Vous êtes sur la page 1sur 11

Progressive Damage Modelling of FMLs.

Implementation in a UMAT subroutine


Jan Hol and Valeria Antonelli
Faculty of Aerospace Engineering,
Delft University of Technology, Kluyverweg 1, 2629 HS Delft, The Netherlands
Abstract: The present paper presents the procedure used to write a user material subroutine
UMAT for shell elements to include property degradation of the composite plies of a Fibre Metal
Laminate (FML). The Hashims three-dimensional piecewise stress-based criteria for in-plane
failure has been implemented in the subroutine. The type of failure and its extension in the prepreg layers can be visualised in ABAQUS-CAE. Case study has been the blunt notch strength of a
GLARE plate. The numerical results were in good agreement with the experimental ones.

1. Introduction
Fibre Metal Laminates (FMLs) are hybrid laminates made of polymer matrix composite (PMC)
plies interspersed with metal plies. The two materials are assembled by curing the PMC plies and
metal foils to form a composite laminate (Figure 1).

Figure 1. Example of a FML.


The benefits of FMLs arise from the ability to tailor material properties so that the attractive
aspects of the two constituent materials are utilised and their weaknesses are avoided. The metal

2003 ABAQUS Users Conference

layers protect the PMC core from environmental effects such as moisture ingress as well as
providing improved impact resistance and bearing properties. The composite core has higher
strength and stiffness to weight ratio than monolithic metal and is known to be less sensitive to
fatigue effects. The combination of the two materials as a hybrid composite could potentially
outperform either of the two constituent materials.
To be able to fully exploit the capabilities of FML in a design stage, it is necessary to model with
accuracy the type of failure and its location within the laminate. To be able to predict with a better
accuracy the blunt notch strength of FML, a user subroutine UMAT has been implemented

2. Material modelling
As fibre metal laminates are normally used in aeronautical applications, to describe the material
behaviour of FML structures, laminated shell elements are used in FE-analyses. Two materials are
therefore defined to describe the FML: an isotropic metal layer and a 2D-orthotropic one for the
composite layer. The isotropic layer is defined as an elastic plastic material using the standard
keywords available in ABAQUS.
In general, for the composite ply, the 2D orthotropic material definition is used. In addition to this,
failure of the different layers can be verified during post processing with the classical failure
criteria as Tsai-Hill, Tsai-Wu, etc. In the present case, the Hashins in-plane stress failure [1] has
been implemented in a material subroutine.
The state of stress is characterised by the three stress components
11
11
= [C ]
22
22
12
12

Where [C] is a 3 x 3 stiffness matrix, whose components in terms of the engineering constants are
given by:
E1
;
(1 12 21)
12 E2
C21 =
;
(1 12 21 )
C31 = 0;
C11 =

21E1
;
(1 12 21 )
E2
C 22 =
;
(1 12 21 )
C 32 = 0;

C12 =

C13 = 0
C 13 = 0
C 33 = G12

with 12 E2 = 21E1 . Where E1 is the longitudinal and E2 is the transverse Youngs modulus, 12 is
the in-plane Poissons ratio and G12 is the in-plane shear modulus.
2.1

Failure criteria and degradation modelling

According to Hashin, failure occurs in different modes and cannot therefore be easily represented
by one smooth function. The failure modes and their analytical description is given by:
2

2003 ABAQUS Users Conference

Tensile fibre mode, 11 0:


2

11 12

+
= 1
X t Sc

Tensile fibre mode (fibre buckling), 11 <0:


2

11

= 1
Xc

Tensile matrix mode, 22 0:


2

22 12

+ = 1
Yt S c

Compressive matrix mode, 22 <0:


2

12
= 1
Sc

where index 1 stands for fibre and 2 for transverse directions, Xt , Xc are the tensile and
compressive strength in the fibre direction, Yt and Yc are the tensile and compressive strength in
the transverse direction and Sc is the in-plane shear strength of the ply.
The Youngs moduli, Poissons ratios and Shear moduli are reduced according to the failure
modes of the single ply:

Fibre breaking / fibre buckling:

E1d = d1 E1 , E 2d = d 2 E2 , G12d = d 3 G12 , d12 = d 4 12

Matrix cracking / matrix failure:

E 2d = d 5 E2 ; G12d = d 6 G12 ; d12 = d 4 12

Where the coefficients d 1 to d6 are degradation factors that can either be derived from tensile
testing of the composite, or taken from values found in literature [2].

3. UMAT implementation
In the input file, two materials are given which compose the laminate, the metal layer, described
with the *ISOTROPIC and *PLASTIC keywords, with the description of the material properties
and the plastic behaviour, and the one defined by the UMAT subroutine which is described by the
*USER MATERIAL and *DEPVAR keywords.

2003 ABAQUS Users Conference

The *USER MATERIAL keyword defines the number of constants that describe the material and
that are called by the user subroutine. This enables any user to use this subroutine without
requiring a background of (Fortran) programming. The *DEPVAR variable defines the number of
variables to be stored for post processing. In this case, each failure mechanism is stored as a single
variable, to enable visualization of every type of failure.
Figure 2 shows the block diagram for the UMAT subroutine.
Equilibrium iteration

Acquire data from input


file: mechanical
properties, degradation
coefficient and damage
array size
Compare stresses with
failure criteria

Failure occurred?
Yes

No

Use degraded properties to


calculate stiffness matrix

Use original properties to


calculate stiffness matrix

New increment

Update Stresses
and strains

Update damage a rray

Exit UMAT

Figure 2. Block diagram of the UMAT subroutine.


The subroutine is called for each element in the model. The stresses reached in the previous
iteration are referenced by the failure criteria to check if failure has occurred and choose the
appropriate properties. In case failure has occurred, the degraded properties are applied and the

2003 ABAQUS Users Conference

element is added to the elements that have failed one of the criteria. In case the element has not
failed any of the criteria, the undamaged properties are applied. At this point, the stresses and
strains as well as the damage array are updated for the new increment.
Care has been taken, as suggested in [2], to upgrade the damage array so that each damaged
element continues having degraded properties. If this was not done, it could happen, in fact, that
the stresses in one of the damaged elements are reduced in a following step due to lower
mechanical properties, causing the element to be again undamaged at a higher load.
More tips for writing user material subroutines can be found in [4].

4. Analysis procedure
Every analysis has been carried out with ABAQUS, using as much as possible ABAQUS /CAE
for pre and post processing.
Advantage of ABAUQS/CAE is that, being a feature-based program, every assignment (material
properties, boundary condition, load application) is made on the geometrical model, allowing easy
mesh changes.
Small drawback is the difficulty of applying line loads. As ABAQUS does not support line load
on an edge, this has to be done through point loads, which imp lies the division of the edges in
which the load is applied at as many sections as elements on that edge.
Unfortunately the used version of ABAQUS/CAE (Version 6.3-1) does not support layered
material yet; the definition of the shell properties has therefore been made by hand, directly in the
input file. On the other hand it supports all the keywords needed for the user material subroutine,
making the implementation of the needed coefficients very easy. Another advantage is the
possibility to view the progression of the damage during post processing, as this output is added to
the standard ones through the *DEPVAR keyword, as mentioned before.

5. Case study
Case study is the blunt notch strength of a GLARE3 3/2 0.3 plate, tested in uniaxial tension.
GLARE3 3/2 0.3 is the definition of a FML plate composed by three layers of aluminium of 0.3
mm thickness each, interspersed by two layers of unidirectional glass fibre reinforced epoxy layers
of 0.125 mm thickness, placed at 0 and 90. The laminate is fabricated to be symmetric with
respect to the middle aluminium layer [Al/0/90/Al/90/0/Al], and its total thickness is 1.4 mm.
Material properties of the two materials are presented in Table 1 and Table 2.
Table 1. Mechanical characteristic of the aluminium layers.
E (MPa)
71340

2003 ABAQUS Users Conference

0.3

yield (MPa)
300

max (MPa)
400

max (%)
14

The aluminium layers are defined as an isotropic material with an elastic-plastic behaviour. Each
GFRP layer is described via the UMAT subroutine through the definition of its mechanical
characteris tics and the degradation coefficients. An orientation is coupled to each layer to define
the principal direction of the layer.
Table 2. Mechanical characteristics of the unidirectional glass fibre reinforced plastic layers.
E1 (MPa)
54000

E2 (MPa)
9400

12
0.33

G12 (MPa)
5548

Xt (MPa)
2430

Y t (MPa)
47

Sc (MPa)
40

The degradation coefficients for these calculations are all taken equal to 0.05.
The model geometry is shown in Figure 3. It is a plate of 300 mm length and 50 mm width with a
central hole of 5mm diameter.
A tension load has been applied on the two short sides of the plate. To prevent rigid body motion,
the x displacement has been fixed along the yz plane as well the y displacement on two point of
the xz plane.

Figure 3. Geometry of the Finite Element Model.

In order to check the mesh sensitivity, three different meshes have been considered in an area of
20 mm x 20 mm around the cut-out, while the mesh seed on the edges of the plate has a typical
length of 3 mm and has remained unchanged for all four calculations. The different meshes are
shown in Figure 4.

2003 ABAQUS Users Conference

Figure 4. Different mesh densities.


All meshes gave the same load at which fibre and resin failure occurred, although the shape of the
fibre failure for the most refined meshes was different from the previous two (Figure 5). This
difference seems to be due to the shape of the elements more than an improvement of the results.
As it has already been pointed out in [2], the triangular mesh around a cut-out is not the best
type for this type of problem. As failure occurs at adjacent element, being the elements very small
in the critical area, failure increases in an unusual way, following the shape of the elements. Since
it is known that the delamination is triangular at cut-outs, and the mesh is becoming straighter in
the first two meshes closer to the cut-out, the second mesh has been used for further calculations.

Figure 5. Area in which fibre failure has occurred for different mesh densities.
The type of element used is S4R5 elements. These are four node elements used for the analysis of
thin shells and have 5 degrees of freedom per node. The same analysis has been carried out also
with S4R elements, which are four nodes elements for general purpose. Although resin and fibre

2003 ABAQUS Users Conference

failure prediction were quite accurate, the S4R, failed to converge quite early in the calculation,
giving a poor prediction of the blunt notch strength.
Figure 6 and Figure 7 show the progression of matrix and fibre failure respectively for the blunt
notch of GLARE 3 3/2 0.3. Matrix failure occurs at first in the layers where the fibres are placed at
90 with respect to the load, while fibre failure occurs in the layers with fibres parallel to the load.

Figure 6. Matrix failure progression for specimens tested at 0 off-axis angle from
starting of matrix failure up to maximum load.

Figure 7. Fibre failure progression for specimens tested at 0 off-axis angle from
starting of fibre failure up to maximum load.
Three more calculations have been carried out for uniaxial tension of GLARE3 3/2 0.3, at off-axis
angles, as uniaxial tests for this configurations were available. In the present work those results
are useful to check whether the subroutine works well also in the case when different fibre angles
are considered. The results obtained with the user material subroutine are compared with the test
result in Table 3.

2003 ABAQUS Users Conference

Table 3. Test versus FEM results.


Off axis angle

0
15
30
45

Test results
(average f our results)
(MPa)
472
435
357
346

ABAQUS with umat


(MPa)

Difference

431
392
328
316

9%
10 %
8%
9%

Although the degradation coefficients are not verified with test results, the results are well in
agreement with the experimental results.
Figure 8 shows the stress-strain diagram for a notched specimen at an off-axis angle of 45 under
uniaxial tension, while Figure 9 and Figure 10 show the fibre and resin failure extension at
maximum load for the same specimen.
The stiffness of the plate is not influenced by matrix failure of the composite layers, while the
fibre failure deeply influences its stiffness. The plate will fail when the aluminium layers have
failed.
350

300

net blunt notch stress (MPa)

250
First fiber
failure
200

150
First resin
failure
100

50

0
0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

strain

Figure 8. Simulated stress-strain curve for a test at an off-axis angle of 45.

2003 ABAQUS Users Conference

Figure 9. Fibre damage extension at maximum load (red part has failed).

Figure 10. Resin damage extension at maximum load (red part has failed).

10

2003 ABAQUS Users Conference

6. Conclusions
A user material subroutine for shell elements allows the prediction of the damage and failure
mechanisms within the composite layers of a fibre metal laminate. ABAQUS/CAE allows the
visualization of the different failure modes and their progression in the structure at increasing load.
The user subroutine is created in such a way to be easily used by an inexperienced user, although
attention has to be paid in the choice of element type and size.
The use of more appropriate degradation coefficient could improve the results.

7. References
1.

Hashin, Z, Failure Criteria for Unidirectional Fibre Composites, Journal of Applied


Mechanics, June 1980

2.

Feih, S, Shercliff H.R, McGrath G, 3-D progressive damage modelling for adhesively
bonded composite peel joints: implementation in a UMAT and related computational issues,
Conference proceedings of UK ABAQUS user conference, 2000

3.

Chang, F, and Chang, K, A progressive damage model for laminated composites containing
stress concentrations. Journal of Composite Materials, 21:834-855, 1987

4.

Writing user Subroutines with ABAQUS, Hibbit, Karsson & Sorensen, Inc., 2001

2003 ABAQUS Users Conference

11

Vous aimerez peut-être aussi