Benchmarking study of

steel-composite structures in CAE

crash applications
Master’s thesis in Applied Mechanics

MADELEINE ANDERSSON
EMMA LARSSON

Department of Applied Mechanics

CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden 2016
 MASTER’S THESIS IN APPLIED MECHANICS

Benchmarking study of steel-composite structures in CAE crash applications

MADELEINE ANDERSSON
EMMA LARSSON

Department of Applied Mechanics

Division of Material and Computational Mechanics
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden 2016
Benchmarking study of steel-composite structures in CAE crash applications
MADELEINE ANDERSSON
EMMA LARSSON

c MADELEINE ANDERSSON, EMMA LARSSON, 2016

Master’s thesis 2016:23

ISSN 1652-8557
Department of Applied Mechanics
Division of Material and Computational Mechanics
Chalmers University of Technology
SE-412 96 Göteborg
Sweden
Telephone: +46 (0)31-772 1000

Cover:
Cross section at midpoint of steel-composite hat profile at initial delamination.

Chalmers Reproservice
Göteborg, Sweden 2016
Benchmarking study of steel-composite structures in CAE crash applications
Master’s thesis in Applied Mechanics
MADELEINE ANDERSSON
EMMA LARSSON
Department of Applied Mechanics
Division of Material and Computational Mechanics
Chalmers University of Technology

Abstract
One of the major issues for decreasing the environmental impact of the automotive industry is lowering the
weight of the vehicles. For weight reduction purposes, fibre reinforced polymer composites are of interest due
to their desirable stiffness per weight properties and the ability to design the material for its specific purpose.
Since the development process of this industry is highly based on computer simulations of the components
performance, it is of importance to develop the simulation models. This in order to obtain reliable results of
the potential failure of the fibre reinforced polymer components and how the composite interacts with metal
materials, e.g. in automotive crash simulations. The work within the project is focusing on increasing the
understanding of how to model the failure behaviour of mixed material automotive components (a combination
of fibre reinforced polymer and ultra high strength steel) at a reasonable computational cost.
In this report, a benchmarking study of modelling composite materials using the explicit FE solver LS-DYNA
is described, focusing on capturing the delamination behaviour in the lamina interface as well as the composite
steel interface. For the study different modelling aspects are investigated, such as composite material models
available in LS-DYNA, the compatibility of element types and material models and two different types of
adhesive modelling (a tiebreak condition and a cohesive zone). Simulations are run to test the mode I and
mode II delamination behaviour.
The simulation model is validated using experimental data. The experiments were conducted, specifically
for the project, using two different composite laminates. The procedure of designing the composite laminate
stackup is described in the report. In both experimental tests the composite shows delamination as well as
failure in the composite. The joining between the steel and composite is undamaged in both test cases.
The recommendations for modelling of steel-composite structures, includes modelling both steel and
composite using thick-shell elements, modelling the adhesive between steel and composite as well as the lamina
interface with cohesive elements. LS-DYNA version R9.0.0. is needed in order for the described model procedure
to work as intended.

Keywords: Fibre composite, Composite failure, Delamination, Crash analysis, Finite Element Method, LS-
DYNA, Composite-steel interface, Comparison Tiebreak and Cohesive zone modelling, Failure modelling using
cohesive zone, Material card comparison

i
ii
 Preface
The work within the project is focusing on increasing the understanding of how to model the failure in mixed
material automotive components (a combination of fibre reinforced polymer and ultra high strength steel). The
project was conducted at Chalmers University of Technology in cooperation with the companies Case5 and
Gestamp HardTech.
Case5 is a consulting company in Gotheburg established in 2010 with a wide range of customers from
different industries, such as automotive, aviation, space, and offshore. The company provides their customers
with structural analysis simulations and help them to achieve weight and performance optimised products by
using state of the art simulation methods.
Gestamp HardTech is a company located in Luleå developing and manufacturing press hardened ultra high
strength steel components for the automotive industry as well as press hardening tools.

Acknowledgements
Firstly, thank you to DYNAmore Nordic AB for providing us with a license to LS-DYNA, and to BETA CAE
Systems supplying us with the pre- and postprocessors ANSA and µETA. An extra thanks to the staff at
DYNAmore Nordic for answering our many questions and going that extra mile and providing us with the
newest BETA version of LS-DYNA.
To our supervisor at Case5, Salar Mostofizadeh, we are very grateful for all your help and support. Also, a
great thanks to all co-workers at Case5, for all the help and the great laughs that have carried us through the
project.
To our examiner and supervisor at Chalmers, Martin Fagerström, we are thankful for your support.
We want to acknowledge our contact at Gestamp Hardtech, Lars Wikström, for being so accommodating
and helpful.
Last but not least, we are thankful for our families and friend being so supportive and understanding.

iii
iv
 Contents

Abstract i

Preface iii

Acknowledgements iii

Contents v

1 Introduction 1
1.1 Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Theory 2
2.1 Failure process of a composite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.1 Compressive failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 Tensile failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.3 Delamination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Non-crimp fabric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Classic laminate theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 FE-modelling of composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Common numerical errors in FE-modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5.1 Hourglass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5.2 Shear locking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Benchmark test-case setup 9

3.1 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Designing the composite laminate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Experimental setup and result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4 Developing and benchmarking the simulation model 20

4.1 Element types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.1 Solid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.2 Thin-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.3 Thick-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.4 Brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Composite material models in LS-DYNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1 Stiffness response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.1.1 Solid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.1.2 Thin-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2.1.3 Thick-shell elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2.2 Behaviour at failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2.2.1 Solid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.2.2 Thin-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.2.3 Thick-shell elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.3 Brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Adhesive modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.1 Tiebreak contact condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.2 Cohesive zone modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.3 DCB test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3.3.1 Solid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.3.2 Thin-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3.3.3 Thick-shell elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

v
 4.3.4 ENF test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.4.1 Solid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.4.2 Thin-shell elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3.4.3 Thick-shell elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3.5 Brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Combining composite material model and adhesive model . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.1 Brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.5 Version study in LS-DYNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.5.1 Brief conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.6 Hat profile model testing procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.6.1 Modelling the steel beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.6.2 Modelling the composite laminate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6.3 Benchmarking the simulation model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6.3.1 Yield property study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5 Recommended model configuration 54

5.1 Element type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Adhesive model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Material models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.4 Mesh discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.5 Contact conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.6 Important settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6 Concluding remarks 56

7 Future work 57

References 58

A Material cards I
A.1 MAT22 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2 MAT54/55 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
A.3 MAT58 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV
A.4 MAT261. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI
A.5 MAT262. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII

B Tiebreak card X
B.1 AUTOMATIC SURFACE TO SURFACE TIEBREAK . . . . . . . . . . . . . . . . . . . . . . . . . X

C Failure criteria XI
C.1 Chang-Chang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI
C.2 Tsai-Wu. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI
C.3 Modified Hashin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XII
C.4 Pinho . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XII
C.5 Camanho . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII

vi
1 Introduction
1.1 Background
The automotive industry strives towards reducing the environmental impact. One of the major issues for this
improvement is to reduce the weight of the vehicles. In order to do so new technologies and design solutions
are required. Fibre reinforced polymer composites are of interest due to their desirable stiffness per weight
properties and the ability to design the material for its specific purpose.
For the automotive industry, a limiting factor for the weight reduction is the crash safety. The energy
absorption prediction needs to be accurate for the crash simulations during the development process, and this
is yet only achievable for metals and polymers with well-known behaviour. An interesting industrial approach
to reduce the vehicle weight is to combine fibre reinforced polymers for increased stiffness (per weight unit)
with ultra high strength steel for energy absorption in the safety structure. Since the development process of
this industry is highly based on computer simulations of the component performance, it is of importance to
develop the simulation models in order to obtain reliable results of the potential failure of the fibre reinforced
polymer components, e.g. in automotive crash simulations.

1.2 Purpose
The purpose of the project has been to establish a modelling method for mixed material (fibre reinforced
polymer and ultra high strength steel) automotive components in crash scenarios, suitable for the needs within
the component development process at Gestamp HardTech.

1.3 Objective
The main objective has been to take a step forward from the simulation models used by Gestamp Hardtech in
level of failure prediction in both cracking and delamination of the composite material. The project should
determine what material model and element type to use in crash simulations of fibre reinforced composites,
how the interaction and interface between steel and composite should be modelled, and which level of detail is
necessary in the simulations in order to achieve a satisfactory prediction of failure and energy in the mixed
material component.

1.4 Limitations
The project did not take manufacturing methods into account, and only focused on the impact simulations
matching the crash scenarios for hat profiles given by Gestamp HardTech. The analysis was performed in
LS-DYNA and the pre- and post-processing carried out in ANSA and µETA respectively. The material models
and failure criteria investigated will therefore be limited to the ones available in LS-DYNA [1].

1
2 Theory
A composite is a material consisting of two or more constituent materials with significantly different physical
properties. The combination of the constituent materials gives a composite with noticeably different properties
compared to the individual constituents. Still, the characteristics of the composite are strongly influenced by the
properties of the included materials, their distribution, and the interaction between them. The constituents may
interact in such a way that the properties of the composite are not so easily provided as by the volume-fraction
sum of the separate material properties. Many factors therefore need to be accounted for when modelling a
composite, such as the reinforcements concentration, concentration distribution, and orientation.
The composite materials are classified based on the geometry of the reinforcement. The basic categories
are particle- and fibre-reinforcement. The particles are of approximately the same length in all directions, e.g.
a sphere or a cube, whereas the fibre is characterised by its length being much greater than its cross section
dimensions. Reinforcing fibres may be short or long compared with its overall dimensions. Composites with
short fibres are called discontinuous-fibre-reinforced composites while composites with long fibres are called
continuous-fibre-reinforced composites.
The fibre reinforced polymer composites have become an important class of composite materials due to
their ability of achieving high stiffness and strength at low weight. The continuous fibres may be aligned in one
direction and bound together and protected by a matrix, e.g. a polymer. This is called a unidirectional (UD)
composite, see Figure 2.1 with axis 1 in the longitudinal direction (along the fibres), axis 2 in the transverse
direction in the lamina plane and axis 3 in the through thickness direction perpendicular to the lamina plane.
These UD-plies have very different properties in the longitudinal and the transverse directions. Therefore,
UD-plies are stacked together with varying fibre directions to form laminates with a desired behaviour. [2]

2 3
Figure 2.1: Schematic representation of a unidirectional composite, with axis 1 and 2 in the longitudinal and
transverse directions respectively, and axis 3 in the through thickness direction.

2.1 Failure process of a composite

The failure process in a composite is significantly different and more complex than the fracture process in
homogeneous and on the global scale isotropic materials such as metal and polymers. In isotropic materials the
crack growth is often a simple enlargement of the crack without branching or directional changes, also known
as self-similar crack growth. In composite laminates on the other hand, the crack growth is generally different
in plies with different fibre orientation. Therefore, it is not possible to define a unique crack length and it
becomes more meaningful to analyse a damage zone ahead of the crack. [2]
The damage zone includes several energy absorbing mechanisms since internal material failure generally
proceeds the macroscopic failure of the specimen. The internal material failure may appear in various forms
separately or combined, such as (i) breaking of fibres, (ii) microcracking of the matrix, (iii) separation of fibres
from the matrix, i.e. debonding and (iv) separation of adjacent plies, i.e. delamination. Generally, due to the
1
internal failure, the material response changes well before the actual failure of the specimen. The failure load
of a UD composite may therefore either be considered as the load at which the material behaviour deviates
from linear stress-strain response or the load at final fracture. Note that most UD composites have a linear
stress-strain response up to failure. [2]
The main failure modes of laminated fibre-reinforced composites are matrix compression failure, fibre
compression failure, matrix tensile failure, fibre tensile failure and delamination [3]. These failure modes are
described further in the subsections below.

2
2.1.1 Compressive failure
When subjected to longitudinal compressive loading the fibres of the composite can be seen as long columns and
therefore are at risk of buckling. When appearing within a composite it is referred to as fibre microbuckling [2].
Kink bands are commonly observed in the compressive failure mode and may be a consequence of microbuckling.
Note that there is a discussion in the literature concerning the kinking being a separate failure mode and not
being seen as a consequent of microbuckling. Kinking is a localized shear deformation of the matrix, along
a band across the fibres. This deformation is triggered by imperfections in the material, particularly initial
fibre misalignments, and by the rotation of the fibres during the compressive loading. The failure mode is also
affected by the resin shear behaviour [3]. Due to these factors, the failure mechanism typically takes place is
regions with a lower level of shear stress and higher compressive stress or when severe local fibre misalignments
take place in the composite [4]. The failure mode is illustrated in Figure 2.2a.
The predominant failure mode of a unidirectional composite subjected to transverse compression is matrix
shear failure with or without constituent debonding and/or fibre crushing. The failure of the specimen may
be accelerated by failure of the fibre-resin bond, resulting in a lower transverse compressive strength than
the longitudinal transverse strength [2]. The failure surface typically has a 53◦ angle to the loading axis
[Pinho2005]. This failure mode is illustrated in Figure 2.2b.

(a) (b)
Figure 2.2: Failure modes when the composite is subjected to compression; (a) shear failure mode due to
longitudinal loading and (b) matrix shear failure due to transverse loading.

2.1.2 Tensile failure

In a UD composite subjected to an increasing longitudinal tensile load, failure is initiated by fibre breakage at
their weakest cross sections. Individual fibres break until the cross section of the composite is unable to sustain
the load and failure of the specimen occurs. This brittle failure may arise with or without additional failure
mechanisms like fibre pullout, interface-matrix shear or constituent debonding. The brittle failure of a UD
composite subjected to longitudinal tension is illustrated in Figure 2.3a.
When subjected to a transverse tensile load, the main failure mode is matrix tensile failure. This is due
to stress concentration at the interface and in the matrix created by the fibres aligned perpendicular to the
loading direction. If the fibres are highly aligned and are weak in the transverse direction, the failure may
originate by fibre transverse tensile failure, i.e. fibre splitting. Another failure mechanism that may occur is
constituent debonding, i.e failure of interfacial bonds between the fibre and the matrix. The matrix tensile
failure is illustrated in Figure 2.3b.[2]

(a) (b)
Figure 2.3: Failure modes when the composite is subjected to compression; (a) shear failure mode due to
longitudinal loading and (b) matrix shear failure due to transverse loading.

2.1.3 Delamination
As stated above, composite laminates consists of distinct plies that are stacked together. One critical failure
mode is delamination, i.e. the plies detaching from each other [3]. Composite laminates are highly susceptible
to crack initiation and propagation along the laminar interfaces, this is one of the most common life-limiting

3
crack growth modes in composite laminates. This may cause severe reduction of in-plane strength and stiffness,
potentially leading to catastrophic failure of the whole structure [5]. Even if the delamination is not visible on
the surface or the free edges it may affect the material properties [3].
When the composite is subjected to a 2D load, 3D stress states still needs to be accounted for. The stresses
σ3 , τ13 and τ23 are called interlaminar stresses and may initiate delamination. Their magnitude is largest at the
lamina interface, i.e. where the plies are stuck together, and at the edges of the lamina. A specimen developing
a tensile value of the interlaminar normal stress, σ3 , shows delamination much prior to the fracture of the
specimen. A specimen developing a compressive normal stress at the free edge shows very little evidence of
delamination even when fracture occurs. [2]
The interlaminar stresses are affected by the angles of the plies and the laminate stacking sequence. The
crack propagation along a laminar interface has three different modes; (i) mode I - pure opening of the crack,
(ii) mode II - pure sliding between the layers and (iii) mode III - tearing. These three modes are illustrated in
Figure 2.4. Mode I propagation is driven by a force or stress that opens the crack in the normal direction to
the crack surface, mode II propagation is driven by a force or stress in the direction of the crack propagation
and mode III is drive by a force or stress perpendicular to the crack propagation direction. [2]

(a) (b) (c)

Figure 2.4: Delamination modes, (a) mode I - opening of the crack in the normal direction, (b) mode II - sliding
between the plies, and (c) mode III - tearing of the composite.

2.2 Non-crimp fabric

Non-crimp fabrics (NCF) are reinforcement fabrics created to combine the accurate fibre alignment of UD
tapes with the easy handling of the woven fibres. If unidirectional tapes are used, either placed by hand or by a
robot, and the structure is cured in an autoclave the fibre placement and local properties can be used efficiently.
However, the manufacturing is inconvenient and costly. On the contrary, if a woven reinforcement is used, and
the curing is handled without an autoclave, the manufacturing process cost can be lowered significantly since
large sheets of fabric are handled. In woven materials the mechanical properties are affected by the waviness of
the fibres that the woven structure imposes, which means that more material is required, compared to the
material cost of a UD tape. [6]
The NCF may be produced in various variants, with the basic idea of bundles of fibres being aligned and
held together by weft threads [6]. This enables a fabric that behaves as a woven material in manufacturing but
structurally similar to a UD material, although with reduced longitudinal stiffness and longitudinal strength.
The name non-crimp comes from that the fibre reinforcement is considered ideally straight in the fibre direction,
which means that the fibres are not prebuckled when subjected to longitudinal compressive loads [7]. A
schematic figure of a unidirectional NCF is illustrated in Figure 2.5.

4
Figure 2.5: A schematic figure of a non-crimp fabric (NCF), with vertically aligned bundles of fibres (gray)
held together by weft threads (white).

2.3 Classic laminate theory

The classic laminate theory includes some general assumptions; plane stress, linearly elastic orthotropic layers,
perfect bonding between fibres and matrix as well as between layers [8]. For classic laminate theory the relation
between the resultant forces (N) and moments (M) acting on a laminate cross section and the midplane strain
(0 ) and the curvature (κ), is presented in Equation (2.1). The matrices are called the extensional stiffness
matrix (A), the coupling stiffness matrix (B), and the bending stiffness matrix (D). Note that for a symmetric
stackup the coupling stiffness matrix becomes zero and therefore there is no coupling between curvature and
strain. [2]     
N A B 0
= (2.1)
M B D κ
The matrices are calculated according to Equation (2.2)-(2.4). The contribution of each layer of the composite
is determined using the distance to the midpoint (h). This component is defined in Figure 2.6.
n
X

Aij = Q̄ij k
(hk − hk−1 ) (2.2)
k=1
n
1 X
h2k − h2k−1



Bij = Q̄ij k
(2.3)
2
k=1
n
1 X
h3k − h3k−1



Dij = Q̄ij k
(2.4)
3
k=1

5
 h0
k=1
h1
k=2
h2

hn−2
k =n−1
hn−1
k=n
hn
y

Figure 2.6: Schematic figure of how the distance from the midpoint (hk ) is defined.

The included parameters are presented in Equations (2.5)-(2.8). Q̄ is the lamina stiffness matrix in the
global x, y, z-coordinate system and is constant for each lamina. It is determined using the lamina stiffness
matrix in the local 1, 2, 3-coordinate system Q and the matrices T1 and T2 which are the stress and the strain
transformation matrices respectively. θ is the angle of the fibres, i.e. the angle between axis 1 and x, see
Figure 2.7. [2]

x
2
y 1
z
3

Figure 2.7: Composite UD ply with the local 1, 2, 3−coordinate system and the global x, y, z−coordinate system.

−1
Q̄ = [T1 ] Q [T2 ] (2.5)
E11 ν12 E11
 

Q11 Q12 Q13
 0
 1 − ν12 ν21 1 − ν12 ν21 
   
   ν E

E22

Q21
Q= Q22 Q23 
= (2.6)

12 22
0 
   1 − ν12 ν21 1 − ν12 ν21
 

Q31 Q32 Q33  
0 0 G12

 2 2 
cos(θ) sin(θ) 2 cos(θ) sin(θ)
2 2
T1 =  sin(θ) cos(θ) −2 cos(θ) sin(θ)  (2.7)
2 2
− cos(θ) sin(θ) cos(θ) sin(θ) cos(θ) − sin(θ)
 2 2 
cos(θ) sin(θ) cos(θ) sin(θ)
2 2
T2 =  sin(θ) cos(θ) − cos(θ) sin(θ)  (2.8)
2 2
−2 cos(θ) sin(θ) 2 cos(θ) sin(θ) cos(θ) − sin(θ)

The stiffness (i.e. relation between stress and strain) of the total composite laminate is calculated using
the extensional stiffness matrix since the extensional stiffness relates the resultant forces to the midplane

6
strains. For bending around the z-axis the x direction is assumed to take up the load (global direction), which
corresponds to the value of the first row and first column in A, i.e. A(1, 1). The Young’s modulus for the entire
laminate may then be approximated by dividing this component with the lamina thickness. [2]

2.4 FE-modelling of composites

When modelling a composite laminate using FE there is two general approaches; (i) joining all plies together in
one element using a summation of the lamina properties or (ii) modelling each ply with individual element
layers and joining the elements together with either a cohesive element or a contact condition [8]. The element
configuration of the two approaches is presented in Figure 2.8

(i) When modelling the laminate using only one element, each lamina has an individual
through thickness integration point in order to capture the stress state in each ply.
This method does not take the delamination into account. This approach has a low
computational cost since only one element through the thickness of the lamina is
required. It is possible to either sum up the properties using classic laminate theory or
to sum the properties in a manner that takes the through thickness stress into account.
[8]
(ii) The method of modelling each ply with a layer of elements is computationally heavy,
but in return the delamination behaviour can be included in the analysis if the joining
between the element layers allows for separation. [8]

(a) (b)
Figure 2.8: Mesh and integration point configuration for a composite element with three layers of (a) all plies
joined into a single layer and (b) the plies modelled with individual elements with the integration points marked
with red points.

2.5 Common numerical errors in FE-modelling

There are some numerical errors that can appear in a finite element analysis. The different types of numerical
errors and why they occur are described below.

2.5.1 Hourglass
Hourglass modes are nonphysical, zero energy modes of deformation corresponding to zero strain and stress,
and can occur when using reduced integration [9]. A schematic figure of an hourglass mode is presented in
Figure 2.9. As can be seen in Figure 2.9 the strain at the midpoint between the nodes remain unchanged. Since
hourglass deformation modes are unnoticed by the integration point, work done by the hourglass resistance is
neglected in the energy equation. This may lead to a small loss of energy [10].

7
Figure 2.9: A schematic figure of a zero energy hourglass mode, for an element with one integration point (red).

2.5.2 Shear locking

Shear locking is a phenomena of an overly stiff response for a first order element in bending. The shape of the
in-plane deflection for a bent element is curved in the ideal case, but for a first order element the deflection
within each element is modelled linearly, see Figure 2.10. The energy required to bend the first order element is
higher than the energy in the actual case since the angle ϕ is changed for the first order element. [11]

ϕ
ϕ

(a) (b)
Figure 2.10: A schematic figure of the deflection of the actual case (a) and the first order element (b). Figures
reproduced and altered from Shear Locking and Hourglassing in MSC Nastran, ABAQUS, and ANSYS [11].

8
3 Benchmark test-case setup
The simulation results were calibrated and validated using physical test data, provided by Gestamp HardTech.
The physical test was conducted at the Gestamp HardTech facilities in Luleå and consisted of a regular three
point bending test where the force-displacement relation was recorded in the impactor. For the experiment
a composite laminate was designed consisting of a well characterised composite material from Chalmers and
Swerea SICOMP [12]. In this section, the properties of the composite material, the procedure of designing the
composite laminate, and the setup of the physical test are described.

3.1 Material properties

The composite material used in the tests and simulations is a well characterised non-crimp fabric (NCF)
material used for research purposes at Chalmers and Swerea SICOMP. The material consists of carbon fibres in
an epoxy matrix, where the carbon fibres are stitched together using glass fibres. The averaged properties of
the composite are obtained from tests conducted by Bru et al. [12] and are presented in Table 3.1. Additionally,
the volume fraction of the carbon fibres is 60%, the thickness of each ply is 0.18 mm, and the shear fracture
angle is 62◦ [12]. Note that the Young’s modulus (E) has different values for compression and tension in both
the longitudinal and the transverse direction. All further use of E11 and E22 is referring to the lower values for
the two cases, i.e. E11c and E22t .

9
 Table 3.1: Average values of physical properties for the material used in the tests and simulation. [12]
Young’s modulus
E11c 132 GP a Longitudinal compression
E11t 140 GP a Longitudinal tension
E22c 9.3 GP a Transverse compression
E22t 9 GP a Transverse tension
Shear modulus
G12 4.4 GP a In-plane
G13 3.7 GP a Through thickness
Strength
Xc 631 M P a Longitudinal compression
Xt 1787 M P a Longitudinal tension
Yc 130 M P a Transverse compression
Yt 29.2 M P a Transverse tension
S12 77.8 M P a Shear
Poissons ratio
ν12 0.28
ν21 0.029
ν31 0.02
ν32 0.43
Interlaminar energy release rate
GI 149 J/m2 Mode I
GII 690 J/m2 Mode II
Strain at failure
ε11cu 0.491 %
ε11tu 1.23 %
ε22cu 1.71 %
ε22tu 0.32 %
γ12u 9.1 %
γ13u 2.9 %
Fracture toughness - initiation value of energy release rate
GIC,lamcomp 51.8 kJ/m2
GIC,0◦ comp 103.1 kJ/m2
GIC,lamtens 33.7 kJ/m2
GIC,0◦ comp 67.1 kJ/m2

The adhesive used to attach the composite to the steel is a glue from Sika (SikaPower R
-MBX Class I),
intended for joining of metals and polymers. The mechanical properties are presented in Table 3.2. During
the curing process the beam was heat treated. The heat treatment consisted of a temperature of 60◦ C for 20
minutes, then a temperature increase to 175◦ C that was kept for 30 minutes.

Table 3.2: Mechanical properties of the glue used in the experiment [13].
Property Value
Young’s modulus 800 MPa
Lap shear strength 20 MPa
Tensile strength 20 MPa
Elongation at failure 20 %
T-Peel strength 10 N/mm

The steel used in the experiment is a steel manufactured by Gestamp HardTech. The mechanical properties
of the steel is presented in Table 3.3, and the yield properties of the steel is presented in Figure 3.1. Note that
this is for the steel without the heat treatment.

10
 Table 3.3: Mechanical properties of the steel used in the experiment.
Property Value
Young’s modulus 206 GPa
Poisson’s ratio 0.3

1800

1700

1600

1500
Stress [MPa]

1400

1300

1200

1100

1000

900

800
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Strain [ ]
Figure 3.1: Yield function of the steel used in the experiment.

3.2 Designing the composite laminate

The design of the composite laminate implemented in the hat profile beam three point bending test, described
in Section 3.3, was made to provide a noticeable increase of the bending stiffness of the beam, compared to the
pure steel beam. However, the design also takes into account that a too stiff composite laminate may lead
to total delamination between the laminate and the steel, resulting in the composite laminate simply falling
off. The design also provokes delamination within the laminate since this is a behaviour the simulation model
should be able to capture.
In order to approximate the effect of the composite laminate on the bending stiffness of the beam, the
bending stiffness of the total beam is compared to the one of the pure steel beam. This is calculated by dividing
the beam into parts with constant Young’s modulus and second moment of inertia, see Figure 3.2. The bending
is assumed to occur around the centre of mass which is in reality not the case for a beam consisting of two
different materials, but is in this case considered a reasonable assumption. It can be seen that the pure steel
beam is divided into three parts while the mixed-material beam has a fourth part - the composite laminate.
Since the bending in the test is around the x-axis, only this bending stiffness component is of interest. Note
that the coordinate system is modified from the one in Section 2 in order to match the coordinate system in
the model provided by Gestamp Hardtech.

11
 3 3
4
2 2
y y
z x 1 z x 1
(a) (b)
Figure 3.2: The cross section of the hat profile beam, with the different parts of the beam with constant Young’s
modulus and moment of inertia marked numbered. The parts are marked for both (a) the pure steel beam and
(b) the combined steel and composite beam. The dash-dotted line represents the symmetry line in the x-direction.
1 1

For each part of the beam, the second moment of inertia for the total centre of rotation is calculated using
Steiners law [14], see Equation (3.1). The original second moment of inertia (I) is calculated according to
Equation (3.2) [14], with the width and the thickness referred to as w and t respectively. A is the area and b
the distance between the centre of rotation for the part and for the total beam. The total bending stiffness is
the sum of the individual parts bending stiffness around the centre of the rotation, meaning that each part
gives a contribution of the Young’s modulus (E) multiplied with the compensated moment of inertia (Icomp )
from Steiners law, according to Equation (3.3) [14]. For approximating Young’s modulus for the composite
laminate some further calculations are needed.
As presented in Section 2.3, for a laminate with a linearly elastic composite material where plane stress
is assumed, the relation between the resultant forces and the midplane strain may be approximated using
components from the extensional stiffness matrix. For bending around the x-axis (see Figure 3.2 for coordinate
system), the stiffness corresponding to the strain in the z-direction is of interest.

Icomp =I + b2 A (3.1)
3
wt
I= (3.2)
12 N
X
(EI)total = E1 Icomp,1 + E2 Icomp,2 + ... = Ei Icomp,i (3.3)
i=1

For the experiment, two different laminate stack-ups are designed. This to secure the simulation model to
not be calibrated for only one case. These two laminates are therefore needed to give a different response in
the three point bending test. However, only one piece of composite laminate is manufactured, from which both
composite plates are to be cut. The design of the composite plate therefore also takes into account that the
stiffness needs to be different depending on the placement of the cut out. The two plates, referred to as plate A
and plate E, are therefore cut from the same composite laminate but rotated with an angle of 90◦ from each
other, see Figure 3.3. The plate is kept symmetric in order to not introduce any coupling between strain and
curvature. Some ±45◦ plies are included in the laminate to prevent a rapid drop in the force-displacement
relation when failure occurs, and also to provoke delamination. Due to the significantly higher Young’s modulus,
the 0◦ plies increase the bending stiffness more than any other fibre angle, and therefore plate E has a higher
bending stiffness than plate A. The stack-ups of plates A and E are presented in Figure 3.4 and a change of the
bending stiffness compared to the pure steel beam is predicted to 11.64% for beam A and 17.41% for beam E.

12
 A w E L

Figure 3.3: The manufactured composite laminate and how the two different plates, A and E, are cut out.

A: θ = 0◦ E : θ = 90◦
A: θ = -45◦ E : θ = 45◦
A: θ = 45◦ E : θ = -45◦
A: θ = 90◦ E : θ = 0◦
A: θ = 90◦ E : θ = 0◦
t x
A: θ = 90◦ E : θ = 0◦
A: θ = 90◦ E : θ = 0◦
A: θ = 45◦ E : θ = -45◦
A: θ = -45◦ E : θ = 45◦
A: θ = 0◦ E : θ = 90◦

Figure 3.4: Cross section view of the laminate stackup. The angles of the fibres (θ) are listed to the left for
plate A and to the right for plate E. The thickness is referred to as t and the width as w.

13
3.3 Experimental setup and result
The three point bending test is performed such that the hat profile beam is placed on two roller supports with
an applied load located at the middle of the beam. The applied load is introduced by an impactor, i.e. a
cylinder shaped body with a prescribed vertical displacement (δ) pressing on top of the beam. A schematic
figure of the test setup is presented in Figure 3.5 along with the beam cross section. As can be seen in the
figure, the composite laminate is attached to the top flange of the hat profile. Note that there are two plates
attached to the bottom of the hat profile, one in each end of the beam. The purpose of the plates is to keep
the bottom flanges together over the supports in order to make the bending of the beam the main deformation
mode. The dimensions are presented in Table 3.4. Note that the composite plate does not span over the entire
length or width of the profile. The plate measurements are included in the table.

δ δ

a l
L

(a)

t
2t

w
W
(b)
1

Figure 3.5: Schematic figure of (a) the three point bending test and (b) the cross section of the beam with the
steel hat profile with the lid (light grey) and the attached composite laminate (dark grey).

14
 Table 3.4: Geometric properties for the hat profile experiment
Dimension Length [mm] Description
D 305 Diameter of impactor
L 490 Length of beam
Lcomposite 300 Length of composite plate
l 130 Length of cover plate
a 45 Placement of supports
r 25 Radius of supports
h 36.6 Height of beam
W 82 Width of beam
w 15 Width of flange
wcomposite 24 Width of composite plate
t 0.8 Thickness of steel

A photo of the actual test setup of the three point bending test is presented in Figure 3.6, and a photo of
test beam A before the experiment is shown in Figure 3.7. The composite laminate (black) is attached to the
top flange of the hat profile using glue (green). The bottom plates have a lighter grey colour in the figure.

Figure 3.6: The setup of the three point bending test. The hat profile beam is placed on two roller supports and
the impactor is aligned in the middle of the beam.

15
Figure 3.7: Test beam A before the experiment. The composite is (black) can be seen attached on the top flange
of the steel hat profile. The light gray plates are attached to the bottom flanges to prevent them from bending
outward along their width. The glue (green) is visible at the edge of the composite.

The loading profile of the applied displacement can be seen in Figure 3.8 as a time-displacement curve. Note
that the displacement starts at a negative value, due to the reference displacement not being set to zero. The
velocity for the experiment is approximately 6 mm/s, which indicates quasi static conditions (i.e. no inertial
effects need to be considered).
The result in form of a force-displacement curve measured in the impactor is presented in Figure 3.9. As
expected, test beam A and test beam E differ in behaviour. Delamination in the composite laminate occurs for
both test beams. Since test beam E has a higher bending stiffness, a higher force (and therefore a higher stress)
is to be expected at the same displacement, compared to test beam A. This higher stress leads to a delamination
at approximately 6 mm for test beam E, between the ±45◦ plies closest to the steel. The delamination can
be seen in the force-displacement relation as a rapid drop in the force corresponding to further displacement.
For test beam A, a similar decrease of the force can be seen at a displacement of approximately 10 mm. The
delamination for test beam A occurs, as for test beam E, in the interface of the top ±45◦ plies. Note that an
additional delamination may be detected between the remaining −45◦ ply and the underlying 0◦ ply, due to
the inability of the −45◦ ply to follow the shape of the steel profile.
After the delamination, the composite plies still attached to the test beam A (0◦ and −45◦ ) have a higher
bending stiffness than the plies attached to test beam E (90◦ and 45◦ ). This is due to a higher ability of
withstanding loads along the length of the plate for a 0◦ ply, compared to a 90◦ ply. This is visible in the
diagram as a higher curve for test beam A.
Both beams experience a full buckling behaviour at a displacement of approximately 21 mm. The force
corresponding to the applied displacement is then distinctly reduced.

16
 60

Displacement [mm] 50

40

30

20

10

0 2 4 6 8 10 12
Time [s]
Figure 3.8: The loading curve of the applied displacement. The displacement starts at a negative value due to
the reference displacement not being set to zero.

12

10
Force [kN]

2 Experiment A
Experiment E
0
0 5 10 15 20 25 30 35 40 45 50 55
Displacement [mm]
Figure 3.9: Experimental result for test beam A (red) and test beam E (blue).

Photos of the test beams after the test are presented in Figure 3.10. The adhesive used to attach the
composite to the steel beam is visible in the photos (green). The delamination, which is observed as a rapid
decrease of the force needed to induce further delamination, is observed as the inability of the composite to
mimic the deformation of the steel beam. In the photo, the delamination is visible along the whole length
of the exposed part of laminate for test beam E. For test beam A, the laminate is intact on one side of the
buckled steel. In Figure 3.10 this is observed on the left side.

17
 (a)

(b)
Figure 3.10: Photo of the hat profile beam after the experiment for both (a) test beam A and (b) test beam E.

Due to the severe delamination, the buckling behaviour of the steel and the breakage of the still attached
composite plies are not visible in Figure 3.10. Therefore the delaminated parts of the composite were removed
and the still attached composite documented, see Figure 3.11. For both test beam A and test beam E, it can
be seen that the delamination occurred at a ply with a fibre angle of 45◦ and that this ply is still attached.
The angle of the plies may be seen from the weft thread (white) that runs perpendicular to the fibre direction.
Note that an additional delamination may be detected between the remaining −45◦ ply and the underlying 0◦
ply, due to the inability of the −45◦ ply to follow the beams shape. Failure of the composite has occurred in
the midpoint of the beam, at the buckling of the steel.

18
 (a)

(b)
Figure 3.11: Photo of the hat profile beam after the experiment with the delaminated pieces removed for both (a)
test beam A and (b) test beam E.

19
4 Developing and benchmarking the simulation model
In this section the process of obtaining the final model presented in Section 5 is described. Since the model
needs to account for many different behaviours, various tests are run in order to determine how to model both
the composite laminate as well as the glue and lamina interface. The delamination behaviour is of interest
and therefore the general approach is to use one layer of elements for each ply. Element types are investigated
in order to determine how to model each individual ply such that the behaviour is reasonable even after
delamination, i.e when there is only one element through thickness. To capture the behaviour of the material,
different composite material models are investigated and compared. Also the compatibility of the element
types and composite materials is studied. To capture the delamination behaviour, both between the steel and
composite parts and within the composite laminate, different types of adhesive modelling are tested. During
the process, both the behaviour of the model and the simulation time is documented in order to get a model
that gives a reasonable response for a justifiable computational cost.

4.1 Element types

This section describes the element types in LS-DYNA and the procedure of determine which types are suited
for the current model. A simple bending test with a cantilever beam is performed with the different el-
ement types in order to rule out element types not giving a correct stiffness response. The setup of the
test is presented in Figure 4.1. The fixed boundary was modelled using the boundary condition BOUND-
ARY SPC(SET) with all degrees of freedom fixed. The applied displacement was imposed using the condition
BOUNDARY PRESCRIBED MOTION, with a linear loading curve at velocity 0.05 m/s.
A short description of the element formulations used and the results of the cantilever bending test is
presented under each subsection. For all simulations one layer of elements through thickness is used, since
the intention is to model each ply with one layer in order to capture the delamination behaviour. Since solid
elements usually are modelled with at least three elements through thickness, employing a configuration of one
layer of elements through thickness will give the solid elements a disadvantage and the element type will most
likely not perform as well as the shell elements [15]. The simulations to investigate the element formulations
are performed to find the types and formulations that can perform well with one layer of elements through
thickness, and no alterations in the mesh configuration is performed between simulations despite the different
recommendations for different element types. Note that for the under-integrated elements hourglass control is
recommended in order to prevent hourglass modes. The energy dissipating by the hourglass forces reacting
against the formations of the hourglass modes is tracked and reported in the output file GLSTAT. [10]

t
L w
(a) (b)
Figure 4.1: A schematic figure of (a) the cantilever beam test setup, performed to test the stiffness response
of the element types, and (b) the beam cross section. The length is referred to as L, the width as w and the
thickness as t.

The expected response is calculated according to Equation (4.1) [14] with the tip displacement δ, the
Young’s modulus E, the second moment of inertia I and the length L. This analytical solution is represented
by a black line in the figures under each subsection. The mechanical and geometric properties for the cantilever
beam are presented in Table 4.1. The mesh consists of 50 elements along the length, 7 elements along the width

20
and one element through thickness.
3EI
F = δ (4.1)
L3

Table 4.1: Mechanical and geometric properties for cantilever beam analysis
Property Value
E 132 GPa
L 150 mm
t 1.98 mm
w 20 mm

4.1.1 Solid elements

The solid element is a three-dimensional finite element that requires no simplifying geometry, constitutive, or
loading assumptions. The boundary conditions are treated more realistic compared to shell or beam elements.
This, however, comes at a high computational cost. Also, very important in this case, solid elements often
has poor performance for thin-walled structures due to shear locking. Hexahedron elements are to be used if
possible. The 8-noded hexahedron solid element is presented in Figure 4.2.

h
z

x
y

Figure 4.2: Schematic figure of a 8-noded hexahedron solid element. The thicknesss of the element is marked
with h.

For hexahedron elements intended to model the composite there are five applicable element formulations
in LS-DYNA; (i) constant stress element, (ii) fully integrated selectively reduced (S/R) element, (iii) fully
integrated quadratic 8 node elements with nodal rotations, (iv) efficient element formulation for poor aspect
ratio, and (v) accurate element for poor aspect ratio [16, 17].

(i) The constant stress element (ELFORM 1) is the default element in LS-DYNA. The
element is underintegrated (has one integration point in the middle of the element),
which allows for hourglass modes. The element formulation is efficient and accurate,
and can be used for large deformations. Hourglass control type 6 is recommended as a
supplement to avoid zero-energy modes. [16, 17]
(ii) The fully integrated selectively reduced (S/R) element (ELFORM 2) is has 8 integration
points and therefore hourglass formation is not an issue. This element type is slower
than the the constant stress solid element and is more unstable in large deformation
applications. It is too stiff in many situations, especially for poor aspect ratios, where
shear locking is a problem. [16, 17]
(iii) Fully integrated quadratic 8 node elements with nodal rotations (ELFORM 3) is an
expensive element with quadratic interpolation between nodes. The element has 14
integration points, and 6 degrees of freedom in each of the 8 nodes. The element is not
compatible with incompressible materials or simulations with plasticity. This element
formulation is not generally recommended due to high computational cost. [16, 17]

21
 (iv),(v) There are two element formulations for poor aspect ratio; one accurate formulation
(ELFORM -2) and one efficient formulation (ELFORM -1). Both formulations are
based on the fully integrated selectively reduced (S/R) element (ELFORM 2). They
result in a slower simulation than ELFORM 2, but since they can handle poor aspect
ratio without shear locking, it can still be the faster choice for thin structures since
fewer elements are needed. The efficient formulation (ELFORM -1) needs to be
supplemented by hourglass control. [16, 17]
The force-displacement relation obtained from the cantilever bending test is presented in Figure 4.3. The
simulation times are listed in Table 4.2. It can be seen that ELFORM 1 behaves too softly in the cantilever
beam analysis. Since the element has only one integration point through thickness and is modelled with only one
element in this direction the behaviour is expected. The element formulation is excluded from further analysis.
Since element formulation 2 over-predicts the stiffness substantially this element formulation is excluded as
well. Element formulations -1 and -2 lies directly on top of each other, and since simulations using ELFORM -2
are substantially slower (three times slower than using ELFORM -1) this formulation is excluded. This means
that ELFORM -1 and ELFORM 3 are still kept for further analysis.

×10−3
4.5
Analytical
4 Solid ELFORM 1
Solid ELFORM 2
3.5 Solid ELFORM 3
Solid ELFORM -1
3 Solid ELFORM -2
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.3: The result from the cantilever beam bending test, where the following element types are tested;
ELFORM 1 (blue), ELFORM 2 (red), ELFORM 3 (green), ELFORM -1 (orange) and ELFORM -2 (purple).
The analytical result is represented by a black line. Note that the curve representing ELFORM -1 is covered by
ELFORM -2.

Table 4.2: The simulation time for the cantilever beam test with solid element formulations.
ELFORM Time [s] Time step [ms]
1 69 1.96 · 10−4
2 125 1.96 · 10−4
3 411 1.96 · 10−4
-1 149 1.96 · 10−4
-2 472 1.96 · 10−4

4.1.2 Thin-shell elements

The shell elements are two-dimensional elements based on a combination of plane stress assumptions and plate
theory. This assumption is considered as reasonable for thin-walled structures [9]. The shell element describes
the mid surface of the structure. A schematic figure of a thin-shell element is presented in Figure 4.4

22
 z
h/2
x
y h/2

Figure 4.4: Schematic figure of a thin-shell element, with the mid surface (gray) and the thickness offset h/2
(dashed).

According to Stelzmann, 3-5 integration points (or more) per element are recommended for a non-linear
material model. Only two of the available shell elements are recommended; (i) the Belytschko-Tsay element
and (ii) the fully integrated shell element. The Belytschko-Tsay element is recommended for simulations that
need to give results at low cost and the fully integrated element is recommended for higher accuracy. [17]

(i) Belytschko-Tsay (ELFORM 2) is a standard element with one point integration, and
should according to Stelzmann be used with 3 through thickness integration points
(NIP) and the hourglass control type 4 to avoid spurious zero energy modes. In
CONTROL SHELL it is recommended to use ISTUPD=4, BWC=1 and PROJ=1.
[17]
(ii) The fully integrated element (ELFORM 16) is intended for use when accuracy is
needed. Simulations using this element formulation are 2.5 times slower than the
Belytschko-Tsay element. The formulation is recommended by Stelzmann to use with
5 through thickness integration points (NIP) and hourglass control type 8 to avoid
spurious energy modes. In CONTROL SHELL it is recommended to use ISTUPD=4.
[17]
The result from the cantilever beam test is presented in Figure 4.5. The simulation times are listed in
Table 4.3. From the simulations it can be seen that element formulations 2 and 16 performs almost equally well
in predicting the force-displacement. Element formulation 16 takes approximately two times longer to run than
element formulation 2 as can be seen in Table 4.3, which is close to the difference predicted by Stelzmann [17].
It is worth noting that the fully integrated thin-shell element (ELFORM 16) performs well in computational
cost compared to the results for solid elements in Table 4.2 and the thick-shell elements in Table 4.4. It is also
interesting to note that the time steps for the thin-shell elements are larger than the time step used for the
solid elements in Table 4.2 and the thick-shell elements in Table 4.4. This indicates that the shell elements offer
a substantially faster simulation when each time step takes longer time to evaluate (e.g. when the material
model is changed from an elastic model to a composite model).

23
 ×10−3
4.5
Analytical
4 Shell ELFORM 2
Shell ELFORM 16
3.5

3
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.5: The result from the cantilever beam bending test, where the following element types are tested;
ELFORM 2 (blue) and ELFORM 16 (red). The analytical result is represented by a black line.

Table 4.3: The simulation time for the cantilever beam test for thin-shell element formulations.
ELFORM Time [s] Time step [ms]
2 46 3.32 · 10−4
16 87 3.32 · 10−4

4.1.3 Thick-shell elements

The thick-shell element is an 8-noded element based on shell theory but with an added strain component
through thickness [10]. The element has 8 corner nodes according to Figure 4.6.

z
h/2
x
y h/2

Figure 4.6: Schematic figure of a thick-shell element, with the mid surface (gray) and the thickness offset h/2.

In LS-DYNA there are four types of thick-shell elements; (i) one point reduced integration, (ii) selectively
reduced 2x2 in plane integration, (iii) assumed strain 2x2 in plane integration, and (iv) assumed strain reduced
integration. [17]

24
 (i) The one point reduced integration element (ELFORM 1) is a plane-stress formulation,
like a thin-shell element and has one in-plane integration point. The thickness stiffness
is modelled by a penalty function between the top and bottom nodes, and thickness
changes are imposed by the membrane strain as for a regular thin-shell element. The
element formulation can behave too softly and hourglass modes are possible. [17]
(ii) The selectively reduced 2 x 2 in plane integration formulation (ELFORM 2) is a fully
integrated element formulation with a plane-stress assumption. The thickness stiffness
is modelled by a penalty function between the top and bottom nodes, and thickness
changes are imposed by the membrane strain as for a regular thin-shell element. The
accuracy of the element is comparable to the Belytschko-Tsay thin-shell element, but
it is 7-8 times more expensive. [17]
(iii) The assumed strain 2 x 2 in plane integration formulation (ELFORM 3) is a 3D
formulation like a solid element and the thickness changes are imposed by the out-
of-plane stress. The formulation is slow (approximately 65 times slower than the
Belytschko-Tsay thin-shell element), and at least two elements in the thickness direction
are needed for an accurate result in bending simulations. [17]
(iv) The assumed strain reduced integration (ELFORM 5) is a 3D formulation like a solid
element and the thickness changes are imposed by the out-of-plane stress. The element
formulation is developed to model thick composites with the option of including
laminate shell theory. The element formulation is fast, approximately 1.5 times slower
than the Belytschko-Tsay thin-shell element and the formulation can handle bending
simulation with only one element through thickness. Hourglass modes and shear
locking are stabilised using an assumed strain method. [17]
The result from the cantilever bending test is presented in Figure 4.7 and the simulation times are presented
in Table 4.4. It can be seen that element formulation 3 is the slowest of the thick shells, just as stated
by Stelzmann [17]. The setup in the simulation only uses one element through thickness, and for element
formulation 3 it is recommended to use at least 2 elements through thickness. This could explain the inaccurate
results for this element type. Since it is both extremely slow and inaccurate this element formulation is left out
from further analysis. Element formulations 1, 2 and 5 are equally good at predicting the force-displacement,
and since neither of the element formulations are exceptionally slow all 3 formulations are kept for further
analysis. ELFORM 5 is the least costly of the three formulations and therefore becomes the preferred choice.
It can also be seen that the time steps in the simulations are larger for the shell-type elements (thick-shell
ELFORM 1 and 2) compared to the time steps for the solid-type elements (thick-shell ELFORM 3 and 5). As
for the case when comparing thin-shell elements and solid elements.

×10−3
4.5
Analytical
4 Tshell ELFORM 1
Tshell ELFORM 2
3.5 Tshell ELFORM 3
Tshell ELFORM 5
3
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.7: The result from the cantilever beam bending test, where the following element types are tested;
ELFORM 1 (blue), ELFORM 2 (red), ELFORM 3 (green) and ELFORM 5 (orange). The analytical result is
represented by a black line.

25
 Table 4.4: The simulation time for the cantilever beam test for thick-shell element formulations.
ELFORM Time [s] Time step [ms]
1 96 2.3 · 10−4
2 248 2.3 · 10−4
3 1069 1.96 · 10−4
5 79 1.96 · 10−4

4.1.4 Brief conclusion

From the result of the cantilever beam test it can be concluded that the most efficient element formulations,
based on the stiffness response and simulations time, are solid ELFORM 3, thin-shell ELFORM 2, and thick-shell
ELFORM 5. These three formulations are used as a starting point for the subsequent simulations. Note
that the performed tests simply provides information about the stiffness response and the simulation time.
The results may therefore only be considered a suggestion of which element formulation that are suitable for
modelling the plies of the composite laminate. If there are indications of the element types not giving a reliable
response in any further analysis, e.g. hourglass modes, the formulation will be changed into any of the other
remaining suitable formulations.

4.2 Composite material models in LS-DYNA

The material data used in the material cards is specified in a report from Bru et al [12] and is listed in
Section 3.1. Since the material used in the experiment is a well characterised composite material, most physical
parameters are available. However, if the data is not available for any of the material cards, the parameters are
estimated based on the data from these given material properties.
In LS-DYNA there are several material models for composites available. These differ in the failure criteria
used and the type of elements accepted. The composite material models applicable for the modelling are
described further below.

26
 MAT22 MAT COMPOSITE DAMAGE
Failure criterion: Chang-Chang, found in Appendix C.1.
Element types accepted: Thin-shell, Thick-shell, Solid
The material card represents an orthotropic material with an optional brittle failure.
[18]

MAT54/55 MAT ENHANCED COMPOSITE DAMAGE

Failure criterion (54): Chang-Chang, found in Appendix C.1.
Failure criterion (55): Modified Tsai-Wu and Chang-Chang, found in Appen-
dices C.2 and C.1
Element types accepted: Thin-shell, Thick-shell, Solid
An enhanced version of MAT22. The model contains a crashfront algorithm, meaning
that once an element has failed it is deleted and the strength of surrounding element
is reduced according to a stated parameter. [18]

MAT58 MAT LAMINATED COMPOSITE FABRIC

Failure criterion: Modified Hashin, found in Appendix C.3. [18]
Element types accepted: Thin-shell, Thick-shell [18]
Material model to represent UD and woven composites as well as complete laminates.
[18]

MAT261 MAT LAMINATED FRACTURE DAIMLER PIHNO

Failure criterion: Pinho, found in Appendix C.4.
Element types accepted: Thin-shell, Thick-shell, Solid
The failure criteria of this model is physically based for all failure modes. It is a
continuum damage model and includes non-linear in-plane shear behaviour. [18]

MAT262 MAT LAMINATED FRACTURE DAIMLER CAMANHO

Failure criterion: Camanho, found in Appendix C.5.
Element types accepted: Thin-shell, Thick-shell, Solid
The failure criteria of this model is physically based for all failure modes. It is a
continuum damage model and includes a simplified non-linear in-plane shear behaviour.
[18]

4.2.1 Stiffness response

With the aim to detect if any of the composite material cards have a different and unwanted behaviour, a
cantilever beam test was performed for the different material cards. The setup of the test is identical to the
one described in Section 4.1. For this test, the given material properties are not used. Instead, the composite
material card was implemented to mimic an isotropic elastic material, i.e. with the same material parameters
in all directions and with the failure parameters set to values large enough such that failure will not occur (e.g.
Xc = 1000 GPa). The results are then compared with the response of the pure elastic material in the previous
cantilever beam test. The result is presented separately for each element type.

4.2.1.1 Solid elements

In Figure 4.3 and Table 4.2 it can be seen that the fully integrated quadratic 8 node element with nodal
rotations (ELFORM 3) gives a good force-displacement response with a reasonable simulation time. This
element is therefore used when implementing the different material models. The force-displacement curves
for the different materials are compared with the pure elastic material in Figure 4.8. The stiffness response is
identical to the elastic material for all the material models.
The simulation times are listed in Table 4.5. It can be seen that all material models increase the simulation
time, but not the time step. This means that each individual step is slower for the composite material model
than for the isotropic elastic material, just as anticipated. The simulation time differs between the material
models, where the simplest model (MAT22) has the shortest simulation time, while MAT261 has the largest
computational cost. This is expected due to the complexity of the models.

27
 ×10−3
4.5
Solid elastic material
4 Solid MAT22
Solid MAT55
3.5 Solid MAT261
Solid MAT262
3
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.8: A comparison of the response of the material models in the cantilever bending test. The lines
represent the pure elastic material (black), MAT22 (blue), MAT55 (red), MAT261 (orange) and MAT262
(purple).

Table 4.5: The simulation time for the cantilever beam test for solid elements and composite material models.
Material model Time [s] Time increase from reference [%] Time step [ms]
Elastic 411 - 1 .96 · 10 −4
22 588 43 1.96 · 10−4
55 661 60 1.96 · 10−4
261 1523 270 1.96 · 10−4
262 868 111 1.96 · 10−4

4.2.1.2 Thin-shell elements

In Figure 4.5 and Table 4.3 it can be seen that the fully integrated shell element (ELFORM 16) gives a good
force-displacement response for a reasonable simulation time. This element is therefore used when implementing
the different material models. The force-displacement curves for the different material models are compared
with the pure elastic material in Figure 4.9. It can be seen that for the lower displacements, MAT262 exhibits a
larger magnitude of the force oscillations. As the displacement increases this behaviour settles and all material
models behave similarly.
The simulation times and time step sizes are compared with the reference case (elastic material model)
in Table 4.6. The time step is decreased for the composite material models MAT22, MAT58, MAT261 and
MAT262, which implies a higher computational cost. For material model MAT55 the time step is of the same
size as for the elastic reference material, but still the simulation time is large. This suggests that each step is
computationally heavy, but that the model is more stable than the other composite material models.

28
 ×10−3
4.5
Shell elastic material
4 Shell MAT22
Shell MAT55
3.5 Shell MAT58
Shell MAT261
3 Shell MAT262
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.9: A comparison of the response of the material models in the cantilever bending test. The lines
represent the pure elastic material (black), MAT22 (blue), MAT55 (red), MAT58 (green), MAT261 (orange)
and MAT262 (purple). Note that MAT58, MAT261 and MAT262 exhibit the same behaviour, and therefore
MAT58 and MAT261 are hidden behind MAT262 in the figure.

Table 4.6: The simulation time for the cantilever beam test for thin-shell elements and composite material
models.
Material model Time [s] Time increase from reference [%] Time step
Elastic 87 - 3 .32 · 10 −4
22 120 38 3.0 · 10−4
55 154 77 3.32 · 10−4
58 269 209 3.0 · 10−4
261 78 359 3.0 · 10−4
262 190 118 3.0 · 10−4

4.2.1.3 Thick-shell elements

In Figure 4.7 and Table 4.4 it can be seen that the assumed strain reduced integration formulation (ELFORM
5) gives a reasonable force-displacement response with the shortest simulation time. This element is therefore
used when implementing the different material models. The force-displacement curves for the different materials
are compared with the pure elastic material in Figure 4.10. It can be seen that all material models behave very
similarly for the thick-shell elements.
The simulation times are compared with the reference case (elastic material model) in Table 4.7. In the
table it can be seen that the time step is not changed for any of the material models. Still, for all models the
simulation time has increased. For the thick-shell elements the increase is generally not as large as for the
solid and thin-shell elements in Table 4.5 and Table 4.6, and the reference time is the lowest of the three as
well. This indicates that from a computational cost aspect the thick-shell element is well suited for composite
materials. Note that the time step for MAT58 is not comparable to the time steps of the other material models
since the time step sizes for the elastic case using thick-shell ELFORM 1 and ELFORM 5 are not the same.

29
Table 4.7: The simulation time for the cantilever beam test for thick-shell elements and composite material
models. ? Note that MAT58 is run with ELFORM 1 and not ELFORM 5. ?? The elastic case for ELFORM 1 is
used as reference case.
Material model Time [s] Time increase from reference [%] Time step
Elastic 79 - 1 .96 · 10 −4
22 108 37 1.96 · 10−4
55 124 77 1.96 · 10−4
58? 139 45?? 2.08 · 10−4
261 266 177 1.96 · 10−4
262 130 86 1.96 · 10−4

×10−3
4.5
Tshell elastic material
4 Tshell MAT22
Tshell MAT55
3.5 Tshell MAT58
Tshell MAT261
3 Tshell MAT262
Force [kN]

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.10: A comparison of the response of the material models in the cantilever bending test. The lines
represent the pure elastic material (black), MAT22 (blue), MAT55 (red), MAT58? (green), MAT261 (orange),
and MAT262 (purple). ? Note that MAT58 is run with ELFORM 1 and not ELFORM 5

4.2.2 Behaviour at failure

The failure behaviour of the material models is investigated by a simply supported three point bending test,
where the specimen is bent until failure with the aim of eliminating combinations of material models and
element types that behave in an unreasonable way. The setup of the test is illustrated in Figure 4.11, where
a displacement is applied to the middle of the beam. The supports are modelled using boundary conditions
(BOUNDARY SPC(SET)), and the mesh consists of 50 elements in the length direction, 7 in the width direction
and 1 element through the thickness. The geometric parameters used in the three point bending to failure are
presented in Table 4.8, together with the material data used for the analytical calculations. The displacement
is increased linearly until failure occurs, with a velocity of 0.05 m/s, and imposed using a boundary condition
(BOUNDARY PRESCRIBED MOTION). The parameters of the material models are listed and described in
Appendix A. Identical tests with the same mesh sizes are performed for all applicable material models, for solid,
thin-shell, and thick-shell elements. Note that the strain failure is captured using the ADD EROSION card.
To give an indication of the performance in both longitudinal and transversal directions of the material
model, the bend to failure is simulated for both fibre angles θ = 0◦ and θ = 90◦ . At failure, the displacement
and the corresponding applied force is compared with analytical results in order to verify the accuracy of the
analysis. Note that the analytical force/displacement is not an accurate result, but gives an indication of the
order of magnitude of the results. A failure at a force or displacement very far from the analytical result is
considered as incorrect. In the analytical solution the material is assumed to be isotropic elastic with the
mechanical properties taken from the desired fibre angle (i.e. E11 /Xc /ε11cu and E22 /Yt /ε22tu respectively).
In the analytical calculations, the failure may be initiated due to reaching the maximally permitted strain or
stress.

30
 F

t
L w
(a) (b)
Figure 4.11: A schematic figure of (a) the three point bending test setup, performed to test the failure of the
matrial models, and (b) the beam cross section. The length is referred to as L, the width as w and the thickness
as t.

Table 4.8: Mechanical and geometric properties for the three point bending analysis
Property Value
L 150 mm
t 1.98 mm
w 20 mm
0◦ 90◦
E 132 GPa 9 GPa
εmax 0.49 % 0.32 %
σmax 631 MPa 29.2 MPa

The force and displacement at failure in the bending test are estimated by an elementary case where the
maximal stress and the maximum bending moment for an applied force, Mmax = F4L and σmax = ± 6M wt2 , are
max

combined to obtain the maximal force of the beam, see Equation (4.2). [19]

2σmax wt2
Fmax,stress = (4.2)
3L
The displacement δmax,stress corresponding to the force to failure Fmax,stress is calculated from an elementary
case, see Equation (4.3). [14]

Fmax L3 σmax L2
δmax,stress = = {insert Fmax,stress and I} = (4.3)
48EI 6Et
With the (incorrect) assumption of isotropic, elastic material Hooke’s law is used to replace the maximum
stress with the maximum strain in Equation (4.2) and Equation (4.3). [14]

2Eεmax wt2
Fmax,strain = (4.4)
3L
εmax L2
δmax,strain = (4.5)
6t
The analytical solution is presented in Table 4.9.

Table 4.9: The analytically obtained displacement and force at failure for both a ply with fibre angle 0◦ (fibres
along the length of the beam) and with fibre angle 90◦ (fibres along the width of the beam).
Failure mode θ = 0◦ θ = 90◦
F = 224.4 [N] F = 10.4 [N]
Stress
δ = 9.0 [mm] δ = 6.1 [mm]
F = 150.2 [N] F = 15.7 [N]
Strain
δ = 6.0 [mm] δ = 9.2 [mm]

31
4.2.2.1 Solid elements
The displacement and force at failure for the different material models, when modelling using solid elements,
are listed in Table 4.10. All material models give reasonable failure responses compared to the analytical
result. However, the result of material model MAT22 differs for the 90◦ ply from the remaining materials. This
indicates that this material model should not to be used with this element type.

Table 4.10: The displacement and force, for the different composite material models, at failure in a bending
test using solid elements.
Material δ0◦ [mm] F0◦ [N] δ90◦ [mm] F90◦ [N]
MAT22 10.8 275 6.16 12.7
MAT55 10.7 274 12.06 19.7
MAT261 10.7 274 11.05 18.6
MAT262 10.8 276 10.73 17.5

4.2.2.2 Thin-shell elements

The displacement and force at failure for the different material models, when modelling using thin-shell elements,
are listed in Table 4.11. The results of the bending test for MAT262 differs significantly from both the analytical
solution and the results of the remaining material models, for the 0◦ ply. This indicated that thin-shell elements
are not compatible with this material model. The material models MAT22 and MAT58 exhibit failure prior to
the remaining models for the 90◦ ply, which also suggests on an inaccuracy of the combination of the element
types and material models.

Table 4.11: The displacement and force, for the different composite material models, at failure in a bending
test using thin-shell elements.
Material δ0◦ [mm] F0◦ [N] δ90◦ [mm] F90◦ [N]
MAT22 8.1 141.1 7.76 13.3
MAT55 9.6 196.0 11.23 17.2
MAT58 9.7 237.5 8.27 14.1
MAT261 9.7 170.4 11.00 16.7
MAT262 3.6 86.4 11.22 17.1

4.2.2.3 Thick-shell elements

The displacement and force at failure for the different material models, when modelling using thick-shell
elements, are listed in Table 4.12. The combination of thick-shell elements and the material model MAT55
resulted in an unrealistic displacement, and no failure was initiated. This implies that thick-shell elements and
MAT55 are not compatible. All remaining material models give reasonable response. However, like for the
thin-shell elements the material models MAT22 and MAT58 exhibit failure prior to the remaining models for
the 90◦ ply. Note that MAT58 is not compatible with thick-shell elements of formulation ELFORM 5. These
results are instead obtained using ELFORM 1.

Table 4.12: The displacement and force, for the different composite material models, at failure in a bending
test using thick-shell elements. ? Note that MAT58 is run with ELFORM 1 and not ELFORM 5
Material δ0◦ [mm] F0◦ [N] δ90◦ [mm] F90◦ [N]
MAT22 11.5 283 7.89 13.1
MAT55 - - - -
MAT58? 11.1 272 8.4 13.8
MAT261 11.3 279 11.96 20.2
MAT262 11.3 279 11.78 16.8

32
4.2.3 Brief conclusion
MAT22 and MAT58 exhibit lower values of the displacement and force at failure for 90◦ for all element types
(except for the combination of MAT58 and solid elements, since this combination is not compatible). However,
this is not the case for the 0◦ ply.
Two different combinations are considered not compatible based on these results; solid elements combined
with MAT22 and thin-shell elements together with MAT262 exhibit significantly difference failure values than
the remaining combinations. From the analysis it can also be concluded that MAT55 does not work as intended
with thick-shell ELFORM 5.
Note that all material models except MAT261 and MAT262, have unphysical parameters that need calibration.
This tuning has not been performed and some results may therefore be misleading. However, since MAT261
and MAT262 are based on physical data, these models are the preferred choice since a new material can
be implemented without tuning and estimations (if all physical parameters are known). Both MAT261 and
MAT262 behave reasonably in these simulations and therefore they have a great advantage over the other
models. MAT262 gives these results at a lower computational cost than MAT261, and is therefore considered
the best choice of material model when taking both accuracy and computational cost into account.

4.3 Adhesive modelling

In order to capture the delamination behaviour of the simulation model, some sort of adhesive modelling
is required both between the steel and the composite laminate, and between the plies of the laminate. In
LS-DYNA, there are two available approaches; tiebreak contact condition and cohesive zone. The first option is
a contact condition which as the name reveals ties the parts together and then breaks the contact at a given
condition. The cohesive zone consists of a certain type of solid elements, that is implemented between the
parts and fails at a given condition. In this section, these adhesive modelling approaches are further described
and a comparison between them is conducted. The results are presented separately for each element type. As
described in Section 2.1.3, the delamination may occur in three different modes. Two of these, mode I and
mode II, are of interest for the adhesive modelling, and will be investigated by a double cantilever beam (DCB)
test and an end notched flexure (ENF) test respectively.

4.3.1 Tiebreak contact condition

The tiebreak condition is a contact condition which allows for separation according to an optional failure
criteria between the two bodies in contact. The contact is penalty based, and no sliding is allowed between
the included elements. The tiebreak conditions in LS-DYNA are contact conditions that transfer both tensile
and compressive forces, unlike standard contact conditions that only transfer compressive forces. The tiebreak
conditions also contain a failure criterion, and when the condition is met the bodies no longer remain in contact.
The failure criteria in the tiebreak contact condition only includes the force/stress to failure. This means that
no energy is considered in the failure, and this is not considered as a physically correct model. [20]
The tiebreak contact condition AUTOMATIC SURFACE TO SURFACE TIEBREAK is used in the simu-
lations. The contact is described in Appendix B. In order to determine the behaviour of the tiebreak condition
in combination with different element types, separate simulations with solid, thin-shell and thick-shell elements
are run.

4.3.2 Cohesive zone modelling

When using a cohesive zone to model the crack propagation, cohesive elements are implemented between the
layers. The nodes of the cohesive elements are merged with the nodes of the adjacent plies. The element may
be regarded as nonlinear springs, representing the behaviour of the adhesive [21], see Figure 4.12a. A zero
thickness is allowed, inducing a two dimensional behaviour. Instead of strains, the deformation is modelled by
the relative displacement between the two surfaces interpolated to the Gauss points. The force per unit area
corresponding to the displacement is determined by a cohesive material model in shape of a traction-separation
curve for each mode. This governs the behaviour of the element. When the separation at failure is reached (i.e.
the traction is 0 for a non-zero separation), the element is eroded.
There are a few cohesive elements in LS-DYNA, all of them are solid elements, since two nodes through
the thickness are required in order to not lock the two bodies together in the same node. As stated earlier,

33
LS-DYNA allows for cohesive elements without thickness, meaning that solids can be modelled with initial
contact with an adhesive between the two bodies (if d = 0). The configuration of a cohesive element between
two solids is shown with a thickness d in Figure 4.12c. For thin-shells the cohesive element spans from the mid
surfaces as can be seen in Figure 4.12b. [22]

h2
h2
d≥0 d>0 d≥0
h1 h1

(a) (b) (c)

Figure 4.12: A schematic figure of (a) 2-dof spring configuration of a cohesive element (yellow) and how it may
be used for connecting both (a) thin-shell elements and (b) solid or thick-shell elements. Note that the cohesive
element may have a thickness d = 0, meaning initial contact between the solid or thick-shell elements. The
structure of the cohesive element consists of vertical and horizontal springs, corresponding to mode I and mode
II respectively. Figures reproduced and altered from Adhesives modelling with LS-DYNA: Recent developments
and future work [22].

In LS-DYNA the solid elements ELFORM 19 and ELFORM 20 are intended for use with cohesive material
models. Element formulation 19 defines the traction in the mid surface as the mid points between the nodes,
and the element has four integration points. The element accepts an initial volume of zero, meaning that it
can be used for adhesive bonding between solid elements. ELFORM 20 is identical but has offsets for shells.
Note that there is an additional cohesive material card, MAT169 MAT ARUP ADHESIVE, that is intended
for adhesive bonding in aluminium structures and can be used with other types of elements than ELFORM 19
and ELFORM 20. [1]

MAT138 MAT COHESIVE MIXED MODE

Damage in the interface is considered in this model using irreversible conditions. [23]

MAT184 MAT COHESIVE ELASTIC

Simple cohesive model with three stress outputs instead of the usual six. [23]

MAT185 MAT COHESIVE TH

Tri-linear traction-separation law. Three stress outputs instead of the usual six. [23]

MAT186 MAT COHESIVE GENERAL

Arbitrary traction separation law. Damage of the interface is considered in this model
using irreversible conditions. [23]
For the cohesive elements, the material type MAT186 is used due to it allowing an arbitrary traction
separation law. Since all data for the material card is provided the general cohesive model is used. The material
data has been supplied by Gestamp HardTech for the glue, and from Bru et al. [12] for the composite interface.
This configuration allows for both movement without failure and separation at failure.

4.3.3 DCB test

The double cantilever beam (DCB) test is intended to investigate the delamination behaviour of mode I.
This test consists of a specimen with an initial crack with length a0 . On each side of the opening there is
a force (F ) opening the crack, with equal size of both sides, see Figure 4.13. The forces are added using
an applied displacement in the end nodes using BOUNDARY PRESCRIBED MOTION. The fixed edge
is implemented using BOUNDARY SPC(SET) with all degrees of freedom fixed. The parameters for the
simulations are presented in Table 4.13, and the results obtained for the different configurations are compared
to the force-displacement curve in Figure 4.14.
In the tests an elastic model with the longitudinal stiffness of the composite is used, and the properties

34
used together with the measurements presented in Table 4.13. Each lamina ply is modelled with one layer of
elements through thickness and each element is 1×1 mm, i.e. a total of 20×150 elements in each ply. This
mesh, including dimensions and mesh refinement, is based on a report from Turon et al. In the Turon et al.
report it can be seen that the performance of the cohesive elements is heavily mesh dependent, and since
no mesh convergence study is performed a mesh already concluded to give accurate results at a reasonable
computational cost is chosen. [24]

Table 4.13: Material properties and measurements for the DCB simulations.
Property Value
E 132 GPa
ν 0.3
ρ 1.5 g/cm3
GIC 149 J/m2
Yt 29.2 MPa
L 150 mm
a0 58 mm
t 1 mm
w 20 mm

t
t
L a0
F

Figure 4.13: Schematic figure of Mode I/DCB test. The two plies (grey) with length L, thickness t and w are
pulled apart by two forces F (red). Between the plies are an adhesive (yellow) and an initial opening of length
a0 .

To predict the behaviour of the crack growth along the interface, an analytical solution is calculated. The
analytical solution of the DCB test is divided into two parts; (i) before the crack grows and (ii) during the
crack growth. The calculations are based on an elementary case, see Equation (4.6) where a is the crack size,
E the Young’s modulus and I the second moment of inertia [14]. The crack size is determined according to
Equation (4.7). The crack starts to grow when the energy release (GI ) reaches the critical value (GIC ) [25].
The behaviour according to the hand calculations, with the properties for the DCB simulations described above,
is presented in Figure 4.14. The force at initial crack propagation is approximately 14 N at a tip displacement
of just above 4 mm.

3EI
F = δ (4.6)
a3
a0 (i)
a =  9EIδ2 1/4 (4.7)

wGIC (ii)

35
 0.015

0.01
Force [kN]

0.005

Before crack growth

Crack growth
0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
Figure 4.14: The force-displacement relation for the analytic Mode I/DCB calculations.

In the analysis for the different element types, the fracture zone is modelled using either a tiebreak contact
condition (using Yt as failure condition) or a cohesive zone (using both GIC and Yt as failure condition). The
two beams are modelled using an elastic material model (MAT1). The configurations are simulated for solid,
thin-shell, and thick-shell elements and the results are presented for each element type separately.

4.3.3.1 Solid elements

The force-displacement relation obtained in the DCB test, when modelling using solid elements, is presented
in Figure 4.15. Both the tiebreak condition and the cohesive zone are included, as well as the analytical
solution. It can be seen that both adhesive modelling methods exhibit the desired behaviour. However, the
curve representing the cohesive zone simulation demonstrates a more stable behaviour that is similar to the
analytical solution. The tiebreak condition exhibits larger drops in force during the crack propagation process,
which could be an indication that the contact is released, i.e. the criteria is met, for a large area simultaneously.
The magnitude of force that is required to initiate crack growth for the tiebreak condition is significantly lower
than the analytically predicted force. This is something that could be solved by tuning the parameters in the
contact card, but since a parameter that requires tuning is undesirable this is considered as a drawback of
the solid element - tiebreak condition combination. For both simulations the response is stiffer than in the
analytical case. This comes from using ELFORM 3, which in Figure 4.3 is shown to be overly stiff.
Although the cohesive zone modelling comes at a higher computational cost, see simulation times listed in
Table 4.14, the cohesive zone is the preferred method based on the obtained response.

36
 0.02
Analytical
0.018
Solid with cohesive zone
0.016 Solid with tiebreak

0.014
Force [kN]

0.012

0.01

0.008

0.006

0.004

0.002

0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
Figure 4.15: Comparison of the force-displacement relation for the DCB/Mode I simulations with cohesive
elements and tiebreak condition as adhesive model for the solid element formulation.

Table 4.14: Simulation times for the DCB test when modelling with solid elements.
Adhesive model Time [s]
Cohesive zone 10596
Tiebreak condition 7628

4.3.3.2 Thin-shell elements

The force-displacement relations for the modelling using thin-shell elements are presented and compared to the
analytical result in Figure 4.16. The configuration with the cohesive zone exhibits a reasonable behaviour, but
vibrates heavily. The tiebreak condition configuration initially behaves as the analytical solution. However, after
this initial behaviour, the failure criteria of the contact is reached earlier than expected (at a displacement of
approximately 3 mm), and a response dissimilar to the analytical solution arises. Neither of the two simulations
gives an acceptable result.
The simulation times are listed in Table 4.15. It may be seen that the cohesive zone has a higher computational
cost than the tiebreak condition. Due to the undesired response, none of the thin-shell configurations in the
DCB test are recommended.

37
 0.02
Analytical
0.018
Thin-shell with cohesive zone
0.016 Thin-shell with tiebreak

0.014
Force [kN]

0.012

0.01

0.008

0.006

0.004

0.002

0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
Figure 4.16: Comparison of the force-displacement relation for the DCB/Mode I simulations with cohesive
elements and tiebreak condition as adhesive model for the thin-shell element formulation.

Table 4.15: Simulation times for the DCB test when modelling using thin-shell elements.
Adhesive model Time [s]
Cohesive zone 3240
Tiebreak condition 2551

4.3.3.3 Thick-shell elements

In Section 4.1, it was recommended to use element formulation ELFORM 5 when modelling thick-shell elements.
However, some unstable behaviour was detected in the testing of this section. The thick-shell elements are
therefore henceforth modelled using element formulation ELFORM 1, with hourglass control activated. The
difference is illustrated in Figure 4.17 where the results for both ELFORM 1 and ELFORM 5 are presented. It
can be seen that while the ELFORM 1 configuration behaves as intended, the ELFORM 5 response is vibrating
heavily.

0.02
Analytical
0.018
Thick-shell ELFORM 1 with cohesive zone
0.016 Thick-shell ELFORM 5 with cohesive zone

0.014
Force [kN]

0.012

0.01

0.008

0.006

0.004

0.002

0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
Figure 4.17: Comparison of the force-displacement relation for cohesive zone using ELFORM 1 (green) and
ELFORM 5 (orange).

38
 For the thick-shell element, the force-displacement relations obtained in the DCB test are presented and
compared to the analytical result in Figure 4.18. It can be seen that both adhesive models exhibit the desired
behaviour, but the force-displacement curve obtained when modelling using tiebreak condition is not as smooth
as the one obtained using a cohesive zone or the analytical results. This indicates that large areas are released
from contact at the same displacement. Additionally, the tiebreak condition does not sustain enough loading
until the crack propagation is initiated. This is something that could be solved by tuning the parameters in the
contact card, but since a parameter that requires tuning is undesirable this is considered as a drawback of the
thick-shell element - tiebreak contact condition combination.

0.02
Analytical
0.018
Thick-shell with cohesive zone
0.016 Thick-shell with tiebreak

0.014
Force [kN]

0.012

0.01

0.008

0.006

0.004

0.002

0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
Figure 4.18: Comparison of the force-displacement relation for the DCB/Mode I simulations with cohesive
elements and tiebreak condition as adhesive model for the thick-shell element formulation.

Table 4.16: Simulation times for the DCB test when modelling using thick-shell elements.
Adhesive model Time [s]
Cohesive zone 3768
Tiebreak condition 1988

4.3.4 ENF test

In order to study the response when a crack propagates along the ply interface according to mode II, an end
notched flexure (ENF) test is performed. A schematic figure of the ENF test is presented in Figure 4.20. The
test consists of a specimen with an initial crack of length a0 . The specimen is placed on two supports, both
modelled using the boundary condition BOUNDARY SPC(SET), with a velocity of 0.05 m/s. For the pinned
support (left) all translational degrees of freedom are fixed, while for the roller support (right) all translational
degrees of freedom are fixed except for the movement along the length of the beam. For the solid and thick-shell
elements the boundary condition is applied on the bottom nodes in the thickness direction. A force (F ) is
applied at the midpoint of the beam. The force is added using an applied displacement in the middle nodes
of the top ply modelled with BOUNDARY PRESCRIBED MOTION. For the solid and thick-shell elements
the upper nodes in the thickness direction are used for the applied displacement. The parameters for the
simulations are presented in Table 4.17. An elastic material (MAT1) is used for modelling the plies. The same
mesh as for the DCB tests is used in the ENF simulations, which means that each layer is modelled with one
ply through thickness and the element size is 1×1 mm. This gives a total of 20×150 elements in each ply.
In the analysis the fracture zone is modelled using either a tiebreak contact condition (using S12 as failure
condition) or a cohesive zone (using both GIIC and S12 as failure condition).

39
 Table 4.17: Material properties and measurements for the ENF simulations.
Property Value
E 132 GPa
ν 0.3
ρ 1.5 g/cm3
GIIC 690 J/m2
S12 77.8 MPa
L 150 mm
a0 58 mm
t 1 mm
w 20 mm

t
t

L a0

Figure 4.19: Schematic figure of Mode II/ENF test. The two plies (gray) with length L, thickness t and width
w with an applied force F (red) in the middle of the beam. Between the plies are an adhesive (yellow) and an
initial opening of length a0 .

To predict the behaviour of the crack growth along the ply interface, an analytical solution is calculated. The
analytical solution of the ENF test is divided into four parts; (i) before the crack grows, (ii) the crack growth
until the crack spans to the midpoint of the beam, (iii) the crack growth from the midpoint until the crack spans
over the entire length and (iv) after the crack spans over the entire length of the beam. The force-displacement
relation for each part can be seen in Equations (4.8)-(4.11). The fourth part of the calculations is after the
crack spans over the entire length, meaning bending of two beams that slide on top of each other [25]. This part
is not included in the solution, due to only the crack propagation being of interest. The behaviour according to
the hand calculations, with the properties for the ENF simulations described above is presented in Figure 4.20.
The force at initial crack propagation is approximated to 135N at a displacement of 9 mm.

2L3 + 3a30
δi = F (4.8)
96EI !
3/2
F 3 (64GIIC wEI)
δii = 2L + √ (4.9)
96EI 3F 3
!
3/2
F 3 (64GIIC wEI)
δiii = 2L − √ (4.10)
24EI 4 3F 3
L3
δiv = F (4.11)
12EI

40
 0.16

0.14

0.12
Force [kN]

0.1

0.08

0.06

0.04
Before crack growth
0.02 Crack growth to midpoint
Crack growth after midpoint
0
0 5 10 15
Displacement [mm]
Figure 4.20: The force-displacement relation for the analytic ENF/Mode II calculations.

4.3.4.1 Solid elements

In Figure 4.21, a comparison of the force-displacement response obtained for the solid elements in the ENF test
is presented for both adhesive modelling methods. The response of the cohesive zone, the tiebreak condition,
and the analytical solution is included. It can be seen that the tiebreak condition does not correlate with the
analytical force-displacement relation. The behaviour of the tiebreak simulation indicates that the criteria for
failure is met immediately, and therefore it behaves as if the two plies are not tied together in any way. This
is not a reasonable response. It can also be seen that the behaviour of the cohesive element configuration is
similar to the analytical predicted response.
The simulation times are listed in Table 4.18. The tiebreak condition is more time efficient than the cohesive
zone, but as stated earlier an unreasonable force-displacement response is obtained for the combination of solid
elements and the tiebreak contact condition.

0.16

0.14

0.12
Force [kN]

0.1

0.08

0.06

0.04
Analytical
0.02 Solid with cohesive zone
Solid with tiebreak
0
0 5 10 15
Displacement [mm]
Figure 4.21: The force-displacement relation obtained from the ENF simulations when modelling the plies using
solid elements. Both a cohesive zone (blue) and a tiebreak condition (red), was used to model the adhesive. The
analytical solution (black) is included for a comparison.

41
 Table 4.18: The simulation time for the ENF test when modelling using solid elements.
Adhesive model Time [s]
Cohesive zone 13984
Tiebreak 11893

4.3.4.2 Thin-shell elements

A comparison of the force-displacement responses obtained for the two different adhesive modelling methods
in the ENF test are presented in Figure 4.22. Thin-shell elements were used for modelling the plies. Both
configurations behave unreasonable compared to the analytical solution. The configuration with cohesive
elements does not sustain enough displacement until failure and is vibrating heavily. The vibrations may
introduce local stress concentrations and this could be the reason for the early failure. Even though the
cohesive elements are eroded too quickly, the behaviour has the same pattern as the analytical result, and with
some modifications this model may be improved to resemble the desired curve. The tiebreak condition fails
immediately and give an unphysical response. This adhesive modelling gave an error and the simulation was
terminated.
The simulation times are listed in Table 4.19. Note that due to the termination of the tiebreak condition
simulation, no simulation time may be listed for this configuration.

0.16

0.14

0.12
Force [kN]

0.1

0.08

0.06

0.04
Analytical
0.02 Thin-shell with cohesive zone
Thin-shell with tiebreak
0
0 5 10 15
Displacement [mm]
Figure 4.22: The force-displacement relation obtained from the ENF simulations when modelling the plies using
thin-shell elements. Both a cohesive zone (blue) and a tiebreak condition (red), were used to model the adhesive.
The analytical solution (black) is included for a comparison.

Table 4.19: The simulation time for the ENF test when modelling using thin-shell elements.
Adhesive model Time [s]
Cohesive zone N/A
Tiebreak 3909

4.3.4.3 Thick-shell elements

A comparison of the response of the adhesive modelling methods (cohesive zone and tiebreak contact condition)
in the ENF is presented in Figure 4.23. It can be seen that the configuration using tiebreak condition behaves
unreasonable and very stiffly compared to the analytical solution. The behaviour of the thick-shell elements
combined with cohesive elements is very similar to the analytical behaviour.
The simulation time for both the cohesive zone and the tiebreak contact condition are listed in Table 4.20.

42
 0.16

0.14

0.12
Force [kN]

0.1

0.08

0.06

0.04
Analytical
0.02 Thick-shell with cohesive zone
Thick-shell with tiebreak
0
0 5 10 15
Displacement [mm]
Figure 4.23: Comparison of the force-displacement relation for the ENF/Mode II simulations with cohesive
elements and tiebreak condition as adhesive model for the thick-shell element formulation.

Table 4.20: The simulation time for the ENF test when modelling using thick-shell elements.
Adhesive model Time [s]
Cohesive zone 5483
Tiebreak 3160

4.3.5 Brief conclusion

The crack propagation along the interface of two plies, i.e. the delamination behaviour, was investigated
using two different tests; DCB and ENF. Based on this, the behaviour of the adhesive modelling methods,
tiebreak condition and cohesive zone, in combination with both solid, thin-shell, and thick-shell elements was
investigated.
For solid elements, the cohesive zone gave a good response for both mode I and mode II. For mode I, the
curve at crack growth is not as smooth as the analytical solution, but clearly the wanted behaviour is reached.
The tiebreak condition gave a reasonable response for mode I, but with a too early initiated crack growth and
an unstable force-displacement relation. For mode II, an unrealistic result was obtained.
In both tests, the thin-shell elements exhibited large vibrations and none of the adhesive modelling methods
gave results that corresponded to the analytical results. The thin-shell are therefore excluded from all further
analysis.
For thick-shell elements, the cohesive zone gave good results close to the analytical solution, especially for
mode II. As for solid elements, the curve is not as smooth as the analytical solution for mode I, but the desired
behaviour is clearly implied. The tiebreak condition gave a reasonable result for mode I crack propagation.
However, as for the solid elements the crack growth was initiated at a displacement lower than indicated in the
analytical calculations. Compared to the cohesive response, the tiebreak condition provides an unstable curve.
For mode II, the tiebreak condition gave unreasonable results.
It is concluded that adhesive modelling using a tiebreak condition does not perform as well as using a
cohesive zone. Note that the tiebreak condition most likely can be calibrated to work as intended for both
modes. For the cohesive zone on the other hand, no calibration was needed, and the initial configuration
gave reasonable results for both thick-shell and solid elements. Therefore, this adhesive modelling method
is to be preferred. Additionally, the cohesive zone may result in a smoother curve if a mesh refinement is
performed. However, this is not of interest in this case due to the higher computational cost the increase in
number elements would induce.
Based on that the combination of thick-shell elements ELFORM 1 and a cohesive zone results in a simulation
time that is less than half of the one using solid elements, this combination is chosen to be further analysed.

43
4.4 Combining composite material model and adhesive model
The purpose of this section, is to determine which combinations of the different composite material models
and the adhesive modelling that are compatible. The same tests as used to analyse the adhesive modelling
behaviour are repeated in this section; DCB and ENF tests.
In Section 4.3, it was determined that the adhesive was best modelled using a cohesive zone. This, together
with thick-shell elements (ELFORM 1) gave the most efficient way of modelling the delamination of mode I
and mode II. Therefore, this configuration is used in the simulations.
In Section 4.2, it can be seen that all listed composite materials are compatible with the thick-shell elements
except for MAT55 that does not exhibit a proper failure behaviour.
The force-displacement relation is illustrated in Figure 4.24 for the elastic case as well as with the different
material models. It can be seen that for the DCB test, three material models gives a reasonable response
similar to the analytical one; MAT22, MAT58 and MAT262. For the ENF test, only two material models give
a reasonable response; MAT261 and MAT262. Note that MAT22 follow the same curve, but exhibit failure.
This material model has earlier been documented to reach the failure criteria in an early stage.

0.02
Analytical
0.018
Elastic
0.016 MAT22
MAT58
0.014 MAT261
MAT262
Force [kN]

0.012

0.01

0.008

0.006

0.004

0.002

0
0 1 2 3 4 5 6 7 8 9 10
Tip displacement [mm]
(a)
0.16
Analytical
0.14 Elastic
MAT22
0.12 MAT58
MAT261
MAT262
Force [kN]

0.1

0.08

0.06

0.04

0.02

0
0 5 10 15
Displacement [mm]
(b)
Figure 4.24: The force displacement relation for (a) the DCB simulations and (b) the ENF simulations using
different composite material models combined with thick-shell elements (ELFORM 1).

44
4.4.1 Brief conclusion
The only material model that exhibit a good response in both the DCB and the ENF test is MAT262. This
strengthens the conclusion of using this material model for modelling the composite laminate.

4.5 Version study in LS-DYNA

During the project, it was concluded that different versions of LS-DYNA gave unexpected differing results.
The presented results are not an accurate comparison between the versions, but a way of alerting users of the
varying results that different versions of LS-DYNA provide. Note that it is unclear if the results differ due
to the versions or the solver types; SMP (symmetric multiprocessing, for running on few CPUs, e.g. desktop
computer) or MPP (massively parallel processing, for running on many CPUs, e.g. a computer cluster).
A comparison of version R7.1.2 SMP and version R8.0.0 MPP has been performed, and a difference of the
simulation results has been discovered. A comparison of the force-displacement relation for a cantilever beam
test with shell elements (ELFORM 16) and MAT261 is presented in Figure 4.25. It can be seen that the results
differ between the versions. It is worth noting that for version R7.1.2 SMP, the vibrating increases without any
extra loading introduced, which is not considered as reasonable.

×10−3
4.5

3.5

3
Force [kN]

2.5

1.5

SMP R7.1.2
0.5
MPP R8.0.0
0
0 0.5 1 1.5 2 2.5
Tip displacement [mm]
Figure 4.25: A comparison of the force-displacement curve from the cantilever beam test with MAT261 and
Shell ELFORM 16 when using version R7.1.2 SMP (green) and version R8.0.0 MPP (orange).

To compare the effect of different versions for the cohesive elements, the ENF test described in Section 4.3.4
is run using the two different versions in LS-DYNA. The force-displacement results are presented in Figure 4.26
and here the difference is visible as well. The simulations are run using solid elements (ELFORM 3) and an
elastic material model (MAT1).

45
 0.16

0.14

0.12
Force [kN]
0.1

0.08

0.06

0.04
Analytical
0.02 SMP R7.1.2
MPP R8.0.0
0
0 5 10 15
Displacement [mm]
Figure 4.26: A comparison of the force-displacement curve from the ENF test when using version R7.1.2 SMP
(green) and version R8.0.0 MPP (orange).

4.5.1 Brief conclusion

From Figure 4.25 and Figure 4.26 it can be seen that for MAT261 in version R8.0.0. MPP seems to behave
more stable with regards to vibration, and for the ENF simulations the behaviour of the R8.0.0 MPP simulation
is closer to the analytical, desired behaviour. Therefore version R8.0.0 MPP is preferred over R7.1.2 SMP.

4.6 Hat profile model testing procedure

In this chapter, the process of implementing the obtained results into the model matching the experimental
setup described in Section 3 is presented. The model is built on a simulation model provided by Gestamp
HardTech.

4.6.1 Modelling the steel beam

In the original geometry, the steel hat profile beam was modelled with thin-shell elements. To be able to
simulate the delamination behaviour between the steel and the composite using a cohesive zone, the steel beam
is modelled with thick-shell elements. None of the cohesive elements in LS-DYNA can be used to connect
thin-shell with thick-shell or solid elements. A comparison between the original element formulation (thin-shell
ELFORM 2) and two different types of thick-shell elements, (ELFORM 1 and 5) was done. The setup of the
simulation was the same for all cases, with a difference in element type and the corresponding hourglass control.
A loading based on the one in the experiment was used, i.e. a constant velocity of the impactor.
The result is presented in Figure 4.27. It can be seen that thick-shell ELFORM 1 has the same buckling
behaviour as thin-shell ELFORM 16 while thick-shell ELFORM 5 does not buckle properly. The buckling
starts (or should start) at the displacement δ = 23mm for the pure steel beam, based on the results from the
thin-shell element simulations. The subsequent decrease of bending stiffness may be seen as a drop in the force
needed to impose the displacement at approximately 23 mm.

46
 18

16

14

12
Force [kN]

10

4
Thin-shell ELFORM 2
Thick-shell ELFORM 1
2
Thick-shell ELFORM 5
0
0 5 10 15 20 25 30
Displacement [mm]
Figure 4.27: The force-displacement relation of the steel beam for (blue) thin-shell ELFORM 2, (red) thick-shell
ELFORM 1, and (green) thick-shell ELFORM 5.

The buckling at a displacement of 30 mm, i.e when the buckling is fully developed, is presented in Figure 4.28.
It can be seen that the geometry modelled with thick-shells ELFORM 1 gives a similar buckling response as the
original modelling method using thin-shell ELFORM 2. The thick-shells ELFORM 5 on the other hand exhibit
a different behaviour. The buckling behaviour of the thin-shell and thick-shell ELFORM 1 approaches are
considered physically plausible, while the behaviour of the thick-shell ELFORM 5 simulation is not. Therefore
the steel beam in the full model is modelled using thick-shell elements ELFORM 1.

47
 (a) (b)

(c) (d)

(e) (f)
Figure 4.28: The buckling behaviour using (a)-(b) the original modelling method (thin-shell elements ELFORM
2), (c)-(d) thick-shell elements ELFORM 1 and (e)-(f ) thick-shell elements ELFORM 5. The behaviour is
captured at a applied displacement of 30 mm.

4.6.2 Modelling the composite laminate

In the middle of the composite plate there are four adjacent plies with the same fibre orientation. Since
delamination does not occur often in the interface between same angled layers [2], and since the computational
cost is important, these four plies are modelled together using one thick ply. By doing this 6 element layers are
removed, 3 composite plies and 3 cohesive plies, which saves computational effort since both composite and
cohesive material models are computationally costly, as has been presented with previous testing.

4.6.3 Benchmarking the simulation model

From the small scale tests it is suggested that the steel should be modelled using thick-shell ELFORM 1, the
composite material using thick-shell ELFORM 1 combined with MAT262, for material card see Appendix A.
The interface between steel and composite and lamina interface should be modelled using cohesive elements
with a size of 1 × 1mm. It was also discovered that for LS-DYNA version R8.0.0. cohesive elements are not
eroded automatically when its neighbouring composite elements are removed. This leads to a heavily distorted
cohesive element, and is not considered as a physical behaviour. This problem is fixed in LS-DYNA version
R9.0.0. Cohesive elements are eroded automatically if the ICOHED flag in the CONTROL SOLID card is used.
Therefore the hat profile simulations use R9.0.0 MPP with the ICOHED flag instead of R8.0.0 MPP. It was
also discovered that the ADD EROSION card gave out-of-bound forces, and therefore the erosion card is left
out of the simulations. During the search for errors regarding the cohesive zone, the material model of the
cohesive elements was changed into the more simple material model MAT138.

48
 The resulting force-displacement for the initial settings with the modifications described above (i.e. no
ADD EROSION, a mesh size of 1 × 1mm in the composite and cohesive elements, and LS-DYNA version
R9.0.0 MPP) is presented in Figure 4.29 together with the experimental results. In Figure 4.30 the deformation
of beam A from three views is shown. It can be seen that the composite exhibits delamination throughout
the entire composite plate, resulting in the bottom part of the plate falling off. The same views for beam E
is shown in Figure 4.31. For beam E, the composite has not delaminated throughout the entire plate, and
therefore the composite is still attached at the edges but has detached at the buckled part of the steel. Note
that for both beams there is still one ply attached to the steel. For beam A the ply has a fibre angle of 0◦ and
for beam E the fibre angle of the ply is 90◦ .

14

12

10
Force [kN]

4
Experiment A
Experiment E
2 Simulation A
Simulation E
0
0 5 10 15 20 25 30
Displacement [mm]
Figure 4.29: The force displacement relations for (red) test beam A, (blue) test beam E, (green) simulation
beam A, and (orange) simulation beam E.

49
 (a)

(b)

(c)

Figure 4.30: Test beam A simulation results from (a) the bottom, (b) from the side with the beam cut through
the middle of the width in order to see the delamination of the beam, and (c) the cross section at the final time
step of the simulation. The plies are colour coordinated according to; 0◦ - pink, 45◦ - orange, −45◦ - green, 90◦
- blue, and cohesive elements are yellow.

50
 (a)

(b)

(c)
Figure 4.31: Test beam E simulation results from (a) the bottom, (b) from the side with the beam cut through
the middle of the width in order to see the delamination of the beam, and (c) the cross section at the final time
step of the simulation. The plies are colour coordinated according to; 0◦ - pink, 45◦ - orange, −45◦ - green, 90◦
- blue, and cohesive elements are yellow.

51
 In Figure 4.29 it can be seen that the simulations give stiffer results than both experimental results. This
indicates that either some material in the simulation is stiffer than the actual case, or that some dimension is
larger in the model than in reality. In the gluing process the composite together with the steel was heated
to 175◦ C and this could have had an effect on the properties of steel, which could be the reason for the
softer experimental results. When comparing the delamination behaviour for the simulations and experiments,
which can be seen in the graph as a sudden drop in the force, it can be seen that beam A delaminates at an
earlier stage in the simulations compared to the experiment. For beam E, the simulation and the experiment
delaminates at the same displacement, but not the same force. This however could be described by the overly
stiff response of the steel in the simulation.
In order to see this effect, the simulation results are compared to the response of a pure steel beam in
Figure 4.32. In the figure it can be seen that initially both beams with composite behave stiffer, but after
delamination all three beams exhibit the same stiffness until buckling. The point of buckling is different for all
three beams.

16

14

12
Force [kN]

10

4
Only steel
2 Simulation A
Simulation E
0
0 5 10 15 20 25 30
Displacement [mm]
Figure 4.32: The force displacement relations for (purple) the pure steel beam, (green) simulation of beam A,
and (orange) simulation of beam E.

4.6.3.1 Yield property study

Since the steel behaves too stiff, the simulations are performed again with lowered yield properties of the steel.
The values in the yield curve presented in Section 3 are lowered by 20% in order to compensate for the overly
stiff steel. The result is presented in Figure 4.33.
In Figure 4.33 it can be seen that lowering the steel properties alters the buckling behaviour, but does not
move the point of delamination much with respect to the displacement. This indicates that the stiffness of
the steel caused the difference in point of buckling for the original setup, and with some tuning of the yield
parameters the correct point of buckling could be obtained. This also indicates that the original assumption of
the overly stiff steel in simulations was correct.

52
 16

14

12
Force [kN]

10

4 Experiment A
Experiment E
2 Simulation A (modified)
Simulation E (modified)
0
0 5 10 15 20 25 30
Displacement [mm]
Figure 4.33: The force displacement relations for (red) test beam A, (blue) test beam E, (green) simulation
beam A with lowered yield parameters, and (orange) simulation beam E with lowered yield parameters.

53
5 Recommended model configuration
The current section describes the modelling procedure in LS-DYNA that is found best to predict the behaviour
of a steel-composite structure in a crash scenario. The described modelling method is formulated with focus
on capturing the delamination between the steel and the composite laminate as well as between the plies in
the composite laminate. When deciding on for example element type, material model and adhesive modelling
method, both the behaviour of the model and the computational cost were taken into account.
In order to capture the delamination behaviour within the composite laminate, each ply of the laminate
requires to be modelled separately. The plies are then connected by an adhesive model.

5.1 Element type

Both the composite plies and the steel should be modelled using thick-shell elements. The recommended
element formulation is ELFORM 1, supplemented with hourglass control. For the simulations in this project,
hourglass control 1 with the default coefficient (0.1) is used. Since no investigation of the different hourglass
control methods has been performed, no specific recommendation can be provided.
If there are two or more adjacent plies with the same fibre angle, these can be modelled as one thick ply.
This in order to reduce the computational cost. If this is done, it is important to keep the aspect ratio of the
elements in mind. If the aspect ratio gets poor when combining many plies into one element, it is possible to use
more layers of elements and merge the nodes instead of using a cohesive element between. This will also reduce
the computational cost. Note that this is not applicable for composites where there are very many plies with
the same orientation stacked together. Since this not has been studied specifically, no exact recommendation as
to how many plies can be joined together without losing delamination information can be given.

5.2 Adhesive model

In Section 4.3, a comparison between different adhesive modelling methods is presented. Based on this
comparison, the cohesive zone is recommended for capturing delamination of both mode I and mode II.
By using a cohesive zone, cohesive elements are implemented between the layers with the nodes merged
to the adjacent plies. A zero thickness is allowed and used in the tested model. No study of the thickness of
the cohesive zone is performed. Since the adhesive is implemented between plies modelled using thick-shells,
element formulation 19 is to be used for the cohesive zone.
The cohesive elements are mesh dependent and the recommendations of the discretisation are described in
Section 5.4.

5.3 Material models

Based on the tests described in Section 4, it can be seen that material model MAT262 has a good behaviour and
comes at a computational cost that, compared to the other composite material models, is reasonable. Therefore,
if there is enough material data provided, this material model is recommended. Note that no calibration of the
unphysical parameters included in the more simple material models was performed. If not enough material data
is provided to implement MAT262, an additional study may be needed. In order to capture the maximum and
minimum strains in the ADD EROSION card should be added if it can be implemented without causing errors.
For the cohesive zone, the material model MAT186 was used in the pre-study. This material card requires
the critical energy release rates (GIC and GIIC ), the ultimate strengths (mode I and mode II) and a traction-
separation curve. Due to a search of an error, the material model was replaced by MAT138 when benchmarking
the full hat profile beam model. When implementing MAT138, the Young’s moduli (Et and En ), energy release
rates (GIC and GIIC ), the ultimate strengths (mode I and mode II), and the ultimate displacements are
possible inputs. Peak traction is not necessary to define and can be calculated using the other inputs, since the
traction-separation relation is bilinear for this material model. However, the error was not due to the cohesive
material model. Nevertheless, due to the fine response of the full model, MAT138 was kept. Both cohesive
material models MAT138 and MAT186 are therefore considered usable.

54
5.4 Mesh discretisation
The recommended adhesive modelling method, the cohesive zone, is mesh dependent. The cohesive elements
require a relatively fine mesh to provide the wanted behaviour. For the simulations in Section 4.3 an element
size of approximately 1 × 1 mm is used. If the mesh is too coarse the delamination behaviour will not be
captured as intended.

5.5 Contact conditions

A general single surface contact is included to handle the contact between the beam and supports, as well as
the contact between the composite plies after delamination. In order to capture the contact force between the
rigid impactor and the steel beam, FORCE TRANSDUCER PENALTY is implemented between the two parts.
For the contacts, BSORT is set to 10, so that the contact is checked every tenth time step. This in order to not
let the contact pass through one of the thin composite elements unnoticed.

5.6 Important settings

It is important to use version R9.0.0 of LS-DYNA, otherwise the cohesive elements can cause problems if
its neighbouring elements are eroded. This problem is solved by setting the parameter ICOHED to 1 in the
CONTROL SOLID card.
In order to allow for initial penetration, the IGNORE parameter in CONTROL CONTACT need to be set
to 1 or 2. If this is not done, an initial penetration can lead to an inaccurate behaviour in the beginning of the
simulation.
If a contact condition is used to model the interface, it is important to use OPTION 4 in the contact card.

55
6 Concluding remarks
The obtained results suggests that the delamination may be captured in a steel-composite structure. The
computational cost is high due to many factors. The composite material models are generally computational
expensive. This is combined with the large amount of elements, both due to modelling each ply with an element
through thickness and due the mesh refinements needed for the cohesive zone to behave as intended. The
recommended model configuration is formulated to keep the computational cost low while still getting the
wanted behaviour.
From the procedure it can be concluded that it is of importance to be observant of what material models are
compatible with what element types. It is also concluded that for modelling delamination where the composite
is at risk of failing prior to the cohesive elements, the LS-DYNA version R9.0.0 should be used. This with the
ICOHED flag activated.
Additional studies are needed in order to possess a simulation model that is reliable enough to be used in
the product development process at Gestamp HardTech. However, this study is a great initial step towards a
simulation model that may predict the failure behaviour of steel-composite components.

56
7 Future work
In order to ensure the level of prediction in the model, more types of experiments are needed. This is to be
carried out at Gestamp HardTech. The maximum and minimum strains need to be included in order to give a
better correlation with the experiments with respect to the failure. A more thorough investigation of which
adjacent fibre angles are possible to model together, either using merged nodes or one thicker ply, and still
maintain a satisfactory level of prediction in the analysis is required. This is needed if laminates with more
plies are to be modelled at a reasonable computational cost.

57
References
[1] LS-DYNA Keyword User’s Manual, Volume I. 971st ed. California: Livermore Software Technology
Corporation, 2007. isbn: 0977854027.
[2] B. D. Agarwal, L. J. Broutman, and K. Chandrashekhara. Analysis and performance of fiber composites.
3rd ed. Hoboken, N.J: John Wiley & Sons, 2006. isbn: 0471268917;9780471268918.
[3] L. Iannucci et al. Failure Models and Criteria for FRP Under In-Plane or Three-Dimensional Stress
States Including Shear Non-Linearity. Tech. rep. 2005.
[4] P. Robinson, E. Greenhalgh, and S. Pihno. Failure mechanisms in polymer matrix composites. Cambridge:
Woodhead Publishing Limited, 2012. isbn: 9781845697501.
[5] J.-K. Kim and M.-L. Sham. Impact and delamination failure of woven-fabric composites. English.
Composites Science and Technology 60.5 (2000), 745–761.
[6] S. Lomov and Knovel. Non-Crimp Fabric Composites: Manufacturing, Properties and Applications.
English. Sawston;LaVergne; Woodhead Publishing Limited, 2011. isbn: 1845697626;9781845697624;
[7] F. Edgren, L. E. Asp, and R. Joffe. Failure of NCF composites subjected to combined compression and
shear loading. Composites Science and Technology 66.15 (2006), 2865–2877.
[8] S. Hartmann, T. Klöppel, and C. Liebold. Introduction to Composite Material Modeling with LS-DYNA.
Dynamore GmbH, 2014.
[9] N. Saabye Ottosen and H. Petersson. Introduction to the finite element method. Prentice Hall, 1992.
[10] J. Hallquist. LS-DYNA theory manual. California: Livermore Software Technology Corporation, 2006.
isbn: 0977854000.
[11] E. Qiuli Sun. Shear Locking and Hourglassing in MSC Nastran, ABAQUS, and ANSYS. Tech. rep. 2006.
[12] T. Bru et al. Characterisation of the mechanical and fracture properties of a uni-weave carbon fibre/epoxy
non-crimp fabric composite. Data in Brief 6 (2016), 680–695.
[13] Sika R
. SikaPower R
-MBX Class I. Product Data Sheet Version 1, 2014.
[14] B. Alfredsson. Handbok och formelsamling i Hållfasthetslära. 11th ed. Instutitionen för hållfasthetslära
KTH, 2014.
[15] M. Fagerström. Associate Professor at Chalmers University of Technology, Department of Applied
Mechanics, Division of Material and Computational Mechanics, 2016.
[16] T. Erhart. Review of Solid Element Formulations in LS-DYNA. Tech. rep. 2011.
[17] U. Stelzmann. Die große Elementbibliothek in LS-DYNA - Wann nimmt man was? Tech. rep. 2010.
[18] LS-DYNA keyword user’s manual, Volume II, Material Models. California: Livermore Software Technology
Corporation, 2016.
[19] K. Björk. Formler och Tabeller för Mekanisk Konstruktion - MEKANIK och HÅLLFASTHETSLÄRA.
5th ed. Karl Björks Förlag HB, 2003.
[20] S. Bala. Tie-Break Contacts in LS-DYNA. Tech. rep. 2007.
[21] Cohesive element formulation. http://www.dynasupport.com/howtos/element/cohesive-element-
formulation. Accessed: 2016-06-20.
[22] T. Graf, A. Haufe, and F. Andrade. Adhesives modeling with LS-DYNA: Recent developments and future
work. Tech. rep. 2014.
[23] LS-DYNA keyword user’s manual, Volume II, Material Models. California: Livermore Software Technology
Corporation, 2015.
[24] A. Turon et al. An engineering solution for mesh size effects in the simulation of delamination using
cohesive zone models. Tech. rep. 2006.
[25] Y. Mi et al. Progressive Delamination Using Interface Elements. Journal of Composite Materials 32.14
(1998), 1246–1272.
[26] Z.Hashin. Failure Criteria for Unidirectional Fiber Composites. Journal of Applied Mechanics 47 (1980),
329–334.

58
Appendices

59
A Material cards
A.1 MAT22

Table A.1: Material card for MAT22

Card 1 MID RO EA EB EC PRBA PRCA PRCB
1500 kg/m3 132 GPa 9 GPa 9 GPa 0.029 0.02 0.43
Card 2 GAB GBC GCA (KFAIL) AOPT MACF
4.3 GPa 3.5 GPa 4.3 GPa 2 1
Card 3 XP YP ZP A1 A2 A3
0 1 0
Card 4 V1 V2 V3 D1 D2 D3 BETA
1 0 0
Card 5 SC XT YT YC ALPH SN SYZ SZX
77.8 MPa 1787 MPa 29.2 MPa 130 MPa 0 29.2 MPa 56.7 MPa 77.8 MPa

Table A.2: Description of abbreviations in Material model MAT22.

Name Description
Card 1
MID Material ID (for the current model)
RO Density (ρ)
EA Young’s modulus - longitudinal direction (Ea )
EB Young’s modulus - transverse direction (Eb )
EC Young’s modulus - normal direction (Ec )
PRBA Poisson’s ratio - ba (νba )
PRCA Poisson’s ratio - ca (νca )
PRCB Poisson’s ratio - cb (νcb )
Card 2
GAB Shear modulus - ab (Gab )
GBC Shear modulus - bc (Gbc )
GCA Shear modulus - ca (Gca )
KFAIL Bulk modulus of failed material (K), necessary for compressive failure.
AOPT Material axes option
MACF Material axes flag, change axes for brick elements
Card 3
[XP YP ZP] Defines coordinate of point p for AOPT=1 and 4
[A1 A2 A3] Defines component vector a for AOPT=2
MANGLE Material angle (degrees) for AOPT=0 and 3
Card 4
[V1 V2 V3] Defines component vector v for AOPT=3 and 4
[D1 D2 D3] Defines component vector d for AOPT=2
BETA Material angle for AOPT=0 and 3 (degrees). Can be overrided by the element card.
Card 5

I
 SC Shear strength - ab-plane (Sc )
XT Tensile strength -longitudinal (Xt )
YT Tensile strength - transverse (Yt )
YC Compressive strength- transverse (Yc )
ALPH Shear stress for the non-linear term
SN Normal tensile strength - solid elements
SYZ Transverse shear strength - solid elements
SZX Transverse shear strength - solid elements

A.2 MAT54/55

Table A.3: Material card for MAT54/55

Card 1 MID RO EA EB EC PRBA PRCA PRCB
1500 kg/m3 132 GPa 9 GPa 9 GPa 0.029 0.02 0.43
Card 2 GAB GBC GCA (KF) AOPT
4.3 GPa 3.5 GPa 4.3 GPa 2
Card 3 A1 A2 A3 MANGLE
0 1 0
Card 4 V1 V2 V3 D1 D2 D3 DFAILM DFAILS
1 0 0
Card 5 TFAIL ALPH SOFT FBRT YCFAC DFAILT DFAILT EFS
-1 1000 GPa 1 0 2 0.0123 -0-0049
Card 6 XC XT YC YT SC CRIT BETA
630 MPa 1787 MPa 130 MPa 29.2 MPa 77.8 MPa 55 1
Card 7 PFL EPSF EPSR TSMD SOFT2

Card 8 SLIMT1 SLIMC1 SLIMT2 SLIMC2 SLIMS NCYRED SOFTG

1
Card 9 LCXC LCXT LCYC LCYT LCSC DT

Table A.4: Description of abbreviations in Material model MAT54/55.

Name Description
Card 1
MID Material ID (for the current model)
RO Density (ρ)
EA Young’s modulus - longitudinal direction (Ea )
EB Young’s modulus - transverse direction (Eb )
EC Young’s modulus - normal direction (Ec )
PRBA Poisson’s ratio - ba (νba )
PRCA Poisson’s ratio - ca (νca )
PRCB Poisson’s ratio - cb (νcb )
Card 2
GAB Shear modulus - ab (Gab )
GBC Shear modulus - bc (Gbc )
GCA Shear modulus - ca (Gca )
(KF) Bulk modulus of failed material (not used) (K)
AOPT Material axes option
Card 3

II
[A1 A2 A3] Defines component vector a for AOPT=2
MANGLE Material angle (degrees) for AOPT=0 and 3
Card 4
[V1 V2 V3] Defines component vector v for AOPT=3
[D1 D2 D3] Defines component vector d for AOPT=2
DFAILM Maximum strain for matrix in tension and compression (only MAT54)
DFAILS Maximum tensorial shear strain (only MAT54)
Card 5
TFAIL Time step criteria for element deletion
ALPH Scear stress parameter for the non-linear term
SOFT Softening reduction factor for material strength in crashfront
FBRT Softening for fibre tensile strength
YCFAC Reduction factor for compressive strength after matrix failure (only MAT54)
DFAILT Maximum strain for fibre tension (only MAT54)
DFAILT Maximum strain for fibre compression (only MAT54)
EFS Effective failure strain (only MAT54)
Card 6
XC Compressive strength - longitudinal
XT Tensile strength -longitudinal
YC Compressive strength- transverse
YT Tensile strength - transverse
SC Shear strength - ab - plane
CRIT Failure criteria (54 or 55)
BETA Weight factor for shear term (only MAT54)
Card 7 Only MAT54
PFL Percentage of layers that must fail before crashfront is initiated
EPSF Damage initiation transverse shear strain
EPSR Final rupture transverse shear strain
TSMD Maximum damage, transverse shear
SOFT2 Softening reduction factor for material strength in crashfront elements
”orthogonal”
Card 8 Only MAT54
SLIMT1 Factor to determine the minimum stress limit after stress maximum - fibre tension
SLIMC1 Factor to determine the minimum stress limit after stress maximum - fibre compression
SLIMT2 Factor to determine the minimum stress limit after stress maximum - matrix tension
SLIMC2 Factor to determine the minimum stress limit after stress maximum - matrix compres-
sion
SLIMS Factor to determine the minimum stress limit after stress maximum - shear
NCYRED Number of cycles for stress reduction from maximum to minimum
SOFTG Softening reduction factor for shear stiffness in crashfront elements - transverse
Card 9
LCXC Load curve ID for XC - strain rate. Overrides XC in Card 6
LCXT Load curve ID for XT - strain rate. Overrides XT in Card 6
LCYC Load curve ID for YC - strain rate. Overrides YC in Card 6
LCYT Load curve ID for YT - strain rate. Overrides YT in Card 6
LCSC Load curve ID for SC - strain rate. Overrides SC in Card 6
DT Strain rate averaging option

III
A.3 MAT58

Table A.5: Material card for MAT58

Card 1 MID RO EA EB (EC) PRBA TAU1 GAMMA
1500 kg/m3 132 GPa 9 GPa 9 GPa 0.029
Card 2 GAB GBC GCA SLIMT1 SLIMC1 SLIMT2 SLIMC2 SLIMS
4.3 GPa 3.5 GPa 4.3 GPa
Card 3 AOPT TSIZE ERODS SOFT FS EPSF EPSR TSMD
2 1
Card 4 XP YP ZP A1 A2 A3 PRCA PRCB
0 1 0 0.02 0.43
Card 5 V1 V2 V3 D1 D2 D3 BETA
1 0 0
Card 6 E11C E11T E22C E22T GMS
0.0049 0.0123 0.0171 0.0032 0.091
Card 7 XC XT YC YT SC
630 MPa 1787 MPa 130 MPa 29.2 MPa 77.8 MPa
Card 8 LCXC LCXT LCYC LCYT LCSC LCTAU LCGAM DT

Card 9 LCE11C LCE11T LCE22C LCE22T LCGMS LCEFS

Table A.6: Description of abbreviations in Material model MAT58

Name Description
Card 1
MID Material ID (for the current model)
RO Density
EA Young’s modulus - longitudinal direction
EB Young’s modulus - transverse direction
EC Young’s modulus - normal direction
PRBA Poisson’s ratio - ba
TAU1 Stress limit of the first slightly non-linear part of the shear stress-shear strain curve
GAMMA Stain limit of the first slightly non-linear part of the shear stress-engineering shear
strain curve
Card 2
GAB Shear modulus - ab
GBC Shear modulus - bc
GCA Shear modulus - ca
SLIMT1 Factor to determine minimum stress limit after stress maximum - fibre tension
SLIMC1 Factor to determine minimum stress limit after stress maximum - fibre compression
SLIMT2 Factor to determine minimum stress limit after stress maximum - matrix tension
SLIMC2 Factor to determine minimum stress limit after stress maximum - matrix compression
SLIMS Factor to determine minimum stress limit after stress maximum - shear
Card 3

IV
AOPT Material axes option
TSIZE Time step for automatic element deletion
ERODS Maximum effective strain for element layer failure
FS Failure surface type
EPSF Damage initiation - transverse shear strain
EPSR Final rupture - transverse shear strain
TSMD Maximum damage - transverse shear
Card 4
[XP YP ZP] Defines coordinates of point p for AOPT=1
[A1 A2 A3] Defines component vector a for AOPT=2
PRCA Poisson’s ratio - ca
PRCB Poisson’s ratio - cb
Card 5
[V1 V2 V3] Defines component vector v for AOPT=3
[D1 D2 D3] Defines component vector d for AOPT=2
BETA Material angle for AOPT = 0 and 3 (◦ )
Card 6
E11C Strain at longitudinal compressive strength
E11T Strain at longitudinal tensile strength
E22C Strain at transversal compressive strength
E22T Strain at transversal tensile strength
GMS Engineering shear stain at shear strength
Card 7
XC Compressive strength - longitudinal
XT Tensile strength -longitudinal
YC Compressive strength- transverse
YT Tensile strength - transverse
SC Shear strength - ab - plane
Card 8
LCXC Load curve ID for XC - strain rate. Overrides XC in Card 7
LCXT Load curve ID for XT - strain rate. Overrides XT in Card 7
LCYC Load curve ID for YC - strain rate. Overrides YC in Card 7
LCYT Load curve ID for YT - strain rate. Overrides YT in Card 7
LCSC Load curve ID for SC - strain rate. Overrides SC in Card 7
LCTAU Load curve ID for TAU1 - strain rate. Overrides TAU1 in Card 1. Only used for
FS=-1
LCGAM Load curve ID for GAMMA - strain rate. Overrides GAMMA in Card 1. Only used
for FS=-1
DT Strain rate averaging option
Card 9
LCE11C Load curve ID for E11C - strain rate. Overrides E11C in Card 6
LCE11T Load curve ID for E11T - strain rate. Overrides E11T in Card 6
LCE22C Load curve ID for E22C - strain rate. Overrides E22C in Card 6
LCE22T Load curve ID for E22T - strain rate. Overrides E22T in Card 7
LCGMS Load curve ID for GMS - strain rate. Overrides GMS in Card 6
LCEFS Load curve ID for ERODS - strain rate. Overrides ERODS in Card 3

V
A.4 MAT261

Table A.7: Material cards for MAT261

Card 1 MID RO EA EB (EC) PRBA PRCA PRCB
1500 kg/m3 132 GPa 9 GPa 9 GPa 0.029 0.02 0.43
Card 2 GAB GBA GCA AOPT DAF DKF DMF EFS
4.3 GPa 3.5 GPa 4.3 GPa 2 0 0 0 0
Card 3 XP YP ZP A1 A2 A3
0 1 0
Card 4 V1 V2 V3 D1 D2 D3 MANGLE
1 0 0
Card 5 ENKINK ENA ENB ENT ENL
103.1 kJ/m2 67.1 kJ/m2 51.8 kJ/m2 33.7 kJ/m2 33.7 kJ/m2
Card 6 XC XT YC YT SL
630 MPa 1787 MPa 130 MPa 29.2 MPa 77.8 MPa
Card 7 FIO SIGY LCSS BETA FPL PUCK SOFT
62 ◦ 1000 GPa 0 0 0 0 1

Table A.8: Description of abbreviations in Material model 261

Name Description
Card 1
MID Material ID (for the current model)
RO Density
EA Young’s modulus - longitudinal direction
EB Young’s modulus - transverse direction
EC Young’s modulus - normal direction
PRBA Poisson’s ratio - ba
PRCA Poisson’s ratio - ca
PRCB Poisson’s ratio - cb
Card 2
GAB Shear modulus - ab
GBA Shear modulus - bc
GCA Shear modulus - ca
AOPT Material axes option
DAF Flag to control failure of integration point - longitudinal tensile fibre failure
DKF Flag to control failure of integration point - longitudinal compressive fibre failure
DMF Flag to control failure of integration point - transverse matrix failure
Card 3
[XP YP ZP] Defines component vector p for AOPT=1 & 4
[A1 A2 A3] Defines component vector a for AOPT=2
Card 4
[V1 V2 V3] Defines component vector v for AOPT=3
[D1 D2 D3] Defines component vector d for AOPT=2
MANGLE Material angle (degrees) for AOPT=0& 3
Card 5
ENKINK Fracture toughness for longitudinal compressive fibre failure
ENA Fracture toughness for longitudinal tensile fibre failure
ENB Fracture toughness for interlaminar tensile matrix failure
ENT Fracture toughness for interlaminar transverse shear matrix failure
ENL Fracture toughness for interlaminar longitudinal shear matrix failure
Card 6

VI
XC Compressive strength - longitudinal
XT Tensile strength -longitudinal
YC Compressive strength- transverse
YT Tensile strength - transverse
SL Longitudinal shear strength
Card 7
FIO Fracture angle in pure transverse compression (◦ )
SIGY In-plane shear yield stress
LCSS Load curve or Table ID
BETA Hardening parameter for in-plane shear plasticity
PFL Percentage of elements that must fail before crashfront is initiated
PUCK Flag for evaluation and postprocessing of the Puck criterion
SOFT Softening reduction factor for material strength in the crashfront

VII
A.5 MAT262

Table A.9: Material cards for MAT262

Card 1 MID RO EA EB (EC) PRBA PRCA PRCB
1500 kg/m3 132 GPa 9 GPa 9 GPa 0.029 0.02 0.43
Card 2 GAB GBA GCA AOPT DAF DKF DMF EFS
4.3 GPa 3.5 GPa 4.3 GPa 2 0 0 0 0
Card 3 XP YP ZP A1 A2 A3
0 1 0
Card 4 V1 V2 V3 D1 D2 D3 MANGLE
1 0 0
Card 5 GXC GXT GYC GYT GLS GXCO GXTO
103.1 kJ/m2 67.1 kJ/m2 51.8 kJ/m2 33.7 kJ/m2 33.7 kJ/m2
Card 6 XC XT YC YT SL XCO XTO
630 MPa 1787 MPa 130 MPa 29.2 MPa 77.8 MPa
Card 7 FIO SIGY ETAN BETA FPL PUCK SOFT
62 ◦ 1000 GPa 0 0 0 0 1

Table A.10: Description of abbreviations in Material model 261

Name Description
Card 1
MID Material ID (for the current model)
RO Density
EA Young’s modulus - longitudinal direction
EB Young’s modulus - transverse direction
EC Young’s modulus - normal direction
PRBA Poisson’s ratio - ba
PRCA Poisson’s ratio - ca
PRCB Poisson’s ratio - cb
Card 2
GAB Shear modulus - ab
GBA Shear modulus - bc
GCA Shear modulus - ca
AOPT Material axes option
DAF Flag to control failure of integration point - longitudinal tensile fibre failure
DKF Flag to control failure of integration point - longitudinal compressive fibre failure
DMF Flag to control failure of integration point - transverse matrix failure
Card 3
[XP YP ZP] Defines component vector p for AOPT=1 & 4
[A1 A2 A3] Defines component vector a for AOPT=2
Card 4
[V1 V2 V3] Defines component vector v for AOPT=3
[D1 D2 D3] Defines component vector d for AOPT=2
MANGLE Material angle (degrees) for AOPT=0& 3
Card 5
GXC Fracture toughness for longitudinal compressive fibre failure
GXT Fracture toughness for longitudinal tensile fibre failure
GYC Fracture toughness for transverse compressive failure
GYT Fracture toughness for transverse tensile failure
GSL Fracture toughness for in-plane shear failure

VIII
GXCO Fracture toughness for longitudinal compression fibre failure - bi-linear damage evolu-
tion
GXTO Fracture toughness for longitudinal tension fibre failure - bi-linear damage evolution
Card 6
XC Compressive strength - longitudinal
XT Tensile strength -longitudinal
YC Compressive strength- transverse
YT Tensile strength - transverse
SL Longitudinal shear strength
XCO Longitudinal compressive strength at inflection point
XTO Longitudinal tensile strength at inflection point
Card 7
FIO Fracture angle in pure transverse compression (◦ )
SIGY In-plane shear yield stress
LCSS Load curve or Table ID
BETA Hardening parameter for in-plane shear plasticity
PFL PErcentage of elements that must fail before crashfront is initiated
PUCK Flag for evaluation and postprocessing of the Puck criterion
SOFT Softening reduction factor for material strength in the crashfront

IX
B Tiebreak card
B.1 AUTOMATIC SURFACE TO SURFACE TIEBREAK

Table B.1: Contact card for AUTOMATIC SURFACE TO SURFACE TIEBREAK

Card 1 SSID MSID SSTYP MSTYP SBOXID MBOXID SPR MPR
3 3 0 0
Card 2 FS FD DC VC VCD PENCHK BT DT
20 0
Card 3 SFS SFM SST MST SFST SFMT FSF VSF

Card 4 OPTION NFLS SFLS PARAM ERATEN ERATES CT2CN CN

4 29.2 MPa 77.8 MPa 1
Card A SOFT SOFTCL LCIDAB MAXPAR SBOPT DEPTH/LCID DEPTH BSORT/LCID
0 0.1 1.025 0 DEPTH 2 BSORT
A, cont BSORT FRCFRQ
10 0

X
C Failure criteria
All information in this section is from the LS-DYNA keyword manual [18]. Variables for all criteria:

σaa Stress from FE analysis - aa σab Stress from FE analysis - ab

σbb Stress from FE analysis - bb
Ea Young’s modulus - longitudinal Eb Young’s modulus - transverse
Ec Young’s modulus - normal νba Poisson’s ratio ba
νca Poisson’s ratio ca νcb Poisson’s ratio cb
Gab Shear modulus ab Gbc Shear modulus bc
Xc Compressive strength - longitudinal Xt Tensile strength - longitudinal
Yc Compressive strength - transverse Yt Tensile strength - transverse
Sc Shear strength Sl Longitudinal shear strength
σ̃ Effective stress tensor φ Fracture plane for general loading
φo Fracture plane for pure compression Ψ Angle to principal stress
Table C.1: Variables used in failure criteria calculations

C.1 Chang-Chang
Tensile fibre mode:
2 (
≥0
  
σaa σab failed
σaa > 0 then e2f = +β −1 (C.1)
Xt Sc <0 elastic

Ea = Eb = Gab = νba = νab = 0

Compressive fibre mode:
2 (
≥0

σaa failed
σaa < 0 then e2c = −1 (C.2)
Xc <0 elastic

Ea = νba = νab = 0
Tensile matrix mode:
2 2 (
≥0
 
σbb σab failed
σbb > 0 then e2m = + −1 (C.3)
Yt Sc <0 elastic

Ea = νba = 0 =⇒ Gab = 0
Compressive matrix mode:
2 " 2 # 2 (
≥0
 
σbb Yc σbb σab failed
σbb < 0 then e2d = + −1 −1 (C.4)
2Sc 2Sc Yc Sc <0 elastic

Eb = νba = νab = 0 =⇒ Gab = 0

Xc = 2Yc for 50% fibre volume

C.2 Tsai-Wu
The tensile and compressive fibre failure criteria for Tsai-Wu are the same as for Chang-Chang, see equations
(C.1) and (C.2). The same equation is used for both tensile and compressive matrix failure, see (C.5).
2 (
2
(Yc − Yt ) σbb ≥ 0 failed

2 σbb σab
emd = + −1 (C.5)
Yc Yt Sc Yc Yt < 0 elastic

XI
C.3 Modified Hashin
[26]
Tensile fibre mode:
2 2 (
≥1
 
σaa τab failed
fa = {σaa > 0} = + (C.6)
Xt Sl <1 no failure

Compressive fibre mode:

2 (
≥1

σaa failed
fa = {σaa < 0} = (C.7)
Xc <1 no failure

Tensile matrix mode: 2 2 (

≥1
 
σbb τab failed
fa = + (C.8)
Yt Sl <1 no failure
Compressive matrix mode:
2 2 ! 2 (
≥1
  
σbb Yc σbb τbb failed
fa = + −1 + (C.9)
2Sc 2Sc Yt Sl <1 no failure

C.4 Pinho
Tensile fibre mode: (
σa ≥1 failed
fa = (C.10)
Xt <1 no failure
Compressive fibre mode (kinking model):
 2  2
τt τl
+ Sl −µ =1 if σbm ≥ 0


St −µt σn σ n
fkink =  2  2  l  2 (C.11)
 σYn + Sτt + Sτl
 =1 if σbm > 0
t t l

with St = Yc
2tan(φo ) ,
1
µt = − tan(2φ o)
, µl = Sl Sµtt
σbm + σc Ψ σbm − σcΨ
σn = + cos(2φ) + τbm cΨ sin(2φ)
2 2

σbm − σcΨ
τt = − sin(2φ) + τbm cΨ cos(2φ)
2

τl = τam bm cos(2φ) + τcΨ am sin(φ)

Tensile matrix mode:
2 2 2 (
≥1
  
σn τt τl failed
if σ ≥ 0 fmat = + + (C.12)
Yt St Sl <1 no failure
with
σb + σc σb − σc
σn = + cos(2φ) + τbc sin(2φ)
2 2

σb − σc
τt = − sin(2φ) + τbc cos(2φ)
2

XII
 τl = τab cos(2φ) + τca sin(φ)
Compressive matrix mode:
2 2 (
≥1
 
τt τl failed
if σ < 0 fmat = + (C.13)
St − µt σn Sl − µl σn <1 no failure

C.5 Camanho
Tensile fibre mode: (
σ11 − ν12 σ22 ≥1 failed
φ1+ = (C.14)
Xt <1 no failure
Compressive fibre mode: (
m m
h|σ12 | + µl σ22 i ≥1 failed
φ1− = (C.15)
Sl <1 no failure
with

Sl cos(2φo )
µl = −
Yc cos2 (φo )
m
σ22 = σ11 sin2 (ϕc ) + σ22 cos2 (ϕc ) − 2|σ12 |sin(ϕc )cos(ϕc )
m
= (σ22 − σ11 ) sin(ϕc )cos(ϕc ) + |σ12 | cos2 (ϕc ) − sin2 (ϕc )


σ12

 r   
Sl Sl
1 − 1−4 X c
+ µl Xc 
ϕc = arctan    
Sl
2 X + µ
 
c
l

Tensile matrix mode (perpendicular to the laminate mid-plane):

 r  2  2
if σ22 ≥ 0 then (1 − g) σ22 + g σ22 + σ12

 (
Yt Yy Sl ≥1 failed
φ2+ = (C.16)
h|σ 12 | + µ σ
l 22 i <1 no failure
if σ22 < 0 then


Sl
Compressive/shear matrix mode:
s 2 2 (
≥1

τt τl failed
if σ22 < 0then φ2− = + (C.17)
St Sl <1 no failure

With
1
µt = −
tan(2φo )
 
cos(φo )
St =Yc cos(φo ) sin(φo ) +
tan(2φo )
−|σ12 |
 
θ =arctan
σ22 sin(φo )
τt = h−σ22 cos(φo ) [sin(φo ) − µt cos(φo )cos(θ)]i
τl = hcos(φo ) [|σ12 | + µl σ22 cos(φo )sin(θ)]i

XIII