Composite Structures 265 (2021) 113692

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Review

Analysis of the hygroscopic and hygroelastic behaviours of water

aged flax-epoxy composite
Belahcen Djellouli ⇑, Wajdi Zouari, Mustapha Assarar, Rezak Ayad
University of Reims Champagne Ardenne, ITheMM EA 7548, 51097 Reims, France

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

Keywords: In this paper, an analysis of the hygroscopic and hygroelastic behaviours of unidirectional flax fibre‐reinforced
Flax‐epoxy composite epoxy composite is presented. Samples of this composite material were elaborated by the vacuum infusion pro-
Water ageing cess, then cut and prepared to be aged in tap water until saturation. To describe their hygroscopic behaviour,
Fick’s model Fick’s model is considered and an optimisation procedure is developed to identify their macroscopic water dif-
Hygroscopic and hygroelastic finite element
fusion parameters. After that, a plane finite element modelling of the hygroscopic behaviour of the aged flax‐
analysis
Internal stress
epoxy samples is conducted to identify the flax fibre radial and longitudinal water diffusion parameters using
an inverse approach. Finally, a plane finite element analysis of their hygroelastic behaviour is proposed allow-
ing the estimation of the hygroscopic internal stress in flax fibres and epoxy matrix. In particular, the hygro-
scopic analysis reveals that water diffusion is more pronounced in the direction of flax fibres compared to
the transverse and thickness directions. In addition, the hygroelastic finite element modelling highlights the
high compressive and tensile stress subjected to flax fibres and epoxy matrix, respectively, due to water
absorption.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. Materials and manufacturing process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. Conditioning of the flax-epoxy samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3. Scanning electronic microscopy observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4. Fick’s analytical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Finite element analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.1. Description of the plane modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2. Hygroscopic and hygroelastic finite element approximations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.1. Water diffusion parameters of the sealed samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.2. 3D water diffusion parameters of the flax-epoxy specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.3. Finite element analysis of the hygroscopic behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.4. Finite element analysis of the hygroelastic behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Declaration of Competing Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

⇑ Corresponding author.
E-mail address: belahcen.djellouli@etudiant.univ-reims.fr (B. Djellouli).

https://doi.org/10.1016/j.compstruct.2021.113692
Received 10 July 2020; Revised 23 December 2020; Accepted 2 February 2021
Available online 12 February 2021
0263-8223/© 2021 Elsevier Ltd. All rights reserved.
B. Djellouli et al. Composite Structures 265 (2021) 113692

1. Introduction mens increases linearly with fibre orientation and decreases with
thickness.
Natural fibre reinforced polymer composites gain ground in sev- Usually, the analytical models can only describe moisture diffusion
eral sectors such as automotive and construction. This increased use in composite materials at the macroscopic scale. In addition, these ana-
of these materials is motivated by the interesting specific mechani- lytical models can’t predict the diffusion behaviour of composite con-
cal properties of natural fibres as well as their several advantages stituents such as the moisture diffusion properties of natural fibres,
related particularly to the environmental impacts [1–7] Indeed, nat- which are relatively difficult to measure experimentally. The use of a
ural fibres are renewable resources, biodegradable, neutral in terms numerical finite element model can be considered as an alternative
of CO2 emissions and require relatively low energy to be produced to overcome the experimental difficulties, but the literature reveals a
[2]. However, the anisotropic nature of natural fibres and their high limited number of contributions dealing with the numerical modelling
sensitivity to the surrounding environment conditions, especially of moisture diffusion in natural fibre‐reinforced composites. Among
humidity (a natural fibre swells and shrinks in presence of moisture) these contributions, Regazzi et al. [28] have modelled water diffusion
and temperature, constitute a major barrier to their wide develop- in Polylactic acid (PLA) composite specimens reinforced with dog
ment. Accordingly, many works have investigated the effect of bone‐shaped short flax fibres. The numerical model has considered
humidity on the mechanical and dynamic characteristics of natural the PLA‐flax composite samples as homogeneous materials. Homoge-
fibre‐reinforced polymer composites [8–19]. They have principally neous numerical modelling has also been considered by Péron et al.
shown that the elastic and failure properties of these materials sig- [38] to describe moisture diffusion and hygroscopic swelling in flax‐
nificantly vary with the surrounding environment, which conse- polypropylene composite. Contrary to these homogeneous finite ele-
quently reduces their durability. ment models, Chilali et al. [39] have considered the heterogeneity of
In order to promote the use of natural fibre‐reinforced polymer twill flax fabric‐reinforced epoxy composites by using two geometric
materials in environmental conditions, it is essential to understand [0/90] and elliptical models to describe the waviness of the flax fab-
their moisture diffusion kinetics. Some physical models that can rics. The authors have estimated the radial water diffusion coefficient
describe this phenomenon, particularly Fick’s and Langmuir’s mod- of the flax fibre by modelling the hygroscopic behaviour of the aged
els [20,21], have been considered in the literature. For natural flax‐epoxy samples. In another study, Zouari et al. [40] have devel-
fibre‐reinforced composites, the one‐dimensional (1D) Fick’s model oped a three‐node triangular plane finite element to analyse the hygro-
has been mainly used to identify their moisture diffusion parameters scopic behaviour of 2/2 twill flax fabric‐reinforced epoxy composite
[6,11,14,18,22–26]. For example, in the work by Christian and aged in tap water at room temperature. To take account of the hetero-
Billington [22], the approximation of Shen and Springer has been geneity of the flax‐epoxy composite in the finite element model, they
considered to calculate the 1D diffusion parameters of hemp/cellu- have considered a sinusoidal description of the flax fabric undulation.
lose acetate (hemp‐cellulose) and hemp/poly (β‐hydroxybutyrate) They have shown that water diffusion coefficients of the epoxy matrix
composites (hemp‐PHB) under various moisture and temperature and the flax fibre follow an exponential evolution as a function of the
conditions. In particular, these authors have found that the moisture composite thickness.
diffusion parameters of hemp‐cellulose and hemp‐PHB composites, In addition to the hygroscopic behaviour, the finite element analy-
hygrothermal aged at 30 °C and 100% relative humidity (RH), are sis has been also considered to estimate the internal stress induced by
respectively 20 and 10 times that of glass‐polyester composite. moisture absorption especially in synthetic fibre‐reinforced composites
Recently, Cheour et al. [26] have analysed the impact of water age- [41–47]. The number of contributions dealing with the estimation of
ing on the free vibration behaviour of flax‐glass hybrid composites. internal hygroscopic stress in natural fibre‐reinforced composite
The authors have supposed that water diffusion is carried out exclu- remains limited in the literature [38,39]. Péron et al. [38] have devel-
sively along the thickness direction of the composite samples and oped a hydromechanical model based on the finite difference method,
used Fick’s 1D model to identify their diffusion parameters. They where the diffusion problem is first solved, which allows determining
have found that the position of the flax layers in the hybrid lami- the distribution of water content in unidirectional flax‐polypropylene
nates directly affects the values of the flexural modulus, the loss fac- laminate samples. Then, the mechanical problem is solved by taking
tor and the mass gain at saturation. In addition, the flexural into account the hygroscopic swelling due to the previously calculated
modulus of the aged flax‐glass hybrid composite is found 4.5 times water content. They have found that most of the plies (about 75% of
higher than that of the aged flax composite. the entire laminate) are subjected to compressive stress across the
Moisture diffusion in natural fibre‐reinforced composite materials transverse direction of the flax fibres. Chilali et al. [39] have consid-
depends on various parameters among others the temperature ered the temperature‐displacement analogy in their finite element
[22,27,28], the type of natural fibres and their morphology [29,30]. model to estimate the internal stress induced by hygroscopic swelling
These parameters indicate that moisture diffusion is a rather complex in flax‐epoxy samples. In particular, they have shown the high stress
phenomenon which could not be only described by 1D Fick’s model. concentration at the fibre/matrix interface induced by the dissimilar
For this purpose, it is necessary to consider a three‐dimensional (3D) swelling between the flax fibres and the epoxy matrix. Stress concen-
analysis that allows describing moisture diffusion in the three direc- tration peaks reaching 300 MPa have been estimated which greatly
tions. 3D moisture diffusion analyses have been mainly considered exceed the ultimate tensile strength of the flax‐epoxy specimens.
in synthetic fibre‐reinforced polymer composites [31–37]. However, Most of the above‐cited works have been interested only in the esti-
only a few works have studied moisture diffusion in natural fibre‐ mation of the flax fibre radial moisture diffusion coefficient. Moreover,
reinforced composites by considering 3D diffusion models. For exam- some of these studies have considered the temperature‐displacement
ple, Saidane et al. [8] have shown that 3D Fick’s model accurately analogy to estimate the hygroscopic internal stress. In the continuation
describes water diffusion kinetics in flax‐epoxy composite. Their of these works, an estimation of the flax fibre radial and longitudinal
results have also shown that water diffusion mainly occurs in the direc- moisture diffusion coefficients is given in this contribution by combin-
tion of the flax fibres and their morphology has an impact on the com- ing numerical and experimental results of water aged flax‐epoxy spec-
posite behaviour. Recently, Chilali et al. [30] have assessed the main imens. In addition to that, a three‐node triangular hygroelastic
3D water diffusion parameters of twill flax fabric‐reinforced thermo- membrane finite element is proposed to analyse the hygroelastic beha-
plastic and thermoset composites, aged in tap water until saturation, viour of the studied materials. This finite element development allows
by considering Fick’s and Langmuir’s 3D models. In particular, they avoiding the use of the temperature‐displacement analogy which over-
have found that the mass at saturation of the aged composite speci- estimates the hygroscopic internal stress.

2
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 1. Schematic of the vacuum infusion process.

Fig. 2. Conditioning of the flax-epoxy samples: (a) unsealed and (b)-(d) sealed samples.

2. Materials and methods During the ageing tests, the flax‐epoxy samples were regularly
weighed using a
1 mg precision balance. Before each mass measure-
2.1. Materials and manufacturing process ment, each sample was dried with a dry wipe to remove water from its
surfaces. The amount of water absorbed after an immersion time t was
This study is devoted to the study of the hygroscopic behaviour of a calculated with the following expression:
unidirectional (UD) flax fibre‐reinforced epoxy composite. The UD flax
Wt  W0
fibres are commercialised by Lineo Company under the commercial M t ð%Þ ¼  100 ð1Þ
W0
name FlaxTape™ 200. They present an aerial weight of 200 g=m2
and a mass density of 1450 kg=m3 . where W 0 is the mass of the dry sample and W t is its mass at time t.
The epoxy resin is the SR8100 and was provided with its associated
hardener SD 8823 by Sicomin. The mass ratio between the hardener
2.3. Scanning electronic microscopy observations
and the resin is 26:100 g. Several flax‐epoxy composite plates com-
posed of six layers of UD flax fibres were prepared by the vacuum infu-
Scanning electron microscopy (SEM) observations were performed
sion process as shown in Fig. 1. The vacuum pressure used in the
on the flax‐epoxy samples in the dry state and at saturation using a
manufacturing process was fixed to 0.7 bar. The infused plates were
“JEOL JSM‐7900F” scanning electron microscope operating at 1 kV.
first cured for 24 h at room temperature and then polymerised at
The flax‐epoxy samples were cut and embedded in a thermoplastic
60 °C and 10% of RH in a climate chamber for 24 h as recommended
resin and all the observed surfaces were polished using a double disc
by the purveyor.
polishing machine. Before the microscopic observations, the flax‐
epoxy samples were dried in an oven at 40 °C to remove water mole-
2.2. Conditioning of the flax-epoxy samples
cules and facilitate SEM observations.
The flax‐epoxy composite plates were cut to obtain square samples
with 25 mm side length. These samples were then polished to have 2.4. Fick’s analytical model
planar and even surfaces. All samples were dried at 60 °C for five days
to remove any trace of water. Several research works have shown that moisture uptake curves of
The thickness and mass of the flax‐epoxy samples were measured in natural fibre‐reinforced polymer composites generally exhibit Fickian
the dry state and are equal to 2.68 ± 0.10 mm and 1.99 ± 0.10 g behaviour [1,8,19,26,30,38]. Fick's law is described by the following
respectively. Their fibre volume fraction and their porosity rate are differential equation which allows to predict water diffusion as a func-
equal to 33.05 ± 1.78% and 4.64 ± 1.76%, respectively. tion of time and space [20]:
The ageing of the square flax‐epoxy samples consisted in totally
@c
immersing them in tap water at room temperature until saturation. ¼ divðD:rcÞ ð2Þ
@t
As shown in Fig. 2 some of the composite samples were sealed with
a waterproof paint to force water diffusion along the principal direc- where c is the moisture concentration and D is the diffusion tensor
tions x, y and z where x is the direction of flax fibres, y is the transverse which is supposed symmetric in this work.
(or width) direction and z is the thickness direction. In our case, x, y and z are the principal directions and Fick's law can
The main objective is to assess the principal water diffusion param- be rewritten as follows if water diffusion parameters are supposed
eters Dx, Dy and Dz as well as the saturation mass gains Mx, My and Mz. constant:

3
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 3. (a) A square flax-epoxy specimen. (b)–(c) Plane modelling of water diffusion in the yz plane using (b) regular and (c) random distributions of flax fibres. (d)
Plane modelling of water diffusion in the xz plane.

n  2
@c @2c @2c @2c q ¼ ∑ M a ðt k Þ  M exp ðt k Þ ð6Þ
¼ Dx 2 þ Dy 2 þ Dz 2 ð3Þ
@t @x @y @z i¼1

The solution of Eq. (3) reads [48,49]: where M a ðt k Þ is the moisture rate at time t k evaluated from the analyt-
   ical solution of Fick’s model (Eqs. (4) and (5)) and M exp ðt k Þ the experi-

3 1 1 1 exp π 2 t D 2iþ12 þ D 2jþ12 þ D 2kþ12 mental moisture content at time t k .
Mt 8 x L y L z h
¼ 1 2 ∑ ∑ ∑
M1 π i¼0 j¼0 k¼0 ðð2i þ 1Þð2j þ 1Þð2k þ 1ÞÞ2
3. Finite element analysis
ð4Þ

where M t is the mass gain at time t, M 1 is the saturation mass, and Land In this section, the geometric plane modelling of flax‐epoxy sam-
h are the flax‐epoxy sample side and thickness. ples is firstly presented before developing the hygroscopic and hygroe-
In the case of sealed samples, water diffusion mainly occurs perpen- lastic finite element approximations. To account for their
dicularly to the unsealed faces. Accordingly, the solution of Eq. (3) heterogeneity, three plane geometric models of the flax‐epoxy speci-
reads for the three principal directions: mens are considered in which the UD flax fibres and the epoxy matrix
  are modelled separately as described in Section 3.1.

 1 exp π 2 tD 2iþ12
Mt 8 l l
¼1 2 ∑ ð5Þ 3.1. Description of the plane modelling
M1 π i¼0 ðð2i þ 1ÞÞ2
where Dl ¼ Dx ; Dy ; Dz and l ¼ L; h. To simplify, a plane modelling of water diffusion in the flax‐epoxy
In order to determine the water diffusion parameters of the sealed samples is considered as depicted in Fig. 3. To model water diffusion
and unsealed flax‐epoxy samples, an optimisation algorithm was along the direction of flax fibres and the thickness direction, three
developed using the fminsearch function of Matlab. This program plane models are considered as depicted in Fig. 3b, 3c and 3d. It is
allows minimising the quadratic error q between the analytical solu- worthy to recall that the flax‐epoxy samples are composed of six layers
tion of Fick’s model and the experimental points: of UD flax fibres which explains the geometric models adopted in

Fig. 4. (a) 2D composite solid with moisture and displacement boundary conditions. (b) The three-node hygroelastic membrane finite element.

4
B. Djellouli et al. Composite Structures 265 (2021) 113692

Z Z Z
Fig. 3b and 3d. Owing to symmetry, only a quarter of each plane sur- _
wðc; c ; u; u Þ ¼ c c dV þ rc  D  rc dV þ c  ϕn dS
face is modelled as shown in Fig. 3. B0 B0 @B0
Z Z
Besides, the UD flax fibres bundles are supposed to present a circu-
þ ɛ   σ dV  u  T dS ¼ 0 8 u ; c ð13Þ
lar cross section with diameter of 250 μm and their number and distri- B0 @B0
bution were chosen so as to respect the fibre volume fraction of the
The resolution of the weak form (13) requires spatial and time dis-
flax‐epoxy composite (V f ¼ 33:05%).
cretization. For the spatial discretization, the triangular hygroscopic
membrane element of Zouari et al. [40] is considered and displace-
3.2. Hygroscopic and hygroelastic finite element approximations
ment degrees of freedom are added to estimate the mechanical strains.
The three‐node triangular hygroelastic element is shown in Fig. 4b.
To model the hygroscopic behaviour of the aged flax‐epoxy speci-
The three shape functions associated with the three‐node triangular
mens, the three‐node hygroscopic membrane finite element developed
membrane element allow us to approximate the displacements along
by Zouari et al. [40] is considered.
x and y, denoted u and v, as well as the moisture concentration:
For a 2D composite domain B0 subjected to moisture boundary con-
3 3 3
ditions applied at Sc as shown in Fig. 4a, the weak form of Fick’s dif-
u ¼ ∑ N i ðζ; ηÞui; v ¼ ∑ N i ðζ; ηÞvi ; c ¼ ∑ N i ðζ; ηÞci ð14Þ
ferential equation reads: i¼1 i¼1 i¼1
Z
_ _ @c By using the approximations (14), it is possible to show that c, rc, u
w1 ðc; c Þ ¼ ðc divðD  rcÞÞc dV ¼ 0 8c with c ¼ ð7Þ
B0 @t and ɛ are related to un ¼ f   ui vi ci    i ¼ 1; 3gt ; the nodal
where c* is a test function that verifies c*=0 on Sc. degrees of freedom vector of the hygroelastic element, as follows:
The weak form of Fick's diffusion equation can be rewritten as: c¼t N c  un ; u ¼ N u  un ; rc ¼ Bc  un ; ɛ ¼ Bm  un ð15Þ
Z Z Z
w1 ðc; c Þ ¼
_
c c dV þ rc  D  rc dV þ c ϕn dV ¼ 0 8c ð8Þ At the element level, the finite element approximation of the weak
B0 B0 @B0 form (13) is written:
Z
Z t
_
where ϕn is the normal diffusion flux and@B0 is the boundary of the 2D we ðun ; un Þ ¼ t un  Nc t Nc dV  un þ
Bc  D  Bc dV
B0 B0
composite domain B0. Z t Z t 
In the following, the formulation of the hygroscopic triangular ele- þ Bm  H  Bm dV  Bm  H  βm t Nc dV  un
ment developed by Zouari et al. [40] is extended to the hygroelastic B
Z0 Z
B0

problem. The main objective is to predict the mechanical stress that

þ Nc  ϕn dS  Nu  TdS ¼ 0 8un ð16Þ
appears in the flax‐epoxy samples when they are subjected to water Sc Sc
ageing. To simplify, an uncoupled hygroelastic model is used and After simplification, the finite element equation is written at the
the term “uncoupled” is employed here to indicate that the mechanical elementary level:
variables have no influence on the hygroscopic parameters. K eq  un þ C  u_ n  F eq ¼ 0 ð17Þ
In addition to the hygroscopic equilibrium Eq. (2), the quasi‐static R R R
mechanical equilibrium is considered which, in the absence of body where C ¼ B0 N c  N c dV; F eq ¼ Sc N c  ϕn dS  Sc N u  T dS and K eq ¼
t
R t R t
force, is written locally for a plane composite solid B0 as: K c þ K m  K mc with K c ¼ B0 Bc  D  Bc dV; K m ¼ B0 Bm  H  Bm dV
 R t
σ xx;x ðx; tÞ þ τxy;y ðx; tÞ ¼ 0 and K mc ¼ B0 Bm  H  βm t N c dV.
8x ∈ B0 ð9Þ For the time integration of Eq. (17), the loading interval ½t; t þ Δt  is
τxy;x ðx; tÞ þ σ yy;y ðx; tÞ ¼ 0
considered. The backward Euler integration of Eq. (17) gives the
where σ xx ; σ yy and τxy are the plane components of the mechanical elementary residual vector at t þ Δt:
stress tensor ðσ xx;x ¼ @σ@xxx Þ: Δun
RtþΔt ¼ K eq  untþΔt þ C   F eq
tþΔt
with Δun ¼ utþΔt  utn ð18Þ
The weak form of the static mechanical equilibrium Eq. (9) is Δt n

obtained by introducing a test function u that verifies u ¼ 0 on Su

The elementary tangent stiffness matrix K tþΔt of the three‐node
the part of @B0 on which displacement boundary conditions are T
hygroelastic membrane element is obtained by differentiating the
applied:
Z Z residual vector RtþΔt with respect toΔun :
w2 ðu; u Þ ¼ ɛ  σ dV  u  T dS ¼ 0 8u ð10Þ
B0 @B0

where σ is the vector of plane stresses, ɛ is the vector of virtual plane

strains and T is the vector of external surface forces applied on @B0 . It is
worthy to note that a vectorial notation of the stress and strain tensors
is used in Eq. (10) instead of tensorial notation to simplify the construc-
tion of the finite element approximation.
The stress vector σ can be written in the following form:
σ ¼ H  ɛ el ! σ ¼ H  ðɛ  ɛ h Þ ð11Þ

where H is the elasticity matrix and ɛ is the vector of hygroscopic

h

strains induced by moisture diffusion within the flax‐epoxy samples.

This vector is written as a function of the moisture content cðx; tÞ using
the following linear expression:
ɛ h ðx; tÞ ¼ βh ðx; tÞcðx; tÞ ð12Þ
h
whereβ is the hygroscopic expansion vector which is considered inde-
pendent of the ageing time in this study.
By combining Eqs. (8) and (10), the weak form of the hygroelastic Fig. 5. Experimental water uptake curves of the sealed flax-epoxy samples
equilibrium of the composite medium B0 is obtained: compared with 1D Fick’s model solution.

5
B. Djellouli et al. Composite Structures 265 (2021) 113692

Table 1
Water diffusion parameters of the sealed flax-epoxy sample.

Weight gains at saturationM 1 ð%Þ

Dx ð107 mm2 =sÞ Dy ð107 mm2 =sÞ Dz ð107 mm2 =sÞ
Rx ¼ DDxz Ry ¼
Dy
M x1 ð%Þ M y1 ð%Þ M z1 ð%Þ Dz

648.65 ± 0.64 285.79 ± 0.01 6.45 ± 0.01 9.14 ± 0.13 7.55 ± 0.12 8.01 ± 0.07 100.56 44.30

Fig. 6. SEM micrographs of cross sections of 60-day aged flax-epoxy samples with water diffusion in the direction of flax fibres.

@RtþΔt C Section 2.4. Fig. 5 shows the water uptake curves of the sealed flax‐
K tþΔt ¼ ¼ K eq  ð19Þ
T
@Δu Δt epoxy samples and Table 1 summarises their principal water diffusion
parameters identified from the optimisation procedure.
The triangular hygroelastic finite element has been implemented in In fact, Dx is found 100 times larger than Dz and approximately 2.3
the commercial finite element code ABAQUS using a user element times greater than Dy . This difference is related to the composition of
subroutine. the flax fibre which promotes water transport along its stem particu-
larly through the lumen [31,50–52]. In fact, the hydrophilic compo-
4. Results and discussion nents of the flax fibre (cellulose, hemicellulose and pectin) promote
water diffusion along its longitudinal direction [4,8,30,53].
4.1. Water diffusion parameters of the sealed samples Besides, the weight gain at saturation in the direction of the flax
fibres is found greater than those in the transverse and thickness direc-
In this part, the water diffusion parameters of the sealed flax‐epoxy tions. Indeed, this could be explained by the increase of diffusion rate
samples are identified using the optimisation procedure described in (defined as the slope of the absorption curve) which further promotes

6
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 7. Experimental water uptake curve of the unsealed flax-epoxy samples compared with 3D Fick’s model solution obtained with: (a) the diffusion parameters
of the sealed samples (red curve) (b) the diffusion parameters of Table 2 (black curve).

Table 2
Water diffusion parameters of the unsealed flax-epoxy samples.

Dx ð107 mm2 =sÞ Dy ð107 mm2 =sÞ Dz ð107 mm2 =sÞ M x1 ð%Þ Rx ¼ DDxz Ry ¼
Dy
Dz

213.09 ± 0.01 107.38 ± 0.08 3.55 ± 0.01 9.15 ± 0,10 60,00 30,22

the swelling of flax fibres and leads to more matrix swelling and prop- of moisture diffusion kinetics, which causes more swelling of the flax
agation of micro‐cracks at the fibre‐matrix interface. This process fur- fibres and the epoxy matrix as previously explained.
ther weakens the aged flax‐epoxy samples and thus increases their
water absorption capacity. These explanations are supported by the 4.2. 3D water diffusion parameters of the flax-epoxy specimens
SEM observations of Fig. 6 on sealed flax‐epoxy samples at saturation.
In particular, Fig. 6a shows micro‐cracks at the fibre‐matrix inter- Most of the moisture diffusion analyses of the literature show that
face which can be induced by the formation of hydrogen bonds water diffusion in bio‐composites is complex and the diffusion param-
between the hydrophilic components of the flax fibre primary wall eters identified from the sealed samples do not allow describing the 3D
(pectin, hemicellulose) and water molecules [54]. Fig. 6b shows water diffusion behaviour. This is also confirmed by Fig. 7a which rep-
micro‐cracks in the bundle of flax fibres mainly caused by the differen- resents a comparison between the experimental water uptake curve of
tial hygroscopic swelling between the fibre‐fibre and fibre‐matrix the unsealed flax‐epoxy samples and the 3D Fick’s model solution
interfaces. The presence of these micro‐cracks increases by capillarity obtained with water diffusion parameters of the sealed specimens
water diffusion in these composite samples. Fig. 6c shows the forma- (Table 1).
tion of micro‐porosities with an average size of 0.2 µm which is smaller Accordingly, the principal water diffusion parameters Dx ; Dy and Dz
than the lumen size (Dlumen =5–10 µm) [55]. of the unsealed flax‐epoxy samples should be identified at the same
This can be explained by the dissolution in water of some flax fibre time by the optimization procedure with two additional constraints
components such as pectins. These voids increase the weight gain at on the principal diffusion parameters: Dy < Dx and Dz < Dy : These
saturation of the flax‐epoxy samples. Fig. 6d shows a partial pull‐out two constraints are deduced from the identification of water diffusion
of a flax fibre and this can be explained by a possible failure due to parameters of the sealed flax‐epoxy samples (Table 1).
the differential hygroscopic swelling of the flax fibres. To this end, a penalty function is added to the quadratic error q and
For water diffusion along the transverse and thickness directions (y the new objective function is written as follows:
and z directions), the weight gain at saturation M y1 is found lower than
f obj ¼ q þ ρmaxð0; Dx  Dy ; Dy  Dz Þ ð20Þ
M z1 while Dy is 44.30 times Dz (Table 1). These variations are mainly
related to the sample’s dimensions where the width (y direction) is where ρ is a penalty factor.
almost 10 times the thickness (z direction). In fact, when the dimen- Table 2 summarises the identified water diffusion coefficients of
sions increase the weight gain at saturation decrease and the water dif- the unsealed flax‐epoxy samples and Fig. 7b shows that these optimal
fusion coefficients increase. The decrease of the saturation weight with parameters allow correctly describing the hygroscopic behaviour of
sample dimensions has been also reported in the case of other compos- these samples.
ite materials and some explanations have been proposed [30,31,56]. It is interesting to note that the water diffusion parameters of the
Bunsell et al. [56] have attributed this variation to a molecular rear- unsealed flax‐epoxy specimens (Table 2) are different from those of
rangement of the polymeric network when samples are thicker which the sealed samples (Table 1).
would slow down the water diffusion in the composite. For the flax‐ Besides, the ratios Rx and Ry of the unsealed samples show a
epoxy samples studied in this work, the variation of the saturation decrease by 41% and 32% respectively compared with those of the
weight with the dimensions could also be accentuated by the increase sealed samples (Table 1). This decrease is probably related to the

7
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 8. Finite element modelling of water diffusion in the unsealed flax-epoxy samples using 56,232 three-node hygroscopic elements.

movement of water particles which are forced to follow one of the and longitudinal diffusion coefficients of the UD flax fibres
three main directions in the case of sealed samples, in particular in (DLf and DRf Þ:
the direction of flax fibres which explains the important value of Rx In fact, the modelling in the yz plane (Fig. 3b) allows considering
for sealed samples. For the unsealed specimens, the privileged direc- only the radial diffusion parameter of the flax fibre which could bias
tion is less pronounced than in the case of sealed samples, and a kind its estimation. It is also important to recall that the aged flax‐epoxy
of hygroscopic balance between the three main directions is obtained, samples are constituted of six layers of UD flax fibres, which leads to
which could explain the relatively large decrease in Rx and Ry . a total thickness of 2.68 mm. Owing to symmetry, one‐quarter of the
At the end of this study, by making the hypothesis that the moisture xz plane is modelled as shown in Fig. 8. The three layers of UD flax
diffusion process is mainly controlled by the unidirectional flax fibres, fibres are modelled in this plane by three rectangles of thickness
a ratio between their longitudinal and radial water diffusion coeffi- 0.147 mm to respect the fibre volume fraction of 33%. The refined
cients (denoted DLf and DRf ) equivalent to that resulting from the opti- mesh of Fig. 8 constituted of 56,232 elements was fixed after a mesh
mization procedure on unsealed samples ðRx ¼ Dx =Dy ¼ 60Þ can be convergence study. The moisture boundary conditions are applied to
deduced. This ratio is considered in Section 4.3 to estimate the longi- the top and left edges of the 2D finite element model
tudinal and radial moisture diffusion coefficients of the flax fibre with (c ¼ M 1 ¼ 9:15%).
the hypothesisDLf ¼ 60 DRf . Fig. 9 shows a comparison between the experimental and the
numerical water uptake curves of the unsealed flax‐epoxy samples
4.3. Finite element analysis of the hygroscopic behaviour using the first set of diffusion parameters (the red curve). It is worthy
to note that the numerical water uptake curve was calculated as the
In this part, the hygroscopic behaviour of the flax‐epoxy samples is arithmetic average of all nodal moisture concentrations of the finite
modelled using the hygroscopic plane finite element of Zouari et al. element model. The obtained results indicate that the numerical curve
[40] (three‐node triangular membrane element with only moisture does not accurately describe the experimental curve. The difference
degrees of freedom). To this end, the water diffusion coefficients of between these two curves reflects the difficulty of predicting the
the epoxy resin and those of the UD flax fibres are needed. The hygro- hygroscopic behaviour of the flax‐epoxy samples using a heteroge-
scopic behaviour of the pure epoxy resin is supposed isotropic and neous model. In the following a better approximation of the radial
characterised by the water diffusion coefficient and longitudinal water diffusion coefficients of the flax fibre is pro-
Dm ¼ 5:42  107 mm2 =s obtained from its experimental water uptake posed to correctly fit the experimental curve of the unsealed flax‐
curve. This value is close to other epoxy moisture diffusion coefficients
reported in the literature [57,58].
For the UD flax fibres, their water diffusion behaviour is anisotro-
pic and should be principally described by a radial and a longitudinal
coefficient denotedDLf and DRf [8,39,40,59,60].
Although some works in the literature have focused on the determi-
nation of the flax fibre radial diffusion coefficient as the work by
Célino et al [61], the determination of its longitudinal diffusion param-
eter remains almost inexistent in the literature. Therefore, an estima-
tion of the radial and longitudinal water diffusion parameters of the
flax fibre is proposed in this section by considering an inverse
approach. As explained at the end of Section 4.2, a ratio of 60 between
DLf and DRf is considered.
To begin, a first approximation of the radial diffusion coefficient of
the flax fibre was obtained by means of Halpin‐Tsai homogenisation
model [62] knowing the value of Dm and the water diffusion
parameter across the thickness of the sealed flax‐epoxy samples
ðDz ¼ 6:45  107 mm2 =sÞ:In this case, the first approximation of the
radial diffusion parameter of the flax fibre is
DRf ¼ 5:45  107 mm2 =s. The water diffusion coefficients DRf ; DLf
Fig. 9. Comparison between the experimental and the numerical water
ðDLf ¼ 60DRf Þ and Dm are now used as inputs of the finite element model uptake curves of the unsealed flax-epoxy samples by considering the first set of
of Fig. 3d to simulate water diffusion in the unsealed flax‐epoxy parameters (red curve): DRf ¼ 5:45  107 mm2 =s and Dm ¼ 5:42
7
specimens. It is worthy to note that the finite element modelling is 10 mm =s and
2
the second set of parameters (black curve)
conducted in the xz plane (Fig. 3d) because it considers the radial :DRf ¼ 16:32  107 mm2 =s,and Dm ¼ 5:66  107 mm2 =s.

8
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 10. (a) Plane finite element modelling of the hygroelastic behaviour of the unsealed flax-epoxy samples with a mesh of 78,452 triangular elements. (b)-(d)
Plane modelling with three random distributions of the flax fibres.

epoxy samples. To this end, Dm and DRf have been varied while always experimental results of the literature [61,64]. For example, Stamboulis
keeping ¼ 60
DLf until fitting the experimental curve (see Zouari
DRf et al. [64] have reported a value of the flax fibre diffusion parameter
et al. [40] for more details about this inverse procedure). Accordingly, equal to 1  104 mm2 =s for a flax fibre bundle of 1.5 mm of length
aged at 100% of relative humidity at room temperature. This value
the following optimal parametersDRf ¼ 16:32  107 mm2 =s,
of DLf is very close to that estimated in this work. In another work,
DLf ¼ 979:21  107 mm2 =s andDm ¼ 5:66  107 mm2 =s, have been
Célino et al. [61] have studied moisture diffusion kinetics in UD flax
obtained and allow an accurate prediction of the experimental curve
samples aged by total immersion in water at room temperature. They
as shown in Fig. 9 (the black curve).
have assumed that moisture diffusion principally occurs across the
It is interesting to note that the water diffusion coefficient of the
radial direction of the flax fibres and have reported a radial diffusion
epoxy matrix in the flax‐epoxy samples (Dm ¼ 5:66  107 mm2 =s) is
parameter of 11:9  107 mm2 =s which is also close to the value esti-
slightly greater than that of the pure epoxy
mated in this study.
resinðDm ¼ 5:42  107 mm2 =s).
This behaviour has been already reported in the case of synthetic
4.4. Finite element analysis of the hygroelastic behaviour
fibre [43,63] and also natural fibre reinforced composites
[39,40,43]. It can be explained firstly by the presence of porosities
In this section, the three‐node triangular hygroelastic finite element
in the flax‐epoxy samples, compared to the epoxy pure resin speci-
presented in Section 3.2 is considered to estimate the internal hygro-
mens, and secondly the appearance of micro‐cracks due to the dis-
scopic stress that appears in the aged flax‐epoxy samples. To this
similar swelling between the flax fibres, very hydrophilic, and the
end, the finite element model of Fig. 10a with displacement and mois-
epoxy matrix, rather hydrophobic compared to the reinforcement.
ture boundary conditions is considered. It is worthy to note that this
These defects can make the epoxy matrix more brittle to moisture
modelling is conducted in the yz plane where the sections of the UD
diffusion and partly explains the increase in its water diffusion
flax fibres are modelled in a first approximation with circles of diam-
coefficient.
eter 0.25 mm (Figs. 10, 11).
Concerning the longitudinal and radial moisture diffusion coeffi-
The main objective is to more accurately estimate the hygroscopic
cients of the flax fibre ðDRf ¼ 16:32  107 mm2 =s and stress caused by water ageing compared to a modelling in the xz plane
DLf ¼ 979:21  107 mm2 =s), they seem in good agreement with some (Fig. 3b and c). The influence of the distribution of the UD flax fibres in

Fig. 11. Distribution of the transverse displacement at saturation (after 66 days of immersion) of the unsealed flax-epoxy samples modelled with a regular
distribution of the flax fibres.

9
B. Djellouli et al. Composite Structures 265 (2021) 113692

Fig. 12. Distribution of the internal stress across the thickness σ zz inside the unsealed flax-epoxy samples after 66 days of immersion: (a) regular distribution of the
flax fibres, (b)-(d) random distributions of the flax fibres.

the flax‐epoxy samples is also studied by considering the random dis- the epoxy matrix is subjected to tensile stress. The compression stress
tributions of Fig. 10b–d in addition to the regular distribution of that appears in the reinforcement can be explained by the fact that the
Fig. 10a. flax fibres and the epoxy matrix don’t swell in the same way, although
For the hygroscopic and elastic material properties of the epoxy they present the same moisture concentration at saturation
matrix and the UD flax fibres to be used with the finite element models ðc ¼ 9:153%Þ. This is due to the difference between their hygroscopic
of Fig. 10, the water diffusion parameters estimated in Section 4.3 in swelling parameters (βRf ≈4βm ). Thus, the flax fibres have tendency to
relation with the unsealed flax‐epoxy samplesðDm ¼ 5:66 swell more than the epoxy matrix and are therefore found constrained
107 mm2 =s and DRf ¼ 16:32  107 mm2 =sÞ are firstly considered. For by the latter. In addition, it is worth noting that the flax fibres distri-
the hygroscopic expansion coefficients, the following epoxy matrix bution can significantly influence the transverse internal stress distri-
coefficientβm ¼ 0:24 is used and constitutes an average of some coeffi- bution within the flax‐epoxy sample as shown in Fig. 12. Indeed, the
cients reported in the literature [65–68]. In addition, the hygroscopic stress peaks, which are mainly localized in the epoxy matrix, vary from
expansion of the epoxy matrix is assumed to be isotropic. For the UD 217.3 MPa (regular distribution) to 463.5 MPa found with the third
flax fibres, their radial hygroscopic expansion coefficient is particu- random finite element model (Fig. 12d). Similar stress peaks have been
larly considered based on the plane models of Fig. 10. already reported in the case of synthetic fibre‐reinforced polymer com-
To this end, the radial expansion coefficient reported by Péron posites where the synthetic fibres have been supposed hydrophobic
et al. [69] (βRf ¼ 0:97) is used which seems to be in agreement with (moisture diffusion and hygroscopic expansion parameters equal to
that given by Le Duigou et al. [70] (βRf ¼ 1:14) and the expansion zero) [41,46]. In the work by Ouled Ahmed et al. [46], stress peaks
close to 300 MPa within glass‐polyester composites aged in distilled
parameter reported by Garat et al. [71] (βRf ¼ 1:07). For the longitudi-
water have been found.
nal hygroscopic expansion coefficient of the UD flax fibres, Le Duigou The relatively high levels of internal stress found in this study,
et al. [72] have explained that it is very negligible with respect to βRf which are certainly overestimated by the current uncoupled hygroelas-
and close to zero as for wood fibres. tic model, exceed in particular the tensile strength of the epoxy matrix
Fig. 12 shows the distribution of the hygroscopic internal stress (around 40 MPa), which can lead to the initiation of damage by matrix
across to the thickness ðσ zz Þ at saturation within the aged flax‐epoxy cracking, delamination and degradation of the flax fibres without add-
samples using regular and random distributions of the flax fibres. It ing external loads.
is important to note that the phenomenon of hygroscopic swelling This could explain the micro‐cracks in the matrix close to the fibre‐
mainly occurs across to the thickness of the aged samples which matrix interface depicted in Fig. 6a as well as cracks in the flax fibre
explains the particular study of σ zz . bundle shown in Fig. 6b. Despite the high difference in the transverse
Fig. 13 also shows the evolution of the mean transverse stress dur- internal stress distribution between the regular and random finite ele-
ing ageing in the matrix (Fig. 13a), in the reinforcement (Fig. 13b) and ment models, the mean value of this hygroscopic stress varies from 40
in the composite sample (Fig. 13c). These results show that, overall, to 50 MPa in the matrix and from 80 to 100 MPa in the reinforcement
the flax fibres are subjected to hygroscopic compression stress while (Fig. 13).

10
B. Djellouli et al. Composite Structures 265 (2021) 113692

m f
Fig. 13. Evolution of the transverse mean stress during ageing inside: (a) the matrix σ zz ; (b) the reinforcement σ zz and (c) the flax-epoxy composite sample.

5. Conclusion which can lead to the initiation of damage inside the flax‐epoxy
samples without external loads.
In this paper, the transient hygroscopic behaviour of unidirec- As a future work, it would be interesting to consider another
tional (UD) flax fibre‐reinforced epoxy composite aged by immer- hygroscopic model to describe water diffusion kinetics within the
sion in tap water at room temperature until saturation was firstly aged samples. In fact, Fick’s model used in this study can’t take into
investigated. To this end, the absorption water uptake curves of account the maximum amount of water absorbed by the epoxy
sealed and unsealed flax‐epoxy samples, which present Fickian beha- matrix since it makes the assumption that water molecules move
viour, allowed identifying their macroscopic water diffusion param- freely within the aged composite samples. This is in contradiction
eters using an optimisation approach. The main conclusion drawn with the water uptake curve of the pure epoxy resin showing a max-
from this first part is that water diffusion is much more predomi- imum water uptake of 1.1% while the moisture boundary conditions
nant in the direction of the flax fibres than the thickness and the applied to the finite element models of this study largely exceed this
width directions. Secondly, the results of the first part were consid- value (~10%). This may partly explain the overestimation of the
ered to estimate the flax fibre radial and longitudinal water diffu- internal hygroscopic stress in the matrix found by the hygroelastic
sion parameters using a numerical inverse procedure. This was finite element modelling. To this end, the use of a more general dif-
conducted by a plane finite element model that takes into account fusion model like that of Langmuir could be considered. In addition
the heterogeneity of the flax‐epoxy samples. The obtained flax fibre to the hygroscopic model, the use of a fully coupled hygroelastic
diffusion parameters seem in good accordance with other literature model between the mechanical and the hygroscopic variables leads
results. Finally, the estimation of the hygroscopic internal stress that to a more accurate estimation of the hygroscopic stress inside the
appears in the flax‐epoxy samples due to water ageing was investi- flax‐epoxy specimens.
gated. To this end, a three‐node triangular membrane finite element
based on an uncoupled hygroelastic model was developed. Finite Declaration of Competing Interest
element simulations with regular and random distributions of the
flax fibres were conducted and important stress peaks were obtained The authors declare that they have no known competing financial
particularly in the matrix domain. These stress levels exceed in par- interests or personal relationships that could have appeared to influ-
ticular the tensile strength of the epoxy matrix (around 40 MPa), ence the work reported in this paper.

11
B. Djellouli et al. Composite Structures 265 (2021) 113692

References [25] Tengsuthiwat J, Asawapirom U, Siengchin S, Karger-Kocsis J. Mechanical,

thermal, and water absorption properties of melamine–formaldehyde-treated
sisal fiber containing poly(lactic acid) composites. J Appl Polym Sci 2018;135
[1] Abd El-baky MA, Attia MA. Water absorption effect on the in-plane shear
(2):45681. https://doi.org/10.1002/app.45681.
properties of jute–glass–carbon-reinforced composites using Iosipescu test. J
[26] Cheour K, Assarar M, Scida D, Ayad R, Gong X-L. Long-term immersion in water of
Compos Mater 2019;53(21):3033–45. https://doi.org/10.1177/
flax-glass fibre hybrid composites: effect of stacking sequence on the mechanical
0021998318809525.
and damping properties. Fibers Polym 2020;21(1):162–9. https://doi.org/
[2] Pickering KL, Efendy MGA, Le TM. A review of recent developments in natural
10.1007/s12221-020-9494-7.
fibre composites and their mechanical performance. Compos A Appl Sci Manuf
[27] Espert A, Vilaplana F, Karlsson S. Comparison of water absorption in natural
2016;83:98–112. https://doi.org/10.1016/j.compositesa.2015.08.038.
cellulosic fibres from wood and one-year crops in polypropylene composites and
[3] George M, Chae M, Bressler DC. Composite materials with bast fibres: structural,
its influence on their mechanical properties. Compos A Appl Sci Manuf 2004;35
technical, and environmental properties. Prog Mater Sci 2016;83:1–23. https://
(11):1267–76. https://doi.org/10.1016/j.compositesa:2004.04.004.
doi.org/10.1016/j.pmatsci.2016.04.002.
[28] Regazzi A, Corn S, Ienny P, Bergeret A. Coupled hydro-mechanical aging of short
[4] Célino A, Freour S, Jacquemin F, Casari P. The hygroscopic behavior of plant
flax fiber reinforced composites. Polym Degrad Stab 2016;130:300–6. https://doi.
fibers: a review. Front Chem 2014;1. https://doi.org/10.3389/fchem.2013.00043.
org/10.1016/j.polymdegradstab.2016.06.016.
[5] Al-Hajaj Z, Sarwar A, Zdero R, Bougherara H. In-situ damage assessment of a novel
[29] Charlet K, Baley C, Morvan C, Jernot JP, Gomina M, Bréard J. Characteristics of
carbon/flax/epoxy hybrid composite under tensile and compressive loading. J
Hermès flax fibres as a function of their location in the stem and properties of the
Compos Mater 2019;53(19):2701–14. https://doi.org/10.1177/
derived unidirectional composites. Compos A Appl Sci Manuf 2007;38
0021998319839129.
(8):1912–21. https://doi.org/10.1016/j.compositesa:2007.03.006.
[6] Moudood A, Rahman A, Khanlou HM, Hall W, Öchsner A, Francucci G.
[30] Chilali A, Assarar M, Zouari W, Kebir H, Ayad R. Effect of geometric dimensions
Environmental effects on the durability and the mechanical performance of flax
and fibre orientation on 3D moisture diffusion in flax fibre reinforced
fiber/bio-epoxy composites. Compos B Eng 2019;171:284–93. https://doi.org/
thermoplastic and thermosetting composites. Compos A Appl Sci Manuf
10.1016/j.compositesb.2019.05.032.
2017;95:75–86. https://doi.org/10.1016/j.compositesa.2016.12.020.
[7] Sergi C, Tirillò J, Seghini MC, Sarasini F, Fiore V, Scalici T. Durability of basalt/
[31] Arnold JC, Alston SM, Korkees F. An assessment of methods to determine the
hemp hybrid thermoplastic composites. Polymers (Basel) 2019;11.;11(4):603.
directional moisture diffusion coefficients of composite materials. Compos A Appl
https://doi.org/10.3390/polym11040603.
Sci Manuf 2013;55:120–8. https://doi.org/10.1016/j.compositesa.2013.08.012.
[8] Saidane EH, Scida D, Assarar M, Ayad R. Assessment of 3D moisture diffusion
[32] Jiang X, Song J, Qiang X, Kolstein H, Bijlaard F. Moisture absorption/desorption
parameters on flax/epoxy composites. Compos A Appl Sci Manuf 2016;80:53–60.
effects on flexural property of glass-fiber-reinforced polyester laminates: three-
https://doi.org/10.1016/j.compositesa.2015.10.008.
point bending test and coupled hygro-mechanical finite element analysis.
[9] Chilali A, Zouari W, Assarar M, Kebir H, Ayad R. Effect of water ageing on the load-
Polymers 2016;8:290. https://doi.org/10.3390/polym8080290.
unload cyclic behaviour of flax fibre-reinforced thermoplastic and thermosetting
[33] Beringhier M, Simar A, Gigliotti M, Grandidier JC, Ammar-Khodja I. Identification
composites. Compos Struct 2018;183:309–19. https://doi.org/10.1016/
of the orthotropic diffusion properties of RTM textile composites for aircraft
j.compstruct.2017.03.077.
applications. Compos Struct 2016;137:33–43. https://doi.org/10.1016/
[10] Baley C, Gomina M, Breard J, Bourmaud A, Davies P. Variability of mechanical
j.compstruct.2015.10.039.
properties of flax fibres for composite reinforcement. A review. Ind Crops Prod
[34] Humeau C, Davies P, Jacquemin F. An experimental study of water diffusion in
2020;145:111984. https://doi.org/10.1016/j.indcrop.2019.111984.
carbon/epoxy composites under static tensile stress. Compos A Appl Sci Manuf
[11] Moudood A, Rahman A, Öchsner A, Islam M, Francucci G. Flax fiber and its
2018;107:94–104. https://doi.org/10.1016/j.compositesa.2017.12.016.
composites: an overview of water and moisture absorption impact on their
[35] Dana HR, Casari P, El Abdi R, Fréour S, Jacquemin F. Characterisation of the
performance. J Reinf Plast Compos 2019;38(7):323–39. https://doi.org/10.1177/
hygro-thermo-mechanical behaviour of organic matrix composites instrumented
0731684418818893.
with optical fibres: a study of interfacial bonding. Int J Adhes Adhes
[12] Omrani F, Wang P, Soulat D, Ferreira M. Mechanical properties of flax-fibre-
2017;77:63–71. https://doi.org/10.1016/j.ijadhadh.2017.04.001.
reinforced preforms and composites: influence of the type of yarns on multi-scale
[36] Beringhier M, Djato A, Maida D, Gigliotti M. A novel protocol for rapid
characterisations. Compos A Appl Sci Manuf 2017;93:72–81. https://doi.org/
identification of anisotropic diffusion properties of polymer matrix composite
10.1016/j.compositesa.2016.11.013.
materials with complex texture. Compos Struct 2018;201:1088–96. https://doi.
[13] Toubal L, Cuillière J-C, Bensalem K, Francois V, Gning P-B. Hygrothermal effect on
org/10.1016/j.compstruct.2018.05.142.
moisture kinetics and mechanical properties of hemp/polypropylene composite:
[37] Sinchuk Y, Pannier Y, Antoranz-Gonzalez R, Gigliotti M. Analysis of moisture
experimental and numerical studies. Polym Compos 2016;37(8):2342–52. https://
diffusion induced stress in carbon/epoxy 3D textile composite materials with voids
doi.org/10.1002/pc.23414.
by µ-CT based Finite Element Models. Compos Struct 2019;212:561–70. https://
[14] Cheour K, Assarar M, Scida D, Ayad R, Gong X-L. Effect of water ageing on the
doi.org/10.1016/j.compstruct.2018.12.041.
mechanical and damping properties of flax-fibre reinforced composite materials.
[38] Péron M, Célino A, Castro M, Jacquemin F, Le Duigou A. Study of hygroscopic
Compos Struct 2016;152:259–66. https://doi.org/10.1016/
stresses in asymmetric biocomposite laminates. Compos Sci Technol
j.compstruct.2016.05.045.
2019;169:7–15. https://doi.org/10.1016/j.compscitech.2018.10.027.
[15] Scida D, Assarar M, Poilâne C, Ayad R. Influence of hygrothermal ageing on the
[39] Chilali A, Assarar M, Zouari W, Kebir H, Ayad R. Analysis of the hydro-mechanical
damage mechanisms of flax-fibre reinforced epoxy composite. Compos B Eng
behaviour of flax fibre-reinforced composites: assessment of hygroscopic
2013;48:51–8. https://doi.org/10.1016/j.compositesb.2012.12.010.
expansion and its impact on internal stress. Compos Struct 2018;206:177–84.
[16] Le Duigou A, Bourmaud A, Baley C. In-situ evaluation of flax fibre degradation
https://doi.org/10.1016/j.compstruct.2018.08.037.
during water ageing. Ind Crops Prod 2015;70:204–10. https://doi.org/10.1016/j.
[40] Zouari W, Assarar M, Chilali A, Ayad R, Kebir H. Numerical modelling of the
indcrop.2015.03.049.
transient hygroscopic behavior of flax-epoxy composite. J Renewable Mater
[17] Nguyen-Duy C, Makke A, Montay G. A Pull-Out Test to Characterize the Fiber/
2019;7:839–53. , https://doi.org/10.32604/jrm.2019.06773.
Matrix Interfaces Aging of Hemp Fiber Reinforced Polypropylene Composites. In:
[41] Peret T, Clement A, Freour S, Jacquemin F. Numerical transient hygro-elastic
Nguyen-Xuan H, Phung-Van P, Rabczuk T, editors. Proceedings of the
analyses of reinforced Fickian and non-Fickian polymers. Compos Struct
International Conference on Advances in Computational Mechanics 2017,
2014;116:395–403. https://doi.org/10.1016/j.compstruct.2014.05.026.
Singapore: Springer Singapore; 2018, p. 477–84. https://doi.org/10.1007/978-
[42] Joliff Y, Belec L, Chailan JF. Modified water diffusion kinetics in an unidirectional
981-10-7149-2_32.
glass/fibre composite due to the interphase area: experimental, analytical and
[18] Malloum A, Mahi AE, Idriss M. The effects of water ageing on the tensile static and
numerical approach. Compos Struct 2013;97:296–303. https://doi.org/10.1016/
fatigue behaviors of greenpoxy–flax fiber composites. J Compos Mater 2019;53
j.compstruct.2012.09.044.
(21):2927–39. https://doi.org/10.1177/0021998319835596.
[43] Joliff Y, Belec L, Heman MB, Chailan JF. Experimental, analytical and numerical
[19] Kollia E, Saridaki X, Karagiannis D, Kostopoulos V. Effect of water aging on the
study of water diffusion in unidirectional composite materials – Interphase impact.
mechanical properties of flax fiber/bio-based resin composites. J Appl Polym Sci
Comput Mater Sci 2012;64:141–5. https://doi.org/10.1016/
2020;137(13):48787. https://doi.org/10.1002/app.v137.1310.1002/app.48787.
j.commatsci.2012.05.029.
[20] Fick A. Ueber Diffusion. Annalen Der Physik 1855;170:59–86. https://doi.org/
[44] Mercier J, Bunsell A, Castaing P, Renard J. Characterisation and modelling of
10.1002/andp.18551700105.
aging of composites. Compos A Appl Sci Manuf 2008;39(2):428–38. https://doi.
[21] Langmuir-Type Model for Anomalous Moisture Diffusion In Composite Resins -
org/10.1016/j.compositesa:2007.08.015.
Harris G. Carter, Kenneth G. Kibler, 1978 n.d. https://journals.sagepub.com/doi/
[45] Meng M, Rizvi MJ, Le HR, Grove SM. Multi-scale modelling of moisture diffusion
10.1177/002199837801200201 (accessed January 13, 2020).
coupled with stress distribution in CFRP laminated composites. Compos Struct
[22] Christian SJ, Billington SL. Moisture diffusion and its impact on uniaxial tensile
2016;138:295–304. https://doi.org/10.1016/j.compstruct.2015.11.028.
response of biobased composites. Compos B Eng 2012;43(5):2303–12. https://doi.
[46] Ouled Ahmed RBA, Chatti S, Ben DH. Modeling of hygrothermal damage of
org/10.1016/j.compositesb:2011.11.063.
composite materials. Mech Adv Compos Struct 2016;3:137–44. , https://doi.org/
[23] Cherif ZE, Poilâne C, Vivet A, Ben Doudou B, Chen J. About optimal architecture of
10.22075/macs.2016.475.
plant fibre textile composite for mechanical and sorption properties. Compos
[47] Peret T, Clement A, Freour S, Jacquemin F. Effect of mechanical states on water
Struct 2016;140:240–51. https://doi.org/10.1016/j.compstruct.2015.12.030.
diffusion based on the free volume theory: numerical study of polymers and
[24] Habibi M, Laperrière L, Hassanabadi HM. Effect of moisture absorption and
laminates used in marine application. Compos B Eng 2017;118:54–66. https://doi.
temperature on quasi-static and fatigue behavior of nonwoven flax epoxy
org/10.1016/j.compositesb.2017.02.046.
composite. Compos B Eng 2019;166:31–40. https://doi.org/10.1016/
[48] Crank J, Crank EPJ. The Mathematics of Diffusion. Clarendon Press; 1979.
j.compositesb.2018.11.131.

12
B. Djellouli et al. Composite Structures 265 (2021) 113692

[49] De Wilde WP, Frolkovic P. The modelling of moisture absorption in epoxies: effects [62] Affdl JCH, Kardos JL. The Halpin-Tsai equations: a review. Polym Eng Sci 1976;16
at the boundaries. Composites 1994;25(2):119–27. https://doi.org/10.1016/ (5):344–52. https://doi.org/10.1002/(ISSN)1548-263410.1002/pen.
0010-4361(94)90005-1. v16:510.1002/pen.760160512.
[50] Aoki Y, Yamada K, Ishikawa T. Effect of hygrothermal condition on compression [63] Vaddadi P, Nakamura T, Singh RP. Transient hygrothermal stresses in fiber
after impact strength of CFRP laminates. Compos Sci Technol 2008;68 reinforced composites: a heterogeneous characterization approach. Compos A Appl
(6):1376–83. https://doi.org/10.1016/j.compscitech.2007.11.015. Sci Manuf 2003;34(8):719–30. https://doi.org/10.1016/S1359-835X(03)00135-0.
[51] Bao L-R, Yee AF. Moisture diffusion and hygrothermal aging in bismaleimide [64] Stamboulis A, Baillie CA, Peijs T. Effects of environmental conditions on
matrix carbon fiber composites—part I: uni-weave composites. Compos Sci mechanical and physical properties of flax fibers. Compos A Appl Sci Manuf
Technol 2002;62(16):2099–110. https://doi.org/10.1016/S0266-3538(02) 2001;32(8):1105–15. https://doi.org/10.1016/S1359-835X(01)00032-X.
00161-6. [65] Karalekas D, Cugnoni J, Botsis J. Monitoring of hygrothermal ageing effects in an
[52] Choi HS, Ahn KJ, Nam J-D, Chun HJ. Hygroscopic aspects of epoxy/carbon fiber epoxy resin using FBG sensor: a methodological study. Compos Sci Technol
composite laminates in aircraft environments. Compos A Appl Sci Manuf 2001;32 2009;69(3-4):507–14. https://doi.org/10.1016/j.compscitech.2008.11.028.
(5):709–20. https://doi.org/10.1016/S1359-835X(00)00145-7. [66] Changsoo Jang, Yoon S, Bongtae Han. Measurement of the hygroscopic swelling
[53] Yan L, Chouw N, Jayaraman K. Flax fibre and its composites – a review. Compos B coefficient of thin film polymers used in semiconductor packaging. IEEE Trans
Eng 2014;56:296–317. https://doi.org/10.1016/j.compositesb.2013.08.014. Comp Packag Technol 2010;33(2):340–6. https://doi.org/10.1109/
[54] Baley C, Busnel F, Grohens Y, Sire O. Influence of chemical treatments on surface TCAPT.2009.2038366.
properties and adhesion of flax fibre–polyester resin. Compos A Appl Sci Manuf [67] Glaskova-Kuzmina T, Aniskevich A, Sevcenko J, Borriello A, Zarrelli M. Cyclic
2006;37(10):1626–37. https://doi.org/10.1016/j.compositesa:2005.10.014. moisture sorption and its effects on the thermomechanical properties of epoxy and
[55] Charlet K, Jernot J-P, Breard J, Gomina M. Scattering of morphological and epoxy/MWCNT nanocomposite. Polymers 2019;11:1383. https://doi.org/
mechanical properties of flax fibres. Ind Crops Prod 2010;32(3):220–4. https:// 10.3390/polym11091383.
doi.org/10.1016/j.indcrop.2010.04.015. [68] Yang S, Kwon S, Lee MY, Cho M. Molecular dynamics and micromechanics study
[56] Bunsell AR. Hydrothermal ageing of composite materials. Rev Inst Fr Pét 1995;50 of hygroelastic behavior in graphene oxide-epoxy nanocomposites. Compos B Eng
(1):61–7. https://doi.org/10.2516/ogst:1995006. 2019;164:425–36. https://doi.org/10.1016/j.compositesb.2019.01.059.
[57] Liu M, Wu P, Ding Y, Chen G, Li S. Two-dimensional (2D) ATR−FTIR [69] Péron M, Célino A, Castro M, Jacquemin F, Duigou A. Biocomposites with
spectroscopic study on water diffusion in cured epoxy resins. Macromolecules asymmetric stacking for the study of hygro-mechanical couplings. RCMA 2019;29
2002;35(14):5500–7. https://doi.org/10.1021/ma011819f. (4):243–52.
[58] Li L, Yu Y, Wu Q, Zhan G, Li S. Effect of chemical structure on the water sorption of [70] le Duigou A, Merotte J, Bourmaud A, Davies P, Belhouli K, Baley C. Hygroscopic
amine-cured epoxy resins. Corros Sci 2009;51(12):3000–6. https://doi.org/ expansion: a key point to describe natural fibre/polymer matrix interface bond
10.1016/j.corsci.2009.08.029. strength. Compos Sci Technol 2017;151:228–33. https://doi.org/10.1016/
[59] Baley C. Analysis of the flax fibres tensile behaviour and analysis of the tensile j.compscitech.2017.08.028.
stiffness increase. Compos A Appl Sci Manuf 2002;33(7):939–48. https://doi.org/ [71] Garat W, Le Moigne N, Corn S, Beaugrand : J, Bergeret A. Swelling of natural fibre
10.1016/S1359-835X(02)00040-4. bundles under hygro- and hydrothermal conditions: determination of hydric
[60] Monti A, El Mahi A, Jendli Z, Guillaumat L. Mechanical behaviour and damage expansion coefficients by automated laser scanning. Composites Part A: Applied
mechanisms analysis of a flax-fibre reinforced composite by acoustic emission. Science and Manufacturing 2020:105803. https://doi.org/10.1016/
Compos A Appl Sci Manuf 2016;90:100–10. https://doi.org/10.1016/ j.compositesa.2020.105803.
j.compositesa.2016.07.002. [72] Joffre T, Neagu RC, Bardage SL, Gamstedt EK. Modelling of the hygroelastic
[61] Célino A, Fréour S, Jacquemin F, Casari P. Characterization and modeling of the behaviour of normal and compression wood tracheids. J Struct Biol 2014;185
moisture diffusion behavior of natural fibers. J Appl Polym Sci 2013;130 (1):89–98. https://doi.org/10.1016/j.jsb:2013.10.014.
(1):297–306. https://doi.org/10.1002/app.39148.

13