Modelling and Simulation of Fibre Metal Laminates

Peter Linde1, Jürgen Pleitner1, Henk de Boer2, Clarice Carmone3

1. Airbus Deutschland GmbH, Kreetslag 10, DE-21129 Hamburg, Germany

2. ALE, Advanced Lightweight Engineering, Rotterdamseweg 145, 2628 AL, Delft,
The Netherlands
3. Bishop Aeronautical Engineers GmbH, Oesterleystraße 3, DE-22587 Hamburg, Germany

Abstract: In Aircraft construction, fibre metal laminates (FML) have obtained an increased use as
a skin material due to its beneficial characteristics in terms of fatigue and fire resistance. The
material, consisting of alternate layers of aluminium and glass fibre reinforced epoxy, is however
complicated to model numerically. This paper deals with the modelling of FML and is focused on
a detailed simulation of the inter rivet buckling behaviour in a stiffened fuselage shell. The skin
material consists of FML and is modelled with solid elements, layer by layer. Each of the
constituents are provided with their respective material model. For the glass fibre reinforced
epoxy the user subroutine UMAT is employed for description of the failure modes, such as matrix
cracking and fibre failure. The behaviour of the delamination between the metal layer and the
fibre reinforced epoxy is also described with a user subroutine UINTER, which is an optional
contact definition. This subroutine contains a failure criterion for the delamination. The rivets
between the stiffener and the skin of a fuselage are modelled by elastic solid elements with a
plastic material model, but without any failure criterion. This reflects the design, where no rivet
failure is allowed prior to buckling of the skin. A specially designed experimental test, which
captures the main characteristics of inter rivet buckling, is modelled and simulated. The
numerical results are compared to the experimental data. Finally, conclusions and
recommendations are given for future research.

Keywords: Advanced Composite, Aircraft, Crack, Delamination, Epoxy, Experimental

Verification, Failure Criterion, Failure Mode, Fuselage, Fiber Metal Laminate (FML), Glass
Fibre, Glare®, Inter Rivet Buckling, Layer, Non-linearity, Numerical Modelling, Prepreg, Skin,
Static Behaviour, Subroutine.

1. Introduction

In modern aircraft fuselage design advanced composite materials are increasingly utilised. A
special family of such materials are the fibre metal laminates or hybrid materials, consisting of

2004 ABAQUS Users’ Conference 421

alternating layers of aluminium and glass fibre reinforced epoxy. This material provides improved
fatigue characteristics, considerable fire resistance and provides improved damage behaviour.
This paper deals with numerical modelling of the material, focusing on the damage behaviour for
the different constituents. Particularly delamination between adjacent layers is modelled and
described. An application, the inter rivet buckling problem is presented and used with the
developed models. The paper is organized as follows:
first the modelling of fibre metal laminates is presented, containing a description of the damage
models. After that inter rivet buckling is introduced and a physical test is presented which is
modelled by finite elements. In this model the above mentioned damage models are implemented.
Subsequently, results from numerical simulations are presented and compared with test data.
Finally, a summary and conclusions are presented.

2. Modelling of Fibre Metal Laminate

2.1 General

Fibre metal laminates are increasingly used as skin material in modern aircraft design. FMLs
consist of aluminium sheets covering the front and the back and, between them, alternating fibre
layers of unidirectional glass fibre reinforced epoxy (prepreg), and aluminium. Typical total skin
thickness may vary between 1.6 mm up to around 3.2 mm, corresponding to a thickness of
monolithic aluminium skin.
In general, there are three different approaches that can be chosen to model fibre reinforced
composite metal laminates: the micro-level approach, in which individual fibre and fibre-matrix
interfaces are studied, the meso-level characterization (Vlot, 2001), in which individual plies are
modelled and, finally, the macro-level one, in which the effect of the complete homogenised
laminate is studied.
Since the aim of this work is to model the failure mechanism on a macroscopic (buckling) level
and a microscopic level (delamination growth), the approach applied in this research is the meso-
level one (Linde, 2002), (Puck, 2002). This choice also leads to an uniform approach for all FML
types possible on the one hand, and on the other hand, to a relatively limited number of elements
used. It implies that both the aluminium layer and the prepreg one are considered as homogeneous
materials, with given constitutive parameters. In particular, orthotropic material properties are
given to the prepreg.
Usually, the adhesive bonding between the aluminium and the prepreg surfaces is of excellent
quality. Indeed, the ultimate interface strength is much higher than the strength in the transition
zone between fibre rich and resin rich zones of the prepreg (Vlot, 2001). Therefore a delamination
crack can be expected in those areas. Moreover, experimental observations have shown that
delamination may grow in the resin rich zone between two plies. It follows that particular
attention must be paid in modelling the interface between mated plies. One method to accurately

422 2004 ABAQUS Users’ Conference

model that area is by means of an additional layer (Figure 1) made of continuum elements.
However, since
resin
Resin rich
rich layer interface
Interfacelayer
pl
ply

continuum
Continuum elements interface
Interfaceelements

Interaction
Interaction surfaces
surfaces
pl
ply

Figure 1. Modelling of FML’s resin-rich interface.

the layer is very thin, the number of elements needed would be very large and, as consequence,
this method is scarcely used in finite element analyses. Another method, which is frequently used,
is by lumping the behaviour of the layer in zero-thickness interface elements (Allix, 1992),
(Hashagen, 1998), (Schipperen, 2001). In these elements only the aspect ratio between the length
and the width of the element has to be taken into account, so the number of elements is defined by
the element size in the individual plies and not by the thickness of the resin-rich interface layer. A
disadvantage however, is that the FE meshes of two adjacent plies have to comply so that the
interface elements can be placed in between. This disadvantage can be removed by modelling the
delamination behaviour using an interaction law between two adjacent plies. This can be
accomplished with the ABAQUS FE program. In this method no elements have to be generated
between matching layers by the user. The program itself finds the displacements between two
adjacent points on the surfaces of the layers and these displacements can be evaluated by a user
defined interaction law.
The present work deals with the development of proper damage models for damages that can
occur in the Fibre Metal Laminate Glare®. Two main failure modes for FML are numerically
modelled: fibre/matrix failure and interlaminar delamination.

2.2 Damage modelling

Damage in fibre metal laminates can be distinguished between damage in the aluminium, damage
in the prepreg, and delamination, see Figure 2.

2004 ABAQUS Users’ Conference 423

In the aluminium layers, the damage typically consists of plasticity in compression, but it can also
consist of cracking under tension.
For damage in the prepreg, a distinction must be made between the epoxy matrix failure and the
fibre failure. In the first case the damage will typically occur under tension by cracking. In the

a) undamaged fibre metal laminate

b) matrix failure c) delamination

Figure 2. Modelling of damage in Fibre Metal Laminates.

second case, the fibre may fail under tensile loading. Some further forms of failure exist, which
involve both matrix and fibre; however since not relevant for this study they are not described
here. The matrix failure and the fibre failure are implemented in the models by means of the user
subroutine UMAT.
Between two layers, delamination may take place when a failure criterion has been exceeded. A
failure criterion is described below. The delamination between two layers is implemented by
means of the user subroutine UINTER. The failure criterion for delamination is described
subsequently.

424 2004 ABAQUS Users’ Conference

3. Fibre/matrix failure

Fibre or matrix failure only occurs inside a glass/epoxy layer of the FML. Matrix failure is a non-
fatal failure mode, since the layers can still carry an increasing load after matrix failure.
A laminate will show stress redistributions after first ply cracking and additional loading of the
laminate will still be possible (Alessandrini, 2003). Therefore, for a correct description of a
laminate structure also the effects of crack propagation have to be described. Crack propagation
influences the material properties of the cracking ply. For a finite element modelling on a meso-
level, the degradation of the properties at a ply level is needed. In papers dealing with numerical
modelling three different options are mentioned for the material degradation:
1. instantaneous unloading;
2. gradual unloading;
3. constant stress at ply failure.
The failure criterion used here is based on a strain-based continuum damage formulation with
different failure criteria for matrix and fibre failure. A gradual degradation of the material
properties is assumed. This gradual degradation is controlled by the individual fracture energies of
matrix and fibre, respectively.
The user-subroutine UMAT has been used for implementation of the damage model. For matrix
failure the following failure criterion is used

2
ε 22t ⎛ t (ε 22 t 2
) ⎞ ⎛ ε 22
t
⎞
f m = c (ε 22 ) + ⎜⎜ ε 22 − c ⎟⎟ε 22 + ⎜⎜ s ⎟⎟ (ε12 ) 2
2
,
ε 22 ⎝ ε 22 ⎠ ⎝ ε12 ⎠
where ε 22t ε 22c are the failure strains perpendicular to the fibre direction in tension and
and
compression, respectively. The failure strain for shear is ε 12 . Failure occurs when f m exceeds its
s

threshold value ε 22 .
t

In case of damage, a damage parameter is calculated as follows

ε 22t (− C t
22 ε 22 ( f m
t
− ε 22 ) / Gm ).
dm = 1 − e
fm
The glass/epoxy layer shows transverse isotropy. Therefore, the modulus stiffnesses read

EL (1 − ν TT
2
) E (1 − ν LTν TL ) E (1 − ν LTν TL )
C11 = ; C22 = T ; C33 = T
a a a

2004 ABAQUS Users’ Conference 425

 ET (ν LT − ν LTν TT ) E (ν − ν LTν TT ) E (ν − ν LTν TL )
C12 = ; C13 = T LT ; C23 = T TT
a a a
C33 = GTT ; C55 = GLT ; C44 = GLT ,
with

a = 1 − 2ν LTν TL − ν TT
2
− 2ν LTν TLν TT .
The failure criterion for fibre failure is given by

ε11t ⎛ t (ε11t ) 2 ⎞
ff = (ε ) 2
+ ⎜⎜ ε11 − c ⎟⎟ε11 .
ε11c 11 ⎝ ε11 ⎠

Here ε 11t ε11c are the failure strains in fibre direction in tension and compression, respectively.
and
Failure occurs when f f exceeds its threshold value ε 11 .

A second damage parameter for fibre damage is introduced

ε11t (− C t
11ε 11 (
t
f f − ε 11 )/Gf )
d f =1− e .
ff

The modulus matrix of the glass/epoxy layer will be reduced according to

⎡(1 − d f )C11 (1 − d f )(1 − d m )C12 (1 − d f )C13 0 0 0 ⎤

⎢ (1 − d m )C 22 (1 − d f )(1 − d m )C 23 0 0 0 ⎥
⎢ ⎥
⎢ C 33 0 0 0 ⎥
C=⎢ ⎥
(1 − d f )(1 − d m )C 44 0 0
⎢ ⎥
⎢ C 55 0⎥
⎢ sym ⎥
⎣ C 66 ⎦

The stresses are computed by

σ = C ⋅ε .

426 2004 ABAQUS Users’ Conference

4. Delamination

Interlaminar delamination might occur at any interface of the FML, i.e. at the interface between
aluminium and glass/epoxy or at the interface between two glass/epoxy layers. Experimental
observations have shown that a delamination may grow at the resin rich layer between two layers.
Its initiation, however, might be due to matrix failure inside the glass/epoxy layer.
An orthotropic delamination model, describing mixed mode delamination, is applied.
If an initial delamination exists this delamination may close under the applied load. To prevent the
two adjacent plies from penetrating, a simple contact model is utilised. In this contact model the
normal stiffness of the interface is given a penalty value if the relative normal displacement
between two adjacent plies is negative. The normal stiffness is zero if opening of the initial
delamination occurs. The shear stiffness of the initial delamination zone is always taken equal to
zero, so no friction is modelled.
The delamination model has been implemented in the ABAQUS FE program, using the surface-
to-surface contact option. In case of surface-to-surface contact, the FE meshes of adjacent plies do
not need to be identical. The contact algorithm of ABAQUS will determine which node of the so-
called master surface is in contact with a given node on the slave surface. The relative
displacement between these two nodes is given in a local coordinate system. The first component
of the relative displacement u1 corresponds to the normal direction of the master surface. The
other two components ( u 2 and u 3 ) are the two out-of-plane shear components. The user-
subroutine UINTER can be used to specify a dedicated relation between the relative displacement
and the corresponding traction forces. Hence, the user can define the interaction between the two
surfaces.
Failure in this model is related to the equivalent relative displacement, κ , defined by
2 2
⎛ u ft ⎞ 2 ⎛ u ft ⎞ 2
κ = u + ⎜⎜ 2
1
⎟ u2 + ⎜
⎟
⎟ u ,
⎜u ⎟ 3
⎝ u fs ⎠ ⎝ fs ⎠
where u ft is the gap opening displacement leading to failure and u fs denotes the maximum shear
displacement. Failure will occur when κ > u ft . Notice that the failure function is based on
relative displacements. The stiffness of the interface is taken equal to the stiffness of the matrix
material. The strength of the interface is taken from experimental results. This leads to

σ t ,max τ max
u ft = , and u fs = .
E ⋅t G ⋅t
Here t denotes the thickness of the resin rich layer, E , G and maximum stresses relate to the
matrix material.

2004 ABAQUS Users’ Conference 427

In case of damage the elastic properties of the interface will reduce. The damage parameters are
defined as

u ft (− C11u ft (κ − u ft ) / Gc ,I )
d1 = 1 − e ,
κ
u ft (− C 22 u fs 2 (κ − u ft ) /(Gc ,II ⋅u ft ) )
d2 = 1 − e ,
κ
u ft (− C 33u fs 2 (κ − u ft ) /( Gc ,II ⋅u ft ) )
d3 = 1 − e ,
κ
where Gc , I and Gc , II denote the fracture energy per unit area for mode I and mode II in [MPa],
respectively.
The modulus stiffnesses read

E G G
C11 = , C22 = , C33 = .
τ τ τ
Notice that damage parameters are equal to zero if κ > u ft . The relation between relative
displacements and tractions is given by

⎡ t1 ⎤ ⎡(1 − d1 )C11 0 0 ⎤ ⎡ u1 ⎤
⎢t ⎥ = ⎢ 0 (1 − d 2 )C22 0 ⎥ ⎢u ⎥ .
⎢ 2⎥ ⎢ ⎥⎢ 2 ⎥
⎢⎣t3 ⎥⎦ ⎢⎣ 0 0 (1 − d 3 )C33 ⎥⎦ ⎢⎣u3 ⎥⎦

Notice that the failure function κ and the damage parameters d i cannot be chosen independently,
since
∞

∫ t du
1
u1 = uft
1 = Gc , I

must hold true. A similar relation applies for Gc,II.

428 2004 ABAQUS Users’ Conference

5. Inter Rivet Buckling simulation

The developed models described above are applied in the simulation of the static behaviour of
fuselage structures. If tested statically until failure these structures typically eventually fail by
column buckling. Among the different types of instability of a stiffened fuselage, the inter rivet
buckling is particularly challenging since physical and geometrical non-linearity needs to be
modelled together with rivet contact behaviour in order to obtain realistic results.
In general, two types of inter-rivet buckling are distinguished:
1. the classical type of a skin riveted on a stringer;
2. the skin riveted to a stringer at the crossing of the frame attachment and the stringer.
This crossing causes locally a larger distance between the last rivet in a frame clip and
the rivet in the crossing stringer, as shown in Figure 3.

Figure 3. Inter rivet buckling: larger rivet pitch where stringer pass between clips.

The physical test of the inter rivet buckling problem is shown in Figure 4. The width of the
specimen is equal to 25 or 32 mm, respectively without and with initial delamination, while the
total length is 226 mm. The specimen consists of either Glare® 3-5/4-0.41 or Glare® 4B-5/4-0.4
with a total thickness of 3.0 mm and 3.5 mm, respectively. Three layers of aluminium stiffener
tabs are added. Rivet Type I is ASNA 2026-T3A, and rivet type II is DAN5-8-8.

1
According to the laminate coding system (Vlot, 2001), the following definitions are used:
Glare®3 and Glare®4B are the Standard Glare grades, 5/4 represent the number of aluminium and
fibre layers respectively and 0.4 is the aluminium layer thickness in mm.

2004 ABAQUS Users’ Conference 429

 Figure 4. Inter rivet buckling test.

The loading at the right hand side of the specimen is applied as a uniform horizontal displacement.
The C3D8I incompatible modes solid elements are used for the Glare® material and the aluminium
2024-T3 tabs.
Delamination and prepreg damage modelling is obtained by using UINTER and UMAT
subroutines.
The behaviour of the aluminium layers in the Glare® is modelled by an orthotropic plasticity
model. It accounts for orthotropic hardening by using the *POTENTIAL option in ABAQUS.
Table 1 gives the plasticity data for Aluminium 2024-T3, which is used in Glare®. Isotropic elastic
properties are: Young’s modulus 73800 MPa and Poisson’s ratio 0.33.

Table1. Plasticity data for aluminium 2024-T3.

Yield 300 320 340 355 375 390 410 430 450 470 484
stress
[MPa]

Plastic 0.00 0.016 0.047 0.119 0.449 1.036 2.130 3.439 5.133 8.00 14.71
strain
[%]

The UD glass-epoxy prepreg shows a transversely isotropic behaviour. The corresponding elastic
properties are listed in Table 2.

430 2004 ABAQUS Users’ Conference

 Table2. Elastic properties and ultimate stresses of UD prepreg.
u,t u,c u,t u,c u u
EL ET GLT GTT νLT νTT σL σL σT σT τLT τTT
[MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa]

53980 9412 5548 3000 0.33 0.33 2430 2000 50 160 50 50

The rivets between the stiffener and the skin of a fuselage are modelled by a plastic material
model, but without any failure criterion. This reflects the intention of the design, according to
which no rivet failure is allowed prior to buckling of the skin.

Figure 5. Numerical model of the inter rivet buckling problem.

2004 ABAQUS Users’ Conference 431

 initial
delamination

Figure 6. Delamination modelling.

Due to symmetry reasons, only a quarter of the specimen is modelled. The different interfaces and
their numbering are depicted in Figure 5.
A section is shown through a model in Figure 6, and an initial delamination is seen at interface 5
(see Figure 5) placed at the most critical location at the edge of the stiffener plate.
Tests on samples with initial defects were carried out on specimens having a width of 32 mm
instead of the 25 mm. The increased width is required to be able to add an initial circular
delamination with a diameter of 12 mm. Firstly an analysis has been carried out on a 32 mm
specimen without an initial delamination. Subsequently, a specimen having an initial delamination
as described above has been analysed. As a reference, also the modelling of a 25 mm wide
specimen without initial delamination is considered.

6. Results

In this section results are presented in form of axial average stress at the center of the specimen
versus axial displacement from numerical simulations of the inter rivet buckling problem
described in the previous section. Moreover a comparison between the numerical results obtained
and experimental results is shown.
The experimental results refer to inter-rivet buckling tests that were carried out on both Glare® 3-
5/4-0.4 and Glare® 4A-5/4-0.4 specimens having a width of 25 mm. Additional tests were carried
out on the effect of a circular initial delamination, using a specimen width of 32 mm.
A Glare® 3-5/4-0.4 specimen having a width of 25 mm is analysed. Initially, the clamped part of
the specimen (50 mm) is not modelled. Instead, it is assumed that the edge of the specimen
remains

432 2004 ABAQUS Users’ Conference

 u = 2 mm

Figure 7. Simplified boundary conditions.

straight and the specimen is loaded by a uniform prescribed displacement of 2 mm at this edge,
see Figure 7. Finally, the clamping conditions were modelled in an improved manner, more
realistic, leading to the boundary conditions as depicted in Figure 8.
Results are shown in Figure 9. Tests (Ypma, 2001), in a batch of 4, are shown with thin lines,
numerical simulations are shown with thick lines. For the simplified boundary conditions it can be
noticed that the FE curve shows a too stiff behaviour and the gross strength is roughly 15%-20%
too high. For the improved modelling of the boundary conditions, both the stiffness and the gross
strength are in good agreement with the experimental results. It should be noticed that these test
conditions are more severe than the conditions in a real aircraft fuselage, and thus conservative.
The numerical simulation showed that damage develops in the specimen during loading. Around
the rivets and at the buckled area matrix cracks are found. No fibre damage occurs during the

2004 ABAQUS Users’ Conference 433

 p = 10 MPa

u = 2 mm

z=0

Figure 8. Improved model of boundary conditions.

434 2004 ABAQUS Users’ Conference

 275

250

225

200

175

150 G3-1

G3-2
125
[MPa
100
] G3-3

G3-4
75 Glare 3-5/4-0.4 L, simplified BC

50 Glare 3-5/4-0.4 L

25

0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25

[mm]
Figure 9. Boundary conditions results.

analysis. At interface 5, in the middle of the specimen delamination takes place upon ultimate
load. Also, there is some delamination around the rivet and just in front of it. However, the main
delamination is found to be a free edge delamination crack. This failure mode was also observed
in the experiments. Moreover, experimental results show that the location of the delamination at
interface 5 at the free edge is also correctly predicted by the numerical simulation, see Figure 10.

2004 ABAQUS Users’ Conference 435

 Figure 10. Inter rivet buckling tests.

436 2004 ABAQUS Users’ Conference

 225

200

175

150

125
width 25mm

100 width 32mm

[MPa] width 32mm + initial delamination

75

50

25

0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2

[mm]

Figure 11. Delamination results.

Figure 11 (all numerical) shows that the initial delamination has hardly any influence on the
ultimate load. The initial delamination did not grow until ultimate load upon which a slight growth
took place.

2004 ABAQUS Users’ Conference 437

7. Summary and conclusions

Based on the results presented, it can be concluded that:

• The damage models for FML developed during the present research are capable of
accurately predicting the inter rivet buckling behaviour both qualitatively and
quantitatively.

• Proper modelling of the boundary conditions for the clamping is important. Improper
boundary conditions can overestimate the gross stress by approximately 20%.

• Upon inter rivet buckling, delaminations occurred at the free edge of the test specimen
which are not relevant in aircraft conditions and in design.

• In these simulations, the circular initial delamination did not grow until ultimate load.

• The initial delamination was not found to have an influence on the gross stress at failure.

438 2004 ABAQUS Users’ Conference

8. References

1. Alessandrini, A., “Delamination Behaviour of Z-pinned Laminates under Shear Loading”,

M.Sc. Thesis, Cranfield University, 2003.
2. Allix, O., Ladeveze, P., “Modelling and Computation of Delamination for Composite
Laminates”, Arch. Mech, vol. 44, pp. 5-13, 1992.
3. Hashagen, F., “Numerical Analysis of Failure Mechanisms in Fibre Metal Laminates”,
Dissertation, Delft University of Technology, Delft, 1998.
4. Linde, P., “Fibre Metal Laminate Simulation”, Outline Proposal of Research and Technology
Project, Internal report, Airbus Deutschland GmbH, 2002
5. Puck, et al., “Guidelines for the Determination of the Parameters in Puck’s Action Plane
Strength Criterion”, Composites Science and Technology, vol. 62, pp. 371-378, 2002.
6. Schipperen, J. H. A., “Computational Modelling of Failure in Fibre Rinforced Plastic”,
Dissertation, Delft University of Technology, Delft, 2001.
7. Vlot, A., Gunnink, J. W., “Fibre Metal Laminates - An introduction”, Kluwer Academic
Publishers, 2001.
8. “Writing user Subroutines with ABAQUS”, Hibbitt, Karlsson & Sorensen, Inc., 2001.
9. Ypma, M.S., Exploratory compression strength tests on skin detail (IRB frame clip), Report
B2V-00-59, Delft University of technology, 2001

2004 ABAQUS Users’ Conference 439