Content uploaded by Dongmin Yang
Author content
All content in this area was uploaded by Dongmin Yang on Dec 08, 2020
Content may be subject to copyright.
Fibre misalignment and breakage in 3D printing of continuous
carbon fibre reinforced thermoplastic composites
Haoqi Zhang1, 2, Jiayun Chen1, Dongmin Yang1, 2, *
1. Institute for Materials and Processes, School of Engineering, University of Edinburgh,
Edinburgh, EH9 3FB, UK
2. School of Civil Engineering, University of Leeds, Leeds, LS2 9JT, UK
Abstract
This paper investigates the formation of manufacturing induced fibre misalignment and
breakage during fused filament fabrication (FFF) 3D printing of 1K continuous carbon
fibre filament. Single stripes at various turning angles and curvatures are printed by a
desktop printer Prusa i3 using a specific brass nozzle and characterised using X-ray
computed micro-tomography (µCT) and optical microscopy. A finite element (FE)
model of the printing process is also established to support the experimental
measurement. It has been found that high porosity and fibre misalignment in the
printed straight stripe result from the weak fibre/matrix interface and the uneven
pressure executed by the nozzle. Increase of turning angle and/or reducing of
curvature radius leads to more aggravated printing defects, including shape
inaccuracy, fibre twisting, folding and misalignment, due to the excessive force from
the nozzle, debonding with the print bed and the unmatched geometry of nozzle outlet
and fibre filament. Severe fibre breakage and significant change of thickness can be
seen in the samples with turning angles larger than 120° or curvature radius smaller
than 5 mm, while the wrinkles of the stripe in the inner periphery appear more
frequently as the curvature radius decreases.
Keyword: 3D printing; continuous carbon fibres; X-ray computed micro-
tomography; fibre breakage; fibre misalignment; Finite element analysis
* Corresponding author. Email: Dongmin.Yang@ed.ac.uk
1. Introduction
Continuous carbon fibre-reinforced polymer (CCFRP) composites have been
increasingly used in airframes and other high-end industrial products [1, 2] due to their
superior mechanical properties such as high strength-weight and stiffness-weight
ratios [3]. Traditional manufacturing methods of CCFRP, such as autoclave and resin
transfer moulding, have difficulties to fabricate composites with complex geometry [4],
since continuous fibres in the fabric preforms are usually aligned in specific directions.
Notches and holes normally need to be created by subtractive machining processes
such as drilling and cutting when mechanical fastening is needed. These machining
processes could induce residual damages including delamination and fibre breakage
which may compromise the structural integrity of the composites [5, 6].
To mitigate the defects caused by subtractive machining processes, alternative
manufacturing methods have been explored to place continuous fibre tows and
fabricate composites with complex geometries. Among them, Automated Fibre
Placement (AFP) uses small fibre tows (typically 8 mm wide or less) to form
composites, which leads to better precision and increased deposition rates when
compared with experienced laminators for hand lay-up. For composite parts with
complex geometries, fibre tows can be steered along the desired fibre paths. However,
the diameter of the compaction roller and the head geometry limits female mould
radius that can be used for parts built with this technology [7]. Also, several defects
may appear during tow steering, which includes out-of-plane wrinkling, blistering, tow
pull-ups and shearing effects. A continuous tow shearing (CTS) technique has been
proposed to shear dry tows to mitigate these defects with success [8], but it has only
been applied to composite plates with relatively large curvatures.
Another recently developed technology, additive manufacturing (also known as 3D
printing) has shown potential to fabricate continuous fibre-reinforced composites with
highly complex shapes in both 2D and 3D. Among them, Fused Filament Fabrication
(FFF) 3D printing melts and deposits small thermoplastic filament (usually with 1k fibre
tows) layer by layer to create the composite parts [3]. In 2014, Markforged® released
a series of 3D printers, e.g., Mark One/Two, which can manufacture composites with
continuous fibre reinforcement [9]. Supported by its dedicated software Eiger, Mark
Two printer places the fibres around the geometry singularity using a concentric
approach. Other researchers also developed in-house printers for CCFRP composites
[10-12], which were generally achieved by impregnating the fibres with a matrix
material before extruding or within the nozzle [13]. Compared with AFP, 3D printing of
CCFRP offers better surface finishing and more design freedom but less efficient [14],
due to a higher resolution of the printing system and a smaller width of the printed fibre
tows (1-2 mm) [15].
The mechanical performance of 3D printed composites with continuous fibre
reinforcements has been evaluated by researchers [16-18]. Some of them developed
numerical simulations to predict the progressive damage of 3D-printed curvilinear
CCFRP composites [19]. It has been revealed that the productivity of the 3D printing
process and the quality of final parts depend on a large number of factors, including
the geometric complexity of the part as well as the process parameters such as printing
speed and fibre orientation [10, 20]. Despite the advantages and capabilities, 3D
printing also comes with its own limitations. For instance, the significant voids content
results in much lower stiffness and strength than expected [21-23], compared with
traditionally manufactured FRP composites. Also, the fibre bundle could be twisted,
folded or even broken when printing a curved section [24]. The printed radius turned
out to be lower than the designed value with a larger fibre bundle size or a smaller set
radius [25]. In previous studies, the effect of these defects on mechanical performance
has yet been investigated and there is still a lack of established relationship between
printing parameters and the formation of those defects. This is largely because the
fundamental mechanisms for the FFF 3D printing of CCFRP composites have not
been fully understood. Therefore, a comprehensive study is needed for the 3D printing
process of CCFRP, especially for the curvilinear paths. In addition, a more detailed
simulation of the printing process will be helpful for the further understanding of the
mechanisms.
In this paper, the printing induced defects including fibre misalignment (wrinkling,
twisting and folding) and breakage are investigated during the FFF 3D printing of
CCFRP composites using 1K carbon fibre filament and a specific brass nozzle. First,
void formation and fibre misalignment in the printed straight stripe are studied. And
then a set of single stripes are printed out at various turning angles and curvatures to
represent the complex geometries of notches and holes, which are characterised
using X-ray micro-tomography and optical microscopy. A finite element (FE) model of
the printing process is also established to aid the understanding of the mechanisms of
continuous fibre printing in terms of local stress distributions.
2. 3D printing of CCFRP
2.1 Printing filaments and nozzle
The printing material 0.375 mm carbon fibre (CF) filament is sourced from Markforged
company. The previous study had evidenced that polyamide 6-I (PA6-I) is the polymer
matrix for CF filaments and the polymer coating is a polyamide 6 (PA6) [26]. A µCT
scan of its internal structure is shown in Figure 1(a) & 1(b). It can be seen that the
cross-section of the filaments is relatively irregular circular with a diameter from 0.35
mm to 0.40 mm and carbon fibres are arranged well straight along the longitudinal
direction of the filament. The volume fraction of carbon fibres 𝑉
𝑓 is measured as 21.34%
(𝑉
𝑓= 𝑣𝑓/𝑣𝑐, 𝑣𝑓 is the volume of fibres and 𝑣𝑐 is the volume of the composite). However,
as shown in the cross-section, the fibres are not evenly distributed in the filament,
instead, they are concentrated into 3 parallel zones. This leads to two regions mainly
composed of PA6-I in the filament, in which noticeable air bubbles are observed. Some
small voids are trapped around the carbon fibres, indicating a weak interface adhesion
between the fibre and matrix. The average porosity of the CF filament is measured as
0.7%. All these pores formed in the fabrication process of CF filament may affect the
mechanical properties of 3D printed finished parts.
Figure 1. µCT scans of filament and nozzle: (a) cross section and (b) 3D view of
continuous carbon fibre filament; (c) brass nozzle
The brass nozzle used for the CF filament printing is also sourced from Markforged.
Dimensions of the nozzle are measured from its µCT scan, which is illustrated in
Figure 1(c). The nozzle has a standard M6 screw thread and an outlet diameter of 1.3
mm. There is a thermostable plastic (Polytetrafluoroethylene, PTFE) tube inside the
nozzle but for the benefit of measuring the dimensions of the nozzle, it is not shown in
the CT image due to its low density compared with the brass. The PTFE plastic tube
is used as a guide of CF filaments while maintaining a more uniform temperature
distribution for the CF filament and avoiding sticking and friction with the inner wall of
the brass nozzle. The large difference between the diameters of the nozzle outlet and
CF filament (1.3 mm and 0.375 mm, respectively) is one of the main reasons causing
the inaccuracy of fibre paths. Unlike other nozzles for the printing of pure thermoplastic
(such as ABS and PLA) as well as short fibre-reinforced thermoplastic, it has rounded
corners at the tip which presses the heated continuous fibre filament on the printer
bed. The continuous carbon fibres are passively pulled out from the nozzle as it moves.
This is the unique mechanism of the continuous fibre printing process and will be
represented in later numerical modelling. Moreover, the friction between the nozzle
and melted CF filament could wear the tip of nozzles, which consequently causes
inaccuracy of the fibre paths in the printing process.
2.2 Desktop 3D printing
The Prusa i3 MK3s printer is used in this study and a schematic diagram of 3D printing
CCFRP is shown in Figure 2. The filament is heated to a temperature of 245°C and
the off-distance between nozzle tip and printer bed is set to 0.1 mm. The speed of
nozzle movement is equivalent to a feed rate of 5 mm/s for the filament, which means
the free end of the CF filament is slightly under tension due to the twine of the spool,
but the feeding of filament would not be affected. Samples are printed onto an
unheated Garolite printer plate which was coated with a layer of PVA, to ensure
adequate adhesion at the start of the printing.
Figure 2. (a) Schematic diagram of 3D printing of continuous carbon fibre filament
and (b) modified print head with the nozzle
A straight stripe of continuous CF is printed first, as a base unit of the finished part.
The internal microstructures are characterised by X-ray µCT scans to explore the
printing quality of continuous CF filaments. Then a set of single stripes of the
composites are printed with different turning angles ranging from 30° to 180° at an
increment of 30°, as well as various curvature radius ranging from 2.5mm to 20mm.
The study of turning angles reveals the manufacturing of composites with complex
geometry and singularity, like sharp corners and notches, while curvature radius more
represents circular holes in the composite structures. Since the filament used in this
study contains continuous fibres, and the toolpaths cannot be the same as those in
traditional thermoplastic printing. Therefore continuous toolpaths (G-code) are
generated through a MATLAB script and then transferred to the printer. The case
studies with different turning angles and curvature radius aim to investigate the
mechanisms of fibre wrinkling and breakage during the 3D printing process.
3. X-ray µCT scans and optical microscopy
X-ray µCT scans of original filament and printed stripes with various turning angles are
carried out on a Zeiss Xradia Versa 410 µCT system. For other samples printed at
various curvature radius, optical microscope characterisation was carried out on a
Zeiss Stemi 2000 Stereo Zoom Microscope. The reason for using optical microscopy
instead of X-ray µCT for those samples is that optical microscopy gives a much larger
view of the printed samples than X-ray µCT so as to provide more useful information
of the fibre wrinkling along the whole printed curve.
The X-ray source with an accelerating voltage of 80 kV and a power of 7 W is used for
all scans. The exposure time and effective pixel size vary for each sample to achieve
the possible best intensity and contrast of the images, as listed in Table 1. Each
sample is fully rotated by 360° during the scan, resulting in thousands of projections
collected on a 1k x 1k pixel, noise suppressed charge-coupled detector. The raw data
are reconstructed using a Zeiss built-in reconstruction software to obtain clear images
and the threshold value is determined by analysing experience.
Table 1. Exposure time and effective pixel size for the samples in X-ray µCT scans
Raw
filament
Straight
single stripe
Single stripes with different turning angles
30°
60°
90°
120°
150°
180°
Exposure
time (s)
5
2
2
2
8
2
2
2
Effective
pixel size
(µm)
1.023
2.638
3.293
3.293
3.123
2.826
2.384
3.293
For this paper, the voids percentage and fibre volume fraction are quantified following
the image processing steps in Avizo software (commercial software for data
visualisation and analysis), as shown in Figure 3. Because the images obtained from
µCT scanning are usually not good enough for later processing such as edge detection,
filtering technique is necessary to remove the noise [27]. The median filtering used in
this paper is considered to be a technique for linear smooth processing, often used to
preserve edges while removing noise, in which the main idea is to run through the
signal entry by entry, replacing each entry with the median of neighbouring entries [28].
As shown in Figure 3, compared with raw images, the noise in the matrix material is
reduced effectively and the boundaries between different materials are easier to detect
in images after median filtering.
Figure 3. Steps of images processing and quantitative analysis
Then the segmentation method is used to separate void, fibres and matrix. The grey
values are sorted from low to high, and the different materials can be segmented by
changing the intensity range. In this study, the lowest grey value threshold is the void,
the middle is the polyamide 6 and the highest is the fibres (in colour black, grey and
purple, respectively), as shown in Figure 3. The application Label Analysis is used to
label all the constituents detected from segmentation method and volume of them can
be calculated, therefore volume fraction of different materials can be obtained.
4. Finite element modelling
A finite element (FE) model is established in ABAQUS/Explicit to further investigate
the influence of turning angle and curvature radius on the printing quality of a single
printed stripe. Also, the parametric studies about filament/tape width and fibre volume
fraction are carried out to inform the design of 3D printing filament.
In real printing experiments, the original filament with a cylindrical shape (as shown in
Figure 1a) is heated up in the heating chamber and passively pulled out as a result of
the movement of the printer head. The printer nozzle exerts a contact pressure on the
heated filament, changing the shape of the filament from cylindrical to roughly
rectangular. As shown in Figure 4 (a), this process is reasonably simplified in the FE
model by using a tape with a rectangular cross-section passing through a guide in the
printer nozzle. The geometry and dimensions of the printer nozzle are informed from
the measurement in µCT image as in Figure 1 (c), while the printing tape is assumed
to have a width of 1 mm and a thickness of 0.1 mm according to the µCT image in
Figure 5. The size of approximate global seeds is 0.05 mm so 20 elements per
millimetre mesh in the longitudinal direction. Shell thickness is set to 0.1 mm in the
thickness direction. In our modelling, the printer nozzle and the guide are both
considered as rigid bodies. A total of 4000 S4R shell elements are adopted for the
printing tape, with predefined cohesive properties for the printer bed to represent the
bonding behaviour.
The printing tape is considered as transversely isotropic and assumed elastic, wherein
the modulus of elasticity in fibre direction (E1) is calculated and determined in light of
weighted average modulus of T300 carbon fibre and polyamide 6. A more
sophisticated elasto-plastic constitutive law for thermoplastic composites could be
considered in the FE model but the challenges to determine the model parameters at
elevated temperature make the model validation extremely difficult. Therefore,
qualitative elastic analysis is mainly focused on the modelling results at this stage. The
modulus in transverse to fibre direction (E2) is four orders of magnitude smaller than
E1, allowing relatively free deformations of fibre tape in the transverse direction,
corresponding to the real printing process at a high temperature. Also, the default set
of contact is used for the interaction between the tape, the nozzle, print bed and the
guide, which is ‘friction factor = 0.01’ and ‘hard contact’ (only allow the transfer of
compressive stress) for tangential and normal behaviour. In order to simulate the
bonding condition between printing tape and printer bed, a cohesive contact is
predefined between the contact surfaces. Previous research of AFP manufacturing
process [4] is used as a source of reference and the material properties of the printing
tape and interfacial parameters under 245°C temperature are listed in Tables 2 & 3.
Since the time-temperature superposition along with viscoelastic material properties
is complicated and hard to decide, only properties measured at appropriate rates and
temperatures for both fibre tape and interface are considered. Among them, interface
parameters are increased by two orders of magnitudes considering the higher
viscosity of thermoplastic PA-6 (approximately 120 Pa·s at 240°C [29]).
The simulation process is divided into 3 steps (Figure 4b & 4c) and the printer bed is
always fixed in all steps. In step 1, the rigid nozzle moves downwards and presses the
filament onto the printer bed. Only two translational degrees of freedoms (DOFs) are
released for the nozzle, and an upward force (1 N) acts at end of filament above the
nozzle to straighten the filament in the PTFE guide. After that, the gap between nozzle
and printing bed reduces to 0.1 mm, and it is maintained for the rest of the simulation.
In step 2, the nozzle moves 2 mm horizontally along the longitudinal direction of the
composite printed on the bed. In step 3, the nozzle moves in another direction
translationally with a specific turning angle or a curvature from those selected in the
printing experiments in Section 2.
The FE model with those selected parameters is first qualitatively validated against
the experimental measurements in terms of tape deformation and stress concentration,
as will be discussed in Section 5. Furthermore, the datasheet provided by TORAY
company shows that the tensile strength of single T300 carbon fibre is 3530 MPa and
strain at failure is 1.5%. According to typical elastic assumptions, the axial strain 𝜀 can
be expressed as |𝜀|=𝑑2𝜌
⁄ when a fibre with a diameter of 𝑑 (7 µm) is bent to a
curvature radius of 𝜌 [30]. Supposing that fibres in the fused filament bend
independently, the minimum curvature radius of T300 carbon fibre can be calculated
as 233 µm. These data will also be compared with the modelling result, which acts as
the supplement of the elastic model.
Figure 4. FE model of the printing process of CCFRP: (a) The assembled model, and
(b) & (c) the simulation processes
Table 2. Mechanical properties of printed carbon fibre filament [4, 31]
Density
E1
E2
v12
G12 & G23 & G13
1.4 g/cm3
31 GPa
4.6 MPa
0.2
30.25 MPa
Table 3. Interface parameters [32]
Cohesive stiffness Knn, Ktt, Kss
Maximum Stress
Fracture Energy
329.23 N/mm3
38.52 MPa
0.868 N/mm
5. Results and discussion
5.1 Experimental observations
The shape of the printed straight stripe is like a flat ribbon because the filament is
pressed on the bed by the tip of the nozzle during the printing process. The shape is
also affected by the printing speed and the off-distance between the nozzle and the
bed. With the printing parameters used in our study, the printed straight stripe is about
1.3 mm in width and 0.10 mm in thickness. The porosity at the central region of the
stripe is measured as 1.39% in Avizo and the fibre volume fraction is approximately
20% (since the edge area of the printed stripe is too irregular to be measured).
As shown in Figure 5 (b), voids mainly appear between the fibres, which indicates the
low adhesion between fibres and matrix in the printed CCFRP composites. And more
pores are found at the edges of the stripe in Figure 5 (b) & (c), which are likely due to
the uneven pressure acting on the filament by the tip of the nozzle. Large gaps
occasionally appear on the bottom surface of printed stripe, indicating a weak
interlaminar strength, which could be caused by the cooling within a very short time
[33], too fast nozzle speed and/or insufficient compaction from the nozzle. The results
reveal a lot of voids of the 3D printed CCFRP composites and indicate the expected
mechanical performance is hard to be achieved by the present 3D printing technique,
compared with that obtained from traditional manufacturing with porosity <1% at a fibre
volume fraction >50% [34, 35]. For instance, in hot compression moulding, the creation
of the interfaces and the diffusion of the material is ensured since the material is kept
under high-pressure levels at temperatures above 𝑇
𝑔 for a long period of time (280 °C
for 3.5 min under a compression pressure of 4 kg/cm2 for CF/polyamide 6 pre-
impregnated sheets [35]). In addition to high porosity, misalignment of continuous
fibres is also a serious issue, even when printing along a straight path, as shown in
Figure 5 (a). All these manufacturing induced defects would dramatically reduce the
stiffness (and strength) of the finished part, and in particular, the fibre misalignment
would easily trigger the buckling failure of composites under compression loading [36,
37].
Figure 5. µCT images of a printed straight stripe: (a) overview and (b) & (c) cross
sections
The µCT images of a single stripe at different turning angle (30°, 60° and 90°) are
shown in Figure 6, wherein the dashed lines show some of the representative fibres
in the bundle of each case and the black points indicate the location where the nozzle
was turned. With a turning angle 30°, the filament is flattened on the bed as a tape, in
which the actual turning angles of most continuous fibres are consistent but slightly
smaller than 30°. The inaccuracy of the printing path is likely caused by the unmatched
diameters between the nozzle and the filament (1.3 mm and 0.375 mm, respectively),
which also makes a curved section rather than a turning point with a specific angle in
this case. It can be seen that no folded section is formed when printing at such a small
turning angle, but different deformation between the continuous fibres at outer and
inner edges of the bundle is observed, i.e., fibres in the inner periphery are slightly
twisted and wrinkled while fibres in the outer periphery are stretched in tension
because of the force received from the print nozzle.
As the turning angle increases, the inaccuracy of the printing path becomes more
apparent. As shown in Figure 6 (b) & (c) for the turning angles of 60° and 90°,
paramount and complex deformation of fibre bundles is observed. Folding of fibre
bundles starts to occur when the turning angle increases to 60°, in which some
continuous fibres at the outer periphery are misaligned and then flipped over to the
inner periphery. The folding scenario is assumed to be caused by the twisting torque
due to the adhesive force from the printer bed and tensile force from the printer nozzle
[24]. However, no noticeable fibre breakage is found in the printed stripes with a
turning angle up to 90°, in which the folding of curved fibres may play a protective role.
Figure 6. 3D printed carbon fibre filament a turning angle of (a) 30° (b) 60° and (c)
90°
As the turning angle further increases to 120°, noticeable folding can be seen at the
turning point of the printed filament, which is accompanied with a complete switch of
the inner and outer circumferences before and after the curved section. Also, the
average width of the printed filament is reduced due to the folding at the turning point,
as shown in Figure 7 (a). This could enlarge the gaps between the printing stripes and
cause large fibreless areas, thus leading to a local weakness in the parts with complex
geometries [38]. At the turning angles of 150° and 180°, the fibres are severely twisted
and misaligned, and more importantly, the fibre breakages are observed. As shown in
Figure 7 (b), a matrix rich area is created and the excessive overlap of continuous
fibres causes the upheaval of printed filament at the turning point. The printing process
could be terminated due to the fibre breakage and upheaval of printed filament. These
results also indicate that even if the filament can be printed out at such turning angles
the mechanical properties would have much deteriorated at the curved section.
In summary, the µCT scans show the error of the printing path increases as the turning
angle alters from 30° to 120°. Also, the printing induced defects such as fibre wrinkling,
twisting and folding become more severe. However, in the printing cases with a turning
angle no more than 120°, no noticeable fibre breakages are seen for the specific
printing parameters and filaments used therein, which may benefit from the folding of
the fibres at the curved section. In the samples with turning angles of 150° and 180°,
fibre breakage and significant change of thickness can be seen at the turning section,
which means such large turning angles should be avoided in the path design of
continuous fibre printing.
Figure 7. 3D printed carbon fibre filament a turning angle of (a) 120° (b) 150° and (c)
180°
For the printed stripes with various curvatures, optical microscopy was used to give a
larger view and more information about manufacturing induced defects. Same printing
parameters are used as in the cases with turning angles. The results with curvature
radii of 2.5, 5, 10 and 20 mm are shown in Figure 8. For the single stripe with a
curvature radius of 20 mm, no obvious surface defects can be observed and the width
of the single stripe is roughly consistent along the printing direction. In the case with
10 mm, fibres in the inner periphery are twisted and wrinkled. When it comes to 5 mm,
these two defects appear more frequently. For the case with a curvature radius of 2.5
mm, the single stripe can hardly be printed properly on the bed as designed paths and
the folding phenomenon with a complete switch of the inner and outer circumferences
is observed. Also, a small number of fibres are broken in the cases with a curvature
radius of 5 and 2.5 mm, although the printing curvatures are smaller than the minimum
curvature (233 µm) obtained from the calculation with typical elastic assumptions.
Therefore, the fibre breakage is more likely to be caused by the shear stress in the
printing process, which will be further investigated in the modelling section.
Figure 8. Optical microscopy images of the printed composites with various
curvature radius (length of the scale bar in figure = 1000 µm)
5.2 FE modelling
The mechanism of 3D printing continuous fibre is further investigated by FE modelling.
As shown in Figure 9, the comparison between the printed straight stripe and the
printed stripe at a turning angle of 30° is presented. The distribution of three stress
components is discussed. The stress in the fibre direction S11 is the component with
the maximum value, which is also used to interpret the deformation of continuous
fibres in the printing process. The shear stress S12 indicates the possibility of fibre
breakage (although no fibre breakage can be observed in this simulation because only
elastic properties of fibre tape are considered). As the tape has much higher strengths
in the fibre direction, fibre breakages are more likely caused by the shear stress in our
studied process. The distribution of pressure reflects the force received from the
nozzle and print bed, which determines the mechanical response of fibre tape in the
printing process.
To better view stress distributions in the fibre tape, the rigid nozzle is hidden from
visualisation. As shown in Figure 9, the two stress components (S11 and S12) are not
evenly distributed in the cross-section transverse to the printing direction, which is
likely the reason for the fibre misalignment as well as the gaps between the tape and
printer bed (as shown in Figure 5). The contact pressure on the fibre tape is studied
further, including compressive force received from the nozzle and cohesive force
received from the print bed (red circle 1 and 2, respectively). As can be seen, the front
edge (black dashed line) of the cohesive area is not perpendicular to the printing
direction, due to the circular outlet of the nozzle which also results in the printing
defects such as void rich area at the edges of the stripe (as shown in Fig. 7 from CT
images). When the fibre tape is printed at an angle, the concentration of S11 and S12
is aggravated in the curved section, indicating the more severe deformation and
potential fibre breakage. Meanwhile, the absolute value of cohesive force received
from the printer bed increases dramatically and concentrates on the outer edge of the
curved section, which indicates the trend of debonding in this area with a high absolute
value of the cohesive force. But the compressive pressure received from the printer
nozzle increases slightly in the case with a turning angle. Since we focus more on the
bonding between the bed and tapes rather than the compressive force from the nozzle
(fibres would not be damaged by pressure in our printing), the pressure shown in the
following discussion will be the absolute value of the cohesive force received from the
print bed.
Figure 9. Mechanism analysis of 3D printing continuous fibre tape straight and
angularly (30°): (a) S11, (b) S12, (c) contact pressure and (d) bottom view.
Figure 10 presents FE modelling results for the distribution of S11 in printed
continuous fibres tape at various turning angles. From 15º to 75º, stress concentration
exists at the curved section and the maximum value gradually increases as the turning
angle becomes larger. With small turning angles such as 15º and 30º, the tape can be
printed entirely on the bed with a more-or-less constant width and the location of
maximum stress is close to the outer edge of the curved section, which is consistent
with the µCT scanning result in Figure 6 (a). With a relatively larger turning angles 45º-
75º, the location of maximum stress gradually shifts to the inner edge of the curved
section, while the outer edge of the tape lifts and no longer sticks on the bed (the black
circle in Figure 10). Also, the values of three stress components have a sharp rise
from the turning angle 45º. This is in good agreement µCT scans in Figure 6 (b),
showing fibre twisting and folding. It is revealed that the folding occurs because of the
debonding between the tape and the printer bed and the excessive tensile force from
the nozzle. When turning angle increases to 90º, the width of printed tape in the
simulation is enlarged because of the excessive tensile force, but the ’swap’ of the
inner and outer edges observed in experiments cannot be simulated. Since the
filament is simplified as a tape with a rectangular cross-section in our FE model and
shell elements are used to simulate the tape, only the initiation of folding phenomenon
can be captured. Similar results occur in the modelling cases with a turning angle
larger than 90º, thus these results are not shown in this discussion section. The S11
(macro) stress obtained from the simulation can be compared with the recalculated
fibre tensile strength based on the fibre volume fraction (3530 * 20% = 706 MPa), in
order to evaluate the possibility of fibre breakage. The maximum value of S11 rises
sharply and exceeds the recalculated strength in the cases with turning angle 60º and
75º, and it decreases to 552 MPa in the case with 90º truing angle due to the tape
model used in the simulation, which all indicates a strong possibility of fibre breakage.
Compared with the experimental result from µCT-scan, it also reveals that the fibre
misalignment in the fused filament and the unmatched diameter do protect continuous
fibres from breakage, although they cause some defects such as the inaccuracy of the
paths.
Figure 10. Stress distributions in printed tapes at various turning angles (from 15° to
90°): (a) Distribution of S11 (b) Maximum values of three stress components
Figure 11 presents the distribution of shear stress S12 in printed tapes at various
curvatures. With the decrease of curvature radius from 20 mm to 2.5 mm, the area
with a high value of shear stress extends gradually. As mentioned before, the increase
of maximum values of shear stress indicates a greater possibility of fibre breakage. It
is also consistent with the result from optical microscopy in Figure 8, wherein the fibre
breakage was captured in the cases with curvature radius 5 mm and 2.5 mm. Since
the curvature radiuses used in the modelling cases are larger than the minimum
curvature radius calculated by the typical elastic assumptions, the fibre breakages are
more likely to be caused by the shear stress rather than the axial stress in the fibre
direction. When the curvature radius comes to 10 mm, 5 mm and 2.5 mm, the wrinkling
of the tape appears in the inner periphery (black circles in Figure 11), also with a higher
frequency as the curvature radius decreases, which is in good agreement with the
experimental result in Figure 8.
Figure 11. Stress distributions in printed tapes at various curvature (Radius of
curvature from 20 mm to 2.5 mm): (a) Distribution of S12 (b) Maximum values of
three stress components
5.3 Parametric studies of tape width and fibre volume fraction
The first parametric study is the influence of the fibre bundle size on the printing
process. In the real 3D printing process, the width of the printed tape is determined by
the fibre bundle size and the off-distance between print bed and nozzle. Printing
experiments of 1 mm, 2 mm and 3 mm tapes are simulated with 30° turning angle and
0.1 mm off-distance (the diameter of the nozzle outlet is scaled up accordingly). As
shown in Figure 12, the increase of width to 2 mm or 3 mm enlarges the area with a
high value of S12. This indicates that fibre misalignment and breakage are more likely
to be induced when using a larger fibre bundle size. And the maximum values of three
stress components increase approximately linearly, but the deformation of the curved
section in these three cases are very similar.
Figure 12. Comparison of FE modelling results at 30° turning angle between different
tape widths: (a) Distribution of S12 (b) Maximum values of three stress components
Another investigated parameter is the fibre volume fraction of filament. Since the
filament is heated to 250°C to melt the PA6 thermoplastic matrix in the printing process,
the fibre volume fraction mainly affects the elastic modulus in fibre direction (E1). In
this parametric study, 0.5-, 1- and 1.5-times elastic modulus are used, representing
equivalent fibre volume fraction 𝑉
𝑓 of 15%, 25% and 40% approximately (The 𝑉
𝑓 of
commercial filaments with 1K continuous carbon fibres is usually between 20% to
30%). As shown in Figure 13, the three stress components increase dramatically as a
higher fibre fraction is used. With a 1.5-times elastic modulus, the widening of the
curved section is similar to that in the cases with large turning angles (> 90º), which
means the increase of fibre volume fraction has a great influence on the defect
formation. The sharp increase of S11 stress to 2001 MPa also indicates the fibre
breakage. It is therefore suggested that when printing composite filaments with high
fibre volume fractions, the printing defects including shape inaccuracy, fibre folding
and breakage, would happen at a relatively small turning angle, even using a small
fibre bundle size.
Figure 13. FE Modelling results at 30° turning angle with different elastic modulus in
fibre direction: (a) Distribution of S12 (b) Maximum values of three stress
components
6. Conclusions
This paper investigates the mechanism of 3D printing continuous carbon fibre filament
from both experiments and numerical simulations. In the straight-printed stripe, high
porosity and fibre misalignment are observed, which is caused by the weak interface
of materials and the uneven pressure from the tip of the nozzle. When fibre tapes are
printed with a small turning angle or curvature, filaments can be flattened on the bed
with nearly constant width and fibre paths. As the turning angle and curvature
increases, lots of printing defects occur and aggravate, including path error, fibre
twisting, folding and misalignment, caused by the excessive tensile force from the
nozzle, debonding with the print bed and the inconsistent diameter of nozzle outlet
and fibre filament.
In the samples with angles >120° and curvature radius < 5mm, fibre breakage and the
significant change of thickness can be seen, which means such turning angles and
curvatures should be avoided in the paths design of the continuous fibre printing. The
good agreement between experimental and numerical results also indicates the FE
model could offer a useful tool for analysing the printing process of continuous fibre
filament. Further parametric studies show that the increase of fibre bundle size would
aggravate the printing defects including fibre misalignment and breakage. But the
deformations of printed stripes are approximately the same in the cases with 30°
turning angle. The increase of fibre volume fraction has a great influence on the defect
formation. Some printing defects observed only in the cases with large turning angles,
such as shape inaccuracy and fibre folding, would appear in the cases with higher
fibre volume fractions when printing at a relatively small turning angle and fibre bundle
size.
Other printing parameters worth studying, such as nozzle speed and off-distance,
would also influence the quality of finished CCFRP composites. Part of printing defects,
for example, filament folding and fibre breakage, can not be observed directly in the
current model, since only elastic properties and shell elements are used to simulate
the simplified tape. Therefore, the FE model should be improved to better investigate
the alignment and breakage of fibres in the printing process. Future research is still
required to address those issues and inform the design of 3D printing process for
continuous fibre reinforced composites.
Acknowledgement
Dongmin Yang would like to acknowledge Royal Society (IEC/NSFC/170418), EPSRC
(EP/P017169/1), EPSRC CIMComp Hub (EP/P006701/1-RIS3718946) for financial
support of this study.
REFERENCES
[1] P. Parandoush, D. Lin, A review on additive manufacturing of polymer-fiber composites,
Composite Structures 182 (2017) 36-53.
[2] B. Jenett, S. Calisch, D. Cellucci, N. Cramer, N. Gershenfeld, S. Swei, K.C. Cheung, Digital Morphing
Wing: Active Wing Shaping Concept Using Composite Lattice-Based Cellular Structures, Soft Robot
4(1) (2017) 33-48.
[3] H. Zhang, D. Yang, Y. Sheng, Performance-driven 3D printing of continuous curved carbon fibre
reinforced polymer composites: A preliminary numerical study, Composites Part B: Engineering 151
(2018) 256-264.
[4] N. Bakhshi, M. Hojjati, An experimental and simulative study on the defects appeared during tow
steering in automated fiber placement, Composites Part A: Applied Science and Manufacturing 113
(2018) 122-131.
[5] J. Xu, C. Li, S. Mi, Q. An, M. Chen, Study of drilling-induced defects for CFRP composites using new
criteria, Composite Structures 201 (2018) 1076-1087.
[6] E.D. Eneyew, M. Ramulu, Experimental study of surface quality and damage when drilling
unidirectional CFRP composites, Journal of Materials Research Technology 3(4) (2014) 354-362.
[7] K. KozaczuK, Automated fiber placement systems overview, Prace Instytutu Lotnictwa (2016).
[8] B.C. Kim, P.M. Weaver, K. Potter, Manufacturing characteristics of the continuous tow shearing
method for manufacturing of variable angle tow composites, Composites Part A: Applied Science and
Manufacturing 61 (2014) 141-151.
[9] S.M.F. Kabir, K. Mathur, A.-F.M. Seyam, A critical review on 3D printed continuous fiber-
reinforced composites: History, mechanism, materials and properties, Composite Structures 232
(2020).
[10] N. Li, Y. Li, S. Liu, Rapid prototyping of continuous carbon fiber reinforced polylactic acid
composites by 3D printing, Journal of Materials Processing Technology 238 (2016) 218-225.
[11] R. Matsuzaki, M. Ueda, M. Namiki, T.K. Jeong, H. Asahara, K. Horiguchi, T. Nakamura, A.
Todoroki, Y. Hirano, Three-dimensional printing of continuous-fiber composites by in-nozzle
impregnation, Sci Rep 6 (2016) 23058.
[12] X. Tian, T. Liu, C. Yang, Q. Wang, D. Li, Interface and performance of 3D printed continuous
carbon fiber reinforced PLA composites, Composites Part A: Applied Science and Manufacturing 88
(2016) 198-205.
[13] F. Baumann, J. Scholz, J. Fleischer, Investigation of a New Approach for Additively Manufactured
Continuous Fiber-reinforced Polymers, Procedia CIRP 66 (2017) 323-328.
[14] J. Frketic, T. Dickens, S. Ramakrishnan, Automated manufacturing and processing of fiber-
reinforced polymer (FRP) composites: An additive review of contemporary and modern techniques
for advanced materials manufacturing, Additive Manufacturing 14 (2017) 69-86.
[15] H. Zhang, A.N. Dickson, Y. Sheng, T. McGrail, D.P. Dowling, C. Wang, A. Neville, D. Yang, Failure
analysis of 3D printed woven composite plates with holes under tensile and shear loading,
Composites Part B: Engineering 186 (2020).
[16] A. Todoroki, T. Oasada, Y. Mizutani, Y. Suzuki, M. Ueda, R. Matsuzaki, Y. Hirano, Tensile property
evaluations of 3D printed continuous carbon fiber reinforced thermoplastic composites, Advanced
Composite Materials 29(2) (2020) 147-162.
[17] H. Mei, Z. Ali, I. Ali, L. Cheng, Tailoring strength and modulus by 3D printing different continuous
fibers and filled structures into composites, Advanced Composites and Hybrid Materials 2(2) (2019)
312-319.
[18] M. Caminero, J. Chacón, I. García-Moreno, G. Rodríguez, Impact damage resistance of 3D
printed continuous fibre reinforced thermoplastic composites using fused deposition modelling,
Composites Part B: Engineering 148 (2018) 93-103.
[19] N. Ichihara, M. Ueda, Y. Urushiyama, A. Todoroki, R. Matsuzaki, H. Hirano, Progressive damage
simulation for a 3D-printed curvilinear continuous carbon fiber-reinforced thermoplastic based on
continuum damage mechanics, Advanced Composite Materials 29(5) (2020) 459-474.
[20] J. Shang, X. Tian, M. Luo, W. Zhu, D. Li, Y. Qin, Z. Shan, Controllable inter-line bonding
performance and fracture patterns of continuous fiber reinforced composites by sinusoidal-path 3D
printing, Composites Science and Technology 192 (2020).
[21] E. Pei, A. Lanzotti, M. Grasso, G. Staiano, M. Martorelli, The impact of process parameters on
mechanical properties of parts fabricated in PLA with an open-source 3-D printer, Rapid Prototyping
Journal (2015).
[22] W. Hao, Y. Liu, H. Zhou, H. Chen, D. Fang, Preparation and characterization of 3D printed
continuous carbon fiber reinforced thermosetting composites, Polymer Testing 65 (2018) 29-34.
[23] J. Justo, L. Távara, L. García-Guzmán, F. París, Characterization of 3D printed long fibre
reinforced composites, Composite Structures (2017).
[24] H. Shiratori, A. Todoroki, M. Ueda, R. Matsuzaki, Y. Hirano, Mechanism of folding a fiber bundle
in the curved section of 3D printed carbon fiber reinforced plastics, Advanced Composite Materials
(2019) 1-11.
[25] R. Matsuzaki, T. Nakamura, K. Sugiyama, M. Ueda, A. Todoroki, Y. Hirano, Y. Yamagata, Effects
of Set Curvature and Fiber Bundle Size on the Printed Radius of Curvature by a Continuous Carbon
Fiber Composite 3D Printer, Additive Manufacturing 24 (2018) 93-102.
[26] G. Chabaud, M. Castro, C. Denoual, A. Le Duigou, Hygromechanical properties of 3D printed
continuous carbon and glass fibre reinforced polyamide composite for outdoor structural
applications, Additive Manufacturing 26 (2019) 94-105.
[27] D. Yang, H. Zhang, J. Wu, E.D. McCarthy, Fibre flow and void formation in 3D printing of short-
fibre reinforced thermoplastic composites: An experimental benchmark exercise, Additive
Manufacturing (2020).
[28] S. Khan, D.-H. Lee, An adaptive dynamically weighted median filter for impulse noise removal,
EURASIP Journal on Advances in Signal Processing 2017(1) (2017).
[29] D.J. Dijkstra, Guidelines for rheological characterization of polyamide melts (IUPAC Technical
Report), Pure and Applied Chemistry 81(2) (2009) 339-349.
[30] K. Ishii, A. Todoroki, Y. Mizutani, Y. Suzuki, Y. Koga, R. Matsuzaki, M. Ueda, Y. Hirano, Bending
fracture rule for 3D-printed curved continuous-fiber composite, Advanced Composite Materials
28(4) (2018) 383-395.
[31] A. Beakou, M. Cano, J.-B. Le Cam, V. Verney, Modelling slit tape buckling during automated
prepreg manufacturing: A local approach, Composite structures 93(10) (2011) 2628-2635.
[32] C. Wohl, F.L. Palmieri, A. Forghani, C. Hickmott, H. Bedayat, B. Coxon, A. Poursartip, B. Grimsley,
Tack measurements of prepreg tape at variable temperature and humidity, Composites and
Advanced Materials Expo (CAMX 2017), 2017.
[33] M. Iragi, C. Pascual-González, A. Esnaola, C.S. Lopes, L. Aretxabaleta, Ply and interlaminar
behaviours of 3D printed continuous carbon fibre-reinforced thermoplastic laminates; effects of
processing conditions and microstructure, Additive Manufacturing 30 (2019).
[34] H.M. El-Dessouky, C.A. Lawrence, Ultra-lightweight carbon fibre/thermoplastic composite
material using spread tow technology, Composites Part B: Engineering 50 (2013) 91-97.
[35] Y. Ma, M. Ueda, T. Yokozeki, T. Sugahara, Y. Yang, H. Hamada, A comparative study of the
mechanical properties and failure behavior of carbon fiber/epoxy and carbon fiber/polyamide 6
unidirectional composites, Composite Structures 160 (2017) 89-99.
[36] C. Leopold, M. Schutt, W.V. Liebig, T. Philipkowski, J. Kurten, K. Schulte, B. Fiedler, Compression
Fracture of CFRP Laminates Containing Stress Intensifications, Materials (Basel) 10(9) (2017).
[37] K. Poulios, C.F. Niordson, Micro-buckling of periodically layered composites in regions of stress
concentration, Composite Structures 157 (2016) 424-435.
[38] L.G. Blok, M.L. Longana, H. Yu, B.K.S. Woods, An investigation into 3D printing of fibre reinforced
thermoplastic composites, Additive Manufacturing 22 (2018) 176-186.
