Optimising The FDM Additive Manufacturing Process - Sympletic
Optimising The FDM Additive Manufacturing Process - Sympletic
TITLE
Optimising the FDM additive manufacturing process to achieve maximum tensile strength: a state-of-
the-art review
AUTHORS
Gordelier, TJ; Thies, P; Turner, L; et al.
JOURNAL
Rapid Prototyping Journal
DEPOSITED IN ORE
16 August 2019
http://hdl.handle.net/10871/38357
A NOTE ON VERSIONS
The version presented here may differ from the published version. If citing, you are advised to consult the published version for pagination, volume/issue and date of
publication
Optimising the FDM additive manufacturing process to
achieve maximum tensile strength:
A state of the art review.
Abstract
Additive manufacturing or ‘3D printing’ is a rapidly expanding sector and
is moving from a prototyping service to a manufacturing service in its own
right. With a significant increase in sales, Fused Deposition Modelling (FDM)
printers are now the most prevalent 3D printer on the market. The increase in
commercial manufacturing necessitates an improved understanding of how to
optimise the FDM printing process for various product mechanical properties.
This paper seeks to identify optimum print parameters for the FDM process
to achieve maximum tensile strength through a review of recent studies in this
field. The effect of the governing printing parameters on the tensile strength of
printed samples will be considered, including: material selection, print orienta-
tion, raster angle, air gap and layer height.
Key findings include material recommendations, such as the use of emerging
print materials like polyether-ether-ketone (PEEK) to produce samples with
tensile strength over 200% that of conventional materials such as acrylonitrile
butadiene styrene (ABS). Amongst other parameters it is shown that printing in
the ‘upright’ orientation should be avoided (samples can be up to 50% weaker in
this orientation) and air gap and raster width should be concurrently optimised
to ensure good ‘inter-raster’ bonding. The optimal choice of raster angle depends
on print material; in ABS for example, selecting a 0◦ raster angle over a 90◦
angle can increase tensile strength by up to 100%.
The paper conclusions provide researchers and practitioners with an up to
date, single point reference, highlighting a series of robust recommendations to
optimise the tensile strength of FDM printed samples. Improving the mechanical
performance of FDM printed samples will support the continued growth of this
technology as a viable production technique.
Keywords: FDM, tensile strength, raster angle, ABS, PLA, print orientation,
air gap.
∗ Correspondingauthor
Email address: [email protected] (T.Gordelier)
2
Figure 1: U.S. 3D printing market forecast by year, 2014-2025. Summarised from [1].
Before detailing the FDM process, it should be noted that this process is
also referred to as Fused Filament Fabrication (FFF). The term FDM was trade-
marked by Stratysys Inc. and hence an alternative name, FFF, was developed
45 by some users to avoid conflict. The two names refer to identical technoloy.
As outlined in Section 1, FDM printing is one of the most widely adopted
forms of AM and many publications outline the printing process in detail; a
3
Figure 2: 3D printing technologies adopted by respondents, from a 2017 global survey of 900+
users. Summarised from [3].
selection of these are included for reference [4, 5, 6, 7]. A detailed explanation
of the physical phenomena behind the FDM printing process is also provided
50 by [8]. In this section, a brief summary of the process will be provided with
reference to Figure 3 which outlines the key components of a FDM printer. The
first step of printing a FDM component is to develop a computer aided design
(CAD) model of the component. The CAD model is commonly exported as
a stereolithography (STL) file which is then ‘sliced’ by specialist 3D printing
55 software and read by the printer to print the component in a series of ‘sliced’
layers. A spool of printing filament is drawn through the extruder head, which
heats the filament to a semi-molten state. It is then forced through an extrusion
nozzle and onto the printer build platform. The extruder head (or the build
platform) is able to move around in the X-Y plane creating a 2D slice of the
60 required part. Once the first layer is complete, the extruder head (or the build
platform) will move in the Z direction, to enable a second layer of 2D material
to be applied on-top of the first. The semi-molten state of the material allows
adjacent layers to fuse together forming the 3D solid layer by layer. Some 3D
structures may have overhangs, and as such require a support structure during
65 the printing process. For this there are two main options: The FDM printer
can print the support structure (or ‘scaffolding’) in a more fragile state so that
it can be clipped off once printing is complete. Alternatively, some printers
use a second nozzle utilising a different material to print the support structure
concurrently with the main component. The support structure is then removed
70 once printing is complete (often the support structure material is soluble for
example).
4
Figure 3: Key components of a FDM printer. Movement is achieved in the x, y and z direction
by either movement of the extrusion head (purple arrows), movement of the build platform
(green arrows) or a combination of the two e.g. x and y direction achieved by movement of
the extrusion head and z direction achieved by movement of the build platform. Developed
from original image by [6].
3. Applied methodologies
5
Figure 4: Dimensions for specimen 1A of ISO 527-2-1996 standard for the determination of
tensile properties of moulding and extrusion plastics [12].
- qualification and certification process for materials and components’ [10], may
help to bridge this lack of guidance, however this publication refers to generic
paths for qualification and cites a range of existing test methodologies rather
95 than specifically creating a new standard for AM.
Despite this lack of guidance, many studies have investigated the effect of nu-
merous printing parameters on the final printed product. Within these studies
a range of methodologies have been adopted to quantify the tensile strength of
printed samples. The approaches fall into one of two categories: The ‘Hourglass’
100 type specimen, utilising a dumbbell or hourglass shaped test piece and covered
in Section 3.1, and the alternative ‘Rectangular’ type specimen, utilising a rect-
angular shaped test piece and covered in Section 3.3. The following sections will
review the two broad approaches adopted by the majority of studies discussed
here and highlight potential weaknesses with these approaches.
6
Figure 5: FEA model demonstrating the equivalent stress in an ISO 527-2 specimen 1A
(dimensions as detailed in Figure 4). Model is subject to a tensile load of 5,000N. The peak
stress in red occurs adjacent to the bend radii.
adopted by [22, 23, 24, 25, 26] and [27]. The ISO 527-3 approach [28] (specifically
125 for films and sheets) also utilises an ‘hourglass’ shaped sample and is utilised
by [27].
There are three further studies that have adopted an ‘hourglass’ type tensile
test specimen, though have not specifically followed either the ASTM D628 or
the ISO 527-2 test methodology. These are [23, 29] and [30].
130 Although the majority of these studies report valid results when adopting the
above approach some issues with this type of methodology have been reported
by [7] and others, and will be highlighted below.
7
Figure 6: Tensile tests conducted by [31] demonstrate issues with premature shear failure at
the sample radius due to stress concentrations developing. Image replicated from [31].
Further to these studies [14] also uses the ASTM D638 approach for a com-
155 parative study of ABS and PLA specimens printed using a ‘RepRap’ type
printer. This study states that “many specimens broke outside of the gauge
length due to assumed stress concentrations in the regions changing geometry”.
Despite observing this during the experimental phase, the data from these in-
vestigations were reported as valid results.
160 It is of concern that further studies utilising the discussed methodologies do
not interpret an issue with the obtained results if the samples fail outside of
the gauge length. Only a limited number of studies publish images of failed
specimens which makes it challenging to establish the validity of results. Help-
fully, a study by [25] utilising the ISO 527-2 approach to characterise properties
165 of FDM printed ABS samples publish an image demonstrating a selection of
failed specimens, this is replicated in Figure 7. Of the 15 specimens detailed
only three demonstrate fracture within the gauge length of the samples, with
the remaining failures occurring at, or very near, the radii. Again, it appears in
this study that all these samples were included as valid results.
170 A single example image of a failed specimen is published by [27] who utilise
an ‘hourglass’ specimen to review a range of materials from both FDM and SLA
printing. This image also demonstrates a failure at the radius of the samples
but no reference is made to this being a rejected sample or out of scope.
For both of these examples, these out of range failures could be considered
175 a ‘worst-case-scenario,’ underestimating the true strength of the material. In-
clusion of these results in the reported data should therefore not lead to over
optimistic conclusions and can perhaps be justified on this basis. However, it is
surprising to see no discussion of this in the presented reports.
Some studies include details of samples that have successfully broken within
180 the sample gauge length including [15, 29, 18] and [6]. It is a valuable contri-
bution that all the publications discussed above do provide details on the failed
samples, whether deemed ‘successful’ or not. Many of the studies however, do
8
Figure 7: Results published in [25] highlighting the range of failure modes when utilising ISO
527-2 to investigate the characteristics of FDM printed ABS. The samples were printed with
a range of raster angles at the following orientations: (a) Horizontal orientation. (b) Vertical
orientation. (c) Perpendicular orientation. All samples excluding row (a) H-60, H-90 and
H-45 demonstrate failure at the sample radius. Image replicated from [25].
not publish this information which raises concerns regarding whether early fail-
ures occurred at the sample radii and if so, how this could affect the validity of
185 the results.
Given these concerns regarding the ‘hourglass’ type specimens, alternative
methodologies are discussed below.
9
the ‘hourglass’ ASTM D638 standard for AM applications suggesting it is “rec-
ommended for highly oriented and/or high modulus fibre reinforced polymer
210 composites” [35].
There is limited adoption of the ISO 527-5 standard, although it is referred to
by [36] who propose a methodology for single layer AM specimens and state that
this standard is applicable due to “structural analogies” between AM specimens
and fibre-reinforced plastics.
215 Having considered the various approaches adopted by the sector for quan-
tifying the tensile strength of samples, the next section will detail the results
obtained across the various studies.
10
In addition to alternative materials, some studies have considered the use
of additives with standard materials to develop composites in order to improve
mechanical properties. A detailed review of FDM fibre reinforced polymers
is provided by [8] who summarise numerous studies that investigate the po-
255 tential advantages of these composite materials. This review concludes that
continuous fibre reinforced composites are significantly stronger than discon-
tinuous fibre composites, achieving strength comparable to aluminium in some
instances. The direction of loading also plays a significant role, with loading
perpendicular to the raster beads resulting in much weaker components. The
260 concept of anisotropy in FDM printed parts will be further explored in Sections
4.2 and 4.3.
One such example of a composite study is conducted by [15] who compare
standard ABS to ABS with 5 wt.% discontinuous jute fibre (Cyclolac , GE ®
ABS resin), ABS with 5 wt.% titanium dioxide (TiO2 ) and an ABS polymeric
265 blend with 5 wt.% of a thermoplastic elastomer (TPE). Utilising the ASTM
D638 standard the tensile strength of each variant is investigated and only the
ABS with TiO2 is shown to improve the tensile strength of the standard ABS
samples, increasing the UTS by 13.2% on average. The jute composite and the
TPE blend actually reduce the UTS by 9% and 16% respectively. This work
270 is continued in a further study, [37], which focuses on ABS polymer matrix
composites and polymer blends. Although this study demonstrates that some
of these additives can lead to a reduction in sample anisotropy, this comes at
the expense of UTS with a reduction in overall sample tensile strength. For
further studies relating to FDM fibre reinforced composites see [8].
275 Even within each material classification there are variations between the
specific feedstock utilised for printing and [38] consider the tensile strength of
specimens printed from a range of different coloured PLA filaments (white,
blue, grey, black and natural). Utilising the ASTM D638 standard, this study
demonstrates that natural PLA (with no added dye) has the highest tensile
280 strength whilst grey PLA has the lowest. On average, the samples printed with
natural PLA have an ultimate tensile strength 12% higher and a yield strength
14% higher than the grey samples. Blue, white and black PLA samples have
tensile strengths scattered between these two extremes.
To summarise the large number of studies considering the effect of material
285 selection on component strength, Table 1 has been included.
In addition to material selection, numerous printing parameters can be varied
to optimise the printing process and the following sections will review these.
11
Table 1: Summary of studies investigating the effect of material selection on printed sample
tensile strength. Table notes: *Denotes studies utilising ‘RepRap’ printer technology
12
Figure 8: Raster angle definition. Blue lines identify raster pattern and large black arrows
identify direction of tensile loading. In this paper, raster angle will be defined from 0◦ being
fully aligned with the tensile loading direction to 90◦ being perpendicular to the tensile loading
direction.
295 considered the raster orientation and will be detailed below in sub-sections re-
lating to printing material. Conventional printing practice generally utilises a
solid outer skin filled with an internal structure to speed up printing time and
minimise material use. This is specified by the pattern of the structure and a
% infill value, with machines often defaulting to 20% infill. However, for the
300 experimental investigations reported below, unless otherwise stated, the sam-
ples are printed at 100% infill i.e. solid blocks. Although this is not normally
adopted in general printing practice, this ensures that it is the material and
printing parameters that are investigated as opposed to the print construction.
13
75% the strength of the 0◦ samples). On average, the 0◦ specimens obtain
325 a tensile strength of 94.8% compared to an injection moulded equivalent. The
study also considers the tension-tension fatigue performance of the samples with
±45◦ demonstrating the highest cycles to failure.
The work by [25] utilises ISO 527-2 to investigate the effect of five different
raster angles (0◦ , 30◦ , 45◦ , 60◦ , 90◦ ) on a range of mechanical properties. This
330 study also considers the effect of print orientation on these mechanical properties
which will be explained in Section 4.3. For standard printing orientations (flat or
‘on-edge’) the 0◦ raster angle demonstrates the highest tensile strength. For the
upright print orientation (which would not usually be adopted in conventional
printing practice) the 90◦ raster angle demonstrates the highest tensile strength.
335 A further study by [39] utilises a multi-layered iterative (MIA) group method
of data handling (GMDH) algorithm to investigate optimal parameters for FDM
printing applying differential evolution (DE). Experimental results applying the
ISO 527-2 process are utilised to verify the mathematical model. Pre-tests are
conducted to inform the algorithm and the full suite of tests consider two raster
340 angles 0◦ and 90◦ . The developed algorithm then calculates the optimal process
parameters for tensile strength. Many different variables are investigated and
although raster angle is shown not to be the most significant factor influencing
tensile strength, the optimal raster angle for tensile strength is found to be 50◦ .
A finite element (FE) model of FDM specimens is developed by [16] who
345 compare FE results to physical testing following the ASTM D638 test procedure.
Three raster angles are considered 0◦ , 45◦ and 90◦ . Although summary data
of results is not provided in the paper, graphical results of the physical tests
clearly demonstrate the maximum UTS is obtained with a raster angle of 0◦ .
Further investigations are conducted by [5] who utilise the rectangular ASTM
350 D3039 specimens to investigate the strength of a range of raster angles and layer
thickness on the tensile strength of FDM printed ABS. Three raster angles are
considered, 0◦ , 45◦ and 90◦ and across the range of variables considered samples
printed with a 0◦ raster angle demonstrated the highest tensile strength.
An investigation by [19] utilises slightly different terminology, but also con-
355 siders the effect of raster angle on the strength of specimens printed to the
ASTM D638 standard. Within this standard a range of specimen sizes are
considered and this investigations reviews the results from ‘hourglass’ Type I,
IV and V specimens. In this work ‘longitudinal’ refers to a raster angle of 0◦ ,
‘cross-hatched’ refers to a raster angle alternating by 90◦ between subsequent
360 print layers i.e. 0◦ /90◦ , and ‘transversal’ refers to a raster angle of 90◦ relative
to the length of the specimen. For the two larger specimens investigated, Type
I and IV, a raster angle of 0◦ produced samples with the highest UTS, however
for the smaller sample Type V, a raster angle of 0◦ /90◦ produced the highest
UTS.
365 Raster angle, build direction and material are investigated by [6] who com-
pare both ABS and PC using ASTM D638. Two different raster angles are
considered for the ABS samples: ±45◦ and 0◦ /90◦ . This study finds very little
difference between the UTS of the samples with different raster angels. A raster
angle of 0◦ /90◦ resulted in a UTS 3.7% and 3.0% higher than the ±45◦ for
14
370 samples printed in a flat and upright orientation respectively. In the on-edge
printing orientation 0◦ /90◦ samples were 2.3% weaker. These small differences
between sample strengths are insignificant, particularly when considering con-
fidence intervals of the data.
The final study to be discussed here by [14] was introduced in Section 4.1 and
375 utilises a ‘RepRap’ type printer to investigate the effect of raster angle on ABS
utilising the ASTM D638 standard. Two raster angles are considered: 0◦ /90◦
and ±45◦ . The results demonstrate very little variation in tensile strength
between the two types of sample investigated, although the strongest samples
were printed with a 0◦ /90◦ raster angle.
380 Due to the large number of studies discussed here, a summary of the results
regarding selection of raster angle when printing in ABS is detailed in Table 2.
This section has summarised the results from investigations looking at ABS
and the next two sections will summarise findings from alternative materials.
15
Table 2: Summary of studies investigating the effect of raster angle on ABS printed sample
tensile strength. Table notes: † Many of these studies consider multiple variables and as such
there is not one raster angle that achieves maximum tensile strength across all variables;
the angle reported here results in maximum tensile strength for the majority of variables
investigated. *Denotes studies utilising ‘RepRap’ printer technology.
16
Table 3: Summary of studies investigating the effect of raster angle on PLA printed sample
tensile strength. Table notes: † Some of these studies consider multiple variables and as such
there is not one raster angle that achieves maximum tensile strength across all variables;
the angle reported here results in maximum tensile strength for the majority of variables
investigated. *Denotes studies utilising ‘RepRap’ printer technology.
415 the higher strain rate, the 45◦ and 0◦ specimens were 24.2% and 11.0% stronger
than the injection moulded specimens respectively. At the lower strain rate
investigated, all three raster angle directions produced samples with strength
exceeding the UTS of the injection moulded samples.
Finally, the study introduced in Section 4.1 by [14] also investigated the
420 relative strength of PLA samples printed using a ‘RepRap’ printer. Again, two
angles were studied: 0◦ /90◦ and ±45◦ . Minimal variation in tensile strength was
observed between these two defined raster angles, though the highest samples
were achieved with a 0◦ /90◦ raster angle.
A summary of the investigations and results for raster angle selection in PLA
425 is detailed in Table 3.
17
Figure 9: Variation in ultimate tensile strength with print raster angle for polycarbonate
(PC) samples tested to the ASTM D638 standard. Average UTS of 6 specimens for each
raster angle detailed. These results are replicated from [4] and revised to align with the raster
nomenclature adopted in this paper. UTS has been converted from PSI to MPa.
alignment with the direction of the tensile load applied, the higher the UTS.
For samples with fully aligned raster angles (0◦ in the other studies we have
considered) the UTS is approximately 3 times the strength of samples where
the load is applied perpendicular to the raster (90◦ ). Figure 9 details the aver-
440 age results from this study, revised to reflect the angle notation adopted in this
review paper.
PC is further considered in a study by [6] which was introduced in Section
4.2.1. Four different raster angles are considered which differ from those in-
vestigated in other studies: ±45◦ , +30◦ / − 60◦ , +15◦ / − 75◦ and 0◦ /90◦ . The
445 study also investigates the effect of print orientation, which affects the optimal
raster orientation selected. For samples printed flat on the printer bed (as de-
tailed in Figure 10) raster angles of ±45◦ and +15◦ / − 75◦ have similar tensile
strength properties and produce significantly stronger samples than the other
angles investigated. For the other print orientations, on-edge and upright, only
450 two raster angles are investigated, ±45◦ and 0◦ /90◦ . Very little difference is
observed between the two raster angles, with a raster angle of ±45◦ producing
samples with very marginally higher UTS.
Further to this [29] consider the effect of raster angle on the tensile strength
of PEEK samples considering 0◦ /90◦ , 30◦ /-60◦ and ±45◦ . The 0◦ /90◦ sam-
455 ples were significantly stronger than the other samples, demonstrating tensile
strength 35.4% and 30.7% higher than 30◦ /-60◦ and ±45◦ respectively.
A summary of the investigations and results for raster angle selection in
alternative materials is detailed in Table 4.
Having considered the effect of raster angle on the tensile strength of FDM
460 printed materials, the next section will consider the effect of print orientation.
18
Table 4: Summary of studies investigating the effect of raster angle on polycarbonate (PC)
and polyether-ether-ketone (PEEK) printed sample tensile strength. Table notes: † Some of
these studies consider multiple variables and as such there is not one raster angle that achieves
maximum tensile strength across all variables; the angle reported here results in maximum
tensile strength for the majority of variables investigated.
19
Figure 10: The different build orientation options available for FDM printed samples as
published by [20].
20
Table 5: Summary of studies investigating the effect of print orientation on printed sample
tensile strength of ABS, PC and PLA. Table notes: † Some of these studies consider multiple
variables and as such the print orientation reported here results in maximum tensile strength
for the majority of variables investigated.
21
thickness of the samples tested from 0.2mm to 7mm. Across the full range
of sample thicknesses investigated, 0.2mm printed layers consistently result in
higher sample UTS. This appears to contradict the findings presented above by
[22] and reasons for this will be addressed in the Discussion section 5.4.
515 As previously discussed [14] utilise ‘RepRap’ type printers to investigate the
tensile strength of both ABS and PLA printed samples. This study also consid-
ered the effect of varying three layer heights: 0.4mm, 0.3mm and 0.2mm. For
the ABS samples, layer height made minimal difference; samples printed with
the lowest height, 0.2mm, demonstrated marginally higher UTS. In the PLA
520 samples however, layer height made a more significant difference; on average
samples adopting a 0.2mm layer height had 24.5% and 10.0% greater UTS than
samples printed with a 0.3mm and 0.4mm layer height respectively.
The effect of layer height on the UTS of PLA samples is also investigated
by [20] who consider 4 layer heights: 0.06mm, 0.12mm, 0.18mm, 0.24mm. This
525 study also investigates print orientation (as discussed in Section 4.3) and fila-
ment feed rate (which will be discussed in Section 4.6). In the upright build
orientation, which has significantly lower UTS than on-edge or flat build orien-
tations, a larger layer height increases the UTS across all feed rates investigated.
For the on-edge and flat print directions, an increase in layer height has lim-
530 ited affect on UTS at the lowest feed rate (20mm/s) but at the higher feed
rates (50mm/s and 80mm/s) an increase in layer height is observed to reduce
the UTS. The strongest samples were printed in the flat orientation with the
thinnest layer height, 0.06mm. Reasons for these observed behaviours will be
further discussed in Section 5.4.
535 Given the number of studies investigating the effect of layer height on tensile
strength, a summary of the results is detailed in Table 6.
22
Table 6: Summary of studies investigating the effect of layer height on printed sample tensile
strength of ABS and PLA. Table notes: † Some of these studies consider multiple variables and
as such the layer height reported here results in maximum tensile strength for the majority of
variables investigated. *Represents studies utilising ‘RepRap’ printer technology.
Figure 11: Air gap and raster width definitions for FDM printing. (a) Details a plan view of
the raster pattern showing a positive air gap between adjacent rasters. (b) Details the effect
of air gap selection on the cross-section through adjacent deposited rasters.
23
(a) Grid (b) Lines (c) Triangles (d) Tetrahedral
Figure 12: Range of infill options available with Ultimaker replicated from [40].
0.5588mm, and two different raster widths, 0.2034mm and 0.5588mm. Other
variables are also considered in this study but consistently the negative air gap
resulted in a higher sample UTS for all combinations of the other variables.
For a negative air gap, an increase in raster width slightly reduces the UTS of
560 the samples across the range of variables tested. Conversely, for a positive air
gap a higher raster width increases the relative UTS of the samples, although
it should be noted that overall the samples are much weaker than the samples
with a negative air gap. Reasons for this will be addressed in Section 5.5.
24
580 45◦ raster angle, or a honeycomb infill structure made from multiple hexagonal
prisms. The infill density selected for the honeycomb structure is unfortunately
not specified in the paper, although printers commonly default to 20% infill.
This study also investigates a range of different printer outputs; here those
samples printed on a Felix 1.0e will be discussed for consistency. For samples
585 printed in ABS the 100% infill samples had over twice the yield strength of
the honeycomb structure, and the PLA samples had around 2.5 times the yield
strength. This is not a surprising result but the values for specific strength
(σS ), calculated by Equation 1 are perhaps more informative. Here σT = ten-
sile strength and ρ = material density.
σT
Specif ic strength, σS = ( ) (1)
ρ
590 In ABS, the specific strength of the 100% infill samples were approximately
1.4 times the strength of the honeycomb structure, and the 100% PLA samples
were approximately 1.3 times the specific strength of the honeycomb structure.
It is clear that the 100% infill has a measurably higher specific strength.
Another potential variable of the FDM printing process is the feed rate of
595 the print filament; a higher feed rate will clearly speed up the printing process,
leading to quicker production times and lower costs. [20] consider the effect
of this on the tensile strength of samples, investigating the impact of three
feed rates: 20mm/s, 50mm/s and 80mm/s. The study also considers print
orientation and layer height. In the upright build orientation, an increase in
600 feed rate results in a measurable drop in tensile strength of the samples across
all layer heights. For the flat and on-edge samples, feed rate appears to have
minimal impact on the tensile strength of the samples.
In addition to investigating the impact of PLA colour on the tensile strength
of samples as discussed in Section 4.1, [38] also consider the effect of varying the
605 extruder temperature when using ‘RepRap’ type printers. Four printing tem-
peratures are investigated: 190◦ C, 200◦ C, 210◦ C and 215◦ C. A large variation
is apparent in the results but a general correlation of increasing yield strength
with increasing extruder temperature is observed, however there is a slight dip
in sample yield strength at a temperature of 200◦ C. Reasons for this observed
610 behaviour will be discussed in Section 5.6.
5. Discussion
5.1. Material selection
Overall, the studies considering the effect of material selection on tensile
strength find that PLA, PC and PEEK all out-perform ABS. Various reasons
615 are discussed for this. In addition to higher tensile strength, [6] suggest that PC
samples showed much lower coefficients of variation (COV), a measure of relative
variability, than the ABS samples. The study points out that the printer used
for manufacturing the PC samples was a professional grade printer, compared
to the ‘hobby-grade’ printer used for the ABS samples; this could have affected
620 variability and quality of the produced samples.
25
The study by [14] utilise a range of ‘Rep-rap’ printers of a similar grade
to compare ABS and PLA and demonstrate that PLA has a higher tensile
strength. The failure surfaces of these samples are assessed and it is found
that PLA specimens demonstrate superior inter-raster and inter-layer bonding,
625 appearing more like a “homogeneous solid than a composition of individual
extruded rasters”. This superior bonding will lead to improved tensile strength
performance. This is further supported by the claim in this study that ‘Rep-Rap’
printed PLA samples can achieve comparable strength to injection moulded
PLA parts (which are a homogeneous solid, without inter-raster or inter-layer
630 bonds). ‘Rep-Rap’ ABS parts however, are generally weaker than injection
moulded alternatives. The fact that ‘Rep-Rap’ printed PLA parts can achieve
tensile strengths equivalent to injection moulded parts demonstrates the quality
of the inter-raster and inter-layer bonding in parts printed with this material.
The study also suggests there is significant variety in filament characteristics
635 even for the same material classification and argues filament vendors should
provide better data including composition and mechanical test data for the
printing filament itself.
It is also suggested by [29] that inter-raster and inter-layer bonding may
be the cause of different tensile strength results. In this study SEM (Scanning
640 Electron Microscope) imaging is used to review fracture surfaces and again finds
that in ABS samples the individual rasters can be clearly seen. However, in the
stronger PEEK samples, the fracture surface appears to be one homogeneous
block. Given the relatively new introduction of PEEK this study suggests that,
in time, the printing parameters with this material can be further optimised
645 leading to an even greater tensile strength improvement over ABS.
The investigation by [15] that considers including additives to create an ABS
composite also utilises SEM to interpret the observed results. For those sam-
ples loaded with jute fibre and printed in the flat plane, the fractured samples
were shown to have “multiple craters and voids”, which would clearly lead to a
650 reduction in tensile strength. This is believed to be due to the decomposition
of the jute fibre at 180◦ C, well below the extrusion temperature used in the
printing process. This process of decomposition leads to by-products and gases
creating the observed voids. ABS samples incorporating TiO2 were shown to
have the highest tensile strength and the study suggests the presence of TiO2
655 limits the ability of the plastic macromolecules to slide over one another, which
can be seen by the brittle fracture surface observed in the samples. Brenken
et al. [8] also cite void formation, in addition to fibre damage during printing,
as reasons for under-performing fibre reinforced composites. This review also
highlights the importance of sufficient wetting of the composite fibres during
660 the printing process to maximise the strength of continuous fibre composites.
Further demonstrating the variation within material classifications [38] show
a significant variation between PLA samples printed with different colour fila-
ments; natural PLA, with no added dye, has the highest tensile strength. X-ray
powder diffraction (XRD) is utilised to understand the reasons for this variation.
665 It is found that the variation in tensile strength is related to the crystallinity of
the printed parts, which is dependent on filament colour; the specimens printed
26
using natural PLA have the lowest % crystallinity when analysed utilising XRD.
27
710 weakness in the upright samples occurs due to the predominant load being
carried across the printed layers rather than along the length of the rasters;
the links between adjacent layers are referred to as inter-layer fusion bonds by
[20]. This inter-layer bonding between consecutive layers is much weaker than
the trans-raster strength along the length of the rasters, and hence samples
715 printed in the upright direction prove to be significantly weaker. This theory is
supported by [20] who suggest the difference in UTS due to print orientation is
attributable to the increased number of the weaker, inter-layer fusion bonds in
the upright samples.
Interestingly, [6] observe very little variation in tensile strength with print
720 orientation in ABS but significant variation with PC samples. This demon-
strates that print orientation needs to be optimised for the selected printing
material (as the strength of the inter-layer and inter-raster bonds will be strongly
affected by printing material as well as print parameters).
Print direction will also effect printing times and hence production costs.
725 This is addressed by [25] and [20] where the data shows both flat and on-edge
samples printed to ASTM D638 dimensions are significantly quicker to print
than upright samples (although print time is also affected by layer height). The
consideration of print time provides further confirmation that upright printing
should be avoided.
730 It is clear from most of the studies that FDM printed samples demonstrate
significant anisotropy and some further studies have looked at techniques for re-
ducing this behaviour. For example, [41] utilise ionizing radiation with copoly-
mer blends to improve the mechanical performance of PLA. Through exposure
to gamma rays post printing, inter-layer adhesion of the samples was improved
735 via the introduction of crosslinks between polymer chains. This study demon-
strates this technique can be used to improve the tensile strength of FDM sam-
ples and reduce anisotropy. Another study by [37] also seeks to reduce the
anisotropy of FDM printed samples by enhancing ABS with additives. Six
polymer matrix and four polymer blends are investigated, and although it is
740 shown that anisotropy can be reduced, this comes at the expense of an overall
reduction in the UTS of the samples.
Further studies of this nature have the potential to both reduce anisotropy
and improve the mechanical performance of FDM printed samples, and are an
area of continued research.
28
755 authors therefore suggest that increasing the individual layer height reduces the
total number of layers required, which, due to the above reasons, increases the
strength of the sample.
The opposite effect is observed in a study by [5], who also consider ABS.
Here, the lower layer height results in higher sample UTS. To understand this be-
760 haviour, this study investigates the printed sample structures by using a Keyence
VHX-600 digital microscope to quantify the air-gap to material ratio for sam-
ples printed with differing layer heights. The sample printed with a greater layer
height, 0.4mm, had a 5.26% air-gap to material ratio compared to the 0.2mm
layer sample, which had a 0.3% ratio. This reduced air-gap to material ratio for
765 the samples with lower layer height results in more material to carry the tensile
load and hence a higher UTS.
These results are further supported by [20] who find that for PLA sam-
ples printed in a flat or on-edge orientation a lower layer height results in a
higher tensile strength. For samples printed in an upright orientation however,
770 the lower layer height reduces the tensile strength. This is due to the previ-
ously discussed issue of inter-layer fusion between consecutive layers resulting
in weaker links than trans-raster loading. In an upright sample, a reduction
in layer height actually increases the number of weak inter-layer fusion bonds
across which the tensile force is experienced, resulting in a reduction in overall
775 tensile strength.
As discussed in Section 5.3, layer height will also affect printing time and
therefore production costs. It is shown by [20] that the lower the layer height,
the longer the printing time; reducing layer height from 0.24mm to 0.06mm
can increase the print time by up to 300%. Despite the fact that reducing
780 layer height can improve product mechanical properties, this must be balanced
against the associated increase in print time and production costs. The optimal
solution will therefore be very project and product dependant.
29
800 consecutive raster roads in a given layer and there is more material to carry the
tensile load for a given unit area; hence the UTS goes up. This paper also
considers the effect of raster width, and demonstrates that optimum tensile
strength is achieved when correctly selecting raster width and air gap to be
complementary, achieving optimum bonding between consecutive raster roads.
805 This paper does warn about ‘overfill’ flaws in samples occurring when a negative
air gap is used with a thicker raster width, essentially resulting in a 12.5%
overlap between raster roads. This ‘overfill’ can create flaws and an unpleasant
surface appearance and is shown to slightly reduce the UTS of the samples
investigated. The correct balance between air gap and raster width must be
810 achieved to ensure minimal air gap and good bonding between consecutive raster
roads without excessive material leading to instances of ‘overfill’.
30
hence stress distribution within them will be much more complex. The results
presented must therefore be interpreted and optimised for any particular ge-
845 ometry. The use of FEA numerical models to identify the stress distribution
and hence allow the user to select the optimum print parameters (such as print
direction and raster angle) for that geometry should be considered.
The majority of the studies considered in this review are empirical studies
presenting results from physical test campaigns. Whilst this satisfies the main
850 objective of this paper, studies addressing the physical processes governing FDM
printing provide further insight into the observed results, and highlight opportu-
nities to use predictive tools and numerical models to predict printed component
properties. This area of study is addressed by [8] who discuss several studies of
interest.
855 For example, the tensile strength of printed components has been shown
to be strongly dependant on the strength of the inter-raster and inter-layer
bonds. Various physical processes govern the formation of these bonds and
are considered in detail by [42, 43, 44, 45], with [16] developing filament and
macro scale FE models to consider the effect or raster angle and air gap in ABS
860 printed parts and comparing results to a physical test campaign. In addition to
this, the heat transfer and solidification of the printed parts will also affect the
tensile strength. Two early papers, [46] and [47], develop 1D and 2D models
respectively of the thermal processes involved in FDM. Further studies have
been conducted to develop the understanding of these processes including [48,
865 49, 50, 51, 52, 53] and [54].
The high temperatures involved in the printing process can also lead to
residual stress accumulation and/or deformation in the produced part. Analytic
solutions for simple geometries have been developed to predict the warpage
of printed components by [48] and [55]. A 3D FEA model of the printing
870 process is developed by [56] and is subsequently used to develop experimental
investigations [57]. 3D simulations of composite printed parts are also presented
by [58] and [59].
Despite this large body of work, investigating and modelling the physical
processes involved in FDM additive manufacturing, the conclusion of the review
875 paper by [8] is that many of the models are overly simplified, and do not fully
describe the processes. It is suggested that future research should seek to develop
further modelling tools to more accurately predict the printing outcomes and
therefore create more certainty on printed product mechanical properties.
6. Conclusions
31
mise the printing process to achieve particular material properties is clearly
advantageous.
This paper has reviewed a range of recent publications aiming to optimise
the FDM printing process in order to improve the tensile strength of printed
890 samples. Reviewing the range of methodologies applied by the different stud-
ies highlights the need for specific methodologies for the mechanical testing of
additive manufactured components. This would allow for a more systematic
approach across the sector. Any new methodology should permit test pieces
of greater thickness to allow for investigation into the many internal structures
895 possible with 3D printed products.
Although not all the studies concur on all printing variables, some key recom-
mendations can be made to optimise the tensile strength of samples as detailed
below.
32
5. Layer height: For flat and on-edge orientations, layer height should be
930 minimised in order to achieve maximum tensile strength. This improves
the strength for both ABS and PLA samples considered, with one PLA
study demonstrating a 24.5% tensile strength improvement when reducing
layer height from 0.4mm to 0.2mm. Reduced layer height increases print
time, so the improvement in tensile strength should be balanced against
935 the increased manufacturing time (which leads to increased product costs).
6. Other variables: The following variables have been considered in this
paper but need more investigation.
(a) Infill: Numerous infill options are available, and the choice of infill
should be selected to balance print time and print materials against
940 the achieved product tensile strength. Clearly 100% infill will result
in higher tensile strength, and one study demonstrated that it also
showed highest specific strength. Further work is required on the
numerous options available to optimise infill patterns and to strike
the desired balance between printing time and materials used Vs.
945 product tensile strength.
(b) Printing temperature: Although, only considered in one study, an in-
crease in extruder temperature was shown to improve tensile strength
performance due to improved inter-raster and inter-layer bonding.
Further research on this topic is required.
950 (c) Feed rate: The feed rate of the printing filament is considered in one
study and is found to have limited effect on flat or on-edge samples
though some effect is observed in upright samples. The feed rate
should therefore be optimised to minimise printing time, rather than
optimise material properties.
955 In summary, there are numerous printing parameters available when utilising
AM technology. The conclusions presented in this paper can act as a robust
and pragmatic reference point for users seeking to optimise their FDM printing
processes regarding product tensile strength.
Research and development of new materials, different post processing tech-
960 niques and alternative printing parameters will continue in this fast moving
technology field. The development of numerical modelling packages to improve
the prediction of material properties for these products is required, and is a
key challenge for the sector given the fast pace of technology development. Un-
doubtedly, the mechanical properties of FDM printed components will continue
965 to improve, making the use of this technology as a large scale production tech-
nique increasingly plausible.
7. Acknowledgements
This work was funded through the European Union under the European Re-
gional Development Funding (ERDF) Marinei, Grant/Award Number: 05R16P00381.
33
970 The work was commissioned by 3D Kernow in order to optimise their printing
processes to better serve the additive manufacturing requirements of the marine
sector in the Cornwall and Isles of Scilly region.
References
[1] Grand View Research, Market research report: 3D Printing (3DP) market
975 analysis by printer type (desktop, industrial), by technology, by software,
by application, by vertical, by region, and segment forecasts, 2018-2025
(2017).
URL https://www.grandviewresearch.com/industry-analysis/
3d-printing-industry-analysis
980 [2] B. Berman, 3-D printing: The new industrial revolution, Business horizons
55 (2) (2012) 155–162. doi:10.1016/j.bushor.2011.11.003.
[3] C. Moreau, The state of 3D printing (2017).
URL https://www.sculpteo.com/media/ebook/State%20of%203DP%
202017_1.pdf
985 [4] N. Hill, M. Haghi, Deposition direction-dependent failure criteria for fused
deposition modeling polycarbonate, Rapid Prototyping Journal 20 (3)
(2014) 221–227. doi:10.1108/RPJ-04-2013-0039.
[5] B. Rankouhi, S. Javadpour, F. Delfanian, T. Letcher, Failure analysis and
mechanical characterization of 3D printed ABS with respect to layer thick-
990 ness and orientation, Journal of Failure Analysis and Prevention 16 (3)
(2016) 467–481. doi:10.1007/s11668-016-0113-2.
[6] J. T. Cantrell, S. Rohde, D. Damiani, R. Gurnani, L. DiSandro, J. An-
ton, A. Young, A. Jerez, D. Steinbach, C. Kroese, Experimental charac-
terization of the mechanical properties of 3D-printed ABS and polycar-
995 bonate parts, Rapid Prototyping Journal 23 (4) (2017) 811–824. doi:
10.1108/RPJ-03-2016-0042.
[7] J. R. C. Dizon, A. H. Espera, Q. Chen, R. C. Advincula, Mechanical char-
acterization of 3D-Printed polymers, Additive Manufacturing 20 (2017)
44–67. doi:10.1016/j.addma.2017.12.002.
34
[10] DNV GL, Additive manufactring - qualification and certification process
for materials and components (2017).
URL https://rules.dnvgl.com/docs/pdf/DNVGL/CG/2017-11/
DNVGL-CG-0197.pdf
1010 [11] ASTM, ASTM D638-14 Standard test method for tensile properties of plas-
tics, ASTM International (2014).
[12] BSI, BS EN ISO 527-2:1996. Plastics - determination of tensile properties.
part 2: Test conditions for moulding and extrusion plasics, BSI Group
(1996).
1040 [19] A. R. Torrado Perez, D. A. Roberson, Failure analysis and anisotropy eval-
uation of 3D-printed tensile test specimens of different geometries and print
raster patterns, Journal of Failure Analysis and Prevention 16 (1) (2016)
154–164. doi:10.1007/s11668-016-0067-4.
35
[20] J. Chacn, M. Caminero, E. Garca-Plaza, P. Nez, Additive manufacturing
1045 of PLA structures using fused deposition modelling: effect of process pa-
rameters on mechanical properties and their optimal selection, Materials &
Design 124 (2017) 143–157. doi:10.1016/j.matdes.2017.03.065.
[21] J. Kotlinski, Mechanical properties of commercial rapid prototyping ma-
terials, Rapid Prototyping Journal 20 (6) (2014) 499–510. doi:10.1108/
1050 RPJ-06-2012-0052.
[22] A. K. Sood, R. K. Ohdar, S. S. Mahapatra, Parametric appraisal of me-
chanical property of fused deposition modelling processed parts, Materials
& Design 31 (1) (2010) 287–295. doi:10.1016/j.matdes.2009.06.016.
36
[32] BSI, BS EN ISO 527-5:1997. Plastics - determination of tensile properties.
part 5: Test conditions for uni-directional fibre-reinforced plastic compos-
ites, BSI Group (2009).
1085 [33] ASTM, ASTM D3039-17 Standard test method for tensile properties of
polymer matrix composite materials, ASTM International (2017).
37
1120 [43] C. Bellehumeur, L. Li, Q. Sun, P. Gu, Modeling of bond formation be-
tween polymer filaments in the fused deposition modeling process, Jour-
nal of Manufacturing Processes 6 (2) (2004) 170–178. doi:10.1016/
S1526-6125(04)70071-7.
[44] Q. Sun, G. Rizvi, C. Bellehumeur, P. Gu, Effect of processing conditions on
1125 the bonding quality of FDM polymer filaments, Rapid Prototyping Journal
14 (2) (2008) 72–80. doi:10.1108/13552540810862028.
[45] P. K. Gurrala, S. P. Regalla, Part strength evolution with bonding between
filaments in fused deposition modelling: This paper studies how coalescence
of filaments contributes to the strength of final FDM part, Virtual and
1130 Physical Prototyping 9 (3) (2014) 141–149. doi:10.1080/17452759.2014.
913400.
[46] M. A. Yardimci, S. Geri, Conceptual framework for the thermal process
modelling of fused deposition, Rapid Prototyping Journal 2 (2) (1996) 26–
31. doi:10.1108/13552549610128206.
1150 [51] S. Costa, F. Duarte, J. Covas, Thermal conditions affecting heat trans-
fer in FDM/FFE: a contribution towards the numerical modelling of the
process, Virtual and Physical Prototyping 10 (1) (2015) 35–46. doi:
10.1080/17452759.2014.984042.
[52] B. Brenken, A. Favaloro, E. Barocio, N. DeNardo, R. Pipes, Development
1155 of a model to predict temperature history and crystallization behavior of
3D printed parts made from fiber-reinforced thermoplastic polymers, in:
SAMPE conference, Long Beach, CA, 2016.
URL http://refhub.elsevier.com/S2214-8604(17)30447-5/
sbref0400
38
1160 [53] Y. Zhou, T. Nyberg, G. Xiong, D. Liu, Temperature analysis in the fused
deposition modeling process, in: Information Science and Control Engi-
neering (ICISCE), 2016 3rd International Conference on, IEEE, 2016, pp.
678–682. doi:10.1109/ICISCE.2016.150.
[54] K. Pooladvand, C. Furlong, Thermo-mechanical investigation of fused
1165 deposition modeling by computational and experimental methods, Me-
chanics of Composite and multi-functional materials 7 (2017) 45–54.
URL https://link.springer.com/chapter/10.1007/
978-3-319-41766-0_6
[55] L. Xinhua, L. Shengpeng, L. Zhou, Z. Xianhua, C. Xiaohu, W. Zhongbin,
1170 An investigation on distortion of PLA thin-plate part in the FDM process,
The International Journal of Advanced Manufacturing Technology 79 (5-8)
(2015) 1117–1126. doi:10.1007/s00170-015-6893-9.
[56] Y. Zhang, Y. K. Chou, 3D FEA simulations of fused deposition modeling
process, in: ASME 2006 International Manufacturing Science and Engi-
1175 neering Conference, American Society of Mechanical Engineers, 2006, pp.
1121–1128. doi:10.1115/MSEC2006-21132.
[57] Y. Zhang, K. Chou, A parametric study of part distortions in fused deposi-
tion modelling using three-dimensional finite element analysis, Proceedings
of the Institution of Mechanical Engineers, Part B: Journal of Engineering
1180 Manufacture 222 (8) (2008) 959–968. doi:10.1243/09544054JEM990.
[58] M. Talagani, S. DorMohammadi, R. Dutton, C. Godines, H. Baid, F. Abdi,
V. Kunc, B. Compton, S. Simunovic, C. Duty, et al., Numerical simulation
of big area additive manufacturing (3D printing) of a full size car, SAMPE
Journal 51 (4) (2015) 27–36.
1185 URL https://tinyurl.com/y6ughkrn
[59] A. J. Favaloro, B. Brenken, E. Barocio, R. B. Pipes, Simulation of
polymeric composites additive manufacturing using Abaqus, Science in
the Age of Experience, Chicago, IL (2017).
URL http://refhub.elsevier.com/s2214-8604(17)30447-5/
1190 sbref0455
39