- Email: [email protected]
Recycling of shredded composites from wind turbine blades in new thermoset polymer composites
Composites: Part A 90 (2016) 390–399
Contents lists available at ScienceDirect
Composites: Part A journal homepage: www.elsevier.com/locate/compositesa
Recycling of shredded composites from wind turbine blades in new thermoset polymer composites Justine Beauson ⇑, Bo Madsen, Chiara Toncelli, Povl Brøndsted, Jakob Ilsted Bech Department of Wind Energy, Technical University of Denmark, Risø Campus Frederiksborgvej 399, 4000 Roskilde, Denmark
a r t i c l e
i n f o
Article history: Received 2 February 2016 Received in revised form 8 July 2016 Accepted 10 July 2016 Available online 11 July 2016 Keywords: A. Recycling B. Mechanical properties C. Analytical modelling E. Vacuum infusion
a b s t r a c t As the energy produced from wind increases every year, a concern has raised on the recycling of wind turbine blades made of glass fibre composites. In this context, the present study aims to characterize and understand the mechanical properties of polyester resin composites reinforced with shredded composites (SC), and to assess the potential of such recycling solution. A special manufacturing setup was developed to produce composites with a controlled content of SC. Results show that the SC in the composites was well distributed and impregnated. The composite stiffness was well predicted using an analytical model, and fibre orientation parameters for strength modelling were established. The stress-strain curves revealed composite failure at unusual low strain values, and micrographs of the fracture surface indicated poor adhesion between SC and matrix. To tackle this problem, chemical treatment of SC or use of an alternative resin, to improve bonding should be investigated. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction One of the targets of the European energy and climate policy is to increase the energy consumed from renewable resources to 20% by 2020. As a result, the European Wind Energy Association (EWEA) predicts that the consumed wind energy will increase from 5,3% in 2010 to 15.7% by 2020 and will continue to rise to 50% by 2050 [1]. In 2010, 72,000 wind turbines [2] were providing electricity across the European Union (EU) countries, representing a capacity of 84,3 GW (GW) [3]. At the end of 2014, this capacity was raised to 128,8 GW [4]. Due to the rapid development of the technology, the blade length has grown from 18 m in a 600 kW turbine in year 1990 to around 85 m in a 8000 kW turbine in year 2015. Altogether, this development means that the mass of decommissioned wind turbine material is ever increasing [5]. While some of the wind turbine materials are recyclable (such as the metal parts in the tower, gearbox, and blade hub) [6,7], the blades made of thermoset polymer composites represent a challenge. As one of the central objectives of the EU are to maximise recycling and to phase out landfilling of non-recyclable waste, it is obvious that wind turbine blades need recycling solutions [8]. Wind turbine blades are typically made from continuous glass fibre reinforced polymer composites. E-glass fibres, a low cost glass fibre grade combining high strength (2 GPa) and moderate stiffness ⇑ Corresponding author. E-mail address: [email protected] (J. Beauson). http://dx.doi.org/10.1016/j.compositesa.2016.07.009 1359-835X/Ó 2016 Elsevier Ltd. All rights reserved.
(76 GPa), are used together with a thermoset polymer matrix such as epoxy, polyester or vinylester. Sandwich constructions, consisting of multiaxial composite laminates, together with balsa wood or polyvinyl chloride (PVC) foam, are used for the outer shell and for the shear webs in the blades, see Fig. 1. Uniaxial composite laminates are used for the load carrying beam. These glass fibre laminates have a typical fibre volume content of 50% (corresponding to a fibre weight content of 70%) [9]. Wind turbine blades have a predicted life time of 20–25 years. End of life solutions for wind turbine blades can be compared by the amount of re-processing needed. Few re-processing steps of the blades are needed to extend their service life. After 20– 25 years, where the blades still have some residual strength [10], they can be taken down, refurbished and re-installed [11]. Without too many further reprocessing steps, sections of blades can also be reused [12]. The field of such applications is however limited. The mechanical properties of the different composite materials in the blade are precisely designed to fulfil the requirements of a blade. In addition, the specific shape and dimensions of blade sections make their reuse in new structural applications challenging. Heavier re-processing of blades is costly, but widens the number of possible applications for the recovered materials. Such solutions include shredding the composite materials, or separating the glass fibres from the polymer matrix using thermal and chemical processes [11,13,14]. Shredding of composite materials, also called mechanical recycling, is the simplest method, and the only solution that has been brought to a commercial level [15–17].
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
391
Fig. 1. Wind turbine blade structure and material. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Mechanical recycling of composites consists of successive grinding or cutting steps to reduce the materials to fragments of few millimetres. The visual aspect of the resulting shredded composite (SC) is a mixture of short individual fibres, longer fibre bundles partly impregnated with matrix material, and clusters of materials, see Fig. 2. In the perspective of using SC in new polymer composites, previous work has found that high SC quality is crucial, and that the most valuable fractions are the ones with a high glass fibre content and with long fibres [16,18]. Research on methods to separate the different fractions present in SC has been conducted and techniques using air, gravity or vibration have been developed [18]. The reuse of SC in new thermoset polymer composites has mainly been studied in order to reduce the amount of virgin glass fibres in existing composite systems. This contributes to a lower cost of the product, without having to change the original composite manufacturing procedure [18–23]. So far, however, only a few studies have studied the performance of thermoset polymer composites manufactured with SC as the only reinforcing part [23–26]. In the context of assessing the potential of mechanical recycling of wind turbine blade materials, the objective of this study is to understand and predict the reinforcement effect of SC in new polymer composites. SC from uniaxial glass fibre composites, originating from a wind turbine blade, will be classified into fractions using manual sieving, and then characterized by combustion and microscopy. Composites will be manufactured using SC as the only reinforcing part. The gravimetric and volumetric composition of the composites will be determined. The mechanical properties of the composites incorporating various amount of SC will be measured, and analysed using micromechanical models for composites.
2. Materials and methods The shredded composite (SC) material in this study comes from the load carrying beam of a 26 m long wind turbine blade manufactured by LM Wind Power. Before shredding, the blade was subjected to dynamic testing from December 2000 to May 2001. Glass fibre composite laminates with dimensions of 200 mm long, 200 mm wide and 22 mm thick were cut out from the load carrying beam, see Fig. 1. The glass fibre volume content of the composite laminate before shredding was measured to be 50 vol%. The composite laminate was shredded using a Bolen Super Tomahawk (grinding process based on rotating cylinder hammer, main frame screen with 20 mm hole size). Fig. 2 shows the SC as received. A thermoset polyester resin was selected to manufacture the new polymer composites. This choice was made in order to avoid mixing different types of resins, since polyester is also the resin used in the composite laminate. An orthophtalic polyester resin (Polylite 413-575) was chosen, and it was pre-accelerated and initiated with 1.5 wt% of Butanox HBO-50.
2.1. Characterisation of SC The SC was divided into three fractions: a fraction containing the SC as received, and two fractions named fine and coarse, separated by manual sieving with a sieve of 1.5 1.5 mm square hole sizes. To determine the matrix coverage on the glass fibres, in addition to the diameter and length of the fibres, the three SC fractions were observed with a scanning electron microscope (SEM, TM1000 Hitachi). The glass fibre content of the SC fractions was determined by combustion, using a so-called burn off test. A material sample was placed in a crucible, it was weighed, and then burnt at 565 °C for 2 h, and finally, it was weighed again. The difference in weight of the sample before and after combustion, allows calculating the glass fibre weight content in the SC fractions.
2.2. Manufacturing of composites
Fig. 2. Shredded composite. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
In order to manufacture composites containing different contents of SC, a special vacuum infusion setup was implemented. The wanted weight contents of SC in the new polymer composites were set to 10, 20 and 30 wt%. In order to obtain these weight contents with a vacuum infusion process, the weight of the manufactured composite plates need to be controlled. For that, a rectangular metal frame with dimensions of 300 400 7 mm, and with holes along the short sides of the frame was used.
392
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
To determine the weight of SC needed to manufacture composite plates with final SC weight content of 10, 20 and 30 wt%, the density of SC was first calculated. To do so, a rule of mixture relationship for composite was used, see Eq. (1).
1
qSC
¼
1
qf
Wf þ
1
qm
The definitions and relationships for the weight contents in the SC and the manufactured composites are given by the following equations:
Weight content of fibres in SC : SC f ¼ Wm
ð1Þ
qSC is the density of SC, qf and Wf the density and weight content of fibre, qm and Wm the density and weight content of matrix. The density of SC was calculated assuming a glass fibre density of 2.55 g/cm3, a polyester matrix density of 1.18 g/cm3 and the measured fibre weight content. Knowing the density of SC and the wanted SC weight content, the density of the final composite was calculated using the same rule of mixture. As the volume of the mould is known, the weight of the final composite was predicted and the weight of SC needed to manufacture the composite deducted/derived from it. The pre-weighed amount of SC was manually placed within the frame and then covered with a 7.5 mm thick aluminium plate to close the frame volume, see Fig. 3. The layup was then packed with a vacuum bag, and vacuum was applied (60 kPa). The setup was tilted to a vertical position in order to promote a stable resin flow front, and the polyester resin was then infused into the bag. At the end of the infusion, the vacuum was lowered to 20 kPa to avoid bending of the aluminium plate. Finally, after curing at room temperature for about 18 h (overnight), the composite plate was demoulded and post-cured for 2 h at 80 °C. Nine composite plates were manufactured using the three SC fractions, and with three SC contents each. In addition, a pure polyester plate was manufactured.
2.3. Characterisation of composites The composite plates were characterized by their gravimetric and volumetric composition, their microstructure and their mechanical properties. To determine the fibre, matrix and porosity content of the composites, six rectangular samples of dimensions 10 20 mm were cut from each composite plate. The samples were roughly polished, and then dried overnight, and weighed. The samples were then sealed with paraffin wax to be able to include the correct porosity volume in the determinations. The sealed samples were weighed first in air, and then in water. Then, the polyester matrix in the samples was burnt off overnight at 565 °C. The weight of the remaining glass fibres was recorded. Based on these measurements, the glass fibre weight and volume content, Wf and Vf, the matrix weight and volume content, Wm and Vm, and the porosity volume content, Vp, were determined using standard equations [27].
mf Wf ¼ mf þ mm1 W SC
Weight content of old matrix in SC : mm1 SC m1 ¼ ¼ 1 SC f mf þ mm1 Weight content of SC in composites : mf þ mm1 Wf ¼ W SC ¼ mf þ mm1 þ mm2 SC f Weight content of fibres in composites : mf Wf ¼ ¼ SC f W SC mf þ mm1 þ mm2 Weight content of old matrix in composites : mm1 ¼ SC m1 W SC W m1 ¼ mf þ mm1 þ mm2 Weight content of new matrix in composites : mm2 W m2 ¼ ¼ 1 W SC mf þ mm1 þ mm2
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
ð7Þ
To characterize the microstructure of the composites, a sample with dimensions 20 10 mm was cut from each plate. The samples were casted in epoxy, and the sample cross-section in the plane normal to the thickness direction was grinded and polished. Observations were made using an optical microscope. To conduct static tensile testing, six dog bone shaped test specimens were cut according to ISO 527-2 from each composite plate. A tensile test machine with 50 kN load cell and hydraulic grips were used. The tests were performed with a displacement rate of 2 mm/min. Based on the measured stress-strain curves, the stiffness (the slope in the strain interval 0.05–0.25%) and the strength (the maximum stress value) were determined. In addition, the fracture surface of the tested specimens were coated with gold and observed by using a SEM. 2.4. Modelling of mechanical properties of composites The measured values of stiffness and strength of the composites were compared to theoretical model predictions. To predict the stiffness of composites with randomly oriented short fibres, the combined rule of mixtures model [28] was used:
Fig. 3. Special vacuum infusion setup used for manufacturing of SC composites. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
Ec ¼ g0 gl1 V f Ef þ ð1 V f ÞEm
ð8Þ
where Ec is the composite stiffness, V f is the fibre volume fraction, Ef is the fibre stiffness, and Em is the matrix stiffness. The effect of a non-unidirectional fibre orientation in composites was addressed by Krenchel [29] by introducing a fibre orientation efficiency factor g0. For a 2D random fibre orientation, this factor is equal to 0.375 (= 3/8), whereas it is equal to 0.20 (= 1/5) for a 3D random fibre orientation. The effect of short fibres in composites was addressed by Cox [30] by introducing a fibre length efficiency factor gl1, which takes into account the reduced efficiency of stress transfer from matrix to fibres, when the fibres are non-continuous:
gl1 ¼ 1
tanh bl2
ð9Þ
bl 2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi bl lu u Gm ¼2 t 2 d E ln k f V
ð10Þ
f
where l is the fibre length, d is the fibre diameter, Gm is the matrix shear modulus, and k is a constant that depends on the geometrical packing of fibres. To predict the strength of composites with randomly oriented short fibres, a modified Kelly-Tyson model was used [31,32]:
rcu ¼ go gl2 V f rfu þ ð1 V f ÞEm rfu =Ef
ð11Þ
where rcu is the composite strength, and rfu is the fibre strength. As for the prediction of stiffness, the non-unidirectional fibre orientation is addressed via the parameter go . In the Kelly-Tyson model, the fibre length efficiency factor (gl2) is defined as:
(
nl2 ¼
l 2lc
;
for l < lc
1 2lcl ;
for l P lc
ð12Þ
where lc is the critical fibre length defined by lc ¼ rfu d=2si where si is the interfacial shear stress between fibres and matrix. The KellyTyson model is describing the situation of brittle fibres in a ductile matrix, i.e. efu < emu, where e is strain. The model is based on the failure criterion that the composite is failing when the fibres are failing. As can be seen from Eq. (11), the stress in the matrix at fibre failure is calculated by rm ¼ Em rfu =Ef , assuming a linear stressstrain behaviour of the matrix. 3. Results and discussion 3.1. Characteristics of SC and the manufactured composites The fine fraction of SC obtained by manual sieving is shown in Fig. 4A. It contains mostly short glass fibres, in addition to some bundles of long glass fibres, and some pieces of polyester matrix. From the SEM images, the average fibre length for this fraction is
393
estimated to be around 0.8 mm. The material retained on the sieve contained two types of fractions: the coarse fraction, see Fig. 4B, and some clusters, see Fig. 4C. The shredded composite was found to be made of 52 wt% of coarse fraction, 44 wt% of fine fraction and 4 wt% of clusters. The clusters were manually removed to get the coarse fraction. In principle, the clusters, made of agglomerated short glass fibres, can be crushed and added to the fine fraction, but their manual breaking is a time consuming process. Therefore, it was decided to discard them. The coarse fraction consists of long glass fibre bundles, and pieces of polyester matrix. By visual inspection, the average fibre length in this fraction was approximated to be 30 mm. The SC fractions were observed with SEM. It was found that large pieces of matrix were attached to the fibre bundles and to the stitching thread (Fig. 5), and that smaller pieces of matrix were attached to the surface of the glass fibres (Fig. 6). The burn off test revealed a similar glass fibre content in the range 72–74 wt% for the fine and coarse fractions (Table 1). Regarding the manufactured composites, the three wanted SC weight contents of 10, 20 and 30 wt% were all almost achieved, except for the composite plates with the wanted weight content of 30 wt%, where the obtained weight contents are slightly lower, see Table 1. For these relatively high fibre contents, the vacuum pressure applied during the manufacturing process was not able to properly close the volume within the metal frame, and this led to a larger volume (and mass) of the plates than expected. Regarding the porosity content, it was estimated to be around 1 vol% for all composites, which is an indication of good quality composites. In Table 1, the SC weight content (WSC) is calculated by Eq. (4) using local measurements of the composite fibre weight content (Wf), based on small samples, and the measured fibre weight content of the SC (SCf). These local values of WSC can be compared to the measured global weight content of SC in the composites obtained by the pre-weighed amount of SC used in the manufacturing process, and the final weight of the composite plate. These values are presented in Table 2. The two sets of results are similar to each other, which is indicating that the SC are well dispersed over the whole composite plates. Fig. 7 shows representative images of the microstructure of the composites with the three SC fractions (as received, fine and coarse). The micrographs show that the glass fibres are well impregnated by the matrix, and with no apparent porosities. It is not possible to distinguish the old matrix from the new one, since the appearance of the matrix is uniform. The few black spots in the micrographs are believed to be due to fibres that have been pulled out from the surface during polishing. The results of the tensile testing are reported in Table 3. The stiffnesses of the composites, ranging from 4.2 to 5.8 GPa, are all higher than the stiffness of the pure polyester plate on 3.4 GPa. For each type of SC composite, the higher the fibre content, the higher the stiffness. Regarding the strength, all composites have a lower tensile strength than the one of the pure polyester plate.
Fig. 4. Images of the obtained SC fractions (A) fine, (B) coarse, (C) clusters. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
394
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
Fig. 5. SEM images of glass fibre bundles in the SC fractions.
Table 2 Local and global weight contents of SC in the manufactured composites. SC fractions
SC weight content in composites [wt%] Local
Global
As received
13 19 24
11 20 27
Fine
12 21 24
11 19 25
Coarse
11 24 29
11 21 30
Fig. 6. SEM image of a glass fibre with small pieces of polyester matrix attached to the surface.
content, whereas the failure strain of the neat polyester plate is equal to 1.7%.
The strengths of the composites are ranging from 15 to 29 MPa, whereas the strength of the neat polyester plate is equal to 49 MPa. Similarly to stiffness, the strength of the composites is increased with the fibre content. Regarding the failure strain, the composites reach significantly lower failure strains than the one of the pure polyester plate. The failure strains of the composites are ranging from 0.3 to 0.6%, with no apparent influence of the fibre
3.2. Modelling of mechanical properties of composites For the prediction of composite stiffness using Eq. (8), the values used to calculate the fibre length efficiency factor (gl1) are: Ef = 78 GPa (as measured on the same type of glass fibres in the study by Beauson et al. [33]), Gm = 1.2 GPa (as calculated from the measured tensile stiffness of 3.4 GPa for the isotropic polyester
Table 1 Characteristics of SC fractions, and the manufactured SC composites. SC fractions
SC characteristics
Composite characteristics
Fibre length [mm]
Fibre content [wt%] SCf
Density [g cm3]
As received
0.8–30
73
Fine
0.8
Coarse
30
Gravimetric composition [wt%]
Volumetric composition [vol%]
SC Wsc
Fibres Wf
Matrix Wm
Old matrix Wm1
New matrix Wm2
Fibres Vf
Matrix Vm
Porosity Vp
1.22 ± 0.02 1.26 ± 0.02 1.29 ± 0.02
13 19 24
9±2 14 ± 3 18 ± 3
91 ± 2 86 ± 3 83 ± 3
3 5 6
87 81 76
4±1 7±1 9±2
94 ± 1 92 ± 1 90 ± 2
1.5 ± 0.4 1.1 ± 0.2 0.8 ± 0.1
72
1.22 ± 0.02 1.27 ± 0.01 1.28 ± 0.02
12 21 24
9±2 15 ± 2 17 ± 3
91 ± 2 85 ± 2 83 ± 3
4 6 7
88 79 76
4±1 8±1 9±1
94 ± 1 91 ± 1 90 ± 1
1.4 ± 0.3 1.2 ± 0.1 1.2 ± 0.2
74
1.22 ± 0.01 1.29 ± 0.03 1.32 ± 0.04
11 24 29
8±1 18 ± 4 21 ± 3
92 ± 1 82 ± 4 79 ± 3
3 6 7
89 76 71
4±1 9±2 11 ± 2
95 ± 1 90 ± 2 88 ± 1
1.3 ± 0.4 0.9 ± 0.1 1.3 ± 1.4
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
395
Fig. 7. Optical microscopy images showing the microstructure of composites with the three SC fractions: (A) as received, (B) fine, and (C) coarse. The cross-sections are in a plane normal to the thickness direction of the composites.
Table 3 Mechanical properties of the manufactured SC composites. Tensile properties Fibre content [vol%]
Stiffness [GPa]
Strength [MPa]
Failure strain [%]
Polyester
0
3.4 ± 0.0
49 ± 6
1.7 ± 0.3
As received
4 7 9
4.2 ± 0.2 4.8 ± 0.1 5.5 ± 0.6
16 ± 3 18 ± 1 20 ± 4
0.4 ± 0.1 0.4 ± 0.0 0.4 ± 0.1
Fine
4 8 9
4.0 ± 0.2 4.7 ± 0.2 5.0 ± 0.2
21 ± 5 23 ± 4 23 ± 3
0.6 ± 0.1 0.5 ± 0.1 0.5 ± 0.1
Coarse
4 9 11
4.3 ± 0.3 5.4 ± 0.8 5.8 ± 0.4
15 ± 1 18 ± 1 29 ± 4
0.4 ± 0.1 0.3 ± 0.1 0.6 ± 0.1
matrix), k ¼ P4 (assuming a square fibre packing), d = 20 lm (as measured in this study), and the fibre lengths as reported in Table 1. The resulting values of gl1 are 0.99, 0.85 and 1.00 for the as received, fine and coarse SC composites, respectively.
Fig. 8 shows the experimental stiffness of the composites as a function of the fibre volume content. As can be observed, the stiffnesses of the as received SC composites and the coarse SC composites are similar, and slightly above the stiffnesses of the fine SC composites. This is expected due to the shorter fibres in the fine SC fraction, which is taken into account in the model predictions by using the above determined values of gl1 on 1.00 and 0.85 for the two groups of composites. By assuming that the fibre orientation is the same for all composites, the fibre orientation factor (go) was determined to be equal to 0.32 by fitting the model lines to the experimental data. This is showing that the fibres are oriented almost in a 2D random fibre orientation. The deviation from a complete 2D random fibre orientation, where go is equal to 0.375, is indicating that the fibres are also oriented in out-of-plane directions, i.e. in the thickness direction of the composite plates. This is supported by the micrographs in Fig. 7, where the longitudinal and transverse cross-sections of fibres demonstrate that the fibres are oriented in both 2D in-plane and 3D out-of-plane directions, respectively. Altogether, the results in Fig. 8 show a good agreement between the experimental data and the theoretical model predictions of the stiffness of the composites. This will form the basis for the strength predictions as presented next.
Fig. 8. Experimental data and model predictions for the stiffness of SC composites with variable glass fibre content.
396
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
For the prediction of composite strength using Eq. (11), the parameters are set as above for the stiffness predictions, i.e., Ef = 78 GPa, Em = 3.4 GPa, and go = 0.32. The glass fibre strength (rfu ) is set to be 1600 MPa. This value is taken from a study by Palmer [18] where the tensile strength of glass fibres recovered by mechanical recycling was measured. Accordingly, the glass fibre failure strain can be calculated to be 2.1% (=100 1.6 GPa/78 GPa). The interfacial shear stress (si) between glass fibres and polyester matrix was set to be 33 MPa, as taken from a study by Chua and Piggott [34]. The critical fibre length (lc) can then be calculated to be equal to 0.48 mm. It can be noted that the mean fibre length in the three SC composites are all above lc. Finally, the fibre efficiency factor (gl2) can be calculated to be 0.99, 0.69 and 0.99 for the as received, fine and coarse SC composites, respectively. Using the modified Kelly-Tyson model (Eq. (11)), the strength of the composites with the variable fibre contents are predicted to be in the ranges 87–109, 81–96 and 87–118 MPa for the as received, fine and coarse SC composites, respectively. As can be seen in Table 3, these model predictions are much above the experimental values of composite strength in the range of 15–29 MPa for all composites. In order to understand the reason for the large difference between the model predictions and the experimental values of composite strength, the stress-strain curves must be examined. Fig. 9 shows (i) a predicted stress-strain curve of glass fibres, assuming a linear stress-strain behaviour with a stiffness of 78 GPa, (ii) the measured curves of neat polyester matrix, and (iii) the measured and predicted stress-strain curves of the coarse SC composites. The predicted stress-strain curves of the composites are calculated by a rule of mixtures relationship for stresses:
rc ¼ go gl2 V f rf þ ð1 V f Þrm ! rc ¼ go gl2 V f Ef ec þ ð1 V f Þrm h2c i
ð13Þ
where rm h2c i is the the stress in the matrix as function of the strain in the composite, which is obtained by reading stress values from the measured stress-strain curve of the polyester matrix. As can be seen in Fig. 9, the shape of the predicted stress-strain curves for the composites with the three levels of fibre contents are well agreeing with the shape of the measured stress-strain curves. The model predictions can however not be used to predict the failure
stress and strain of the composites, i.e. the end points of the curves. Accordingly, a literature survey was performed on the strength properties of comparable composite systems. In the literature survey, it was decided to focus on studies of composites with randomly oriented glass fibres in the form of chopped strand mat (CSM), and with a polyester matrix. Studies on composites made by bulk moulding compound (BMC) and sheet moulding compound (SMC) were not taken into account due to the considerable amount of calcium carbonate fillers typically used for such composites. Literature data on strength properties was selected only when information was given for the volumetric or gravimetric composition of the composites. If the glass fibre volume content was needed to be calculated from the reported fibre weight content, a glass fibre density of 2.55 g/cm3 and a polyester matrix density of 1.18 g/cm3 were used. Next, the three selected studies from the literature survey are presented. In the study by Barton and Soden [35], CSM glass fibre/polyester composites were manufactured. Based on the reported thickness of the composite plates, the number of CSM layers, and the CSM area weight, the glass fibre volume content was calculated. Data is presented for composites having a fibre volume content of 18 vol%. In the study by Hancox and Mayer [36], polyester composites were manufactured using two different types of CSM: an emulsion bonded and a powder bonded. Based on the reported fibre weight contents, the fibre volume content of the manufactured composites was calculated to be 16 vol%. In the study by Johnson [37], the fibre volume content of the manufactured CSM glass fibre/polyester composites was calculated to be in the range 13–19 vol%. The failure strain of the composites is not reported, and it was estimated here using the reported strength and stiffness of the composites assuming a linear stressstrain behaviour. This approach is expected to lead to a slight underestimation of the failure strain values due to the convex shape of the stress-strain curves (see Fig. 9). The experimental data on failure stress and strain from the three literature studies is shown in Fig. 9 together with the predicted stress-strain curves for composites with fibre volume fractions of 16, 18 and 20 vol%. For these model predictions, a fibre orientation factor of 0.375 is used to reflect the expected inplane random orientation of CSM fibres. In general, the experimental data points from the literature studies are well located in the
Fig. 9. Experimental and predicted stress-strain curves for coarse SC composites with variable fibre content, in addition to curves for neat glass fibres and polyester matrix. Shown are also experimental data points for failure stress and strain of comparable composite systems, as obtained from three selected literature studies. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
range given by the predicted stress-strain curves, taking into consideration the approximated fibre volume contents and experimental scatter. The lower experimental data points from the study by Hancox and Mayer [36] can be explained by a slight incorrect determination on the fibre volume content on 16%, e.g. due to porosity in the composites, which is known to lead to both a lower fibre volume content, and to lower mechanical properties [38]. The failure strain of the CSM composites are varying in the range 1.16– 1.81%, which is lower than the estimated failure strain of the glass fibres on about 2.1%. Thus, the assumption in the Kelly-Tyson model of composite failure at the failure strain of the fibres is not valid for these composites. On the other hand, the failure strain of the CSM composites is much above the failure strain of the SC composites. Since the two types of composites are comparable, apart from a slight difference in fibre orientation, it should be expected that the SC composites would show similar failure strain as the CSM composites. Altogether, based on the above analyses of the results in Fig. 9, it is demonstrated that the SC composites are showing the expected stress-strain behaviour, i.e. they are following the predicted shape of the curves, but they are failing prematurely. In order to understand the reason for this, the fracture surface of the composites were inspected. Fig. 10 shows representative SEM images of fracture surfaces of the failed tensile specimens. The micrographs show large debonding cracks between glass fibre bundles and
397
polyester matrix. It is reasonable to believe that these cracks are critically contributing to the composite failure. The micrographs show also examples of fibre pull-out which typically is seen as a sign of low fibre/matrix interface bonding strength [32]. Altogether, it is indicated that the SC and the new polyester matrix are not sufficiently bonded to each other, and this initiates crack formation, and ultimately, it is leading to that the composites fail prematurely. Future work is needed to address this issue, e.g. by physical or chemical treatments to activate the surface of the old polyester matrix for promoting a better bonding to the new polyester matrix. Research on this topic has not yet given any conclusive results. DeRosa et al., [22] used hydrolysis in order to create a AOH groups on the surface of the shredded composite. The purpose of the AOH groups was to react with anhydrides or acids similar to those used in unsaturated polyesters resin. The treatment did not improve the mechanical properties of the composite manufactured. An alternative solution is to burn off the remaining matrix material from the surface of the glass fibres and apply a new sizing layer to the fibres as demonstrated by Thomasson et al. [39]. All these surface treatment methods will naturally add to the cost of the final product. The type of SC in the composites was also found to have an effect on the stress-strain curves. Fig. 11 shows the measured stress-strain curves for the as received, fine and coarse SC composites, and with the highest fibre content. As can be observed in the
Fig. 10. SEM images of fracture surface of SC composites.
Fig. 11. Stress-strain curves of the as received, fine and coarse SC composites with the highest fibre content on about 10 vol%. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
398
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
2D in-plane and 3D out-of-plane directions. The experimental strengths in the range 15–29 MPa were found to be significantly lower than the theoretical predictions in the range 81–118 MPa. In order to understand the reason for this difference, the stressstrain curves of the composites were evaluated. The analysis of the curves enabled to point out that the SC composites are showing the expected stress-strain behaviour, i.e. they are following the predicted shape of the curves, but they are failing prematurely. This is supported by a literature survey on the mechanical properties of comparable composite systems made of glass fibre chopped strand mats (CSM) and polyester. The SC composites in the present study show failure strain in the range from 0.3 to 0.6% which is much below the failure strain in the range 1.2–1.8% for the literature CSM composites. Based on observations of large fibre/matrix debonding cracks at the fracture surfaces of the SC composites it is indicated that the low failure strength and strain of the composites is due to insufficient bonding between the SC and the new polyester matrix. To tackle this problem, the application of a physical or chemical treatment of the SC or the use of an alternative resin such as epoxy, to improve bonding could be investigated.
Fig. 12. Failure characteristics of tensile specimens of fine (three specimens to the left) and coarse (three specimens to the right) SC composites.
figure, there is a tendency that the coarse SC composites fail at higher values of strength and strain than the as received and fine SC composites. In addition, the curves for the coarse SC composites show an erratic behaviour with several peaks before final failure. This indicates the development of smaller cracks in the coarse SC composites, which presumably are bridged by the long fibres to suppress final failure. For the as received SC composites, some curves show similar characteristics, whereas all the curves for the fine SC composites fail in a brittle manner, presumably due to the fibres being too short to prevent the cracks to grow. The difference between the stress-strain curves is supported by observations of the failure characteristics of the test specimens. Fig. 12 shows representative test specimens of the fine and coarse SC composites. Several cracks can be seen on the coarse SC composites, whereas only a single crack can be identified for the fine SC composites.
4. Conclusions In the context of recycling composite materials from wind turbine blades this study has been conducted to assess the reinforcing potential of shredded glass fibre/polyester composites in new polymer composites. Shredded composite (SC) obtained from the load carrying beam of a wind turbine blade was manually sorted into two fractions by sieving. A fraction named fine and containing the shortest fibres, and a fraction named coarse containing the longest fibres. A vacuum infusion setup was developed to manufacture composites with a controlled amount of SC. Composites were manufactured using three different SC fractions (as received, fine and coarse) as the only reinforcing part. The manufacturing process successfully enabled to produce composites with the wanted amount of SC, and the SC was measured to be well distributed over the whole plate. The porosity content was determined to be 1 vol% for all composites indicating high materials quality. The measured stiffness and strength of the SC composites were compared to theoretical predictions based on rules of mixture models. The experimental stiffnesses in the range 4.0–5.8 GPa were well predicted by the model using an established fibre orientation factor of 0.32 indicating that the fibres are oriented in both
Acknowledgements The authors would like to thank the innovation consortium GenVind on recycling of composite materials from the industry, LM Wind Power, especially Karin Magelund Møller and Jacob Boutrup for providing the material to this study and Force Technology, Jens Henriksen for shredding the material. This research has been supported by The Danish Council for Technology and Innovation under the Ministry of Science, Innovation and Higher Education, Denmark. References [1] The European Wind Energy Association. Research note outline on recycling windturbine blade; 2011. p. 10. [2] The European Wind Energy Association. The European Wind Energy Association website [accessed 26 January 2016]. [3] Wilkes J, Moccia J. Wind in power: 2010 European statistics; 2011. p. 12. [4] Pineda I, Wilkes J. Wind in power: 2014 European statistics; 2015. p. 12. [5] Larsen K. Recycling wind turbine blades. Renew Energy Focus 2009:70–3. [6] Martínez E, Sanz F, Pellegrini S, Jiménez E, Blanco J. Life cycle assessment of a multi-megawatt wind turbine. Renew Energy 2009;34:667–73. [7] Schleisner L. Life cycle assessment of a wind farm and related externalities. Renew Energy 2000;20:279–88. [8] The European Commission for the Environment. The European Commission for the environment website [accessed 26 January 2016]. [9] Brøndsted P, Lilholt H, Lystrup A. Composite materials for wind power turbine blades. Annu Rev Mater Res 2005;35:505–38. [10] Sayer F, Bürkner F, Blunk M, Van Wingerde AM, Busmann H. Influence of loads and environmental conditions on material properties over the service life of rotor blades. DEWI Mag 2009:24–31. [11] Beauson J, Bech JI, Brøndsted P. Composite recycling : characterizing end of life wind turbine blade materials. In: Proceeding of ICCM-19 conference, Montréal. p. 8. [12] Jongert J, Peeren C, Bergsma J, Schiferli F, de Krieger J, van Rooyen K, et al. Superuse studios website [accessed 26 January 2016]. [13] Beauson J, Lilholt H, Brondsted P. Recycling solid residues recovered from glass fibre-reinforced composites – a review applied to wind turbine blade materials. J Reinf Plast Compos 2014;33:1542–56. [14] Pickering SJ. Recycling technologies for thermoset composite materials— current status. Compos Part A Appl Sci Manuf 2006;37:1206–15. [15] Phoenix Fibreglass Inc. Fibreglass composite recycling: report. Ontario: Ministry of Environment and Energy; 1994. [16] Bledzki AK, Goracy K. Use of recycled fibre composites as reinforcement for thermosets. Mekhanika Kompoz Mater 1993;29:473–7. [17] Petterson J, Nilsson P. Recycling of SMC and BMC in standard process equipment. J Thermoplast Compos Mater 1994;7:56–63. [18] Palmer J. Mechanical recycling of automotive composites for use as reinforcement in thermoset composites. PhD thesis, University of Exeter; May 2009. p. 224. [19] Jutte R, Graham D. Recycling SMC. In: Proceeding of 46th Annu Conf Compos Institute, Soc Plast Conference. p. 6. [20] Palmer J, Ghita OR, Savage L, Evans KE. Successful closed-loop recycling of thermoset composites. Compos Part A Appl Sci Manuf 2009;40:490–8.
J. Beauson et al. / Composites: Part A 90 (2016) 390–399 [21] Derosa R, Telfeyan E, Mayes S. Expanding the use of recycled SMC in BMCs. In: Proceeding of Glob Plast Environ Conf 2004—Help grow a greener Environ Conference, Detroit. p. 371–83. [22] Derosa R, Telfeyan E, Gaustad G, Mayes S. Strength and microscopic investigation of unsaturated polyester BMC reinforced with SMC-recyclate. J Thermoplast Compos Mater 2005;18:333–49. [23] Silva LV, Angrizani CC, Souza JR, Amico SC, Medeiros JTN. Use of polyester/glass-fiber residues as fillers for composites. J Appl Polym Sci 2012;124:302–10. [24] Kouparitsas CE, Kartalis CN, Varelidis PC, Tsenoglou CJ, Papaspyrides CD. Recycling of the fibrous fraction of reinforced thermoset composites. Polym Compos 2002;23:682–9. [25] Zheng Y, Shen Z, Ma S, Cai C, Zhao X, Xing Y. A novel approach to recycling of glass fibers from nonmetal materials of waste printed circuit boards. J Hazard Mater 2009;170:978–82. [26] Dangtungee R, Somchua S, Siengchin S. Recycling glass fiber/epoxy resin of waste printed circuit boards: morphology and mechanical properties. Mech Compos Mater 2012;48:325–30. [27] Madsen B B, Løgstrup Andersen T, Lilholt H. Volumetric composition: fiber content and porosity. In: Oksman K, Mathew AP, Bismarck A, Rojas O, Mohini S, editors. Handb Green Mater Biobased Compos Mater their Process Prop Ind Appl, vol. 4. World Scientific Publishing Co Pte Ltd.; 2014. p. 137–53. [28] Bunsell AR, Renard J. Fundamentals of fibre reinforced composite materials. London: Institute of Physics Publishing; 2005. [29] Krenchel H. Fibre reinforcement. In: Theoretical and practical investigations of the elasticity and strength of fibre-reinforced materials. Copenhagen: Akademisk Forlag; 1964.
399
[30] Cox HL. The elasticity and strength of paper and other fibrous materials. Br J Appl Phys 1952;3:72–9. [31] Joffe R, Madsen B, Nättinen K, Miettinen A. Strength of cellulosic fiber/starch acetate composites with variable fiber and plasticizer content. J Compos Mater 2015;49:1007–17. [32] Kelly A, Tyson WR. Tensile properties of fibre-reinforced metals: copper/tungsten and copper/molybdenum. J Mech Phys Solids 1965;13:329–50. [33] Beauson J, Fraisse A, Toncelli C, Ilsted Bech J, Brøndsted P. Thermoset composite recycling : properties of recovered glass fiber. In: Proceeding of ICCM-20 conference, Copenhagen. p. 9. July. [34] Chua PS, Piggott MR. The glass fibre-polymer interface: II—work of fracture and shear stresses. Compos Sci Technol 1985;22:107–19. [35] Barton DC, Soden PD. Short-term in-plane stiffness and strength properties of CSM-reinforced polyester laminate. Composites 1982:66–78. [36] Hancox NL, Mayer RM. Design data for reinforced plastics – a guide for engineers and designers. London: Chapman & Hall; 1994. [37] Johnson AF. Engineering design properties of GRP. London: British Plastics Federation; 1979. [38] Madsen B, Thygesen A, Lilholt H. Plant fibre composites – porosity and stiffness. Compos Sci Technol 2009;69:1057–69. [39] Thomason JL, Nagel U, Yan L, Sáez E. Regenerating the strength of thermally recycled glass fibres using hot sodium hydroxide. Compos Part A Appl Sci Manuf 2016;87:220–7.
Contents lists available at ScienceDirect
Composites: Part A journal homepage: www.elsevier.com/locate/compositesa
Recycling of shredded composites from wind turbine blades in new thermoset polymer composites Justine Beauson ⇑, Bo Madsen, Chiara Toncelli, Povl Brøndsted, Jakob Ilsted Bech Department of Wind Energy, Technical University of Denmark, Risø Campus Frederiksborgvej 399, 4000 Roskilde, Denmark
a r t i c l e
i n f o
Article history: Received 2 February 2016 Received in revised form 8 July 2016 Accepted 10 July 2016 Available online 11 July 2016 Keywords: A. Recycling B. Mechanical properties C. Analytical modelling E. Vacuum infusion
a b s t r a c t As the energy produced from wind increases every year, a concern has raised on the recycling of wind turbine blades made of glass fibre composites. In this context, the present study aims to characterize and understand the mechanical properties of polyester resin composites reinforced with shredded composites (SC), and to assess the potential of such recycling solution. A special manufacturing setup was developed to produce composites with a controlled content of SC. Results show that the SC in the composites was well distributed and impregnated. The composite stiffness was well predicted using an analytical model, and fibre orientation parameters for strength modelling were established. The stress-strain curves revealed composite failure at unusual low strain values, and micrographs of the fracture surface indicated poor adhesion between SC and matrix. To tackle this problem, chemical treatment of SC or use of an alternative resin, to improve bonding should be investigated. Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction One of the targets of the European energy and climate policy is to increase the energy consumed from renewable resources to 20% by 2020. As a result, the European Wind Energy Association (EWEA) predicts that the consumed wind energy will increase from 5,3% in 2010 to 15.7% by 2020 and will continue to rise to 50% by 2050 [1]. In 2010, 72,000 wind turbines [2] were providing electricity across the European Union (EU) countries, representing a capacity of 84,3 GW (GW) [3]. At the end of 2014, this capacity was raised to 128,8 GW [4]. Due to the rapid development of the technology, the blade length has grown from 18 m in a 600 kW turbine in year 1990 to around 85 m in a 8000 kW turbine in year 2015. Altogether, this development means that the mass of decommissioned wind turbine material is ever increasing [5]. While some of the wind turbine materials are recyclable (such as the metal parts in the tower, gearbox, and blade hub) [6,7], the blades made of thermoset polymer composites represent a challenge. As one of the central objectives of the EU are to maximise recycling and to phase out landfilling of non-recyclable waste, it is obvious that wind turbine blades need recycling solutions [8]. Wind turbine blades are typically made from continuous glass fibre reinforced polymer composites. E-glass fibres, a low cost glass fibre grade combining high strength (2 GPa) and moderate stiffness ⇑ Corresponding author. E-mail address: [email protected] (J. Beauson). http://dx.doi.org/10.1016/j.compositesa.2016.07.009 1359-835X/Ó 2016 Elsevier Ltd. All rights reserved.
(76 GPa), are used together with a thermoset polymer matrix such as epoxy, polyester or vinylester. Sandwich constructions, consisting of multiaxial composite laminates, together with balsa wood or polyvinyl chloride (PVC) foam, are used for the outer shell and for the shear webs in the blades, see Fig. 1. Uniaxial composite laminates are used for the load carrying beam. These glass fibre laminates have a typical fibre volume content of 50% (corresponding to a fibre weight content of 70%) [9]. Wind turbine blades have a predicted life time of 20–25 years. End of life solutions for wind turbine blades can be compared by the amount of re-processing needed. Few re-processing steps of the blades are needed to extend their service life. After 20– 25 years, where the blades still have some residual strength [10], they can be taken down, refurbished and re-installed [11]. Without too many further reprocessing steps, sections of blades can also be reused [12]. The field of such applications is however limited. The mechanical properties of the different composite materials in the blade are precisely designed to fulfil the requirements of a blade. In addition, the specific shape and dimensions of blade sections make their reuse in new structural applications challenging. Heavier re-processing of blades is costly, but widens the number of possible applications for the recovered materials. Such solutions include shredding the composite materials, or separating the glass fibres from the polymer matrix using thermal and chemical processes [11,13,14]. Shredding of composite materials, also called mechanical recycling, is the simplest method, and the only solution that has been brought to a commercial level [15–17].
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
391
Fig. 1. Wind turbine blade structure and material. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Mechanical recycling of composites consists of successive grinding or cutting steps to reduce the materials to fragments of few millimetres. The visual aspect of the resulting shredded composite (SC) is a mixture of short individual fibres, longer fibre bundles partly impregnated with matrix material, and clusters of materials, see Fig. 2. In the perspective of using SC in new polymer composites, previous work has found that high SC quality is crucial, and that the most valuable fractions are the ones with a high glass fibre content and with long fibres [16,18]. Research on methods to separate the different fractions present in SC has been conducted and techniques using air, gravity or vibration have been developed [18]. The reuse of SC in new thermoset polymer composites has mainly been studied in order to reduce the amount of virgin glass fibres in existing composite systems. This contributes to a lower cost of the product, without having to change the original composite manufacturing procedure [18–23]. So far, however, only a few studies have studied the performance of thermoset polymer composites manufactured with SC as the only reinforcing part [23–26]. In the context of assessing the potential of mechanical recycling of wind turbine blade materials, the objective of this study is to understand and predict the reinforcement effect of SC in new polymer composites. SC from uniaxial glass fibre composites, originating from a wind turbine blade, will be classified into fractions using manual sieving, and then characterized by combustion and microscopy. Composites will be manufactured using SC as the only reinforcing part. The gravimetric and volumetric composition of the composites will be determined. The mechanical properties of the composites incorporating various amount of SC will be measured, and analysed using micromechanical models for composites.
2. Materials and methods The shredded composite (SC) material in this study comes from the load carrying beam of a 26 m long wind turbine blade manufactured by LM Wind Power. Before shredding, the blade was subjected to dynamic testing from December 2000 to May 2001. Glass fibre composite laminates with dimensions of 200 mm long, 200 mm wide and 22 mm thick were cut out from the load carrying beam, see Fig. 1. The glass fibre volume content of the composite laminate before shredding was measured to be 50 vol%. The composite laminate was shredded using a Bolen Super Tomahawk (grinding process based on rotating cylinder hammer, main frame screen with 20 mm hole size). Fig. 2 shows the SC as received. A thermoset polyester resin was selected to manufacture the new polymer composites. This choice was made in order to avoid mixing different types of resins, since polyester is also the resin used in the composite laminate. An orthophtalic polyester resin (Polylite 413-575) was chosen, and it was pre-accelerated and initiated with 1.5 wt% of Butanox HBO-50.
2.1. Characterisation of SC The SC was divided into three fractions: a fraction containing the SC as received, and two fractions named fine and coarse, separated by manual sieving with a sieve of 1.5 1.5 mm square hole sizes. To determine the matrix coverage on the glass fibres, in addition to the diameter and length of the fibres, the three SC fractions were observed with a scanning electron microscope (SEM, TM1000 Hitachi). The glass fibre content of the SC fractions was determined by combustion, using a so-called burn off test. A material sample was placed in a crucible, it was weighed, and then burnt at 565 °C for 2 h, and finally, it was weighed again. The difference in weight of the sample before and after combustion, allows calculating the glass fibre weight content in the SC fractions.
2.2. Manufacturing of composites
Fig. 2. Shredded composite. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
In order to manufacture composites containing different contents of SC, a special vacuum infusion setup was implemented. The wanted weight contents of SC in the new polymer composites were set to 10, 20 and 30 wt%. In order to obtain these weight contents with a vacuum infusion process, the weight of the manufactured composite plates need to be controlled. For that, a rectangular metal frame with dimensions of 300 400 7 mm, and with holes along the short sides of the frame was used.
392
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
To determine the weight of SC needed to manufacture composite plates with final SC weight content of 10, 20 and 30 wt%, the density of SC was first calculated. To do so, a rule of mixture relationship for composite was used, see Eq. (1).
1
qSC
¼
1
qf
Wf þ
1
qm
The definitions and relationships for the weight contents in the SC and the manufactured composites are given by the following equations:
Weight content of fibres in SC : SC f ¼ Wm
ð1Þ
qSC is the density of SC, qf and Wf the density and weight content of fibre, qm and Wm the density and weight content of matrix. The density of SC was calculated assuming a glass fibre density of 2.55 g/cm3, a polyester matrix density of 1.18 g/cm3 and the measured fibre weight content. Knowing the density of SC and the wanted SC weight content, the density of the final composite was calculated using the same rule of mixture. As the volume of the mould is known, the weight of the final composite was predicted and the weight of SC needed to manufacture the composite deducted/derived from it. The pre-weighed amount of SC was manually placed within the frame and then covered with a 7.5 mm thick aluminium plate to close the frame volume, see Fig. 3. The layup was then packed with a vacuum bag, and vacuum was applied (60 kPa). The setup was tilted to a vertical position in order to promote a stable resin flow front, and the polyester resin was then infused into the bag. At the end of the infusion, the vacuum was lowered to 20 kPa to avoid bending of the aluminium plate. Finally, after curing at room temperature for about 18 h (overnight), the composite plate was demoulded and post-cured for 2 h at 80 °C. Nine composite plates were manufactured using the three SC fractions, and with three SC contents each. In addition, a pure polyester plate was manufactured.
2.3. Characterisation of composites The composite plates were characterized by their gravimetric and volumetric composition, their microstructure and their mechanical properties. To determine the fibre, matrix and porosity content of the composites, six rectangular samples of dimensions 10 20 mm were cut from each composite plate. The samples were roughly polished, and then dried overnight, and weighed. The samples were then sealed with paraffin wax to be able to include the correct porosity volume in the determinations. The sealed samples were weighed first in air, and then in water. Then, the polyester matrix in the samples was burnt off overnight at 565 °C. The weight of the remaining glass fibres was recorded. Based on these measurements, the glass fibre weight and volume content, Wf and Vf, the matrix weight and volume content, Wm and Vm, and the porosity volume content, Vp, were determined using standard equations [27].
mf Wf ¼ mf þ mm1 W SC
Weight content of old matrix in SC : mm1 SC m1 ¼ ¼ 1 SC f mf þ mm1 Weight content of SC in composites : mf þ mm1 Wf ¼ W SC ¼ mf þ mm1 þ mm2 SC f Weight content of fibres in composites : mf Wf ¼ ¼ SC f W SC mf þ mm1 þ mm2 Weight content of old matrix in composites : mm1 ¼ SC m1 W SC W m1 ¼ mf þ mm1 þ mm2 Weight content of new matrix in composites : mm2 W m2 ¼ ¼ 1 W SC mf þ mm1 þ mm2
ð2Þ
ð3Þ
ð4Þ
ð5Þ
ð6Þ
ð7Þ
To characterize the microstructure of the composites, a sample with dimensions 20 10 mm was cut from each plate. The samples were casted in epoxy, and the sample cross-section in the plane normal to the thickness direction was grinded and polished. Observations were made using an optical microscope. To conduct static tensile testing, six dog bone shaped test specimens were cut according to ISO 527-2 from each composite plate. A tensile test machine with 50 kN load cell and hydraulic grips were used. The tests were performed with a displacement rate of 2 mm/min. Based on the measured stress-strain curves, the stiffness (the slope in the strain interval 0.05–0.25%) and the strength (the maximum stress value) were determined. In addition, the fracture surface of the tested specimens were coated with gold and observed by using a SEM. 2.4. Modelling of mechanical properties of composites The measured values of stiffness and strength of the composites were compared to theoretical model predictions. To predict the stiffness of composites with randomly oriented short fibres, the combined rule of mixtures model [28] was used:
Fig. 3. Special vacuum infusion setup used for manufacturing of SC composites. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
Ec ¼ g0 gl1 V f Ef þ ð1 V f ÞEm
ð8Þ
where Ec is the composite stiffness, V f is the fibre volume fraction, Ef is the fibre stiffness, and Em is the matrix stiffness. The effect of a non-unidirectional fibre orientation in composites was addressed by Krenchel [29] by introducing a fibre orientation efficiency factor g0. For a 2D random fibre orientation, this factor is equal to 0.375 (= 3/8), whereas it is equal to 0.20 (= 1/5) for a 3D random fibre orientation. The effect of short fibres in composites was addressed by Cox [30] by introducing a fibre length efficiency factor gl1, which takes into account the reduced efficiency of stress transfer from matrix to fibres, when the fibres are non-continuous:
gl1 ¼ 1
tanh bl2
ð9Þ
bl 2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi bl lu u Gm ¼2 t 2 d E ln k f V
ð10Þ
f
where l is the fibre length, d is the fibre diameter, Gm is the matrix shear modulus, and k is a constant that depends on the geometrical packing of fibres. To predict the strength of composites with randomly oriented short fibres, a modified Kelly-Tyson model was used [31,32]:
rcu ¼ go gl2 V f rfu þ ð1 V f ÞEm rfu =Ef
ð11Þ
where rcu is the composite strength, and rfu is the fibre strength. As for the prediction of stiffness, the non-unidirectional fibre orientation is addressed via the parameter go . In the Kelly-Tyson model, the fibre length efficiency factor (gl2) is defined as:
(
nl2 ¼
l 2lc
;
for l < lc
1 2lcl ;
for l P lc
ð12Þ
where lc is the critical fibre length defined by lc ¼ rfu d=2si where si is the interfacial shear stress between fibres and matrix. The KellyTyson model is describing the situation of brittle fibres in a ductile matrix, i.e. efu < emu, where e is strain. The model is based on the failure criterion that the composite is failing when the fibres are failing. As can be seen from Eq. (11), the stress in the matrix at fibre failure is calculated by rm ¼ Em rfu =Ef , assuming a linear stressstrain behaviour of the matrix. 3. Results and discussion 3.1. Characteristics of SC and the manufactured composites The fine fraction of SC obtained by manual sieving is shown in Fig. 4A. It contains mostly short glass fibres, in addition to some bundles of long glass fibres, and some pieces of polyester matrix. From the SEM images, the average fibre length for this fraction is
393
estimated to be around 0.8 mm. The material retained on the sieve contained two types of fractions: the coarse fraction, see Fig. 4B, and some clusters, see Fig. 4C. The shredded composite was found to be made of 52 wt% of coarse fraction, 44 wt% of fine fraction and 4 wt% of clusters. The clusters were manually removed to get the coarse fraction. In principle, the clusters, made of agglomerated short glass fibres, can be crushed and added to the fine fraction, but their manual breaking is a time consuming process. Therefore, it was decided to discard them. The coarse fraction consists of long glass fibre bundles, and pieces of polyester matrix. By visual inspection, the average fibre length in this fraction was approximated to be 30 mm. The SC fractions were observed with SEM. It was found that large pieces of matrix were attached to the fibre bundles and to the stitching thread (Fig. 5), and that smaller pieces of matrix were attached to the surface of the glass fibres (Fig. 6). The burn off test revealed a similar glass fibre content in the range 72–74 wt% for the fine and coarse fractions (Table 1). Regarding the manufactured composites, the three wanted SC weight contents of 10, 20 and 30 wt% were all almost achieved, except for the composite plates with the wanted weight content of 30 wt%, where the obtained weight contents are slightly lower, see Table 1. For these relatively high fibre contents, the vacuum pressure applied during the manufacturing process was not able to properly close the volume within the metal frame, and this led to a larger volume (and mass) of the plates than expected. Regarding the porosity content, it was estimated to be around 1 vol% for all composites, which is an indication of good quality composites. In Table 1, the SC weight content (WSC) is calculated by Eq. (4) using local measurements of the composite fibre weight content (Wf), based on small samples, and the measured fibre weight content of the SC (SCf). These local values of WSC can be compared to the measured global weight content of SC in the composites obtained by the pre-weighed amount of SC used in the manufacturing process, and the final weight of the composite plate. These values are presented in Table 2. The two sets of results are similar to each other, which is indicating that the SC are well dispersed over the whole composite plates. Fig. 7 shows representative images of the microstructure of the composites with the three SC fractions (as received, fine and coarse). The micrographs show that the glass fibres are well impregnated by the matrix, and with no apparent porosities. It is not possible to distinguish the old matrix from the new one, since the appearance of the matrix is uniform. The few black spots in the micrographs are believed to be due to fibres that have been pulled out from the surface during polishing. The results of the tensile testing are reported in Table 3. The stiffnesses of the composites, ranging from 4.2 to 5.8 GPa, are all higher than the stiffness of the pure polyester plate on 3.4 GPa. For each type of SC composite, the higher the fibre content, the higher the stiffness. Regarding the strength, all composites have a lower tensile strength than the one of the pure polyester plate.
Fig. 4. Images of the obtained SC fractions (A) fine, (B) coarse, (C) clusters. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
394
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
Fig. 5. SEM images of glass fibre bundles in the SC fractions.
Table 2 Local and global weight contents of SC in the manufactured composites. SC fractions
SC weight content in composites [wt%] Local
Global
As received
13 19 24
11 20 27
Fine
12 21 24
11 19 25
Coarse
11 24 29
11 21 30
Fig. 6. SEM image of a glass fibre with small pieces of polyester matrix attached to the surface.
content, whereas the failure strain of the neat polyester plate is equal to 1.7%.
The strengths of the composites are ranging from 15 to 29 MPa, whereas the strength of the neat polyester plate is equal to 49 MPa. Similarly to stiffness, the strength of the composites is increased with the fibre content. Regarding the failure strain, the composites reach significantly lower failure strains than the one of the pure polyester plate. The failure strains of the composites are ranging from 0.3 to 0.6%, with no apparent influence of the fibre
3.2. Modelling of mechanical properties of composites For the prediction of composite stiffness using Eq. (8), the values used to calculate the fibre length efficiency factor (gl1) are: Ef = 78 GPa (as measured on the same type of glass fibres in the study by Beauson et al. [33]), Gm = 1.2 GPa (as calculated from the measured tensile stiffness of 3.4 GPa for the isotropic polyester
Table 1 Characteristics of SC fractions, and the manufactured SC composites. SC fractions
SC characteristics
Composite characteristics
Fibre length [mm]
Fibre content [wt%] SCf
Density [g cm3]
As received
0.8–30
73
Fine
0.8
Coarse
30
Gravimetric composition [wt%]
Volumetric composition [vol%]
SC Wsc
Fibres Wf
Matrix Wm
Old matrix Wm1
New matrix Wm2
Fibres Vf
Matrix Vm
Porosity Vp
1.22 ± 0.02 1.26 ± 0.02 1.29 ± 0.02
13 19 24
9±2 14 ± 3 18 ± 3
91 ± 2 86 ± 3 83 ± 3
3 5 6
87 81 76
4±1 7±1 9±2
94 ± 1 92 ± 1 90 ± 2
1.5 ± 0.4 1.1 ± 0.2 0.8 ± 0.1
72
1.22 ± 0.02 1.27 ± 0.01 1.28 ± 0.02
12 21 24
9±2 15 ± 2 17 ± 3
91 ± 2 85 ± 2 83 ± 3
4 6 7
88 79 76
4±1 8±1 9±1
94 ± 1 91 ± 1 90 ± 1
1.4 ± 0.3 1.2 ± 0.1 1.2 ± 0.2
74
1.22 ± 0.01 1.29 ± 0.03 1.32 ± 0.04
11 24 29
8±1 18 ± 4 21 ± 3
92 ± 1 82 ± 4 79 ± 3
3 6 7
89 76 71
4±1 9±2 11 ± 2
95 ± 1 90 ± 2 88 ± 1
1.3 ± 0.4 0.9 ± 0.1 1.3 ± 1.4
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
395
Fig. 7. Optical microscopy images showing the microstructure of composites with the three SC fractions: (A) as received, (B) fine, and (C) coarse. The cross-sections are in a plane normal to the thickness direction of the composites.
Table 3 Mechanical properties of the manufactured SC composites. Tensile properties Fibre content [vol%]
Stiffness [GPa]
Strength [MPa]
Failure strain [%]
Polyester
0
3.4 ± 0.0
49 ± 6
1.7 ± 0.3
As received
4 7 9
4.2 ± 0.2 4.8 ± 0.1 5.5 ± 0.6
16 ± 3 18 ± 1 20 ± 4
0.4 ± 0.1 0.4 ± 0.0 0.4 ± 0.1
Fine
4 8 9
4.0 ± 0.2 4.7 ± 0.2 5.0 ± 0.2
21 ± 5 23 ± 4 23 ± 3
0.6 ± 0.1 0.5 ± 0.1 0.5 ± 0.1
Coarse
4 9 11
4.3 ± 0.3 5.4 ± 0.8 5.8 ± 0.4
15 ± 1 18 ± 1 29 ± 4
0.4 ± 0.1 0.3 ± 0.1 0.6 ± 0.1
matrix), k ¼ P4 (assuming a square fibre packing), d = 20 lm (as measured in this study), and the fibre lengths as reported in Table 1. The resulting values of gl1 are 0.99, 0.85 and 1.00 for the as received, fine and coarse SC composites, respectively.
Fig. 8 shows the experimental stiffness of the composites as a function of the fibre volume content. As can be observed, the stiffnesses of the as received SC composites and the coarse SC composites are similar, and slightly above the stiffnesses of the fine SC composites. This is expected due to the shorter fibres in the fine SC fraction, which is taken into account in the model predictions by using the above determined values of gl1 on 1.00 and 0.85 for the two groups of composites. By assuming that the fibre orientation is the same for all composites, the fibre orientation factor (go) was determined to be equal to 0.32 by fitting the model lines to the experimental data. This is showing that the fibres are oriented almost in a 2D random fibre orientation. The deviation from a complete 2D random fibre orientation, where go is equal to 0.375, is indicating that the fibres are also oriented in out-of-plane directions, i.e. in the thickness direction of the composite plates. This is supported by the micrographs in Fig. 7, where the longitudinal and transverse cross-sections of fibres demonstrate that the fibres are oriented in both 2D in-plane and 3D out-of-plane directions, respectively. Altogether, the results in Fig. 8 show a good agreement between the experimental data and the theoretical model predictions of the stiffness of the composites. This will form the basis for the strength predictions as presented next.
Fig. 8. Experimental data and model predictions for the stiffness of SC composites with variable glass fibre content.
396
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
For the prediction of composite strength using Eq. (11), the parameters are set as above for the stiffness predictions, i.e., Ef = 78 GPa, Em = 3.4 GPa, and go = 0.32. The glass fibre strength (rfu ) is set to be 1600 MPa. This value is taken from a study by Palmer [18] where the tensile strength of glass fibres recovered by mechanical recycling was measured. Accordingly, the glass fibre failure strain can be calculated to be 2.1% (=100 1.6 GPa/78 GPa). The interfacial shear stress (si) between glass fibres and polyester matrix was set to be 33 MPa, as taken from a study by Chua and Piggott [34]. The critical fibre length (lc) can then be calculated to be equal to 0.48 mm. It can be noted that the mean fibre length in the three SC composites are all above lc. Finally, the fibre efficiency factor (gl2) can be calculated to be 0.99, 0.69 and 0.99 for the as received, fine and coarse SC composites, respectively. Using the modified Kelly-Tyson model (Eq. (11)), the strength of the composites with the variable fibre contents are predicted to be in the ranges 87–109, 81–96 and 87–118 MPa for the as received, fine and coarse SC composites, respectively. As can be seen in Table 3, these model predictions are much above the experimental values of composite strength in the range of 15–29 MPa for all composites. In order to understand the reason for the large difference between the model predictions and the experimental values of composite strength, the stress-strain curves must be examined. Fig. 9 shows (i) a predicted stress-strain curve of glass fibres, assuming a linear stress-strain behaviour with a stiffness of 78 GPa, (ii) the measured curves of neat polyester matrix, and (iii) the measured and predicted stress-strain curves of the coarse SC composites. The predicted stress-strain curves of the composites are calculated by a rule of mixtures relationship for stresses:
rc ¼ go gl2 V f rf þ ð1 V f Þrm ! rc ¼ go gl2 V f Ef ec þ ð1 V f Þrm h2c i
ð13Þ
where rm h2c i is the the stress in the matrix as function of the strain in the composite, which is obtained by reading stress values from the measured stress-strain curve of the polyester matrix. As can be seen in Fig. 9, the shape of the predicted stress-strain curves for the composites with the three levels of fibre contents are well agreeing with the shape of the measured stress-strain curves. The model predictions can however not be used to predict the failure
stress and strain of the composites, i.e. the end points of the curves. Accordingly, a literature survey was performed on the strength properties of comparable composite systems. In the literature survey, it was decided to focus on studies of composites with randomly oriented glass fibres in the form of chopped strand mat (CSM), and with a polyester matrix. Studies on composites made by bulk moulding compound (BMC) and sheet moulding compound (SMC) were not taken into account due to the considerable amount of calcium carbonate fillers typically used for such composites. Literature data on strength properties was selected only when information was given for the volumetric or gravimetric composition of the composites. If the glass fibre volume content was needed to be calculated from the reported fibre weight content, a glass fibre density of 2.55 g/cm3 and a polyester matrix density of 1.18 g/cm3 were used. Next, the three selected studies from the literature survey are presented. In the study by Barton and Soden [35], CSM glass fibre/polyester composites were manufactured. Based on the reported thickness of the composite plates, the number of CSM layers, and the CSM area weight, the glass fibre volume content was calculated. Data is presented for composites having a fibre volume content of 18 vol%. In the study by Hancox and Mayer [36], polyester composites were manufactured using two different types of CSM: an emulsion bonded and a powder bonded. Based on the reported fibre weight contents, the fibre volume content of the manufactured composites was calculated to be 16 vol%. In the study by Johnson [37], the fibre volume content of the manufactured CSM glass fibre/polyester composites was calculated to be in the range 13–19 vol%. The failure strain of the composites is not reported, and it was estimated here using the reported strength and stiffness of the composites assuming a linear stressstrain behaviour. This approach is expected to lead to a slight underestimation of the failure strain values due to the convex shape of the stress-strain curves (see Fig. 9). The experimental data on failure stress and strain from the three literature studies is shown in Fig. 9 together with the predicted stress-strain curves for composites with fibre volume fractions of 16, 18 and 20 vol%. For these model predictions, a fibre orientation factor of 0.375 is used to reflect the expected inplane random orientation of CSM fibres. In general, the experimental data points from the literature studies are well located in the
Fig. 9. Experimental and predicted stress-strain curves for coarse SC composites with variable fibre content, in addition to curves for neat glass fibres and polyester matrix. Shown are also experimental data points for failure stress and strain of comparable composite systems, as obtained from three selected literature studies. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
range given by the predicted stress-strain curves, taking into consideration the approximated fibre volume contents and experimental scatter. The lower experimental data points from the study by Hancox and Mayer [36] can be explained by a slight incorrect determination on the fibre volume content on 16%, e.g. due to porosity in the composites, which is known to lead to both a lower fibre volume content, and to lower mechanical properties [38]. The failure strain of the CSM composites are varying in the range 1.16– 1.81%, which is lower than the estimated failure strain of the glass fibres on about 2.1%. Thus, the assumption in the Kelly-Tyson model of composite failure at the failure strain of the fibres is not valid for these composites. On the other hand, the failure strain of the CSM composites is much above the failure strain of the SC composites. Since the two types of composites are comparable, apart from a slight difference in fibre orientation, it should be expected that the SC composites would show similar failure strain as the CSM composites. Altogether, based on the above analyses of the results in Fig. 9, it is demonstrated that the SC composites are showing the expected stress-strain behaviour, i.e. they are following the predicted shape of the curves, but they are failing prematurely. In order to understand the reason for this, the fracture surface of the composites were inspected. Fig. 10 shows representative SEM images of fracture surfaces of the failed tensile specimens. The micrographs show large debonding cracks between glass fibre bundles and
397
polyester matrix. It is reasonable to believe that these cracks are critically contributing to the composite failure. The micrographs show also examples of fibre pull-out which typically is seen as a sign of low fibre/matrix interface bonding strength [32]. Altogether, it is indicated that the SC and the new polyester matrix are not sufficiently bonded to each other, and this initiates crack formation, and ultimately, it is leading to that the composites fail prematurely. Future work is needed to address this issue, e.g. by physical or chemical treatments to activate the surface of the old polyester matrix for promoting a better bonding to the new polyester matrix. Research on this topic has not yet given any conclusive results. DeRosa et al., [22] used hydrolysis in order to create a AOH groups on the surface of the shredded composite. The purpose of the AOH groups was to react with anhydrides or acids similar to those used in unsaturated polyesters resin. The treatment did not improve the mechanical properties of the composite manufactured. An alternative solution is to burn off the remaining matrix material from the surface of the glass fibres and apply a new sizing layer to the fibres as demonstrated by Thomasson et al. [39]. All these surface treatment methods will naturally add to the cost of the final product. The type of SC in the composites was also found to have an effect on the stress-strain curves. Fig. 11 shows the measured stress-strain curves for the as received, fine and coarse SC composites, and with the highest fibre content. As can be observed in the
Fig. 10. SEM images of fracture surface of SC composites.
Fig. 11. Stress-strain curves of the as received, fine and coarse SC composites with the highest fibre content on about 10 vol%. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
398
J. Beauson et al. / Composites: Part A 90 (2016) 390–399
2D in-plane and 3D out-of-plane directions. The experimental strengths in the range 15–29 MPa were found to be significantly lower than the theoretical predictions in the range 81–118 MPa. In order to understand the reason for this difference, the stressstrain curves of the composites were evaluated. The analysis of the curves enabled to point out that the SC composites are showing the expected stress-strain behaviour, i.e. they are following the predicted shape of the curves, but they are failing prematurely. This is supported by a literature survey on the mechanical properties of comparable composite systems made of glass fibre chopped strand mats (CSM) and polyester. The SC composites in the present study show failure strain in the range from 0.3 to 0.6% which is much below the failure strain in the range 1.2–1.8% for the literature CSM composites. Based on observations of large fibre/matrix debonding cracks at the fracture surfaces of the SC composites it is indicated that the low failure strength and strain of the composites is due to insufficient bonding between the SC and the new polyester matrix. To tackle this problem, the application of a physical or chemical treatment of the SC or the use of an alternative resin such as epoxy, to improve bonding could be investigated.
Fig. 12. Failure characteristics of tensile specimens of fine (three specimens to the left) and coarse (three specimens to the right) SC composites.
figure, there is a tendency that the coarse SC composites fail at higher values of strength and strain than the as received and fine SC composites. In addition, the curves for the coarse SC composites show an erratic behaviour with several peaks before final failure. This indicates the development of smaller cracks in the coarse SC composites, which presumably are bridged by the long fibres to suppress final failure. For the as received SC composites, some curves show similar characteristics, whereas all the curves for the fine SC composites fail in a brittle manner, presumably due to the fibres being too short to prevent the cracks to grow. The difference between the stress-strain curves is supported by observations of the failure characteristics of the test specimens. Fig. 12 shows representative test specimens of the fine and coarse SC composites. Several cracks can be seen on the coarse SC composites, whereas only a single crack can be identified for the fine SC composites.
4. Conclusions In the context of recycling composite materials from wind turbine blades this study has been conducted to assess the reinforcing potential of shredded glass fibre/polyester composites in new polymer composites. Shredded composite (SC) obtained from the load carrying beam of a wind turbine blade was manually sorted into two fractions by sieving. A fraction named fine and containing the shortest fibres, and a fraction named coarse containing the longest fibres. A vacuum infusion setup was developed to manufacture composites with a controlled amount of SC. Composites were manufactured using three different SC fractions (as received, fine and coarse) as the only reinforcing part. The manufacturing process successfully enabled to produce composites with the wanted amount of SC, and the SC was measured to be well distributed over the whole plate. The porosity content was determined to be 1 vol% for all composites indicating high materials quality. The measured stiffness and strength of the SC composites were compared to theoretical predictions based on rules of mixture models. The experimental stiffnesses in the range 4.0–5.8 GPa were well predicted by the model using an established fibre orientation factor of 0.32 indicating that the fibres are oriented in both
Acknowledgements The authors would like to thank the innovation consortium GenVind on recycling of composite materials from the industry, LM Wind Power, especially Karin Magelund Møller and Jacob Boutrup for providing the material to this study and Force Technology, Jens Henriksen for shredding the material. This research has been supported by The Danish Council for Technology and Innovation under the Ministry of Science, Innovation and Higher Education, Denmark. References [1] The European Wind Energy Association. Research note outline on recycling windturbine blade; 2011. p. 10. [2] The European Wind Energy Association. The European Wind Energy Association website [accessed 26 January 2016]. [3] Wilkes J, Moccia J. Wind in power: 2010 European statistics; 2011. p. 12. [4] Pineda I, Wilkes J. Wind in power: 2014 European statistics; 2015. p. 12. [5] Larsen K. Recycling wind turbine blades. Renew Energy Focus 2009:70–3. [6] Martínez E, Sanz F, Pellegrini S, Jiménez E, Blanco J. Life cycle assessment of a multi-megawatt wind turbine. Renew Energy 2009;34:667–73. [7] Schleisner L. Life cycle assessment of a wind farm and related externalities. Renew Energy 2000;20:279–88. [8] The European Commission for the Environment. The European Commission for the environment website [accessed 26 January 2016]. [9] Brøndsted P, Lilholt H, Lystrup A. Composite materials for wind power turbine blades. Annu Rev Mater Res 2005;35:505–38. [10] Sayer F, Bürkner F, Blunk M, Van Wingerde AM, Busmann H. Influence of loads and environmental conditions on material properties over the service life of rotor blades. DEWI Mag 2009:24–31. [11] Beauson J, Bech JI, Brøndsted P. Composite recycling : characterizing end of life wind turbine blade materials. In: Proceeding of ICCM-19 conference, Montréal. p. 8. [12] Jongert J, Peeren C, Bergsma J, Schiferli F, de Krieger J, van Rooyen K, et al. Superuse studios website [accessed 26 January 2016]. [13] Beauson J, Lilholt H, Brondsted P. Recycling solid residues recovered from glass fibre-reinforced composites – a review applied to wind turbine blade materials. J Reinf Plast Compos 2014;33:1542–56. [14] Pickering SJ. Recycling technologies for thermoset composite materials— current status. Compos Part A Appl Sci Manuf 2006;37:1206–15. [15] Phoenix Fibreglass Inc. Fibreglass composite recycling: report. Ontario: Ministry of Environment and Energy; 1994. [16] Bledzki AK, Goracy K. Use of recycled fibre composites as reinforcement for thermosets. Mekhanika Kompoz Mater 1993;29:473–7. [17] Petterson J, Nilsson P. Recycling of SMC and BMC in standard process equipment. J Thermoplast Compos Mater 1994;7:56–63. [18] Palmer J. Mechanical recycling of automotive composites for use as reinforcement in thermoset composites. PhD thesis, University of Exeter; May 2009. p. 224. [19] Jutte R, Graham D. Recycling SMC. In: Proceeding of 46th Annu Conf Compos Institute, Soc Plast Conference. p. 6. [20] Palmer J, Ghita OR, Savage L, Evans KE. Successful closed-loop recycling of thermoset composites. Compos Part A Appl Sci Manuf 2009;40:490–8.
J. Beauson et al. / Composites: Part A 90 (2016) 390–399 [21] Derosa R, Telfeyan E, Mayes S. Expanding the use of recycled SMC in BMCs. In: Proceeding of Glob Plast Environ Conf 2004—Help grow a greener Environ Conference, Detroit. p. 371–83. [22] Derosa R, Telfeyan E, Gaustad G, Mayes S. Strength and microscopic investigation of unsaturated polyester BMC reinforced with SMC-recyclate. J Thermoplast Compos Mater 2005;18:333–49. [23] Silva LV, Angrizani CC, Souza JR, Amico SC, Medeiros JTN. Use of polyester/glass-fiber residues as fillers for composites. J Appl Polym Sci 2012;124:302–10. [24] Kouparitsas CE, Kartalis CN, Varelidis PC, Tsenoglou CJ, Papaspyrides CD. Recycling of the fibrous fraction of reinforced thermoset composites. Polym Compos 2002;23:682–9. [25] Zheng Y, Shen Z, Ma S, Cai C, Zhao X, Xing Y. A novel approach to recycling of glass fibers from nonmetal materials of waste printed circuit boards. J Hazard Mater 2009;170:978–82. [26] Dangtungee R, Somchua S, Siengchin S. Recycling glass fiber/epoxy resin of waste printed circuit boards: morphology and mechanical properties. Mech Compos Mater 2012;48:325–30. [27] Madsen B B, Løgstrup Andersen T, Lilholt H. Volumetric composition: fiber content and porosity. In: Oksman K, Mathew AP, Bismarck A, Rojas O, Mohini S, editors. Handb Green Mater Biobased Compos Mater their Process Prop Ind Appl, vol. 4. World Scientific Publishing Co Pte Ltd.; 2014. p. 137–53. [28] Bunsell AR, Renard J. Fundamentals of fibre reinforced composite materials. London: Institute of Physics Publishing; 2005. [29] Krenchel H. Fibre reinforcement. In: Theoretical and practical investigations of the elasticity and strength of fibre-reinforced materials. Copenhagen: Akademisk Forlag; 1964.
399
[30] Cox HL. The elasticity and strength of paper and other fibrous materials. Br J Appl Phys 1952;3:72–9. [31] Joffe R, Madsen B, Nättinen K, Miettinen A. Strength of cellulosic fiber/starch acetate composites with variable fiber and plasticizer content. J Compos Mater 2015;49:1007–17. [32] Kelly A, Tyson WR. Tensile properties of fibre-reinforced metals: copper/tungsten and copper/molybdenum. J Mech Phys Solids 1965;13:329–50. [33] Beauson J, Fraisse A, Toncelli C, Ilsted Bech J, Brøndsted P. Thermoset composite recycling : properties of recovered glass fiber. In: Proceeding of ICCM-20 conference, Copenhagen. p. 9. July. [34] Chua PS, Piggott MR. The glass fibre-polymer interface: II—work of fracture and shear stresses. Compos Sci Technol 1985;22:107–19. [35] Barton DC, Soden PD. Short-term in-plane stiffness and strength properties of CSM-reinforced polyester laminate. Composites 1982:66–78. [36] Hancox NL, Mayer RM. Design data for reinforced plastics – a guide for engineers and designers. London: Chapman & Hall; 1994. [37] Johnson AF. Engineering design properties of GRP. London: British Plastics Federation; 1979. [38] Madsen B, Thygesen A, Lilholt H. Plant fibre composites – porosity and stiffness. Compos Sci Technol 2009;69:1057–69. [39] Thomason JL, Nagel U, Yan L, Sáez E. Regenerating the strength of thermally recycled glass fibres using hot sodium hydroxide. Compos Part A Appl Sci Manuf 2016;87:220–7.