Additive Manufacturing 33 (2020) 101171

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Effect of density and unit cell size grading on the stiffness and energy T
absorption of short fibre-reinforced functionally graded lattice structures
János Plocher, Ajit Panesar*
Department of Aeronautics, Imperial College London, SW7 2AZ, UK

A R T I C LE I N FO A B S T R A C T

Keywords: Architectured structures, particularly functionally graded lattices, are receiving much attention in both industry
Functionally grading and academia as they facilitate the customization of the structural response and harness the potential for multi-
Lattices functional applications. This work experimentally investigates how the severity of density and unit cell size
Material extrusion grading as well as the building direction affects the stiffness, energy absorption and structural response of ad-
Lightweight structures
ditively manufactured (AM) short fibre-reinforced lattices with same relative density. Specimens composed of
Composites
tessellated body-centred cubic (BCC), Schwarz-P (SP) and Gyroid (GY) unit cells were tested under compression.
Compared to the uniform lattices of equal density, it was found, that modest density grading has a positive and
no effect on the total compressive stiffness of SP and BCC lattices, respectively. More severe grading gradually
reduces the total stiffness, with the modulus of the SP lattices never dropping below that of the uniform
counterparts. Unit cell size grading had no significant influence on the stiffness and revealed an elastomer-like
performance as opposed to the density graded lattices of the same relative density, suggesting a foam-like be-
haviour. Density grading of bending-dominated unit cell lattices showcased better energy absorption capability
for small displacements, whereas grading of the stretching-dominated counterparts is advantageous for large
displacements when compared to the ungraded lattice. The severity of unit cell size graded lattices does not
affect the energy absorption capability. Finally, a power-law approach was used to semi-empirically derive a
formula that predicts the cumulative energy absorption as a function of the density gradient and relative density.
Overall, these findings will provide engineers with valuable knowledge that will ease the design choices for
lightweight multi-functional AM-parts.

1. Introduction industrial production. Particularly for fused deposition modelling

(FDM) processes, short and continuous fibre-reinforced (FR) filaments
As the industry is increasingly employing architectured structures in offer yet another means of improving the specific stiffness [9–17]. As
the design for additive manufacturing in recent years, aspects of the performance is a function of the fibre length and orientation [15],
structural performance and response are progressively emerging at the which in FDM is dictated by print direction [15,18–21], the inherent
forefront of academic research. Additive manufacturing (AM) has made porosity in a layered structure, compromising the performance, can be
latticing a viable lightweighting practice that is increasingly being offset to even match the performance of FR compression moulded parts
adopted in the product design process to-date [1]. Having the ability to [15].
produce intricate structures with ease, opens new possibilities for the In contrast to foams, representing an unstructured (i.e. stochastic
design and structural optimisation and therefore sets AM into the po- structures) geometry, AM lends itself for fabricating periodically-con-
sition of being a serious alternative to conventional manufacturing trolled cellular structures with repeating unit cells. These types of lat-
techniques like CNC machining. Hence, a number of useful application tices, commonly composed of either strut- or surface-based unit cells
have been presented for the biomedical sector [2–4], aerospace [5] or like e.g. the implicitly defined Triply Periodic Minimal Surfaces
consumer products [6,7]. (TPMS), has immediate benefits for multiphysics problems. Compared
Simultaneously, the AM-industry is - among others - benefitting to continuum solid solutions, they provide - inter alia - a good strength-
from a growing pallet of print materials [8], pushing the boundaries of to-weight ratio [22–24], greater energy absorption and damping cap-
intrinsic mechanical properties and hence progressing towards abilities [25–31], enable partial permeability or improved thermal

⁎
Corresponding author.
E-mail address: a.panesar@imperial.ac.uk (A. Panesar).

https://doi.org/10.1016/j.addma.2020.101171
Received 16 December 2019; Received in revised form 29 February 2020; Accepted 7 March 2020
Available online 20 March 2020
2214-8604/ © 2020 Elsevier B.V. All rights reserved.
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

conduction [32–35], in exchange for a reduced stiffness. Hence, they discussed in [62,63]. Central to answering this question is to identify all
are particularly suitable for multi-objective optimisations, yielding the individual influencing variables that can be altered and create ex-
parts with functional integration in which the geometrical configura- perimentally or numerically derived datasets that allow us to provide
tion has more than a structural purpose. better consultancy when it comes to designing effectively for AM. This
As functional graded lattices (FGLs) are recently not only receiving work sits at the core of this question and aims at answering the question
much attention in research [36,37] but also in industry through more of “how the severity of grading and build direction relates to the stiff-
readily available modelling software like nTopology [38], Altair [39] or ness and energy absorption response of FGLs”.
Materialise [40], specifically tailored solutions, providing a carefully More specifically, this work seeks to experimentally investigate how
weighted trade-off between structural and functional performance, the severity of density and unit cell size grading affects the mechanical
become even more tangible. From a practical standpoint, AM enables us performance of lattices with the same relative density tested in com-
to manufacture FGLs with ease, providing the capability to create pression. In view of DfAM constraints, the effect of the printing direc-
structures with a microstructural, composition and porosity gradient, as tion i.e. prevailing fibre direction on the properties is also elucidated for
reviewed in detail in [4]. Previous works have studied aspects of the density graded lattices. The properties of ungraded (i.e. uniform
modelling [33,36,41–46] as well as manufacturing and testing density from one unit cell to another) lattices with various densities
[2,37,45–48] to explore the mechanical properties of such architected have been obtained to derive trends i.e. apply scaling laws for the
structures. Most studies to-date, conducting finite element analyses, are density graded FGLs. Moreover, the ungraded lattice of the same re-
limited to linear-elastic investigations, whereas models predicting the lative density functioned as reference for the FGLs. Both the stiffness
behaviour beyond the yield point, as shown e.g. in [49], would be more and the energy absorption behaviour are the key performance para-
valuable for understanding the behaviour of FGLs. The key to fathom meters investigated in both types of FGLs. The properties were com-
the underlying mechanisms and improve these models in the future are pared in part to other engineering materials as well as ideally bending-
experimental investigations. and stretching dominated behaviour, following the categorization of
Fundamental insights have been offered through comparison be- Gibson-Ashby. Predictions for the stiffness of lattices with higher fibre
tween the compressive performance of uniform and graded lattices volume fraction were made using the Halpin-Tsai criterion. Semi-em-
composed e.g. graded strut- [37,47,48,50,51] or TPMS-based lattices pirical energy absorption curves were derived with respect to unit cell
[2,52,53]. A major conclusion from most of these initial studies was an type and severity of grading. Moreover, the remit of this work includes
improved energy absorption capability [37,47,48,50,52] of the graded providing insight into the failure mechanisms/structural response.
over the ungraded lattices of the same density and progressive failure of Overall, this work provides valuable guidelines for the design of fibre-
unit cell regions, often exhibiting 45° shear band failure [2,48,54] in reinforced FGLs and sheds light on the potential of grading for func-
rigid polymers or metal specimens. Yang et al. [55] have also conducted tional lightweight AM-parts.
research into the failure of specimens tested transverse to the gradient The outline of the paper is as follows. First, the design-to-test pro-
direction and found a similar behaviour as the uniform counterparts cedure is outlined in the methodology, constituting everything from the
with favourable stiffness and strength values over specimens tested fundamentals for creating the FGLs to the approaches for assessing their
parallel to the density gradient. Recently, light was also shed on the mechanical performance. Secondly, the results of the compression tests
failure behaviours induced by differently radially graded lattices [49] are presented for both baseline (ungraded) and graded specimens.
as well as the potential of this type of grading to outperform uniform Subsequently, these are discussed considering the observed structural
lattices both in terms of mechanical performance and permeability response/failure and recorded performance before concluding by
[56]. So far only little research was focusing on the aspects of strength summarizing the main findings and providing an outlook.
in FGLs [53,57] and the effective rather than nominal stiffness of FGLs.
Besides being partly limited by a step-wise rather than true linear 2. Methodology
gradient or a sole comparison between a graded and a uniform lattice,
studies have yet to provide an in-depth understanding about the effect The methodology of this work is comprised of the design-to-test
of e.g.: 1) Severity of grading, 2) Strut- vs surface-based lattices, 3) Print workflow illustrated in Fig. 1. The work is hereby limited to short-fibre
materials, 4) Manufacturing, etc. on the overall performance. Re- reinforced nylon as the print material and cubic tests specimens with an
garding the evaluation of the effective stiffness of FGLs, studies have edge length of 30 mm for both density and unit cell size graded lattices.
made use of the Voigt model [53,55,57], however the assumption of a The former set of specimens is composed of 6 × 6 × 6 unit cells,
continuous change in stiffness as a function of the gradient direction, as whereas the latter is made up of a constant initial 3 × 3 unit cell ar-
proposed in [41], is likely more accurate for truly linearly graded rangement at the bottom and increases smoothly into up to 9 × 9 unit
specimens. cells per layer.
Another increasingly investigated type of architected structures
utilize morphing approaches to realize a smooth transition between
boundaries of e.g. dissimilar unit cell sizes or unit cell types. Yang et al. 2.1. Functionally graded lattice structures
[58] have presented two methods to achieve the required hybridisation
i.e. morphing of TPMS lattices, by employing a Sigmoid function or a The approach for generating the density graded lattice structures is
Gaussian radial basis function in order to connect two or multiple cell based on the work of Panesar et al. [36], which allows a true func-
type regions, respectively. Similarly, Yoo and Kim [59] demonstrated tionally governed grading, and the method for the unit cell size grading
how such TPMS-based multi-morphology structures can be achieved was obtained from Yang et al. [58]. Both functionalities were im-
through the use of a volumetric distance field and beta growth func- plemented into the in-house software LatTess (Lattice Tessellation),
tions. By changing the morphology locally, these structures lend developed upon [33,36] by Panesar and co-workers, which served as a
themselves for precisely customizing the mechanical performance means of creating the FGLs in this work.
alongside e.g. permeability or heat conduction, paving the way for even In the case of the TPMS lattices, the geometric surface representa-
more advanced engineering solutions. Despite the tools, making the tion is hereby defined as an implicit trigonometric function
modelling of these lattices feasible, only little experimental investiga- f (x , y, z ) ≤ t (1)
tions were conducted yet [60].
For an increased application of FGLs in industry, databases will be whereby the isovalue t serves as a control parameter for the offset
required to build confidence in predicting their performance [61]. How from the zero level-set. The corresponding solid or double variant re-
to use, classify and apply architected structures has recently been presentation is expressed as

2
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 1. Overview of the methodology followed for investigating the mechanical performance of FGLs.

f 2 (x , y, z ) ≤ t 2 (2) (see Fig. 2) i.e. each lattice is defined by this maximum ρmax and
minimum ρmin local density.
in which the interval [−t, t] governs the density bounds for the
solid phase, which is determined by the space between the two mani-
folds of the network phase defined for −t < f > t. In this study, lattices 2.1.2. Unit cell size grading
composed of the strut-based Body Centred Cubic (BCC) unit cells and Structures with graded unit cell size i.e. a hybridization of unit cells
the surface-based Schwarz-P (SP) and Gyroid (GY) unit cells, are in- of different edge lengths have been generated based on the algorithm
vestigated. The equations for the latter two TPMS cells are as follows proposed in [58], using a sigmoid function with a constant transition
width across all specimens, ensuring a smooth morphing/merging be-
fSP (x , y, z ) = cos(λ x x ) + cos(λ y y ) + cos(λ z z ) (3) tween dissimilar cell counts. For this purpose the edge length of the GY
and SP lattices were divided into three segments of 10 mm each, con-
fGY (x , y, z ) = cos(λ x x )*sin(λ y y ) + cos(λ y y )*sin(λ z z ) stituting the design space for a set unit cell size (see Fig. 3). The severity
+ cos(λ z z )*sin(λ x x ) (4) of grading is hereby defined by the increase in the number of unit cells
from one segment to another. Hence, the expression for the gradient 3-
where λi = 2π * ni / Li governs the periodicity in three dimensions with 6-9 for the lattices shown in Fig. 3. It is of note that the average density
ni corresponding to the number of cell tessellations/repetitions along between layers of different unit cell size is kept constant.
the lattice edge length Li.

2.2. Design of experiments

2.1.1. Density grading
Grading of the unit cell density is conducted as in [36], where t
2.2.1. Test case matrix
controls variation of the volume fraction within 3D-space such that the
Initially, a set of ungraded SP and BCC lattices with average den-
condition
sities of 0.2, 0.35, 0.5, 0.65 and 0.8 were printed and tested parallel and
1, single variant (network phase ) transverse to the build-direction (defined as z-direction for the re-
f n (x , y, z ) ≤ t n (x , y, z ) n = ⎧ mainder of the work). This established baseline values to classify the
⎨
⎩ double variant (matrix phase )
2, (5)
performance of the density graded counterparts. Fig. 4 displays the
is fulfilled. The above mentioned approach is conceptionally coupons printed for investigations into the density and cell size grading.
equivalent for the strut-based counterparts, however, the actual im- Building up from the author's previous work [64], Table 1 sum-
plementation for generating those is relying on a database of multiple marizes the entire test case matrix of this work. Five different severities
unit cells with different uniform density gradients. In this study both SP of density grading ( ρΔ ) with the same relative density (i.e. average
and BCC lattices were linearly graded between two parallel surfaces density ρavg ) as the ungraded counterpart were chosen, printed in both

Fig. 2. Isometric and front view of a (a) SP and (b) BCC lattice with a linear density gradient in the z-direction.

3
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 3. Isometric and front view of a (a) GY and (b) SP lattice with a unit cell size gradient in the z-direction, realized through a cell size hybridization at two
equidistant positions along the edge length. The unit cell count per edge length in the three segments changes from 3 to 6 to 9, hence the nomenclature convention 3-
6-9.

z- and x-direction (parallel and transverse to the build direction) to approximately 0.5 mm was applied on the surfaces of the specimens.
investigate the effect of printing direction on the mechanical perfor- Individual strain measurements in the density graded specimens were
mance. On the other hand, three different unit cell size gradients (UCΔ ) taken at the centre of the front-facing cubic specimens and 2D strain
were investigated for the SP and GY lattices while keeping ρavg = 0.5. In maps were used to showcase the overall distribution in the cell size
all test cases, the loading direction aligns with the direction of the graded lattices. Another important issue to mention is the edge effect in
gradient. cellular solids stemming from the unit cell count per edge length.
Previous studies [66,67] have conducted convergence studies on cubic
2.2.2. Manufacturing and quasi-static experimental testing TPMS lattices, determining the number of unit cells required per edge
The specimens were printed with the Markforged Inc. MarkTwo fused length such that the upper bound for e.g. the modulus is matched. The
deposition modelling (FDM) machine, using short carbon fibre-re- error for a 3 × 3 × 3 and a 4 × 4 × 4 lattice was found to be in the
inforced nylon (material properties as obtained from the manufacturer realm of 1 % and 0.2 %, respectively [66,67], substantiating negligible
[65]). The stl. files were sliced using the Markforged’s Eiger software, influence even for the cell size graded lattices in this study, composed of
whereby the specimens were intended to be self-supported. Two roof-/ 9 unit cells in the bottom layer.
floor- and wall-layers were selected, and the remaining volume was
printed with a ± 45° infill. 2.3. Assessment of moduli and energy absorption capability
As the lattices are modelled in a voxel environment, there is always
a small error between the target and numerically achievable value for The elastic modulus of the five ungraded lattices of dissimilar
the density. The subsequent slicing of the model i.e. the transformation density was determined from the initial linear slope in the nominal
into a gcode that defines - among others - the infill pattern, can lead to stress-strain curve i.e. the stress was calculated from the load on the
further deviations. This is the case with most slicers, as the bead width nominal surface area of the lattice (see Fig. 5). If not expressly specified,
is not changed in-situ, using an adaptive extrusion value to accom- the remainder of this work will report on nominal values i.e. capture
modate for passages that do not match a multiple of the bead width. the global lattice response and average properties, neglecting the local
Finally, the layer-by-layer extrusion process is naturally associated with material behaviour. Furthermore, it is of note, that the stress and strain
producing inter-bead pores causing a discrepancy from design-to-print. distribution varies due to the shape of the specimen and the free edges.
The subsequent results were therefore normalized using the actual mass On the basis of the nominal moduli E nom of the ungraded lattices,
of the printed specimens and the material’s density. If not otherwise the standard scaling law (see Eq. (6)), as described by Gibson and Ashby
specified, the term “density” refers to the theoretical value in the re- [68], was employed to relate the experimentally determined relative
nom
mainder of this work. moduli Erel (nominal moduli E nom normalized by the bulk modulus of
It is important to note that the cubic shape of the specimens pre- the material ES ) to the relative lattice density ρrel . From this data of the
vents a true uniaxial stress state. In order to provide further insight into baseline lattices, the performance of the graded equivalent was derived.
the deformation behaviour of the FGLs, recordings with an optical As the relative density changes continuously along the z-direction in the
strain gauge were made for which a speckled pattern with a dot size of graded lattices, every single one of the n unit cell regions has a different

Fig. 4. Fabricated lattice specimens with unit cell (a) density grading and (b) cell size grading.

4
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Table 1
Test case matrix summarizing the number of specimens tested for each configuration with respect to the average density ρavg , the density and unit cell size gradient ( ρΔ
and UCΔ ) as well as the test direction.
Ungraded Graded

Avg. density: ρavg 0.2 0.35 0.5 0.65 0.8 0.5 0.5
Gradient: n.a. 0.6 0.6–0.35 0.7–0.3 0.7–0.25 0.8–0.2 3-4-5 3-5-7 3-6-9
ρΔ & UCΔ –
0.4
No. of tests P 2 2 2 2 2 2 2 2 2 2 2 2 2
T 2 2 2 2 2 2 2 2 2 2 / / /

Note: “P” and “T” refer to the number of tests conducted parallel and transverse to the build direction.

average unit cell density and hence a disparate modulus, as illustrated integration along the gradient as proposed in [41]:
in Fig. 5. The curve fit for relative modulus can thus be expressed as
1 1 L 1
nom
Erel (z ) = C1 (ρ (z )rel )m (6) nom
Etot
=
L
∫0 eff
Erel (z )
dz
(8)
with C1 representing a constant governing the geometrical features
As the average cross-section area between the individual unit cell
and taking a typical value between 0.1–4.0, whereas the exponent m
regions of the ungraded lattice is equal, the nominal stress-strain curves
lies around 2 [68]. As the relative properties are considered, the stiff-
functioned as the basis for the determination of the energy absorption
ness is zero for ρrel = 0 and maximal for ρrel = 1, i.e. equal to the bulk
per unit volume. However, due to the continuous change of the cross-
modulus ES . Hence, these bounds have been considered in fitting the
section in the graded counterparts, the stress distribution is no longer
experimental data. With the change in density Δρrel , defined as absolute
max min uniform. Therefore, it is only adequate to plot the load-displacement
difference between the maximum ρrel and the minimum ρrel density at
curves from which the absorbed energy is directly computed.
the two parallel surfaces (see Fig. 5), the density can be expressed as a
As the fibre volume fraction of the filament used for printing is
function of z
relatively small (∼9 %), the Halpin-Tsai model [69] was used to predict
Δρrel (L − z ) the potential for this type of FGLs assuming the theoretically highest
min
ρ (z )rel = ρrel + achievable fibre volume fraction for FDM processes, which is estimated
L (7)
to be around 40 % [70]. The model for predicting the Young’s modulus
where L is the total edge length of the specimen. On this basis, the total of the composite is
nominal stiffness of the density graded lattices was determined through

Fig. 5. Schematics of a linearly graded cubic SP

lattice of edge length L composed of n different
unit cell regions of different average density.
(a) Isometric and cross-section view, high-
lighting the nominal (global) vs the effective
(local) surface area. (b) Frontal view with a
close-up of graded and ungraded single unit
cells of identical density, illustrating the con-
tinuity of the density gradient throughout in-
dividual unit cell regions.

5
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

⎧ η = ( Ef − 1)/( Ef − ξ ) density. Generally layered structured can be considered transversely

Ec 1 + ηξφf ⎪ Em Em isotropic and with the introduction of fibres the difference between in-
= with:
Em 1 − ηφf ⎨ l plane and out-of-plane properties becomes even greater. In FDM pro-
⎪ξ = 2 cesses high aspect-ratio short fibres are always aligned with the print/
⎩ d (9)
track direction due to the shear force in the nozzle [15,21]. Micro-
where the ratio between the modulus of the composite Ec and the graphs presented in [74–76], confirm that this is also the case for the
matrix Em is determined from the fibre volume fraction φf , the modulus short-fibre reinforced polymer under investigation (aspect ratio
of the fibre Ef as well as the length l and diameter d of the fibre. ∼14:1), using a nozzle with an internal diameter of ∼0.36 mm. For
Investigations into the energy absorption capability generally em- specimens tested transverse to the build direction, an increasing lattice
ploy nominal stress-strain data (if not otherwise specified) and utilize density ρavg results in a larger percentage of fibres and print tracks being
the densification onset strain εDO as a reference value, describing the aligned with the load-direction (segment length and continuity together
point after which the slope of the load-displacement is rapidly ap- with consistent double wall layer) and more effective design space
proaching the one of the initial linear-elastic regions, as the cell walls becomes occupied by the ± 45° infill (see Fig. 8). Analogously to the
start to coalesce, until the densification strain εD is reached. This strain behaviour described through the standard laminate composite theory -
is determined using the energy efficiency method [71], whereby the to which 3D printed specimens conform well [77,78] - it is assumed,
efficiency κ can be defined as that this is the reason for the better performing high-density SP lattice
1 ε 1 u tested transverse to the build direction.
κ (ε ) =
σ (ε )
∫0 σ (ε ) dε i. e. κ (u) =
P (u)
∫0 P (u) du
(10)
3.2. Graded lattices
as derived from either the stress-strain (σ-ε) or load-displacement
(P-u) curves, respectively. εDO is subsequently determined at the plateau 3.2.1. Lattices with graded unit cell density
i.e. Fig. 9 illustrates the individual deformation stages of the graded SP
dκ (ε ) dκ (u) and BCC lattices up to 60 % nominal strain, at which the specimens take
= 0 i. e. =0 up a trapezoidal shape. In comparison, the ungraded counterpart dis-
dε du (11)
plays no such tilt of the edges at this strain. As the Poisson’s ratio is
independent of the unit cell relative density the tilt in the graded lat-
3. Results tices is due to the dissimilar stiffness of the individual unit cell layers.
The associated load-displacement curves of the graded lattices are
3.1. Ungraded baseline lattices: Physical properties and mechanical displayed in Fig. 10, displaying distinct load-drops and plateaus for the
performance most severe density gradients between the linear elastic and the plateau
region. These represent the failure of lowest-density unit cell layers
Fig. 6 illustrates the nominal stress-strain curves, indicating gen- before transitioning into plateau strain region, which is only really
erally higher compressive moduli of the stretching-dominated SP lat- pronounced for the ungraded SP lattice, whereas the remaining speci-
tices than the bending-dominated BCC lattices. This is independent of mens illustrate a positive plateau strain rate. It is noteworthy that the
ρavg , which is in line with earlier findings [72], and the build direction. SP lattices tested parallel to the build direction show a distinct point of
However, the high-density BCC lattices tested transverse to the build intersection at around 10 mm displacement before which the most se-
direction, display a greater dispersion compared to the ones tested verely graded lattices take lower loads for a given displacement than
parallel to the build direction. the ungraded counterpart and after which this state is reversed. The SP
As a summary, Fig. 7 highlights two key parameters for the un- lattices tested transverse to the build direction display a similar trend,
graded lattices: 1) the relation between the theoretical and actual lat- without having such a unique point of inversion. The ungraded BCC
tice density and 2) the nominal compressive modulus of the ungraded lattice tested parallel to the build direction outperformed the graded
lattices with respect to cell type, average lattice density and build di- counterparts and vice versa for the specimens tested transverse to the
rection. It was found that the numerical density values are in close build direction. Generally, the load-displacement curves of identical
agreement with the targeted (i.e. theoretical) values, whereas the actual specimen types are very tight showing great consistency, only the
density is systematically lower than anticipated, owing to the pores specimens tested transverse to the build direction displayed greater
created during printing (see Fig. 7a). The experimental density value variability at large deformation. These findings provide a good guide as
was hereby determined by means of the print material density of 1.2 g/ to the usefulness of grading with respect to cell type, loading direction
cm3 [73] and the measured specimen mass. Likewise, the overall re- and envisaged deformation range and can, therefore, lend itself to infer
lative density of the graded BCC and SP lattices was calculated and improved AM-designs.
came to 0.42 ± 0.005 and 0.43 ± 0.013, respectively. The following cumulative energy absorption (see Fig. 11), derived
Fig. 7(b) highlights the nominal compressive modulus of the un- from the load-displacement curves, visualize the concrete performance
graded lattices with respect to cell type, average lattice density and of the lattices for a given lattice strain. Except the BCC lattices tested
build direction. Generally, at low ρavg up to 0.5, there is only little parallel to the build direction, the ungraded lattice underperforms for
improvements in stiffness, however, notable increases were observed greater lattice strain. The gradual increase in resistance in the graded
from ρavg = 0.5 to ρavg = 0.65. This growth is bigger for the SP lattice lattices enforces a greater displacement that needs to be covered before
which can most likely be attributed to the transition from open to the lattices have the same effective stiffness as their ungraded equiva-
closed unit cells, whereby the share of cell wall stretching to bending lent and could hence lend themselfs greatly for applications considering
increases in favour of the former, providing greater axial cell wall impacts.
stiffness [68]. With increasing cell wall thickness from ρavg = 0.65 to The SP lattices displayed a much tighter set of curves, particularly
ρavg = 0.8 this effect is reinforced. for the specimens tested parallel to the build direction, as compared to
Similarly, it was observed that the for a low average density up to the BCC lattices. Grading was found to have a more detrimental effect
ρavg = 0.5, the differences between the building directions were mar- on the energy absorption capability of the BCC than for the SP coun-
ginal and not coherent, however, the BCC lattices displayed a slightly terparts. The most severely graded SP lattice was found to outperform
higher stiffness for specimens of ρavg = 0.65 and ρavg = 0.8, tested the BCC equivalent throughout, and similarly, the graded SP lattices
transverse to the build direction. However, the most significant differ- were found to generally absorb more energy up to the densification
ence was observed in the SP specimens with the highest average onset of the respective ungraded lattice than the graded BCC

6
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 6. Nominal stress-strain curves of the uniform (a)/(c) BCC and (b)/(d) SP lattices of different density, tested (a)/(b) parallel and (c)/(d) transverse to the build
direction.

Fig. 7. (a) Relation between the theoretical, numerical and actual (i.e. experimentally determined) lattice density of the ungraded lattices. (b) Nominal compressive
Young’s moduli of the ungraded BCC and SP lattices as a function of the theoretical volume fraction and build direction (z-axis) including error bars capturing the
standard deviation.

7
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 8. Comparison of print slices of uniform SP lattices with different average density taken at identical positions (hight of specimen). High density lattices illustrate
a greater percentage of 0° paths (aligned with load direction) and ± 45° infills.

counterparts. However, the latter surpass the cumulative energy ab- 3.2.2. Lattices with graded unit cell size
sorption capability of the surface-based lattices beyond this strain. Fig. 12 displays the load-displacement and the deduced cumulative
Considering the cumulative energy absorption results for the tests energy absorption curves for the SP and GY lattices with three different
conducted transverse to the build direction, it becomes apparent, that severities of unit cell size grading. Generally, the SP lattices showcase
the BCC lattices perform slightly better than in parallel direction, less dispersion both between the individual tests and the different cell
whereas no significant difference was observed in the SP lattices. size configurations compared to the GY lattices. Moreover, the former
The coherent intersection points between the graded SP specimens displays a lower strain rate in the plastic plateau region with a sharp
tested parallel with the ungraded counterpart (see Figs. 10 & 11) reveal increase in load at densification, as opposed to the GY lattices which
a superior energy absorption capability of graded lattices compared to illustrate a higher strain rate after the linear-elastic region with a less
the ungraded counterpart for large deformations i.e. lattice strains distinct load increase beyond 15 mm displacement.
above 50 %. The average break-even strain εBE i.e. intersection point for The average Young’s moduli, as determined from the linear-elastic
the graded lattices with the ungraded lattice (see Table 2), with regard region in the corresponding nominal stress strain curve are displayed in
to the cumulative energy absorption, lies within the realm of the den- Fig. 13. Apart from generally higher moduli in the GY lattices compared
sification onset strain εDO of the uniform lattice (49.3 % ± 2.3) as to the SP lattices, no clear trend can be derived from these values.
shown in Table 2. However, due to the constant relative density through the thickness of
these specimens a significant influence of the severity of unit cell size

Fig. 9. Deformation stages of the most severely graded (0.8-0.2) BCC and SP lattices during compression for different nominal lattice strains up to 60 %. Each final
two images highlight the trapezoidal shape of the graded vs the unchanged shape of the uniform counterpart.

8
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 10. Load-displacement curves for the graded (a)/(c) BCC and (b)/(d) SP lattices compared to the ungraded equivalents with equal mass tested (a)/(b) parallel
and (c)/(d) transverse to the build direction.

grading on the Young’s modulus is not to be expected, as proven by intermediately graded lattices. The most severely graded specimen
Gibson and Ashby’s scaling law [68]. Reasons for variability could however, shows an interruption in the middle region (intermediate unit
possibly stem from manufacturing (e.g. wall thickness influences infill) cell size with lower strain values). The latter suggests that the local
and the change in the morphology of the hybridized zones, but further stiffness is higher for the cell size corresponding to 6 × 6 unit cell
investigations are required to shed light on those influencing para- segment as opposed to 3 × 3 and 9 × 9, however, the transverse strain
meters. highlights the opposite effect i.e. higher strains in the middle sections.
The average densification onset εDO strain was found to be 54.4 Hence, the lack of unit cell symmetry in GY can result in a change in
% ± 0.6 for the SP lattice and 55.7 % ± 1.0 for the GY lattice, re- parallel and transverse stiffness dependent on the unit cell size. This
spectively. Moreover, the SP lattices displayed a more severe load in- needs further investigations as this suggests that there is not a gradual
crease and a minor rise in the most severely graded lattice prior to εDO , convergence towards a stiffness value that is independent of the edge
indicating an intermediate densification of some sort. Generally, the SP effect and therefore requires attention when selecting cell size and type.
lattices displayed lower variability in the plastic-densification region On the basis of the deformed shapes, the unit cell size graded SP lattices
compared to the GY lattices but overall, the performances are com- demonstrate the same trapezoidal-like shape as the density graded
parable. specimens, whereas the GY lattices retain their shape similar to the
In Fig. 14 the 2D axial (εz ) and transverse (εy ) strain maps are illu- ungraded lattices.
strated for 10 % and 30 % lattice strain to better identify the differences
in the structural behaviour. For both the SP and GY lattices, the strain
4. Discussion
distributions are homogeneous across the front surface of the specimen
in both direction for a lattice strain of 10 %, indicating a more evenly
4.1. Ungraded baseline lattices: Densification onset and energy absorption
distributed load. However, thin sections of lower axial strain were
identified around the hybridized sections and/or the topmost layer of
As shown in Table 3, the densification onset strain εDO could only be
the most severely graded lattices. These sections become more pro-
determined up to ρavg = 0.5 and ρavg = 0.65 for the BCC and SP speci-
nounced at 30 % strain while the GY lattices demonstrate a strain
mens, respectively. Lattices with a higher average density have led to a
gradient towards the bottom layer (greatest unit cell size) for low and
convergence of the energy efficiency curve without providing a distinct

9
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 11. Cumulative energy absorption curves for the ungraded and graded (a)/(c) BCC and (b)/(d) SP lattices of equal relative density tested (a)/(b) parallel and (c)/
(d) transverse to the build direction.

Table 2 4.2. Graded lattices

Break-even strain εBE for the cumulative energy absorption between the graded
and ungraded lattice with dissimilar density gradient ρΔ . 4.2.1. Failure and structural response
ρΔ 0.6–0.4 0.65–0.35 0.7–0.3 0.75–0.25 0.8–0.2
As a nominal stress-strain curve would not accurately represent the
εBE [%] 51.3 ± 2.4 46.1 ± 2.3 49.3 ± 1.7 46.4 ± 1.1 48.6 ± 1.1 individual stresses in each unit cell layer of the graded lattice, Fig. 15
illustrates the local unit cell strain for the BCC and SP lattice graded
from 0.8 to 0.2. The initial displacement-range up to ∼0.5−1 mm
optimum. This implies, that the cell wall interactions are muted at high (linear-elastic region up to the onset of lattice yielding i.e. load pla-
volume fraction, resulting in a smooth transition into actual material teau), reveals a highly different strain distribution in the z-direction
densification. The BCC lattices showcased a drop in εDO with increasing (i.e. loading direction) with the BCC specimen taking a concave shape
ρavg for both sets of specimens, whereas no such trend can be derived whereas the SP specimen displays a convex distribution. It is of note,
from the SP lattices, which demonstrate a fairly constant εDO . Table 3 that the increase in strain at the unit cell layer with the highest density
also displays the corresponding cumulative energy absorption up to the (i.e. between 25 and 30 mm) are relicts caused by measurement in-
densification strain wεDO . It was observed, that a rise in average lattice accuracies. Hence, when the strain has already gone to zero in layers of
density results in increased energy absorption and that the transverse lower density it can be ignored for these unit cell regions. At dis-
build direction results in a lower absorption capability except for the placements up to around ∼3.5 mm, the vertical lattice strains in the
BCC specimen at ρavg = 0.2 . A significant increase i.e. a multiplication of BCC specimens go to zero earlier than in the SP specimen, implying a
the cumulative energy absorption by a factor of 2–3 was observed from less uniform load distribution in the bending dominated unit cell lattice.
ρavg = 0.5 to ρavg = 0.65 in the SP specimens. Overall, SP specimens This more sequential collapse in the BCC lattice, as indicated by the
outperform BCC counterparts independent of ρavg and build direction. higher overall unit cell strain values in the first third of the lattice
(0−10 mm), explains the lower stiffness compared to the SP lattice. At
displacements above ∼4 mm, the distributions between the two lattices
correspond better. The transverse unit cell strains up to ∼3.5 mm
showcase a similar trend as the parallel ones and match well with the

10
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 12. (a)/(b) Load-displacement and (c)/(d) cumulative energy absorption curves of unit cell graded (a)/(c) SP and (b)/(d) GY lattices.

observations of the deformation stages shown in Fig. 9.

Some deformed specimens are illustrated in Fig. 16, exemplifying
the crack formation with respect to cell type and build direction. It was
found that the SP specimens tested parallel to the build direction have
shown hardly any signs of cracks, whereas, the corresponding BCC
lattices displayed horizontal cracks along the unit cell interfaces. This
can be explained by x-shaped struts acting like a hinge when loaded and
hence promoting tensile stress concentrations at the corners, where the
geometry changes abruptly. With the interlaminar strength being lower
than the in-plane strength in layered composites fabricated with FDM,
commonly due to the inter-bead porosity and other manufacturing-re-
lated issues such as wetting, cracks are likely to develop along the build
plane as observed in the test specimens. Another factor - frequently
observed in composites - is crack-steering through the fibres i.e. cracks
can easily propagate in fibre direction rather than having to circumvent
them. However, it is important to note, that these are assumptions for
the cause of the crack formation, requiring further investigations for
absolute conclusiveness.
Similarly, vertical cracks were observed in the BCC lattices tested
transverse to the build direction (i.e. along the build plane). As the BCC
lattices are bending dominated the minimum fibre length gains im-
Fig. 13. Average nominal Young’s moduli for the unit cell size graded SP and portance, especially in conjunction with the ductile nylon matrix,
GY lattices, including individual standard deviation and overall unit cell type
however, significant growth in stiffness should not be expected as the
average as indicated by the horizontal bands.
interlaminar properties will be unaffected. The same applies to the
specimens tested transverse to the build direction and thus the stiffness

11
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 14. 2D strain maps in z- and x-direction at 10 % and 30 % lattice strain for the SP and GY unit cells with different severities of unit cell size grading. White arrows
and brackets highlight the extremes.

values are equal for a range of average lattice densities (see Fig. 7). in the circular openings of the specimens tested transverse to the build
Despite the stretching-dominated nature of the SP specimens, the direction where the material strength is only governed by the inter-
lattice will inevitably undergo local bending particularly when the cells plane bonding strength i.e. polymer matrix. In return, this explains that
are open. This has led to cracks beyond a lattice strain of 50 %, initiated cracks fail to appear in the lattices loaded parallel to the build direction

Table 3
Densification onset strain εDO and cumulative energy absorption up to the densification onset WεDO for BCC and SP lattices tested parallel and transverse to the build
direction.
Build direction BCC SP

εDO WεDO εDO WεDO εDO WεDO εDO WεDO

Parallel Transverse Parallel Transverse

ρavg = 0.2 51 ± 0.2 977 ± 64 52 ± 64 1531 ± 64 58 ± 64 2546 ± 64 46 ± 64 2236 ± 64

ρavg = 0.35 42 ± 0.1 3429 ± 64 44 ± 64 2744 ± 64 59 ± 64 4961 ± 64 57 ± 64 4607 ± 64
ρavg = 0.5 38 ± 0.2 4979 ± 64 39 ± 64 3597 ± 64 49 ± 64 5967 ± 64 49 ± 64 5597 ± 64
ρavg = 0.65 n.a. n.a. n.a. n.a. 45 ± 64 17153 ± 64 48 ± 64 13387 ± 64
ρavg = 0.8 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

12
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 15. Representation of the average unit cell strain distribution for different displacement ranges fitted with an n-1 order polynomial for n data points. Displayed
are both the strain in x- and z-direction for the BCC and SP specimen with a relative density gradient from 0.8-0.2 measured in the centre of the specimen surface.

because the layers and fibres are arranged transverse to the potential tested transverse to the build direction (see Fig. 10). Overall, this work
crack direction and therefore need to overcome a higher crack initiation shows smoother load-displacement curves compared to e.g. a previous
resistance. study by Maskery et al. [37] in which unreinforced FGLs with a piece-
Consequently, a greater variability, particularly for the BCC speci- wise variation of constant unit cell density ρavg , were tested. This can
mens, was observed in the load-displacement curves for specimens most likely be attributed to a true linear density gradient through the

Fig. 16. Crack formation (highlighted in red) in the density graded SP (left) and BCC (right) lattices (70 % to 30 %) tested parallel and transverse to the building
direction z, accompanied by schematic drawings.

13
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 17. Crack formation (highlighted in red) in unit cell size graded SP (left) and GY (right) lattices (3-5-7 unit cell count gradient).

thickness (i.e. density gradient within individual unit cells), avoiding 4.2.2. The effect of grading on the stiffness
brittle and catastrophic failure. By applying the scaling law between the nominal Young’s modulus
All the SP lattices with unit cell size gradient demonstrated hor- and the relative density of the ungraded baseline lattices, trendlines
izontal and vertical cracks in the front surface of the largest unit cells were generated for both tests conducted parallel and transverse to the
(see Fig. 17) and occasionally signs of small horizontal cracks in layers build direction, as shown in Fig. 18a). Based on the Eq.7 the total
of intermediate unit cell size. The horizontal cracks in the GY lattices stiffness of the density graded lattices was subsequently determined
were significantly smaller and occurred exclusively in the lowest layer (see Fig. 18b).
i.e. in the largest unit cells. The failure in the region of large unit cells is For the bending-dominated BCC lattices, it is interesting to note,
indeed in accordance with findings in [60], where authors referred to that moderate grading (Δρ = 0.6 − 0.4 ) yields a similar total stiffness
possible manufacturing-related causes such as inferior wetting in large compared to the ungraded lattices, whereas more severe grading
cells [79]. This could also explain the observed cracks in the bottom (Δρ = 0.8 − 0.3) gradually reduces the stiffness to ∼50 % of the un-
layer, as individual print paths have more time to solidify before an graded counterpart. Δρ = 0.65 − 0.35. On the other hand, the moder-
adjacent path is being printed and thus ensure worse wettability. Fur- ately graded stretching-dominated SP lattices display a significant two-
thermore, as opposed to [60], no shear band failures were observed in fold increase in stiffness and the performance for the most severely
this work, which is most likely owed to the plasticity of the nylon graded lattice settles at a similar or even higher value as the uniform
specimens. lattice for specimens tested parallel and transverse, respectively. The
Another aspect to consider is the edge effect, which is theoretically observed effect for moderate grading is in line with the findings illu-
higher in the bottom layer (i.e. lower amount of unit cells), possibly strated in Fig. 7, demonstrating a threefold increase in stiffness from the
constituting a lower local stiffness. As no significantly higher or con- lattice with ρavg = 0.5 to ρavg = 0.65, while the stiffness for the lattice
sistent difference in strain between the bottom layers and the remaining with ρavg = 0.35 is only marginally lower than that with ρavg = 0.5. For
two layers was observed (see Fig. 14) for up to 30 % lattice strain and in both the graded BCC and SP lattices, the specimens in transverse to the
light of the marginal error expected for a 3 × 3 × 3 lattice config- build direction outperform the once tested parallel to it. This stems
uration as reported in [66,67], the edge effect is assumed to be negli- from the higher stiffness values obtained from ρavg = 0.65 and ρavg = 0.8
gible. Thus, values obtained in the linear-elastic and major parts of the as illustrated in Fig. 7. As these values are derived from the fitted
plastic region are only marginally - if at all - influenced by this effect. In function, it is important to note that they constitute an approximation
fact, as the failure in the SP lattices occurred at high strains (> 50 %) of the total stiffness and should not be confused with the ‘initial’ stiff-
and followed the same principle as explained above for the density ness, which is still governed by the lowest relative density unit cells of
graded lattices, it seems plausible that a higher bending moment at the the graded lattice.
circular opening compared to the smaller unit cells, stemming from the The different effect of density grading on the bending- and
load distribution, could be a reason for the local crack development. stretching-dominated lattices is shedding light on the importance of
It is important to mention that the fibre content in the filament is establishing baseline values which clearly demonstrate the individual
very low (∼ 8 %), potentially supplying insufficient overlaps and performance of lattices at a given relative density. Therefore, lattices
therefore compromising bending stiffness. However, based on the with an even more modest grading between ungraded and 0.6–0.4
above findings, it is assumed that a higher fibre volume fraction will could potentially yield an even higher stiffness, providing scope for
only really improve the stiffness of the SP lattices. It is therefore con- further fine-tuning. Similarly, knowing to which level the stiffness de-
cluded, that the inter-plane bonding strength and the unit cell geometry grades in comparison to the ungraded counterpart for a given severity
are the key influencing variables for the lattice performance. It is also of of grading is crucial, as it allows e.g. for better tailoring of multi-ob-
note, that the fibre-reinforced nylon lattices did not experience cracks jective applications such as e.g. local permeability [81] while adhering
i.e. failure at the necks of the SP unit cells as found in the lattices to a required stiffness value.
manufactured from unreinforced nylon 12 [67]. Apart from delamina- From the initial slopes of the load-displacement curves as well as the
tions, experienced in specimens tested transverse to the build direction, general curve characteristics of the unit cell size graded lattices (see
no severe failures like shear band deformations, as recently reported for Fig. 12), it can be concluded, that the elastic and momentary stiffness,
density graded metal-based TPMS [80], were found in specimens tested respectively, are very similar. Thus, unit cell graded lattices can be
parallel to the build direction. Similar to the density graded lattices this tuned to e.g. a specific conductivity rate (possible application for heat
is assumed to stem from the inherent material plasticity of nylon sinks) by increasing the surface area per volume through control of the
compared to the more brittle metals and could also be influenced by the severity of the unit cell size gradient while preserving the stiffness of
introduction of fibres to the matrix. However, further investigations are the structure.
required to shed light on the influence of fibres on the structural re- Fig. 19a) illustrates how the ungraded and density graded lattices
sponse of these lattices. perform in comparison to the empirical data of Ashby [72]. It was
observed, that the data points of the ungraded BCC lattices lie slightly
underneath the trendline for an ideal bending-dominated behaviour

14
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 18. (a) Fitting of Gibson Ashby model to determine constants of proportionality between relative Young’s modulus (modulus of the lattice E divided by the bulk
modulus ES) and the relative lattice density. (b) The relative Young’s moduli of the density graded lattices with respect to their ungraded counterparts, as determined
from Eq.16.

while having a similar slope, whereas the ungraded SP lattices do not ungraded lattices investigated in this study were integrated into a
match the ideal stretching-dominated behaviour while showing a Gibson-Ashby plot, highlighting their performance with respect to other
slightly steeper trendline than ideally stretching-dominated lattices. engineering materials. Interestingly, the density graded lattices fall into
With higher relative density both lattice types indicate convergence the realm of foams and natural materials, whereas the unit cell size
towards the corresponding ideal behaviour. The graded lattices show- graded lattices behave similarly to elastomers. This insight can inform
case the potential for fine-tuning the relative stiffness of the lattices, designers and engineers about the suitability of either density or unit
providing a greater spectrum for potential applications. The perfor- cell graded lattices for their application from a mechanical standpoint
mance of the SP lattices can be enhanced such that they approximate an but also allows contrasting those with aspects of multi-functionality
ideally bending-dominated behaviour and the moderately graded BCC FGLs have on offer, aiding the decision-making process.
lattices provide scope for matching the stiffness of a uniform SP lattice Using the Halpin-Tsai model and the rule of mixture, the specific
of the same density. It is of note, that the differences between those two Young’s modulus was included in Fig. 19b) for a fibre volume fraction
lattice types are generally small for such a high relative density, but an of 40 % instead of ∼9 %, as determined in [74]. The fibre length and
even greater scope for fine-tuning of the lattice performance through diameter were assumed to be ∼7 μm and ∼100 μm , respectively
density grading is to be expected for lower relative density lattices. [74,75]. Based on the material data provided by the manufacturer of
As shown in Fig. 19b), the stiffness values of the graded and the filaments [65], it is calculated, that the composite tensile and

Fig. 19. a) Relative modulus-density curves of bending- and stretching-dominated lattices as redrawn from Ashby [72] (with permission from The Royal Society).
Included are the values of the ungraded as well as density (Δρ ) and unit cell size (ΔUC ) graded lattices investigated in this study. b) Gibson-Ashby plot, as taken and
modified from [82], illustrating the relation between Young’s modulus and density for a range of engineering materials and showcasing the nominal compressive
moduli of the ungraded and graded lattices investigated in this study as well as the Halpin-Tsai estimated for fibre-volume fraction of ∼40 %, indicated by the dotted
lines.

15
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 20. Efficiency curves for the graded (a)/(c) BCC and (b)/(d) SP lattices compared to the ungraded counterparts with equal mass tested (a)/(b) parallel and (c)/
(d) transverse to the build direction.

flexural stiffness increases almost eightfold and more than threefold, It is of note, that the graded SP lattices showcase an efficiency
respectively. The moduli where hence multiplied by this factor as- plateau from 25 % lattice strain onwards (see Fig. 20), making it po-
suming the same relative ratio between the modulus of the lattice and tentially very attractive for applications in which a constant material
the bulk modulus. This is certainly only an estimate, not considering behaviour is desired. It can generally be concluded, that a unique op-
microstructural or manufacturing aspects and should not be regarded as timum cannot be derived from the efficiency curves of graded lattices
absolute values but more as an indication. Through a joint optimisation and hence there is no clear densification onset strain. This is to be ex-
of the mesostructure and the intrinsic material properties a much wider pected, as each individual unit cell layer would be expected to have
spectrum of specific stiffnesses can eventually be achieved. Thus, the their own onset strain which is further explaining the load-drops and
data illustrated in Fig. 19 is intended to serve as an additional insight waviness of the curves. It is overall positive to note how grading results
into potential parameters which can be considered when optimising a in a constant or increasing efficiency curve, highlighting how these
design with mechanical and functional behaviour in mind. structures can outperform their ungraded counterparts at higher lattice
strains (recall Fig. 11) and their scope for tailoring the lattice design to
4.2.3. The effect of grading on the energy absorption capability match an envisaged load profile.
From the graphs in Fig. 20 it can be observed that the SP lattices As proposed by Gibson and Ashby [68], the normalized cumulative
generally have a greater energy absorption efficiency than the BCC energy absorption is plotted over the normalized peak stress (see
lattices with the same density gradient. In fact, for tests conducted Figs. 21 & 22 ), to better identify lattices which absorb the highest
parallel and transverse to the build direction, more severe grading re- energy at the lowest possible stress. The cumulative energy absorption
sults in an earlier deviation from the initial linear region (mainly dic- and peak stress were normalized by the materials Young’s modulus ES.
tated by the ungraded lattice) in the BCC lattices. However, this de- In Fig. 21, the values were displayed together with the envelope curves
tachment occurs more gradual in the BCC lattices tested transverse to derived from the five ungraded lattices and plotted as a second-order
the build direction as opposed to the ones tested parallel to it, sug- polynomial.
gesting better performance at even lower strains as confirmed in Typically, a significant increase in energy absorption with a small
Fig. 11. rise in stress is to be expected in the plateau region (denoted by “B” in

16
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 21. Normalized nominal energy absorption diagram of the graded (a)/(c) BCC and (b)/(d) SP-lattice tested (a)/(b) parallel and (c)/(d) transverse to the build
direction compared to the ungraded counterpart, showing the global response in the elastic realm (A), the plastic yielding corresponding to the plateau region in the
load-displacement curve (B) and stress threshold initiating the onset of densification (C). Note that the marked values for the yield and densification onset strain as
well as the envelope curve are taken form the ungraded lattice.

Fig. 22. Normalized effective energy absorption diagram of the most severely graded (a) BCC and (b) SP-lattice ( ρΔ = 0.8–0.2) tested parallel to the build direction
and compared to the ungraded counterpart. Note that these values are directly obtained from the unit cell effective stress-strain values recorded with an optical strain
gauge and that the plotted yield and densification onset points are derived from the ungraded lattice of the same density. The regions A, B and C represent the linear-
elastic realm, the plateau and the densification region, respectively.

17
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 21) of the ungraded lattices, which can be confirmed by the graphs. controlled debonding, etc.) that are pivotal for the superior toughness
In the graded lattices this effect occurs earlier as the layers with a lower or energy absorption [83], thus not affecting the performance with
theoretical unit cell density of 0.5 have surpassed the yield point al- change in grading severity.
ready. This has led to significantly better energy absorption capability, Overall, these findings can help to better harness the potential of
particularly in the BCC lattices, prior to the yield stress of the corre- FGLs for specific applications, depending on whether particular stress-
sponding ungraded equivalent of the same density. In case of the BCC, constraints (see Figs. 21–23) need to be considered or whether only the
lattices tested parallel to the build direction, the curves of the most cumulative energy absorption is of interest (recall Fig. 11).
severely graded lattice demonstrate a peak, matching the values of the In an attempt to find a semi-empirical formula to predict the cu-
ungraded envelope curve equivalent to a uniform lattice of lower mulative energy absorption behaviour of the lattice types under in-
average density. In turn, it was found that they fail to achieve the same vestigation for a given density gradient and lattice strain, the curves in
performance by the onset of densification of the ungraded counterpart. Fig. 11 were fitted by power-law expressions, providing a goodness-of-
The same effect, although much more muted, was recorded for the fit measure of R2 > 0.98. The average parameters of these initial fits are
graded SP lattices and the curves match the ungraded lattice very well plotted in Fig. 24 over the density gradient and were subsequently fitted
in the corresponding linear-elastic range (see Fig. 21). again to derive a trend. This facilitated the determination of the power-
It is important to mention, that the abovementioned nominal values law constant and exponent for describing the energy absorption beha-
must be treated with caution, as the peak stresses are systematically viour of the lattices with the same mass for various severities of
underestimated due to a constant change in the effective surface area grading. These consolidated in the following formula:
through the thickness of the specimen. Hence, Fig. 22 provides further BCC _ 0.5 2.20Δ ρ + 1.99
insight into differences between the normalized energy absorption Wparallel = 1.94Δρ−0.46*εLatt (12)
capability at a certain peak unit cell stress for individual unit cell layers
SP _ 0.5 1.71Δ ρ + 1.78
in the FGLs. For this purpose, the effective unit cell stress-strain values W parallel = 5.96Δρ−0.41*εLatt (13)
(stress as load over the effective unit cell area and strain from the op-
BCC _ 0.5 0.86Δ ρ + 2.21
tical strain gauge) of the most severely graded lattices were compared Wtransverse = 8.67Δρ−0.03*εLatt (14)
to the average effective values of the ungraded counterpart. It becomes
SP _ 0.5 1.36Δ ρ + 1.65
evident, that the graded BCC lattices demonstrate higher energy ab- Wtransverse = 21.4Δρ−0.22*εLatt (15)
sorption at lower stresses prior to yielding. Conversely, the SP lattices
It is however of note, that the BCC lattices tested transverse to the
demonstrate a good agreement with the ungraded counterpart up to
build direction imply very low accordance with the regression model
yielding with only small peaks prior to that. Hence, grading is only
and therefore should not be considered or needs at least be treated with
advantageous for BCC lattices if stresses are within the linear-elastic
caution. This stems from the greater variability in the experimental data
region of the corresponding ungraded lattice with the same relative
for this set of specimens as illustrated in Fig. 11 and supports the ob-
density.
servations in the deformed specimen that an increased number of cracks
At last Fig. 23 illustrates the normalized nominal energy absorption
have occurred along the print-plane (recall Fig. 16), as the transverse
of the unit cell graded lattices, showcasing no significant differences
strength of the material is solely governed by the inter-bead bonding
with respect to either the severity of grading nor the unit cell type. This
strength i.e. the polymer matrix. With the BCC lattice failing in bending
is in line with the abovementioned results on this family of FGLs. It can,
this effect is exacerbated, suggesting that the stretching-dominated unit
therefore, be concluded that for selecting the correct unit cell type and
cell types should be favoured to guarantee repeatable performance.
severity of grading in an AM-design, solely aspects of multi-function-
Moreover, this sheds light on the importance of optimal processing
ality (e.g. local permeability, etc.) need to be considered as the me-
conditions, allowing sufficient fusion between layers to minimize the
chanical performance is equal for lattices with the same relative den-
occurrence of weak spots aiding the propagation of cracks.
sity. It should be pointed out, that despite the nature-like appearance of
The cumulative energy absorption curves of the unit cell size graded
these cell-size graded lattices, resembling a hierarchical structure and
lattices were also fitted with a power-law expression, providing an
suggesting - among others - improved energy absorption, this is merely
equal goodness-of-fit measure as above. As shown in Fig. 12, no sig-
a copy and not a replicate of nature. In fact, the lattice is not strictly a
nificant differences can be observed between the differently graded
hierarchical configuration but presents only a meso-structure and has at
lattices, however, a greater variability between equal specimens was
best one second length-scale level (counting the fibre-reinforcement). In
observed. For this purpose, the average i.e. mean values are presented
addition, it was not actually engineered to replicate the underlying
in Table 4. Compared to the density graded lattices no distinct trend
mechanisms (e.g. crack deflection, bridging, interfacial hardening,
could be observed. It is however of note that SP lattices with the lowest

Fig. 23. Normalized nominal energy absorption diagram of the cell size graded (a) SP and (b) GY lattices.

18
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

Fig. 24. Average power law constant and exponent derived from a power-law fit for the cummulative energy absorption curves of the density graded lattices.
Trendlines of the constants help predict the energy absorption capability as a function of the density gradient, assuming an absorption following the power-law.

Table 4 lattice of same relative density. In contrast, even severely graded SP

Power-law constants of cumulative energy absorption over the lattice strain fit lattices do not fall below the uniform threshold.
of lattices with different severities of unit cell size gradient. Note C is the
constant and m is the exponent.
• Specimens tested transverse to the build direction generally show-
cased a higher variability in performance compared to the ones
m
W = C *εLatt CSP CGY mSP mGY tested parallel to the build direction. Moreover, it was found, that a
higher average lattice density yields greater stiffness values for lat-
3-4-5 173 68 1.67 1.95 tices tested transverse to the build direction. This is assumed to stem
3-5-7 94 45 1.82 2.04
from a higher percentage of fibres being orientated favourably, i.e.
3-6-9 83 54 1.84 2.03
in 0° and ± 45° to the loading direction, thus highlighting the in-
fluence of the infill and manufacturing.
and the GY with the highest grading severity performed slightly better • It was found that on average the GY lattices of the same relative
than the two remaining configurations. density and different severity of unit cell size grading are stiffer than
the SP counterparts. No clear trend between the stiffness and the
5. Conclusion severity of grading was observed for these types of lattices, leading
to assume no significant influence of unit cell size grading.
In this article, the effect of the severity of grading on the com- • A categorisation of specific performance of the lattices with respect
pressive stiffness, energy absorption and structural response was in- to the corresponding ideally stretching- and bending-dominated
vestigated for additively manufactured short fibre-reinforced function- behaviour revealed certainly room for improvement but more im-
ally graded lattices. The work features both linear unit cell density and portantly highlighted the effect of grading and its potential for fine-
unit cell size grading of lattices with the same relative density, com- tuning the relative modulus for a given density. In comparison to
posed of Schwarz-P and body-centred-cubic as well as Schwarz-P and other common engineering materials, both the ungraded and den-
Gyroid unit cells, respectively. For the former, baseline values were sity graded lattices fall in the realm of foams and natural materials,
established from a range of uniform i.e. ungraded lattices, which were whereas the unit cell size graded lattices can be regarded as elas-
utilized to derive and evaluate the properties of the graded counter- tomeric. An estimate of the stiffness for 40 % fibre volume fraction
parts. Besides, all the uniform baseline and density graded lattices were (maximum theoretically achievable with extrusion-based AM) using
tested both parallel and transverse to the build direction to provide the Halpin-Tsai model, predicts a three- to eightfold increase.
further insight into manufacturing-related aspects of the performance. • The work has highlighted, that a more severe grading has a detri-
In conclusion, mechanical properties and behaviours of the lattices mental effect on the cumulative energy absorption of BCC lattices,
were summarized and embedded into Gibson-Ashby plots providing a whereas, on the contrary, severely graded SP lattices should be fa-
categorization of the FGLs. Furthermore, trends with respect to the voured for large deformations, beyond the densification onset strain
energy absorption capability for a given density gradient were derived of the corresponding ungraded lattice. However, with respect to the
and stiffness estimations with respect to a higher fibre volume fraction normalized energy absorption for a certain peak stress, it was found,
centred on the Halpin-Tsai model were provided. In conclusion, the that the density graded BCC lattices outperform the ungraded
major findings include: counterpart up to their respective yield point, while an equal per-
formance was observed for the density graded and ungraded SP
• Moderate grading of the density results in a significantly improved lattice.
total stiffness of the SP lattices compared to the ungraded counter- • The unit cell size graded lattices (SP and GY) showcased almost
part, whereas the BCC lattices yield a similar modulus as the uni- identical cumulative and normalized energy absorption curves, de-
form equivalent. In turn, more severe grading results in a reduction monstrating no effect of unit cell type and severity of grading of
in stiffness. This decrease is considerable in the BCC lattice, resulting lattices with same relative density.
in a modulus that is only a fraction of the corresponding uniform • Regarding the failure of the lattices, it was found, that cracks oc-
curred along the print plane, with more severe delaminations for

19
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

specimens tested transverse to the build direction. Hence, the col- 10.1016/j.compositesb.2016.11.034.
lected data displays a greater variability for this set of specimens. [10] Z. Quan, A. Wu, M. Keefe, et al., Additive manufacturing of multi-directional pre-
forms for composites: opportunities and challenges, Mater. Today 18 (2015)
However, the SP lattices tested parallel to the build direction did not 503–512, https://doi.org/10.1016/j.mattod.2015.05.001.
show any visible cracks owing to their stretch-dominated nature. [11] P. Parandoush, D. Lin, A review on additive manufacturing of polymer-fiber com-
The unit cell size graded lattices exhibited almost exclusively cracks posites, Compos. Struct. 182 (2017) 36–53, https://doi.org/10.1016/j.compstruct.
2017.08.088.
in the region with the largest cell size. [12] B. Brenken, E. Barocio, A. Favaloro, et al., Fused filament fabrication of fiber-re-
• A semi-empirical formula for the cumulative energy absorption ca- inforced polymers: a review, Addit. Manuf. 21 (2018) 1–16, https://doi.org/10.
1016/j.addma.2018.01.002.
pacity as a function of the density gradient was provided.
[13] L.G. Blok, M.L. Longana, H. Yu, et al., An investigation into 3D printing of fibre
reinforced thermoplastic composites, Addit. Manuf. 22 (2018) 176–186, https://
This work shall not only inform better AM-designs but is also in- doi.org/10.1016/j.addma.2018.04.039.
tended to spark interest for future research on analytical models, cap- [14] T. Hofstätter, D.B. Pedersen, G. Tosello, et al., State-of-the-art of fiber-reinforced
polymers in additive manufacturing technologies, J. Reinf. Plast. Compos. 36
turing the material behaviour of FGLs, which would make a ground-up
(2017) 1061–1073, https://doi.org/10.1177/0731684417695648.
property prediction and hence design-choices more straight-forward. [15] H.L. Tekinalp, V. Kunc, G.M. Velez-Garcia, et al., Highly oriented carbon fiber-
Further studies into the structural response of both unreinforced and polymer composites via additive manufacturing, Compos. Sci. Technol. 105 (2014)
more significantly reinforced lattices (i.e. fibre volume fraction greater 144–150, https://doi.org/10.1016/j.compscitech.2014.10.009.
[16] F. Ning, W. Cong, J. Qiu, et al., Additive manufacturing of carbon fiber reinforced
than 20 %) in conjunction with a consideration of infill patterns and thermoplastic composites using fused deposition modeling, Compos. Part B Eng. 80
build direction would be a trajectory worthwhile exploring, to fully (2015) 369–378, https://doi.org/10.1016/j.compositesb.2015.06.013.
understand their influence on the performance. Overall, the presented [17] M. Ivey, G.W. Melenka, J.P. Carey, et al., Characterizing short-fiber-reinforced
composites produced using additive manufacturing, Adv. Manuf. Polym. Compos.
findings highlight the great potential for fibre-reinforced FGLs and help Sci. 3 (2017) 81–91, https://doi.org/10.1080/20550340.2017.1341125.
identify key parameters for fine-tuning their performance to better [18] J.R. Raney, B.G. Compton, J. Mueller, et al., Rotational 3D printing of damage-
harness the potential AM has on offer for functional lightweight tolerant composites with programmable mechanics, Proc. Natl. Acad. Sci. (2018)
1–6, https://doi.org/10.1073/pnas.1715157115.
structures. [19] B.G. Compton, J.A. Lewis, 3D-printing of lightweight cellular composites, Adv.
Mater. 26 (2014) 5930–5935, https://doi.org/10.1002/adma.201401804.
CRediT authorship contribution statement [20] Z. Quan, Z. Larimore, A. Wu, et al., Microstructural design and additive manu-
facturing and characterization of 3D orthogonal short carbon fiber/acrylonitrile-
butadiene-styrene preform and composite, Compos. Sci. Technol. 126 (2016)
János Plocher: Conceptualization, Methodology, Validation, 139–148, https://doi.org/10.1016/j.compscitech.2016.02.021.
Formal analysis, Investigation, Writing - original draft, Writing - review [21] J.P. Lewicki, J.N. Rodriguez, C. Zhu, et al., 3D-printing of meso-structurally ordered
carbon Fiber/Polymer composites with unprecedented orthotropic physical prop-
& editing. Ajit Panesar: Conceptualization, Methodology, Validation,
erties, Sci. Rep. 7 (2017) 1–14, https://doi.org/10.1038/srep43401.
Formal analysis, Writing - review & editing, Supervision. [22] M. Helou, S. Kara, Design, analysis and manufacturing of lattice structures: an
overview, Int. J. Comput. Integr. Manuf. 31 (2017) 243–261, https://doi.org/10.
Declaration of Competing Interest 1080/0951192X.2017.1407456.
[23] X. Wendy Gu, J.R. Greer, Ultra-strong architected Cu meso-lattices, Extrem. Mech.
Lett. 2 (2015) 7–14, https://doi.org/10.1016/j.eml.2015.01.006.
The authors declare that they have no known competing financial [24] M.C. Messner, Optimal lattice-structured materials, J. Mech. Phys. Solids 96 (2016)
162–183, https://doi.org/10.1016/j.jmps.2016.07.010.
interests or personal relationships that could have appeared to influ- [25] X. Wang, T.J. Lu, Optimized acoustic properties of cellular solids, J. Acoust. Soc.
ence the work reported in this paper. Am. 106 (1999) 756–765, https://doi.org/10.1121/1.427094.
[26] P. Göransson, Acoustic and vibrational damping in porous solids, Philos. Trans. R.
Soc. A Math. Phys. Eng. Sci. 364 (2006) 89–108, https://doi.org/10.1098/rsta.
Acknowledgements 2005.1688.
[27] T. Tancogne-Dejean, A.B. Spierings, D. Mohr, Additively-manufactured metallic
This work was supported and funded by the Department of micro-lattice materials for high specific energy absorption under static and dynamic
loading, Acta Mater. 116 (2016) 14–28, https://doi.org/10.1016/j.actamat.2016.
Aeronautics at Imperial College London and the Engineering and 05.054.
Physical Sciences Research Council (EPSRC) with the project reference [28] M. Vesenjak, L. Krstulović-Opara, Z. Ren, et al., Cell shape effect evaluation of
number 2091639. polyamide cellular structures, Polym. Test. 29 (2010) 991–994, https://doi.org/10.
1016/j.polymertesting.2010.09.001.
[29] C. Bonatti, D. Mohr, Large deformation response of additively-manufactured FCC
References metamaterials: from octet truss lattices towards continuous shell mesostructures,
Int. J. Plast. 92 (2017) 122–147, https://doi.org/10.1016/j.ijplas.2017.02.003.
[1] J. Plocher, A. Panesar, Review on design and structural optimisation in additive [30] P. Qiao, M. Yang, F. Bobaru, Impact mechanics and high-energy absorbing mate-
manufacturing: towards next-generation lightweight structures, Mater. Des. 183 rials: review, J. Aerosp. Eng. 21 (2008) 235–248, https://doi.org/10.1061/(ASCE)
(2019) 108164, , https://doi.org/10.1016/j.matdes.2019.108164. 0893-1321(2008)21.
[2] M. Afshar, A.P. Anaraki, H. Montazerian, et al., Additive manufacturing and me- [31] L. Cheng, X. Liang, E. Belski, et al., Natural frequency optimization of variable-
chanical characterization of graded porosity scaffolds designed based on triply density additive manufactured lattice structure: theory and experimental valida-
periodic minimal surface architectures, J. Mech. Behav. Biomed. Mater. 62 (2016) tion, J. Manuf. Sci. Eng. 140 (2018) 105002, , https://doi.org/10.1115/1.4040622.
481–494, https://doi.org/10.1016/j.jmbbm.2016.05.027. [32] E. Handler, A. Sterling, J. Pegues, et al., Design and process considerations for ef-
[3] C. Yan, L. Hao, A. Hussein, et al., Ti-6Al-4V triply periodic minimal surface struc- fective additive maufacturing of heat exchangers, Proc. 28th Annu. Int. Free. Fabr.
tures for bone implants fabricated via selective laser melting, J. Mech. Behav. Symp. Austin (Texas), 2017, pp. 2632–2640.
Biomed. Mater. 51 (2015) 61–73, https://doi.org/10.1016/j.jmbbm.2015.06.024. [33] A.O. Aremu, J. Brennan-Craddock, A. Panesar, et al., A voxel-based method of
[4] D. Mahmoud, M. Elbestawi, Lattice structures and functionally graded materials constructing and skinning conformal and functionally graded lattice structures
applications in additive manufacturing of orthopedic implants: a review, J. Manuf. suitable for additive manufacturing, Addit. Manuf. 13 (2017) 1–13, https://doi.org/
Mater. Process. 1 (2017) 13, https://doi.org/10.3390/jmmp1020013. 10.1016/j.addma.2016.10.006.
[5] L. Zhu, N. Li, P.R.N. Childs, Light-weighting in aerospace component and system [34] H. Wadley, T. Queheillalt, Thermal applications of cellular lattice structures, Mater.
design, Propul. Power Res. 7 (2018) 103–119, https://doi.org/10.1016/j.jppr.2018. Sci. Forum. 539–543 (2007) 242–247.
04.001. [35] L. Cheng, J. Liu, X. Liang, et al., Coupling lattice structure topology optimization
[6] J. Wu, W. Wang, X. Gao, Design and Optimization of Conforming Lattice Structures, with design-dependent feature evolution for additive manufactured heat conduc-
(2019), pp. 1–14 http://arxiv.org/abs/1905.02902. tion design, Comput. Methods Appl. Mech. Eng. 332 (2018) 408–439, https://doi.
[7] Z. Doubrovski, J.C. Verlinden, J.M.P. Geraedts, Optimal design for additive man- org/10.1016/j.cma.2017.12.024.
ufacturing: opportunities and challanges, Proc. ASME 2011 Int. Des. Eng. Tech. [36] A. Panesar, M. Abdi, D. Hickman, et al., Strategies for functionally graded lattice
Conf. Comput. Inf. Eng. Conf. IDETC/CIE, Washington, DC, USA, 2011, pp. 1–12. structures derived using topology optimisation for additive manufacturing, Addit.
[8] D. Bourell, J.P. Kruth, M. Leu, et al., Materials for additive manufacturing, CIRP Manuf. 19 (2018) 81–94, https://doi.org/10.1016/j.addma.2017.11.008.
Ann. Manuf. Technol. 66 (2017) 659–681, https://doi.org/10.1016/j.cirp.2017.05. [37] I. Maskery, A. Hussey, A. Panesar, et al., An investigation into reinforced and
009. functionally graded lattice structures, J. Cell. Plast. 53 (2017) 151–165, https://doi.
[9] X. Wang, M. Jiang, Z. Zhou, et al., 3D printing of polymer matrix composites: a org/10.1177/0021955X16639035.
review and prospective, Compos. Part B Eng. 110 (2017) 442–458, https://doi.org/ [38] Architected Materials, NTopology, (2019) (Accessed 19 November 2019), https://
ntopology.com/architected-materials/.

20
J. Plocher and A. Panesar Additive Manufacturing 33 (2020) 101171

[39] Altair OptiStruct® Revolutionizes Lattice Structures for 3D Printing, Altair, (2017) using volumetric distance field and beta growth function, Int. J. Precis. Eng. Manuf.
(Accessed 14 December 2017), http://www.altairhyperworks.com/newsdetail. 16 (2015) 2021–2032, https://doi.org/10.1007/s12541-015-0263-2.
aspx?news_id=11109&news_country=en-US. [60] F. Liu, Z. Mao, P. Zhang, et al., Functionally graded porous scaffolds in multiple
[40] Materialise 3-matic Lightweight Structures Module, Materialise, (2017) (Accessed patterns: new design method, physical and mechanical properties, Mater. Des. 160
14 December 2017), http://www.materialise.com/en/software/3-matic/modules/ (2018) 849–860, https://doi.org/10.1016/j.matdes.2018.09.053.
lightweight-structures-module. [61] Islam M. El-Galy, Bassiouny I. Saleh, Mahmoud H. Ahmed, Functionally graded
[41] I. Maskery, A.O. Aremu, L. Parry, et al., Effective design and simulation of surface- materials classifications and development trends from industrial point of view, SN
based lattice structures featuring volume fraction and cell type grading, Mater. Des. Appl. Sci. 1 (2019) 1378–1401, https://doi.org/10.1007/s42452-019-1413-4.
155 (2018) 220–232, https://doi.org/10.1016/j.matdes.2018.05.058. [62] D. Bhate, C. Penick, L. Ferry, et al., Classification and selection of cellular materials
[42] R. Gabbrielli, I.G. Turner, C.R. Bowen, Development of modelling methods for in mechanical design: engineering and biomimetic approaches, Designs 3 (2019)
materials to be used as bone substitutes, Key Eng. Mater. 361–363 (2009) 903–906, 19, https://doi.org/10.3390/designs3010019.
https://doi.org/10.4028/www.scientific.net/kem.361-363.903. [63] D. Bhate, Four questions in cellular material design, Materials (Basel) 12 (2019),
[43] C. Liu, Z. Du, W. Zhang, et al., Additive manufacturing-oriented design of graded https://doi.org/10.3390/ma12071060.
lattice structures through explicit topology optimization, J. Appl. Mech. 84 (2017) [64] J. Plocher, A. Panesar, Mechanical performance of additively manufactured fiber-
081008, , https://doi.org/10.1115/1.4036941. reinforced functionally graded lattices, JOM (2019), https://doi.org/10.1007/
[44] P. Terriault, V. Brailovski, Modeling and simulation of large, conformal, porosity- s11837-019-03961-3.
graded and lightweight lattice structures made by additive manufacturing, Finite [65] Composite 3D Printing, Markforged, Inc., 2017 (Accessed 14 December 2017),
Elem. Anal. Des. 138 (2018) 1–11, https://doi.org/10.1016/j.finel.2017.09.005. https://markforged.com/composites/.
[45] Y. Wang, L. Zhang, S. Daynes, et al., Design of graded lattice structure with opti- [66] I. Maskery, A.O. Aremu, L. Parry, et al., Effective design and simulation of surface-
mized mesostructures for additive manufacturing, Mater. Des. 142 (2018) 114–123, based lattice structures featuring volume fraction and cell type grading, Mater. Des.
https://doi.org/10.1016/j.matdes.2018.01.011. 155 (2018) 220–232, https://doi.org/10.1016/j.matdes.2018.05.058.
[46] S. Daynes, S. Feih, W.F. Lu, et al., Optimisation of functionally graded lattice [67] I. Maskery, L. Sturm, A.O. Aremu, et al., Insights into the mechanical properties of
structures using isostatic lines, Mater. Des. 127 (2017) 215–223, https://doi.org/ several triply periodic minimal surface lattice structures made by polymer additive
10.1016/j.matdes.2017.04.082. manufacturing, Polymer (Guildf) 152 (2018) 62–71, https://doi.org/10.1016/j.
[47] S.Y. Choy, C.N. Sun, K.F. Leong, et al., Compressive properties of functionally polymer.2017.11.049.
graded lattice structures manufactured by selective laser melting, Mater. Des. 131 [68] I. Gibson, M. Ashby, Cellular Solids: Structure and Properties - Second Edition, 2nd
(2017) 112–120, https://doi.org/10.1016/j.matdes.2017.06.006. ed., Cambridge University Press, Cambridge, 1997, https://doi.org/10.1017/
[48] D.S.J. Al-Saedi, S.H. Masood, M. Faizan-Ur-Rab, et al., Mechanical properties and CBO9781139878326.
energy absorption capability of functionally graded F2BCC lattice fabricated by [69] J.C. Halpin, J.L. Kardos, The Halpin-Tsai equations: a review, Polym. Eng. Sci. 16
SLM, Mater. Des. 144 (2018) 32–44, https://doi.org/10.1016/j.matdes.2018.01. (1976) 344–352.
059. [70] R. Matsuzaki, M. Ueda, M. Namiki, et al., Three-dimensional printing of continuous-
[49] M. Afshar, A. Pourkamali Anaraki, H. Montazerian, Compressive characteristics of fiber composites by in-nozzle impregnation, Sci. Rep. 6 (2016) 1–7, https://doi.org/
radially graded porosity scaffolds architectured with minimal surfaces, Mater. Sci. 10.1038/srep23058.
Eng. C 92 (2018) 254–267, https://doi.org/10.1016/j.msec.2018.06.051. [71] Q.M. Li, I. Magkiriadis, J.J. Harrigan, Compressive strain at the onset of densifi-
[50] S. Wang, J. Wang, Y. Xu, et al., Compressive behavior and energy absorption of cation of cellular solids, J. Cell. Plast. 42 (2006) 371–392, https://doi.org/10.1177/
polymeric lattice structures made by additive manufacturing, Front. Mech. Eng. 0021955X06063519.
(2019), https://doi.org/10.1007/s11465-019-0549-7. [72] M.F. Ashby, The properties of foams and lattices, Philos. Trans. R. Soc. A Math.
[51] I. Maskery, N.T. Aboulkhair, A.O. Aremu, et al., A mechanical property evaluation Phys. Eng. Sci. 364 (2006) 15–30, https://doi.org/10.1098/rsta.2005.1678.
of graded density Al-Si10-Mg lattice structures manufactured by selective laser [73] Markforged Material Datasheet - Composites, Markforged, Inc., 2018 (Accessed 14
melting, Mater. Sci. Eng. A 670 (2016) 264–274, https://doi.org/10.1016/j.msea. May 2019), https://static.markforged.com/markforged_composites_datasheet.pdf.
2016.06.013. [74] M.J. Sauer, Evaluation of the Mechanical Properties of 3D Printed Carbon Fiber
[52] D. Li, W. Liao, N. Dai, et al., Comparison of mechanical properties and energy Composites, (2018) https://openprairie.sdstate.edu/etd/2436.
absorption of sheet-based and strut-based gyroid cellular structures with graded [75] J. Naranjo-Lozada, H. Ahuett-Garza, P. Orta-Castañón, et al., Tensile properties and
densities, Materials (Basel) 12 (2019), https://doi.org/10.3390/ma12132183. failure behavior of chopped and continuous carbon fiber composites produced by
[53] C. Han, Y. Li, Q. Wang, et al., Continuous functionally graded porous titanium additive manufacturing, Addit. Manuf. 26 (2019) 227–241, https://doi.org/10.
scaffolds manufactured by selective laser melting for bone implants, J. Mech. 1016/j.addma.2018.12.020.
Behav. Biomed. Mater. 80 (2018) 119–127, https://doi.org/10.1016/j.jmbbm. [76] T. Isobe, T. Tanaka, T. Nomura, et al., Comparison of strength of 3D printing objects
2018.01.013. using short fiber and continuous long fiber, IOP Conf. Ser. Mater. Sci. Eng. 406
[54] J. Kadkhodapour, H. Montazerian, S. Raeisi, Investigating internal architecture (2018), https://doi.org/10.1088/1757-899X/406/1/012042.
effect in plastic deformation and failure for TPMS-based scaffolds using simulation [77] C. Casavola, A. Cazzato, V. Moramarco, et al., Orthotropic mechanical properties of
methods and experimental procedure, Mater. Sci. Eng. C 43 (2014) 587–597, fused deposition modelling parts described by classical laminate theory, Mater. Des.
https://doi.org/10.1016/j.msec.2014.07.047. 90 (2016) 453–458, https://doi.org/10.1016/j.matdes.2015.11.009.
[55] L. Yang, R. Mertens, M. Ferrucci, et al., Continuous graded Gyroid cellular struc- [78] M. Sugavaneswaran, G. Arumaikkannu, Additive Manufactured Multi-Material
tures fabricated by selective laser melting: design, manufacturing and mechanical Structure with Directional Specific Mechanical Properties Based Upon Classical
properties, Mater. Des. 162 (2019) 394–404, https://doi.org/10.1016/j.matdes. Lamination Theory, (2018), https://doi.org/10.1108/RPJ-06-2017-0118.
2018.12.007. [79] C. Yan, L. Hao, A. Hussein, et al., Evaluations of cellular lattice structures manu-
[56] H. Montazerian, M.G.A. Mohamed, M.M. Montazeri, et al., Permeability and me- factured using selective laser melting, Int. J. Mach. Tools Manuf. 62 (2012) 32–38,
chanical properties of gradient porous PDMS scaffolds fabricated by 3D-printed https://doi.org/10.1016/j.ijmachtools.2012.06.002.
sacrificial templates designed with minimal surfaces, Acta Biomater. 96 (2019) [80] O. Al-ketan, D. Lee, R. Rowshan, et al., Functionally graded and multi-morphology
149–160, https://doi.org/10.1016/j.actbio.2019.06.040. sheet TPMS, J. Mech. Behav. Biomed. Mater. (2019) 103520, https://doi.org/10.
[57] S. Li, S. Zhao, W. Hou, et al., Functionally graded Ti-6Al-4V meshes with high 1016/j.jmbbm.2019.103520.
strength and energy absorption, Adv. Eng. Mater. 18 (2016) 34–38, https://doi.org/ [81] J. Plocher, C. Lee, A. Panesar, Additive manufacturing of bone-inspired structural
10.1002/adem.201500086. power composites, 22nd Int. Conf. Compos. Mater. Melbourne, 2019.
[58] N. Yang, Z. Quan, D. Zhang, et al., Multi-morphology transition hybridization CAD [82] Material Property Charts, Granta Des. (2019) (Accessed 10 November 2019),
design of minimal surface porous structures for use in tissue engineering, CAD https://grantadesign.com/education/students/charts/.
Comput. Aided Des. 56 (2014) 11–21, https://doi.org/10.1016/j.cad.2014.06.006. [83] A.R. Studart, Towards high-performance bioinspired composites, Adv. Mater. 24
[59] D.J. Yoo, K.H. Kim, An advanced multi-morphology porous scaffold design method (2012) 5024–5044, https://doi.org/10.1002/adma.201201471.

21