You are on page 1of 16

Original Article

Journal of Reinforced Plastics


and Composites

Impregnated fibre bundle test for natural 0(0) 1–16


! The Author(s) 2017
Reprints and permissions:
fibres used in composites sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0731684417695461
journals.sagepub.com/home/jrp

F Bensadoun1, I Verpoest1, J Baets1, J Müssig2, N Graupner2,


P Davies3, M Gomina4, A Kervoelen5 and C Baley5

Abstract
In this study, the impregnated fibre bundle test, a common method used by carbon and glass fibre manufacturers to
determine the properties of fibres used in composites, was adapted for natural fibres and validated by a round robin test
on one type of natural fibres, namely flax fibres. Five European laboratories have carried out in parallel the impregnated
fibre bundle test, on the same batch of hackled flax (long fibres), to check the applicability and reliability of this modified
method on natural fibres. The results were compared to the more traditional single fibre test on elementary fibres.
The back-calculated fibre stiffness shows a very low scatter between the five laboratories of less than 5%
(59.8  2.4 GPa, as measured between 0 and 0.1% strain). The fibre ultimate tensile strength of 527  138 MPa has a
higher scatter, compared to stiffness values, as this property is highly sensitive to imperfections and flaws.

Keywords
IFBT, flax fibre, hemp fibre, mechanical properties, single fibre test

Introduction data sheets to report the fibre properties.4 There are


Natural fibres play an increasing role as reinforcement two intrinsic advantages of this method. The first is
in polymer composites not only because they are that a large number of fibres is tested simultaneously,
‘natural’ but also because they have interesting proper- resulting in a rather direct measurement of the average
ties, which makes them competitive with or complemen- ‘‘effective’’ back-calculated fibre stiffness and strength.
tary to glass and carbon fibres. When designers have to Second, the test is less complicated than a single fibre
decide which reinforcement to use in their composite test (SFT), where a delicate preparation of the fibres is
products, they first look at the mechanical performance required. Moreover, a very large number of individual
of the fibre-reinforced composite. Basic mechanical fibres have to be tested to provide statistically reliable
properties, like strength and stiffness, are dominated by results.1,5
the fibre properties; hence, it is important to characterize
the fibre properties and measure them in a reliable way.
As of today, there are three distinct ways to characterize
1
the fibre properties. Either, the fibres are tested as such, Department of Materials Engineering, KU Leuven, Leuven, Belgium
using single fibre tensile tests (SFT)1 or dry fibre bundle
2
HSB – University of Applied Sciences Bremen, Faculty 5 – Biomimetics/
Biological Materials, Bremen, Germany
tests.2 Alternatively, the fibre properties are back-calcu- 3
Centre de Bretagne, Marine Structures Group, IFREMER, Plouzané,
lated from the results of the impregnated fibre bundle France
test (IFBT).3 In this last method, a unidirectional com- 4
Laboratoire CRISMAT, UMR 6508 CNRS/ENSICAEN, Université de
posite is loaded in the fibre direction, and the fibre stiff- Caen Basse-Normandie, Caen Cedex, France
5
ness and strength are back-calculated from the measured Université de Bretagne Sud, Lorient, France
composite properties, using straightforward microme-
Corresponding author:
chanical equations such as the rule of mixture. I Verpoest, Department of Materials Engineering, KU Leuven,
The IFBT is well established for carbon and glass Kasteelpark, Arenberg 44, 3001 Leuven, Belgium.
fibres and many companies refer to it in their technical Email: ignaas.verpoest@kuleuven.be
2 Journal of Reinforced Plastics and Composites 0(0)

A possible drawback of the IFBT is that the back- elementary fibres. As the number of elementary fibres
calculated properties depend on the validity of the varies over the length of a technical fibre, the diameter
micromechanical equations used. This is not a problem of a technical fibre will also vary over its length.13 This
for the stiffness where the influence of matrix and inter- cross-section variation renders the determination
face is negligible, at least when continuous fibres, or dis- of stresses, strength, and Young’s modulus problematic
continuous fibres with a very high aspect ratio, are used. when one technical fibre is tested in an SFT.14 Different
However, for strength, the influence of the matrix, methods have been proposed to measure the cross-
the possible (local) fibre misorientations, the fibre section variation, but the procedure is tedious and the
volume fraction and the fibre-matrix adhesion have to variation has a bigger effect on strength (sensitive to
be considered. The matrix and interface properties and the weakest cross-section) than on modulus (a property
fibre volume fraction will influence the mutual inter- averaged over the tested fibre length) of technical fibres.
action of fibre failures; hence, in order to provide On the contrary, elementary fibres can be tested indi-
detailed stochastic information on the fibre strength, vidually and have the advantage of a limited diameter
more elaborate data reduction schemes should be variation over their length.
used.6,7 In this paper, however, the simple ‘rule of mix- The main difference between glass and carbon fibres
ture’ will be used; it can be argued that the fibre on one hand and flax fibres on the other hand is the fact
strength values, back-calculated from the IFBT, can that the latter are not continuous. Whereas an elemen-
be considered as reliable values, when the fibre strength tary flax fibre can be considered as a continuous
data are used for predicting the properties of composite filament with limited length, the technical flax fibres
where matrices with similar properties and similar are composed of several elementary fibres assembled
interface strengths are used. Extrapolations to com- together to create a discontinuous long fibre. When
pletely different matrix materials have to be handled these technical fibres are subjected to tensile testing, it
with care. For instance, if the matrix would have a has been found that they have a lower ‘apparent’ stiff-
very low yield stress or interfacial strength, then the ness than elementary fibres.15 This is caused by the
rule of mixture would not apply directly and another elementary fibres sliding over each other because of
micromechanics calculation method (certainly for the the weak shear resistance of the pectin-rich layer
strength) needs to be used. (‘middle lamella’, see Figure 1).14 These shear stresses
are highest at the elementary fibre ends, which will
The particularities of natural fibres and their be present if the technical fibre length tested is longer
than the maximum elementary fibre length, which is
potential effects on the IFBT results normally the case. This adds an amount of strain to
Natural fibres, because of their complex morphology the intrinsic strain in the elementary fibres, which will
and microstructure, distinguish themselves from glass be dependent on the rigidity of the pectin-rich layer and
and carbon fibres which have continuous filaments with on the number of elementary fibres in the technical
a constant cross-section. The heterogeneity of the glass fibre. This sliding effect can be minimised by embedding
and carbon fibre microstructure happens at the nano- the technical fibres into a resin, like in the IFBT.
scale (for instance, orientation of graphitic layers in The outer elementary fibres will be adhered to the
carbon fibres), and hence the mechanical properties matrix,9,16,17 and hence their sliding relative to the
can be considered as homogeneous on the micro or other (outer and inner) elementary fibres will be limited.
fibre scale. Various types of defects will, however, In this way, this in situ behaviour of the technical fibre
create a scatter in the strength values, leading to the is much more realistic and relevant for composite appli-
well-known stochastic nature of the strength of cations than its behaviour when tested as an individual
carbon and glass fibres. On the other hand, natural technical fibre (without resin impregnation), as
fibres are not homogeneous as they have a complex, described above.
multiscale morphology as seen in Figure 1.8,9 As another alternative to the SFT, sometimes the
In composite applications, the technical plant fibres ‘‘dry fibre bundle test’’ or ‘‘unimpregnated’’ fibre
are mostly used to create continuous preforms. These bundle test is proposed. The pertinence of using IFBT
technical fibres are composed of several elementary testing instead of ‘‘dry fibre bundle test’’ for natural
fibres bonded together by a middle lamella mainly com- fibre arises considering the facts that:
posed of pectin.10,11 These elementary fibres have an
average length of 25–30 mm and an average diameter . It is practically impossible to carry out a ‘dry fibre
of 19 mm.12 Furthermore, they are composed of differ- bundle test’ on elementary fibres, as they have dif-
ent layers made of crystalline cellulose microfibrils with ferent lengths (<3 cm) and diameters. Moreover,
different orientations and embedded in an amorphous they are difficult to extract and to bring together
matrix. The technical fibres consist of 10 to 40 into a bundle.
Bensadoun et al. 3

Figure 1. The multi-scale structure of the flax fibre: (a) cross-section of the stem of a flax plant and of the position of the bundles of
elementary fibres, forming the technical fibres after extraction8 and (b) ESEM Image of a ‘‘technical fibre’’ and its constituting
elementary fibres.9

. Technical fibres are discontinuous, and as such, test- discontinuous fibres with a multi-scale microstructure, as
ing a bundle of dry fibres would inevitably mean that described above). It generates effective and average prop-
some fibres (an unknown number) are not loaded, erties, as in each sample hundreds to thousands of rein-
because they have their fibre end somewhere within forcing fibres are tested. Furthermore, the manufacturing
the tested bundle length. of the specimens and the tensile testing are common prac-
tice in composite labs.
Hence, the use of the impregnated fibre bundle A well-established standard is followed, enhancing
method on technical fibres becomes more interesting as in this way the confidence the composites industry has
all the fibre ends inside the bundle length will be loaded in the reliability of technical data provided by the flax
thanks to the interfacial shear stresses with the matrix. fibre producers. Finally, the IFBT can also be used to
assess flax fibres at different stages of the refinement
process (scutching, hackling, carding, combing, etc.)
The IFBT method for natural fibres for long fibres as well as tows and yarns.
The aim of this paper is to show that there are many
advantages in using the IFBT, from which the fibre stiff-
Motivation of the study
ness and strength are back-calculated. It should represent
correctly the behaviour of flax fibres and, by extension, of In 2013, the European Scientific Committee of the
other natural fibres in a composite (because they are also Technical Section of the European Confederation on
4 Journal of Reinforced Plastics and Composites 0(0)

Linen and Hemp (CELC)18 started the elaboration of separates the bast fibres from the shives, and hackling.
technical data sheet templates for flax and hemp fibre After these two steps, hackled tapes of long flax fibres
preforms (weaves, non-crimp fabrics, random mats, and tows were obtained; all fibres are parallel to each
yarn/roving, etc.).19 These technical data sheets follow other, and no twist is induced (as would be the case
a similar structure as the well-established ones for glass during spinning flax yarns used for weaving). It has
and carbon fibre preforms. When coming to fibre prop- to be noted that the sampling area of fibres in the
erties, or to properties of unidirectional tapes or rov- plant height is a parameter that influences their mech-
ings, the question arose which standardized method anical properties and that, if the fibres from only one
could be used, either the SFT (on elementary fibres) or area are tested, a reduction in the variability of strength
the IFBT (on technical fibres). For the scientific reasons and stiffness has been observed.24,25 In this research,
presented above, it was decided to investigate the possi- however, the hackled tape contains fibres from all pos-
bility of adapting the IFBT methodology standardized itions along the plant height, and hence homogenised
for continuous carbon filaments (ISO 10618:2004) to the or averaged properties with somewhat higher scatter
specificities of natural fibres, and in this particular case can be expected. A batch of 20 kg of tapes of hackled
to flax fibres.3,20 fibres was randomly split up into tapes of 10 meters.
A working group of five European laboratories sup- These hackled flax tapes were sent to each of the five
ported by CELC’s European Scientific Committee research institutes.
studied the existing ISO standard3 and proposed some
modifications to the specimen-manufacturing proced-
The matrices
ures, whereas the testing and data reduction were left
unchanged. Then, a limited round-robin experiment The IFBT was performed by several research institutes,
was set up, in which five European laboratories parti- thus, several polymer matrix resins were used. Their
cipated. One batch of hackled long fibres type Aramis compositions, curing cycles and mechanical properties
was chosen. This is a recent variety of spring flax, which are shown in Table 1. All resins have been chosen
was sowed for the first time in 2011 and has a good according to a number of criteria: a low viscosity to
fibre yield and good disease resistance.21,22 In 2013, the allow a good impregnation of the fibres and a good
Aramis variety represented 4.9% of the cultivated area fibre/matrix adhesion, which is an important parameter
of 58.500 hectares of flax in France.22 The choice of to obtain a good fibre-matrix stress transfer in the
only one batch of fibres is justified, because the aim composite.26–28
of this round-robin experiment was to verify that To stay close to the fibre performance in ‘‘real’’ com-
the method works independently of the operator. posites, where the matrix and interface are optimised
Such comparison cannot be made if different varieties towards optimum composite performance, the ISO
are used. standard advises to use a matrix (preferably epoxy)
Five laboratories manufactured the specimens and with a good adhesion to the fibre, and with a sufficiently
performed the tests and data reduction following as high failure strain.3 This means that the strain to failure
much as possible the ISO-standard with its proposed of the polymer matrix should be at least twice that of
modifications. The results of these tests are presented in the fibre, preferably three times. All resins chosen had
this paper and critically assessed taking into account Young’s moduli between 2.5-3.5 GPa with a failure
the differences in matrix resins and testing parameters. strain of at least 4% (beside HSB) and a low curing
Finally, they are compared to tensile tests on elemen- temperature (<150 C) to avoid degradation of the
tary fibres, carried out on the same batch of hackled fibres. In the Ensicaen laboratory, a vinyl ester was
long fibres. used. Including another type of resin, but still fulfilling
the requirements on stiffness and failure strain as stipu-
lated in the ISO-standard, allowed the influence of the
Experimental matrix type on the back-calculated fibre properties to
be assessed.
The flax fibres
In this study, hackled flax fibres type Aramis from the
Terre de Lin company (France) were used. The flax
Manufacturing of the IFBT samples
plants were harvested mature in 2013 in Normandy The IFBT on technical flax fibres has been performed
and then field retted (dew-retting). The dew-retted based on the ISO 10618:2004 standard for carbon
flax was selected by the producer for its good quality, fibres. In this research, bundles of unidirectional,
i.e. optimal retting, as it is known that the retting influ- hackled fibres are used instead of rovings (untwisted
ences the mechanical properties of the fibres.23 The yarns) and the manufacturing method has been slightly
retted flax subsequently underwent scutching, which modified following the parameters below.
Bensadoun et al. 5

Table 1. Thermoset matrix characteristics.


Tensile Tensile Hardener
Matrix Commercial modulus strength Strain at Density Hardener ratio in
Institute type name (GPa) (MPa) failure (%) (g/cm3) type resin (phr) Cure cycle Supplier


KU LEUVEN Epoxy Epikote 828 LVEL 2.7 70 4 1.16 Dytek DCH-99 15.2 2 h at 150 C Resolution
(amine) performance
products
UBS Epolam 2020 3 69 4.2 1.17 Epolam 2020 34 24 h at room Axson
hardener temperature +
3 h at 40 C + 2 h
at 60 C + 2 h at
80 C + 5 h at 100 C
IFREMER SR 1500/2505 3.1 71 5.4 1.13 SD2505 (amine) 33 4 h at 50 C Sicomin
HSB RIM 135 3 53 1.8 1.15 RIMH 137i 30 48 h at room Lange+Ritter
temperature GmbH
ENSICAEN Vinyester Atlac 580 3.5 83 4.2 1.11 MEKP Butanox 2 10 h at 40 C DSM
Urethane M50

First, the fibres were pre-dried for 24 h at 60 C, cut using the nylon cord. The pultrusion die was then
into a length of 25 cm and weighed in order to calculate closed on one side and was mounted vertically.
the fibre volume fraction. In method I, the fibres are Within 10 minutes, additional epoxy matrix was
placed in the mould cavity where a vacuum film was filled into the mould using a syringe. After 10 minutes,
previously laid (see Figure 2(a) to (c)) to create a uni- the mould was fully closed, inverted and re-opened on
directional composite. The resin is then degassed and the other side. More resin was injected for the next 10
poured on top of the fibres. The vacuum film is then minutes. Thereafter, the mould was closed and laid
folded and the counter-mould is placed on top. The horizontally. The fibre volume fraction of the sample
whole set-up is then positioned inside a heated press was set to 31.4%. The fibres have a nearly unidirec-
for consolidation. Once the curing is completed, the tional orientation in the parallel section of the pul-
samples are demoulded and conditioned for testing at truded test specimens. No further compression
21 C  2 C and 50  2% relative humidity. The set-up consolidation was applied using this technique. After
in Figure 2 is a modified version of the one studied by 48 hours at room temperature, the sample was cured
Coroller et al. and Trujillo et al.26,29 and the samples and demoulded.
are comparable in geometry. It has to be noted that all samples should have a
Because the mass of the fibres placed in the mould resin volume content of at least 35%,3 a good orienta-
is precisely controlled, the fibre volume fraction can be tion of the fibres in the unidirectional direction (0 ) and
controlled by imposing a precise specimen thickness, a limited porosity.
using spacers as displayed in Figure 2(c). The speci- This was indeed realised by the first method, where
men width is determined by the mould width. In this a wet impregnation was followed by compression to
way unidirectional composite samples with small cross consolidate the composite resulting in the required
sections have been manufactured, which normally volume fraction, a good fibre orientation and an
leads to a limited number of defects and a good con- acceptable porosity (&1.7% for UBS and <1% for
trol of the orientation of the fibres. They are moreover KU Leuven (KUL – University of Leuven)). The
distributed evenly throughout the composite, as shown void content in the IFREMER samples was limited
in the cross-sections (Figure 3) obtained by scanning to values below 0.1%. The void content of the
electron microscopy (SEM) and micro-tomography. Ensicaen samples was not measured, but because
All manufacturing parameters such as curing they used a processing method and a resin which
conditions and resin/hardener ratio are described were very similar to the one used by KUL and
in Table 1. UBS, it is believed that their porosity level will be
A second method, manual ‘‘pultrusion’’ was used similar as well. In the second, pultrusion-like method
by HSB to manufacture the samples (see Figure 2(d)). however, the lack of consolidation pressure could have
In this method, the long flax fibres, assembled together caused a misorientation of the fibres (due to the dog-
at one end by a nylon cord, were impregnated by bone shape obtained through pultrusion) and led to a
immersion in a bath, containing the epoxy resin, for slightly higher porosity content (2.9%). It has to be
at least 10 minutes. The impregnated bundle was then noted that the porosity levels were calculated by either
drawn through a pultrusion die with a dog-bone shape SEM, Micro-CT or density measurements.
6 Journal of Reinforced Plastics and Composites 0(0)

Figure 2. Impregnated fibre bundle test method I, (a) top view of six mould cavities and (b) Front view of one cavity of the mould
(spacer not included, the vacuum film surrounds all the flax fibre bundle), (c) longitudinal cross section view of the counter-mould with
the spacers and (d) side and top view of the pultrusion mould.
Bensadoun et al. 7

Figure 3. UD/epoxy IFBT sample cross-section (a) SEM cross-section from UBS and (b) micro-CT image of the KU Leuven
specimen.

Single fibre testing – Elementary fibres IFBT sample testing – Technical fibres
Single fibre tensile (SFT) tests on elementary fibres The tensile tests were performed in the longitudinal
were carried out and the results will be compared to direction of the IFBT samples. A gauge length of
the back-calculated properties obtained through the 150 mm was used. The samples were rectangular with
IFBT test. The elementary fibres were manually 2 mm thickness, 10 mm width and 200 mm length. At
selected and the testing was done according to the least 6 samples were tested by each laboratory in order
AFNOR NF T 25-501-1,30 AFNOR NF T 25-501- to ensure reproducibility and reliable standard devi-
231 and AFNOR NF T 25-501-332 standards. They ations. Only the HSB laboratory used 35 mm gauge
were adhesively bonded onto a paper frame with a length because of the dog-bone shape of 2 mm
gauge length of 10 mm. The average apparent diam- x2.8 mm x60 mm of the sample manufactured accord-
eter was determined under a microscope from 6 meas- ing to the DIN EN ISO 527-2 standard. All samples in
urements. The frame, as described in Trujillo De Los all labs were conditioned for at least 24 h at a controlled
Rı́os,33 was clamped on a universal MTS-type tensile- temperature of 23 C  1 C and a relative humidity of
testing machine equipped with a 2 N capacity load cell 50%  2% (RH ¼ 48% for UBS).
and a span length of 10 mm. The load was applied at a Abrasive paper was placed without any glue in-
constant crosshead-displacement rate of 1 mm/min up between the testing clamps and the samples to avoid
to failure, in a controlled environment with a tempera- slippage. However, in the tests carried out by UBS,
ture of 23 C and relative humidity of 48%. The deter- glass fibre composite end tabs were used. These were
mination of the tensile properties was carried out in bonded and cured for 24 h at room temperature followed
accordance with the NF T 25-501-2 (flax fibres) and by 14 h at 50 C. Finally, the HSB laboratory did not use
the NF EN 1007-4 (advanced technical ceramics) either end tabs or abrasive paper, because the samples
standards which take into account the compliance of had a dog-bone shape. The porosity content in the
the loading frame. Fifty valid tensile tests were per- samples (and the volume fraction) was calculated by
formed to ensure reliable results. In addition to the image analysis from the picture shown in Figure 3.
stiffness defined according to standard NF T 25-501- All remaining parameters are described in Table 2.
2, the stress-strain curves were studied to understand The calculations of the composite properties were
the evolution of tangent modulus with increasing made according to the standards ISO 527, DIN EN
strain, as seen in Figure 4. ISO 527-2 and ASTM D3039, which are standards
8 Journal of Reinforced Plastics and Composites 0(0)

Figure 4. Example of a tensile test on an elementary flax fibre. The dotted line is the behaviour (stress/strain curve), and the full line
is the tangent modulus. The red dot represents the yield point.

Table 2. Tensile testing parameters.

Specimen shape
and dimensions Strain Extensometer
thickness  width rate Load type and gage
Institute  length (mm) Standard Equipment (% min1) cell (kN) length (mm)

KU Leuven Rectangular ASTM D3039 Instron 4505 4 30 Clip-on 50


2  10  200
UBS Rectangular ISO 527 MTS Synergie 4 10 (preload 25
2  10  200 RT 1000 80 N)
IFREMER Rectangular ASTM D3039 Instron 5566 4 10 Clip-on, 50
2  10  200
HSB Dog-bone DIN EN Zwick Z020 2 20 (preload 50
4  10  15 ISO 527-2 10 N)
ENSICAEN Rectangular ASTM D3039 Instron 3366 4 10 25
2  10  200

for tensile testing of unidirectional and symmetric and The reason for calculating two values for the stiff-
balanced laminate composites. The main difference ness is that it has been observed that, in flax fibre com-
between these standards and the present characterisa- posites, a decrease in stiffness can occur at strains
tion method is that two stiffnesses are calculated from higher than 0.2%,34 this was also observed in our
the stress-strain data: experiments (see Figure 5, IFBT results for technical
fibres section). It is important to provide both values
. Ef,1: the stiffness is calculated between 0% and 0.1% as they can be relevant for design purposes. The
strain strength was calculated at the maximum strain and
. Ef,2: the stiffness is calculated between 0.3% and only the samples which failed between the grips were
0.5% strain deemed valid.
Bensadoun et al. 9

Figure 5. Stress–strain curve of flax-epoxy IFBT specimens from KU Leuven (Vf ¼ 52–62%), UBS (Vf ¼ 35–55%) and IFREMER
(Vf ¼ 25–29%). The zones were the moduli E1 and E2 are calculated are indicated.

(ISO 10618:20043). From the measured composite


Fibre volume fraction calculation properties on this impregnated fibre bundle, the fibre
The calculation of the fibre volume fraction (Vf) was properties can be back-calculated using the following
performed in the following way and according to the formulae, which are in fact the inverted formulae of the
ISO 14127:2008 standard:35 well-known ‘‘rule of mixture’’:36

mf EC  Em ð1  Vf Þ
f Ef ¼ ð3Þ
vf ¼ ð1Þ Vf
Vc
where mf is the mass of the fibres (mf ¼ vf  Vc  f Þ, VC  C   0 m ð1  Vf Þ
 u, f ¼ ð4Þ
is the volume of composite and rf the density of the Vf
fibres. A density of flax of 1.4 g/cm3 was used for the
calculations. This method was used by KUL, Ensicaen where Ef is the longitudinal modulus of the fibre, Em is
and UBS. A second method, used by HSB and the modulus of the matrix and Vf is the fibre volume
IFREMER, determines the Vf using the mass of the fraction, sf is the longitudinal strength of the fibre and
fibre and the mass of the composite with the hypothesis s’m the stress in the matrix at the failure strain, which
that there is no porosity (equation (2) can be calculated assuming linear elastic deformation of
the matrix up to the moment of fibre (and hence com-
mf posite) failure: s’m ¼ Em * "u, c. Equation (4) is valid
 only when the fibres fail first, meaning that they have a
Vf ð%Þ ¼ m mf  m  100 ð2Þ
f c f lower failure strain than the matrix. For this reason, the
þ
f res matrix needs to have a sufficiently high ductility.

Statistical analysis – One way ANOVA


Back-calculation of the fibre properties To evaluate in a ‘quantitative’ way whether the results
From the strength and stiffness of the impregnated fibre for each measured property were significantly different
bundles, the properties of the flax fibres, as they are or not, the data generated by the 5 labs, (i.e. stiffness
present and behave in a composite, can be determined, Ef,1 and Ef,2,and strength su,f) were statistically evalu-
using the methodology developed for carbon fibres ated using a a one-way analysis of variance statistical
10 Journal of Reinforced Plastics and Composites 0(0)

technique, ANOVA, with a confidential interval of Table 3. Tensile tests on elementary flax fibres.
95% (a ¼ 5%). The post hoc single-step multiple com-
Ef1 Strength
parisons (grouping) were done using the Tukey
Diameter (GPa) Ef2 s f,u
method. The calculations were performed using the (mm) (Etan) (GPa) (Mpa) " (%)
JMP 11 statistical software.37
Average 19 57 44.5 791 1.8
Standard 3 12.8 13.6 319 0.5
Results and discussion deviation

Elementary single fibre testing Note: The average fibre diameter and average mechanical properties are
listed. Etan max is the maximal tangent modulus for low strain represent-
As the unidirectional flax composite in the IFBT is con- ing Ef1. Ef2 is the tangent modulus for the second part of the curve
stituted of flax technical fibres and within them elemen- (strains > 0.5%) according to the standard NF T25-502.
tary fibres, tensile tests were carried out in order to
assess the behaviour of both constituents. The technical varying from 52% to 62% depending on the sample.
fibre properties, back-calculated from the IFBT results, In order to determine the back-calculated fibre proper-
are compared with the results from the SFT on elemen- ties, individually obtained by each of the participating
tary fibres. The tensile tests on elementary fibres are laboratories, equations (3) and (4) have been applied
delicate, due to their small diameter and their modest separately to the data sets of each of the laboratories,
length. It is not straightforward to keep the elementary while taking into account the fibre volume fractions of
fibres perfectly straight and aligned with great precision, each of the specimens individually.
once clamped in the tensile machine. A misalignment Figure 6 presents all the individual data on stiffness
could lead to an inaccuracy in the measurement of the and strength of technical fibres as a function of the
strain at small deformations. measured fibre volume fraction. It can be observed
Table 3 shows the average diameter and average ten- that these data follow a trend, as predicted by the
sile mechanical properties of the elementary fibres, rule of mixture for stiffness and strength, taking into
while Figure 4 shows the full stress-strain curve of an account an average modulus and stress at fibre failure
elementary fibre. A non-linear behaviour was observed, for the matrix of Eavg ¼ 3 GPa and s’m ¼ 35 MPa
thus confirming the need to calculate two stiffnesses respectively. When these equations are fitted through
according to the strain region. The moduli (Ef,1 and all the experimental data, the following average values
Ef,2) of the elementary fibres are calculated, using the are found for the fibre stiffnesses and strength: E1,f ¼
stress-strain curves and the modulus-strain curves pre- 59.8 GPa, E2,f ¼ 40.8 GPa and su,f ¼ 527 MPa. These
sented in Figure 4. values can be considered as an overall average for the
The yield point corresponds to the point coinciding back-calculated properties, and are represented in
with a drop in tangent modulus, i.e. the deviation from Figure 7 by the dotted horizontal lines.
linear-elastic behaviour. It is represented by the red dot in The results, presented in the diagrams in Figure 7(a)
Figure 4. The tangent modulus is calculated as the ratio and (b), clearly show that the difference between the
of the stress increment to the strain increment for succes- data of the five laboratories for the back-calculated
sive readings of the stress and strain. This tangent modu- fibre stiffnesses E1,f and E2,f is limited, as the average
lus (Etan) technique is also used to calculate the initial values fall within a scatter band of  5% around the
composite modulus, from which Ef,1 is back-calculated. overall average. This was expected as the modulus
The tangent modulus (Etan) value is not constant and is known to be less sensitive to imperfections in the
there is a maximum value for low deformations. Hence, fibre-matrix interface, in the fibres and in the compos-
the value of Ef,1 mentioned in Table 3 is the average of ites than the strength. As expected, the average value
the maximum values of Etan, measured in the initial strain for the stiffness at higher strains (> 0.3%) E2, f is about
interval (< 0.3%). For the calculation of Ef,2, the NF 31% lower than for the stiffness at very low strains
T25-501-2 standard31 was used. (<0.1%) E1, f, which is in fact the ‘real’ Young’s modu-
lus according to its strict definition. The reason for this
non-linearity, which starts at strains between 0.2 and
IFBT results for technical fibres 0.3%, has been related to several possible phenomena.
A set of typical stress-strain curves for the IFBT speci- When all elementary fibres in a technical fibre are well
mens with technical fibres, tested in tension, is pre- bonded to the matrix, the non-linearity should be
sented in Figure 5. also here, the non-linearity can be linked to the intrinsic non-linearity of the elementary
clearly observed, and the slope changes at a strain of fibres, which is mostly explained by the start of visco-
about 0.25%. It has to be noted that the scatter in plastic deformation of the amorphous materials inside
properties is caused by the variation in fibre volume the different lamellar layers or S-walls38 (see flax fibre
Bensadoun et al. 11

Figure 6. Composite properties vs. fibre volume fraction: (a) modulus and (b) strength.

internal wall structure in Célino et al.39), and which was resulting in a gradual increase of the stiffness, and
also observed in the elementary fibre test (see Figure 4). hence of the slope of the stress-strain-curve.
At much higher deformations (> 0.3%) this could grad- However, not all elementary fibres are well bonded
ually lead to a rotation of the crystalline cellulose to the matrix. When a technical fibre contains a large
microfibrils towards the loading (longitudinal) axis, number of elementary fibres, the interior ones are not in
12 Journal of Reinforced Plastics and Composites 0(0)

Figure 7. Back-calculated properties of the technical flax fibre: (a) modulus and (b) strength. Dotted horizontal lines (E1 - - - and E2
) represent the average values, as calculated by fitting the rule-of-mixture equation through all the experimental data.

contact with the matrix. Then a second mechanism can but a fundamental characteristic of the behaviour of the
lead to the decrease in stiffness and hence to the observed technical fibre inside a composite. Hence, the measured
‘‘yield point’’ (as seen in Figure 5): when some ‘interior’ E2,f is a correct representation of the behaviour of tech-
elementary fibres are bonded to their neighbours by only nical flax fibres in composites at higher strain levels.
a weak pectin-rich layer, shear deformation in these Thus, it can be stated that the IFBT will provide a reli-
layers could occur. This will reduce the contribution of able, albeit average, fibre stiffness for both E1,f and E2,f.
these weakly bonded elementary fibres to the resistance The variation in strength is higher than the stiffness
of the embedded technical fibre against deformation, variations, as shown in Figures 6(b) and 7(b)). This is
and hence to the measured and back-calculated stiffness not unexpected; it is also the case for traditional com-
of the embedded technical fibre. This is not an artefact, posites, because the strength depends on the
Bensadoun et al. 13

distribution of defect sizes and types. The reasons for pultrusion as described in section 2.3) and a higher por-
the systematic difference between the results of some of osity level (2.9%), which may cause a further reduction
the laboratories are discussed in the next section. of the longitudinal strength. Hence, the variation in
strength, observed between the different laboratories,
could be related to a difference in composite manufac-
Flax fibre properties - Statistical analysis
turing, leading to a different quality of the test samples.
In this study, it was observed that some of the fibre For instance, a small misorientation of the fibres during
properties, obtained by back calculation from the composite manufacturing could lead to a locally
IFBT data, presented a certain variation between increased stress, and hence to a decrease of the longi-
the five participating laboratories, as shown again in tudinal strength, which in turn leads to a reduction of
Figure 7. The question however remains whether this the back-calculated fibre strength.40
scatter is acceptable and whether there was a significant There may also have been small variations in retting
difference between the results of the different labora- and defects related to the hackling, although these
tories. In order to quantify this in a reliable way, a effects are expected to be minimal as the retting was
statistical ANOVA study was performed using carefully controlled and the scutching and hackling
Tukey’s grouping method (see section 2.8). Table 4 were performed under optimum conditions. Finally,
summarises the results obtained for the moduli (E1,f also the manipulation of the fibres by the operator
and E2,f) and the strength su,f. It was confirmed, as and the way the fibres are stored before the composite
was mentioned before in a qualitative way, that there manufacturing will play a role.
were no significant differences in modulus values (95% Because strength is a defect sensitive property, the
confidence level) for all laboratories, as they all share larger variation in strength suggests that even more
the letter A (Tukey’s grouping method, see Table 4). attention should be paid to a fully randomised fibre
As for the strength significant differences were found selection and a defect free specimen preparation.
between the laboratories: only KUL and Ensicaen seem
to be similar (sharing letter A), with UBS having a
slight but significantly lower strength value. Usually,
Comparison with results on elementary fibres
a large variation in flax fibre strength is attributed to Two points need to be considered if the correlation is to
the flax fibre extracting conditions and refining, the be made between the elementary fibre properties and
crop quality, the environmental conditions during the back-calculated properties of the technical fibres
growing, etc. . . In this study however, all the fibres from the IFBT. First, the span length of the elementary
came from the same batch, and hence these explan- fibre tests was 10 mm and it is known that this length
ations are unlikely. The low value for the IFREMER influences the mechanical properties and in particular
tests is probably related to the manufacturing method the tensile strength.26,33 Secondly, the technical fibres
as a mechanical clamping system rather than a press in the IFBT tests are a combination of several
was used to apply pressure, and this may encourage elementary fibres.
heterogeneous fibre distribution and lower strength. Comparing the back-calculated technical fibre prop-
The lack of consolidation pressure in the HBS sample erties obtained with the IFBT method to the elementary
could also have caused a misorientation of the fibres fibre properties measured by the SFT (see Table 5),
(due to the dog-bone shape obtained through some interesting observations can be made. First, the
scatter in the back-calculated data from the IFBT is
much smaller than the elementary fibre test data
Table 4. Statistical analysis of tensile properties of flax back-
calculated properties.

E1 E2 Strength Table 5. Elementary flax fibre properties vs back-calculated


technical fibre properties.
KU Leuven A A A
UBS A A B Strength
IFREMER A A C Ef1 (GPa) Ef2 (GPa) s f,u (Mpa)
HSB A A BC Elementary 57.0  13 44.5  14 791  319
ENSICAEN A A A fibre test
Back-calculated 59.8  2.4 40.8  3.5 527  138
from IFBT
Note: All the grouping was obtained from the Tukey’s method. A, B and C
are just names given to the three groups which have been distinguished IFBT: impregnated fibre bundle test.
when applying Tukey’s method. Note: Mean values and standard deviations are indicated.
14 Journal of Reinforced Plastics and Composites 0(0)

(+/- 4% vs. +/- 23% for E1,f). This is logical as the However, the back-calculated strength data from the
IFBT sample contains a few tens of technical fibres, and five laboratories show larger differences. The two
hence a few hundreds of elementary fibres and each laboratories presenting lower strength values used a
measured composite data point provides an average different manufacturing process, potentially resulting
value in itself. Even for strength, the scatter in the in a lower specimen quality in terms of fibre orientation
IFBT-derived values is about half of that of the elem- and fibre distribution. This highlights on one hand the
entary fibre test (+/- 26% vs. +/- 40%). necessity to prepare the samples with great care,
Second, the average values of fibre stiffness follow according to the proposed method. On the other
the same trend in both the IFBT and the SFT results, hand, it confirms that measurement of strength
Ef,2 being about 30% lower than Ef,1. It is interesting to always leads to a higher scatter, compared to stiffness
observe that Ef,1is almost identical for both test meth- as it depends on critical defects, whereas stiffness is an
ods, suggesting that, at these low strains and hence low average property where small defects are averaged out
loads, the elementary fibres in the technical fibres, and have much less impact. Once the data are com-
embedded in the IFBT-samples, do not yet slip over pared to the elementary fibre properties, the average
each other. This seems however to be different for the strength of the technical fibres is lower, although the
stiffness Ef,2 measured at higher strains. In this case the scatter in the SFT results for the elementary fibres was
IFBT results in about 10% lower values than the elem- higher.
entary fibre test. This is consistent with the assumption The IFBT-method is well suited for assessing in a
that, at higher strains, the elementary fibres inside the quantitative way the influence of the different extrac-
technical fibres, embedded in the IFBT-samples, can tion and refining steps on the effective stiffness and
indeed slip over each other, leading to a lower back- strength of flax and hemp fibres, and of other natural
calculated stiffness as previously mentioned in section fibres in composites. Finally, this method could be used
3.2. in the future for the characterization of the properties
Third, despite the large scatter, it can be stated that of twisted yarns in order to study the effect of twist on
the average strength of the elementary fibres is higher the yarn properties. These data can also be used as
than that of the technical fibres. This was not unex- input to model woven flax composites and predict
pected, because the technical fibres are composed of their performance according to the quality of the
shorter elementary fibres, and hence fracture can be yarns as proposed by Straumit et al.41
induced by debonding of the elementary fibres, starting
at their fibre ends. On the contrary, elementary fibres
can be considered as short but continuous filaments, Acknowledgments
and hence fracture has to be introduced by all kinds Special Thanks to the remaining members of the European
of defects (for instance at kink bands), leading to a Scientific Committee of the Technical Section of CELC: Hans
higher average strength but also a larger scatter. Lilholt, Joris Van Acker, Mark Hughes and Gerhard
Ziegmann, for their valuable input in the development of
this new standard, as well as to Julie Pariset (CELC) for the
Conclusions efficient coordination of this study. Many thanks to members
of the natural fibre team within the Composite Materials
In this study, the applicability and the reliability of the Group at KU Leuven, Eduardo Trujillo, Dieter Perremans,
IFBT technique to determine the flax fibre properties Aart W. van Vuure and others, for valuable discussions
by back-calculation from the composites properties was during the development of the IFBT method. The authors
investigated. The results show a very low scatter also acknowledge Birgit Uhrlaub, Fabian Uhrlaub and
between the five laboratories for the two stiffness Tim Dunter for the sample preparation and the tensile testing
values Ef,1 and Ef,2, back-calculated from the IFBT- of the composites. Special mention to Jean-Paul Trouvé from
results, typically less than  5%. These values are Terre de Lin for providing the flax fibres samples.
comparable with the stiffness values measured on elem-
entary fibres in the two strain regions. At low strains Declaration of Conflicting Interests
(>0.1%), Ef,1 values are very similar for both SFT and The authors declared no potential conflicts of interest with
IFBT meaning that the elementary fibres in the tech- respect to the research, authorship, and/or publication of this
nical fibres (embedded in the IFBT-samples) do not yet article.
slip over each other. The larger difference observed for
Ef,2 can be explained by the fact that in the IFBT-sam- Funding
ples not all elementary fibres (in particular the ones at The authors disclosed receipt of the following financial
the interior of the technical fibre) might be well bonded support for the research, authorship, and/or publication of
to the matrix, and hence the stress transfer to them is this article: This research is sponsored by the European
less efficient, especially at higher strains. Confederation of Linen and Hemp (CELC).
Bensadoun et al. 15

References 18. European Linen and Flax, www.mastersoflinen.com/eng/


1. Yao J and Yu W. Tensile strength and its variation for (accessed 2016).
PAN-based carbon fibers. II. Calibration of the variation 19. Pariset J. European flax/linen & hemp fibers,advanced &
from testing. J Appl Polym Sci 2007; 104: 2625–2632. natural material for the composite industry. JEC
2. Chi ZF, Chou TW and Shen GY. Determination of sin- Composites Magazine, March 2015, Vol 95, pp.34–6.
gle fibre strength distribution from fibre bundle testing. Paris: JEC 2015.
20. ASTM D2343 – 09. Standard test method for tensile prop-
J Mater Sci 1984; 19: 3319–3324.
erties of glass fiber Strands, yarns, and rovings used in
3. ISO 10618:2004. Carbon fibre – determination of tensile
reinforced plastics. West Conshohocken, PA: ASTM
properties of resin-impregnated yarn. Geneva: ISO, 2004.
International, 2009.
4. Toray Carbon Fibers. Technical data sheet – Torayca
21. Terre de Lin. Aramis, The new yield champion. http://uk.
T700s.
tdl.isagri-ingenierie.fr/internet-2015/art-of-linen/techni-
5. Swolfs Y, Verpoest I and Gorbatikh L. Issues in strength
cal-information/technical-information-1137.aspx
models for unidirectional fibre-reinforced composites
(accessed 2016).
related to Weibull distributions, fibre packings and
22. Saneco. Lin á usage textile en France - Evolution des
boundary effects. Compos Sci Technol 2015; 114: 42–49.
ventes de semences par variétés. http://www.saneco.
6. Swolfs Y, Morton H, Scott AE, et al. Synchrotron radia-
com/en/, 2015.
tion computed tomography for experimental validation
23. Martin N, Mouret N, Davies P, et al. Influence of the
of a tensile strength model for unidirectional fibre-rein-
degree of retting of flax fibers on the tensile properties of
forced composites. Composites Part A: Applied Science
single fibers and short fiber/polypropylene composites.
and Manufacturing 2015; 7: 106–113.
Ind Crops Prod 2013; 49: 755–767.
7. Swolfs Y, McMeeking RM, Verpoest I, et al. Matrix
24. Charlet K, Baley C, Morvan C, et al. Characteristics of
cracks around fibre breaks and their effect on stress redis-
Hermès flax fibres as a function of their location in the
tribution and failure development in unidirectional com- stem and properties of the derived unidirectional compos-
posites. Compos Sci Technol 2015; 108: 16–22. ites. Compos A Appl Sci Manuf 2007; 38: 1912–1921.
8. Baley C, Le Duigou A, Bourmaud A, et al. Reinforcement 25. Charlet K, Jernot J, Gomina M, et al. Influence of an
of polymers by flax fibres: Bio-Based Composites for High- Agatha flax fibre location in a stem on its mechanical,
Performance Materials: From Strategy to Industrial chemical and morphological properties. Compos Sci
Application. Boca Raton, USA: CRC press, 2014, p.87. Technol 2009; 69: 1399–1403.
9. Depuydt D. Characterization of the impact behaviour of 26. Trujillo E, Moesen M, Osorio Serna L, et al. Bamboo
flax fibre composites. Leuven, Belgium: Departement of fibres for reinforcement in composite materials: strength
Material Engineering, KU Leuven, 2013, p.129. Weibull analysis. Compos A Appl Sci Manuf 2014; 61:
10. Bos HL. The potential of flax fibres as reinforcement for 115–125.
composite materials. Eindhoven, Netherlands: Technische 27. Zhao F and Takeda N. Effect of interfacial adhesion and
Universiteit Eindhoven, 2004. statistical fiber strength on tensile strength of unidirec-
11. Bos HL, Van den Oever MJA and Peters O. Tensile and tional glass fiber/epoxy composites. Part I: experiment
compressive properties of flax fibres for natural fibre results. Compos A Appl Sci Manuf 2000; 31: 1203–1214.
reinforced composites. J Mater Sci 2002; 37: 1683–1692. 28. Lorenzo L and Hahn HT. Fatigue failure mechanisms in
12. Baley C. Analysis of the flax fibres tensile behaviour and unidirectional composites. Symposium on Composite
analysis of the tensile stiffness increase. Compos A Appl Materials: Fatigue and Fracture was held in Dallas,
Sci Manuf 2002; 33: 939–948. Texas, 24–25 October 1984.
13. Smole MS, Hribernik S, Kleinschek KS, et al. Plant fibres 29. Coroller G, Lefeuvre A, Le Duigou A, et al. Effect of flax
for textile and technical applications. Adv Agrophys Res fibres individualisation on tensile failure of flax/epoxy
2013; 10: 52372. unidirectional composite. Compos A Appl Sci Manuf
14. Charlet K and Beakou A. Mechanical properties of inter- 2013; 51: 62–70.
faces within a flax bundle part I: experimental analysis. 30. NF T25-501-1: Fibres de renfort - Fibres de lin pour
Int J Adhes Adhes 2011; 31: 875–881. composites plastiques - Partie 1 : terminologie et carac-
15. Perremans D, Hendrickx K, Verpoest I and Van Vuure térisation des fibres de lin. AFNOR standards, 2015.
AW. Investigation of the behaviour of flax fibre composites 31. NF T25-501-2: Fibres de renfort - Fibres de lin pour
in humid conditions. 2nd International Conference on composites plastiques - Partie 2 : détermination des pro-
Natural Fibers Sao Miguel, Portugal, 27–28 April 2015. priétés en traction des fibres élémentaires. AFNOR stan-
16. Müssig J and Haag K. The use of flax fibres as reinforce- dards, 2015.
ments in composites. In: Faruk O and Sain M (eds) 32. NF T25-501-3: Fibres de renfort - Fibres de lin pour
Biofiber reinforcements in composite materials. composites plastiques - Partie 3 : détermination des pro-
Cambridge (UK): Woodhead Publishing, 2015, pp.35–85. priétés en traction des fibres techniques. AFNOR stan-
17. Verpoest I, Baets J, Van Acker J, et al. Flax and hemp dards, 2015.
fibres: A natural solution for the composite industry, 1st 33. Trujillo De Los Rı́os E. Polymer composite materials based
edn. Paris, France: Prepared for JEC by the European on bamboo fibres (Polymere composietmaterialen gebaseerd
Scientific Committee of CELC – JEC Group/CELC, op bamboevezels). Leuven, Belgium: Departement of
2012. Materials Engineering, KU Leuven, 2014.
16 Journal of Reinforced Plastics and Composites 0(0)

34. Hughes M, Carpenter J and Hill C. Deformation and 39. Célino A, Fréour S, Jacquemin F and Casari P. The
fracture behaviour of flax fibre reinforced thermosetting hygroscopic behavior of plant fibers: A review.
polymer matrix composites. J Mater Sci 2007; 42: Frontiers in chemistry 2013; 2014; 43: 1–12.
2499–2511. 40. Berthelot JM. Effect of fibre misalignment on the elastic
35. ISO 14127:2008. Carbon-fibre-reinforced composites – properties of oriented discontinuous fibre composites.
determination of the resin, fibre and void contents. Fibre Sci Technol 1982; 17: 25–39.
Geneva: ISO, 2011. 41. Straumit, Ilya, S. Lomov, Farida Bensadoun, and
36. Gibson RF. Principles of composite material mechanics. Martine Wevers. Automatic transformation of 3D
Boca Raton, USA: CRC Press, 2011. micro-CT images into finite element models with aniso-
37. JMP. Oneway Analysis- Compare Means. http://www. tropic local properties. In: ECCM 16: European confer-
jmp.com/support/help/Compare_Means.shtml, 2012. ence on composite materials. 22–26 June 2014.
38. Giancane S, Panella FW and Dattoma V.
Characterization of fatigue damage in long fiber epoxy
composite laminates. Int J Fatigue 2010; 32: 46–53.

You might also like