Composites: Part A 139 (2020) 106135

Contents lists available at ScienceDirect

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

A specific 3D shell approach for textile composite reinforcements under

large deformation
Renzi Bai , Julien Colmars , Naim Naouar , Philippe Boisse *
Université de Lyon, LaMCoS, CNRS, INSA-Lyon, F-69621, France

A R T I C L E I N F O A B S T R A C T

Keywords: The deformation of textile composite reinforcements is strongly conditioned by their fibrous composition.
A. Fabrics/textiles Standard plate and shell theories are based on kinematic assumptions that are not verified for textile re­
A. Layered structures inforcements. A 3D shell approach specific to fibrous reinforcements is proposed. It is based on two specificities:
C. Finite element analysis (FEA)
the inextensibility of the fibres and the possible slippage between the fibres. The approach is developed in a
E. Forming
Fibrous shell
continuum-based shell element. The form of the virtual work reflects the specificities of the deformation of the
fibrous reinforcements. It takes into account the tensile and bending stiffness of the fibres. Friction between fibres
is taken into account in a simple way in connection with bending. The present approach is based on the actual
physics of the deformation of the textile reinforcements. It makes it possible to simulate the 3D deformations of
textile reinforcements and provides displacements and strains for all points in the fabric thickness and the proper
rotations of the material normal.

1. Introduction methods are purely geometric and are fast. However, they do not take
into account the mechanical behaviour of the materials or the exterior
The use of composite materials has led to weight reductions and loads on the reinforcements. In view of the low bending stiffness of
consequently decrease in fuel consumption in the transport industry, textile reinforcements, some membrane approaches have been proposed
particularly in the aeronautical and automotive sectors. Composites can [13–17]. They neglect the bending stiffness. They take into account the
be adapted so that their characteristics meet the intended applications. in-plane mechanical behaviour of the fabric, in particular the in-plane
However, the manufacturing processes to obtain these materials are shear behaviour which is specific and has a major importance in
complex and achieving a defect-free composite part is a difficult chal­ draping. Nevertheless, it has been shown that bending stiffness has an
lenge. To enable the increasing use of composite materials, it is neces­ important role during draping. In particular, it conditions the onset of
sary to replace costly development with experimental methods based on wrinkling and the size of the wrinkles [18,19]. Taking bending stiffness
trial and error by optimising manufacturing parameters by means of into account is not straightforward. A standard shell approach gives a
simulations and virtual manufacturing. The manufacture of textile- bending stiffness that depends on the membrane rigidities and the
reinforced composites often requires the preforming of a dry textile thickness. This leads to a bending stiffness that is much too large for the
reinforcement and the subsequent injection of a resin in LCM processes textile reinforcement. This is due to the fibrous composition of the
(Liquid Composite Moulding) [1–3]. The composite can also be pro­ reinforcement, which makes slippage possible between the fibres. This is
duced by thermoforming a prepreg consisting of a textile reinforcement an important point that is taken into account in this article. Several
incorporating the unhardened matrix, so that the composite can be approaches have been proposed to address this difficulty. Textile rein­
formed [4–8]. In both cases (LCM and prepreg), the forming process is forcement can be considered as a laminate material with different
driven by the deformation of the textile reinforcement. The basic physics thickness properties that can be adjusted to achieve both correct mem­
of the deformation is the same and is that of the deformation of textile brane and bending stiffnesses [20–24]. Stress resultant shell approaches
reinforcement made of continuous fibres. that relate the resulting forces along a normal (Tensile and shear forces,
Kinematic drape models were the first approaches developed for the bending moments) to membrane and bending strains naturally decouple
simulation of the forming of woven textile reinforcements [9–12]. These membrane and bending stiffnesses [25,26]. Finally, the combination of a

* Corresponding author.
E-mail address: philippe.boisse@insa-lyon.fr (P. Boisse).

https://doi.org/10.1016/j.compositesa.2020.106135
Received 5 July 2020; Received in revised form 17 September 2020; Accepted 19 September 2020
Available online 28 September 2020
1359-835X/© 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 1. (a) Bending experiment of multilayer reinforcements; (b) 3 points bending test of multilayer reinforcement; (c) 3D Bending due to an imposed displacement at
the corner; (d) Simulation using Mindlin shell S3 element in Abaqus; (e) Bending of a thin reinforcement. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

bending finite element (e.g. DKT) with a membrane finite element is also thick textile reinforcements. Simulations of large deformations of textile
used [27,28]. reinforcements in 3D cases are presented and validated by comparisons
These different methods make it possible to decouple the membrane with experiments.
deformation energy from the bending energy. However, some aspects
are artificial, and these methods are not based on the physics of the 2. Specificities of the mechanical behaviour of fibrous materials
deformation of a textile reinforcement. Moreover, they do not provide
the displacements and strains for points in the thickness of the fabric. Plate and shell approaches concern solids whose geometry is close to
Verification of the inextensibility of the fibres is not assured and the a middle surface and thin enough to simplify the kinematics i.e. it de­
rotations of the material normals, which are related to this inex­ pends on a smaller number of variables than 3D solids. In an orthogonal
tensibility, are not known correctly. The approach that is proposed in coordinate system of x,y,z coordinates, a plate is considered to have a
this article has this objective: to define a 3D shell approach, specific to thickness h and a middle surface z = 0. The displacements along x,y,z,
fibrous reinforcement, which gives the displacements and strain in any and the components of the rotation from the normal to the plate are
point of the textile reinforcement as well as the rotation of the material noted u,v,w, and θx , θy respectively. The hypothesis is made that the
normals (This is what a shell theory does). It will be shown that standard points along a segment oriented by the normal initially perpendicular to
shell approaches such as Kirchhoff and Mindlin are not relevant for the middle surface remain on a straight segment that consists of the
fibrous reinforcements. The proposed specific shell approach is based on deformed normal. Consequently:
the quasi-inextensibility of the fibres and the possibility of slippage ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
u u θy
between the fibres. These two points are the major specificities of the ⎣ v ⎦ = ⎣ v ⎦ + z⎣ − θ x ⎦ (1)
physics of the deformation of a fibrous reinforcement. The feasibility of
w w 0
the approach in the case of a 2D two node fibrous element in the plane
has been presented in [29]. The aim of the present article is to develop Here u, v, w are the displacements of the point of the middle surface. The
an approach to simulate all cases of 3D deformation of textile re­ strains are as follows:
inforcements. The formulation is implemented in the framework of the
3D continuum-based shell elements [30–32]. It concerns both thin and

2
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 2. Geometry of the 3D fibrous shell element. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

1 1 to the middle surface) which reflects the physics of the deformation of

ε = e + zχ εxz = γxz εyz = γyz (2)
2 2 standard materials, in particular when they are thin, is abandoned here
and replaced by the constraint of inextensibility of the fibres. The
where e is the membrane strain in the mean surface, χ is the curvature Mindlin approach, by the transverse shear Γ (Eq. (3)) makes it possible
and Γ is the transverse shear strain [33,34]. that the material normal does not remain necessarily perpendicular to
⎡
∂θ y
⎤ the middle surface. In Mindlin’s model, the transverse shear strain is
⎢ ∂x ⎥ ⎡ ⎤ defined by a constitutive law that relates it to the shear stress [33–35].
⎡ ⎤ ⎢ ⎥ ∂w
χ xx ⎢ ⎥ [ ] ⎢ θy + This does not correspond to the physics of fibre reinforcement defor­
⎢ ∂θ x ⎥ γ ⎢ ∂x ⎥⎥
χ = ⎣ χ yy ⎦ = ⎢
⎢ − ∂y ⎥
⎥ Γ = xz = ⎢ ⎥ (3) mation. The position of the material normals after deformation is
γ yz ⎣ ∂w ⎦
χ xy ⎢
⎢
⎥
⎥ − θ x + defined by the quasi inextensibility of the fibres and the possibility of
⎣ ∂θy ∂θx ⎦ ∂y
− slippage between them. Fig. 1c and d shows the deformation of a textile
∂y ∂x reinforcement composed of 11 layers of carbon reinforcement G1151
If the transverse shear strain is zero or very low, which is often (Hexcel). Fig. 1c shows the experimental deformation where the mate­
verified when the plate is very thin, then Γ = 0. The Kirchhoff theory is rial normals were marked and Fig. 1d shows the simulation performed
thus obtained where the directions initially perpendicular to the middle using the Mindlin S3 shell finite element of Abaqus software. The tensile
surface remain perpendicular to the deformed middle surface. and transverse shear properties have been optimized so that the middle
These plate approaches are very efficient for the analysis of thin surface is in agreement with the experiment. However, the material
structures because the kinematics of the deformation is given by a normals obtained by simulation do not correspond to the experiment.
reduced number of variables (u, v, w, θx , θy ) that are function of (x,y). Furthermore, an example of bending of thin textile reinforcement
Nevertheless, in order to use them, it is necessary that the kinematic (Fig. 1e) shows that the normals do not tend to be perpendicular to the
hypothesis on which they are based is verified. It will be seen that this is middle surface after deformation when the thickness is small. Unlike
not the case for fibrous reinforcements. Mindlin shells, textile reinforcements do not tend towards Kirchhoff’s
Fig. 1 shows examples of deformation of composite fibre re­ theory when the thickness is small.
inforcements. It can be seen from all the examples that the material The objective of the present work is to develop a 3D shell approach
normals (direction initially perpendicular to the middle surface and that is tailored to the deformation physics of the fibre reinforcements.
fixed to the material) do not remain perpendicular to the middle surface The objective of this shell approach is to determine all the kinematics
of the textile reinforcement. The deformations do not correspond to and stresses in the entire thickness of the textile reinforcement. It is an
Kirchhoff’s theory. The angle between the material normals and the alternative to the 3D finite element approaches (solid elements) that
middle surface is often very different from 90◦ . The deformation of the have been developed for this purpose [36–39] but with a better nu­
fibrous reinforcements has two main features. The fibres of which they merical efficiency, the number of degrees of freedom being much lower.
are composed (carbon fibres in the examples in Fig. 1) are almost Therefore, a continuum mechanics-based 3D shell element is devel­
inextensible. Especially during a forming process, the fibres practically oped. The kinematics associated with the form of the virtual work taken
do not elongate. Besides, some slippage may occur between the fibres. into account reflect the specificity of the deformation modes of the fibre
These two aspects form the physical basis for the deformation of these reinforcements (quasi inextensibility of the fibres and the possibility of
materials and are specific to textile reinforcements. They are the basis of slippage between fibres). Examples of 3D deformation of textile re­
the proposed specific shell approach. inforcements are analysed both experimentally and simulated with good
Kirchhoff’s hypothesis (the material normal remains perpendicular agreement using the proposed approach.
The fibre reinforcements under consideration are composed of two

3
R. Bai et al. Composites Part A 139 (2020) 106135

directions of quasi inextensible fibres. This is an idealized situation. In

practice, woven composite fibre reinforcements and their stacking are 3.2. Conservation of the thickness in the direction of the normal
close to this situation. The weaving creates a crimp and the condition of
inextensibility is not strictly observed. Nevertheless, on the one hand the As in the classic shell models, the approach presented is for cases
geometries of the textile reinforcements are very flat in order to give where there are no external forces in the thickness direction or where
good mechanical characteristics to the composite, on the other hand, these are not taken into account. The thickness of the reinforcement in
during a process, the deformations of the reinforcement are large and the direction of the normal to its mean surface is the sum of the thick­
the stresses are moderate. Consequently, in practice, for these composite nesses of the fibres which are assumed to be in contact. This thickness in
woven reinforcements, the condition of inextensibility is a correct the direction of the normal remains constant during deformation. It can
assumption. This can be verified in all the tests presented in this article, be seen, especially in Fig. 1, that this hypothesis is verified in the ex­
in particular Fig. 1. periments. This assumption complete the kinematics. Consequently:
h
3. Continuum mechanics-based 3D shell element for fibrous hkm = (10)
Vkm ⋅n
reinforcement
where h is the thickness along the direction normal to the mid-surface of
3.1. Geometry of the 3D shell element
the shell, hkm is the thickness along the material director Vkm at node k.

The proposed shell is developed within the framework of continuum-

based shell elements which was widely used to define effective shell 3.3. Kinematics of the fibrous shell
elements [30–32]. In this article, this fibrous triangular shell element
possesses 3 nodes, the geometry is shown in Fig. 2, the position vector of By considering the kinematic equation of continuum-based element,
a point M(ξ, η, ζ) is defined as: the expression of the incremental displacement for each point in the
element between the geometry at time it and i+1t is developed:
x(M) = x(H) + y(M) (4)
Δu(ξ, η, ζ) = i+1
x− i
x (11)
H is a point defined at the middle surface. ξ, η, ζ are the element
natural coordinates (Fig. 2) with 0⩽ξ⩽1, 0⩽η⩽1, − 1⩽ζ⩽1. The co­ Taking Eq. (5) into account:
ordinates ξ,η give the in-plane position. The element edges coincide with ∑
3 ∑
3
ζ
warp and weft directions, ζ is along the material normal direction which Δu(ξ, η, ζ) = Nk ( i+1 xk − i
xk ) + Nk ( i+1 hkm i+1 V km − i k i k
hm V m )
2
joints the top-bottom layer. k=1 k=1
(12)
The position interpolation in the element is given by: ∑
3 ∑
3
ζ
= Nk Δuk + Nk ( i+1 hkm i+1 V km − i k i k
hm V m )
∑
3 ∑
3
ζ 2
(5)
k=1 k=1
x(ξ, η, ζ) = Nk xk + Nk hkm V km
2
k=1 k=1 The rotation of material director is given by two rotation components
α and β. On the time step from it to i+1t:
where xk is the position vector of node k, hkm is the thickness along the
direction of material director Vkm which is the unit material director
i+1
Vkm − i
Vkm = Δαk i Vk1 − Δβk i Vk2 (13)
defined at each node k. Nk is the shape function at node k. The update of thickness along material director direction is:
N1 = 1 − ξ − η ; N2 = ξ; N3 = η (6) h h
Δhkm = i+1 k
hm − i k
hm = i+1 n( t Vk
− (14)
A local orthogonal frame (V k1 , V k2 , V k3 ) is defined at each node m + Δαk i Vk1 − Δβk i Vk2 ) i n i Vk
m

k(Fig. 2). e1 , e2 , e3 is the global unit base vector, n is the unit vector
where the expression of the updated unit normal vector i+1n is.
normal to the element’s mid-surface.
i+1 i+1
g1 × g2 ( i g + Δu2 − Δu1 ) × ( i g2 + Δu3 − Δu1 )
e2 × V k3 i+1
n= = i 1 (15)
V k3 = V km , Vk1 = ⃒⃒ ⃒, V k = V k3 × V k1 (7) ‖ i+1 g1 × i+1 g ‖ ‖( g1 + Δu2 − Δu1 ) × ( i g2 + Δu3 − Δu1 )‖
e2 × V k3 ⃒ 2 2

Taking Eqs. (13)–(15) into Eq. (12), the displacement increment is:
A point with position x in the element gives the covariant vectors
with respect to natural coordinates:
v
The above formulation results in five degrees of freedom per node.
Δuk is the nodal incremental translation displacement vector, the other

∑
3 ∑
3
ζ ∑3
ζ
Δu(ξ, η, ζ) = Nk Δuk + Nk ( i hkm + Δhkm )(Δαk i V k1 − Δβk i V k2 ) + Nk Δhkm i V km (16)
k=1 k=1
2 k=1
2

i+1
two DOFs are two rotations components. The configuration at t
∂x ∂x ∂x enable the calculation of the internal nodal loads at i+1t.
g1 = , g2 = , g3 = (8)
∂ξ ∂η ∂ζ
4. Internal virtual work of the textile reinforcement
In order to avoid locking due to fibre inextensibility, the vector along
the direction of the warp and weft fibres k1 , k2 are equal to the covariant For any virtual displacement equal to zero on the boundary with
vectors g1 , g2 [40,41]. prescribed displacement, the virtual work theory is written:
g1 = k1 , g2 = k2 (9)

4
R. Bai et al. Composites Part A 139 (2020) 106135

δWext − δWint = δWacc (17)

The internal virtual work is separated into three parts:
Ten
δWint = δWint Bend
+ δWint Shear
+ δWint (18)

In Eq. (18), δWint denotes internal virtual work, Shear

δWint Ten
δWint , Bend
δWint ,
are the tension, bending, in-plane shear internal virtual works
respectively.

∑ ∫
Nfibres ∑∫
Nfibres
Ten
δWint = T 11f δεf11 dL + T 22f δεf22 dL (19)
f =1 Lf f =1 Lf

∑ ∫
Nfibres ∑∫
Nfibres
Bend
δWint = M 11f δχ f11 dL + M 22f δχ f22 dL (20)
f =1 Lf f =1 Lf

∑ ∫
Nfibres Fig. 3. Moment produced by internal force at top-bottom position of node k.
Shear
δWint = M sf δγf dΩ (21) (For interpretation of the references to colour in this figure legend, the reader is
f =1 Ω referred to the web version of this article.)

The superscript f indicates the fibre in consideration. T 11 , T22 are the

tensions in the fibres in the warp and weft direction, δε11 ,δε22 are virtual
tensile strains; M11 ,M22 are the bending moment on the fibres in warp
and weft directions, δχ 11 ,δχ 22 are the virtual curvature; Ms is the in-plane
shear moment, δγ is the virtual in-plane shear angle.
The form (Eqs. (19)–(21)) of the internal virtual work corresponds to
the specific mechanical behaviour of the fibrous reinforcements. The
virtual tension work takes into account the high tensile rigidity of the
fibres and leads to the quasi-inextensibility of the fibres. It controls the
deformation of the fibrous medium. The quasi-inextensibility of the fi­
bres at different altitudes in the thickness of the reinforcement leads to
Fig. 4. In-plane shear angle and virtual displacement. (For interpretation of the
specific positions of the normals as shown in the different examples (e.g. references to colour in this figure legend, the reader is referred to the web
Fig. 1). The virtual bending work considers the bending stiffness of each version of this article.)
fibre. It will also be shown in Section 5.2 that it makes it possible to take
into account the friction between the fibres for certain materials. The in-
n ∫ n ∫
plane shear stiffness in the plane is taken into account in a standard ∑ fT
∑ fT
FTen = (BTen
11 ) T
11f
dL + (BTen
22 ) T
22f
dL (24)
manner for textiles [42–46]. No other stiffness is taken into account thus f =1 Lf f =1 Lf
making it possible for the fibres to slip between the fibres whereas it is
not possible in a standard shell approach. ( )f
Here BTen matrix is the fibre tension strain interpolation matrix, in
Although most forming processes are quasi-static, the simulation is αα
which the right subscript αα represent the component in different fibre
often based on an explicit dynamic approach. [47–49]. The principle of
directions, α takes value 1 or 2. n is the number of fibres in the thickness
virtual work and the finite element approximation leads to:
considered for numerical integration. In the present work n = 3. The
Mü + Cu̇ = Fext − Fint (22) tensile virtual strain in direction α for the fibre f is consequently:
( )f f
M and C are the mass and damping matrices respectively. Fint and Fext (δεαα )f = BTen
αα δu (25)
are the internal and exterior nodal loads. Fint is specific to the textile
reinforcements and is composed of three parts: the tensile nodal loads uf is the single column displacement matrix for fibre end points of the
int , the bending nodal loads Fint
FTen Bend
and the in-plane shear nodal loads segment f. It is obtained from nodal displacements and rotations matrix
FShear . The calculation of F for the specific form of virtual works given u by using Eq. (16). With the Eq. (9), assigning the fibre direction as the
int int
in Eqs. (17)–(19) is a main point of the approach presented and is the direction of element edge, the virtual strain is shown (Eq. (26)) with
subject of Section 4.1 below. ξ1 = ξ ; ξ2 = η ; ξ3 = ζ
The central difference scheme on a time step i Δt = i+1 t − i t gives the
nodal displacement increment: ⎧ ⎫
f
〈 〉⎪
⎪
⎨ δuf1 ⎪
⎪
⎬
g ∂N1 (ξ, η) ∂N2 (ξ, η) ∂N3 (ξ, η)
i+1
uN = i uN + ( i− 1/2 1
u̇N + ( i− 1 Δt + i Δt)M− 1 ( i Fext − i
Fint )) i Δt (23) (δεαα )f = ⃦ α⃦2 ⋅ α α α δu f
2
(26)
2 ⃦gα ⃦
f ∂ξ ∂ξ ∂ξ ⎪
⎩ δuf ⎪
⎪ ⎪
⎭
3

M is the lumped matrix for node k of the shell. It is given in appendix A ( )

Ten f
[50]. The Bαα is given as:
〈 f
〉
( Ten )f g ∂N1 (ξ, η) ∂N2 (ξ, η) ∂N3 (ξ, η)
Bαα = ⃦ α⃦2 ⋅ α α α (27)
4.1. Calculation of internal nodal force ⃦gfα ⃦ ∂ξ ∂ξ ∂ξ

4.1.1. Tension The second part is the load provided by the force at top and bottom
fibre (Fig. 3). The tensile force in the different fibres in the thickness
The nodal tensile internal loads FTen
int for the element consists of two
generate moments at node k:
parts. The first part denoted by FTen concerns the displacement degrees
of freedom. The second part denoted by MTen concerns the rotation
degrees of freedom.

5
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 5. (a) Triangular element mid-surface (b) Parameters defined in two neighbouring elements. (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)

∑n
ζhkm (( Ten )f k ) ∑n
ζhkm (( Ten )f k ) In Fig. 4, considering the fibre in direction α = 1 for example, the
MTen
α k = Fk ⋅V1 , MTen
β k = − Fk ⋅V2 (28) angle between dx1 and δx1 is denoted by γ 11 . Consequently γ22 will
2 2
represent the angle between dx2 and δx2 . The virtual angle between
f=1 f=1

[
MTen = MTen Ten
]
(29) warp and weft direction is given as the combination of these two angles
α Mβ ( )f
[25] in form of Eq. (31) which gives BShear :
Eq. (28) shows the two nodal moment components corresponding ⃦ 2⃦ ⃦ 1⃦
with rotations α, β. The tensile internal load is obtained by assembling ∂uf k2 ⋅k1 ∂uf ⃦k ⃦ ∂uf k1 ⋅k2 ∂uf ⃦k ⃦
δγf = ( ⋅k1 ) ⃦ ⃦ +( ⋅k ) − ( ⋅k ) ⃦ ⃦ − ( ⋅k )
two parts FTen and MTen .
1 2 2
∂ξ ⃦ 2⃦
k ‖k1 ‖ ∂η ‖k1 ‖ ∂η ⃦ 1⃦
k ‖k2 ‖ ∂ξ ‖k2 ‖
(31)
4.1.2. In-plane shear
By using the same method as described in Section 4.1.1 for tension,
The shear angle γf for the element at position f is:
the internal loads FShear
int will also be divided into two parts, and they
( )f
δγ f = BShear δuf (30) ( )f
could be calculated from matrix BShear and the thickness along

Fig. 6. Bending test on a multilayer textile reinforcement (a) Test condition (b) Experiment (c) Simulation by 3D shell element. (d) Position of mid-surface. (e) Angles
between material directors and horizontal direction. (f) Thickness along material director. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

6
R. Bai et al. Composites Part A 139 (2020) 106135

Table 1 4.1.3. Bending

Mechanical properties of single layer in multilayer reinforcement (Section 5.1). The curvature is calculated from the position of the neighbouring
Tension stiffness (per unit width) 1200 N/mm elements. This method has been developed to obtain rotation free shell
element [51–53].
Bending stiffness (per unit width) B = 7.5 N⋅mm when |χαα |⩽0.001
The virtual curvature in warp and weft direction is interpolated from
With Mαα = Bχ αα
the nodal virtual displacement:
0.5 N⋅mm when χ αα ⩾0.001
With Mαα = Bχ αα + 0.0075 δχ αα =BBend
αα δu (34)
0.5 N⋅mm when χ αα ⩽ − 0.001
With Mαα = Bχ αα − 0.0075
Fig. 5 shows some parameters defined in the element. The height
from node k is denoted as hk , nc is the exterior normal to side in the
element’s plane, the relative rotation angle between the principal
material director. element plane and neighbour element around the side s is denoted by θs ,
s vary in (1,2,3), s’ is the corresponding node number in neighbour
∑
n
FShear =
T
(BShear )f (Cγ )f (32) element. Denoting ̂ g α the unit vector of gα , the curvature in the fibre
f =1 direction is obtained [25]:
( )f (
The moment produced by force FShear at different position in the ∑ g α ncs )2 hs θs
3
2(̂
χ αα = − ) with δθs = δφs’ + δφs (35)
thickness is: s=1
hs hs + hs’
∑ ζhkm (( Shear )f k ) ∑ ζhkm (( Shear )f k ) φs and φs′ are respectively the rotation angle of the principal element
n n
MShear
α k = Fk ⋅V1 MShear
β k = − Fk ⋅V2
f=1
2 f=1
2 and the neighbour element around side s. They are represented by the
(33) incremental nodal displacements where p and q are the complements of s
in (1,2,3): (Fig. 5)
The internal in-plane shear nodal internal loads are FShear , MShear .

Fig. 7. (a) 3-Points bending test on a single ply and twenty plies of Hexcel G1151 (b) Moment-curvature data for multilayer reinforcement. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

7
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 8. (a) Schema of 3D experiment (b) Boundary condition of Test 6.1.1 (c) Boundary condition of Test 6.1.2. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

δus ⋅ns cosβq cosβp been carried out and simulated by the 2D approach presented in [29]
δφs = − δup ⋅ns − δuq ⋅ns (36)
hs hp hq with a good agreement between tests and simulations. The 3D shell
approach proposed in this paper leads to strictly identical results in these
Thus θs can be given by the nodal displacements δu, this gives the cases.
bending strain interpolation matrix BBend
αα , consequently, the nodal
bending internal loads are:
∫ ∫ 5.2. Influence of friction between fibres
(37)
T T
FBend
int = (BBend 11
11 ) M dL + (BBend 22
22 ) M dL
L L In the approach proposed above in Sections 3 and 4, friction does not
Mαα is a function of the curvature measured in bending experiment appear explicitly. Nevertheless, the friction between the fibres exists and
[54]. The detail of the calculation of the bending nodal internal loads has an influence that may not be negligible [56–58]. In order to high­
can be found in [25,52]. light the influence of friction on bending stiffness, 3-point bending tests
are carried out Fig. 7 on two specimens consisting of a single carbon
5. Numerical simulations and experimental comparisons in 2D fabric layer (Hexcel G1151®), and twenty layers of the same rein­
forcement respectively.
5.1. Bending test on a multilayer textile reinforcement The measured bending moment versus curvature is shown in Fig. 7b.
The measured bending stiffness of the stack made of twenty G1151
The specimen consists of 20 plies of Hexcel G986® carbon twill layers is larger than the summation of the bending rigidities of twenty
weave stacked in the same orientation (Fig. 6). The horizontal and individual fabrics. This difference is due to friction between the plies and
vertical displacements are imposed at the right end of specimen, is far from negligible. In order to take friction into account in a simple
meanwhile the other is clamped. The experimental deformed shape is way in the simulations, the bending stiffness taken into account in the
shown in Fig. 6b. The simulation based on the proposed shell element is proposed shell element is that of the overall stack (and not the sum of the
shown in Fig. 6c. The material properties of Hexcel G986 are given in stiffness of the individual plies or fibres). From the point of view of
Table 1 [55]. Comparisons between numerical and experimental results experimental identification, the measurement of the stiffness of the
is given in Fig. 6d, e and f. The deflection, the rotation of the material global stack is no more complex and sometimes simpler than that of each
director and the change in thickness along material director are in good ply.
agreement with experiment. Friction within a woven reinforcement is a complex problem and
Furthermore, some other tests like cantilever bending test, have also depends on a number of factors, in particular the clamping forces
applied during forming [59]. Nevertheless, the proposed shell approach

Fig. 9. Deformed shape along side 1. (a) Experiment. (b) Simulation. Deformed shape along side 2. (c) Experiment. (d) Simulation. (For interpretation of the ref­
erences to colour in this figure legend, the reader is referred to the web version of this article.)

8
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 10. (a) (d) Mid-surface along side 1 and side 2 (b) (e) Angles between the material directors and the horizontal direction along side 1 and side 2 (c) (f) Thickness
along with material director along side 1 and side 2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

Fig. 11. Deformed shape along side 1. (a) Experiment (b) Simulation. Deformed shape along side 2. (c) Experiment (d) Simulation. (For interpretation of the ref­
erences to colour in this figure legend, the reader is referred to the web version of this article.)

is macroscopic with only one element in the thickness of the reinforce­ material directors and their rotations.
ment or reinforcement stack. The approach used to take into account the
friction between the fibres makes it possible to remain within this
6.1. Bending test of a fibrous specimen
framework and to be quite effective. Furthermore, measuring the
bending stiffness of the reinforcement is quite simple. This approach has
The specimen is a stack of 130 sheets of paper. It is not exactly a
its limitations, especially in the case of complex boundary conditions
textile composite reinforcement but a model material which, with re­
during a process. In such cases, it may be necessary to consider 3D or
gard to bending, corresponds to the problem under consideration. The
mesoscopic modelling.
in-plane shear stiffness is large and the sheets of paper do not deform in
in-plane shear.
6. Numerical simulations and experimental comparisons in 3D

6.1.1. Bending due to an imposed displacement of a corner

In this section, the proposed approach is applied to 3D deformation
A rigid support imposes a vertical displacement at the corner of the
cases. The part above the diagonal 1–3 of a rectangular fibrous specimen
rectangle specimen (Figs. 8b and 9). The experimental deformed
is clamped and the lower part is subject to bending (Fig. 8). The defor­
configuration of the two sides are shown in Fig. 9a and c and the results
mation of the two sides are captured by two cameras (Figs. 8a and 9).
of the corresponding simulations using the proposed shell approach are
Some straight lines are drawn on both two sides in the through-thickness
displayed in Fig. 9b and d.
direction, they are initially normal to the mid-surface. They show the
Fig. 10 compares the experimental and numerical deformed middle

9
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 12. (a) (d) Mid-surface along side 1 and side 2 (b) (e) Angles between the material directors and the horizontal direction along side 1 and side 2 (c) (f) Thickness
along with material director along side 1 and side 2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

surface positions, material director rotation and thickness. The corre­

Table 2
spondence between the simulation and the experiment is pretty good. In
Mechanical properties of G1151 multilayer reinforcement (Section 6.2).
this test, the curvature of the shell is non-zero and yet the material di­
Tension stiffness (per unit width) 2300 N/mm rectors remain almost parallel. This is not in accordance with standard
Bending stiffness (per unit width) Warp B = 72 N⋅mm shell theory.
Weft B = 68 N⋅mm
6.1.2. Buckling bending test
The diagonal of the specimen is clamped (Fig. 8). An in-plane
displacement is imposed at the corner of the rectangle specimen (see

Fig. 13. Bending of a G1151 laminate due to an imposed displacement of a corner. Deformed shape along side 1. (a) Experiment. (b) Simulation. Deformed shape
along side 2. (c) Experiment. (d) Simulation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

10
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 14. Bending of a G1151 laminate due to an imposed displacement of a corner. (a) (d) Mid-surface along side 1 and side 2 (b) (e) Angles between the material
directors and the horizontal direction along side 1 and side 2 (c) (f) Thickness along with material director along side 1 and side 2. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 15. Buckling bending test of a G1151 laminate. Deformed shape along side 1. (a) Experiment. (b) Simulation. Deformed shape along side 2. (c) Experiment. (d)
Simulation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Figs. 8c and 11). The deformed configuration is shown in Fig. 11. Fig. 12. 6.2.1. Bending due to an imposed displacement of a corner
compares the experimental and numerical deformed shapes with regard The G1151 laminate reinforcement is subjected to an imposed
to mean surface position, material director rotation and thickness. The displacement of a corner (Figs. 8b and 13). The experimental deformed
correspondence between the simulation and the experiment is pretty configuration captured on two sides are shown in Fig. 13a and c, the
good. corresponding simulation result is shown in Fig. 13b and d. The position
of the mid-surface, the material director directions and thicknesses are
6.2. Bending of a carbon textile reinforcement shown in Fig. 14. The simulation shows a good agreement with exper­
iment, which was not be obtained with the Abaqus S3 Mindlin shell
The considered multilayer reinforcement is made up by 11 layers of element.
G1151 carbon weaves. The dimension of the laminate is 200mm ×
150mm × 15mm. The mechanical properties of G1151 have been 6.2.2. Buckling bending test
determined in several previous studies [60–62]. The bending stiffness of An in-plane displacement is imposed at the corner of the rectangle
this multilayer specimen is measured in three point bending test, the G1151 laminate reinforcement with a clamped diagonal (Figs. 8c and
influence of friction is taken into account (Table 2). 15). The deformed configuration is shown in Fig. 15a and c, the

11
R. Bai et al. Composites Part A 139 (2020) 106135

Fig. 16. Buckling bending test of a G1151 laminate. Deformed shape along side 1. (a) Experiment. (b) Simulation. Deformed shape along side 2. (c) Experiment. (d)
Simulation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

corresponding simulation is presented (Fig. 15b and d). The deformed coherence with the experiments. A simple approach was used to take
middle surface, the material director directions and thicknesses are into account friction between layers. The bending stiffness was
shown in Fig. 16. The simulation shows a fairly good agreement with measured on the overall reinforcement, and friction increases the
experiment. bending stiffness, which is taken into account. Other ways can be
considered to take friction into account, for example by adding a specific
6.3. Scope and limits of the approach term in the virtual work. The proposed approach makes it possible to
greatly reduce the cost of the calculations compared to an approach
The proposed approach is developed for fibrous reinforcements with where each layer is described by a layer of shell finite elements in
two directions of inextensible fibres in the plane of the reinforcement. contact with its neighbours. Some aspects remain to be studied and
This is an idealized situation. A real reinforcement, more or less deviates modelled, in particular the case of the different orientations of the fibre
from this situation. Simulations based on this approach will be more plies in the textile reinforcement and the consideration of thickness
relevant if one is close to this situation. Deformations of woven re­ variations during consolidation.
inforcements and stacks of woven reinforcements are simulated in this
paper with a correct agreement with the experiments (Figs. 6, 13, 15). CRediT authorship contribution statement
The modelling can be satisfactorily extended to more complex re­
inforcements, e.g. thick interlocks [29]. However, the deformation of Renzi Bai: Investigation, Methodology, Software, Validation, Visu­
some 3D reinforcements containing tows in the thickness direction alization. Julien Colmars: Investigation, Methodology, Supervision.
cannot be simulated with the proposed shell approach. Naim Naouar: Supervision. Philippe Boisse: Conceptualization,
Methodology, Supervision.
7. Conclusion

A shell approach specific to fibrous reinforcements is necessary Declaration of Competing Interest

because classical theories such as Kirchhoff and Mindlin are based on
kinematic assumptions that are not verified for textile reinforcements. The authors declare that they have no known competing financial
The deformation of these fibrous fabrics has two major features: the interests or personal relationships that could have appeared to influence
inextensibility of the fibres and the possible slippage between the fibres. the work reported in this paper.
A shell approach has been developed to take these specificities into ac­
count. An Ahmad shell finite element has been developed. This element Acknowledgements
has been validated on some 3D deformation tests where it has been
shown that it allows determining the displacement, the strains at all This work was supported by the French Ministry of Higher Education
points in the thickness and the rotation of the material normals in good and Research.

Appendix A

The lumped matrix for node k of the shell is given bellow, where k is the index number of node, ρ is the mass density of the element material, V is the
volume of the element [50].

12
R. Bai et al. Composites Part A 139 (2020) 106135

⎡ ⎤
mkk 0 0 0 0
⎢ 0 mkk 0 0 0 ⎥
⎢ ⎥
[Mk ] = ⎢
⎢ 0 0 mkk 0 0 ⎥⎥ (38)
⎣ 0 0 0 Ikk 0 ⎦
0 0 0 0 Ikk
∫ ∫
ρNi (ξ, η)T Ni (ξ, η)dV mii i 2
mii = ∑3 V ∫ ρdV and Iii = (h ) (39)
T
k=1 V ρNk (ξ, η) Nk (ξ, η)dV V 4 m

References [24] Yu F, Chen S, Viisainen JV, Sutcliffe MPF, Harper LT, Warrior NA. A macroscale
finite element approach for simulating the bending behaviour of biaxial fabrics.
Compos Sci Technol 2020;191.
[1] Arbter R, Beraud JM, Binetruy C, Bizet L, Bréard J, Comas-Cardona S, et al.
[25] Hamila N, Boisse P, Sabourin F, Brunet M. A semi-discrete shell finite element for
Experimental determination of the permeability of textiles: a benchmark exercise.
textile composite reinforcement forming simulation. Int J Numer Meth Eng 2009;
Compos A Appl Sci Manuf 2011;42(9):1157–68.
79(12):1443–66.
[2] Sozer EM, Simacek P, Advani SG. Resin transfer molding (RTM) in polymer matrix
[26] Schirmaier FJ, Dörr D, Henning F, Kärger L. A macroscopic approach to simulate
composites. In: Manufacturing techniques for polymer matrix composites (PMCs).
the forming behaviour of stitched unidirectional non-crimp fabrics (UD-NCF).
Woodhead Publishing; 2012. p. 245–309.
Compos A Appl Sci Manuf 2017;102:322–35.
[3] Caglar B, Orgéas L, Du Roscoat SR, Sozer EM, Michaud V. Permeability of textile
[27] Haanappel SP, Ten Thije RHW, Sachs U, Rietman B, Akkerman R. Formability
fabrics with spherical inclusions. Compos A Appl Sci Manuf 2017;99:1–14.
analyses of uni-directional and textile reinforced thermoplastics. Compos A Appl
[4] Lukaszewicz DJ, Potter KD. The internal structure and conformation of prepreg
Sci Manuf 2014;56:80–92.
with respect to reliable automated processing. Compos A Appl Sci Manuf 2011;42
[28] Dörr D, Henning F, Kärger L. Nonlinear hyperviscoelastic modelling of intra-ply
(3):283–92.
deformation behaviour in finite element forming simulation of continuously fibre-
[5] Guzman-Maldonado E, Hamila N, Boisse P, Bikard J. Thermomechanical analysis,
reinforced thermoplastics. Compos A Appl Sci Manuf 2018;109:585–96.
modelling and simulation of the forming of pre-impregnated thermoplastics
[29] Liang B, Colmars J, Boisse P. A shell formulation for fibrous reinforcement forming
composites. Compos A Appl Sci Manuf 2015;78:211–22.
simulations. Compos A Appl Sci Manuf 2017;100:81–96.
[6] Lessard H, Lebrun G, Benkaddour A, Pham XT. Influence of process parameters on
[30] Ahmad S, Irons BM, Zienkiewicz OC. Analysis of thick and thin shell structures by
the thermostamping of a [0/90] 12 carbon/polyether ether ketone laminate.
curved finite elements. Int J Numer Meth Eng 1970;2(3):419–51.
Compos A Appl Sci Manuf 2015;70:59–68.
[31] Dvorkin EN, Bathe KJ. A continuum mechanics based four-node shell element for
[7] Levy A, Hubert P. Vacuum-bagged composite laminate forming processes:
general non-linear analysis. Eng Computations 1984;1:77–88.
Predicting thickness deviation in complex shapes. Compos A Appl Sci Manuf 2019;
[32] Parisch H. A continuum-based shell theory for non-linear applications. Int J
126.
Numerical Methods Eng 1995;38(11):1855–83.
[8] Henning F, Kärger L, Dörr D, Schirmaier FJ, Seuffert J, Bernath A. Fast processing
[33] Auricchio F, Taylor RL. A triangular thick plate finite element with an exact thin
and continuous simulation of automotive structural composite components.
limit. Finite Elem Anal Des 1995;19(1–2):57–68.
Compos Sci Technol 2019;171:261–79.
[34] O.C. Zienkiewicz, R.L. Taylor, The finite element method for solid and structural
[9] Van Der Weeën F. Algorithms for draping fabrics on doubly-curved surfaces. Int J
mechanics, Elsevier, 2005.
Numerical Methods Eng 1991;31(7):1415–26.
[35] J.L. Batoz, G. Dhatt, Modélisation des structures par éléments finis: Volume 2 :
[10] Long AC, Rudd CD. A simulation of reinforcement deformation during the
Poutres et plaques, Presses Université Laval, 1990.
production of preforms for liquid moulding processes. Proc Inst Mech Engineers,
[36] Creech G, Pickett AK. Meso-modelling of non-crimp fabric composites for coupled
Part B: J Eng Manufacture 1994;208(4):269–78.
drape and failure analysis. J Mater Sci 2006;41(20):6725–36.
[11] Potluri P, Ciurezu DP, Ramgulam RB. Measurement of meso-scale shear
[37] Pazmino J, Mathieu S, Carvelli V, Boisse P, Lomov SV. Numerical modelling of
deformations for modelling textile composites. Compos A Appl Sci Manuf 2006;37
forming of a non-crimp 3D orthogonal weave E-glass composite reinforcement.
(2):303–14.
Compos A Appl Sci Manuf 2015;72:207–18.
[12] Hancock SG, Potter KD. The use of kinematic drape modelling to inform the hand
[38] Mathieu S, Hamila N, Bouillon F, Boisse P. Enhanced modeling of 3D composite
lay-up of complex composite components using woven reinforcements. Compos A
preform deformations taking into account local fiber bending stiffness. Compos Sci
Appl Sci Manuf 2006;37(3):413–22.
Technol 2015;117:322–33.
[13] Cherouat A, Billoët JL. Mechanical and numerical modelling of composite
[39] El Said B, Daghia F, Ivanov D, Hallett SR. An iterative multiscale modelling
manufacturing processes deep-drawing and laying-up of thin pre-impregnated
approach for nonlinear analysis of 3D composites. Int J Solids Struct 2018;132:
woven fabrics. J Mater Process Technol 2001;118(1–3):460–71.
42–58.
[14] Ten Thije RHW, Akkerman R, Huétink J. Large deformation simulation of
[40] Ten Thije RHW, Akkerman R. Solutions to intra-ply shear locking in finite element
anisotropic material using an updated Lagrangian finite element method. Comput
analyses of fibre reinforced materials. Compos A Appl Sci Manuf 2008;39(7):
Methods Appl Mech Eng 2007;196(33–34):3141–50.
1167–76.
[15] Lin H, Wang J, Long AC, Clifford MJ, Harrison P. Predictive modelling for
[41] Hamila N, Boisse P. Locking in simulation of composite reinforcement
optimization of textile composite forming. Compos Sci Technol 2007;67(15–16):
deformations. Analysis and treatment. Compos A Appl Sci Manuf 2013;53:109–17.
3242–52.
[42] Harrison P, Clifford MJ, Long AC. Shear characterisation of viscous woven textile
[16] Chen S, Harper LT, Endruweit A, Warrior NA. Formability optimisation of fabric
composites: a comparison between picture frame and bias extension experiments.
preforms by controlling material draw-in through in-plane constraints. Compos A
Compos Sci Technol 2004;64(10–11):1453–65.
Appl Sci Manuf 2015;76:10–9.
[43] Boisse P, Zouari B, Daniel JL. Importance of in-plane shear rigidity in finite element
[17] Chen S, McGregor OPL, Harper LT, Endruweit A, Warrior NA. Defect formation
analyses of woven fabric composite preforming. Compos A Appl Sci Manuf 2006;37
during preforming of a bi-axial non-crimp fabric with a pillar stitch pattern.
(12):2201–12.
Compos A Appl Sci Manuf 2016;91:156–67.
[44] Cao J, Akkerman R, Boisse P, Chen J, Cheng HS, De Graaf EF, Lee W.
[18] Boisse P, Hamila N, Vidal-Sallé E, Dumont F. Simulation of wrinkling during textile
Characterization of mechanical behavior of woven fabrics: experimental methods
composite reinforcement forming. Influence of tensile, in-plane shear and bending
and benchmark results. Compos A Appl Sci Manuf 2008;39(6):1037–53.
stiffnesses. Compos Sci Technol 2011;71(5):683–92.
[45] Na W, Ahn H, Han S, Harrison P, Park JK, Jeong E, Yu WR. Shear behavior of a
[19] Dangora LM, Mitchell CJ, Sherwood JA. Predictive model for the detection of out-
shear thickening fluid-impregnated aramid fabrics at high shear rate. Compos B
of-plane defects formed during textile-composite manufacture. Compos A Appl Sci
Eng 2016;97:162–75.
Manuf 2015;78:102–12.
[46] Hamila N, Boisse P. A meso–macro three node finite element for draping of textile
[20] Lomov SV, Truevtzev AV, Cassidy C. A predictive model for the fabric-to-yarn
composite preforms. Appl Compos Mater 2007;14(4):235–50.
bending stiffness ratio of a plain-woven set fabric. Text Res J 2000;70(12):
[47] T. Belytschko, An overview of semidiscretization and time integration procedures,
1088–96.
Computational methods for transient analysis(A 84-29160 12-64), Amsterdam,
[21] Yu WR, Zampaloni M, Pourboghrat F, Chung K, Kang TJ. Analysis of flexible
North-Holland, 1983, pp. 1-65.
bending behavior of woven preform using non-orthogonal constitutive equation.
[48] Hughes TJR, Belytschko T. A precis of developments in computational methods for
Compos A Appl Sci Manuf 2005;36(6):839–50.
transient analysis. ATJAM 1983;50:1033–41.
[22] Döbrich O, Gereke T, Diestel O, Krzywinski S, Cherif C. Decoupling the bending
[49] M.A. Crisfield, Non-linear Finite Element Analysis of Solids and Structures:
behavior and the membrane properties of finite shell elements for a correct
Advanced Topics. Bol. 2, John Wiley & Sons, 1997.
description of the mechanical behavior of textiles with a laminate formulation.
[50] Surana KS. Lumped mass matrices with non− zero inertia for general shell and
J Ind Text 2014;44(1):70–84.
axisymmetric shell elements. Int J Numer Meth Eng 1978;12(11):1635–50.
[23] Nishi M, Hirashima T, Kurashiki T. Dry fabric forming analysis considering the
[51] Onate E, Zárate F. Rotation-free triangular plate and shell elements. Int J
influence of tensions on in-plane shear behavior. J Soc Mater Sci, Japan 2014;63
Numerical Methods Eng 2000;47(1–3):557–603.
(5):380–5.

13
R. Bai et al. Composites Part A 139 (2020) 106135

[52] Sabourin F, Brunet M. Detailed formulation of the rotation-free triangular element [58] Guzman-Maldonado E, Wang P, Hamila N, Boisse P. Experimental and numerical
“S3” for general purpose shell analysis. Eng Computations 2006. analysis of wrinkling during forming of multi-layered textile composites. Compos
[53] Brunet M, Sabourin F. Analysis of a rotation-free 4-node shell element. Int J Struct 2019;208:213–23.
Numerical Methods Eng 2006;66(9):1483–510. [59] Chakladar ND, Mandal P, Potluri P. Effects of inter-tow angle and tow size on
[54] Boisse P, Colmars J, Hamila N, Naouar N, Steer Q. Bending and wrinkling of carbon fibre friction. Compos A Appl Sci Manuf 2014;65:115–24.
composite fiber preforms and prepregs. A review and new developments in the [60] De Bilbao E, Soulat D, Hivet G, Gasser A. Experimental study of bending behaviour
draping simulations. Compos B Eng 2018;141:234–49. of reinforcements. Exp Mech 2010;50(3):333–51.
[55] Naouar N, Vidal-Sallé E, Schneider J, Maire E, Boisse P. Meso-scale FE analyses of [61] Allaoui S, Boisse P, Chatel S, Hamila N, Hivet G, Soulat D, Vidal-Salle E.
textile composite reinforcement deformation based on X-ray computed Experimental and numerical analyses of textile reinforcement forming of a
tomography. Compos Struct 2014;116:165–76. tetrahedral shape. Compos A Appl Sci Manuf 2011;42(6):612–22.
[56] Vanclooster K, Lomov SV, Verpoest I. Simulation of multi-layered composites [62] Gherissi A, Abbassi F, Ammar A, Zghal A. Numerical and experimental
forming. Int J Mater Form 2010;3(1):695–8. investigations on deep drawing of G1151 carbon fiber woven composites. Appl
[57] Fetfatsidis KA, Jauffrès D, Sherwood JA, Chen J. Characterization of the tool/fabric Compos Mater 2016;23(3):461–76.
and fabric/fabric friction for woven-fabric composites during the thermostamping
process. Int J Mater Form 2013;6(2):209–21.

14