Composites: Part A 42 (2011) 612–622

Contents lists available at ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Experimental and numerical analyses of textile reinforcement forming

of a tetrahedral shape
S. Allaoui a, P. Boisse b,⇑, S. Chatel c, N. Hamila b, G. Hivet a, D. Soulat a, E. Vidal-Salle b
a
Université d’Orléans, PRISME, Orléans, F-45072, France
b
Université de Lyon, LaMCoS, Insa Lyon, F-69621, France
c
EADS France, Innovation Works, Centre Technocampus EMC2, Bouguenais, F-44340, France

a r t i c l e i n f o a b s t r a c t

Article history: An experimental device for textile composite reinforcement forming is presented. A strongly double
Received 18 April 2010 curved tetrahedral shape is formed by punch and die. This shape is those of the corner fitting parts used
Received in revised form 18 October 2010 as corner brackets. The device shows that is possible to obtain such geometry by punch and die forming
Accepted 1 February 2011
thanks to strong blank holder loads and an appropriate reinforcement. There is no wrinkle in the tetra-
Available online 25 February 2011
hedral part of the formed shape but the six blank holders create wrinkles in the plane part of the preform.
The shear angles reach 60° but there is no wrinkling in this zone. The presented forming process enables
Keywords:
the experimental validation of a semi-discrete simulation approach. It is shown that shear angles and
A. Fabrics/textiles
C. Finite element analysis
wrinkle shapes obtained by this numerical approach are in good agreement with the forming experiment.
E. Forming The computation of the shape of wrinkles after forming is necessary to check that these wrinkles do not
E. Preform expand to the useful part of the preform. This needs to take the bending stiffnesses into account. This is
not the case when the simulation is based on a membrane approach.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction importance for the understanding of deformation modes and for

the analysis of the defects that can occur during the forming.
In Liquid Composite Moulding (LCM) processes, a resin is in- Concerning punch and die forming of textile preforms, several
jected on a textile preform [1–3]. The shape of this preform can experimental forming system have been designed for hemispherical
be complex. It can be obtained by a punch and die forming process forming [7–15] (and extended hemispherical shape [16]). The
from initially flat textile reinforcement. In case of double curved hemispherical shape is much studied because it is rather simple,
shapes, the forming process may be delicate because it needs in- it is double curved and leads to large shear angles (about 45°). In
plan deformations of the textile reinforcement and especially in- the present paper an experimental device is presented for the
plane shear strains. A blank holder is then generally necessary in forming of a tetrahedral shape which is those of corner fitting parts
order to obtain a preform without defects (an especially with no (or corner brackets). This shape is strongly double curved and is
wrinkles). more difficult to form than an hemispherical part especially when
The experimental analysis and the numerical simulation of this the radius of curvature are small. It is shown that it is possible to
forming process of the textile reinforcement are important in order manufacture this preform by punch and die forming thanks to a
to determine the conditions for a successful manufacturing of a strong blank holder load and appropriate mechanical properties
preform (without defects). In addition they can determine the of the textile reinforcement (the thin interlock G1151Ò in the pres-
direction and the density of the fibres at any point on the preform ent work). There is no wrinkle in the tetrahedral part of the preform
after forming. Those directions are much depending on the forming but many wrinkles in the plane zone due to the six blank holders. A
process because shear angles between warp and weft yarns due to main results given by this forming experiment concern the ‘locking
manufacturing can be large when the final shape is double curved. angle’ This locking angle is commonly considered as the angle cor-
These directions condition the mechanical behaviour of the final responding to the onset of wrinkling [17–20]. It is shown that very
solid composite, but also the filling of the resin in the case of a large shear angle are necessary for the tetrahedral forming but
liquid moulding process [4–6]. there is no wrinkle in this zone. Conversely there are many wrinkles
The experimental devices for forming of textile preforms with in other areas where the shear angle are much smaller.
deformation and loads measurements facilities are of main A main interest of the experimental forming device is validation
of numerical forming simulations. These simulations permit to
⇑ Corresponding author. Tel.: +33 4 72 43 63 96. avoid the trial–error approaches for the development of forming
E-mail address: [email protected] (P. Boisse). processes and they also compute the directions and densities of

1359-835X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2011.02.001
 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622 613

the fibres before the injection stage. There are many simulation
methods that are developed today based on different approaches
(see Section 3) and it is important to know their abilities and their
limits. The tetrahedral forming device is used to validate the semi-
discrete simulation approach presented in [21]. In this approach
(that is briefly described in Section 3) a finite element is made
up of woven unit cells. The tensile, in-plane shear and bending
internal virtual work are directly computed from the nodal dis-
placements and from the mechanical properties obtained by
experimental tests specific to textile composite reinforcements.
This finite element is a shell element because bending stiffness is
important for wrinkling simulations. In order to improve the
numerical efficiency, a rotation-free formulation is used [22,23].
There are no rotation degrees of freedom but the curvatures are
computed from the position of the neighbouring elements.
The numerical/experimental comparison concerns principally
two points: the shear angles and the deformed shape, especially
the shape of the wrinkles. The agreement is good. In particular,
the simulation gives a realistic description of the wrinkles after
forming. This is possible with the semi-discrete approach because
the bending strain energy and the bending behaviour of the rein-
forcement (that is very specific) are considered. Some numerical
analyses have been presented on onset of wrinkles [10,13], but
they do not depict the shape of the wrinkles because they are
based on membrane approaches and neglect the bending stiffness Fig. 1. Example of double curved shape for aeronautical piece. (For interpretation of
that conditions the wrinkle shape. However in a forming process the references to colour in this figure legend, the reader is referred to the web
such as the tetrahedral forming presented in this paper, it is not version of this article.)
possible to avoid all wrinkles. The numerical simulation must
check that the process conditions insure that those wrinkles do
not expand in the useful part of the preform. In that goal, the com- loading parameters (loads on the tools, on the blank holders, speed
putation of the shape of the wrinkles after forming, as it is per- of the tools . . .) and also to measure the geometry of the formed
formed in the present study, is important. part. The state of the preform after forming is mainly characterised
by the directions of the fibres (or in-plane shear angles) and the
shape of possible wrinkles. In case of a double curved part, it is of-
2. Experimental forming ten impossible to avoid wrinkles anywhere in the woven fabric but
these wrinkles must not extend to the useful part of the preform.
2.1. Motivation Consequently it is important that the simulation tools forecast effi-
ciently these wrinkles.
The development of one specific device able to preform woven The comparison of the wrinkle shapes obtained by the simula-
textile reinforcement has two main objectives. First, such a device tion with the experimental ones is one of the objectives of this pa-
permit to experimentally analyze the possibilities to manufacture per. This comparison needs an accurate description of the textile
a double curved composite structure without defects in the useful
part with a given textile reinforcement. The role of the blank hold-
ers, of the pretensions, of the speed of the tools. . .can be investi-
gated. This is very useful because the composite forming
processes do not benefit from so much experience than in the case
of sheet metal forming. Furthermore the composite reinforcements
are very numerous and different and it is not simple to extrapolate A A’
the results of a forming process to another one. An example corre-
sponding to an angle bracket used in aeronautical applications is
shown Fig. 1 [24,25]. The design of the part using a composite
material leads to a minimum of 30% weight decrease. Nevertheless
the manufacturing of this part in a composite material is difficult in
particular making the preform in a RTM process with a good homo-
geneity of the fibre density and without defaults especially wrin-
kles. Other manufacturing processes are proposed for this shape
for instance weaving on three-dimensional surfaces [26]. The
forming process is faster.
Secondly, analyze of the state of the composite reinforcement C B
after forming is an essential mean of validating numerical simula-
tions. The numerical codes developed to simulate the textile
reinforcement forming can significantly reduce component devel-
opment time. But their validation by comparison with experimen-
tal forming processes is important for the confidence in their
results. That is the purpose of the present paper. The experimental Fig. 2. The preforming device. (For interpretation of the references to colour in this
forming device described in this section permits to measure the figure legend, the reader is referred to the web version of this article.)
614 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622

Fig. 3. Evolution of the tetrahedron shape during the preforming of the woven reinforcement. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

reinforcement bending. That is not possible for a large number of field is computed using the marker tracking software Deftac3D [31].
the proposed methods that are based on membrane approaches. The markers are plotted on the fabric sample in the initial state
(Fig. 4). This method is efficient in the follow-up of the yarns direc-
2.2. Description of the forming device tions which is important in preforming processes.

The device developed for experimental textile reinforcement

2.3. Mechanical properties of the woven reinforcement during forming
forming is presented Fig. 2. It is made up of two modules. A
mechanical module creates the displacement of the punch and
The tests presented in this paper are carried out on a composite
applies pressure on blank holders. An optical module permits
woven reinforcement used in aeronautics. The G1151Ò interlock
full-field strain measurements. The device is equipped with a
fabric is produced by the Hexcel company and constituted by an
tetrahedral punch (Fig. 3.) in order to form a corner fitting shape
interlock weaving of 6 K carbon yarns (630 g/m2, 7.5 yarns/cm)
(Fig. 1). The die is open (Fig. 3) for the acquisition of the digital
(Fig. 5). This textile reinforcement is an interlock fabric. Three yarn
images by the optical module, located at the top of the device
layers are linked by the weaving (Fig. 5). This woven composite
(Fig. 2A and A0 ). Under the device (Fig. 2B) an electric jack gives
reinforcement has been studied within the European project ITOOL
the movement to the punch. A multi-part blank holder is com-
[32]. Its mechanical properties during forming have been mea-
posed of six independent blank holders actuated by pneumatic
sured by tensile tests, picture frame tests and bending tests. The
jacks that enable to impose and sensor independently a variable
tensile properties of woven fabrics are complex because the weav-
pressure (between 0 and 6 bars) on each of them (Fig. 2C). Fig. 3
ing leads to non-linearities due to the decrease of crimp at the
shows the evolution of the preform shape during the process.
beginning of the loading. Furthermore it is a biaxial phenomenon
The optical module consists in two numerical cameras that are
because the warp and weft axial strains are related by the weaving
located at the top of the device (Fig. 2). They can be positioned in
[33,34,30]. Nevertheless because the G1151Ò is a balanced fabric
function of the specific zone of the fabric that has to be analysed.
with strong tensile stiffness in both directions, the extensions in
They can be used to determine the strains of the preform by digital
the yarn direction are very small and these tensile non-linearities
image correlation [27–30]. In the present study, the 3D displacement
are not of main importance in the forming. For this reason the
woven reinforcement will be considered as linear with the same
tensile stiffness in warp and weft direction (Table 1).
The in-plane shear behaviour of the G1151Ò reinforcement has
been intensively analysed by picture frame [35] and bias-extension
test [29]. The in-plane shear curve is identified as a polynomial of

Fig. 4. Sample with markers in the initial state. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this
article.) Fig. 5. Woven carbon reinforcement (G1151Ò).
 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622 615

Table 1
Mechanical properties of the G1151Ò interlock fabric.

Tensile stiffness in warp direction the tension is assumed to be T1 = C1e11 C1 is assumed to be constant: C1 = 2300 N/yarn
Tensile stiffness in weft direction. The tension is assumed to be T2 = C2e22 C2 is assumed to be constant: C2 = 2300 N/yarn
In-plane shear stiffness: M s ðcÞ ¼ k1  c þ k3  c3 þ k5  c5 k1 = 0.371 mm N, k3 = 0.841 mm N, k5 = 1.03 mm N
Bending stiffness in weft direction. The moment is assumed to be M 1 ¼ B1 v11 þ C 1 v311 B1 = 0.1 N mm, C1 = 0.005 N mm
Bending stiffness in weft direction. The moment is assumed to be M 2 ¼ B2 v22 þ C 2 v322 B2 = 0.1 N mm, C2 = 0.005 N mm

order five (Table 1). Finally two bending tests in the warp and weft neighbouring (Fig. 6). The following resultants of these loads are
direction are preformed in order to identify the bending properties. considered:
A cantilever bending test based on the deformation of the fabric
under its own weight is used [36]. The identified bending moments – The tensions T1 and T2 are the resultants of the loads respec-
in function of the curvature are given Table 1. tively on warp and weft yarns in the direction f 1 and f 2 of these
yarns (Fig. 7a)
– The in-plane shear moment Ms is the moment resulting from
3. Simulation of the forming process using a semi-discrete shell
the loads on the unit woven cell at the centre of the woven cell,
finite element
in the direction of the normal to the fabric (Fig. 7b).
– The bending moments M1 and M2 resulting on the warp and
3.1. Different approaches for preforming simulations
weft yarns respectively (Fig. 7c)

To simulate draping of textile composite reinforcement, several

These Figs. 6 and 7 have been drawn in the case of a plain weave
packages are commercially available based on ‘‘kinematical mod-
for simplicity but the type of weaving can be different.
els’’ [37–39]. This method is very fast but does not account for
In any virtual displacement field g such as g = 0 on the bound-
the mechanical behaviour of the woven reinforcement for possible
ary with prescribed displacements, the principle of virtual work
sliding of the fabric with regard to the tools and for load boundary
imposes on the virtual inertial work to be equal to the difference
conditions. These points are very important in the case of preform-
between virtual internal and external works:
ing with punch and die. A complete physical simulation of a com-
posite forming process needs a model of the mechanical behaviour W ext ðgÞ  W int ðgÞ ¼ W acc ðgÞ ð1Þ
woven reinforcement and a numerical method, for instance, the
finite element method. The mechanical behaviour of fabrics is with
complex due to the intricate interactions of the yarns. It is a W int ðgÞ ¼ W tint ðgÞ þ W sint ðgÞ þ W bint ðgÞ ð2Þ
multi-scale problem. The macroscopic behaviour is very dependent
on the interactions of yarns at the meso-scale (scale of the woven W tint ðgÞ, W sint ðgÞ, W bint ðgÞ are the virtual internal works of tension, in-
unit cell) and at the micro-scale (level of the fibres constituting yarns). plane shear and bending respectively with:
Despite the many works in the field, there is no widely accepted
model that accurately describes all the aspects of a fabric mechan- X
ncell
W tint ðgÞ ¼ p
e11 ðgÞp T 1 p L1 þ p e22 ðgÞp T 2 p L2 ð3Þ
ical behaviour during forming. A first family of models considers p¼1
the fabric as an anisotropic continuum [40–47], the mechanical
behaviour of which has to take into account the influence of the X
ncell
underlying meso-structure. While these models can be easily inte- W sint ðgÞ ¼ p
cðgÞp Ms ð4Þ
grated in F.E. standard shell or membrane elements, the identifica- p¼1
tion of homogenized material parameters is difficult, especially
because these parameters change when the fabric is subjected to X
ncell

the large strains due to forming. At the opposite, some authors W bint ðgÞ ¼ p
v11 ðgÞp M1 p L1 þ p v22 ðgÞp M2 p L2 ð5Þ
p¼1
present discrete models of fabrics based on modelling of the woven
yarns or fibres usually described by simplified elements such as where ncell is the number of woven cell. The quantity A is denoted
beams and springs [48,49]. A major difficulty lies in the very large p
A when it concerns the unit woven cell number p. L1 and L2 are the
number of components at the scale of the yarn or of the fibre and
the very large number of contacts with friction between them. The
simulations presented in this paper are based on the semi-discrete
approach [21]. As in the discrete approach, the components at the f2 f1
meso-scale are considered (woven unit cells in the present work).
But in the case of the semi-discrete approach, they are part of a
finite element and their strains are given by the node displacements.
The corresponding internal virtual works are computed from ten-
sion, in-plane shear and bending behaviour of the reinforcement.
These properties are directly obtained from experimental tests
specific to the textile composite reinforcements (tension, picture
frame or bias extension and bending tests).

3.2. The semi-discrete approach for textile composite reinforcements

forming simulation

The textile composite reinforcement is seen as a set of a discrete Fig. 6. Loads on a unit woven cell. (For interpretation of the references to colour in
number of unit woven cells submitted to the loads of their this figure legend, the reader is referred to the web version of this article.)
616 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622

(a) (b) (c)

Fig. 7. Load resultants on a unit woven cell. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

length of unit woven cell in warp and weft directions. e11 ðgÞ and function of this quantity. The load that is conjugated to this shear
e22 ðgÞ are the virtual axial strain in the warp and weft directions. angle is the moment Ms (Eq. (4) and Fig. 7b).
cðgÞ is the virtual in-plane shear angle i.e. the virtual angle variation
between warp and weft directions. v11 ðgÞ and v22 ðgÞ are the virtual 3.3. Triangular finite element
curvatures of warp and weft directions. e11 ðgÞ, e22 ðgÞ, cðgÞ, v11 ðgÞ
and v22 ðgÞ are function of the gradient of the virtual displacement The three node triangle shown Fig. 8 is composed of ncelle wo-
field. T1, T2, Ms, M1 and M2 are the load results on the woven cell ven cells. This finite element has been proposed in [21]. The calcu-
as presented above and Fig. 7. lations of the nodal internal loads are briefly presented below.
Experimental tests specific to textile composite reinforcements These nodal internal loads are the only quantities that are neces-
are used to obtain these mechanical properties. The biaxial tensile sary in an explicit finite element simulation.
test gives the tensions T1 and T2 in function of the axial strain e11 The natural material coordinates n1, n2 are defined along the
and e22 [33,34], the picture frame or the bias-extension test gives sides of the element. These coordinates have the following values
the shear moment Ms in function of the angle variation c between at the nodes of the triangle: M1(0, 0), M2(1, 0), M3(0, 1). The dis-
warp and weft yarns [50–54,20,35] and the bending test give the placement u and the position x of a point P within the element
bending moments M1 and M2 in function respectively of v11 and are interpolated from the values at nodes:
v22 [36,55] An alternative to these experimental tests consists in
X
3 X
3
virtual tests i.e. in 3D simulations of the deformation of a unit wo- uðPÞ ¼ Ni uðMi Þ xðPÞ ¼ Ni xðM i Þ ð6Þ
ven cell submitted to elementary loadings such as biaxial tensions i¼1 i¼1
or in-plane shear [56,45].
where Ni are the standard linear interpolation functions:
The approach presented in Eqs. (1)–(5) is simplified since it as-
sumes that the internal load state in the material is given by mem- N1 ¼ 1  n1  n2 N 2 ¼ n1 N 3 ¼ n3 ð7Þ
brane and bending resultant loads. It is restricted to thin
reinforcements i.e. when the thickness is small in comparison with
in-plane dimensions of the part. That is true for a large part of com-
posite reinforcements.
The five load resultants T1, T2, Ms, M1 and M2 can depend on the
M5
five kinematical quantities e11, e22, c, v11 and v22. Such knowledge
is generally not available (and probably not often necessary). The
mechanical properties that are considered for the G1151Ò rein-
forcement in the present study are given in Table 1.
M6
Concerning in-plane shear, some studies have shown that the
shear force can depend on the tension state [57,35]. Nevertheless
such data are usually not available and it is assumed that the
picture frame (or the bias-extension test) gives Ms only depend-
ing on c.
In addition, the virtual bending work (5) should include a
v12 ðgÞM12 term coming from the twisting curvature. This term is
neglected in the present approach mainly because of the lack of
experimental data. Nevertheless it is probable that this twisting
term is small in case of woven material because of the structure
of the fabric made of two warp and weft set of yarns.
Finally, a remark can be done concerning the in-plane shear mo-
ment. In the case of textile deformation, the shear angle c i.e. the
M4
angle variation between warp and weft directions) is a significant
and clearly defined quantity. It is used in all the studies concerning Fig. 8. Three node finite element made of woven cells. (For interpretation of the
in-plane shear behaviour of textile reinforcement. Consequently it references to colour in this figure legend, the reader is referred to the web version of
is relevant to express the virtual internal in-plane shear work in this article.)
 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622 617

The material coordinates r1, r2 are defined along the warp and weft
directions (Fig. 8). r1 is equal to zero on M1M3 and is equal to 1 in
M2. r2 is equal to zero on M1M2 and equal to 1 in M3.
The material vectors k1, k2 are defined from r1, r2:

@x @x
k1 ¼ k2 ¼ and consequently k1 ¼ AM2 k2 ¼ BM 3 ð8Þ
@r 1 @r 2
The elementary internal virtual work of tension define the nodal
tensile internal loads

W te eT te
int ðgÞ ¼ g F int ð9Þ

Taking (Eq. (3)) into account, the internal tensile load compo-
nents can be related to tensions:
!
L1 L2
ðF te
int Þij ¼ ncelle B1ij T 1 þ B2ij T 2 ð10Þ
kk1 k2 kk2 k2

i is the index of the direction (i = 1–3), j is the index of the node

(j = 1–3). B1ij and B2ij are components of the strain interpolation.
They are constant over the element because of the linear interpola-
tion (6).
Eq. (4) leads to the following nodal elementary shear loads:
Fig. 9. Geometry of the tools. View from the top. (For interpretation of the
ðF se
int Þij ¼ ncelleBcij M s ðcÞ ð11Þ references to colour in this figure legend, the reader is referred to the web version of
this article.)
The strain interpolation components Bbij are given in Table 2.
In order to avoid supplementary degrees of freedom and conse-
quently for numerical efficiency, the bending stiffness is taken into
account within an approach without rotational degree of freedom 4.1. Experimental/numerical comparison of the deformed shape and of
[22,23]. In theses approaches the curvatures of the element are the wrinkles
computed from the positions and displacements of the nodes of
the neighbouring elements (Fig. 8). From this relation, the internal Fig. 10 shows the experimental and numerical shapes of the
bending virtual work give the bending nodal loads: preform at the end of the forming process i.e. after a 150 mm dis-
placement of the tetrahedral punch. First the forming device shows
!
L1 L2 that the forming of the corner fitting shape that comprises rather
ðF be
int Þkm ¼ ncelle Bb1km M 1 þ Bb2km M 2 ð12Þ small radius of curvature is possible thanks to the blank holders
kk1 k2 kk2 k2
that impose high tensions in the fabric and to the good behaviour
The details of the calculations of the curvature interpolation can of the G1151Ò reinforcement in forming processes. There are no
be found in [58,21]. wrinkles in the tetrahedral part of the preform. The experimental
and numerical shapes are in correct agreement. Especially the
shapes of the wrinkles are correctly described by the simulation
4. Comparison between forming simulations and forming although some differences exist. The comparison between the
experiments experimental and numerical heights of two wrinkles on the left
and right sides of the preform (W1 and W2 in Fig. 11) is made in
The forming experiments have been described in Section 2. The Table 3 (in this table the height of these two wrinkles are given
geometry of the tools used in the simulations is given in Fig. 9. The for three repeated experiments).
tetrahedral punch is in green, the die is in blue and the six blank
holders are in red. The fabric is the G1151Ò interlock fabric de- 4.2. Experimental/numerical comparison of the shear angles
scribed in Section 2.3. The mechanical properties used in the pres-
ent simulations for membrane and bending behaviour are given in The map of the shear angles in the preform is an important re-
Table 1. The pressure on the blank holder is 1 bar on 32 mm diam- sult of the simulation since it strongly influences the permeability
eter pistons. This load is important since the total force on the of the preform and the mechanical properties of the final compos-
blank holders is 480 N. The friction coefficient between the tools ite part. Fig. 11 gives the computed shear angle in the tetrahedral
and the textile reinforcement is assumed to be equal to 0.2 [59– part of the preform. The shear angles are nearly constant in large
61]. The total vertical displacement of the punch is 150 mm. zones of the tetrahedron with values close to 20°. There are narrow
zones where they are close to zero and they reach a maximum va-
lue at the base of the tetrahedron near the corners. This is the con-
sequence of the tetrahedral shape of the punch. In order to
Table 2 estimate the experimental dispersion, the measured shear angles
Membrane strain interpolation components.
are given in Table 4 for three repeated experiments. The results
B1i1 = (a  1)k1i B1i2 = k1i B1i3 = ak1i are reasonably close and the difference between the shear angles
B2i1 = (b  1)k2i B2i2 = bk2i B2i3 = k2i measured in the three tests are smaller than 2°. The experimental
B3i1 = (b  1)k1i B3i2 = bk1i B3i3 = k1i
and numerical values are compared in some points in Fig. 12 and
B4i1 = (a  1)k2i B4i2 = k2i B4i3 = ak2i
with the average angles in Table 3. The agreement between numer-
i is the direction index (i = 1–3), a and b are the lengths given in Fig. 8. ical and experimental values of the shear angles is good.
618 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622

Right view Left view

Front view

W2
W1

Fig. 10. Global comparison of numerical and experimental shapes and wrinkles. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)

Table 3
Heights of wrinkles (experiment and simulation).

Test 1 Test 2 Test 3 Average Maximum Simulation

(mm) (mm) (mm) (mm) difference (mm) result (mm)
Wrinkle 1 (W1)
20 28 25 24 8 23
Wrinkle 2 (W2)
30 30 25 28.3 5 32

The zone with the maximum shear angle is presented in Fig. 13.
It shows that very large shear angles, close to 60° are experimen-
tally obtained in the zone A. This value and its localisation are in
good agreement with the maximum shear angle computed in the
simulation (Fig. 11) (It can be noticed that this area with large
shear angles is not in the useful part of the preform i.e. the tetrahe-
dral part).

4.3. The limitation of the shear locking angle notion

The notion of ‘locking angle’ has been introduced from picture

frame tests [17–20]. The in-plane shear stiffness of a woven textile
composite reinforcement is weak until it reaches a critical shear
Fig. 11. Computed shear angles. (For interpretation of the references to colour in angle called ‘locking angle’. For larger shear angle the shear
this figure legend, the reader is referred to the web version of this article.) stiffness is much more important because of the lateral contact
 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622 619

Table 4
Shear angles for three repeated experiments.

Name of the zone (Fig. 12) Shear angle test 1 (°) Shear angle test 2 (°) Shear angle test 3 (°) Average shear angle (°) Maximum difference (°) Simulation result (°)
Z2-H 17 17 19 17.7 2 18
Z2-B 22 21 23 22 2 23
Z1-H 21 22 22.5 21.8 1.5 22
Z1-B 25 25 25 25 0 25

17° 22°

25°
18°

23°

3.5°

Fig. 12. Shear angles: numerical–experimental comparison (in yellow the experimental results, in blue the numerical ones). (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

good agreement: there is no winkling in zone A and the shear an-

gles reach 60°.
The three components of the internal load state (tension, in-
plane shear and bending) (Eqs. (1)–(4)) play a role in buckling. In
zone A of Fig. 13 the yarn tensions due to the blank holder are very
A important and avoid wrinkling in that zone.

4.4. Experiment and simulation in the case of weak blank holder forces

The objective of this analysis is to compare the experimental

forming process and the simulation when the process parameters
do not lead to a correct forming process [62]. In the present case
the pressure on the blank holders has been decreased to 0.2 bar
(instead of 1 bar in the previous experiment). Because of this lack
of blank holder force, the experimental preform shown Fig. 14a
presents wrinkles in the useful tetrahedral region. These wrinkles
in the useful region are also present in the deformed shape ob-
Fig. 13. Zone of the maximum shear angle. (For interpretation of the references to tained by the numerical simulation (Fig. 14b). If there is some dif-
colour in this figure legend, the reader is referred to the web version of this article.) ferences between the experimental and numerical shape of the
wrinkles, this simulation shows that the blank holder force (corre-
between yarns. For textile reinforcement this locking angle is of the sponding to a 0.2 bar pressure) is not sufficient. One objective of
order of 40–45° [54]. It is also commonly admitted that this ‘lock- the simulation is to avoid a trial–error adjustment of the forming
ing angle’ corresponds to the onset of wrinkling. This locking angle’ process parameters and to determine a set of parameters that will
is used as an indicator of wrinkling when forming a woven fabric. lead to (or at least will be closed to) a satisfactory experimental
The forming experiment presented in this section clearly shows manufacturing process.
this is not correct. In-plane shear angles reach 60° in the zone A of
the preform (Fig. 13). But there is no wrinkle in that area. Con- 4.5. Simulation of the forming process using a continuum FEA
versely, there are many wrinkles in the plane zone of the preform approach
where the in-plane shear angle is much lower. (The shear curve of
the G1151Ò fabric is rather progressive but the ‘locking angle’ can The forming process described in Section 2 has also been simu-
be estimated at 40°.) The results of the forming simulation are in lated using the Abaqus/Explicit commercial code. A material model
620 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622

(a)

Wrinkles in
the useful Fig. 15. Process simulation using a continuum mechanics FEA simulation. (For
region. interpretation of the references to colour in this figure legend, the reader is referred
(b) to the web version of this article.)

Table 6
Shear angles computed with the Vfabric simulation. Comparison with the experimental
results and with the semi-discrete approach.

Name of Shear angle Simulation semi- Simulation Vfabric

the zone experimental discrete element (continuum
(Fig. 12) (average) (°) (Section 3) (°) approach) (°)
Z2-H 17.7 18 21
Z2-B 22 23 19
Z1-H 21.8 22 19
Z1-B 25 25 21
Fig. 14. Experimental (a) and numerical (b) forming processes in the case of weak BG 3 3.5 2.7
blank holder forces. (For interpretation of the references to colour in this figure Zone A (Fig. 13) 60 59 54
legend, the reader is referred to the web version of this article.)

material are reasonable but they are further from the experimental
results than those obtained by the semi-discrete approach.
Table 5
Tensile and in-plane shear stiffnesses of the continuum material.

Tensile stiffness in warp direction r11 = E1e11 E1 = 66,300 N/mm2 4.6. Remark on the nature of the simulation
Tensile stiffness in weft direction r22 = E2e22 E2 = 66,300 N/mm2
In-plane shear stiffness: s = s(c) s = 0 (N/mm2) c = 0 (rd) The geometry obtained after the forming process (Fig. 10) is
s = 1.54 c = 0.1 mainly depending on the loads due to the blank holder. Moreover
s = 2.69 c = 0.2
there are wrinkles in the plane shape of the preform and it is
s = 5.77 c = 0.40
s = 11.54 c = 0.70 important to check that these wrinkles do not expand to the tetra-
s = 26.92 c = 0.87 hedral part.
s = 76.92 c = 1.05 Among the different approaches developed for fabric forming
s = 173.08 c = 1.22 simulations, the kinematical models [37–39] do not take the exter-
nal loads (such as blank holder forces) into account and cannot be
used in the present case. Among the finite element approaches the
membrane approach is very common [8–11,13,46]. The lack of
called ⁄FABRIC has been introduced in the 6.8 version of this code bending stiffness does not permit to describe correctly the wrin-
in order to capture the nonlinear anisotropic behaviour of woven kles. This is possible with the proposed semi-discrete approach
fabrics [63]. in which the tensile, in-plane shear and bending behaviour are
The data requested for this material model do not include bend- described.
ing stiffness. This one is linked to the membrane stiffness for a gi-
ven thickness in a standard plate theory. In order to obtain a
bending stiffness corresponding to those given in Table 1 (limited 5. Conclusion
to the linear term), the thickness is fixed to 0.026 mm. With this
thickness, the membrane rigidities corresponding to those given An experimental device for textile composite preforming has
in Table 1 for the semi-discrete approach are given in Table 5. been presented. It has been shown that a tetrahedral shape used
The result of the simulation is presented in Fig. 15. Globally the for corner brackets can be formed by punch and die without wrin-
results are consistent with the experimental ones and with the kles in the useful part. This forming process has been possible
numerical results obtained by the semi-discrete approach. There thanks to strong loads on the blank holders and because the
is no wrinkle in the useful tetrahedral region. Nevertheless the G1151Ò reinforcement is well suitable to the forming that need
shapes of the wrinkles in the other zones are not well described. large shear angles. This experimental device is of main importance
The computed shear angle are compared to the experimental in order to test and to validate the software’s used to simulate the
values and to the angles obtained from the semi-discrete method forming processes. The tedrahedral forming process has shown
in Table 6. The zone with very small shear angle in the tetrahedral that the proposed semi-discrete simulation approach compute
region and the zone of maximal shear in zone A (defined Fig. 13) the shear angle with a good agreement with experiments. Above
are well obtained. The shear angles computed using the ⁄FABRIC all, this numerical method computes correct shapes of wrinkles.
 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622 621

These shapes are important to check that wrinkles do not extend to [24] Vu Duong Anh. Etude expérimentale du comportement mécanique de renforts
composites tissés lors de la mise en forme sur géométries non développables.
the useful part of the preform.
PhD thesis, University of Orléans; 2008.
In this forming process many wrinkles are present in the final [25] Allaoui S, Launay J, Soulat D, Chatel S. Experimental tool of woven
deformed shape. This is not the case in the ‘‘double dome’’ forming reinforcement forming. In: Proceedings of esaform conference, Lyon; 2008.
chosen as a benchmark for composite reinforcement forming sim- [26] Legrand X, Boussu1 F, Blot P, Guitard D. A new technique of weaving 3d surface
application to carbon/epoxy corner fitting plies. In: Proceedings of esaform
ulations [64]. The tetrahedral forming process could be added to conference, Twente; 2009.
this benchmark in order to test the capability of the different codes [27] Dumont F, Hivet G, Rotinat R, Launay J, Boisse P, Vacher P. Field measurements
to simulate wrinkling development. for shear tests on woven reinforcements. Mécanique Ind 2003;4:627–35.
[28] Zhu B, Yu TX, Tao XM. Large deformation and slippage mechanism of plain
woven composite in bias extension. Composites A 2007;38:1821–8.
Acknowledgements [29] Lomov S, Boisse P, Deluycker E, Morestin F, Vanclooster K, Vandepitte D, et al.
Full field strain measurements in textile deformability studies. Composites A
2008;39:1232–44.
The work reported here has been carried out in the scope of the [30] Willems A, Lomov SV, Verpoest I, Vandepitte D. Optical strain fields in shear and
project ITOOL (European Commission), of the project LCM3M (ANR, tensile testing of textile reinforcements. Compos Sci Technol 2008;68:807–19.
[31] Bremand F, Dupre JC, Lagarde A. Mesure de déformations sans contact par
French National Research Agency) and has been supported by analyse d’images. In: Proceedings of photomechanics; 1995. p. 171–7.
EADS IW company. [32] ITOOL ‘Integrated Tool for Simulation of Textile Composites’. European specific
targeted, research project, sixth framework programme, aeronautics and
space. <http://www.itool.eu>.
References [33] Kawabata S, Niwa M, Kawai H. The finite deformation theory of plain weave
fabrics, part I: the biaxial deformation theory. J Text Inst 1973;64(1):21–46.
[1] Advani SG. Flow and rheology in polymeric composites manufacturing. [34] Buet-Gautier K, Boisse P. Experimental analysis and modeling of biaxial
Amsterdam: Elsevier; 1994. mechanical behavior of woven composite reinforcements. Exp Mech
[2] Rudd CD, Long AC. Liquid molding technologies. Woodhead Publishing 2001;41(3):260–9.
Limited; 1997. [35] Launay J, Hivet G, Duong AV, Boisse P. Experimental analysis of the influence of
[3] Parnas RS. Liquid composite molding. Hanser Garner Publications; 2000. tensions on in plane shear behaviour of woven composite reinforcements.
[4] Hammami A, Trochu F, Gauvin R, Wirth S. Directional permeability Compos Sci Technol 2008;68:506–15.
measurement of deformed reinforcement. J Reinf Plast Compos 1996;15: [36] de Bilbao E, Soulat D, Hivet G, Gasser A. Experimental study of bending
552–62. behaviour of reinforcements. Exp Mech 2010;50:333–51.
[5] Bickerton S, Simacek P, Guglielmi SE, Advani SG. Investigation of draping and [37] Van Der Ween F. Algorithms for draping fabrics on doubly curved surfaces. Int
its effects on the mold filling process during manufacturing of a compound J Numer Methods Eng 1991;31:1414–26.
curved composite part. Composites A 1997;28:801–16. [38] Long AC, Rudd CD. A simulation of reinforcement deformation during the
[6] Loix F, Badel P, Orgéas L, Geindreau C, Boisse P. Woven fabric permeability: production of preform for liquid moulding processes. Proc Inst Mech Eng J Eng
from textile deformation to fluid flow mesoscale simulations. Compos Sci Manuf 1994;208:269–78.
Technol 2008;68:1624–30. [39] Cherouat A, Borouchaki H, Billoet JL. Geometrical and mechanical draping of
[7] Gelin JC, Cherouat A, Boisse P, Sabhi H. Manufacture of thin composite composite fabric. Eur J Comput Mech 2005;14(6–7):693–708.
structures by the RTM process, numerical simulation of the shaping operation. [40] Spencer AJM. Theory of fabric-reinforced viscous fluid. Composites A
Compos Sci Technol 1996;56(7):711–8. 2000;31:1311–21.
[8] Dong L, Lekakou C, Bader MG. Processing of composites: simulations of the [41] Peng X, Cao J. A continuum mechanics-based non-orthogonal constitutive
draping of fabrics with updated material behaviour law. J Compos Mater model for woven composite fabrics. Composites A 2005;36:859–74.
2001;35:38–163. [42] Yu W, Harrison P, Long A. Finite element forming simulation for non-crimp fabrics
[9] Boisse P, Zouari B, Daniel J-L. Importance of in-plane shear rigidity in finite using a non-orthogonal constitutive equation. Composites A 2005;36:1079–93.
element analyses of woven fabric composite performing. Composites A [43] King MJ, Jearanaisilawong P, Socrate S. A continuum constitutive model for the
2006;37(12):2201–12. mechanical behavior of woven fabrics. Int J Solids Struct 2005;42:3867–96.
[10] Lin H, Wang J, Long AC, Clifford MJ, Harrison P. Predictive modelling for [44] Ten Thije RHW, Akkerman R, Huetink J. Large deformation simulation of
optimization of textile composite forming. Compos Sci Technol anisotropic material using an updated Lagrangian finite element method.
2007;67:3242–52. Comp Methods Appl Mech Eng 2007;196(33–34):3141–50.
[11] Lee J, Hong S, Yu W, Kang T. The effect of blank holder force on the stamp [45] Badel P, Gauthier S, Vidal-Salle E, Boisse P. Rate constitutive equations for
forming behaviour of non-crimp fabric with a chain stitch. Compos Sci Technol computational analyses of textile composite reinforcement mechanical
2007;67(3–4):357–66. behaviour during forming. Composites A 2009;40:997–1007.
[12] Molnar P, Ogale A, Lahr R, Mitschang P. Influence of drapability by using [46] Khan MA, Mabrouki T, Vidal-Sallé E, Boisse P. Numerical and experimental
stitching technology to reduce fabric deformation and shear during analyses of woven composite reinforcement forming using a hypoelastic
thermoforming. Compos Sci Technol 2007;67(15–16):3386–93. behaviour. Application to the double dome benchmark. J Mater Process
[13] Skordos AA, Monroy Aceves C, Sutcliffe MPF. A simplified rate dependent Technol 2010;210:378–88.
model of forming and wrinkling of pre-impregnated woven composites. [47] Aimène Y, Vidal-Sallé E, Hagège B, Sidoroff F, Boisse P. A hyperelastic approach
Composites A 2007;38:1318–30. for composite reinforcement large deformation analysis. J Compos Mater
[14] Vanclooster K, Van Goidsenhoven S, Lomov SV, Verpoest I. Optimizing the 2010;44(1):5–26.
deepdrawing of multilayered woven fabric composites. Int J Mater Form [48] Durville D. Modélisation par éléments finis des propriétés mécaniques de
2009;2(Suppl. 1):153–6. structures textiles: de la fibre au tissu. Eur J Comput Mech 2002;11(2–3–
[15] Li X, Bai S. Sheet forming of the multi-layered biaxial weft knitted fabric 4):463–77.
reinforcement. Part I: On hemispherical surfaces. Composites A [49] Pickett AK. Review of finite element methods applied to manufacturing and
2009;40:766–77. failure prediction in composite structures. Appl Compos Mater 2002;9:43–58.
[16] Vanclooster K, Lomov SV, Verpoest I. Experimental validation of forming [50] Wang J, Page JR, Paton R. Experimental investigation of the draping properties
simulations of fabric reinforced polymers using an unsymmetrical mould of reinforcement fabrics. Compos Sci Technol 1998;58:229–37.
configuration. Compos A: Appl Sci Manuf 2009;40(4):530–9. [51] Potter K. Bias extension measurements on cross-plied unidirectional prepreg.
[17] Prodromou AG, Chen J. On the relationship between shear angle and wrinkling Composites A 2002;33:63–73.
of textile composite preforms. Composites A 1997;28A:491–503. [52] Potluri P, Perez Ciurezu DA, Ramgulam RB. Measurement of meso-scale shear
[18] Lebrun G, Bureau MN, Denault J. Evaluation of bias-extension and picture- deformations for modelling textile composites. Composites A 2006;37:
frame test methods for the measurement of intraply shear properties of PP/ 303–14.
glass commingled fabrics. Compos Struct 2003;61:341–52. [53] Lomov SV, Willems A, Verpoest I, Zhu Y, Barburski M, Stoilova Tz. Picture frame
[19] Sharma SB, Sutcliffe MPF, Chang SH. Characterisation of material properties for test of woven composite reinforcements with a full-field strain registration.
draping of dry woven composite material. Composites A 2003;34:1167–75. Text Res J 2006;76(3):243–52.
[20] Harrison P, Clifford MJ, Long AC. Shear characterisation of viscous woven [54] Cao J, Akkerman R, Boisse P, Chen J, et al. Characterization of mechanical
textile composites: a comparison between picture frame and bias extension behavior of woven fabrics: experimental methods and benchmark results.
experiments. Compos Sci Technol 2004;64:1453–65. Composites A 2008;39:1037–53.
[21] Hamila N, Boisse P, Sabourin F, Brunet M. A semi-discrete shell finite element [55] Kawabata S. The standardization and analysis of hand evaluation. 2nd
for textile composite reinforcement forming simulation. Int J Numer Methods ed. Osaka, Japan: The Textile Machinery Society of Japan; 1986.
Eng 2009;79:1443–66. [56] Badel P, Vidal-Sallé E, Boisse P. Computational determination of in plane shear
[22] Onate E, Zarate F. Rotation-free triangular plate and shell elements. Int J mechanical behaviour of textile composite reinforcements. Comput Mater Sci
Numer Methods Eng 2000;47:557–603. 2007;40:439–48.
[23] Sabourin F, Brunet M. Detailed formulation of the rotation-free triangular [57] Lomov SV, Verpoest I. Model of shear of woven fabric and parametric
element ‘S3’ for general purpose shell analysis. Eng Comput 2006;23(5): description of shear resistance of glass woven reinforcements. Compos Sci
469–502. Technol 2006;66:919–33.
622 S. Allaoui et al. / Composites: Part A 42 (2011) 612–622

[58] Hamila N. Simulation of single and multi plies composite forming. PhD thesis, [61] Gorczyca J, Sherwood J, Chen J. A friction model for thermostamping
INSA Lyon; 2007. commingled glass–polypropylene woven fabrics. Composites A 2007;38:
[59] Gorczyca Jennifer, Sherwood James, Liu Lu, Chen Julie. Modeling of friction 393–406.
and shear in thermo-stamping process—part I. J Compos Mater 2004;38: [62] Skordos AA, Sutcliffe MPF. Stochastic simulation of woven composites forming.
1911–29. Compos Sci Technol 2008;68:283–96.
[60] Liu L, Chen J, Gorczyza JL, Sherwood JA. Modelling of friction and shear in [63] Abaqus Analysis User’s Manual. Simulia Corp, Dassault Systems; 2008.
thermostamping of composites – part II. J Compos Mater 2004;38(21): [64] WC. Woven composites benchmark forum; 2008. <http://www.wovencomposites.
1931–47. org/> [accessed 04.09.08].