Composites Science and Technology 128 (2016) 94e103

Contents lists available at ScienceDirect

Composites Science and Technology

journal homepage: http://www.elsevier.com/locate/compscitech

Evaluating vibration assisted vacuum infusion processing of hexagonal

boron nitride sheet modied carbon fabric/epoxy composites in terms
of interlaminar shear strength and void content
Reinhold Meier b, Ilhan Kahraman a, A. Tugrul Seyhan a, c, *, Swen Zaremba b,
Klaus Drechsler b
a
Department of Materials Science and Engineering, Anadolu University (AU), Iki Eylul Campus, 26550, Eskisehir, Turkey
b t Mnchen (TUM), Faculty of Mechanical Engineering, Boltzmannstrae 15, D-85748, Garching,
Institute for Carbon Composites, Technische Universita
Germany
c
Composite Materials Manufacturing Science Laboratory (CMMSL), Research and Application Center of Civil Aviation (RACCA), Anadolu University (AU), Iki
Eylul Campus, 26550, Eskisehir, Turkey

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

Article history: The results of an investigation of the effect of voids on the interlaminar shear strength (ILSS) of carbon
Received 16 December 2015 fabric/epoxy composites consolidated from fabric preforms containing hexagonal boron nitride sheets
Received in revised form (h-BNSs) were presented. To produce the preforms, h-BNSs were rst dissolved in ethanol through
16 March 2016
sonication and the solution obtained was then used for through-the-thickness impregnation of carbon
Accepted 19 March 2016
fabrics. The composites with h-BNSs were produced by membrane added vacuum infusion process
Available online 21 March 2016
(MAVIP) with application of mechanical vibrations at various different low frequencies, including 10, 25
and 60 Hz. For the sake of property comparison, identical static samples were also produced. A facile
Keywords:
Hybrid composites
thermogravimetric analysis (TGA) based methodology was utilized for composite ber volume fraction
Porosity/voids and void content measurements. Analysis was carried out on specimen-by-specimen basis, and the
Thermogravimetric analysis (TGA) ndings obtained were correlated with the ILSS of the composites. It was determined that the com-
Mechanical properties posites produced at 10 Hz exhibited the lowest void content, whereas the highest one was obtained from
Low mechanical vibration the composites produced at 25 Hz. The highest ber volume fraction, density, and ILSS values were
obtained from the composites that were produced at 60 Hz. The ndings obtained were discussed,
placing particular emphasis on the interrelationships among ILSS, mechanical vibration frequency, and
composite ber volume fraction and void content.
2016 Elsevier Ltd. All rights reserved.

1. Introduction
Fiber-reinforced polymer-matrix composites (FRPCs) have
aroused great attention in various industrial applications due to
their high specic stiffness and strength. Conventional vacuum
infusion process (CVIP) has been employed to manufacture large
FRPCs [1]. In CVIP, the vacuum is used to draw resin in and air out to
reduce the incidence of voids in resulting laminates. However, even
despite additional attachments, such as bleeders and breathers,
evacuated air cannot be effectively removed from the resin used in
the process. On the other hand, a pressure gradient that develops
* Corresponding author. Department of Materials Science and Engineering,
Anadolu University (AU), Iki Eylul Campus, 26550, Eskisehir, Turkey.
E-mail address: atseyhan@anadolu.edu.tr (A.T. Seyhan).
during infusion causes a thickness gradient across the part, which
adversely inuences the ber volume fraction, void content, and
thus mechanical property variation across the part [1e6]. There-
fore, when it comes to producing composites for aerospace appli-
cations, CVIP remains vastly lacking in terms of repeatability of the
robustness. Therefore, vacuum bagging combined with autoclave is
used instead. Autoclave pressure aids vacuum in removing excess
resin while impeding the growth of voids, thus giving rise to nal
composite mechanical properties. However, this manufacturing
route is futile in terms of cost and processing time and thus
emerged as a driving force for the composite community to focus
on alternative out-of-autoclave (OoA) processing routes. In an
effort to near CVIP to autoclave process, it has been recently
modied with a gas-permeable membrane, a thin, exible lm or
fabric typically made from either polytetrauoroethylene (PTFE) or

http://dx.doi.org/10.1016/j.compscitech.2016.03.022
0266-3538/ 2016 Elsevier Ltd. All rights reserved.
 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
thermoplastic polyurethane (TPU), that permits uniform vacuum
distribution during infusion. It is expected that the use of such
membranes improve the consistency in the resin ow front and
removes dry spots at no expense of reducing vacuum during the
process [2]. More importantly, dry-spot removal during membrane
added vacuum infusion process (MAVIP) is accomplished without
optimized vent placement, as opposed to CVIP wherein it is inev-
itable. W Li et al. [2] studied the effect of a membrane layer on the
composites properties. For this purpose, they produced E-glass
reinforced epoxy composites through vacuum infusion with and
without a membrane layer. They showed that composite parts
produced by MAVIP exhibited a void content at a level comparable
to autoclave process, and that more undeviating ber volume
fraction, and less void content accompanied with 77% reduction in
overall thickness gradient were accomplished across the parts that
were produced by MAVIP.
Void formation is among the most common aws encountered
during the manufacturing of advanced composites. Voids form
primarily as a result of air that is entrapped in the resin itself during
the mixing [2e7]. Mechanical void entrapment during the lay-up in
a humid air is also possible. Moreover, leakage from the vacuum
bag, poor vacuum source, and application of inadequate tempera-
ture during resin cure cycle can also contribute to their formation
[8e11]. Voids are detrimental to particularly various matrix-
dominated mechanical properties of composites, including inter-
laminar shear, compressive, and exural strengths [10e13]. It was
also revealed that even if the same manufacturing process is used
for production of the two identical composites, their own resulting
mechanical properties would be different as a function of void
content. This is because the effect of voids on the composite me-
chanical properties is inuenced by shape, size and location of
voids. Inuence of void on the mechanical properties of fabric
reinforced composites was studied in the literature. Olivier et al. [3]
examined the effects of various curing pressures on some me-
chanical properties of carbon/epoxy laminates. Their prime target
was to obtain different void content levels as a function of cure
cycles performed at different pressure levels. For this purpose,
water content and temperature dependent viscosity values of the
prepregs were also taken into account to have various different void
contents in the resulting composites. Subsequently, they dened
the pressure conditions and produced unidirectional laminates
with void contents ranging from 0.3 to 10 vol%. They concluded
that, in addition to the exural modulus and strength, the com-
posite interlaminar shear strength was also reduced by 20% when
the void content reached 2 vol%. Hancox [4] studied the shear
properties of articial aw containing pultruded and lament
wound carbon ber/epoxy composite tubes and molded solid rods,
using a torsion rig. To increase void content, they soaked the carbon
bers in methyl ethyl ketone before saturated with the resin. They
found that solid rods were insensitive to aws, and that the pres-
ence of voids (up to 5% in volume) reduced the shear properties of
the composite tubes by 30%, as compared to the shear properties of
an identical void-free composite tube. They also concluded that, as
real void size and spatial distributions were incomparable, as
opposed to the way they were pre-assumed, the principles of
fracture mechanics cannot be properly used to mathematically
express or predict the shear properties of void containing materials.
Application of low frequency vibration during the processing of
composites has been considered a promising approach to altering
lling time, resin ow velocity, and the void content in nal com-
posites. Ghiorse and Jurta [5] subjected vacuum-bagged carbon -
ber epoxy pre-preg to low frequency vibrations (50 Hz) during cure
and compared the resulting void content, the number and size of
the voids, and the density of the laminates produced with and
without vibration. The void volume content of vibrated specimens
95
was found to be decreased by more than 50%. (2.5 and 1.2% for the
static and the vibrated samples, respectively). The number of voids
was found to be decreased by 59%, whereas the average void size
increased by 10% for the vibrated samples. 0.5 wt % increase in
density was found for the vibrated samples. They attributed the
increase in density to the increased ber volume content. J. Muric-
Nesic et al. [6] produced composite laminates of E-glass bers and
vinyl-ester resin by hand lay-up and then applied low mechanical
vibrations during cure. Their goal herein was to choose the most
appropriate from among the frequencies ranging from 10 to 50 Hz.
They also took into account the time for vibration application
during cure to minimize the number of voids in the nal part. As a
result, application of a long period of time (around 30 min) at a
frequency of 30 Hz emerged to be the most effective for reducing
the number of voids.
The major contribution of this study is to suggest using low
frequency mechanical vibration to produce ller modied com-
posites through MAVIP. To our best knowledge, this study is the rst
study in the literature that sheds a light on the mechanical vibra-
tion assistance in processing composites consolidated from pre-
forms with llers. Please note that we used (h-BNSs) as ller
constituent because there are many promising studies in the
literature that showed the improved mechanical and thermal
properties of the polymers with h-BNSs addition [8,9]. For pro-
ducing the preforms, a facile process based on the through the
thickness impregnation of the fabric plies was conducted. The
composites with h-BNSs were produced by MAVIP with the appli-
cation of mechanical vibrations at various different low fre-
quencies; including 10, 25 and 60 Hz. Static samples were also
produced. Fiber volume fraction and void content measurements
were successfully carried out employing a modest TGA based pro-
cedure. The ndings obtained were then correlated with the ILSSs
of the composites.
2. Experimental
2.1. Materials
Epoxy resin system (EPIKOTE Resin MGS RIMR 135 and EPI-
KURE Curing Agent MGS RIMH 1366) were purchased from
Momentive. They were mixed together by a jiffy mixer at the
specied weight ratio (100:30 w/w) for 5 min. The resin was
degassed for a relatively short time (10 min) afterwards, as the use
of the membrane layer was already shown to bring up a full vacuum
gradient combined with a continued degassing across the part
surface during infusion process [2]. T-300 carbon twill weave fab-
rics with an areal density of 200 g/m2 (balanced through weft and
warp directions) were provided by TELATEKS Corp. Turkey. Hex-
agonal boron nitride sheets (h-BNSs) were obtained from BORTEK
Corp, Turkey. Fig. 1a, b, c, and d are the images that give clues to the
morphology of the h-BNSs used in this study. Light microscopes
(Olympus BH2), SEM (SEM-Zeiss SUPRA 50 VP) at an acceleration
voltage of 15 kV, and TEM (TEM-Jeol2100F) at an acceleration
voltage of 200 kV were used to get the corresponding images. More
information regarding their synthesis and processing can be found
elsewhere [14].
2.2. Preform consolidation
A method based on through-the-thickness impregnation of the
fabrics was successfully conducted to consolidate the preforms
with h-BNSs. Fig. 2 presents the schematic illustration of the pro-
cess used for inltrating the carbon fabric preforms with h-BNSs in
this study. Ethanol solution containing h-BNSs (1 wt % of the total
fabric weight) was subjected to sonication for 2 h in a stainless steel
96 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103

Fig. 1. Morphology of h-BNSs a) SEM image b) TEM image c) In between carbon ber bundles after the through thickness impregnation d) TEM image at a higher magnication.

Fig. 2. Schematic illustration of the preform consolidation.

holder by employing a horn type sonicator (SONICS 750 VCF) at a
frequency of 20 kHz, running in a pulse operation mode (on 10 s, off
30 s) with the power set at 750 W. The fabric preforms composed of
20 individual plies with an initial thickness of (hi) were properly
placed in the sonicated solution by hand. The upper press plate
slowly squeezes the preform to the specied thickness (hf) at a
pressure of 0.25 MPa for 5 min. After consolidation, ethanol con-
taining preforms were placed in a vacuum oven at 80
C until they
become almost ethanol-free. This facile method is believed to be
superior over in-plane impregnation of the preforms under vacuum
wherein a solvent solution containing any ller is forced to wet the
preforms, at the end of which a serious ltration problem is likely to
 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
take place in most cases. Note that ethanol was chosen on purpose
as a solvent to keep the damage to the texture of carbon fabrics as
low as possible.
2.3. Composite production
To produce composites with h-BNSs, fabric preforms were rst
placed on a at surface of 10 mm thick aluminum mold followed by
their impregnation with the catalyzed resin under vacuum. Fig. 3
depicts the illustration of the experimental set up with the de-
tails of the processing constituents. Prior to each infusion, whether
or not there was any vacuum leakage was checked by closing the
valves and measuring the pressure increase inside the vacuum bag.
A maximum increase of 0.05 kPa per minute was allowed. Once the
infusion process was terminated, the inlet and outlet lines were left
connected to the vacuum pump to ensure constant part thickness. A
vibration engine (MVE 60/30 M) obtained from OLI-
Vibrationstechnik was mounted on the bottom side in the middle
of the aluminum tool. The rotational axis of the engine was
adjusted perpendicular to the resin ow front direction. The whole
system (mold and vibration engine) was supported by four coil-
springs to eliminate damping as much as possible and to guar-
antee a dened oscillatory movement. The frequency of the engine
is proportional to the frequency of the power supply and adjust-
ment was accordingly made using a frequency converter (BLEMO-
ER12). The vibration engine was mounted 10 mm away from the
aluminum tool to get rid of the engine-generated waste heat
through a cooling fan. Moreover, a thermocouple was placed in a
point on the tool surface, very close to the engine body in order to
monitor temperature rise, if any. Three different low mechanical
frequencies, including 10, 25 and 60 Hz were applied during vac-
uum infusion until the parts were fully cured. At least two com-
posite plates for each case were manufactured to accurately make
the statistical assessment of the ndings. After infusion, the com-
posite parts were allowed to cure at room temperature followed by
post curing at 80
C for 12 h.
2.4. Composite density, ber volume fraction, and void content
measurements
Composite density measurements were made according to
Fig. 3. Schematic illustration of the experimental-set up conguration.
97
ASTM D792-13. The density of each specimen was rst measured
and interlaminar shear strength measurements were then carried
out. The ber content of each was measured last. A facile TGA
based methodology was suggested to more accurately determine
the carbon ber content of the composite specimens. A TA In-
struments Q-500 Thermogravimetric Analyzer (TGA) was used for
this purpose. Tested ILSS specimens were sectioned into tiny cy-
lindrical shapes, large enough to t a TGA pan, to ensure the
maximum heat transfer surface area. Having searched for several
specimen sizes, we decided to use an average representative
specimen size of around 25 mg. The epoxy matrix, carbon ber
constituent, and the composite specimens cut in the required size
were then separately heated at constant heating rate of 20
C/min
under nitrogen atmosphere with a ow rate of 100 ml/min until
observing a relatively at region of the TGA due to the pyrolysis of
the epoxy resin. Uniform sample geometry was taken care of to
obtain reproducible results. For the complete removal of air from
the TGA heating chamber, (as its presence causes the oxidation of
the carbon bers and the residual weight in that case would be
lower than the heating in the absence of air), the furnace around
the balance of the TGA was purged with nitrogen for 15 min at
room temperature before the start of the heating program [15].
Through the vertically intersected point at the selected reference
temperature at which the matrix decomposition was pronounced
stable, by considering mass loss of carbon ber and composites,
resin matrix weight fraction, by extension ber weight fractions,
were computed according to Equation (1) where % Wresin content
refers to, in weight percentage, the amount of the resin present in
the composite sample and Mass losscomposite % and
Mass losscarbon fiber % refer to, in percentage, the lost mass of
composite and carbon bers, respectively, until at a temperature
after which mass loss rate remains unchanged and becomes sta-
ble. The ber volume fractions were then accordingly calculated
from the TGA data obtained for each sample by using the ber and
matrix densities (1.17 and 1.77 gr/cm3 for epoxy and carbon ber,
respectively). Void contents % Vv were then calculated for the
specimens subjected to the resin burn-out testing in the TGA
chamber by using Equation (2) where rcomposite refers to the
composite densities, experimentally determined according to
Archimedes' principle and Wmatrix and Wfiber are matrix and ber
weight fractions, respectively.
98 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103


Mass losscomposite %  Mass losscarbon fiber %
% Wresin content    
Mass lossepoxy matrix Mass losscarbon fiber
100  100
" " ##
Wmatrix Wfiber
% Vv 1  rcomposite x x100
rmatrix rfiber
For the sake of comparison among the composites produced
with and without vibration, samples for TGA measurements were
sectioned from adjacent locations in an attempt to capture the
same local constituent content for gaining more insight into vi-
bration induced effects on the composite properties. For the pur-
pose of measuring the accuracy of the corresponding method, ber
content of some randomly selected composite samples was deter-
mined through the matrix digestion method according to ASTM
D2734-09. Prior to acid digestion, the density of each sample was
measured according to ASTM D792-13 to keep the experimental
conditions the same as in TGA based methodology. For this pur-
pose, the whole ILSS specimen (40  6  6 mm), weight of which is
around 1.9 g, was reconsidered after tested. It was cut into ve
equal small parts in an effort to increase the surface area to ease
digestion. The samples were successively placed in the 500 ml
beaker and 300 ml of the nitric acid was poured in it under a
ventilated fume-hood. By the way, the extent of neat carbon ber
digestion (around 0.3 wt%) was found trivial and neglected in the
calculations. The beaker was heated afterwards, while being stirred
occasionally by the glass stirring rod to avoid sudden boiling of the
acid, until the acid fumed. Heating continued (for approximately
6 h) until the matrix was entirely dissolved and the sample dis-
integrated, eventually leaving bare carbon bers. Following ltra-
tion, the bers were several times washed with distilled water. The
bers obtained were then allowed to dry in a vacuum oven at 90
C
for at least 2 h and then weighed. Based on the ndings obtained
from the both methods, it was evaluated and discussed how
representative the TGA measurement was in comparison to acid
digestion measurement.
2.5. Short beam test measurements
The apparent interlaminar shear strengths of all the composite
specimens were determined by performing the testing according to
ASTM D2344-84. The SBS specimens 40 mm in length and 6 mm in
width were sectioned from the composite laminates. Prior to the
testing, all the samples were dried in a vacuum oven at 70
C to
ensure sample mass constancy to lower the noise in data. To
enhance the comparison accuracy of the ndings, the specimens
were taken from across the same regions of each composite plate.
The length to thickness ratio and span to thickness ratio were kept
constant at 7 and 5, respectively. Testing was conducted at a
crosshead displacement rate of 1 mm/min by using an Instron-
5944 universal testing machine. Load displacement curves for
each specimen were recorded until fracture. The maximum loading
point was then taken into account to calculate the ILSSs according
to the below equation where P is the maximum load recorded
immediate before fracture, while w and t are the specimen width
and thickness, respectively.
3P
t13
4wt
As outlined in detail in the introduction part, ILSS of composites
are highly vulnerable to void content and shape. Earlier studies in
the literature revealed that there are two possible void congura-
tions in composites, namely cylindrical and spherical. In parallel to
this, the theoretically derived equations for cylindrical and spher-
ical voids were further used to t the experimental ndings. They
all assume that the voids, regardless of cylindrical or spherical, are
(1) arranged in a regular square pattern within the composite part
[10,11].
2 30:5
(2)

4Vv
tcylindrical void Vv ; Vf tvoid free 41 
5 (4)
p 1  Vf
2 2 32 3

 p 6Vv
3
tspherical void Vv ; Vf tvoid free 41  4
5 5 (5)
4 p 1V
f
where tvoid free refers to the theoretical ILSS for the void free com-
posite. Assuming that there is considerable linearity between ILSS
and void content, the relation in between was represented by the
following regression line.
t a Vv b (6)
where a and b are constants. Applying the method of least squares
to the experimental values, a and b were predicted for each com-
posite. At least 20 samples were used for each case for the tting.
Here a stands for the ideal strength (the average values of ILSS for
the void-free compositestvoid free , whereas b stands for void con-
tent in the composites. The values for the composites produced
with and without vibration were predicted and interpreted in the
results and discussion section.
3. Results and discussion
Fig. 4 gives typical mass loss rate with respect to temperature for
epoxy, carbon ber and a sample of composite, separately. The
mineral ller content determination is relatively easy in composites
when the surrounding matrix resin undergoes a complete com-
bustion and no additional inorganic llers are present in the system
of interest. However, in contrast to mineral llers, carbon bers
decompose and their content cannot be therefore directly deter-
mined from the residue left, following complete decomposition of
the resin, even by switching from nitrogen to oxygen at a certain
elevated temperature. As seen in the gure, the carbon bers
initially lose weight up to 316
C (%1), which is most probably due
to the sizing accompanied by a small amount of moisture, but do
not show a decomposition drop-off yet, indicating that it combusts
at a higher temperature. The neat epoxy resin rst loses moisture
(2 wt% till around 300
C) followed by pyrolysis in a single step
process beginning from 373 through 595
C after which their mass
loss rate remains unchanged. On the other hand, the composite
material rst gives off a mass loss of about 2% due to loss of
moisture and then decomposes in a single step, as expected, since
the carbon ber does not start decomposing at the given temper-
ature range. The pyrolysis of the resin up to 595
C was taken as
reference to calculate the resin content. In fact, what is going on
here is burning off the different types of the carbon in sequence.
Pyrolysis of the epoxy matrix resin produces active carbon (carbon
black) with a large surface area that combusts at lower tempera-
tures than graphite (carbon bers) with a relatively great packing
(3) density. So the method utilized herein enables one to distinguish
between the carbon types unless the pyrolysis of the resin and the
bers falls into the same temperature zone. As the carbon com-
posites contain a certain amount of h-BNSs, their amounts were
 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
Fig. 4. Typical TGA curves of carbon ber, the matrix resin, and a composite sample.
also determined by combustion of the pyrolysis residues left from
the composites. The fact that more or less the same h-BNSs content
was found in the composite residues indicates that the preform
consolidation was productively executed. A correlation was made
with respect to moisture content of each constituent, and the
weight fractions of the resin, and the values computed on the dry
basis were taken into account. Fig. 5 gives the corresponding values
calculated according to the TGA based methodology together with
the ILSS of the composites. As seen in the gure, the ber volume
fraction values scattered more or less around the same value of 0.56
for static composites and those produced at 10 and 25 Hz, while
slightly increasing to 0.59 for the composites produced at 60 Hz. In
99
other words, ber volume fraction turned out to be 5% high at 60 Hz
compared to the other samples. This can be accounted for with an
improved nesting as a result of vibration assistance. When looking
at the measured density and void content values, similar tendency
is also observable such that ramping up the frequency to 60 Hz led
to an improvement in the corresponding values. By the way, it is
worth pointing out that the value for void volume content is
dependent on the measurement accuracy of both the theoretical
and actual densities of the composites. In other words, the
cornerstone of that approach is based on the use of Equation (2)
wherein small errors might bring about large errors in the pre-
dicted percent void volume in such a manner that change of 0.01 in
the actual density value would cause a change of 1% in the void
volume [15]. This is because the values given by the method for
void content are vastly sensitive to standard deviation of the
composite density. Regardless of the frequency applied, the void
and ber volume fraction contents obtained from the TGA based
methodology were, respectively, found to be 6 and 1.5% higher than
those obtained from the standard acid digestion methodology. This
implies that the utilized TGA based methodology entailing less
material and effort than acid digestion methodology proved to be a
highly reliable approach to determining the ber content. Provided
that accurate density data are available for the composite, the resin,
and the ber, the TGA based methodology can be safely used to
calculate void content as well as ber volume fractions. However, as
opposed to the conclusion of many studies [5e8,13] where a me-
chanical vibration was used for the preforms without any ller
involved, no linear relationship herein upraised between the fre-
quency level applied and the corresponding nal composite prop-
erties. For example, as the ber volume fraction increases, the
density of the composites is supposed to linearly increase too, with
the void content decreasing. Herein, the lowest void content
(0.69%) was obtained from the composites produced at 10 Hz,
whereas the highest one (1.68%) obtained from the composites
produced at 25 Hz. That trend can also be considered the same for
composite density values. It was concluded [6] that lower fre-
quencies, especially those close to the natural frequency of the
mold set-up are the most effective frequencies to be applied for
reducing the void content. The natural frequency of the experi-
mental test set-up used in this study without bers and vacuum
bagging was calculated as 7.3 Hz. In the practical tests, the highest
amplitudes were measured at frequencies around 9.5 Hz, which is
almost the same as 10 Hz, application of which the lowest void
inclusions were accomplished with. This implies that the resin ow
characteristics prove to be more complicated for the preforms with
ller than the preforms without any ller. It was concluded that the
driving force for void formation in composite processing is asso-
ciated with the heterogeneous nature of the preform at the micro-
structural level, where it is composed of arrays of ber bundles or
tows, which affect their permeability, besides the stacking
sequence selected [12]. Lundstrom et al. [13] demonstrated that the
major void formation cause is related to the different permeability
in the ber tows and pore space. During infusion, the resin ows
around the tows before permeating into the tows. This is because,
as pressure in the surrounding resin increases, the void is com-
pressed and enables resin to penetrate until equilibrium is ach-
ieved. From this point of view, one can review that preform
characteristics plays a pivotal role in formation, geometry, and
transportation of the voids. In our case, preform fabrics consist of
bundles of bers which most of the h-BNs were found penetrated
into. As is the case, the different channels that formed among the
bundles might have different directions and inconstant cross
sectional areas due to inhomogeneous distribution of llers among
the adjacent layers, whether inside or outside the bundles. This
would most probably bring about the ow front between the ber
100 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
Fig. 5. Fiber volume fractions, (from the TGA based methodology), void contents (according to ASTM D2734), and experimentally determined density of the composites (according
to ASTM D-792) produced with and without mechanical vibration, and the resulting ILSS of the composites.
bundles to slowly move forward in a random manner, thus
resulting in macro voids, as air would be entrapped as a result of
transverse ow from inside the bundles into the ller modied
channels among the ber bundles. Hence, the ow velocity be-
tween the ber bundles is most probably different than the ow
velocity within the ber bundles. To minimize voids, these two
corresponding ow velocities are to be the same. However, equal-
izing them is not as easy as it sounds when llers are present in
between the adjacent plies, as it would be vastly impossible to
predict whether the llers are located inside or outside the ber
bundles. In the case of mechanical vibration application during
infusion, this would create a boundary, resulting in a continuous
nesting across the part. On the other hand, buoyancy of the bubbles
might pose a more-than-expected inuence on the results ob-
tained, as the volume of the air bubbles that surround the h-BNSs is
vastly dependent on the extent of h-BNSs agglomeration. Note that
buoyancy increases with increasing bubble size and is affected by
the applied vibration such that the biggest bubbles occur at fre-
quencies very close to the natural frequencies of the vibrating mold
set-up where the vibration amplitudes become the largest [5e7].
This implies that the degree of h-BNSs agglomeration could be as at
least important as how homogenously they are distributed around
the tows across the part. In fact, level of the bubbles around and
within the agglomerates may vary depending on the number of
stacked layers of h-BNSs. Application of mechanical vibration is
capable to aid small voids in coagulating into bigger ones to make
possible their removal through vacuum port. Theoretically
speaking, once their diameter remains below the onset value, the
voids cannot be effectively removed and remain intact until the end
of the process [10e13]. From this perspective, the fact that the
composite void content is higher at 25 Hz than at 10 Hz might
witness evolution of the coagulated micro-voids associated with
the vibration induced wiggling of agglomerated llers. Neverthe-
less, compared to the static samples, application of vibration at
60 Hz that resulted in the highest ber volume fraction exhibited
also higher ILSS values with an acceptable void content. In brief,
incorporation of ller into the fabric preforms makes it more
complicated to entirely comprehend what is really happening in
the tiny rooms of the bundles. Parallel to this, ller incorporation
also stimulates a non-linear relationship between the mechanical
vibration application and the resulting composite void and ber
volume fraction contents. As reported in the literature before [4],
since actual void size and their spatial distributions do not near the
reection of the real picture, the principles of fracture mechanics
cannot be properly used to mathematically express and correlate
the properties of void containing materials, which would be even
more complicated for composites made with ller modied pre-
forms. In other words, it would be most probably anybody's guess
what if any other frequency was applied between 25 and 60 Hz in
terms of the corresponding properties. Fig. 6 gives the experi-
mentally measured ILSS of the composites (static and those pro-
duced at different frequencies) with respect to the predictions of
the models. As seen in the gure, regardless of whether they are
static or vibrated, the cylindrical void shape model constitutes the
upper bound, while the spherical void shape model does the lower
bound for the change of ILSS as a function of void content. It might
be theoretically concluded that the composites show propensity to
have spherical shaped voids rather than the cylindrical shaped
ones, as the predicted ILSS of the composites according to spherical
shaped void model tends to get lower in a similar manner as the
experimentally determined ILSS of the composites. In general, it is
established for investigations of the effect of voids on mechanical
properties of composites to show light microscopy images as well
as to perform ultrasonic studies. Both methods would help to get a
better understanding about the void distribution, whether in-plane
or out-of-plane [16]. In this study, in an attempt to delve into the
surroundings of shape and position of the voids present in the
composites, light (LM) and scanning electron microscopy (SEM)
images were taken into account. Theoretically speaking, voids
could be differentiated as interlaminar and intralaminar aws.
 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
Fig. 6. ILSS of the composites produced with and without mechanical vibration as a function of void content at different frequencies.
Interlaminar (delamination) entails fracture between the layers
whereas intralaminar fracture entails through-thickness fracture
among the bers. Intralaminar voids form among bers showing a
high length/width ratio while interlaminar voids are observed
located in and around resin-rich areas squeezed between ber
layers, ending up with an irregular shape [16]. Fig. 7a gives the LM
image of a void inclusion (prior to the testing) that were observed
squeezed between a longitudinal and transverse cut plies of the
composites, while Fig. 7b shows the SEM image of the same void
inclusion. Note that at the top the gure is the illustration of the
approximate positions within the sample (before and after the
testing) where the images were taken from (Each alphabet nearby
the cylindrical spots in the illustration corresponds to the same
alphabet designated to each image underneath it). This void shape
can be considered a good example to the region where an intra-
laminar fracture is likely expected to initiate from. A similar void
inclusion squeezed between two longitudinal cut plies was bum-
ped into after the testing (Fig. 7c, compression side), wherein the
micro-buckling initiates adjacent to the off-axis ply split, and
spreads across the band of 0
bers until intercepting the subse-
quent longitudinal ply split by ending up with the characteristic
kink-band morphology, as seen clearer in Fig. 7d. Note that these
two corresponding modes were, after the testing, typically ob-
tained from almost every composite produced, whether static or
vibrated. Fig. 7e depicts delamination arising from embedded void
defects at the ply interface, as seen clearer in Fig. 7f. This type of
failure (interlaminar failure) may most probably take place thanks
to the fact that the voids do not remain within the original defect
101
plane by displaying propensity to migrate through the laminate
until it reaches a preferential ply interface. Their migration degree
might be highly dependent on the applied frequency and the extent
of agglomerated h-BNSs left stuck in between the adjacent plies.
Fig. 7g is an example to the morphology of a nearly void-free
specimen, whereas Fig. 7h is the failure modes of that sample
that occurred after the testing. In tension side of the composite
specimen, delamination (interlaminar failure) seemed to be the
major failure mode, such that longitudinal plies were split into two
parts, as highlighted by the two red arrows. Through the zone
squeezed between two longitudinal cut plies, as highlighted in a
dashed rectangular shape, intralaminar fracture take places. At the
upper side is another delaminated zone, as highlighted in a smaller
dashed rectangular shape.
4. Conclusion
Utilization of low frequency mechanical vibration was pro-
posed to produce ller modied composites. Carbon fabric pre-
forms were impregnated through the thickness with h-BNSs.
Composites with h-BNSs were produced through MAVIP at three
different mechanical frequencies including 10, 25 and 60 Hz. The
highest ber volume fraction, density, and ILSS values were ob-
tained from the composites produced at 60 Hz. In fact, establishing
a mathematical expression or any simple linear relationship be-
tween the frequency level and the nal composite properties was
found more complicated when they are consolidated from pre-
forms modied with llers. This is because the extent of
102 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
Fig. 7. Light microscopy and SEM images showing the void inclusions and the corresponding failure modes around the voids.
agglomeration of llers, randomly distributed between the plies
across the composite part, and their inuence on the composite
mechanical properties are quite difcult to predict prior to vacuum
infusion. Fillers might act like natural void resource, as they are
capable to hold some air in their galleries, which is linked to the
degree of agglomeration. However, despite not so much substan-
tial, application of low mechanical vibration is worth applying
during cure of the composite parts consolidated from preforms
with llers. In the next study, we focus our attention on the ame
retardant and thermal conductivity behavior of h-BNS lled epoxy
resin, followed by investigation on whether or not low mechanical
vibration application has also impact on the ammability and
thermal response of carbon fabric/h-BNS lled epoxy resin
composites.
Acknowledgments
Authors thank Anadolu University (AU) grant numbered
BAP:1306F270 for funding this work to encourage a long bilateral
collaboration with Institute for Carbon Composites in Technical
University of Munich (TUM) in a future manner.
References
[1] M. Tanog lu, A.T. Seyhan, Investigating the effects of a polyester performing
binder on the mechanical and ballistic performance of E-glass ber reinforced
polyester composites, Int. J. Adhes. Adhes. 23 (2003) 1e8.
[2] W. Li, J. Krehl, J.W. Gillespie, D. Heider, M. Endrulat, K. Hochrein,
M.G. Dunham, C.B. Dubois, Process and performance evaluation of the vacuum
assisted process, J. Compos. Mater. 38 (2004) 20e27.
[3] P. Olivier, J. Cottu, B. Ferret, Effects of cure cycle pressure and voids on some
mechanical properties of carbon/epoxy laminates, Composites 26 (1995)
509e516.
[4] N.L. Hancox, The effects of aws and voids on the shear properties of CFRP,
J. Mater. Sci. 12 (1977) 884e892.
[5] S.R. Ghiorse, R.M. Jurta, Effects of low frequency vibration processing on
carbon/epoxy laminates, Composites 22 (1991) 3e8.
[6] J. Muric-Nesic, P. Compston, N. Noble, Z.H. Stachurski, Effect of low frequency
vibrations on void content in composite materials, Compos. Part A Appl. Sci.
Manuf. 40 (2009) 548e556.
[7] J. Muric-Nesic, P. Compston, Z.H. Stachurski, On the void reduction mecha-
nisms in vibration assisted consolidation of bre reinforced polymer
 R. Meier et al. / Composites Science and Technology 128 (2016) 94e103
composites, Compos. Part A Appl. Sci. Manuf. 42 (2011) 320e327.
[8] Ziyin Lin, Andrew Mcnamara, Yan Liu, Kyoung-sik Moon, Ching-Ping Wong,
Exfoliated hexagonal boron nitride-based polymer nanocomposite with
enhanced thermal conductivity for electronic encapsulation, Compos. Sci.
Technol. 90 (2014) 123e128.
[9] Hong Jun Ahn, Young Jun Eoh, Sung Dae Park, Eung Soo Kim, Thermal con-
ductivity of polymer composites with oriented boron nitride, Thermochim.
Acta 590 (2014) 138e144. http://www.sciencedirect.com/science/article/pii/
S0040603114003141-aff0005.
[10] L.B. Greszczuk, Effect of voids on strength properties of lamentary com-
posites, in: Proceedings, 22nd Annual Meeting of the Reinforced Plastics Di-
vision, Society of the Plastics Industry, 1967, pp. 20-A.1 to 20-A.10.
[11] S. Hernandez, F. Sket, J.M. Molina-Aldaregua, C. Gonz
alez, J. LLorca, Effect of
curing cycle on void distribution and interlaminar shear strength in polymer-
103
matrix composites, Compos. Sci. Technol. 71 (2011) 1331e1339.
[12] N. Pantelelis, A. Kanarachos, Vibration assisted resin transfer molding
(VIARTM), in: Proceedings of ICCM-13 Conference, 2001.
[13] T.S. Lundstrom, B.R. Gebart, C.Y. Lundemo, Void formation in RTM, J. Reinf.
Plast. Compos. 12 (December 1993) 1339e1349.
[14] Hilmi Yurdakul, Yapncak Go nc, Oya Durukan, Atakan Akay, A. Tugrul Sey-
han, Nuran Ay, Servet Turan, Nanoscopic characterization of two-dimensional
(2D) boron nitride nanosheets (BNNSs) produced by microuidization, Ceram.
Int. 38 (2012) 2187e2193.
[15] R.Y. Yee, T.S. Stephens, A TGA technique for determining graphite ber con-
tent in epoxy composites, Thermochim. Acta 272 (1996) 191e199.
[16] W.V. Liebig, C. Viets, K. Schulte, B. Fiedler, Inuence of voids on the
compressive failure behavior of bre-reinforced composites, Compos. Sci.
Technol. 117 (2015) 225e233.