XML Template (2011) [20.8.

2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

JOURNAL OF
COMPOSITE
Article M AT E R I A L S
Journal of Composite Materials
0(0) 1–15
! The Author(s) 2011
Recyclability and reutilization of Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
carbon fiber fabric/epoxy composites DOI: 10.1177/0021998311420604
jcm.sagepub.com

Paolo Feraboli1, Hirohide Kawakami1,2, Bonnie Wade1,

Federico Gasco1, Luciano DeOto3 and Attilio Masini3

Abstract
Solid carbon fiber/epoxy laminates manufactured by liquid resin infusion of twill fabric reinforcement are recycled in a
bath of boiling sulfuric acid to separate the fibers from the matrix. The recycled reinforcement consists of long fibers
arranged in a random, entangled mat. Using the same epoxy matrix and infusion materials and process, the recycled
fibers are reutilized to manufacture solid laminates. The physical properties of the recycled laminates are evaluated by
means of pulse-echo ultrasound, visual microscopy, and fiber volume content. The average fiber volume content of the
recycled laminates is 33%, compared to the 62% of the twill laminates. The mechanical properties of the recycled
composite include tension modulus and strength, compression modulus and strength, three-point bend flexure strength,
and short beam shear strength. The properties are compared against the values of the twill reinforcement with quasi-
isotropic stacking sequence. Results show that the recycled material offers promising elastic properties and strength
values, similar to those of advanced carbon fiber sheet molding compounds, and therefore can be used as structural
material.

Keywords
recycling, liquid resin infusion, mechanical testing

Introduction
because of the large volumes of parts comprised of
A wide range of composite material forms are finding CFRP that are entering service. The vast majority of
use in today’s aerospace, automotive, and other trans- CFRP for aerospace and automotive structural appli-
portation industry segments.1–5 These materials are cations consists of thermoset resins (typically epoxy and
finally fulfilling the promise of providing manufacturers vinylester), which cannot be re-shaped or re-formed
with a cost-competitive alternative to aluminum alloys. once processed. The chemical structure of thermosets,
The Boeing 787 Dreamliner, due to join the world’s which imparts them the desirable attributes of high
active fleet by 2010, features over 50% carbon fiber mechanical properties and environmental durability,
reinforced polymers (CFRP) by structural weight.1 is associated to an irreversible cross-linking process
Beside the direct benefits resulting from the greater spe- that takes place during curing. Hence, once a part is
cific mechanical properties, such as increased fuel effi-
ciency, and reduced pollutant and acoustic emissions,
1
other indirect advantages of a CFRP-intensive airframe Automobili Lamborghini Advanced Composite Structures Laboratory,
Department of Aeronautics & Astronautics, University of Washington,
are reduced maintenance requirements, and increased
Seattle, WA, 98195-2400, USA.
passenger comfort due to the superior fatigue- and cor- 2
Japan Ministry of Defense, Tokyo, Japan.
rosion-resistance characteristics of these materials. 3
Advanced Composites Research Center, Automobili Lamborghini S.p.A.,
However, the introduction of composites in the primary Sant’Agata Bolognese, Italy.
structure of modern aircraft presents new challenges,
Corresponding author:
such as meeting crashworthiness and lightning strike
Paolo Feraboli, Automobili Lamborghini Advanced Composite Structures
protection requirements, and developing a mainte- Laboratory, Department of Aeronautics & Astronautics, University of
nance, inspection, and repair strategy. End-of-life con- Washington, Seattle, WA, 98195-2400, USA
siderations are also becoming increasingly important, Email: [email protected]

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

2 Journal of Composite Materials 0(0)

shaped and cured, it is not easy to physically or technology but later discarded as the primary process
mechanically separate the matrix from the reinforce- because at a temperature of 500 C, the carbon fibers
ment. Even if careful effort is taken to reduce produc- retain oxidation residue or char.
tion waste during manufacturing, there is still a need to Another thermal process being proposed is catalytic
recycle CFRP material in order to avoid incinerating or conversion, whereby a depolymerizing catalyst is used
disposing it in landfills. to convert the cured resin into low–molecular-weight
Since the 1990s, industry has been investigating ways hydrocarbons.6–8 Adherent Technologies Inc. uses a
to recycle CFRP products.6,7 Boeing and other indus- heat transfer fluid (e.g., phenol), which is heated
trial organizations have come together in 2006 to estab- between 150 C and 300 C, to remove the bulk of the
lish a common industry working group, Aircraft Fleet resin and some contaminants. Vacuum pyrolysis is used
Recycling Association (AFRA), with a mutually shared as posttreatment to eliminate the remaining resin and
commitment to improving older fleet asset management contaminants, and achieve 99% purity. The process,
and fostering the recovery and reuse of aerospace mate- though still in the early phases of development by
rials. Boeing is working with AFRA-affiliated compa- Adherent Technologies, yields more pure carbon fibers
nies to develop technologies and processes capable of than pyrolysis. However, challenges arise if the parts
recovering CFRP parts after the end of their service contain scrap contaminants (such as metal, paint,
life. The recycling of composites is normally a two- wires), which are typically present in all structures.
step procedure that involves first, a mechanical separa- Within the framework of AFRA, shredded compos-
tion process of composites from other aircraft materials ite scrap from retired or damaged Boeing F/A-18 hor-
during an aircraft’s retirement, and second, a recycling izontal stabilizers was sent to both Milled Carbon and
process that recovers fibers of sufficient quality so that Adherent Technologies.6–8 The reclaimed AS4 fibers
they can be re-introduced as a materials source in underwent visual and X-ray surface quality assessment
manufacturing. To date, several recycling technologies plus single-fiber mechanical property testing. The
have been suggested for aerospace-grade CFRP mate- recycled fibers were chopped and successfully incorpo-
rials,8–10 which can be broadly divided into mechanical, rated into SMC, BMC, or as resin injection molding
thermal, and chemical processing. In mechanical pro- feedstock, and incorporated into fiber preforms for
cessing, the cured parts undergo crushing in a series of the compression molding of an automotive component
mills, through which the material is reduced to particles (Corvette C6 fender well).
ranging in size from 50 mm to 10 mm. In this process, all A third type of thermal processing being proposed at
the constituents of the original CFRP appear in the the laboratory scale uses fluidized bed treatments or
resulting recyclate, which is therefore a mixture of supercritical fluid treatments to successfully separate
matrix, and reinforcement. These recyclates are typi- the matrix from the fibers. In the fluidized bed treat-
cally reutilized as powder fillers in new composites, ment,8,9 the scrap composite is placed in a bed of silica
not for structural applications. For example, they sand fluidized with heated air at temperatures of 450–
have been used to replace calcium carbonate in com- 500 C. During this thermal process, the polymer matrix
mercial-grade sheet molding compounds (SMC) or material volatizes leaving recovered carbon fibers,
bulk molding compounds (BMC), the advantage of which are separated through a cyclone from the gas
the recyclates being that they are lighter than the mate- stream. In the supercritical fluid treatment,13 which is
rials they replace. a thermal and chemical process, the pieces of material
In thermal processing, it is possible to separate the to be recycled (approximately 25 mm long) are
fibers from the matrix, since the fibers have the most infringed by a supercritical flow (e.g., propanol or
recoverable value in a CFRP. The bulk of the efforts to glycol), at temperatures between 450 C and 500 C,
date have focused on pyrolysis and vacuum pyrolysis as pressure of 50 bars, and velocities up to 1 m/s, which
the primary process to extract fibers for reutiliza- removes the epoxy resin from the surface of the fibers.
tion.8–12 Heat is applied in the absence of oxygen, typ- The recycled product is fluffy and random, and com-
ically in furnaces using inert gases or under vacuum. prises individual fiber filaments with minimal contami-
Typical temperatures of 500 C and 600 C require sev- nation and mean length from 6 to 10 mm. Although
eral hours for the pyrolysis to take place. Single-fiber processed in air, the carbon fibers seem to show negli-
mechanical tests seem to show that contradicting results gible oxidation, and small reduction in surface oxygen
have been obtained, with some authors reporting more content, indicating a good potential for subsequent
or less degradation in strength. Milled Carbon Ltd6–8 bonding to a polymer matrix. The fibers exhibit a strength
uses a pyrolytic continuous flow process to burn off the degradation of 20%, while the modulus remains unaf-
resin and additives and separating the fibers, which can fected. Advantages of the fluidized bed process is that it
then be processed by chopping or milling for various is suitable to process parts containing mechanical
applications. Pyrolysis was initially deemed a viable inserts, surface paints, and foam cores. The recyclates

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 3

could be utilized as reinforcement for nonstructural currently used recycling processes. The precursor is a
composite applications such as BMC or nonwoven fabric, not tape as in the majority of the recycling stud-
veils. ies performed to date, and the recycling process gener-
Microwave, thermal shock, and molten salt bath ates a mat of long, entangled, random fibers which can
processes have also been used with partial success, be used as reinforcement without the need for second-
but they have a tendency to produce recycled fibers ary chopping or milling. Finally, it proposes a direct
with residues of resin or other contaminants on the comparison in terms of physical and mechanical prop-
surface.8,10 erties between the composites made of pristine fiber
Limited effort has been directed toward chemical fabric and those made of the reclaimed random mat.
processing. Little success was achieved in Jody et al.10 Using the same liquid resin and infusion process to
where a glycol/water solution at 240 C was used to manufacture the composite laminates with the pristine
treat composite scrap containing a urethane-based sub- and recycled reinforcements, it is possible to extrapo-
strate. A more successful approach was conducted in late the immediate influence of the architecture of the
ref.,14 where neat epoxy resin and epoxy-based compos- reinforcement on the laminate-level properties. These
ites were dissolved in a solution of nitric acid at tem- tests yield significant information on the possibility to
perature 80 C and concentrationof 4 M (mol/L). The use the recycled fibers as reinforcement for structural
extract was then neutralized and utilized subsequently applications.
for repolymerization. The study showed that it is pos-
sible to recover the epoxy resin following chemical pro-
cessing, and the thermal and mechanical tests show that Experimental approach
strength and glass transition temperature for the
Pristine material
recycled resin are much higher than the pristine resin
due to a higher density network structure. Another Parts are manufactured by vacuum-assisted resin trans-
promising method of recycling of carbon fibers from fer molding (VaRTM) at the Advanced Composites
carbon/epoxy composites with a nitric acid solution Research Center of Automobili Lamborghini S.p.A.
was demonstrated in Liu et al.15 Three variables were in Sant’Agata Bolognese, Italy. The parts include
controlled during the experiment: decomposition tem- embedded surface-treated aluminum inserts for bolted
perature (90 C), acid solution concentration (8 M), and connections and are representative of typical primary
ratio of specimen weight to acid solution volume automotive structures used by the super luxury car
(6 g : 100 mL). The epoxy matrix in the carbon/epoxy manufacturer (Figure 1(A) and (B)). Flat plates are
composites was completely decomposed in 12 h. Then, also manufactured using the same method in order to
the insoluble carbon fibers were separated from the obtain control specimens for comparative mechanical
solution and washed with acetone, and dried in an performance evaluation. Two different quasi-isotropic
oven at 60 C. The carbon fiber surface was clean, stacking sequences are investigated, yielding an upper
with very little contamination. Gel permeation and bound and a lower bound for flexural strength. These
gas chromatography showed that the epoxy resin are [(0/90)/(45)]2s and [(45)/(0/90)]2s. The reinforce-
could entirely decompose into low-molecular-weight ment used is a Toray T700 carbon fiber, 12 k tow, 2  2
compounds, with negligible contamination on the twill architecture, 380 gr fiber areal weight (FAW)
fibers. Electron probe microscopy showed no damage (Figure 2(A)). The dry fabric reinforcement comes in
to the recycled carbon fibers, and single-fiber tension a roll with a pre-applied binder, or tackifier, used for
tests showed that the strength loss of the recycled manual preforming or automated thermoforming
carbon fibers was negligible. (Figure 2(B)). The binder is present in a nominal con-
A summary of the studies performed to date on the tent of 5% by weight. It should be noted that woven
possibility to recycle and reuse carbon fibers shows that fabric comprises over 60% of Europe’s carbon fiber
the majority of the efforts call for chopping or milling waste.7 The resin used is a Huntsman two-part liquid
of the recycled fibers, with a particular effort aimed epoxy with XB 3518 BD resin and Aradur 22962 hard-
toward fine particulates, whiskers, and nonwoven mats. ener, in a mix ratio of 100:26 parts by weight (Table 1).
These materials are not well suited for structural appli- The untoughened resin is a low-viscosity system
cations, but may be utilized for their high thermally or designed for infusion and has a pot life of over 3 h at
electrically conductive properties. This study focuses on room temperature but only 15 min at 60 C. The oven
assessing the possibility to use reclaimed fibers for cure cycle calls for 1 h at 100 C with vacuum, followed
structural applications. The reclaiming process is by 2 h postcure at 140 C, giving a dry Tg of 135 C.
based on chemical solution of the matrix, which gener- The infusion process utilizes a distribution media,
ates pure recycled fibers, without residues, contami- which is essential to achieve good resin distribution in
nants, or surface modification, unlike many of the the panel. The size of the distribution media, which

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

4 Journal of Composite Materials 0(0)

Figure 1. A, B. Example of parts manufactured by vacuum-

assisted resin transfer molding (VaRTM) and containing aluminum
inserts. Figure 2. A, B. Dry fabric reinforcement and detailed view of
thermoset binder.

does not cover the entire preform, is determined by trial

and error. The preform coverage by the distribution test method utilized to determine the fiber volume con-
media is given by a complex interaction between pre- tent Vf of a composite material. The specimen is placed
form permeability, resin viscokinetic characteristics, in a glass acid- and heat-resistant container (i.e., a large
and part geometry, but unfortunately it can only be beaker); then sulfuric acid is added and the solution
determined through experimentation. Other infusion heated on a burner to 110 C. When the solution starts
disposable materials include breather, perforated and to boil, hydrogen peroxide is added to accelerate the
nonperforated barriers, vacuum bag, and sealant. A reaction and oxidize the matrix. The process takes place
caul plate is not used to manufacture the panels with under a fume hood to remove the harmful gases.
the pristine twill reinforcement. Mechanical stirring throughout the duration of the
entire process is essential. If the parts are left in boiling
acid without stirring, the separation of the matrix from
Extraction of fibers from the matrix the fibers does not take place, and only the surface of
The automotive parts are received at the Automobili the matrix gets removed. The process typically takes
Lamborghini Advanced Composite Structures several hours from beginning to end. When the diges-
Laboratory at the University of Washington, Seattle, tion is complete, the container is removed from the
WA, where they are cut into smaller sections burner in order to cool down, at which point the
(200  200 mm2) for ease of processing. This step is solution is separated from the reinforcement through
not required, but it is used due to the constraints of a series of filters. A vacuum pump is used to provide
the experimental setup that is described below. For the aspiration force for the liquid to go through the
industrial applications, this step would be avoided, series of filters. Unlike the actual acid digestion
which would streamline the operation and yield standard, which requires filters in the order of a
longer fibers. The procedure used in this study to few microns, for this purpose, it is possible to use
extract the carbon fibers from the CFRP parts is loosely much coarser filters, since the objective is to recover
based on the principles of acid digestion,16 which is a only the large fibers for subsequent reutilization. The

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 5

Table 1. Summary of test families

Family Fiber Architecture Stacking sequence Processing Epoxy resin/ hardener

A T700 12 k 2  2 Twill [(0/90)/(45)]2S VaRTM XB3518BD/ 22962

B T700 12 k 2  2 Twill [(45)/(0/90)]2S VaRTM XB3518BD/ 22962
R T700 12 k Long random fiber mat recycled Quasi-isotropic VaRTM XB3518BD/ 22962

Figure 3. Separation of fibers from resin and insert by acid digestion. The solution progressively changes color from transparent (A),
to brown (B), and to black (C).

filtration process requires approximately 30 min. After the weight loss at 5.7%, as an average of three mea-
filtration, the fibers are washed with acetone and dis- surements. The nominal content of binder is 5% by
tilled water to remove traces of sulfuric acid as well as weight, but a 1% variation is expectable according
other contaminants. They are subsequently dried in an to the manufacturer. The coloration of the mixture fol-
oven at 100 C for 30 min. lowing the acid digestion is black, as it was in the case
During the digestion, the transparent solution of sul- of the cured panels. Based on these results, the black
furic acid and hydrogen peroxide changes color from coloration can be attributable to a minor loss of carbon
light brown to dark brown, and eventually to ‘pitch’ fiber dust/powder that is left around the fibers after
black (Figure 3(A)–(C)). This suggests that not only processing.
the resin, which is transparent, but also other materials The result of the separation process is a mass of
are dissolved during the process. These may include the entangled carbon fibers, with a random distribution.
binder (also known as tackifier) or carbon fiber dust The mass of fibers is fluffy and fuzzy, with the majority
(microparticles of carbon fibers resulting from fiber of filaments maintaining the form of single tows (Figure
processing). It should be noted that the binder has an 4). The aluminum inserts are preserved and can be reu-
opaque, white color, and hence it should not be respon- tilized in other applications, though the protective coat-
sible for the black coloration. The carbon fibers are ing will likely need to be reapplied prior to embedding
insoluble in acid; so they do not contribute to the in the CFRP. The 2  2 twill architecture is mostly lost,
black coloration by mass loss. A test is performed to but it remains preserved in few areas (Figure 4).
measure the weight loss of the dry fabric reinforcement, The fibers are continuous and approximately 200 mm
which also includes the binder and the sizing. Following long, which is the size of the composite specimen panel
the steps of ASTM D3171,16 it is possible to measure from which they were extracted.

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

6 Journal of Composite Materials 0(0)

Figure 4. Carbon fiber reinforcements after acid digestion (a) and separation of aluminum insert (b); the fibers appear mostly as
single tows, but in a few places, the twill architecture is still visible.

Using a scanning electron microscope (SEM), it is not homogeneous, since the FAW varies from point to
possible to evaluate the surface purity of the recycled point. The variation in fiber density results in different
fibers, and compare them to the pristine ones. The pris- ‘preform’ thicknesses when subjected to vacuum pres-
tine fibers of the dry twill fabric exhibit a rough, irreg- sure. After infusion and curing, these regions result in
ular surface at 1000 (Figure 5(A)), which upon greater large thickness variations on the bag side. A 2-mm
magnification (Figure 5(B) and (C)) reveals to be asso- thick aluminum caul plate is used to reduce the thick-
ciated to a liquid or spray coating, most likely the fiber ness variation. The caul plate is perforated with
sizing or the binder. The recycled fibers are free of resin, 6.35 mm holes to allow for the vacuum to distribute
and exhibit a much more regular and smooth surface evenly across the preform, to facilitate resin flow, and
(Figure 6(A)–(D)). Very few traces of micron-sized to impart flexibility to the caul plate itself. However,
resin residue can be seen in few regions on the fiber since the FAW of the preform remains variable from
surface, but overall it can be confirmed that chemical point to point, the resulting fiber volume content of the
processing yields very pure fibers with little or no resin laminate varies accordingly. The setup under vacuum
residue.15 prior to infusion also shows line ports for resin injec-
tion and aspiration (Figure 7(B)). A total of three plates
for each family (A, B, and R) are manufactured by
Reutilization of fibers
VaRTM, having dimensions 300  300 mm2 and nomi-
The recycled reinforcement is used, without additional nal thickness 4.0 mm. The families are described in
processing, as the preform for VaRTM infusion. The Table 1, and they include the two quasi-isotropic stack-
same resin, infusion process, and materials are utilized ing sequences described above (families A and B) and
as in the case of the pristine twill architecture material the recycled mat (family R). The average measured
(Table 1). This ensures that the only difference between thickness for families A and B is 4.0 mm (0.2 mm),
pristine and recycled specimens is the reinforcement. It without the use of a caul plate. For family R, the aver-
is acknowledged that the binder is lost during this oper- age thickness measured thickness is 4.3 mm (0.4 mm),
ation, and for industrial purposes the recycled fiber mat with the use of a caul plate. A representative surface of
would have to be reapplied with the binder prior to the mold side of a composite panel of family A is shown
utilization. The infusion procedure for the recycled in Figure 8(A) and exhibits the familiar twill pattern. A
random mat is nearly identical to the one used for the representative mold side surface for the recycled com-
pristine twill reinforcement, with the difference that a posite panel, family R, is shown in Figure 8(B), and
semi-rigid perforated caul plate is used to obtain panels shows the long random fiber pattern with some evi-
of constant thickness (Figure 7(A)). The random mat is dence of the original twill architecture.

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:29pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 7

Figure 5. Scanning electron microscope (SEM) images (at different levels of magnification) of pristine carbon fibers show the
presence of a distributed fiber coating as well as discrete larger particles, both of which can be attributed to either the binder or the
sizing: (A) 1000, (B) 5000, and (C) 2300.

Results attenuation of the signal. For this material form, there

are point-to-point variations in material properties due
Comparative evaluation of the physical properties
to the random fiber distribution and the presence of
Nondestructive inspection is performed on 100% of the resin-rich areas, which generate signal attenuation
plates (Table 2) by means of pulse-echo ultrasound often not corresponding to the presence of a traditional
using a C-scan system with a 5 MHz sensor. Typical defect (void or porosity). Therefore, it becomes difficult
amplitude images are shown in Figure 9(A) and (B) to isolate the presence of a small defect from the back-
for families A and R, respectively. The multichromatic ground noise. These observations are consistent with
scale ranges from red, which indicates no signal loss, to the ones observed by the authors for chopped fiber sys-
blue, or complete loss of signal. The signal threshold is tems for compression molding.17,18
set at 6 dB loss, which corresponds to 80% of the Two micrographic coupons are extracted for each
emitted signal. Family A panel indicates a very homo- family (Table 2) in order to better understand the qual-
geneous structure, with yellow/orange/and red colora- ity of the infused panels as well as the overall meso-
tion (Figure 9(A)). Family R panel is comprised of a structure. The microscopic specimen is first vacuum-
diffused mix of colors (Figure 9(B)), which indicates a impregnated with the epoxy, then pressure-cured to
highly heterogeneous structure. The signal is noisy minimize formation of air bubbles. The specimen is
and difficult to interpret compared to family A. With polished with a six-step process: 180 grit, 600 grit,
ultrasonic inspection, any discontinuity in material 1200 grit, 9-micron, 3-micron silk, and 3-micron nonnap
properties, such as the presence of a defect, thickness polyester, ending up with 1 h polishing with a nonnap poly-
transition, or change of modulus is characterized by an ester cloth and 10% alumina solution.19 A representative

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:30pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

8 Journal of Composite Materials 0(0)

Figure 6. Scanning electron microscope (SEM) images (at different levels of magnification) of recycled carbon fibers show a clean and
smooth surface, with very limited traces of resin or coating residue: (A) 250, (B) and (C) 1000, and (D) 5000.

cross-section of a panel of family A (Figure 10(A)) shows which is collected in a catcher. After desiccation, the
an ordered distribution of the plies and tows, tight ply left over fibers are weighed to measure the remaining
compaction, low resin content, as well as absence of volume. The difference between the composite specimen
macro-voids (entrapped air). The cross-section of the volume before acid digestion and the volume of the
panel of family R (Figure 10(B)) shows uneven distribution leftover fibers gives a measure of the resin volume con-
of fibers and resin, a relatively planar distribution of fibers tent. It should be emphasized that this process is very
(two-rather than three-dimensional), odd-shaped groupings slow and lengthy, and selection of the appropriate fil-
of carbon fibers, and a much higher resin content. On the ters as a compromise between process time and mea-
other hand, there are very few voids, of small size, indicat- surement accuracy is key. The average values for the
ing a successful infusion process. four combined specimens extracted from families A and
From the composite panels, specimens weighing B indicate a fiber volume content Vf ¼ 62  1%, while
only 1 g are evaluated for fiber volume content by the average of the five measurements for family R speci-
means of ASTM standard D3171 for acid digestion16 mens shows a fiber volume content Vf ¼ 33  2%. The
(Table 2). After measuring the volumes of the compos- results are summarized in Table 3. The volume content
ite specimens by immersion in distilled water, the speci- for the recycled panel is extremely low, and future
mens are subjected to a procedure essentially identical research should aim at improving the fiber volume con-
to the one used to recycle the fibers. The residue is tent. This can be achieved by first disentangling and
cooled down in cold water and filtered in 15 mm ceramic rearranging the mass of fibers after the recycling oper-
filters using a vacuum pump to separate the liquid, ation with a series of rakes, thus generating a more

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:30pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 9

Figure 7. Vacuum-assisted resin transfer molding (VaRTM)

materials utilized in this study (A), and setup under vacuum at Figure 8. Visual aspects of the mold side of a VaRTM panel
infusion (B), also showing a sketch of the progression of flow. manufactured using the pristine twill reinforcement (A) and the
one using the recycled long random fiber reinforcement (B).

uniform mat. Second, during the preforming operation,

by reapplying a binder (tackifier) and thermoforming,
the recycled mat in a heated press. These preforming compression fixture,21 typically used for open-hole
operations will lead to a higher fiber volume content compression tests but used by Boeing for unnotched
and more uniform panel thickness distribution after compression tests as well.19 Flexure specimens are
curing. machined to ASTM D79022 three-point bend flexure
tests specifications. It should be noted that the support
span s for the three-point bend fixture is determined to
Comparative evaluation of the mechanical properties be s ¼ 40  t, where t is the specimen thickness. While
The panels obtained from recycled fibers have a for traditional composites, the error introduced by
random fiber distribution, which imparts them in- keeping the span constant based on the panel average
plane quasi-isotropic properties similar to the chopped thickness, for the recycled panel this assumption could
fiber systems17 used for compression molding. lead to gross errors. It is therefore necessary to pains-
Although this assumption has not been verified exper- takingly change the span for each specimen in order to
imentally in this study, the following discussion refers maintain the constant s/t ratio. Finally, the short beam
to tests performed on coupon machined in the 0 direc- shear specimen is machined to the specifications of the
tion, but applies to all four principal directions. From ASTM D2344 standard,23 using an s/t ratio of 4.0. All
the molded panels, tensile specimens are machined to specimens are loaded at a rate of 1.3 mm/min in a 2-grip
the ASTM D303920 straight-sided rectangular geome- Instron hydraulic tension/compression test frame.
try with dimensions 304.80  38.10 mm2. Since glass/ Modulus is measured using a 25.4 mm gage extensom-
epoxy fabric tabs of length 50.8 mm are bonded to eter. A total of five repetitions per family per test type
the specimen using 3 M Scotch-Weld film adhesive, are tested (Table 2).
the effective gage length is reduced to 203.2 mm. Tensile modulus results show that family R retains
Compression specimens have the same geometry, but approximately 55% of the modulus of the quasi-iso-
are left untabbed and placed in the ASTM D6484 tropic pristine twill material (Figure 12(B)). Specimens

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:30pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

10 Journal of Composite Materials 0(0)

Table 2. Summary of tests performed and repetitions

Family C-scan Microscopy Acid digestion Tension Compression Three-point bend flexure Short beam shear

A 3 2 2 5 5 5 5
B 3 2 2 5 5 5 5
R 3 2 5 5 5 5 5

specimens of families A and B fail by traditional brittle

fiber fracture, with a clean, more defined fracture front
which divides the specimen in two halves. Tensile
strength results, reported in Table 3 and Figure
12(A), show that family R retains approximately 25–
30% of the tensile strength of families A and B.
Although the as-measured results indicate that the
recycled material form is much less performing than
the pristine form, these results should be interpreted
carefully. First of all, the failure mechanisms are com-
pletely different between family R and families A and
B. For the recycled fibers, failure is matrix-dominated,
while for twill fibers it is fiber-dominated. Second, the
fiber volume content of family R specimens is nearly
half that of families A and B. With these consider-
ations in mind, it becomes important to understand
that with improved preforming operations, it is possi-
ble to obtain a more homogeneous recycled fiber
mat, with fewer thickness variations and higher fiber
volume content. If the results for family R (Vf ¼ 33%)
and families A and B (Vf ¼ 62%) are normalized24 to a
nominal fiber volume content of 50%, the results for
the recycled material become more appealing, with a
tensile strength of approximately 50% (Figure 12(A))
and a modulus of over 100% (Figure 12(B)) that of the
pristine material. Unfortunately, the normalization
process in this context cannot be used in such straight-
forward manner. While tension and compression prop-
erties are fiber-dominated for families A and B, they
are not so in the case of family R. Therefore, it is not
correct to use the standard normalization process, in
Figure 9. C-scan images of a vacuum-assisted resin transfer the same way that it cannot be used for other nonfiber-
molding (VaRTM) panel manufactured using the pristine twill
dominated properties, such as flexure, or short beam
reinforcement (A) and the one using the recycled long random
shear.24 However, it is expected that increasing the
fiber reinforcement (B).
fiber volume content and improving the homogeneity
of the preform will result in fewer and smaller resin-
rich areas, which are known to be the locations of
of family R fail in tension by shear of the intertwined failure initiation for chopped fiber systems.17,18 With
tows, which pull out from the matrix (Figure 11(A) that in mind, and based on the experience of the
and (B)). This resin-dominated failure mechanism is authors, it is more likely that we can expect a tensile
typically observed in discontinuous fiber systems, strength of 240–270 MPa and a modulus of 38.0–
such as the chopped carbon fiber SMC discussed by 40.0 GPa if a fiber volume content of 50% is achieved.
the authors in ref. 17. Some fiber breakage is observed, Compression failure consists of a similar shearing
but the main failure mechanism is not fiber-dominated. mechanism as the one seen in tension (Figure 13), but
After failure, the specimen retains partial integrity and takes the advantage of the higher resin compressive
remains somewhat intact. On the other hand, strength over the resin’s tension and shear strengths.

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:30pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 11

Figure 10. Optical microscopy images of a vacuum-assisted resin transfer molding (VaRTM) panel manufactured using the pristine
twill reinforcement (A) and the one using the recycled long random fiber reinforcement (B).

Table 3. Summary of test results – unnormalized

Family Test type Average thickness (mm) Average fiber volume (%) Strength (MPa) CoV (%) Modulus (GPa) CoV (%)

A Tension 4.0 62 758.4 0.7 51.0 1.5

Compression 4.0 62 450.9 1.1 46.2 3.1
Flexure 4.0 62 697.0 3.3 – –
Short beam shear 4.0 62 50.3 6.3 – –
B Tension 4.0 62 715.7 1.7 49.6 0.9
Compression 4.0 62 422.6 4.0 45.5 2.3
Flexure 4.0 62 522.6 3.6 – –
Short beam shear 4.0 62 52.4 4.7 – –
R Tension 4.3 33 196.5 3.2 29.0 15.8
Compression 4.3 33 190.3 5.9 27.6 9.5
Flexure 4.3 33 365.4 16.0 – –
Short beam shear 4.3 33 37.2 3.4 – –

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:31pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

12 Journal of Composite Materials 0(0)

Figure 11. (A) Postfailure image and details of tensile specimen extracted from a vacuum-assisted resin transfer molding (VaRTM)
panel manufactured using the recycled long random fiber reinforcement and (B) surface details showing failure morphology by shearing
and pull-out of tensile specimen.

As-measured compression strength results, reported in compressive strength for this material following an
Table 3 and Figure 14(A), show that family R retains improved preforming operation would lead to a com-
approximately 42% of the compressive strength of fam- pressive strength of 250–280 MPa and a modulus of
ilies A and B. It is typical for resin-dominated material 38.0–40.0 GPa if a fiber volume content of 50% is
systems, such as chopped carbon fiber SMC, that com- achieved.
pressive strength is equal or greater than tensile Flexural failure manifests on the tensile side, and
strength,17,25 which is in direct contrast with the typical propagates as a shear crack up through the thickness
behaviors observed for unidirectional tapes or fabrics, toward the compressive side (Figure 15). Three-point
whereby tension strength always greatly exceeds the bend flexure strength results for family R are 70%
compressive strength. Compressive modulus results and 52%, respectively, of the flexural strengths of fam-
show that family R retains approximately 60% of the ilies A and B, respectively (Figure 16). The flexure
modulus of the quasi-isotropic pristine twill material strength of the stacking sequence with the (0/90) ply
(Figure 14(B)). Normalization to a nominal fiber on the outer surface is expectedly higher than the one
volume content Vf ¼ 50% would suggest that the com- with the (45) ply. Normalization cannot be performed
pressive strength for family R would reach 80% of the for flexure strength, according to MIL-HDBK-17
strength of families A and B, while the modulus would guidelines,24 but the authors expect that flexure
exceed 100% of the pristine twill (Figure 14(A) and strength will increase significantly if the fiber volume
(B)). Following the same considerations, discussed for content is increased. It is expected that an average flex-
tension, it is expected that a more realistic value for ural strength of approximately 480–500 MPa should be

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:31pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 13

Figure 12. Tension strength (A) and modulus (B) results for all Figure 14. Compression strength (A) and modulus (B) results
three families. For each family, the light bar represents the as- for all three families. For each family, the light bar represents the
measured values, while the dark bar represents a ‘normalized’ as-measured values, while the dark bar represents a ‘normalized’
value based on a nominal fiber volume content of 50%. value based on a nominal fiber volume content of 50%.

Figure 13. Postfailure image and details of compressive specimen extracted from a vacuum-assisted resin transfer molding (VaRTM)
panel manufactured using the recycled long random fiber reinforcement.

achieved with a fiber volume content of 50%. of the same fiber and resin system.17,25 This is justified
Interlaminar shear strength for family R (Figure 17) by the absence of well-defined interlaminar planes, as
is 70% of the strength of families A and B. It is typical well as the pre-existing kinking of the ‘plies,’ which
of random fiber systems to have short beam shear leads to premature slipping of the ‘plies’ with respect
strength lower than the continuous fiber system made to each other. Increasing fiber volume content will

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:31pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

14 Journal of Composite Materials 0(0)

Figure 15. Postfailure image and details of compressive specimen extracted from a vacuum-assisted resin transfer molding (VaRTM)
panel manufactured using the recycled long random fiber reinforcement.

Figure 17. Measured short beam shear strength results for all
Figure 16. Measured flexure strength results for all three three families.
families.

likely not improve the short beam shear strength, but a the twill at 62%. The preform has lower permeability,
more homogeneous distribution of the reinforcement and is more difficult to infuse using the VaRTM pro-
(more flat and well-arranged recycled fiber mat) may cess, but the end quality of laminates is good in terms of
lead to some improvements. void content. Both physical properties (ultrasonic and
micrographic images) and mechanical properties (ten-
sion, compression, flexure, and short beam shear) of the
recycled panels show trends and traits similar to those
Conclusions
of advanced carbon fiber SMC. However, with further
A recycling process that allows for the separation of development efforts, it is possible to realign the recycled
carbon fibers from the epoxy matrix was demonstrated. long fibers and re-establish their original performance.
The process is based on immersion in boiling sulfuric Furthermore, by optimizing the degree of compac-
acid, similar to the ASTM acid digestion standard test tion and homogenization of the recycled fiber mat, it
method. The carbon fibers, which originally featured will be possible to increase the fiber volume content to
twill architecture, take the appearance of a fuzzy, values around 45–50%, and dramatically reduce the
entangled, random, long-fiber mat following the recy- thickness variation. Finally, though VaRTM is a suit-
cling process. VaRTM liquid resin infusion is used to able process for liquid resin infusion of the preform, a
manufacture panels of both the pristine twill fabric as pure RTM process will lead to further increases in the
well as the recycled fiber mat. The panels manufactured quality of the laminate thanks to the higher injection
using the recycled mat exhibit high thickness variations, pressures and control of both inner and outer mold
and a fiber volume content of 33%, much lower than lines.

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012

XML Template (2011) [20.8.2011–7:32pm] [1–15]
//blrnas1/Journals/application/sage/JCM/JCM 420604.3d (JCM) [PREPRINTER stage]

Feraboli et al. 15

Acknowledgments 13. Jiang G, Pickering SJ, Lester EH, Turner TA, Wonga
The authors thank Maurizio Reggiani (Sr Vice-President & KH and Warrior NA. Characterisation of carbon fibres
Chief Technical Officer of Automobili Lamborghini S.p.A.) for recycled from carbon fibre/epoxy resin composites using
supporting the study and for promoting the environmentally supercritical n-propanol. Compos Sci Technol 2009; 69:
conscious use of advanced composite materials. They also 192–198.
acknowledge Prof. Brian Flinn (UW Materials Science and 14. Dang W, Kubouchi M, Yamamoto S, Sembokuya H and
Engineering Department) for the help with the SEM images. Tsuda K. An approach to chemical recycling of epoxy
Graduate Students Francesco Deleo and Marco Ciccu are resin cured with amine using nitric acid. Polymer 2004;
also gratefully acknowledged. 43: 2953–2958.
15. Liu Y, Meng L, Huang Y and Du J. Recycling of carbon/
epoxy composites. J Appl Polym Sci 2004; 94(5):
References
1912–1916.
1. Miller A. The Boeing 787 Dreamliner. In: Keynote 16. ASTM D3171-09. Standard test methods for constituent
Address, 22nd American Society for Composites content of composite materials. Vol. 15.03, West
Technical Conference, September 2007, Seattle, WA. Conshohocken, PA: ASTM International, 2008.
2. Feraboli P and Masini A. Development of carbon/epoxy 17. Feraboli P, Peitso E, Deleo F, Cleveland T and Stickler P.
structural components for a high performance vehicle. Characterization of prepreg-based discontinuous carbon
Composites (Part B) 2004; 35(4): 323–330. fiber/epoxy systems. J Reinf Plast Compos 2009; 28(10):
3. Feraboli P, Masini A, Taraborrelli L and Pivetti A. Integrated 1191–1214.
development of CFRP structures for a topless high perfor- 18. Feraboli P, Cleveland T, Ciccu M, Stickler P and Deoto
mance vehicle. Compos Struct 2007; 78(4): 495–506. L. Defect and damage analysis of advanced discontinu-
4. Feraboli P, Masini A and Bonfatti A. Advanced compos- ous carbon/epoxy composite materials. Composites (Part
ites for the body and chassis of a production high perfor- A) 2010; 41(7): 888–901.
mance car. Int J Veh Des 2007; 44(3–4): 1–14. 19. Feraboli P and Miller M. Damage resistance and toler-
5. Feraboli P, Deleo F, Wade B, Rassaian M, Higgins M, ance of carbon/epoxy composite coupons subjected to
Byar A, et al. Predictive modeling of an energy-absorbing simulated lightning strike. Composites (Part A) 2009;
sandwich structural concept using the building block 40(6–7): 954–967.
approach. Composites (Part A) 2010; 41(6): 774–786. 20. ASTM D3039. Standard test method for tensile properties
6. Boeing. Aircraft & composite recycling. (2007 Boeing com- of polymer matrix composite materials. Vol. 15.03, West
mitment update). Environmental Technotes 2007; 12(1): 1–4. Conshohocken, PA: ASTM International, 2008.
7. McConnell VP. Launching the carbon fibre recycling 21. ASTM D6484. Standard test method for open-hole com-
industry. Reinf Plast 2010; 54(2): 33–37. pressive strength of polymer matrix composite laminates.
8. Allen BE. Characterization of reclaimed carbon fibers and Vol. 15.03, West Conshohocken, PA: ASTM
their integration into new thermoset polymer matrices via International, 2008.
existing composite fabrication techniques. Master Thesis, 22. ASTM D790. Standard test methods for flexural proper-
North Carolina State University, Raleigh, NC, USA, 2008. ties of unreinforced and reinforced plastics and electrical
9. Pickering SJ. Recycling technologies for thermoset com- insulating materials. Vol. 08.01, West Conshohocken, PA:
posite materials—current status. Composites: Part A ASTM International, 2008.
2006; 37: 1206–1215. 23. ASTM D2344. Standard test method for short-beam
10. Jody BJ, Pomykala JA, Daniels EJ and Greminger JL. A strength of polymer matrix composite materials and their
process to recover carbon fibers from polymer matric laminates. Vol. 15.03, West Conshohocken, PA: ASTM
composites in end-of-life vehicles. J Mater 2004; 8: 43–47. International, 2008.
11. Ushikoshi K, Komatsu N and Suqino M. Recycling of 24. MIL-HDB-17. Guidelines for property testing of com-
CFRP by pyrolysis method. J Soc Mater Sci Jpn 1995; posites: data Normalization. Rev. F, Vol. 1, Ch. 2, Par.
44(499): 428–431. 2.4.3, 2002, pp. 2–70–2–75.
12. Meyer LO, Schulte K and Grove-Nielsen E. CFRP-recy- 25. Hexcel Corp. HexMC technical datasheet. Dublin, CA:
cling following a pyrolysis route: process optimization Hexcel Corp., 2005.
and potentials. J Compos Mater 2009; 43(9): 1121–1132.

Downloaded from jcm.sagepub.com at UNIV WASHINGTON LIBRARIES on February 13, 2012