- Email: [email protected]
Exploiting cyclic softening in continuous lattice fabrication for the additive manufacturing of high performance fibre-reinforced thermoplastic composite materials
Composites Science and Technology 164 (2018) 248–259
Contents lists available at ScienceDirect
Composites Science and Technology journal homepage: www.elsevier.com/locate/compscitech
Exploiting cyclic softening in continuous lattice fabrication for the additive manufacturing of high performance fibre-reinforced thermoplastic composite materials
T
Martin Eichenhofera,∗, Joanna C.H. Wonga,b, Paolo Ermannia a
Laboratory of Composite Materials and Adaptive Structures, Institute of Design, Materials and Fabrication, ETH Zurich, Tannenstrasse 3, 8092 Zurich, Switzerland Department of Mechanical and Manufacturing Engineering, Schulich School of Engineering, University of Calgary, 856 Campus Place NW, Calgary, AB, T2N 4V8, Canada b
A R T I C LE I N FO
A B S T R A C T
Keywords: A. Polymer-matrix composites (PMCs) A. Carbon fibres C. Residual stress E. Pultrusion Additive manufacturing (AM)
Continuous lattice fabrication (CLF) was recently introduced as a new additive manufacturing (AM) technology capable of printing continuous fibre-reinforced thermoplastic composites along desired trajectories in threedimensional space. In a systematic attempt to maximize the mechanical properties of the printed extrudate by minimizing the residual void content, this study investigates the thermal deconsolidation behaviour observed in pultruded unidirectional fibre-reinforced thermoplastic composite material when it is reheated above its melting point and exposed to ambient pressure. Fibre decompaction, generally accepted to be the primary cause for deconsolidation in fibre-reinforced thermoplastics, was investigated to assess the influence of cyclic softening of the fibrous media on the residual void content of the extruded material. The magnitude and rate of fibre decompaction were observed to decrease with the number of consolidation-deconsolidation cycles to which the material was subjected. A model was developed to predict the degree of deconsolidation in the CLF process as a function of temperature, processing speed, and processing history. Based on the deconsolidation behaviour observed, a multi-stage pultrusion module was designed that exploits cyclic softening and was demonstrated to reduce the residual void content of the printed extrudate by over 80%.
1. Introduction Additive manufacturing (AM) technologies have the advantages, in comparison to casting and subtractive manufacturing processes, of eliminating some of the costs associated with complex geometries and individually tailored designs, while also reducing waste [1]. The emergence of AM as a viable approach for the production of end-use components remains the target for many low-volume applications [2]. The AM of high performance fibre-reinforced polymer composites (FRPC) is particularly attractive for enabling a new design space for ultra-lightweight structures in aerospace, medical engineering and robotics amongst others. Although efforts to apply layer-by-layer AM strategies to anisotropic FRPC materials have met with moderate success [3–10], in order to fully harness the potential of FRPC materials in engineering structures, alternatives to the planar layer-by-layer approach must be developed which allow for the orientation of anisotropic materials along all relevant vectors, including those positioned out-of-plane. Recently, we introduced continuous lattice fabrication (CLF) as an
∗
Corresponding author. E-mail address: [email protected] (M. Eichenhofer).
https://doi.org/10.1016/j.compscitech.2018.05.033 Received 13 March 2018; Received in revised form 16 May 2018; Accepted 18 May 2018 Available online 19 May 2018 0266-3538/ © 2018 Elsevier Ltd. All rights reserved.
AM-based solution for freely depositing continuous fibre-reinforced thermoplastic composites in three-dimensional space [11]. CLF is capable of depositing free-standing self-supporting filaments by spatial extrusion without the use of supporting sacrificial structures (see Fig. 1A), as well as laying consecutive layers directly onto a substrate (see Fig. 1B). Examples of applications in which these manufacturing capabilities are practical are in the production of lattice cores for ultralightweight sandwich panels (see Fig. 1C) and for locally-reinforcing structures with stiffening elements (see Fig. 1D). CLF functions by means of a serial pultrusion-extrusion system (see Fig. 2A) that enables the in situ consolidation of cost-effective unconsolidated feedstock materials, e.g. commingled yarns. A feeder system pulls the incoming bundle of commingled yarns through a temperature-controlled tapered pultrusion module where the thermoplastic composite intermediate material is melted, consolidated, and cooled. The upper two optical microscopy images in Fig. 2B show crosssectional images of the material before and after being processed in the pultrusion module. These images indicate that an excellent degree of consolidation and very low void content in the pultruded material is
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 1. The CLF process is applied to shape in situ consolidated continuously reinforced thermoplastic filaments. The CLF print path trajectory is determined by the relative movement between the CLF processing head and the substrate in all directions, including the out-of-plane direction. CLF is capable of (A) printing suspended free-form structures as well as (B) directly depositing fibre strands onto a base substrate. Structures made by CLF include amongst others, (C) ultralightweight sandwich cores composed of structures with slender lattice networks and (D) multi-layered stiffening elements onto a FRPC plate.
Fig. 2. The CLF head is comprised of a two-stage pultrusion-extrusion system. (A) Photograph of CLF head. (B) Schematic of the CLF process, illustrating the active thermal management of the heated and cooled pultrusion module as well as the heated extrusion stage. Heat flow into and out of the composite material is represented by Q˙ . Optical microscopy images show the unconsolidated feedstock material prior to being pultruded, the consolidated semi-finished prepreg material before extrusion and the final extruded material indicating some level of deconsolidation. (C) Schematic of CLF head with integrated pulling force sensor.
the deconsolidation of fibre-reinforced thermoplastic composites: (i) the expansion of trapped gases, (ii) bubble coarsening and coalescence, and (iii) the decompaction of the fibrous media [13–15]. Of these mechanisms, the decompaction of the fibrous media has been reported to dominate the deconsolidation behaviour [16]. When fibre-reinforced thermoplastic composites are processed under heat and applied pressures, the fibrous media experiences elastic and inelastic deformations [17–19]. These deformations and the associated stresses are frozen into the consolidated material upon cooling and solidification of the thermoplastic matrix. When the thermoplastic matrix is re-melted, the stored elastic energy may be released, resulting in the expansion of the material and the formation of voids. The degree to which the fibre network decompresses is dependent on the fibre properties, the fibre network configuration, the fibre volume content,
achieved. The fully consolidated material is then fed into a temperature-controlled extrusion module similar to those used in fused filament fabrication [12], where it is reheated and discharged from the extrusion nozzle at a temperature above the melting point of the polymer so that it can be formed into the desired shape. However, upon reheating, significant deterioration in material quality is observed as shown by the increase in void content found in the optical microscopy image provided in the lower micrograph of Fig. 2B. This formation of voids is attributed to the deconsolidation of the FRPC material which occurs when residual stresses are released as the temperature of the thermoplastic matrix rises above its melt/glass transition temperature. The deconsolidation of fibre-reinforced thermoplastic composite materials upon reheating has been studied in literature. Three mechanisms have been identified as being responsible for providing the driving force for 249
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
and the processing history of the material [20]. In particular, processing history has been shown to play a significant role in the deconsolidation of dry fibre beds and it has been reported that the degree of fibre decompaction can be reduced by subjecting the fibrous materials to repeated cycles of compaction and decompaction [21,22]. Cycling the load on a fibre bed produces non-elastic deformations resulting from increased fibre alignment and fibre nesting which reduce the total magnitude of the residual stresses in the compacted material; this phenomenon is known as cyclic softening. In addition to dry fibre beds, cyclic softening has also been observed in fibre-reinforced thermoplastic composite materials [23], indicating that residual stresses can be reduced by exposing thermoplastic composites to several compaction and decompaction cycles. Because the deconsolidation of fibre-reinforced thermoplastic composites leads to higher void contents and thus lower mechanical performance, reducing the degree of deconsolidation is an imperative if AM is to be viable for manufacturing high performance end-use components from FRPC [24–26]. While the thermal deconsolidation and fibre decompaction of FRPC have been investigated in literature for different manufacturing processes [13,16], the implications of deconsolidation in AM processes have not yet been addressed. CLF, like other AM processes, is non-stationary and an understanding of how processing parameters affect deconsolidation during the build process must still be developed. In this study, we investigate the deconsolidation in the extrudate from the CLF process as function of the initial void content and the processing history of the material. Furthermore, the effects of the extrusion speed and temperature of the CLF process on the dynamics of deconsolidation is studied, and a model that describes thermal deconsolidation in the CLF process is presented. Based on the deconsolidation behaviour observed and modelled, a strategy to suppress the deconsolidation effect in CLF is implemented and a significant increase in the quality of the CLF processed FRPC material is demonstrated.
2.3. Material characterization
2. Materials and methods
A χvoid = ⎛ v ⎞·100% ⎝ At ⎠
2.3.1. Thermal properties Differential scanning calorimetry (DSC) (DSC 1, STAR System, Mettler Toledo) was used to determine the specific heat capacity, melt temperature, and glass transition temperature of the material. Specific heat capacity values were taken from the reversing curve of an alternating differential scanning calorimetry (ADSC) measurement by dividing the heat flow by the mass of the sample and the heating rate [28]. A saw-tooth waveform temperature modulation was used to perform a temperature sweep from 25 °C to 190 °C with stepwise inclinations of +20 °C and −10 °C, at a constant rate of ± 5 °C/min. The effects of the sample crucible were removed by subtracting the curve of a blank (empty) specimen, see Fig. A1. The transverse and longitudinal thermal conductivities of the consolidated FRPC material were measured using a custom-developed guarded hot plate device designed for low thermal conductivity materials on a compression moulded sample measuring a 50 mm × 50 mm x 12 mm. The transverse and longitudinal conductivities as a function of temperature are plotted in Fig. A2 and Fig. A3, respectively. 2.3.2. Void content Optical microscopy was used to determine the void content and morphology on cross-sectional samples of the processed material. The samples were prepared by embedding them in resin (SpeciFix 20, Struers, USA), and polishing (Abramin, Struers, Denmark) using coarse and finegrained grinding discs (MD Piano 120 to MD Nap, Struers, Denmark) combined with appropriate polishing solutions (DiaPro Allegro/Largo 9 μm to Diapro Nap R 1 μm) to obtain a smooth surface. Digital images were taken using an optical microscope (DM RXA, Leica, Germany). The images were analyzed using the software package Leica QWin (Leica, Germany) to determine void content and highlight morphological details. Void contents were calculated using the following equation (Eq. (1)): ⎜
2.1. Process description and setup
⎟
(1)
where At is the total measured cross-sectional area and Av is the measured void area. To ensure statistical relevance, at least 5 individual measurements for each processing configuration were analyzed to determine the void content values presented in this study.
The CLF setup is comprised of an in-house built extrusion head, a robotic manipulator system (Kuka KR2, Kuka AG, Germany) and a corresponding software system (KRL, Kuka AG, Germany). The heated sections of the hybrid pultrusion-extrusion system are made of brass with a nominal cross-sectional area of 1.42 mm in diameter. The pultrusion module is comprised of a tapered die with the aspect ratio 1/50 and a total heated length of 25 mm. In this study, thermoplastic fibre composite rods are processed by the CLF system at a constant pultrusion temperature of 250 °C and a constant extrusion temperature of 230 °C. Extrusion speeds were varied from 52 to 315 mm/min. Cooling of the discharged extrudate is achieved by convective cooling with the ambient air at a temperature of approximately 23 °C.
2.3.3. Volumetric expansion The volumetric expansion of the test samples at elevated temperatures was determined by an optical dilatometer (DIL 806, Texas Instruments, USA) operated in the setup shown in Fig. 3. The temperature range investigated was 25 °C–200 °C and a constant heating rate of 20 °C/min was used. The extremely high viscosity of the fibrefilled melt [29] prevented the molten polymer composite from collapsing and provided the mechanical stability that allowed the radial expansion of the material to be measured. Given that the material is quasiisotropic and that the reinforcement fibres are unidirectionally aligned, the radial expansion of the test specimen is used to quantify the extent of the deconsolidation. The radial expansion was recorded as a function of temperature and time. The void contents of the deconsolidated specimens were calculated from the radial expansion measurements using the following equation (Eq. (2)):
2.2. Feedstock material The CLF process design is adaptable to a variety of different intermediate materials, such as unconsolidated powder-coated, sheathed or commingled yarns as well as pre- or fully impregnated tapes [27]. In this work, commingled yarns from Schappe Technologies consisting of stretch-broken STS40 carbon fibres (Toho Tenax) and polyamide 12 (Grilamid®, EMS Chemie) were used as feedstock material. The constituent yarn has a high fibre volume content of 52% v/v and has been reported to have a tensile stiffness of 83 GPa and a tensile strength of 560 MPa when fully consolidated [11]. The melt (Tm ) and glass transition (Tg ) temperature of the polyamide 12 was measured to be 176.15 °C and 41 °C, respectively. The CLF process is not limited to this specific polymer-reinforcement combination and can be adapted to other materials by varying the processing parameters.
⎛ χvoid = χvoid, i + ⎜1 − ⎜ ⎝
⎞
1
(
D% 100
)
2⎟
+1 ⎟ ⎠
·100 %
; with T > Tm (2)
where D% is the radial expansion measured by the dilatometer as a percentage of the initial radius and χvoid, i is the initial void content of the specimen. To ensure statistical relevance, at least 5 individual measurements for each processing configuration were analyzed to estimate the void content values presented in this study. 250
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
2.4. Surface imaging A scanning electron microscope (SEM) (LEO 1530, Zeiss, Germany) was used to visualize the surface morphology of the extrudate. No sputtering of the samples was required. 2.5. Numerical simulation A quantitative assessment of the temperature profile of the extrudate in the extrusion nozzle was made by a conjugate heat transfer simulation in Comsol Multiphysics®. The extrusion process is modelled as a rigid body extrusion nozzle made of brass and a boundary representation in which the fluid body may undergo laminar flow. All nozzle dimensions and the material characteristics are provided in the results section. The interaction length is introduced as the longitudinal contact length between extrudate and nozzle, which can be used in combination with the extrusion diameter to calculate the contact area. The model required inputs for material properties such as thermal conductivities and heat capacity which were experimentally measured. 2.6. Pulling force measurements The pulling force for the pultrusion module was measured by a force sensor (Typ 8438-5500, Burster, Germany) that was integrated into the pultrusion module, see Fig. 2C. Loose guiding pins were used to mount the pultrusion unit onto the force sensor in order to minimize the effect of thermal expansion of the different components. Fig. 3. An optical dilatometer is used to measure the radial expansion of a fibre composite profile. The test sample was placed vertically in a temperature-controlled chamber and the thermal expansion in the radial direction was tracked by a perpendicularly orientated laser curtain.
3. Results and discussion 3.1. Deconsolidation of CLF profiles To study the deconsolidation behaviour of the FRP material during the extrusion process, the effects of the extrusion module had to be first
Fig. 4. Comparison of the consolidated composite material before and after reheating at ambient pressure shows that significant deconsolidation occurs. (A) SEM image of the material obtained directly from the pultrusion module showing a smooth surface and (B) optical image of the cross-section indicating resin-rich regions between the yarns. (C) SEM image of the FRP material after deconsolidation showing significant delamination on the surface of the rod and (D) optical microscopy image of the cross-section of the deconsolidated material showing the appearance of large internal voids along the yarn interfaces. (E) Comparison of average void content of the pultruded material obtained by optical microscopy with that reheated and deconsolidated material obtained by optical microscopy and optical dilatometry. Samples shown here were processed at a speed of 105 mm/min.
251
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
the fibre-reinforced thermoplastic composite rod. Note that at the maximum temperature of 200 °C, the radial expansion of the sample reaches a plateau after which expansion stops. Examination of the dilated samples under microscopes indicated that significant deconsolidation occurred when the samples were reheated. Most significant were the cracks which appeared on the SEM surface images of the samples which are shown in Fig. 4C. These cracks were primarily the result of delamination occurring in the resin-rich regions between the individual yarns as shown in Fig. 4D. However, many large voids were also detected within the individual yarn segments. The average void content of the deconsolidated samples was measured using optical dilatometry to be 24.26% (SD 2.28%). However, the optical dilatometry-based method of determining void content overestimates the volume of voids due to distortions in the profile of the specimen; Eq. (2) assumes a cylindrical profile and this assumption may not always be accurate for describing the shape of deconsolidated profiles as seen in Fig. 4D. When the void contents of the dilated specimens were assessed using optical microscopy, an average void content of 14.97% (SD 1.69%) was found. A graphical comparison between the total void contents measured in the pultruded materials and the deconsolidated pultruded materials is provided in Fig. 4E. The void contents measured here are consistent with values found in literature which report void contents of 10%–55% after deconsolidation [14,32]. The remainder of this paper looks at how the characteristics of the material fed into the CLF extrusion module affect the quality of the material extruded. Specifically, the effects of the initial void content of the material and the processing history on how the material deconsolidates when reheated in the CLF extrusion module are investigated. Unless otherwise indicated, void content measurements based on both the optical dilatometry and the optical microscope image analysis are provided to allow accurate comparisons. To allow a conversion between the void contents obtained by optical microscopy and optical dilatometry, a correction factor δ is introduced and represents the fraction of the measured void content by optical microscopy with respect to the measured void content by optical dilatometry.
isolated from those of the pultrusion module. To achieve this, samples of FRPC materials were assessed for surface morphology, microstructure, and void content directly after cooling in the pultrusion module, i.e. before entering the extrusion module. The unconsolidated pultruded sample rods appeared fully consolidated. Fig. 4A indicates that the surface of the material exiting the pultrusion module was smooth with no noticeable cracks. The cross-sectional imaging of the pultruded material, as shown in Fig. 4B, indicated that the interfaces between the seven individual commingled yarns are characterized by resin rich regions which can be vulnerable to delamination. Based on images like Fig. 4B, the average void content of the material exiting the pultrusion module was calculated to be 0.70% (SD 0.27%). The voids were measured to have radii less than 10 μm. The material quality obtained from the pultrusion module is considered excellent and well below the standard acceptable void content of 1% for thermoplastic composites used in the aerospace industry [30]. The void content for the pultruded rods did not vary significantly when line speeds up to 315 mm/min were used. To observe the thermal deconsolidation of the composite in a controlled environment, the pultruded rods were exposed to elevated temperatures and ambient pressure inside an optical dilatometer. The extent to which a composite material undergoes deconsolidation is affected by the temperature, the amount of time at the elevated temperature, the viscosity of the polymer, and the geometry of the sample [31]. The samples were allowed to expand under controlled conditions until thermal equilibrium was reached and all residual stresses underwent relaxation, i.e. no further expansion was observed. A maximum deconsolidation temperature of 200 °C was selected because this temperature is between the temperature of the extrusion stage at 230 °C and the melting point of the polymer at 176 °C, and is therefore representative of the average temperatures experienced by the material during the extrusion process. To ensure ample time to reach thermal equilibrium, the samples were held at the maximum temperature of 200 °C for 5 min. Fig. 5 is an example of a typical curve obtained from the dilatometer. The samples were observed to expand radially at a moderate rate until shortly after the system surpasses the melting temperature of 176 °C. The moderate rate of expansion is attributed to the thermal expansion of the materials, while the more sudden and increased rate of radial expansion is attributed to the deconsolidation of the fibre-reinforced thermoplastic composite. The time delay between the time at which the system surpasses the melt temperature of the matrix and the sudden expansion of the material is attributed to the time needed for the thermal energy to conduct through the thickness of
3.2. Influence of initial void content on deconsolidation Voids present in the material entering the extrusion module can be modelled as trapped gasses which expand upon heating and reduced pressures. Such gas expansion can be estimated using the YoungLaplace equation (Eq. (3)) and the ideal gas law (Eq. (4)). The YoungLaplace equation estimates the equilibrium pressure inside the trapped
Fig. 5. Optical dilatometry was used to measure the evolution of thermal expansion during deconsolidation. The dilatometer raw data provides the radial elongation of the sample and the applied temperature profile as a function of time. The sample shown here is of a pultruded material processed at 105 mm/min. 252
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
longitudinal to the fibres, thus preventing a build-up of high bubble pressures [34]. The results of the parametric study suggest that the initial void content may only have moderate influence on the expected total void content after deconsolidation and is not sensitive to the morphology of the void distribution and size. The effect of deconsolidation is expected to be greater in materials with higher initial void contents. For example, according to Fig. 6, given an initial void content of 8% and initial void size of 10 μm, the expansion of the trapped gasses may result in a final void content of approximately 12.2% (+Δ 4.2%) compared to only 1.60% (+Δ 0.6%) for an initial void content of 1%. Experimentally, the initial void contents and void sizes in the CLF pultruded material were found to be ∼0.7% and ∼10 μm, respectively. To experimentally investigate the contribution of expanding trapped gases on the total deconsolidation of the thermoplastic composites, the deconsolidation behaviour of two different extrudates with different initial void contents and sizes were compared. If the contribution of trapped gasses to the deconsolidation is statistically significant, then a difference in the final void contents of the deconsolidated specimens would be observed. The first set of extrudates had initial void contents of 0.70% (SD 0.27) and initial average void radius of approximately 10 μm, while the second set of extrudates had initial void contents of 3.17% (SD 0.72) and an initial averaged void radius of approximately 22 μm, as determined by optical microscopy. The sample with the higher initial void contents was created by a pultrusion die with an outlet diameter of 1.44 mm, which is 0.02 mm larger than that used to produce the samples with the lower void contents. Typical cross-sectional images that illustrate the initial void size and distribution which were taken before the samples were deconsolidated can be found in the supplementary material as Fig. A4. An additional set of pultruded specimens with a void content of 3.17% were produced and degassed in a vacuum oven at 70 °C for 48 h before deconsolidation to investigate whether degassing the material would have any impact on the deconsolidation behaviour. All three types of samples were observed to expand comparably upon reheating and were measured to have void contents of approximately 15% after deconsolidation regardless of the initial void content, void size, or degassing procedures as shown in Fig. 7. If, in the CLF process, the influence of the initial void content of the material on the deconsolidation behaviour of the extrudate were to be high, a measurable difference between the two sets of materials would be expected. However, the experimental validation did not indicate a
spherical bubbles as a function of surface tension and bubble size. The ideal gas law (Eq. (4)), relates the volume of an amount of gas to its temperature and pressure.
2γ R
(3)
Tf V pf = ⎛ ⎞ ⎜⎛ i ⎟⎞ pi ⎝ Ti ⎠ ⎝ Vf ⎠
(4)
Δ p= −
⎜
⎟
In Eq. (3), Δp is the pressure difference across the bubble interface, γ is the surface tension of the polymer melt, and R is the bubble radius. In Eq. (4), Vf represents the expected volume of the expanded gas at the final deconsolidation temperature Tf and pressure pf as calculated based on the initial void volume Vi found at the initial temperature Ti and pressure pi . The total pressure ptot of a spherical void at the deconsolidation conditions can then be approximated as the sum of the pressures that act on the bubble as shown in Eq. (5). 3
Tf R 2γ ptot = pf + Δ p= ⎛ ⎞ ⎛⎜ i ⎞⎟ pi − Rf T R i f ⎝ ⎠⎝ ⎠ ⎜
⎟
(5)
where Ri represents the initial radius of the void and Rf is the final radius of the void after deconsolidation. Eq. (5) can be rewritten as:
Tf ptot Rf3 + 2γRf2 − ⎛ ⎞ Ri3 pi = 0 ⎝ Ti ⎠ ⎜
⎟
(6)
After Eq. (6) has been solved for Rf using a graphical method, the final void content χvoid, f can be given as a function of Rf , Ri , and the initial void content χvoid, i as shown in Eq. (7):
Χvoid, f =
Rf3 Ri3
(
1 Χvoid, i
)
− 1 + Rf3
·100% (7)
A parametric study based on Eq. (6) and Eq. (7) was conducted by varying the initial void content from 1 to 8% and the initial average radius of the voids from 10 to 30 μm to investigate their effects on the expected void content of the CLF material after deconsolidation. The results, which are depicted in Fig. 6, were calculated using a total pressure ptot = pi = 1bar , a final temperature of Tf = 230°C , an initial ambient temperature of Ti = 25°C and a polymer surface tension of γ = 30mN / m [33]. This model assumes that, in the pultrusion of commingled yarns, trapped air may freely evacuate through channels
Fig. 6. Simulations show moderate influence of initial void content on final void content after deconsolidation. The Young-Laplace equation and the ideal gas law was used to calculate the expected final void content χvoid, f of the deconsolidated materials based on different initial void contents χvoid, i and void size Ri . The gas pressure ptot = pi was kept constant at 1bar and the final deconsolidation temperature Tf was set to 230 °C. 253
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
consolidation-deconsolidation cycle the CLF extrudates were subjected to, reductions in the pulling force required to draw the material through the pultrusion die were observed. Measurements of the average pulling force as a function of the number of consolidation-deconsolidation cycles are plotted in Fig. 8 and show that the average pulling force was reduced from 63 N to 10 N when the number of consolidation-deconsolidation cycles was increased from 1 to 4. Assuming no material was lost during processing, this decrease in pulling force is a good indication of inelastic deformations and the presence of cyclic softening in the unidirectionally reinforced CLF material. The second indicator of cyclic softening was the extent to which the number of processing cycles affected the final void content measured when the samples were reheated in the optical dilatometer. The results, which are plotted in Fig. 9, show that increasing the number of consolidation-deconsolidation cycles results in lower void contents after deconsolidation. The greatest degree of deconsolidation was observed in the samples that underwent only one processing cycle and were found to have an average void content of 14.97% (SD 1.69%). The lowest degree of deconsolidation was observed in samples that underwent four processing cycles, these materials had average void contents of 4.64% (SD 0.45%). The reductions in void content are greatest between the first three processing cycles and seem to approach a plateau by the fourth, indicating that some elastic deformations persist in the material. The effects of cyclic softening on the pulling force and the deconsolidation reported here conform to similar studies in literature [23], and indicate that fibre decompaction plays a considerable role in the deconsolidation of the CLF extrudates.
Fig. 7. The variation of the initial void content did not show any statistically significant influence on the final void content after deconsolidation by experimental testing of pultruded test specimen in a heated chamber. Bar graph showing the void contents from deconsolidated pultruded rods with different initial void contents. Results shown here were taken from samples processed at a speed of 105 mm/min.
3.4. Effect of cyclic softening on deconsolidation rates significant change in the deconsolidation behaviour. Furthermore, degassing the materials also did not lead to measurable differences. This suggests that the initial void content does not play a significant role in the deconsolidation behaviour of the fibre-reinforced thermoplastic composites in the CLF process.
The rate at which deconsolidation occurs is important in CLF processing because it determines the amount of time the extruded material can be kept at an elevated temperature for forming without deconsolidating the material beyond a certain degree. In this respect, the deconsolidation rate effectively determines the working time of the CLF extrudate as it exits the extrusion nozzle and begins to simultaneously deconsolidate and solidify. Deconsolidation rates in thermoplastic composites have been previously reported to be affected by the viscosity of the polymer and hence are temperature-dependent [31]. In the CLF process, the processing history of the material, i.e. cyclic softening, was also observed to have a measurable effect on the deconsolidation rates of the CLF extrudates. To investigate the effect of cyclic softening on the deconsolidation rates, the optical dilatometry curves for the samples measured in Fig. 9 were analyzed further to extract their time response behaviour. Representative optical dilatometer curves are provided in Fig. 10 for CLF materials that have undergone one and four consolidation-deconsolidation cycles. The onset of deconsolidation was taken to be at the time when the temperature reached the melting point of the matrix and the thermal expansion from room temperature to melting temperature is d not considered for the void calculation. The deconsolidation rate dt χvoid was taken as the slope measured between onset of deconsolidation time = 0 s, and the value at time = 5s. The reported values were averaged over all five samples with the same processing history and the results are tabulated in Table 2. These results show that cyclic softening can be used in the CLF processing not only to reduce the total degree of deconsolidation in the extrudate material, but also the rate at which it occurs. By minimizing the rate and maximum degree of deconsolidation in the reheated fibrereinforced material through cyclic softening as well as minimizing the time the CLF extrudate is above the melt temperature of the thermoplastic polymer, it should be possible to minimize the final void content of the extruded material and hence maximize the mechanical properties of the printed structures.
3.3. Influence of cyclic softening on deconsolidation To investigate the contribution of fibre decompaction on the deconsolidation of the CLF extrudates, the unidirectionally reinforced material was subjected to multiple consolidation (fiber compaction in the pultrusion module) and deconsolidation (fiber decompaction in the extrusion module) cycles to observe the effects, if any, of cyclic softening. This was achieved by manually reloading the extruded material into the CLF pultrusion module and passing it through the extrusion module up to 4 times. The speed was set to 105 mm/min and not varied. Table 1 summarizes the processing history of the different sets of CLF samples that were investigated. Note that the samples were subjected to a final deconsolidation step inside the optical dilatometer. Two important indicators of cyclic softening were observed when the different sets of CLF materials were compared for the effects of processing history. The first was, that with each additional Table 1 Processing history of the CLF materials before final deconsolidation in the heated chamber of the dilatometer. # of consolidationdeconsolidation cycles
# of consolidation cycles through CLF pultrusion module
# of deconsolidation cycles through CLF extrusion module
1 2 3 4
1 2 3 4
0 1 2 3
254
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 8. Pulling force measurements indicate that cyclic softening occurs when the material is processed multiple times. Pulling force measurements as a function of the number of consolidation-deconsolidation cycles. The number indicates how many times the fibre-filled material has seen a consolidation by pultrusion and consequent deconsolidation by thermal treatment above melting point. Samples shown here were processed at a speed of 105 mm/ min.
Table 2 Deconsolidation rates of composites with different processing history. # of consolidation-deconsolidation cycles Averaged deconsolidation rate [%/s]
Standard deviation [%/s]
1 1 4 4
0.21 0.13 0.09 0.07
without correction factor with correction factor δ = 0.62 without correction factor with correction factor δ = 0.77
2.45 1.52 0.44 0.34
3.5. Modelling of deconsolidation in CLF Based on the observations above, a model was developed in order to predict the deconsolidation behaviour of the carbon fibre-reinforced PA12 material during CLF processing. The model is comprised of a numerical model to simulate the temperature distribution within and after the extrusion module as well as a temperature-dependent empirical model to anticipate the state of deconsolidation. A summary of the experimentally measured material and geometrical properties used in the model are provided in Table 3. Fig. 11 shows the temperature profiles and temperature evolutions of the CLF extrudate after exiting the extrusion nozzle as a function of displacement and corresponding travel time for different line speeds as predicated by the Comsol Multiphysics® model. Note that the temperature profiles of the extrudates vary significantly depending on the line speed due to the time allowed for energy to transfer in and out of the material. For example, at a line speed of 52.5 mm/min, the extrudate has a uniform temperature of 230 °C at the exit of the extrusion nozzle and the core of the extrudate is predicted to cool at a slightly
Fig. 9. Optical dilatometer measurements show a significant reduction in the maximum degree of deconsolidation in materials after multiple processing cycles. Samples shown here were processed at a speed of 105 mm/min.
Table 3 Extrusion nozzle dimensions and implemented material characteristics for conjugated heat transfer model in Comsol Multiphysics®.
Fig. 10. The dynamic deconsolidation behaviour is affected by the processing history of the fibre-filled material. The diagram shows the deconsolidation curves of a one-time and a four-time processed test specimen recorded by an optical dilatometer.
255
Property
Unit
Value
heated length (Lh) extrusion diameter (D) heat conductivity brass heat conductivity air heat conductivity extrudate (transverse) heat conductivity extrudate (longitudinal) heat transfer coefficient air-extrudate specific heat capacity extrudate
mm mm W/(m·K) W/(m·K) W/(m·K) W/(m·K) W/(m·K) J/(kg·K)
3.5 1.40 120 [35] 0.05 [35] Fig. A2 Fig. A3 50 [35] Fig. A1
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 11. The deconsolidation time is one of the major processing parameters that determine void formation in the CLF extrusion process. The deconsolidation time describes the duration in which the extrusion material is heated above melting point. It starts when the material leaves the extrusion nozzle and finishes when the average temperature of the extrudate drops below melting temperature. The process is simulated using Comsol Multiphysics®.
Table 4 Comparison between predicted and measured void contents for the CLF extrudate processed under difference conditions. Processing speed [mm/min]
52.5 105 210 315
Cooling length [mm]
5.68 9.41 11.13 3.67
Deconsolidation time (tc) [s]
6.49 5.38 3.18 0.70
max. average polymer temperature (Ta,max) [°C]
221.39 220.06 205.84 190.47
Predicted void content [%]
Measured void content [%]
1 time processed
4 time processed
1 time processed
4 time processed
6.48 5.34 2.56 0.79
1.89 1.64 1.01 0.62
7.47 4.31 2.63 0.84
1.49 1.17 0.87 0.84
(SD (SD (SD (SD
1.25) 0.86) 0.74) 0.43)
(Tm − Ta, max ) ⎞⎟· d χ (Γ, υ )· Tm χvoid, p = χvoid, i + ε·⎜⎛1 − e f void ⎝ ⎠ dt tc (V , Tn, Lh , D , α, λ, cp)
lower rate than the surface. Conversely, at a line speed of 315 mm/min, only the surface of the extrudate is heated to 230 °C by the time the material exits the extrusion nozzle, with the core of the extrudate never fully reaching the melt temperature despite thermal conduction from the heated surface. Therefore, under these CLF processing conditions, materials processed at higher line speeds will experience lower average temperatures during the extrusion process than those processed at lower line speeds. As a result, the degree to which the CLF extrudate will deconsolidate upon exiting the extrusion is expected to differ depending on the processing conditions. Assuming that deconsolidation occurs whenever the average temperature of the thermoplastic composite is above the melting temperature, the deconsolidation time tc can be defined for each set of processing conditions as the cooling length required for the average temperature to fall below the melt temperature divided by the appropriate line speed. The final void content in the deconsolidated material can then be estimated using Eq. (8).
(SD (SD (SD (SD
0.36) 0.34) 0.13) 0.16)
(8)
where ε is an experimentally determined scaling factor which is determined to be 2.64 by linear regression using the measured data. Note d that the deconsolidation rate dt χvoid is expected to be a function of polymer temperature T , processing history Γ , and fibre volume fraction υf , and tc is assumed to be a function of extrusion speed V , nozzle temperature Tn , the heated length of the nozzle Lh and the extrusion diameter D , as well as the material properties such as the heat transfer coefficient α (air-extrudate), heat conductivity λ and the specific heat d capacity cp . Deconsolidation rates of dt χvoid = 1.52%/ s and d χ dt void
= 0.34%/ s were used for the materials subjected to one and four processing cycles, respectively. The exponential term in Eq. (8) is used d to accommodate the temperature dependency of dt χvoid and considers the melt temperature Tm and the maximum average polymer temperature Ta, max at the extrusion exit. 256
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 12. A four stage pultrusion module was developed and integrated into the CLF system to exploit the advantages of cyclic softening. The heated section of the pultrusion is designed to gradually compact and decompact the fibre-filled melt. (A) Schematic showing the design and design parameters of the multi-stage pultrusion module. (B) Photograph of the implemented multi-stage pultrusion system.
Fig. 13. Simulations were carried out to determine the maximum deconsolidation reached between the individual stages of the four-stage pultrusion module. Simulations were conducted at a constant maximum average polymer temperature Ta, max of 250 °C. (A-C) Calculated cross-sectional areas of the individual pultrusion stages and the simulated cross-sectional areas of the pultruded material at a given position within the pultrusion module. (D) Maximum deconsolidation reached in-between the different pultrusion stages of the pultrusion module for different pultrusion speeds.
The void contents predicted for the CLF extrudate by the model are compared to the final void contents measured by optical microscopy in Table 4. A good correlation between the model and the experimental
results was found. Both studies showed that the materials that were subjected to cyclic softening expressed lower void contents when extruded, with a reduction of more than 80% recorded when the 257
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 14. Processing history has a significant effect on deconsolidation during CLF processing. (A) Investigations by optical microscopy revealed that the four-time compacted and deconsolidated material shows a significantly lower degree of void formation (deconsolidation) compared to the material which was only one time processed. (B) Cross-sectional images by optical microscopy illustrate the microstructure inside the manually loaded samples before and after deconsolidation for different processing conditions, voids are shown as dark areas.
point in the die. The distance and the associated time elapsed where the cross-sectional area of the die and the pultrudate match were used to calculate the maximum deconsolidation that occurs in-between the individual pultrusion stages. Fig. 13D summarizes the maximum deconsolidation by v/v % void for the different pultrusion speeds. These results show that the geometry of the dies and the pultrusion speed highly affect the maximum degree of deconsolidation under isothermal conditions. Narrower die geometries and higher pultrusion speeds lead to reduced deconsolidation times. At low pultrusion speeds of 52.5 mm/min, deconsolidation is expected to result in void contents greater than 14%, while at high pultrusion speeds of 315 mm/min, the expected void content is reduced to 3.5%. Experimental investigations showed that the quality of the extruded CLF materials that were cyclically softened using the integrated multistage pultrusion module was comparable to the materials that were manually fed through the single pultrusion module four times. Fig. 14A compares the void content measured from the manual process to those measured from the integrated multi-stage pultrusion setup over a range of line speeds along with the predictions from the simulations. All void contents were measured by optical microscopy analysis, and typical cross-sectional images from the manually loaded samples are provided in Fig. 14B. The integrated four-stage pultrusion module tended to produce materials with slightly higher void contents than manually feeding the material into the single pultrusion module 4 times. This effect may be attributed to the incomplete deconsolidation of the thermoplastic composite as both time and space for expansion are limited in the conical sections. Nonetheless, the integrated multi-stage pultrusion system proves that cyclic softening of high fibre-filled polymeric materials can be used in situ to significantly reduce the deconsolidation behaviour of AM FRPC materials.
commingled yarns were consolidated and deconsolidated 4 times at line speeds of 52.5 mm/min. The model slightly deviates from the experimentally measured data at low line speeds up to 150 mm/min. These minor differences between the model and the experimental results can be explained by simplifications made which ignore the temperature dependence of the deconsolidation rate.
3.6. Exploiting cyclic softening in CLF processing To exploit the advantages of cyclic softening to reduce the degree of deconsolidation found in CLF processing and to increase the mechanical properties of the extruded materials, a multi-stage pultrusion design was developed and integrated into the CLF system. The multi-stage pultrusion module is comprised of four conical dies which gradually force the commingled yarns to its final nominal diameter of 1.42 mm. However, the entrance of each die is designed to be larger than the exit of the preceding die so that the thermoplastic composite has the space to locally deconsolidate and relax before undergoing further compaction. The effect of this cascaded multi-stage pultrusion die is to cyclically soften the thermoplastic composite through repeated loading and unloading. A schematic of the multi-stage pultrusion die and a photograph of the integrated setup are presented in Fig. 12. The model developed for the extrusion process was used to estimate the degree of deconsolidation that occurs in-between the individual pultrusion stages of the multi-stage pultrusion module. The maximum average polymer temperature Ta, max was set to 250 °C and kept constant for all pultrusion speeds. Fig. 13A–C show the cross-sectional areas of the individual pultrusion stages and the corresponding cross-sectional areas simulated for the carbon fibre-reinforced PA12 material at a given
258
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
4. Conclusion
1016/j.compscitech.2016.02.021. [10] R. Matsuzaki, M. Ueda, M. Namiki, T.-K. Jeong, H. Asahara, K. Horiguchi, T. Nakamura, A. Todoroki, Y. Hirano, Three-dimensional printing of continuousfiber composites by in-nozzle impregnation, Nat. Publ. Gr (2016), http://dx.doi. org/10.1038/srep23058. [11] M. Eichenhofer, J.C.H. Wong, P. Ermanni, Continuous lattice fabrication of ultralightweight composite structures, Addit. Manuf. 18 (2017) 48–57, http://dx.doi. org/10.1016/j.addma.2017.08.013. [12] I. Gibson, D. Rosen, B. Stucker, Additive Manufacturing Technologies, Springer New York, New York, NY, 2015, http://dx.doi.org/10.1007/978-1-4939-2113-3. [13] L. Ye, M. Lu, Y.-W. Mai, Thermal de-consolidation of thermoplastic matrix composites—I. Growth of voids, Compos. Sci. Technol. 62 (2002) 2121–2130, http://dx.doi.org/10.1016/S0266-3538(02)00144-6. [14] J. Wolfrath, V. Michaud, J.-A.E. Månson, Deconsolidation in glass mat thermoplastics: influence of the initial fibre/matrix configuration, Compos. Sci. Technol. 65 (2005) 1601–1608, http://dx.doi.org/10.1016/j.compscitech.2005.02.001. [15] J. Wolfrath, V. Michaud, J.-A.E. Månson, Deconsolidation in glass mat thermoplastic composites: analysis of the mechanisms, Compos. Part A Appl. Sci. Manuf 36 (2005) 1608–1616, http://dx.doi.org/10.1016/j.compositesa.2005.04.001. [16] M. Brzeski, P. Mitschang, Deconsolidation and its interdependent mechanisms of fibre reinforced polypropylene, Polym. Polym. Compos. 23 (2015) 515–524, http:// dx.doi.org/10.1017/CBO9781107415324.004. [17] T.G. Gutowski, Z. Cai, S. Bauer, D. Boucher, J. Kingery, S. Wineman, Consolidation experiments for laminate composites, J. Compos. Mater. 21 (1987) 650–669, http://dx.doi.org/10.1177/002199838702100705. [18] M. Danzi, F. Klunker, P. Ermanni, Experimental validation of through-thickness resin flow model in the consolidation of saturated porous media, J. Compos. Mater. 51 (2017) 2467–2475, http://dx.doi.org/10.1177/0021998316671574. [19] D.M. Binding, K. Fisher, R.S. Jones, Compression of thermosetting fabric materials, Compos. Part A Appl. Sci. Manuf 31 (2000) 1323–1330 https://doi.org/10.1016/ S1359-835X(00)00032-4 , Accessed date: 10 October 2017. [20] T.G.P. Gutowski, Advanced Composites Manufacturing, Wiley, 1997. [21] A.A. Somashekar, S. Bickerton, D. Bhattacharyya, Exploring the non-elastic compression deformation of dry glass fibre reinforcements, Compos. Sci. Technol. 67 (2007) 183–200, http://dx.doi.org/10.1016/j.compscitech.2006.07.032. [22] J.J. Cheng, P.A. Kelly, S. Bickerton, A rate-independent thermomechanical constitutive model for fiber reinforcements, J. Compos. Mater. 46 (2011) 247–256, http://dx.doi.org/10.1177/0021998311410506. [23] M. Lu, L. Ye, Y.-W. Mai, Thermal de-consolidation of thermoplastic matrix composites—II, “Migration” of voids and “re-consolidation”, Compos. Sci. Technol. 64 (2004) 191–202, http://dx.doi.org/10.1016/S0266-3538(03)00233-1. [24] H. Huang, R. Talreja, Effects of void geometry on elastic properties of unidirectional fiber reinforced composites, Compos. Sci. Technol. 65 (2005) 1964–1981, http:// dx.doi.org/10.1016/j.compscitech.2005.02.019. [25] M. Hou, L. Ye, Y.W. Mai, Materials processing technology manufacturing process and mechanical properties of thermoplastic composite components, J. Mater. Process. Teclmol. 63 (1997) 334–338, http://dx.doi.org/10.1016/S0924-0136(96) 02644-1. [26] P.-O. Hagstrand, F. Bonjour, J.-A.E. Månson, The influence of void content on the structural flexural performance of unidirectional glass fibre reinforced polypropylene composites, Compos. Part A Appl. Sci. Manuf 36 (2005) 705–714, http:// dx.doi.org/10.1016/j.compositesa.2004.03.007. [27] C. Schneeberger, J.C.H. Wong, P. Ermanni, Hybrid bicomponent fibres for thermoplastic composite preforms, Compos. Part A (2017), http://dx.doi.org/10.1016/ j.compositesa.2017.09.008. [28] Z. Jiang, J.M. Hutchinson, C.T. Imrie, Temperature-modulated differential scanning calorimetry. Part II. Determination of activation energies, Polym. Int. 47 (1998) 72–75, http://dx.doi.org/10.1002/(SICI)1097-0126(199809)47:1<72::AIDPI999>3.0.CO;2-N. [29] R.B. Pipes, A.J. Beaussart, J.T. Tzeng, R.K. Okine, Anisotropic viscosities of oriented discontinuous fiber laminates, J. Compos. Mater. 26 (1992) 1088–1099, http://dx. doi.org/10.1177/002199839202600801. [30] D. Zhang, D. Heider, J.W. Gillespie, Void reduction of high-performance thermoplastic composites via oven vacuum bag processing, J. Compos. Mater. 0 (2017) 1–12, http://dx.doi.org/10.1177/0021998317700700. [31] L. Ye, Z.-R. Chen, M. Lu, M. Hou, De-consolidation and re-consolidation in CF/PPS thermoplastic matrix composites, Compos. Part A Appl. Sci. Manuf 36 (2005) 915–922, http://dx.doi.org/10.1016/j.compositesa.2004.12.006. [32] K. Henninger, F. Ye, Lin Friedrich, Deconsolidation behaviour of glass fibre-polyamide 12 composite sheet material during post-processing, Plast. Rubber Compos. Process. Appl. 27(1998) 287–292. HYPERLINK "https://scholar.google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5" \o "https://scholar. google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5" https:// scholar.google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5 (accessed November 24, 2017). [33] K. Wudy, D. Drummer, M. Drexler, Characterization of polymer materials and powders for selective laser melting, AIP Conf. Proc. 1593 (2014) 702–707, http:// dx.doi.org/10.1063/1.4873875. [34] D.-H. Kim, W. Il Lee, K. Friedrich, A model for a thermoplastic pultrusion process using commingled yarns, Compos. Sci. Technol. 61 (2001) 1065–1077 https://doi. org/10.1016/S0266-3538(00)00234-7 , Accessed date: 17 October 2017. [35] K. Langeheinecke, P. Jany, G. Thieleke, K. Langeheinecke, A. Kaufmann, Wärmeübertragung, Thermodyn. Für Ingenieure, Springer Fachmedien Wiesbaden, Wiesbaden, 2013, pp. 242–289, , http://dx.doi.org/10.1007/978-3-658-031695_10.
The thermal deconsolidation of unidirectional fibre-reinforced thermoplastic composites observed in CLF processing was found to be significantly reduced through cyclic softening. Thermoplastic composites that had undergone cyclic softening were observed to have both lower degrees and rates of deconsolidation compared to materials that were not repeatedly consolidated (compacted) and deconsolidated (decompacted). A numerical-empirical model was developed which was able to accurately predict the deconsolidation observed during the CLF process based on the processing history, heat distribution, and thermal evolution of the extrudates. Based on these fundamental studies into the deconsolidation behaviour of the thermoplastic composites, a multistage pultrusion module was designed and implemented into CLF that allowed the final void content of the extrudate to be reduced by more than 80%. This work represents significant advancements in the ability of CLF to print fibre-reinforced composites with mechanical properties suitable for use in high performance end-use components by 1) enlarging the processing window in which material with acceptably low void contents (< 1%) can be deposited and 2) providing a simulation tool that can be used to select the appropriate processing conditions for thermoplastic composites in extrusion-based AM processes. Acknowledgments This work was supported by ETH Zurich's internal research funding program (ETH Research Grant ETH-33 15-2) and the Commission for Technology and Innovation (CTI) through the Swiss Competence Center for Energy Research (SCCER), Efficient Technologies and Systems for Mobility. Commingled yarns were kindly provided by Schappe Technologies. The authors thank Dr. Samuel Brunner of the Swiss Federal Laboratories for Materials Science and Technology, and Prof. André Studart and Dr. Kunal Masania of the Complex Materials group at ETH Zurich for their support received throughout this study. Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx. doi.org/10.1016/j.compscitech.2018.05.033. References [1] A. Gebhardt, Understanding Additive Manufacturing, Carl Hanser Verlag GmbH & Co, KG, München, 2011, http://dx.doi.org/10.3139/9783446431621. [2] S.H. Huang, P. Liu, A. Mokasdar, L. Hou, Additive manufacturing and its societal impact: a literature review, Int. J. Adv. Manuf. Technol. 67 (2013) 1191–1203, http://dx.doi.org/10.1007/s00170-012-4558-5. [3] X. Wang, M. Jiang, Z. Zhou, J. Gou, D. Hui, 3D printing of polymer matrix composites: a review and prospective, Compos. Part B 110 (2017) 442–458, http://dx. doi.org/10.1016/j.compositesb.2016.11.034. [4] B.G. Compton, J.A. Lewis, 3D-Printing of lightweight cellular composites, Adv. Mater. 26 (2014) 5930–5935, http://dx.doi.org/10.1002/adma.201401804. [5] J.P. Lewicki, J.N. Rodriguez, C. Zhu, M.A. Worsley, A.S. Wu, Y. Kanarska, J.D. Horn, E.B. Duoss, J.M. Ortega, W. Elmer, R. Hensleigh, R.A. Fellini, M.J. King, 3D-Printing of meso-structurally ordered carbon fiber/polymer composites with unprecedented orthotropic physical properties, Nat. Publ. Gr (2017), http://dx.doi. org/10.1038/srep43401. [6] P. Parandoush, D. Lin, A review on additive manufacturing of polymer-fiber composites, Compos. Struct. 182 (2017) 36–53, http://dx.doi.org/10.1016/j. compstruct.2017.08.088. [7] W. Zhang, A.S. Wu, J. Sun, Z. Quan, B. Gu, B. Sun, C. Cotton, D. Heider, T.-W. Chou, Characterization of residual stress and deformation in additively manufactured ABS polymer and composite specimens, Compos. Sci. Technol. 150 (2017) 102–110, http://dx.doi.org/10.1016/j.compscitech.2017.07.017. [8] A.N. Dickson, J.N. Barry, K.A. Mcdonnell, D.P. Dowling, Fabrication of continuous carbon, glass and Kevlar fibre reinforced polymer composites using additive manufacturing, Addit. Manuf. 16 (2017) 146–152, http://dx.doi.org/10.1016/j.addma. 2017.06.004. [9] Z. Quan, Z. Larimore, A. Wu, J. Yu, X. Qin, M. Mirotznik, J. Suhr, J.-H. Byun, Y. Oh, T.-W. Chou, Microstructural design and additive manufacturing and characterization of 3D orthogonal short carbon fiber/acrylonitrile-butadiene-styrene preform and composite, Compos. Sci. Technol. 126 (2016) 139–148, http://dx.doi.org/10.
259
Contents lists available at ScienceDirect
Composites Science and Technology journal homepage: www.elsevier.com/locate/compscitech
Exploiting cyclic softening in continuous lattice fabrication for the additive manufacturing of high performance fibre-reinforced thermoplastic composite materials
T
Martin Eichenhofera,∗, Joanna C.H. Wonga,b, Paolo Ermannia a
Laboratory of Composite Materials and Adaptive Structures, Institute of Design, Materials and Fabrication, ETH Zurich, Tannenstrasse 3, 8092 Zurich, Switzerland Department of Mechanical and Manufacturing Engineering, Schulich School of Engineering, University of Calgary, 856 Campus Place NW, Calgary, AB, T2N 4V8, Canada b
A R T I C LE I N FO
A B S T R A C T
Keywords: A. Polymer-matrix composites (PMCs) A. Carbon fibres C. Residual stress E. Pultrusion Additive manufacturing (AM)
Continuous lattice fabrication (CLF) was recently introduced as a new additive manufacturing (AM) technology capable of printing continuous fibre-reinforced thermoplastic composites along desired trajectories in threedimensional space. In a systematic attempt to maximize the mechanical properties of the printed extrudate by minimizing the residual void content, this study investigates the thermal deconsolidation behaviour observed in pultruded unidirectional fibre-reinforced thermoplastic composite material when it is reheated above its melting point and exposed to ambient pressure. Fibre decompaction, generally accepted to be the primary cause for deconsolidation in fibre-reinforced thermoplastics, was investigated to assess the influence of cyclic softening of the fibrous media on the residual void content of the extruded material. The magnitude and rate of fibre decompaction were observed to decrease with the number of consolidation-deconsolidation cycles to which the material was subjected. A model was developed to predict the degree of deconsolidation in the CLF process as a function of temperature, processing speed, and processing history. Based on the deconsolidation behaviour observed, a multi-stage pultrusion module was designed that exploits cyclic softening and was demonstrated to reduce the residual void content of the printed extrudate by over 80%.
1. Introduction Additive manufacturing (AM) technologies have the advantages, in comparison to casting and subtractive manufacturing processes, of eliminating some of the costs associated with complex geometries and individually tailored designs, while also reducing waste [1]. The emergence of AM as a viable approach for the production of end-use components remains the target for many low-volume applications [2]. The AM of high performance fibre-reinforced polymer composites (FRPC) is particularly attractive for enabling a new design space for ultra-lightweight structures in aerospace, medical engineering and robotics amongst others. Although efforts to apply layer-by-layer AM strategies to anisotropic FRPC materials have met with moderate success [3–10], in order to fully harness the potential of FRPC materials in engineering structures, alternatives to the planar layer-by-layer approach must be developed which allow for the orientation of anisotropic materials along all relevant vectors, including those positioned out-of-plane. Recently, we introduced continuous lattice fabrication (CLF) as an
∗
Corresponding author. E-mail address: [email protected] (M. Eichenhofer).
https://doi.org/10.1016/j.compscitech.2018.05.033 Received 13 March 2018; Received in revised form 16 May 2018; Accepted 18 May 2018 Available online 19 May 2018 0266-3538/ © 2018 Elsevier Ltd. All rights reserved.
AM-based solution for freely depositing continuous fibre-reinforced thermoplastic composites in three-dimensional space [11]. CLF is capable of depositing free-standing self-supporting filaments by spatial extrusion without the use of supporting sacrificial structures (see Fig. 1A), as well as laying consecutive layers directly onto a substrate (see Fig. 1B). Examples of applications in which these manufacturing capabilities are practical are in the production of lattice cores for ultralightweight sandwich panels (see Fig. 1C) and for locally-reinforcing structures with stiffening elements (see Fig. 1D). CLF functions by means of a serial pultrusion-extrusion system (see Fig. 2A) that enables the in situ consolidation of cost-effective unconsolidated feedstock materials, e.g. commingled yarns. A feeder system pulls the incoming bundle of commingled yarns through a temperature-controlled tapered pultrusion module where the thermoplastic composite intermediate material is melted, consolidated, and cooled. The upper two optical microscopy images in Fig. 2B show crosssectional images of the material before and after being processed in the pultrusion module. These images indicate that an excellent degree of consolidation and very low void content in the pultruded material is
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 1. The CLF process is applied to shape in situ consolidated continuously reinforced thermoplastic filaments. The CLF print path trajectory is determined by the relative movement between the CLF processing head and the substrate in all directions, including the out-of-plane direction. CLF is capable of (A) printing suspended free-form structures as well as (B) directly depositing fibre strands onto a base substrate. Structures made by CLF include amongst others, (C) ultralightweight sandwich cores composed of structures with slender lattice networks and (D) multi-layered stiffening elements onto a FRPC plate.
Fig. 2. The CLF head is comprised of a two-stage pultrusion-extrusion system. (A) Photograph of CLF head. (B) Schematic of the CLF process, illustrating the active thermal management of the heated and cooled pultrusion module as well as the heated extrusion stage. Heat flow into and out of the composite material is represented by Q˙ . Optical microscopy images show the unconsolidated feedstock material prior to being pultruded, the consolidated semi-finished prepreg material before extrusion and the final extruded material indicating some level of deconsolidation. (C) Schematic of CLF head with integrated pulling force sensor.
the deconsolidation of fibre-reinforced thermoplastic composites: (i) the expansion of trapped gases, (ii) bubble coarsening and coalescence, and (iii) the decompaction of the fibrous media [13–15]. Of these mechanisms, the decompaction of the fibrous media has been reported to dominate the deconsolidation behaviour [16]. When fibre-reinforced thermoplastic composites are processed under heat and applied pressures, the fibrous media experiences elastic and inelastic deformations [17–19]. These deformations and the associated stresses are frozen into the consolidated material upon cooling and solidification of the thermoplastic matrix. When the thermoplastic matrix is re-melted, the stored elastic energy may be released, resulting in the expansion of the material and the formation of voids. The degree to which the fibre network decompresses is dependent on the fibre properties, the fibre network configuration, the fibre volume content,
achieved. The fully consolidated material is then fed into a temperature-controlled extrusion module similar to those used in fused filament fabrication [12], where it is reheated and discharged from the extrusion nozzle at a temperature above the melting point of the polymer so that it can be formed into the desired shape. However, upon reheating, significant deterioration in material quality is observed as shown by the increase in void content found in the optical microscopy image provided in the lower micrograph of Fig. 2B. This formation of voids is attributed to the deconsolidation of the FRPC material which occurs when residual stresses are released as the temperature of the thermoplastic matrix rises above its melt/glass transition temperature. The deconsolidation of fibre-reinforced thermoplastic composite materials upon reheating has been studied in literature. Three mechanisms have been identified as being responsible for providing the driving force for 249
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
and the processing history of the material [20]. In particular, processing history has been shown to play a significant role in the deconsolidation of dry fibre beds and it has been reported that the degree of fibre decompaction can be reduced by subjecting the fibrous materials to repeated cycles of compaction and decompaction [21,22]. Cycling the load on a fibre bed produces non-elastic deformations resulting from increased fibre alignment and fibre nesting which reduce the total magnitude of the residual stresses in the compacted material; this phenomenon is known as cyclic softening. In addition to dry fibre beds, cyclic softening has also been observed in fibre-reinforced thermoplastic composite materials [23], indicating that residual stresses can be reduced by exposing thermoplastic composites to several compaction and decompaction cycles. Because the deconsolidation of fibre-reinforced thermoplastic composites leads to higher void contents and thus lower mechanical performance, reducing the degree of deconsolidation is an imperative if AM is to be viable for manufacturing high performance end-use components from FRPC [24–26]. While the thermal deconsolidation and fibre decompaction of FRPC have been investigated in literature for different manufacturing processes [13,16], the implications of deconsolidation in AM processes have not yet been addressed. CLF, like other AM processes, is non-stationary and an understanding of how processing parameters affect deconsolidation during the build process must still be developed. In this study, we investigate the deconsolidation in the extrudate from the CLF process as function of the initial void content and the processing history of the material. Furthermore, the effects of the extrusion speed and temperature of the CLF process on the dynamics of deconsolidation is studied, and a model that describes thermal deconsolidation in the CLF process is presented. Based on the deconsolidation behaviour observed and modelled, a strategy to suppress the deconsolidation effect in CLF is implemented and a significant increase in the quality of the CLF processed FRPC material is demonstrated.
2.3. Material characterization
2. Materials and methods
A χvoid = ⎛ v ⎞·100% ⎝ At ⎠
2.3.1. Thermal properties Differential scanning calorimetry (DSC) (DSC 1, STAR System, Mettler Toledo) was used to determine the specific heat capacity, melt temperature, and glass transition temperature of the material. Specific heat capacity values were taken from the reversing curve of an alternating differential scanning calorimetry (ADSC) measurement by dividing the heat flow by the mass of the sample and the heating rate [28]. A saw-tooth waveform temperature modulation was used to perform a temperature sweep from 25 °C to 190 °C with stepwise inclinations of +20 °C and −10 °C, at a constant rate of ± 5 °C/min. The effects of the sample crucible were removed by subtracting the curve of a blank (empty) specimen, see Fig. A1. The transverse and longitudinal thermal conductivities of the consolidated FRPC material were measured using a custom-developed guarded hot plate device designed for low thermal conductivity materials on a compression moulded sample measuring a 50 mm × 50 mm x 12 mm. The transverse and longitudinal conductivities as a function of temperature are plotted in Fig. A2 and Fig. A3, respectively. 2.3.2. Void content Optical microscopy was used to determine the void content and morphology on cross-sectional samples of the processed material. The samples were prepared by embedding them in resin (SpeciFix 20, Struers, USA), and polishing (Abramin, Struers, Denmark) using coarse and finegrained grinding discs (MD Piano 120 to MD Nap, Struers, Denmark) combined with appropriate polishing solutions (DiaPro Allegro/Largo 9 μm to Diapro Nap R 1 μm) to obtain a smooth surface. Digital images were taken using an optical microscope (DM RXA, Leica, Germany). The images were analyzed using the software package Leica QWin (Leica, Germany) to determine void content and highlight morphological details. Void contents were calculated using the following equation (Eq. (1)): ⎜
2.1. Process description and setup
⎟
(1)
where At is the total measured cross-sectional area and Av is the measured void area. To ensure statistical relevance, at least 5 individual measurements for each processing configuration were analyzed to determine the void content values presented in this study.
The CLF setup is comprised of an in-house built extrusion head, a robotic manipulator system (Kuka KR2, Kuka AG, Germany) and a corresponding software system (KRL, Kuka AG, Germany). The heated sections of the hybrid pultrusion-extrusion system are made of brass with a nominal cross-sectional area of 1.42 mm in diameter. The pultrusion module is comprised of a tapered die with the aspect ratio 1/50 and a total heated length of 25 mm. In this study, thermoplastic fibre composite rods are processed by the CLF system at a constant pultrusion temperature of 250 °C and a constant extrusion temperature of 230 °C. Extrusion speeds were varied from 52 to 315 mm/min. Cooling of the discharged extrudate is achieved by convective cooling with the ambient air at a temperature of approximately 23 °C.
2.3.3. Volumetric expansion The volumetric expansion of the test samples at elevated temperatures was determined by an optical dilatometer (DIL 806, Texas Instruments, USA) operated in the setup shown in Fig. 3. The temperature range investigated was 25 °C–200 °C and a constant heating rate of 20 °C/min was used. The extremely high viscosity of the fibrefilled melt [29] prevented the molten polymer composite from collapsing and provided the mechanical stability that allowed the radial expansion of the material to be measured. Given that the material is quasiisotropic and that the reinforcement fibres are unidirectionally aligned, the radial expansion of the test specimen is used to quantify the extent of the deconsolidation. The radial expansion was recorded as a function of temperature and time. The void contents of the deconsolidated specimens were calculated from the radial expansion measurements using the following equation (Eq. (2)):
2.2. Feedstock material The CLF process design is adaptable to a variety of different intermediate materials, such as unconsolidated powder-coated, sheathed or commingled yarns as well as pre- or fully impregnated tapes [27]. In this work, commingled yarns from Schappe Technologies consisting of stretch-broken STS40 carbon fibres (Toho Tenax) and polyamide 12 (Grilamid®, EMS Chemie) were used as feedstock material. The constituent yarn has a high fibre volume content of 52% v/v and has been reported to have a tensile stiffness of 83 GPa and a tensile strength of 560 MPa when fully consolidated [11]. The melt (Tm ) and glass transition (Tg ) temperature of the polyamide 12 was measured to be 176.15 °C and 41 °C, respectively. The CLF process is not limited to this specific polymer-reinforcement combination and can be adapted to other materials by varying the processing parameters.
⎛ χvoid = χvoid, i + ⎜1 − ⎜ ⎝
⎞
1
(
D% 100
)
2⎟
+1 ⎟ ⎠
·100 %
; with T > Tm (2)
where D% is the radial expansion measured by the dilatometer as a percentage of the initial radius and χvoid, i is the initial void content of the specimen. To ensure statistical relevance, at least 5 individual measurements for each processing configuration were analyzed to estimate the void content values presented in this study. 250
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
2.4. Surface imaging A scanning electron microscope (SEM) (LEO 1530, Zeiss, Germany) was used to visualize the surface morphology of the extrudate. No sputtering of the samples was required. 2.5. Numerical simulation A quantitative assessment of the temperature profile of the extrudate in the extrusion nozzle was made by a conjugate heat transfer simulation in Comsol Multiphysics®. The extrusion process is modelled as a rigid body extrusion nozzle made of brass and a boundary representation in which the fluid body may undergo laminar flow. All nozzle dimensions and the material characteristics are provided in the results section. The interaction length is introduced as the longitudinal contact length between extrudate and nozzle, which can be used in combination with the extrusion diameter to calculate the contact area. The model required inputs for material properties such as thermal conductivities and heat capacity which were experimentally measured. 2.6. Pulling force measurements The pulling force for the pultrusion module was measured by a force sensor (Typ 8438-5500, Burster, Germany) that was integrated into the pultrusion module, see Fig. 2C. Loose guiding pins were used to mount the pultrusion unit onto the force sensor in order to minimize the effect of thermal expansion of the different components. Fig. 3. An optical dilatometer is used to measure the radial expansion of a fibre composite profile. The test sample was placed vertically in a temperature-controlled chamber and the thermal expansion in the radial direction was tracked by a perpendicularly orientated laser curtain.
3. Results and discussion 3.1. Deconsolidation of CLF profiles To study the deconsolidation behaviour of the FRP material during the extrusion process, the effects of the extrusion module had to be first
Fig. 4. Comparison of the consolidated composite material before and after reheating at ambient pressure shows that significant deconsolidation occurs. (A) SEM image of the material obtained directly from the pultrusion module showing a smooth surface and (B) optical image of the cross-section indicating resin-rich regions between the yarns. (C) SEM image of the FRP material after deconsolidation showing significant delamination on the surface of the rod and (D) optical microscopy image of the cross-section of the deconsolidated material showing the appearance of large internal voids along the yarn interfaces. (E) Comparison of average void content of the pultruded material obtained by optical microscopy with that reheated and deconsolidated material obtained by optical microscopy and optical dilatometry. Samples shown here were processed at a speed of 105 mm/min.
251
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
the fibre-reinforced thermoplastic composite rod. Note that at the maximum temperature of 200 °C, the radial expansion of the sample reaches a plateau after which expansion stops. Examination of the dilated samples under microscopes indicated that significant deconsolidation occurred when the samples were reheated. Most significant were the cracks which appeared on the SEM surface images of the samples which are shown in Fig. 4C. These cracks were primarily the result of delamination occurring in the resin-rich regions between the individual yarns as shown in Fig. 4D. However, many large voids were also detected within the individual yarn segments. The average void content of the deconsolidated samples was measured using optical dilatometry to be 24.26% (SD 2.28%). However, the optical dilatometry-based method of determining void content overestimates the volume of voids due to distortions in the profile of the specimen; Eq. (2) assumes a cylindrical profile and this assumption may not always be accurate for describing the shape of deconsolidated profiles as seen in Fig. 4D. When the void contents of the dilated specimens were assessed using optical microscopy, an average void content of 14.97% (SD 1.69%) was found. A graphical comparison between the total void contents measured in the pultruded materials and the deconsolidated pultruded materials is provided in Fig. 4E. The void contents measured here are consistent with values found in literature which report void contents of 10%–55% after deconsolidation [14,32]. The remainder of this paper looks at how the characteristics of the material fed into the CLF extrusion module affect the quality of the material extruded. Specifically, the effects of the initial void content of the material and the processing history on how the material deconsolidates when reheated in the CLF extrusion module are investigated. Unless otherwise indicated, void content measurements based on both the optical dilatometry and the optical microscope image analysis are provided to allow accurate comparisons. To allow a conversion between the void contents obtained by optical microscopy and optical dilatometry, a correction factor δ is introduced and represents the fraction of the measured void content by optical microscopy with respect to the measured void content by optical dilatometry.
isolated from those of the pultrusion module. To achieve this, samples of FRPC materials were assessed for surface morphology, microstructure, and void content directly after cooling in the pultrusion module, i.e. before entering the extrusion module. The unconsolidated pultruded sample rods appeared fully consolidated. Fig. 4A indicates that the surface of the material exiting the pultrusion module was smooth with no noticeable cracks. The cross-sectional imaging of the pultruded material, as shown in Fig. 4B, indicated that the interfaces between the seven individual commingled yarns are characterized by resin rich regions which can be vulnerable to delamination. Based on images like Fig. 4B, the average void content of the material exiting the pultrusion module was calculated to be 0.70% (SD 0.27%). The voids were measured to have radii less than 10 μm. The material quality obtained from the pultrusion module is considered excellent and well below the standard acceptable void content of 1% for thermoplastic composites used in the aerospace industry [30]. The void content for the pultruded rods did not vary significantly when line speeds up to 315 mm/min were used. To observe the thermal deconsolidation of the composite in a controlled environment, the pultruded rods were exposed to elevated temperatures and ambient pressure inside an optical dilatometer. The extent to which a composite material undergoes deconsolidation is affected by the temperature, the amount of time at the elevated temperature, the viscosity of the polymer, and the geometry of the sample [31]. The samples were allowed to expand under controlled conditions until thermal equilibrium was reached and all residual stresses underwent relaxation, i.e. no further expansion was observed. A maximum deconsolidation temperature of 200 °C was selected because this temperature is between the temperature of the extrusion stage at 230 °C and the melting point of the polymer at 176 °C, and is therefore representative of the average temperatures experienced by the material during the extrusion process. To ensure ample time to reach thermal equilibrium, the samples were held at the maximum temperature of 200 °C for 5 min. Fig. 5 is an example of a typical curve obtained from the dilatometer. The samples were observed to expand radially at a moderate rate until shortly after the system surpasses the melting temperature of 176 °C. The moderate rate of expansion is attributed to the thermal expansion of the materials, while the more sudden and increased rate of radial expansion is attributed to the deconsolidation of the fibre-reinforced thermoplastic composite. The time delay between the time at which the system surpasses the melt temperature of the matrix and the sudden expansion of the material is attributed to the time needed for the thermal energy to conduct through the thickness of
3.2. Influence of initial void content on deconsolidation Voids present in the material entering the extrusion module can be modelled as trapped gasses which expand upon heating and reduced pressures. Such gas expansion can be estimated using the YoungLaplace equation (Eq. (3)) and the ideal gas law (Eq. (4)). The YoungLaplace equation estimates the equilibrium pressure inside the trapped
Fig. 5. Optical dilatometry was used to measure the evolution of thermal expansion during deconsolidation. The dilatometer raw data provides the radial elongation of the sample and the applied temperature profile as a function of time. The sample shown here is of a pultruded material processed at 105 mm/min. 252
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
longitudinal to the fibres, thus preventing a build-up of high bubble pressures [34]. The results of the parametric study suggest that the initial void content may only have moderate influence on the expected total void content after deconsolidation and is not sensitive to the morphology of the void distribution and size. The effect of deconsolidation is expected to be greater in materials with higher initial void contents. For example, according to Fig. 6, given an initial void content of 8% and initial void size of 10 μm, the expansion of the trapped gasses may result in a final void content of approximately 12.2% (+Δ 4.2%) compared to only 1.60% (+Δ 0.6%) for an initial void content of 1%. Experimentally, the initial void contents and void sizes in the CLF pultruded material were found to be ∼0.7% and ∼10 μm, respectively. To experimentally investigate the contribution of expanding trapped gases on the total deconsolidation of the thermoplastic composites, the deconsolidation behaviour of two different extrudates with different initial void contents and sizes were compared. If the contribution of trapped gasses to the deconsolidation is statistically significant, then a difference in the final void contents of the deconsolidated specimens would be observed. The first set of extrudates had initial void contents of 0.70% (SD 0.27) and initial average void radius of approximately 10 μm, while the second set of extrudates had initial void contents of 3.17% (SD 0.72) and an initial averaged void radius of approximately 22 μm, as determined by optical microscopy. The sample with the higher initial void contents was created by a pultrusion die with an outlet diameter of 1.44 mm, which is 0.02 mm larger than that used to produce the samples with the lower void contents. Typical cross-sectional images that illustrate the initial void size and distribution which were taken before the samples were deconsolidated can be found in the supplementary material as Fig. A4. An additional set of pultruded specimens with a void content of 3.17% were produced and degassed in a vacuum oven at 70 °C for 48 h before deconsolidation to investigate whether degassing the material would have any impact on the deconsolidation behaviour. All three types of samples were observed to expand comparably upon reheating and were measured to have void contents of approximately 15% after deconsolidation regardless of the initial void content, void size, or degassing procedures as shown in Fig. 7. If, in the CLF process, the influence of the initial void content of the material on the deconsolidation behaviour of the extrudate were to be high, a measurable difference between the two sets of materials would be expected. However, the experimental validation did not indicate a
spherical bubbles as a function of surface tension and bubble size. The ideal gas law (Eq. (4)), relates the volume of an amount of gas to its temperature and pressure.
2γ R
(3)
Tf V pf = ⎛ ⎞ ⎜⎛ i ⎟⎞ pi ⎝ Ti ⎠ ⎝ Vf ⎠
(4)
Δ p= −
⎜
⎟
In Eq. (3), Δp is the pressure difference across the bubble interface, γ is the surface tension of the polymer melt, and R is the bubble radius. In Eq. (4), Vf represents the expected volume of the expanded gas at the final deconsolidation temperature Tf and pressure pf as calculated based on the initial void volume Vi found at the initial temperature Ti and pressure pi . The total pressure ptot of a spherical void at the deconsolidation conditions can then be approximated as the sum of the pressures that act on the bubble as shown in Eq. (5). 3
Tf R 2γ ptot = pf + Δ p= ⎛ ⎞ ⎛⎜ i ⎞⎟ pi − Rf T R i f ⎝ ⎠⎝ ⎠ ⎜
⎟
(5)
where Ri represents the initial radius of the void and Rf is the final radius of the void after deconsolidation. Eq. (5) can be rewritten as:
Tf ptot Rf3 + 2γRf2 − ⎛ ⎞ Ri3 pi = 0 ⎝ Ti ⎠ ⎜
⎟
(6)
After Eq. (6) has been solved for Rf using a graphical method, the final void content χvoid, f can be given as a function of Rf , Ri , and the initial void content χvoid, i as shown in Eq. (7):
Χvoid, f =
Rf3 Ri3
(
1 Χvoid, i
)
− 1 + Rf3
·100% (7)
A parametric study based on Eq. (6) and Eq. (7) was conducted by varying the initial void content from 1 to 8% and the initial average radius of the voids from 10 to 30 μm to investigate their effects on the expected void content of the CLF material after deconsolidation. The results, which are depicted in Fig. 6, were calculated using a total pressure ptot = pi = 1bar , a final temperature of Tf = 230°C , an initial ambient temperature of Ti = 25°C and a polymer surface tension of γ = 30mN / m [33]. This model assumes that, in the pultrusion of commingled yarns, trapped air may freely evacuate through channels
Fig. 6. Simulations show moderate influence of initial void content on final void content after deconsolidation. The Young-Laplace equation and the ideal gas law was used to calculate the expected final void content χvoid, f of the deconsolidated materials based on different initial void contents χvoid, i and void size Ri . The gas pressure ptot = pi was kept constant at 1bar and the final deconsolidation temperature Tf was set to 230 °C. 253
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
consolidation-deconsolidation cycle the CLF extrudates were subjected to, reductions in the pulling force required to draw the material through the pultrusion die were observed. Measurements of the average pulling force as a function of the number of consolidation-deconsolidation cycles are plotted in Fig. 8 and show that the average pulling force was reduced from 63 N to 10 N when the number of consolidation-deconsolidation cycles was increased from 1 to 4. Assuming no material was lost during processing, this decrease in pulling force is a good indication of inelastic deformations and the presence of cyclic softening in the unidirectionally reinforced CLF material. The second indicator of cyclic softening was the extent to which the number of processing cycles affected the final void content measured when the samples were reheated in the optical dilatometer. The results, which are plotted in Fig. 9, show that increasing the number of consolidation-deconsolidation cycles results in lower void contents after deconsolidation. The greatest degree of deconsolidation was observed in the samples that underwent only one processing cycle and were found to have an average void content of 14.97% (SD 1.69%). The lowest degree of deconsolidation was observed in samples that underwent four processing cycles, these materials had average void contents of 4.64% (SD 0.45%). The reductions in void content are greatest between the first three processing cycles and seem to approach a plateau by the fourth, indicating that some elastic deformations persist in the material. The effects of cyclic softening on the pulling force and the deconsolidation reported here conform to similar studies in literature [23], and indicate that fibre decompaction plays a considerable role in the deconsolidation of the CLF extrudates.
Fig. 7. The variation of the initial void content did not show any statistically significant influence on the final void content after deconsolidation by experimental testing of pultruded test specimen in a heated chamber. Bar graph showing the void contents from deconsolidated pultruded rods with different initial void contents. Results shown here were taken from samples processed at a speed of 105 mm/min.
3.4. Effect of cyclic softening on deconsolidation rates significant change in the deconsolidation behaviour. Furthermore, degassing the materials also did not lead to measurable differences. This suggests that the initial void content does not play a significant role in the deconsolidation behaviour of the fibre-reinforced thermoplastic composites in the CLF process.
The rate at which deconsolidation occurs is important in CLF processing because it determines the amount of time the extruded material can be kept at an elevated temperature for forming without deconsolidating the material beyond a certain degree. In this respect, the deconsolidation rate effectively determines the working time of the CLF extrudate as it exits the extrusion nozzle and begins to simultaneously deconsolidate and solidify. Deconsolidation rates in thermoplastic composites have been previously reported to be affected by the viscosity of the polymer and hence are temperature-dependent [31]. In the CLF process, the processing history of the material, i.e. cyclic softening, was also observed to have a measurable effect on the deconsolidation rates of the CLF extrudates. To investigate the effect of cyclic softening on the deconsolidation rates, the optical dilatometry curves for the samples measured in Fig. 9 were analyzed further to extract their time response behaviour. Representative optical dilatometer curves are provided in Fig. 10 for CLF materials that have undergone one and four consolidation-deconsolidation cycles. The onset of deconsolidation was taken to be at the time when the temperature reached the melting point of the matrix and the thermal expansion from room temperature to melting temperature is d not considered for the void calculation. The deconsolidation rate dt χvoid was taken as the slope measured between onset of deconsolidation time = 0 s, and the value at time = 5s. The reported values were averaged over all five samples with the same processing history and the results are tabulated in Table 2. These results show that cyclic softening can be used in the CLF processing not only to reduce the total degree of deconsolidation in the extrudate material, but also the rate at which it occurs. By minimizing the rate and maximum degree of deconsolidation in the reheated fibrereinforced material through cyclic softening as well as minimizing the time the CLF extrudate is above the melt temperature of the thermoplastic polymer, it should be possible to minimize the final void content of the extruded material and hence maximize the mechanical properties of the printed structures.
3.3. Influence of cyclic softening on deconsolidation To investigate the contribution of fibre decompaction on the deconsolidation of the CLF extrudates, the unidirectionally reinforced material was subjected to multiple consolidation (fiber compaction in the pultrusion module) and deconsolidation (fiber decompaction in the extrusion module) cycles to observe the effects, if any, of cyclic softening. This was achieved by manually reloading the extruded material into the CLF pultrusion module and passing it through the extrusion module up to 4 times. The speed was set to 105 mm/min and not varied. Table 1 summarizes the processing history of the different sets of CLF samples that were investigated. Note that the samples were subjected to a final deconsolidation step inside the optical dilatometer. Two important indicators of cyclic softening were observed when the different sets of CLF materials were compared for the effects of processing history. The first was, that with each additional Table 1 Processing history of the CLF materials before final deconsolidation in the heated chamber of the dilatometer. # of consolidationdeconsolidation cycles
# of consolidation cycles through CLF pultrusion module
# of deconsolidation cycles through CLF extrusion module
1 2 3 4
1 2 3 4
0 1 2 3
254
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 8. Pulling force measurements indicate that cyclic softening occurs when the material is processed multiple times. Pulling force measurements as a function of the number of consolidation-deconsolidation cycles. The number indicates how many times the fibre-filled material has seen a consolidation by pultrusion and consequent deconsolidation by thermal treatment above melting point. Samples shown here were processed at a speed of 105 mm/ min.
Table 2 Deconsolidation rates of composites with different processing history. # of consolidation-deconsolidation cycles Averaged deconsolidation rate [%/s]
Standard deviation [%/s]
1 1 4 4
0.21 0.13 0.09 0.07
without correction factor with correction factor δ = 0.62 without correction factor with correction factor δ = 0.77
2.45 1.52 0.44 0.34
3.5. Modelling of deconsolidation in CLF Based on the observations above, a model was developed in order to predict the deconsolidation behaviour of the carbon fibre-reinforced PA12 material during CLF processing. The model is comprised of a numerical model to simulate the temperature distribution within and after the extrusion module as well as a temperature-dependent empirical model to anticipate the state of deconsolidation. A summary of the experimentally measured material and geometrical properties used in the model are provided in Table 3. Fig. 11 shows the temperature profiles and temperature evolutions of the CLF extrudate after exiting the extrusion nozzle as a function of displacement and corresponding travel time for different line speeds as predicated by the Comsol Multiphysics® model. Note that the temperature profiles of the extrudates vary significantly depending on the line speed due to the time allowed for energy to transfer in and out of the material. For example, at a line speed of 52.5 mm/min, the extrudate has a uniform temperature of 230 °C at the exit of the extrusion nozzle and the core of the extrudate is predicted to cool at a slightly
Fig. 9. Optical dilatometer measurements show a significant reduction in the maximum degree of deconsolidation in materials after multiple processing cycles. Samples shown here were processed at a speed of 105 mm/min.
Table 3 Extrusion nozzle dimensions and implemented material characteristics for conjugated heat transfer model in Comsol Multiphysics®.
Fig. 10. The dynamic deconsolidation behaviour is affected by the processing history of the fibre-filled material. The diagram shows the deconsolidation curves of a one-time and a four-time processed test specimen recorded by an optical dilatometer.
255
Property
Unit
Value
heated length (Lh) extrusion diameter (D) heat conductivity brass heat conductivity air heat conductivity extrudate (transverse) heat conductivity extrudate (longitudinal) heat transfer coefficient air-extrudate specific heat capacity extrudate
mm mm W/(m·K) W/(m·K) W/(m·K) W/(m·K) W/(m·K) J/(kg·K)
3.5 1.40 120 [35] 0.05 [35] Fig. A2 Fig. A3 50 [35] Fig. A1
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 11. The deconsolidation time is one of the major processing parameters that determine void formation in the CLF extrusion process. The deconsolidation time describes the duration in which the extrusion material is heated above melting point. It starts when the material leaves the extrusion nozzle and finishes when the average temperature of the extrudate drops below melting temperature. The process is simulated using Comsol Multiphysics®.
Table 4 Comparison between predicted and measured void contents for the CLF extrudate processed under difference conditions. Processing speed [mm/min]
52.5 105 210 315
Cooling length [mm]
5.68 9.41 11.13 3.67
Deconsolidation time (tc) [s]
6.49 5.38 3.18 0.70
max. average polymer temperature (Ta,max) [°C]
221.39 220.06 205.84 190.47
Predicted void content [%]
Measured void content [%]
1 time processed
4 time processed
1 time processed
4 time processed
6.48 5.34 2.56 0.79
1.89 1.64 1.01 0.62
7.47 4.31 2.63 0.84
1.49 1.17 0.87 0.84
(SD (SD (SD (SD
1.25) 0.86) 0.74) 0.43)
(Tm − Ta, max ) ⎞⎟· d χ (Γ, υ )· Tm χvoid, p = χvoid, i + ε·⎜⎛1 − e f void ⎝ ⎠ dt tc (V , Tn, Lh , D , α, λ, cp)
lower rate than the surface. Conversely, at a line speed of 315 mm/min, only the surface of the extrudate is heated to 230 °C by the time the material exits the extrusion nozzle, with the core of the extrudate never fully reaching the melt temperature despite thermal conduction from the heated surface. Therefore, under these CLF processing conditions, materials processed at higher line speeds will experience lower average temperatures during the extrusion process than those processed at lower line speeds. As a result, the degree to which the CLF extrudate will deconsolidate upon exiting the extrusion is expected to differ depending on the processing conditions. Assuming that deconsolidation occurs whenever the average temperature of the thermoplastic composite is above the melting temperature, the deconsolidation time tc can be defined for each set of processing conditions as the cooling length required for the average temperature to fall below the melt temperature divided by the appropriate line speed. The final void content in the deconsolidated material can then be estimated using Eq. (8).
(SD (SD (SD (SD
0.36) 0.34) 0.13) 0.16)
(8)
where ε is an experimentally determined scaling factor which is determined to be 2.64 by linear regression using the measured data. Note d that the deconsolidation rate dt χvoid is expected to be a function of polymer temperature T , processing history Γ , and fibre volume fraction υf , and tc is assumed to be a function of extrusion speed V , nozzle temperature Tn , the heated length of the nozzle Lh and the extrusion diameter D , as well as the material properties such as the heat transfer coefficient α (air-extrudate), heat conductivity λ and the specific heat d capacity cp . Deconsolidation rates of dt χvoid = 1.52%/ s and d χ dt void
= 0.34%/ s were used for the materials subjected to one and four processing cycles, respectively. The exponential term in Eq. (8) is used d to accommodate the temperature dependency of dt χvoid and considers the melt temperature Tm and the maximum average polymer temperature Ta, max at the extrusion exit. 256
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 12. A four stage pultrusion module was developed and integrated into the CLF system to exploit the advantages of cyclic softening. The heated section of the pultrusion is designed to gradually compact and decompact the fibre-filled melt. (A) Schematic showing the design and design parameters of the multi-stage pultrusion module. (B) Photograph of the implemented multi-stage pultrusion system.
Fig. 13. Simulations were carried out to determine the maximum deconsolidation reached between the individual stages of the four-stage pultrusion module. Simulations were conducted at a constant maximum average polymer temperature Ta, max of 250 °C. (A-C) Calculated cross-sectional areas of the individual pultrusion stages and the simulated cross-sectional areas of the pultruded material at a given position within the pultrusion module. (D) Maximum deconsolidation reached in-between the different pultrusion stages of the pultrusion module for different pultrusion speeds.
The void contents predicted for the CLF extrudate by the model are compared to the final void contents measured by optical microscopy in Table 4. A good correlation between the model and the experimental
results was found. Both studies showed that the materials that were subjected to cyclic softening expressed lower void contents when extruded, with a reduction of more than 80% recorded when the 257
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
Fig. 14. Processing history has a significant effect on deconsolidation during CLF processing. (A) Investigations by optical microscopy revealed that the four-time compacted and deconsolidated material shows a significantly lower degree of void formation (deconsolidation) compared to the material which was only one time processed. (B) Cross-sectional images by optical microscopy illustrate the microstructure inside the manually loaded samples before and after deconsolidation for different processing conditions, voids are shown as dark areas.
point in the die. The distance and the associated time elapsed where the cross-sectional area of the die and the pultrudate match were used to calculate the maximum deconsolidation that occurs in-between the individual pultrusion stages. Fig. 13D summarizes the maximum deconsolidation by v/v % void for the different pultrusion speeds. These results show that the geometry of the dies and the pultrusion speed highly affect the maximum degree of deconsolidation under isothermal conditions. Narrower die geometries and higher pultrusion speeds lead to reduced deconsolidation times. At low pultrusion speeds of 52.5 mm/min, deconsolidation is expected to result in void contents greater than 14%, while at high pultrusion speeds of 315 mm/min, the expected void content is reduced to 3.5%. Experimental investigations showed that the quality of the extruded CLF materials that were cyclically softened using the integrated multistage pultrusion module was comparable to the materials that were manually fed through the single pultrusion module four times. Fig. 14A compares the void content measured from the manual process to those measured from the integrated multi-stage pultrusion setup over a range of line speeds along with the predictions from the simulations. All void contents were measured by optical microscopy analysis, and typical cross-sectional images from the manually loaded samples are provided in Fig. 14B. The integrated four-stage pultrusion module tended to produce materials with slightly higher void contents than manually feeding the material into the single pultrusion module 4 times. This effect may be attributed to the incomplete deconsolidation of the thermoplastic composite as both time and space for expansion are limited in the conical sections. Nonetheless, the integrated multi-stage pultrusion system proves that cyclic softening of high fibre-filled polymeric materials can be used in situ to significantly reduce the deconsolidation behaviour of AM FRPC materials.
commingled yarns were consolidated and deconsolidated 4 times at line speeds of 52.5 mm/min. The model slightly deviates from the experimentally measured data at low line speeds up to 150 mm/min. These minor differences between the model and the experimental results can be explained by simplifications made which ignore the temperature dependence of the deconsolidation rate.
3.6. Exploiting cyclic softening in CLF processing To exploit the advantages of cyclic softening to reduce the degree of deconsolidation found in CLF processing and to increase the mechanical properties of the extruded materials, a multi-stage pultrusion design was developed and integrated into the CLF system. The multi-stage pultrusion module is comprised of four conical dies which gradually force the commingled yarns to its final nominal diameter of 1.42 mm. However, the entrance of each die is designed to be larger than the exit of the preceding die so that the thermoplastic composite has the space to locally deconsolidate and relax before undergoing further compaction. The effect of this cascaded multi-stage pultrusion die is to cyclically soften the thermoplastic composite through repeated loading and unloading. A schematic of the multi-stage pultrusion die and a photograph of the integrated setup are presented in Fig. 12. The model developed for the extrusion process was used to estimate the degree of deconsolidation that occurs in-between the individual pultrusion stages of the multi-stage pultrusion module. The maximum average polymer temperature Ta, max was set to 250 °C and kept constant for all pultrusion speeds. Fig. 13A–C show the cross-sectional areas of the individual pultrusion stages and the corresponding cross-sectional areas simulated for the carbon fibre-reinforced PA12 material at a given
258
Composites Science and Technology 164 (2018) 248–259
M. Eichenhofer et al.
4. Conclusion
1016/j.compscitech.2016.02.021. [10] R. Matsuzaki, M. Ueda, M. Namiki, T.-K. Jeong, H. Asahara, K. Horiguchi, T. Nakamura, A. Todoroki, Y. Hirano, Three-dimensional printing of continuousfiber composites by in-nozzle impregnation, Nat. Publ. Gr (2016), http://dx.doi. org/10.1038/srep23058. [11] M. Eichenhofer, J.C.H. Wong, P. Ermanni, Continuous lattice fabrication of ultralightweight composite structures, Addit. Manuf. 18 (2017) 48–57, http://dx.doi. org/10.1016/j.addma.2017.08.013. [12] I. Gibson, D. Rosen, B. Stucker, Additive Manufacturing Technologies, Springer New York, New York, NY, 2015, http://dx.doi.org/10.1007/978-1-4939-2113-3. [13] L. Ye, M. Lu, Y.-W. Mai, Thermal de-consolidation of thermoplastic matrix composites—I. Growth of voids, Compos. Sci. Technol. 62 (2002) 2121–2130, http://dx.doi.org/10.1016/S0266-3538(02)00144-6. [14] J. Wolfrath, V. Michaud, J.-A.E. Månson, Deconsolidation in glass mat thermoplastics: influence of the initial fibre/matrix configuration, Compos. Sci. Technol. 65 (2005) 1601–1608, http://dx.doi.org/10.1016/j.compscitech.2005.02.001. [15] J. Wolfrath, V. Michaud, J.-A.E. Månson, Deconsolidation in glass mat thermoplastic composites: analysis of the mechanisms, Compos. Part A Appl. Sci. Manuf 36 (2005) 1608–1616, http://dx.doi.org/10.1016/j.compositesa.2005.04.001. [16] M. Brzeski, P. Mitschang, Deconsolidation and its interdependent mechanisms of fibre reinforced polypropylene, Polym. Polym. Compos. 23 (2015) 515–524, http:// dx.doi.org/10.1017/CBO9781107415324.004. [17] T.G. Gutowski, Z. Cai, S. Bauer, D. Boucher, J. Kingery, S. Wineman, Consolidation experiments for laminate composites, J. Compos. Mater. 21 (1987) 650–669, http://dx.doi.org/10.1177/002199838702100705. [18] M. Danzi, F. Klunker, P. Ermanni, Experimental validation of through-thickness resin flow model in the consolidation of saturated porous media, J. Compos. Mater. 51 (2017) 2467–2475, http://dx.doi.org/10.1177/0021998316671574. [19] D.M. Binding, K. Fisher, R.S. Jones, Compression of thermosetting fabric materials, Compos. Part A Appl. Sci. Manuf 31 (2000) 1323–1330 https://doi.org/10.1016/ S1359-835X(00)00032-4 , Accessed date: 10 October 2017. [20] T.G.P. Gutowski, Advanced Composites Manufacturing, Wiley, 1997. [21] A.A. Somashekar, S. Bickerton, D. Bhattacharyya, Exploring the non-elastic compression deformation of dry glass fibre reinforcements, Compos. Sci. Technol. 67 (2007) 183–200, http://dx.doi.org/10.1016/j.compscitech.2006.07.032. [22] J.J. Cheng, P.A. Kelly, S. Bickerton, A rate-independent thermomechanical constitutive model for fiber reinforcements, J. Compos. Mater. 46 (2011) 247–256, http://dx.doi.org/10.1177/0021998311410506. [23] M. Lu, L. Ye, Y.-W. Mai, Thermal de-consolidation of thermoplastic matrix composites—II, “Migration” of voids and “re-consolidation”, Compos. Sci. Technol. 64 (2004) 191–202, http://dx.doi.org/10.1016/S0266-3538(03)00233-1. [24] H. Huang, R. Talreja, Effects of void geometry on elastic properties of unidirectional fiber reinforced composites, Compos. Sci. Technol. 65 (2005) 1964–1981, http:// dx.doi.org/10.1016/j.compscitech.2005.02.019. [25] M. Hou, L. Ye, Y.W. Mai, Materials processing technology manufacturing process and mechanical properties of thermoplastic composite components, J. Mater. Process. Teclmol. 63 (1997) 334–338, http://dx.doi.org/10.1016/S0924-0136(96) 02644-1. [26] P.-O. Hagstrand, F. Bonjour, J.-A.E. Månson, The influence of void content on the structural flexural performance of unidirectional glass fibre reinforced polypropylene composites, Compos. Part A Appl. Sci. Manuf 36 (2005) 705–714, http:// dx.doi.org/10.1016/j.compositesa.2004.03.007. [27] C. Schneeberger, J.C.H. Wong, P. Ermanni, Hybrid bicomponent fibres for thermoplastic composite preforms, Compos. Part A (2017), http://dx.doi.org/10.1016/ j.compositesa.2017.09.008. [28] Z. Jiang, J.M. Hutchinson, C.T. Imrie, Temperature-modulated differential scanning calorimetry. Part II. Determination of activation energies, Polym. Int. 47 (1998) 72–75, http://dx.doi.org/10.1002/(SICI)1097-0126(199809)47:1<72::AIDPI999>3.0.CO;2-N. [29] R.B. Pipes, A.J. Beaussart, J.T. Tzeng, R.K. Okine, Anisotropic viscosities of oriented discontinuous fiber laminates, J. Compos. Mater. 26 (1992) 1088–1099, http://dx. doi.org/10.1177/002199839202600801. [30] D. Zhang, D. Heider, J.W. Gillespie, Void reduction of high-performance thermoplastic composites via oven vacuum bag processing, J. Compos. Mater. 0 (2017) 1–12, http://dx.doi.org/10.1177/0021998317700700. [31] L. Ye, Z.-R. Chen, M. Lu, M. Hou, De-consolidation and re-consolidation in CF/PPS thermoplastic matrix composites, Compos. Part A Appl. Sci. Manuf 36 (2005) 915–922, http://dx.doi.org/10.1016/j.compositesa.2004.12.006. [32] K. Henninger, F. Ye, Lin Friedrich, Deconsolidation behaviour of glass fibre-polyamide 12 composite sheet material during post-processing, Plast. Rubber Compos. Process. Appl. 27(1998) 287–292. HYPERLINK "https://scholar.google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5" \o "https://scholar. google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5" https:// scholar.google.ch/scholar?cluster=4828173141021934845&hl=en&as_sdt=0,5 (accessed November 24, 2017). [33] K. Wudy, D. Drummer, M. Drexler, Characterization of polymer materials and powders for selective laser melting, AIP Conf. Proc. 1593 (2014) 702–707, http:// dx.doi.org/10.1063/1.4873875. [34] D.-H. Kim, W. Il Lee, K. Friedrich, A model for a thermoplastic pultrusion process using commingled yarns, Compos. Sci. Technol. 61 (2001) 1065–1077 https://doi. org/10.1016/S0266-3538(00)00234-7 , Accessed date: 17 October 2017. [35] K. Langeheinecke, P. Jany, G. Thieleke, K. Langeheinecke, A. Kaufmann, Wärmeübertragung, Thermodyn. Für Ingenieure, Springer Fachmedien Wiesbaden, Wiesbaden, 2013, pp. 242–289, , http://dx.doi.org/10.1007/978-3-658-031695_10.
The thermal deconsolidation of unidirectional fibre-reinforced thermoplastic composites observed in CLF processing was found to be significantly reduced through cyclic softening. Thermoplastic composites that had undergone cyclic softening were observed to have both lower degrees and rates of deconsolidation compared to materials that were not repeatedly consolidated (compacted) and deconsolidated (decompacted). A numerical-empirical model was developed which was able to accurately predict the deconsolidation observed during the CLF process based on the processing history, heat distribution, and thermal evolution of the extrudates. Based on these fundamental studies into the deconsolidation behaviour of the thermoplastic composites, a multistage pultrusion module was designed and implemented into CLF that allowed the final void content of the extrudate to be reduced by more than 80%. This work represents significant advancements in the ability of CLF to print fibre-reinforced composites with mechanical properties suitable for use in high performance end-use components by 1) enlarging the processing window in which material with acceptably low void contents (< 1%) can be deposited and 2) providing a simulation tool that can be used to select the appropriate processing conditions for thermoplastic composites in extrusion-based AM processes. Acknowledgments This work was supported by ETH Zurich's internal research funding program (ETH Research Grant ETH-33 15-2) and the Commission for Technology and Innovation (CTI) through the Swiss Competence Center for Energy Research (SCCER), Efficient Technologies and Systems for Mobility. Commingled yarns were kindly provided by Schappe Technologies. The authors thank Dr. Samuel Brunner of the Swiss Federal Laboratories for Materials Science and Technology, and Prof. André Studart and Dr. Kunal Masania of the Complex Materials group at ETH Zurich for their support received throughout this study. Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx. doi.org/10.1016/j.compscitech.2018.05.033. References [1] A. Gebhardt, Understanding Additive Manufacturing, Carl Hanser Verlag GmbH & Co, KG, München, 2011, http://dx.doi.org/10.3139/9783446431621. [2] S.H. Huang, P. Liu, A. Mokasdar, L. Hou, Additive manufacturing and its societal impact: a literature review, Int. J. Adv. Manuf. Technol. 67 (2013) 1191–1203, http://dx.doi.org/10.1007/s00170-012-4558-5. [3] X. Wang, M. Jiang, Z. Zhou, J. Gou, D. Hui, 3D printing of polymer matrix composites: a review and prospective, Compos. Part B 110 (2017) 442–458, http://dx. doi.org/10.1016/j.compositesb.2016.11.034. [4] B.G. Compton, J.A. Lewis, 3D-Printing of lightweight cellular composites, Adv. Mater. 26 (2014) 5930–5935, http://dx.doi.org/10.1002/adma.201401804. [5] J.P. Lewicki, J.N. Rodriguez, C. Zhu, M.A. Worsley, A.S. Wu, Y. Kanarska, J.D. Horn, E.B. Duoss, J.M. Ortega, W. Elmer, R. Hensleigh, R.A. Fellini, M.J. King, 3D-Printing of meso-structurally ordered carbon fiber/polymer composites with unprecedented orthotropic physical properties, Nat. Publ. Gr (2017), http://dx.doi. org/10.1038/srep43401. [6] P. Parandoush, D. Lin, A review on additive manufacturing of polymer-fiber composites, Compos. Struct. 182 (2017) 36–53, http://dx.doi.org/10.1016/j. compstruct.2017.08.088. [7] W. Zhang, A.S. Wu, J. Sun, Z. Quan, B. Gu, B. Sun, C. Cotton, D. Heider, T.-W. Chou, Characterization of residual stress and deformation in additively manufactured ABS polymer and composite specimens, Compos. Sci. Technol. 150 (2017) 102–110, http://dx.doi.org/10.1016/j.compscitech.2017.07.017. [8] A.N. Dickson, J.N. Barry, K.A. Mcdonnell, D.P. Dowling, Fabrication of continuous carbon, glass and Kevlar fibre reinforced polymer composites using additive manufacturing, Addit. Manuf. 16 (2017) 146–152, http://dx.doi.org/10.1016/j.addma. 2017.06.004. [9] Z. Quan, Z. Larimore, A. Wu, J. Yu, X. Qin, M. Mirotznik, J. Suhr, J.-H. Byun, Y. Oh, T.-W. Chou, Microstructural design and additive manufacturing and characterization of 3D orthogonal short carbon fiber/acrylonitrile-butadiene-styrene preform and composite, Compos. Sci. Technol. 126 (2016) 139–148, http://dx.doi.org/10.
259