
result form the FE solution (are not element quantities) if the constraint
equation method for contact is used. Hence they have to be derived (in the
program) from the stress solution in the elements. If this capability is not
available, the stress components in the elements can be used.

3.6

Parameter study

It is also interesting to present solutions for:


Effect of changing the contact body detection (master and slave)
Results of a 3D model with a 30 degree segment instead of a 90 degree
segment

FENET-Advanced Contact Benchmarks

25/62

4.

Benchmark 3: 3D Sheet Metal Forming

4.1

Summary

Ref. No.

Contact Benchmark - 3

Title

3D Sheet metal forming

Contact
features

Rigid and deformable bodies


Mesh dependency
Elasticity, plasticity and spring back
Sliding contact around circular surface
2D plane strain elements or shell elements

Geometry

Punch

Sheet
Initial position

R2
R3

Die
Punch radius = 23.5 mm
Die radius R2 = 25.0 mm
Die shoulder R3 = 4.0 mm
Width of tools = 50.0 mm
Length of sheet (initially) =120.0 mm
Thickness of sheet = 1.0 mm
Width of sheet = 30.0 mm
Punch stroke = 28.5 mm

Material
properties

Youngs modulus:
Poissons ratio:
Initial yield stress:

Final
position

E = 70.5 kN / mm2
= 0.342
0 = 194 N / mm 2

Hollomon hardening = K n
K = 550.4 N / mm2
n = 0.223
Analysis type Quasi-static analysis
Elastic plastic material (isotropic hardening)
Geometric non-linearity
Non-linear boundary conditions
Displacement Symmetry displacement restraints (half symmetry)
boundary
Bottom surface fixed
conditions
Prescribed vertical displacement for the punch
Applied
No applied forces
loads
Element type
2D plane strain 4 node linear elements
shell 4 node shell elements

FENET-Advanced Contact Benchmarks

26/62

Contact
properties
FE results

2 different cases
coefficient of friction = 0.0
coefficient of friction = 0.1342
1. Forming angle
2. Angle after release
3. Plot of punch force against punch displacement

FENET-Advanced Contact Benchmarks

27/62

4.2

Introduction

This benchmark problem is an approximation of the Numisheet 2002 Benchmark B


problem. Simulations are requested to determine the angle before and after spring
back. Experimental results are available for this benchmark, but it is noted that the
sheet is slightly anisotropic. The current problem uses an isotropic elastic plastic
hardening behaviour.

Source:
FREE BENDING BENCHMARK TESTING OF 6111-T4 ALUMINUM ALLOY SAMPLE
John C. Brem*, Frdric Barlat**, Joseph M. Fridy**Alcoa Technical Center, Pennsylvania,

Numisheet 2002 Conference, Korea

4.3

Required solutions

Two solutions, one using friction coefficient 0.1342 (Coulomb friction model)
between the sheet and both tools and one frictionless solution are requested for:
- Forming angle (the angle at the end of the punch stroke)
- Angle after release (the angle after tool removal)
- Punch force - punch displacement diagram

Figure 24 Requested angles for Benchmark 3

The solutions, obtained with shell elements and plane strain elements, include the
following:
- Element size (in particular near the curved zones)
- Method used in discretisation of the tools
- Method for normal contact detection (penalty stiffness) or hard/direct contact
- Method for stick slip approximation (true stick-slip, penalty based, other)

FENET-Advanced Contact Benchmarks

28/62

4.4

Experimental results

As described in the Numisheet Benchmark 2002, experimental results have been


determined for this problem as well using steel and aluminium material. In the
benchmark the material model has to be described with an orthothropic yield function
based on the experimental R-values. Although the material model to be applied for the
benchmark problem described here is slightly different (here an isotropic yield
function is applied for the steel material), the numerical results can be compared with
the mentioned experimental ones. Four different experiments have been carried out
and the range in experimental results is shown in Table 4. The experimental punch
force punch displacement curve from experiment BE-04 has been selected as
reference and is shown in Figure 25.
Table 4 Experimental values from the Numisheet 2002 (steel material) (Benchmark 3)
Forming angles
19.6- 21.0

Angles after release


53.4 - 55.8

250

Experiment BE04

Punch force [N]

200

150

100

50

0
0

10

15

20

25

30

Punch displacement [m]

Figure 25 Experimental punch force punch displacement result Numisheet 2002 benchmark,
steel BE04 (Benchmark 3)

4.5

FEM solutions

FEM solutions have been obtained with a plane strain approach using MSC.MARC
and Abaqus/Standard.
The 2D geometry, including the positions of the reference points is shown in Figure
26. Both the punch and die can be modelled either in analytical form or in a numerical
form with discrete line segments.

FENET-Advanced Contact Benchmarks

29/62

Figure 26 Two-dimensional geometry applied with Abaqus/Standard (Benchmark 3)

In the presented numerical solution, a very fine curve subdivision is applied in


modelling the tool. The number of applied subdivisions is listed in Table 5.
Table 5 Number of elements applied in tool discretisation (Benchmark 3)

Part
Punch
Die shoulder
Die

Number of elements
720 for 360 degrees
150 for 90 degrees
20 for 90 degrees

Angle [degrees]
0.5
0.6
4.5

The sheet is modelled with quadrilateral plane strain (Abaqus type CPE4I - fully
integrated with incompatible modes) elements with 5 elements over the thickness.
Only half of the sheet is modelled and a fine mesh is applied. The applied element
lengths can be determined from Table 6.
Table 6 Number of elements in length direction (Benchmark 3)

position
0 x 27 mm
27 x 40.2 mm
40.2 x 60 mm

Number of elements
50
100
20

The total mesh consists of 850 elements. The smallest element size is 0.132 mm and
consequently a ratio of element size to radius of curvature of the die shoulder
is l R3 = 0.132 4 = 0.033 . The ratio of element size to radius of curvature of the punch
is l R = 0.54 23.5 = 0.029 .
A detail of the mesh of the undeformed structure is shown in Figure 27

FENET-Advanced Contact Benchmarks

30/62

Figure 27 Detail of undeformed mesh near die shoulder (Benchmark 3)

The following set of boundary conditions has been applied:


Symmetry conditions (i.e. no displacement in horizontal direction) have been
applied to the left size of the strip
A reference point is specified on each rigid surface. For the reference point on
the die surface all degrees of freedom (i.e. displacement in horizontal and
vertical directions and rotation) have been suppressed. For the reference point
on the punch surface the displacement component in horizontal direction and
the rotation is suppressed, while the displacement of the punch in vertical
direction is specified using a table as a function of the time (see Table 7)
The springback phase can be modelled in two ways.
- A gradual release: the boundary conditions as given above.
- An instantaneous release: the middle of the sheet held fixed after the
forming stage and both the punch and the die are removed from the
sheet. Contact has been deactivated during this step. The springback
occurs in one increment (or in a number of increments, if convergence
problems occur as a consequence of removing the reaction forces or
contact forces in one step)
Table 7 Vertical displacement of punch as a function of time (Benchmark 3)

Time
0.0
1.0
1.5

Displacement
0
-28.5
0

The benchmark problem has been analysed both without friction and with a Coulomb
friction model with friction coefficient 0.1348. The Coulomb friction model is a true
stick-slip model and often a variant is implemented in FE codes, which reflects the
elastic stiffness. These approximations are then called tangential stiffness, elastic
stiffness, penalty friction model, smoothing function, etc. The default penalty
formulation in Abaqus/Standard has been applied. The true stick-slip Coulomb model
as implemented with a Lagrange multiplier approach often did not converge.
Characteristic deformed stages for the analysis without friction and with friction
during forming are shown in Figure 28 and during the release in Figure 29. In the
FENET-Advanced Contact Benchmarks

31/62

analysis without friction, contact is initially present between the sheet and the lower
section of the punch. Near the end of the deformation the sheet separates at the lower
section of the punch and gets in contact with the lower section of the die. As soon as
this contact is detected, the sheet is further bent into the final shape and the required
force in the force displacement history curve increases (Figure 28).

Figure 28 Various stages in the forming history (no friction left; with friction right)
(Benchmark 3)

FENET-Advanced Contact Benchmarks

32/62

Figure 29 Various stages in tool release process (no friction left; with friction right)
(Benchmark 3)

In the analysis with friction, the deformation behaviour is different. The tangential
forces due to friction result in a stretching of the sheet, causing contact between the
punch and the sheet to be present during the complete forming history.
The characteristic load displacement curves for the analysis without friction and with
friction are shown in Figure 30. The differences in the shape of the curves are caused
by the different contact conditions at the end of the forming stage.
Observe that the unloading stage is analysed by removing the punch slowly in the
upward direction. A high number of steps are required for this unloading (in particular
for the problem with friction; more steps required for the unloading than the loading
phase).

FENET-Advanced Contact Benchmarks

33/62

Sheet forming - plane strain


250
Abaqus/Standard - no friction
Abaqus/Standard - friction

Punch force [N]

200

150

100

50

0
0

10

15

20

25

30

Punch displacement [m]

Figure 30 Load displacement diagram plane strain analysis (friction and frictionless)
(Benchmark 3)

The characteristic values of the angles at the end of the forming stage and after
removal of the tool are listed in Table 8.
Table 8 Characteristic angles during process (Benchmark 3)
Friction coefficient
0

Forming angle
21.88

Angle after release


48.38

0.1348

21.84

54.45

A comparison of the results obtained with MSC.MARC (direct contact setting or hard
contact) and Abaqus/Standard (penalty based contact with numerical value 200
MPa/mm) is shown in Figure 31 (no friction) and Figure 32 (friction). In this last
figure also a comparison with the experimental result is made. Good agreement can be
obtained with both FE codes. The MSC.MARC results exhibit more oscillations in the
load displacement curve and this is caused by the use of hard contact instead of the
more weak approach applied in Abaqus/Standard. Good agreement with the
experimental result is obtained. The results can still be improved by modifying the
material behaviour of using different contact settings. It should be noted that no
experimental data points are available for the unloading.

FENET-Advanced Contact Benchmarks

34/62

Sheet forming - plane strain


250
Abaqus/Standard - no friction
MSC.MARC - no friction

Punch force [N]

200

150

100

50

0
0

10

15

20

25

30

Punch displacement [m]

Figure 31 Comparison of load displacement curve obtained with MSC.MARC and


Abaqus/Standard (no friction) (Benchmark 3)

Sheet forming - plane strain


250
Abaqus/Standard - friction
MSC.MARC - friction
Experimental data points

Punch force [N]

200

150

100

50

0
0

10

15

20

25

30

Punch displacement [m]

Figure 32 Comparison of load displacement curve obtained with MSC.MARC and


Abaqus/Standard (friction) and comparison with experimental values (Benchmark 3)

The problem has also been analysed with a 3D shell approach. The applied coordinate
system, the geometrical entities and position of the reference points for the tools are
shown in Figure 33. Only half of the plate has been modelled, with appropriate
symmetry conditions (i.e. no displacement I 1 direction and rotation about 2 and 3
direction zero) at the middle of the plate.
FENET-Advanced Contact Benchmarks

35/62

It is worth noting that in some FE codes the shell thickness may be ignored in
calculating the contact thickness. Consequently, the tool radii may have to be
increased by half the shell thickness in the current problem.

Figure 33 Geometrical entities 3D Abaqus/Standard model (Benchmark 3)

An analytical description of the tool geometry has been chosen for the analysis with
shell elements.
The applied element mesh is shown in Figure 34. Six elements in the width direction
have been applied and a very fine mesh in the length direction is used (Table 9). The
smallest element size in x direction is set to 0.25 mm. The applied Abaqus element
type is S4R: a doubly curved 4-noded thin or thick shell element with reduced
integration, hourglass control and allowing finite membrane strains.
Table 9 Number of elements in length direction shell analysis (Benchmark 3)

Position
0 x 40 mm
40 x 60 mm

Number of elements
160
10

The penalty method for contact analysis (both for the normal and the tangential
contact) has been chosen.

FENET-Advanced Contact Benchmarks

36/62

Figure 34 Applied mesh density in shell model (Benchmark 3)

The shell analyses have been compared with the plane strain analysis for penalty
contact (equivalent numerical value 1000 MPa/mm) for both the cases with and
without friction. The load displacement curves are shown in Figure 35 and Figure 36.

Shell versus plane strain_no friction


300
shell_1000
plane_strain_1000
250

Force [N]

200

150

100

50

10

15

20

25

30

Displacement [mm]

Figure 35 Comparison plane strain and shell analysis (no friction) (Benchmark 3)
Shell versus plane strain_ friction
250

shell_1000
Plane strain 1000
200

Force [N]

150

100

50

0
0

10

15

20

25

30

Displacement [mm]

Figure 36 Comparison plane strain versus shell analysis (friction = 0.1348) (Benchmark 3)

FENET-Advanced Contact Benchmarks

37/62

Table 10 Comparison of angles plane strain and shell approach (no friction) (Benchmark 3)

Plane strain
Shell

Forming angle
20.94
20.89

Angle after release


47.69
43.30

Table 11 Comparison of angles plane strain versus shell (friction 0.1348) (Benchmark 3)

Plane strain
Shell

Forming angle
20.90
20.92

Angle after release


53.55
53.27

The resulting values of the characteristic angles are listed in


Table 10 and Table 11. For the case with friction a difference of 0.02 is present in the
forming angle whereas the difference after release is 0.28.
The frictionless situation clearly deviates more due to the bending of the sheet around
the axis of the sheet perpendicular to the major bending direction under the punch.
This bending effect in the analysis with friction is less pronounced due to the contact
between the sheet and the punch.

FENET-Advanced Contact Benchmarks

38/62

4.6

Modelling tips

One of the complicating characteristics in this benchmark problem is a very local


contact between the plate and the curved shoulders of the die. In fact the contact is
almost a point (2D) or line (3D) contact with a large amount of sliding. Contact is
only verified between the nodes of the plate and the rigid dies. Hence, in the discrete
steps of the displacement history, points can be identified where no contact is
detected, especially if large elements are used near the shoulder of the die.
The following are some guidelines and tips for modelling this benchmark:
A fine mesh has to be used to describe the contact of the nodes of the sheet
with the die properly
A smooth representation of the die has to be chosen, either in an analytical
form or by a piecewise linear curve using a high number of segments
If a penalty based method of contact is applied the contact between the sheet
and the tools is modelled with springs (stiffness and forces) and depending
upon the applied penalty function value (i.e. value of the spring stiffness) the
contact will be less local at the cost of some penetration of the sheet between
the tools.

Figure 37 Nodal penetrations in penalty contact (200 MPa/mm) (Benchmark 3)

A too low value of the penalty stiffness causes too much penetration; a too
high value of the penalty stiffness causes a harder contact condition with more
local contact and potential numerical instabilities. In the presented solution a
numerical value of 200 MPa/mm is applied. If in a plane strain solution with
Abaqus/Standard, the thickness of the plane strain element is specified as the
width of the sheet (30 mm), the value of the penalty stiffness has to be
multiplied by this width (hence a value of 6000 in a plane strain simulation
corresponds to 200 in the shell approach).
The unloading behaviour is characterized by removal of the tools and at the
same time adding boundary conditions preventing the possibility of rigid body
movement.
The unloading behaviour should preferably be done in a number of steps. Note
that in these steps low values of the normal and consequently the friction
forces are present which makes it difficult to obtain a converged solution

FENET-Advanced Contact Benchmarks

39/62

4.7

Numerical damping is often recommended to stabilize the solution, but it can


be shown that this greatly influences the accuracy of the solution.

Parameter study

Interesting phenomena can be observed if the analyst varies the following parameters:
- Plane strain element versus true shell element solutions
- Analytical tool description versus discretised tool description
- Spring back in one step instead of gradual tool release

FENET-Advanced Contact Benchmarks

40/62

5.1

Summary

Ref. No.

Contact Benchmark - 4

Title

3D Loaded pin

Contact
features

Receding contact area


Curved contact surfaces
Deformable-deformable contact
Friction stick-slip along the contact surface
3D continuum.
L1 = 200 mm
L2 = 20 mm
t
R1 = 50 mm
R2 = 100 mm
H = 100 mm
t = 10 mm

Geometry

F
2

Benchmark 4: 3D Loaded Pin

5.

L1
R2
H

R1

Material
properties

E pin = 210 kN / mm 2
Esheet = 70 N / mm 2
sheet = pin = 0.3

Analysis type Quasi-static analysis


Linear elastic material
Geometric non-linearity
Non-linear boundary conditions
Displacement Symmetry displacement constraints (quarter symmetry)
Left side of the sheet is fixed
boundary
conditions
Applied
Two equal point forces, resulting in a total force on the pin of100 kN .
loads
Element type 3D continuum
20-node quadric elements
or 8-node linear elements
Contact
coefficient of friction = 0.1
properties
FE results
1. Plot of contact pressure against angle
2. Plot of tangential stress against angle
3. Plot of relative tangential slip against angle

FENET-Advanced Contact Benchmarks

41/62

5.2

Introduction

This benchmark problem evaluates the performance of contact algorithms at curved


boundaries. A cylindrical pin is located in the cylindrical hole of a strip. The
diameters of the hole and the pin are identical. Two equal point loads are applied to
the centre of the pin, resulting to a loss of contact at one side of the pin and a localized
contact area on the other side. It is assumed that the tangential contact can be
described with a Coulomb friction model using friction coefficient 0.1. For a 2D
solution it is sufficient to model only half of the assembly, while for a 3D solution
sufficient a quarter of the assembly is sufficient, provided that appropriate symmetry
conditions are applied.

5.3

Required solutions

A number of different quantities are requested along the curved boundary of the pin.
It is interesting to compare the numerical values obtained at the nodes of the strip with
those at the nodes of the pin. The solution is obtained with linear and quadratic
elements. Both the x and y component of the displacement are requested. In addition
the contact normal and the contact tangential stresses as well as the tangential slip
have to be generated as functions of the angle.

Figure 38 Angle definition in requested displacement field (Benchmark 4)

The solutions, obtained with 3D continuum elements (or 2D plane strain elements),
include the following:
- Element size at (in particular along the curved contact surface)
- Method used in discretisation of the tools
- Method for normal contact detection (penalty stiffness) or hard/direct contact
- Method for stick slip approximation (true stick-slip, penalty based, other)

5.4

FEM solutions

Numerical solutions have been obtained with Abaqus/Standard version 6.5 and
MSC.MARC version 2005. The typical mesh density is shown in Figure 39. Along
the curved strip boundary 18 elements (element edge 8.72 mm) have been applied.
Along the curved pin 40 elements (element edge size 3.93 mm).
Fully integrated elements have been applied in the numerical 2D plane strain
simulations with MSC.MARC and Abaqus/Standard. The applied element types are
listed in Table 12.
FENET-Advanced Contact Benchmarks

42/62

Table 12 Applied element types in numerical solutions (Benchmark 4)

linear
quadratic

Abaqus/Standard
Type CE4
Type CE8

MSC.MARC
Type 11
Type 27

The nodes along the pin boundary are selected as slave (contacting) nodes, while the
nodes along the strip are specified to be the master (contacted) nodes. Observe that a
relatively coarse mesh density is used for the master surface, thus illustrating the
differences in approaches when nodes are considered to be in contact with the
segment between the master nodes. The normal contact is described with the
constraint equation method and the tangential contact is described with a true stick
slip model.

Figure 39 Applied mesh density in numerical solutions (Benchmark 4)

The left edge of the strip is fully clamped while at the bottom edge of both the pin and
the strip symmetry conditions have been applied.
The displacement solution obtained with the linear elements is shown in Figure 40,
while the solution obtained with quadratic elements is shown in Figure 41). The
numerical solutions obtained with the linear and quadratic elements using 2 different
FE codes are only slightly different.
The relatively coarse mesh of the strip is specified as the master (contacted) surface in
the presented solutions. Consequently the nodes of the pin touch the segments
between these master nodes. For Abaqus/Standard, linear segments with smoothing
between the intersecting segments are used, while in the selected analytical contact
algorithm in MSC.MARC first a spline between these (master) nodes is constructed
and the nodes of the pin are considered to be in contact with the spline.

FENET-Advanced Contact Benchmarks

43/62

0.8
MARC_linear_x
MARC_linear_y
Abaqus_linear_x
Abaqus_linear_y

0.7

displacement [mm]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

30

60

90

120

150

180

-0.1
angle (degrees)
Figure 40 Displacement as a function of the angles obtained with linear elements in MSC.MARC
and Abaqus/Standard for the nodes of the master (contacted) surface (Benchmark 4)

0.8
MARC_quadratic_x
Abaqus_quadratic_x
MARC_quadratic_y
Abaqus_quadratic_y

0.7

displacement [mm]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

30

60

90

120

150

180

-0.1
angle (degrees)
Figure 41 Displacement as a function of the angles obtained with quadratic elements in
MSC.MARC and Abaqus/Standard for the nodes of the master (contacted) surface
(Benchmark 4)

The displacements along the circular edge of the pin (slave or contacting surface) as
function of the angle are shown in Figure 42 and Figure 43.
FENET-Advanced Contact Benchmarks

44/62

0.800

x-displacement [mm]

0.780

0.760
MARC_linear
MARC_quadratic
Abaqus_linear
Abaqus_quadratic
0.740

0.720

0.700
0

30

60

90

120

150

180

angle (degrees)

Figure 42 Displacement in x-direction for nodes along the pin (slave/contacting nodes) as a
function of the angle (Benchmark 4)

0.004

MARC_linear
MARC_quadratic
Abaqus_linear
Abaqus_quadratic

0.002

0.000

y-displacement [mm]

30

60

90

120

150

180

-0.002

-0.004

-0.006

-0.008

-0.010
angle (degrees)

Figure 43 Displacement in y-direction for nodes along the pin (slave/contacting nodes) as a
function of the angle (Benchmark 4)

FENET-Advanced Contact Benchmarks

45/62

The resulting contact pressure is more sensitive to specific program settings for the
internal contact handling. Already mentioned are the spline approximation of the
contacted surface in MSC.MARC and the smoothing parameter of the intersecting
surface approximations in Abaqus/Standard. Additional options are stress-free
repositioning of the nodes at the start of the analysis (both MSC.MARC and
Abaqus/Standard) and choosing between default contact and small sliding contact
(Abaqus/Standard). A number of results for different contact parameter settings are
shown. An overview of the applied parameters in the various figures is shown in
Table 13.
Table 13 Applied contact parameter settings for various results (Benchmark 4)
Figure 40
Figure 41
Figure 42
Figure 43
Figure 44
Fiigure 45
Figure 46
Figure 47
Figure 48
Figure 49
Figure 50
Figure 51
Figure 52
Figure 53
Figure 54
Figure 55

element type
linear
quadratic
linear
quadratic

Abaqus/Standard
smoothing node repositioning
on
on
on
on
on
on
on
on

sliding type
large
large
large
large

linear
quadratic

on
on

on
on

large
large and small

quadratic
linear
quadratic

on and off off


on
on
on
on

large
large
large and small

element type
linear
quadratic
linear
quadratic
linear and quadratic
linear
quadratic
linear
quadratic
linear

MSC.MARC
contact descrition
analytical
analytical
analytical
analytical
analytical
analytical
analytical
on
on
analytical and discrete

node repositioning
on
on
on
on
on
on
on
on and off
on and off
off

linear
quadratic
quadratic and linear
3D linear
3D linear

analytical
analytical
analytical
analytical
analytical

on
on
on
on
on

The numerical values of the contact pressure along the slave (contacting) nodes and
along the master (contacted) edges are plotted in Figure 44. The global difference by a
factor of 2 between the pin and strip nodes is due to the difference in the thickness
applied in the plane strain approximation. In MSC.MARC the calculated pressure is
determined from the thickness values specified at the element attached to the node,
while in Abaqus/Standard the analyst has to specify the thickness value (here the
value of the pin thickness of 20 mm is used instead of the value of the strip thickness
of 10 mm).
Oscillations in the contact pressure are present and the magnitudes of the oscillations
are influenced by specific contact settings. In the solution presented in Figure 44 the
stress free projection of the nodes to the surface at the start of the analysis is applied.

FENET-Advanced Contact Benchmarks

46/62

300.000

pin_linear
strip_linear
pin_quadratic
strip_quadratic

Contact pressure [N/mm2]

250.000

200.000

150.000

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 44 Contact pressure along strip nodes and pin nodes obtained with linear and quadratic
plane strain elements in MSC.MARC (Benchmark 4)

The contact pressure along the slave (contacting) nodes is shown in Figure 45 (both
codes using linear elements) and Figure 46 (both codes using quadratic elements). For
the results obtained with Abaqus/Standard both the solutions obtained with small
sliding contact and the one obtained with large sliding contact is shown.

150.000

Contact pressure [N/mm2]

MARC_linear
Abaqus_linear

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 45 Contact pressure comparison for linear elements using Abaqus/Standard and
MSC.MARC (linear elements) (Benchmark 4)

FENET-Advanced Contact Benchmarks

47/62

150.000

Contact pressure [N/mm2]

MARC_quadratic
Abaqus_quadratic
Abaqus_quadratic_small

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 46 Contact pressure comparison using Abaqus/Standard and MSC.MARC (quadratic


elements) (Benchmark 4)

It is clear that the presented solutions contain strong oscillations in the calculated
contact pressure variable. This oscillation is even larger if no stress free projection of
the nodes to the contact surface is applied at the start of the analysis, as is shown for
both the linear and quadratic elements using MSC.MARC in Figure 47 and Figure 48.
150.000

Contact pressure [N/mm2]

MARC_linear_no adjust
MARC_linear

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 47 Comparison of contact pressure as a function adjusting the nodes at the start of the
analysis for linear elements (Benchmark 4)

FENET-Advanced Contact Benchmarks

48/62

150.000

Contact pressure [N/mm2]

MARC_quadratic_no adjust
MARC_quadratic

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 48 Comparison of contact pressure as a function of adjusting the nodes at the start of the
analysis for quadratic elements (Benchmark 4)

The same tendency is present in results obtained with Abaqus/Standard. Clearly


adjusting the nodes improves the results, but if the pin is for instance subsequently
subjected to a rotation, the improvement in calculated contact pressure variable is not
guaranteed.
It is interesting to show the optimised result using the linear elements and compare
these with the results of the default contact settings. This is shown in Figure 49 and
Figure 50. It is clear that due to the lack of contact of some nodes with the straight
segments (default parameters: discrete contact in MSC.MARC and default smoothing
in Abaqus/Standard, with no stress free positioning of the nodes) large oscillations in
the contact pressure are present.
300.000

250.000

Contact pressure [N/mm2]

MARC_linear
MARC_linear_discrete
200.000

150.000

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 49 Comparison of contact pressure obtained with optimised parameter setting and default
parameter setting using MSC.MARC (Benchmark 4)

FENET-Advanced Contact Benchmarks

49/62

300.000

250.000

Contact pressure [N/mm2]

Abaqus_linear
Abaqus_linear_def contact
200.000

150.000

100.000

50.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 50 Comparison of contact pressure obtained with optimised parameter setting and default
parameter setting using Abaqus/Standard (Benchmark 4)

The contact tangential stresses show the same tendency in the sensitivity to specific
parameter settings. For the same parameter values as applied in Figure 45 and Figure
46 the calculated tangential stress is plotted in Figure 51 and Figure 52. The tangential
slip is shown as function of the angle in Figure 53.

20.000

MARC_linear
Abaqus_linear

Contact shear [N/mm2]

15.000

10.000

5.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 51 Contact tangential stress as a function of angle using linear elements (Benchmark 4)

FENET-Advanced Contact Benchmarks

50/62

20.000

MARC_quadratic
Abaqus_quadratic
Abaqus_quadratic_small

Contact pressure [N/mm2]

15.000

10.000

5.000

0.000
0

30

60

90

120

150

180

angle (degrees)

Figure 52 Contact tangential stress as a function of contact angle using quadratic elements
(Benchmark 4)

0.050

0.000
0

30

60

90

120

150

180

-0.050

tangential slip

-0.100

-0.150

-0.200
Abaqus_linear
Abaqus_quadratic
-0.250

-0.300

-0.350
angle (degrees)

Figure 53 Tangential slip as function of the angle (Abaqus/Standard) (Benchmark 4)

FENET-Advanced Contact Benchmarks

51/62

A 3D simulation has been performed with linear elements type 7 in MSC.MARC. The
applied mesh is a simple expansion in z-direction of the meshes applied in the 2D
analysis. The mesh and a contour plot of the s11 component are shown in Error!
Reference source not found..

Figure 54 Element mesh and results for 3D simulations with MSC.MARC (Benchmark 4)

In order to describe contact properly, a smoothing procedure is applied along the


master (contacted) surface when the analytical contact is specified. The smoothed
surfaces, which are internally used in the contact procedure, can be visualized in the
post-processor. The surfaces used along the curved outline are displayed in Error!
Reference source not found.. Observe that the algorithm needs to allow for
discontinuities in the edges to avoid smoothing over two intersecting surfaces.

Figure 55 Smoothed surfaces through the boundary nodes used internally in the contact
algorithm in MSC.MARC (Benchmark 4)

The resulting contact pressure is shown in Figure 56. The peak value in the contact
pressure is found to be around 260 N/mm2. Note that in z-direction a relatively coarse
mesh density is applied to capture the 3D effect near the edge of the pin and the strip.
It is also noticed that a comparison with Figure 44 reveals that the use of the 2D plane
strain nominal thickness of 20 mm for the pressure calculation may affect the value of
the 3D contact pressure. A better approximation of the 3D contact pressure would
have been obtained if the thickness of the strip is used since only half of the length of
FENET-Advanced Contact Benchmarks

52/62

the pin is in contact.

Figure 56 Contact pressure in 3D calculation using analytical contact with linear elements
applied in MSC.MARC (Benchmark 4)

5.5

Modelling tips

Contact analysis with contact between two different curved boundaries is difficult to
analyse. Linear elements are relatively simple in the numerical treatment of contact
but fail to describe a curved boundary accurately, in particular if coarse meshes are
used. Quadratic elements can in principle better describe the curved boundary but the
evaluation of the contact pressure is more difficult to handle.
Along a curved boundary the following rules can be defined:
The less-stiff material should be discretised with the finest mesh to provide the
maximum number of nodes in contact (the slave or contacting nodes)
The coarser mesh results in a less accurate description of the geometric
contour, thus projecting the contacting node on the wrong boundary contour
If the code allows a numerical procedure constructing a spline (2D) or a
surface (3D) through the master surface, it is recommended to use this
procedure

5.6

Parameter study

It is interesting to observe the consequence of the numerical values of the solutions if


the analyst varies the following parameters:
- Slave and master surface definitions
- Different mesh densities in both slave and master definitions
- Analytical tool description versus discretised tool description

FENET-Advanced Contact Benchmarks

53/62

6.

Benchmark 5: 3D Steel Roller on Rubber

6.1

Summary

Ref. No.

Contact Benchmark - 5

Title

3D Steel roller on rubber (Reynolds, 1874)

Contact features

3D deformable-deformable contact
Rolling contact
Incompressible material
3D continuum
Steel width= 20 mm
Mat width = 22 mm
R = 30 mm
R
H = 20 mm
steel
L1 = 60 mm
roller
L2 = 300 mm

Geometry

A
H

Rubber
L1

Material
properties

L2

Esteel = 210 kN / mm 2

C10,rubber = 10 N / mm 2 (Neo Hookean material description)

steel = 0.3
Analysis type

Displacement
boundary
conditions

Applied loads
Element type

Contact
properties
FE results

Quasi-static analysis
Linear elastic material
Geometric non-linearity
Non-linear boundary conditions
Displacement history:
Centre of the roller is fixed in horizontal and vertical direction
Time period 0-1 second:
- no rotation of roller
- move bottom surface of rubber 3 mm up
- sheet x-displacement fixed
Time period 1-2 second
- prescribed rotation of steel roller (360 degrees)
- bottom surface of rubber sheet held at 3 mm y-displacement
- sheet is free to move in horizontal direction
No applied forces
3D continuum
20-node quadratic elements
or 8-node linear elements
coefficient of friction = 0.3
Horizontal displacement of the point A after 360 degrees motion

FENET-Advanced Contact Benchmarks

54/62

6.2

Introduction

If a cylindrical steel roller rotates on a rubber base, it is found that for a complete
revolution of the cylinder, it will have moved a horizontal distance less than the
circumference of the cylinder. This is because the rubber base will stretch under the
stresses induced by the cylinder. Friction will also play a part. This phenomenon was
looked at by Osborne Reynolds in 1874. Also, as Reynolds was the first Professor of
Engineering at The University of Manchester, this experiment has been re-created for
an Exhibition of his life.
This benchmark will test the ability of the algorithms to model the movement of the 2
surfaces when they are in contact because the rubber will stretch compared to the steel
cylinder.
In the current benchmark problem, the centre of rotation of the cylinder is held fixed.
The rubber sheet is moved a distance of 3 mm in an upwards direction, thus causing
sufficient pre-strain in the rubber sheet. The rubber sheet is subsequently free to move
in horizontal direction and the amount of horizontal displacement needs to be
determined.
It is assumed that the material behaviour of the rubber can be described with a NeoHookean material model with numerical parameter C10,rubber = 10 N / mm 2 ,

6.3

Required solutions

The cylindrical roller with radius 30 is rotating over 360 degrees. The circumference
of the cylinder with radius 30 mm is 188.5 mm. Point A is the point on the rubber
sheet, which is at the start of the analysis just below the lowest point of the steel
cylinder. The displacement component of this point in the x-direction needs to be
determined. Solutions are obtained with both linear and quadratic displacement field
elements.
The solutions, obtained with 3D solid elements and 2D plane strain elements, include
the following:
- Element size along top of rubber mat and along circumference of cylinder
- Method for normal contact detection (penalty stiffness) or hard/direct contact
- Method for stick slip approximation (true stick-slip, penalty based, other)
- Element types used in modelling the steel and the rubber

FENET-Advanced Contact Benchmarks

55/62

6.4

FEM solutions

Numerical solutions have been obtained with a 2D plane strain approach (applied
nominal thickness of 1 mm) using MSC.MARC and Abaqus/Standard. In addition a
simulation with MSC.MARC has been performed with a 3D approach using the
thicknesses mentioned in the summary. In the last analysis only half of the assembly
has been analysed with appropriate symmetry conditions.

Figure 57 Applied mesh density in obtaining target solutions with MSC.MARC (Benchmark 5)

The rubber mat is meshed with 10 elements in the height direction and 120 elements
in the horizontal direction, thus causing the element length along the surface to be 2.5
mm. The cylinder is meshed with 40 elements along the circumference resulting in an
element length 4.71 mm.
Different element types are applied for the rubber mat and the steel roller. The applied
elements are listed in Table 14.
Table 14 Applied element types in Benchmark 5

Linear rubber
Linear steel
Quadratic rubber
Quadratic steel
3D Linear rubber
3D Linear steel

Abaqus/Standard
CPE4H
CPE4
CPE8H
CPE8

MSC.MARC
Type 80
Type 11
Type 27
Type 32
Type 84
Type 7

The nodes on the outer surface of the roll have been set as contacted or master nodes,
the nodes on the top surface of the mat are specified as the contacting or slave nodes.
Both in Abaqus/Standard and MSC.MARC the direct method for contact checking is
applied. In MSC.MARC the true stick-slip model is applied, whereas in
Abaqus/Standard the penalty tangential slip formulation is applied.
The vertical motion of the mat has been specified in 10 increments with fixed size.
The rotation of the roll has been specified with 90 increments of equal size. Typical
deformed structure plots are shown in Figure 58.
FENET-Advanced Contact Benchmarks

56/62

Figure 58 Deformed structure obtained with MSC.MARC (Benchmark 5)

The x- displacement of point A in the sheet has been obtained with MSC.MARC and
Abaqus/Standard for various element types and is listed in Table 15. Note that the
circumference is 188.5 mm and hence the effect of contact with a stretched surface is
clearly visible.
Table 15 Displacement of point A after one rotation of the roll (Benchmark 5)
Horizontal
displacement
Linear
elements
(discrete contact)
Linear elements
(analytic contact)
Quadratic elements
3D linear elements
(discrete contact)

MSC.MARC

Abaqus/Standard

175.8 mm

175.2 mm

175.3 mm
175.0 mm

175.0 mm

182.9 mm

The 3D solution differs from the 2D solution due to the displacement in z-direction
under the roller. The 2D plane strain approach has no freedom to deform in zdirection; hence the material will be slightly more stretched in x-direction.
The vertical reaction force on the roller (based on thickness dimensions 1 mm) has
been plotted for the two FE codes and the two element types in Figure 59. The time
period 0-1 seconds corresponds to the movement of the mat in the roller direction.
The time period 1-2 seconds corresponds to the rotation. Only small differences are
present between the linear and the quadratic solutions.
Figure 61 shows the vertical forces for the 2D plane strain simulation (thickness 1
mm) and the force on the roll for the 3D simulation divided by the width of the rubber
mat. It is clear that the 3D solution is less stiff.

FENET-Advanced Contact Benchmarks

57/62

vertical force on roller


300
MSC.MARC_linear
MSC.MARC_quadratic
Abaqus_linear
Abaqus_quadratic

250

Force [N]

200
150
100
50
0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

time [sec]

Figure 59 Vertical forces on the roller versus time (Benchmark 5)

Figure 60 3D analysis - undeformed and contour plots of contact pressure on deformed structure
(Benchmark 5)

vertical force on roller


300
250

MSC.MARC_linear
MSC.MARC_linear3D

Force [N]

200
150
100
50
0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

time [sec]

Figure 61 Comparison plane strain and 3D analysis: force on roller versus time (Benchmark 5)

FENET-Advanced Contact Benchmarks

58/62

6.5

Modelling tips

A key issue in modelling of structural behaviour with 3D elements is to keep the


number of elements within reasonable limits to avoid excessive computer times.
When comparing results of 3D analysis with essential 2D solutions the same mesh
density has to be applied. In the design of the mesh density, the following rules apply:
The weakest material (the mat) should have the highest number of nodes along
the surface
The steel roller does not need a high number of elements to describe the
deformation. However if the nodes of the mat are considered in contact with
the segments between the master nodes, a high mesh density along the roller
improves the solution
For modelling the (nearly) incompressible behaviour of the rubber special
elements have to be used to describe the incompressibility correctly

6.6

Parameter study

Interesting phenomena can be observed if the analyst varies the following parameters:
Effect of the magnitude of the applied displacement in upwards direction
Analytical tool description versus discretised tool description
Rigid steel cylinder instead of a deformable one

FENET-Advanced Contact Benchmarks

59/62

7. Concluding Remarks
At first glance, the proposed benchmark problems seem relatively simple. However, a
close study of the benchmark problems reveals that the obtained FE solutions can be
very sensitive to the parameters chosen in generating a particular FE solution. By
comparing the results of two different FE codes, well known for their ability to solve
complex contact problems, it became also clear that specific parameters or contact
control settings specified by the user can influence the obtained result.
Contact is mostly treated numerically as nodes getting in contact with segments, and
as a result the obtained solutions depend on the number of elements specified along
the boundary. Some general recommendations can be given:

A rule of thumb is to use a finer mesh for the weakest (less-stiff) part. This
surface with the fine mesh should then be used as slave (contacting) surface.
This will ensure that as many nodes as possible get in contact and will avoid
unnecessary penetrations of nodes in the surface.

For contact of a curved part with a straight edge, it is generally better to


specify the curved part as the slave surface.

For contact between two curved parts, the user should check to see if some
type of surface smoothing technique can be applied. If not available, quadratic
elements may be used since these elements can describe curved contours
better.

All solutions presented in this report have been analysed with the default
tolerance settings for global convergence. Some FE codes allow the user to
specify tolerance values, e.g. for contact detection and separation values.

In the verification of the FE solutions, a number of checks can be made:


- Verify, using the post-processor program, if nodal penetration can be
observed (ensure that deformation magnification factor is set to 1)
- Check which nodes are in contact
- Check if tensile contact normal forces or stresses are present at the
surface

After verification of the displacement field, the stress field inside the element
is usually right as well. The verification of the numerically obtained contact
pressure, tangential stress and slip is most effectively performed with a path
plot in a 2D analysis and a contour plot along the surface in a 3D analysis.

Other checks typically depend on the problem at hand and often some feeling
of the importance of a specific numerical parameter is obtained by performing
a sensitivity analysis for this parameter.

The proper meshing technique near the contact zone is extremely important in
benchmarks 1 and 3.

Different handling of the normal contact (direct constraint based approach or

FENET-Advanced Contact Benchmarks

60/62

penalty method based method) is illustrated in benchmark 3.

Different implementations of the tangential contact


implementations) result in different solutions in benchmark 2.

The difficulty in handling contact between two curved boundaries, as well as


the problem with contact using quadratic elements is illustrated in benchmark
4.

Large sliding contact is clearly present in benchmarks 3 and 5.

Difficulties with extending 2D solutions to 3D solutions are illustrated in both


benchmarks 3 and 5.

Elastic-plastic behaviour is illustrated in benchmark 3.

Incompressible rubber behaviour is illustrated in benchmark 5.

Benchmarks 1, 2 and 4 can be analysed using a small strain and small


displacement approach assuming a linear elastic material.

In the presented solutions a geometrically non-linear solution is presented. The


main reason for this choice is that in Abaqus/Standard the selection of a
geometrically non-linear step ensures the proper iterative procedure needed for
the strongly non-linear contact algorithm.

FENET-Advanced Contact Benchmarks

(stick-slip

61/62

References
[1]

Mottershead, J E , FE Analysis of Contact and Friction- A Survey,


NAFEMS Report R0025, 1993.

[2]

Konter, A.W.A , How to undertake Contact and Friction Analysis,


NAFEMS Booklet HT15, 2000.

[3]

Feng, Q and Prinja, N K, NAFEMS Benchmark Tests for Finite Element


Modelling of Contact, Gapping and Sliding, NAFEMS Report R0081, 2001.

[4]

Hertz, H., ber die Berhrung fester elasticher Krper. J. Reine Angew.
Mathm. 92, 156-171, 1881.

FENET-Advanced Contact Benchmarks

62/62