Professional Documents
Culture Documents
Experimental Analysis and Numerical Modelling of Dry Carbon Woven Reinforcement Preforming
Experimental Analysis and Numerical Modelling of Dry Carbon Woven Reinforcement Preforming
Abstract
In this paper, an experimental and numerical study of the preforming process of the G1151 carbon woven fabric
reinforcement is presented. The experimental analysis was based on tensile and shear tests. These tests were important
to analyse the behaviour of this particular woven reinforcement and to figure out some key phenomena related to its
deformation process. The numerical modelling is implemented in the commercial FEM software (Abaqus) using a
discrete finite element approach. The model is built using the concept of a “unit cell” formulated with a Hencky
linear elastic shell/membrane elements coupled with axial connectors. The connectors replace bar and beam elements
used in previous works and can have a linear or a non-linear behaviour. Shell finite elements are chosen to describe the
in-plane shear stiffness and to manage contact phenomena. The model parameters identification technique is based on
experimental constitutive tests and an inverse optimisation procedure. The model has been experimentally validated for
the case of hemispherical single layer preforming of the G1151 woven fabric.
Keywords
Woven carbon fabric, forming simulation, G1151 reinforcement
Material characterization
The material studied in this work is the carbon fibre
woven reinforcement G1151 (Figure 1). This reinforce-
ment, which is produced by HexcelVR 19–23 is a thin three-
layered powdered interlock in which the thickness is
relatively small (lesser than 1 mm). It is used in the
aeronautical industry.
A previous article by the authors has investigated
the frictional behaviour of this specific reinforcement.20
Figure 2. Response of G1151 single yarn in tension.
In the current paper, the behaviour of the G1151 rein-
forcement is experimentally analysed, and an emphasis
is given to the study of the shear behaviour of this
textile. This analysis leads to a phenomenological
description of G1151 interlock reinforcement behav-
iour. A numerical study of this reinforcement preform-
ing has been then carried out using a model which has
been previously introduced by the authors.19,21,22
Tensile analysis
Behaviour of a yarn. The first test which has been per-
formed is a tensile test of a single yarn of G1151, the
length of this yarn was 200 mm. According to the sup-
plier catalogue, the yarn is 6K.23 The result of this test
is presented in Figure 2. Figure 3. Fabric directions.
The curve shows two sectors, the first one is linear; it
starts from the beginning till a strain percentage equal Uniaxial tensile test of the G1151 fabric. Tensile tests of a
to 0.7%, and it describes the elastic extension of the specimen of G1151 have been done in both weft and
carbon fibre yarn. The second sector is characterized by warp direction (Figure 3).
the damage initiation and the stiffness loss of the yarn. A Zwick tensile test machine equipped by a force
The curve illustrates the very low ductility and the brit- sensor of 100 KN is used to perform the tests. To
tle character of the used carbon fibre yarns. avoid sliding during test, the fabric reinforcement
Najjar et al. 3
Figure 11. Repeatability of the force versus displacement Modelling strategy and choices
curves in bias test for the 25 mm/min velocity. To meet the objectives of the established model (accu-
racy and simplicity), the developments have been per-
formed within the concept of the discrete approach.
This approach was initially introduced by Cherouat
and Billo€et26 and Sidhu et al.27 The works of Sharma
et al.28 and Skordos et al.15 were remarkable in this
field. Some important developments were also pro-
posed by the team of Sherwood.16,29,30 Harrison pro-
posed a version of this model which takes into account
the bending stiffness of the reinforcement.31
This family of models has these advantages:
beam elements have been used to model the 1D tensile (Force ¼ f (displacement)) behaviour of the connec-
entities, and a material behaviour and inertial cross tor has to be determined.
section characteristics have to be identified, to describe • Reduced calculation cost. Indeed the connector is
correctly the mechanical responses and represent the not an actual finite element which the response has
tension stiffness of these 1D elements. to be calculated in the finite element solving process
The software Abaqus offers a comprehensive library but an additional constraint between the corner
of special elements to model connections. These con- nodes of the membrane element.
nectors give the possibility to apply kinematic or kinet- • The use of connectors allows simulating the forming
ic constraint, and to define specific behaviour of of several layers. Connectors do not involve in the
interaction between two nodes.32 interface contact, only the (shell/membrane) ele-
An axial connector (Figure 14) is used to define the ments will be involved in the contact calculation
axial force–displacement relationship between the two
nodes. The behaviour of the connector can be also Finally, it is important to mention that in the pro-
complex to describe physical non-linearities. This ele- posed approach, the mesh using the hybrid element has
ment type is used to materialize the 1D element in the to be aligned with the fibre directions.
developed model.
This choice has many advantages:
Analysis of unit cell behaviour
• Using the connectors, only few parameters In this section, the proposed hybrid unit cell mechani-
have to be identified. Actually only the local cal behaviour is analytically analysed in both tension
Najjar et al. 7
The unit cell contains two connectors aligned with The total force can be then expressed as a function
the direction of loading. The load taken by connectors of the extension Dl and material parameters
is then:
E t0
Ftotal ¼ ð2:K0 þ ÞDl (10)
Fconn ¼ 2K0 Dl (3) ð1 2 Þ
In the other hand, the plane stress state of the elastic The ratio between the total force and the force car-
membrane can be written in case of infinitesimal strain ried out by the connectors can also be calculated
as follows
Ftotal
2 3 ¼1þa (11)
2 3 2 3 Fconn
rxx 1 0
6 7 exx
6 7 E 6 1 0 7 6 7
4 ryy 5 ¼ :6 7:4 eyy 5 (4) with
ð1 2 Þ 4 ð1 Þ 5
rxy 0 0 2exy
2 E:t0
a¼ (12)
K0 2 ð1 2 Þ
Then
where a is the contribution of the membrane in the
E E
ryy ¼ exx þ eyy (5) general tensile stiffness of the hybrid element.
ð1 Þ
2 ð1 2 Þ It depends mainly on the ratio Et K0 . This ratio is,
0
hypothesis (i). This statement will be confirmed later on Let’s consider c as a shear angle, and c can be
when E; t and K0 values will be identified. The factor expressed as follows
a is then small too; the ratio FFconn
total
is then almost equal to
1 which means that the tension stiffness can be consid- p
c¼ h (13)
ered as totally carried out by the connectors. 2
Figure 16. The unit cell shear analysis. (a) The unit cell in its actual environment. (b) Stress state of the unit cell and the different
coordinate systems which can be affected to the element. (c) Simplified equivalent model which the behaviour can be expressed as
F ¼ f(d).
Najjar et al. 9
where G ¼ 2ð1þÞ
E ðshear modulusÞ and cL the
logarithmic-Hencky finite shear strain which is
expressed as
p c Figure 17. Simulation of unit cell in pure shear.
c ¼ ln tan
L
(21)
4 2
Then
0rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
pffiffi 2
B 1 dþ2l 2l0 C
B C
r12 ðdÞ ¼ G ln@ A
0
pffiffi (23)
dþ 2l0
2l0
made by a parallel association of four springs. Kfabric are b ¼ 48 mmL ¼ 200 mm. The direction of the test
can be written as a function of Kspring was the weft direction. The result of this test is shown
X
nþ1 in Figure 22. Kfabric value calculated from the curve is
Kfabric ¼ Kspringi ¼ ðn þ 1ÞKspring (25) equal to 5508 N/mm.
i¼1 It is then possible to identify the stiffness value of
On the other hand, the spring (red line) is formed by the connector as a function of the selected mesh size
N serial connectors in which the elementary stiffness is using equation (28).
equal to K0 . As a function of K0 ; Kspring can be written as
Case of non-linear connector
K0
Kspring ¼ (26) The first approach considered a linear behaviour of the
N
fabric in tension. However, the actual tensile response of
According to equations (25) and (26), K0 can be then the reinforcement shows a significant non-linearity at
calculated as follows the start of loading. To account for this phenomenon,
N a non-linear (polynomial) model of the behaviour of
K0 ¼ Kfabric (27) connectors is used to describe the response of the fabric.
ðn þ 1Þ
As a function of the specimen dimensions and the
connector initial length l0 , K0 can be expressed as
Identification of the global behaviour
(specimen behaviour)
ðlL0 Þ The overall behaviour of the tensile test can be defined
K0 ¼ Kfabric (28)
ðlb0 þ 1Þ by two parts (Figure 24):
The tensile test used for the identification is the one • The first sector is associated with a polynomial
h i
presented in Figure 4, where the specimen dimensions defined within the interval 0; 4DL
5 mm .
• The second sector is a straight line, which starts at a
displacement equal to 4DL
5 mm.
a4 a3 a2 a1 a0
a01 a00
5:5E3 5:78E3
8
> a4 a3 a2
>
> Flocal ¼ ðNdlocal Þ4 þ ðN dlocal Þ3 þ ðN dlocal Þ2
>
> n þ 1 n þ 1 n þ 1 4Dl
>
< for dlocal 2 0; mm
a1 a0 5N
þ ðN dlocal Þ1 þ (33)
>
> nþ1 nþ1
>
> 0 0
>
> a1 a 4Dl
: Flocal ¼ ðNdlocal Þ þ 0 for dlocal > mm
nþ1 nþ1 5N
8
>
> ð3:22E2Þ ð1:08E3Þ ð7:90E2Þ
>
> Flocal ¼ ðN dlocal Þ4 þ ðNdlocal Þ3 þ ðNdlocal Þ2
>
> n þ 1 n þ 1 n þ 1 2
>
< dlocal 2 0;
ð1:85E2Þ ð3:13E1Þ N
þ ðNdlocal Þ þ (34)
>
> nþ1 nþ1
>
>
>
> ð 5:5E3 Þ ð5:78E3Þ 2
>
: Flocal ¼ ðNdlocal Þ ; dlocal >
nþ1 nþ1 N
Najjar et al. 13
(
Flocal ¼ 1:288E9d4local þ 4:32E7 d3local þ 3:16E5 d2local þ 7:4E2dlocal þ 1:25; dlocal 2 ½0; 0:02 mm
(35)
Flocal ¼ 2:2E4 dlocal 1:82 E2; dlocal > 0:02 mm
16 Journal of Reinforced Plastics and Composites 0(0)
Figure 39. Experimental results. Figure 41. Experimental result (Measured angles).
Figure 47. Zoom of the surrounded area in Figure 46. with Gong et al38 which demonstrated that the cou-
pling tension-shear has a little effect on the draw-in.
The shape of the preform oriented 45 directions is
also studied. Figures 48 and 49 show the simulated and Conclusion
the experimental profiles. These figures illustrate that
the simulations predict well the global experimen- In this paper, an experimental and numerical analysis
tal profiles. of the preforming of woven carbon fibre has been con-
However, The zoom of the surrounded section in ducted. To better understand the behaviour of the spe-
Figure 48 (Figure 49) show, in coherence with the result cific reinforcement (G1151), some constitutive tests
in the case of the (0, 90 ) oriented perform, that the have been realized, and it was particularly interesting
non-linear connector model gives a smoother profile to show the effect of the shear deformation rate on the
which is closer to the experimental profile than the deformability of this fabric. A numerical hybrid model
linear predicted profile. which was previously introduced19–21 has been used to
Although these results show that the non-linear simulate the preforming operation, the results are then
model generates curves slightly closer locally to the compared to experimental data, and a good agreement
experimental profiles, it seems that there is no signifi- has been found. The tensile stiffness of reinforcement
cant impact of the non-linear connector on the global has been modelled in two ways, the first considering in
draw-in shapes, and this conclusion is in agreement a linear stiffness, and the second uses non-linear
20 Journal of Reinforced Plastics and Composites 0(0)
connectors to model the un-crimping of the fabric. 10. Zhang W, et al. A non-orthogonal material model of
Results clearly show that the non-linear woven composites in the preforming process. CIRP Ann
connectors have an impact on the shear angles distri- 2017; 60: 257–260.
bution, but a little effect on the shape and the draw-in 11. Charmetant A, Vidal-Sallé E and Boisse E. Hyperelastic
modelling for mesoscopic analyses of composite rein-
of the preformed part.
forcements. Compos Sci Technol 2011; 71: 1623–1631.
12. Liang B, Colmars J and Boisse P. A shell formulation for
Acknowledgements fibrous reinforcement forming simulations, Composites
Authors would like to thank: OSEO for its financial support, Part A 2017; 100: 81–96.
Airbus-France, LoireTech and EADS IW for provid- 13. Khan MA, Mabrouki T, Vidal-Sallé E, et al. Numerical
ed facilities. and experimental analyses of woven composite reinforce-
ment forming using a hypoelastic behaviour. Application
Declaration of conflicting interests to the double dome benchmark. J Mater Process Technol
2010; 210: 378–388.
The author(s) declared no potential conflicts of interest with
14. Denis Y, Guzman-Maldonado E, Hamila N, et al.
respect to the research, authorship, and/or publication of A dissipative constitutive model for woven
this article. composite fabric under large strain. Composites Part A
2018; 105: 165–179.
Funding 15. Skordos AA, Monroy Aceves C and Sutcliffe MPF.
The author(s) disclosed receipt of the following financial A simplified rate dependent model of forming and wrin-
kling of preimpregnated woven composites. Composites
support for the research, authorship, and/or publication of
Part A 2007; 38: 1318–1330.
thisarticle: This work has been financially supported by
16. Jauffrès D, Sherwood JA, Morris CD, et al. Discrete
OSEO within the frame of Défi Composite project. mesoscopic modeling for the simulation of woven-fabric
reinforcement forming. Int J Mater Form 2010;
ORCID iD 3: 1205–1216.
W Najjar https://orcid.org/0000-0003-1661-5264 17. D’Agostino MV, Giorgio I, et al. Continuum and
D Soulat https://orcid.org/0000-0001-7911-4243 discrete models for structures including (quasi-) inexten-
sible elasticae with a view to the design and modeling of
composite reinforcements. Int J Solids Struct
References
2015; 59:1–17.
1. Potter KD. Resin transfer moulding. London: Chapman 18. Hamila N, Boisse P, Sabourin F, et al. A semi-discrete
& Hall, 1997. shell finite element for textile composite reinforcement
2. Kruckenberg TM and Paton R (eds). Resin transfer forming simulation. Int J Numer Meth Eng 2009;
moulding for aerospace structures. Dordrecht: Kluwer 79: 1443–1466.
Academic Publishers, 1998. 19. Najjar W. Contribution to the simulation of composite
3. Parnas RS. Liquid composite moulding. Berlin: Hanser woven fabric reinforcements forming (in French). PhD
Gardner Publications, 2000. Dissertation, Arts et Métiers Paristech CER Angers,
4. Badel P, Vidal-Salle E and Boisse P. Computational France, 2012.
determination of in-plane shear mechanical behavior of 20. Najjar W, Legrand X, Pupin C, Boude S, et al. A simple
textile composite reinforcements. Comput Mater Sci discrete method for the simulation of the preforming
2007; 40: 439–448. of woven fabric reinforcement. Key Eng Mater 2012;
5. Bodaghi M, Vanaerschot A, Lomov SV, et al. On the 504–506: 213–218.
stochastic variations of intra-tow permeability induced 21. Najjar W, Legrand X, Dal Santo P, et al. Analysis of the
by internal geometry variability in a 2/2 twill carbon blank holder force effect on the preforming process using
fabric. Compos Part A 2017; 101: 444–458. a simple discrete approach. Key Eng Mater 2013; 554–
6. Arbter R, et al. Experimental determination of the per- 557: 441–446.
meability of textiles: a benchmark exercise. Composites 22. Najjar W, Pupin C, Legrand X, et al. Analysis of fric-
Part A 2011; 42: 1157–1168. tional behaviour of carbon dry woven reinforcement.
7. Vernet N, et al. Experimental determination of the per- J Reinf Plast Compos 2014; 33: 1037–1047.
meability of engineering textiles: benchmark II. 23. Hexel fabrics data sheets website, http://hexply.com/hex
Composites Part A 2014; 61: 172–184. force/database/web/front/main/index.php (accessed 11
8. Boisse P. Finite element analysis of composite forming. June 2019).
In: Long AC (ed) Composite forming technologies. UK: 24. Allaoui S, Boisse P, et al. Experimental and numerical
Woodhead Publishing Limited and CRC press analyses of textile reinforcement forming of tetrahedral
LLC, 2007. shape. Composite Part A 2011; 42: 612–622.
9. Xue P, Peng X and Cao J. A non-orthogonal constitutive 25. De Bilbao E, Soulat D, Hivet G, et al. Experimental
model for characterizing woven composites. Composites study of bending behaviour of reinforcements. Exp
Part A 2003; 34: 183–193. Mech 2010; 50: 333–351.
Najjar et al. 21
26. Cherouat A and Billo€et JL. Impregnated woven fabric by 34. Launay J, Hivet G, Duong A.V, et al. Experimental anal-
deep drawing and laying up processes. J Adv Mater 2000; ysis of the influence of tensions on in plane shear behav-
32: 42–53. iour of woven composite reinforcements. Compos Sci
27. Sidhu RMJS, Averill RC, Riaz M, et al. Finite element Technol 2008; 68: 506–515.
analysis of textile composite preforms stamping. Compos 35. Nosrat-Nezami F, Gereke T, Eberdt C, et al.
Struct 2001; 52: 483–497. Characterisation of the shear–tension coupling of
28. Sharma SB and Sutcliffe MPF. A simplified finite element carbon-fibre fabric under controlled membrane tensions
model for draping of woven material. Composites Part A for precise simulative predictions of industrial preforming
2004; 35: 637–643. processes. Composites Part A 2014; 67: 131–139.
29. Li X, Sherwood J, Liu L, et al. A material model for 36. Komeili M. Multi-scale characterization of shear-tension
woven commingled glass-propylene composite using a interaction in woven fabrics for composite forming and
hybrid finite element approach. Int J Mater Product applications. PhD Dissertation, University of British
Technol 2004; 21: 59–70. Columbia, Vancouver, Canada, 2014.
30. Jauffres D, Morris CD, Kremer J, et al. Simulation of the 37. Haghi Kashania M, Hosseini A, Sassani F, et al.
thermostamping of woven composites: mesoscopic Understanding different types of coupling in mechanical
modeling using FEA explicit codes. Int J Mater Form behaviour of woven fabric reinforcements: a critical
2009; 2: 173–176. review and analysis. Compos Struct 2017; 179: 558–579.
31. Harrison P. Modelling the forming mechanics of engi- 38. Gong Y, et al. An anisotropic hyperelastic constitutive
neering fabrics using a mutually constrained pantograph- model with tension–shear coupling for woven composite
ic beam and membrane mesh. Composites Part A 2016; reinforcements. Int J Appl Mech 2017; 9:175083
81: 145–157. (17 pages).
32. Dassault Systémes. Abaqus analysis user’s manual version 39. Xiao H and Chen L.S. Hencky’s elasticity model and
6.12. Providence, RI, USA , 2012. linear stress-strain relations in isotropic finite hyperelas-
33. Komeili M and Milani AS. On the effect of shear-tension
ticity. Acta Mech 2002; 157: 51–60.
coupling in forming simulation of woven fabric reinforce-
ments. Composites Part B 2016; 99: 17–29.