International Journal of Solids and Structures 246-247 (2022) 111599

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

Uniaxial tensile behaviour of additively manufactured elastically isotropic

truss lattices made of 316L
K. Somlo ∗, S.S. Chauhan, C.F. Niordson, K. Poulios
Department of Mechanical Engineering, Technical University of Denmark, Lyngby, Denmark

ARTICLE INFO ABSTRACT

Keywords: Metal additive manufacturing facilitates the production of complex light-weight structures, such as lattice
Tensile behaviour structures, using a wide range of materials. In this work, elastically isotropic truss lattice structures are
DIC investigated, produced by laser powder bed fusion (LPBF) from austenitic 316L stainless steel. The lattice
FE modelling
specimens are used to explore the effect of different unit cell sizes, orientations and volume fractions on the
Truss lattice
mechanical behaviour.
Stainless steel
Quasi-static, uni-axial tensile experiments are carried out with 3D digital image correlation (DIC) measure-
ments. The experiments are validated by finite element (FE) simulations in order to unravel the governing
mechanism of the tensile strength and deformation behaviour for the tested specimens. Comparisons are
performed by means of force–displacement curves and strain distribution maps.
For all lattice types, the geometrical deviations due to manufacturing are quantified and their impact on the
mechanical properties is investigated using numerical models. The high resolution DIC measurements provide a
detailed insight into the failure mechanisms of the specimens, which are also captured numerically for specific
cases. The capabilities of experimental and computational methods are combined to reveal the uncertainties
in the mechanical properties of LPBF produced truss lattice structures.

1. Introduction Truss lattices are periodic, cellular 3D networks of hollow or solid

truss elements (Tancogne-Dejean and Mohr, 2018). The widely investi-
Additive manufacturing (AM) has become a production technology gated representatives of truss unit cells are the simple cubic (SC), body-
of great importance over the past decades, due to its unique ability to centred cubic (BCC), face-centred cubic (FCC) and octet structures,
produce complex, light-weight structures at small-scales (Maconachie or slight modifications of these (du Plessis et al., 2018). Such rela-
et al., 2019). Metal AM enables geometries that are not feasible with tively simple metamaterials are well-known for their elastic anisotropy
traditional manufacturing, including for example lattices or topology and it has been demonstrated that SC and BCC lattices have bend-
optimised porous structures (Flores et al., 2020). These components ing dominated failure modes. On the other hand, for FCC lattices,
offer high potential in functional engineering applications, primarily failure modes are stretching dominated, resulting in superior perfor-
in the aerospace, automobile and biomedical sectors. Nevertheless, mance (Tancogne-Dejean et al., 2016). Plate lattice structures have
the basic microstructure and mechanical properties of AM metals are been found to outperform truss lattice structures in terms of both
highly active research fields, due to the diversity and many influencing
stiffness and strength properties (Tancogne-Dejean et al., 2018; Bonatti
factors of this manufacturing technology (Herzog et al., 2016; Charmi
and Mohr, 2019). Despite clear conclusions of the mostly theoretical
et al., 2021; Funch et al., 2020). The specific AM technology, powder
studies, these fields are continuously revisited and their conclusions
material, scanning strategy, building parameters and direction all have
are refined. For example, Andersen et al. (2021) concluded that the
considerable impact on the microstructure, mechanical properties and
stiffness and strength competition between plate and truss lattices is
geometric accuracy of the printed parts (Liverani et al., 2017). At the
not settled yet, due to the dependence on the specific structural loading
same time, progress in AM has intensified theoretical investigations of
various so-called architected materials, such as truss and plate lattice conditions.
structures, tailored to obtain certain functionalities. This work focuses Besides the numerical and theoretical studies, experimental investi-
on elastically isotropic truss lattices produced by the laser powder bed gations of different lattice structures have also gained high attention.
fusion (LPBF) process with 316L stainless steel. Among the numerous metal AM technologies, especially the LPBF

∗ Corresponding author.
E-mail address: kinga.somlo@ntnu.no (K. Somlo).

https://doi.org/10.1016/j.ijsolstr.2022.111599
Received 3 December 2021; Received in revised form 25 February 2022; Accepted 22 March 2022
Available online 30 March 2022
0020-7683/© 2022 Published by Elsevier Ltd.
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

process has become popular, mainly due to its geometric accuracy (Ma-
conachie et al., 2019; Köhnen et al., 2018). A wide range of metal AM
powder materials can be used for producing lattices, but Ti-6Al-4V,
316L and Al-Si10-Mg are the primary choice of materials (Zhang et al.,
2019; Xu et al., 2020; Leary et al., 2016). Most of the experimental
studies have been limited to uniaxial compression, partly because these
tests can be carried out with the simplest specimen geometry. For
tensile testing, the unavoidably rigid gripping part constrains the lattice
cross-section, introducing boundary edge effects, and specimens are
more time consuming and expensive to build. Additionally, as also
proposed by Köhnen et al. (2018) tensile lattice specimens require a
more complicated graded porous structure, in order to avoid failure at
the boundary between the solid gripping block and the lattice (Hanzl
et al., 2019).
Detailed studies of deformation and failure behaviour of metal
lattice structures under tension are currently limited, and there are sig-
nificant uncertainties due to the high variation of unit cell geometries
and materials. Nevertheless, digital image correlation (DIC) and finite
element (FE) analyses enable deep insight into local deformations,
contributing to the understanding of the mechanical behaviour at the
meso and macro scales. In a comparative study for Ti-6Al-4V SC unit
cells by Drücker et al. (2021) a fairly good agreement was achieved
between 3D DIC, conventional FE models and homogenised models.
Zhang et al. (2019) applied similar methods for Ti-6Al-4V diamond
cells, but only a FE model, including manufacturing imperfections, such Fig. 1. Geometric models of (a) graded lattice specimen (b) SC-BCC and (c) SC-Octet
as strut diameter deviations, could bring experimental and numerical unit cells with 𝜌 = 20% over-all volume fraction.
results in close agreement.
Considering the real printed geometry accounting for manufacturing
errors, rather than the idealised designed geometry, is an essential 30 mm with a uniform lattice with the target volume fraction, and two
factor for evaluating the potential of AM produced lattices in real graded zones of 40 mm on each side, as illustrated in Fig. 1a. The
applications. The deviations between nominal and printed dimensions lattices in the graded zones have linearly increasing strut diameters
as well as defects, such as broken trusses, depend not only on the towards the solid gripping parts, with the final strut diameters being
manufacturing technology and material but also on the truss diameter, approximately twice as large as in the central region. The models have
unit cell size and orientation. Although lattice structures is a field of been created with the nTopology software (nTopology, 2022).
intense research focus, only a limited number of studies have quanti- In the present work, elastically isotropic lattices have been applied,
fied the impact of actual manufacturing imperfections. Similarly, few based on the work of Tancogne-Dejean and Mohr (2018). As reported
studies have dealt simultaneously with various unit cell geometries, by Tancogne-Dejean and Mohr (2018), elastic isotropy can be obtained
orientations and volume fractions. with an adequate combination of the commonly used elementary cubic
The present work investigates elastically isotropic truss lattice struc- cells. The hereby investigated compound lattices are based on unit cells
tures each composed of two different elastically anisotropic, elementary combining the SC lattice with BCC and Octet lattices, as shown in
unit cells (Tancogne-Dejean and Mohr, 2018). Specifically, SC-BCC and Fig. 1b and c. respectively. Note that any combination of these three
SC-Octet unit cells are considered, for which elastic isotropy is obtained elementary cells in a single compound lattice exhibits elastic isotropy
based on the specific relative volume fractions derived in the work under the following condition (Tancogne-Dejean and Mohr, 2018),
of Tancogne-Dejean and Mohr (2018). Furthermore, the effects of the 𝑐SC 𝑐Octet 𝑐BCC 𝜌
unit cell sizes, the orientations and the volume fractions are studied. + + = (1)
3 6 9 5
Quasi-static tensile experiments are carried out with high resolution where 𝑐SC , 𝑐BCC and 𝑐Octet are the volume fractions of the SC, BCC
3D DIC measurements, and the results obtained are validated by FE and Octet lattices, respectively, and 𝜌 is the macroscopic target volume
simulations. For the FE analyses, both unit cell models with periodic fraction such that the total volume fraction is
boundary conditions and full-scale models of the lattice specimens
geometries are considered. In addition, models of the ‘‘as-fabricated’’ 𝑐SC + 𝑐Octet + 𝑐BCC = 𝜌 (2)
geometries are employed to account for manufacturing inaccuracies Considering Eqs. (1) and (2), the combination of SC and BCC cells leads
and defects. Inaccuracies are accounted for by correcting the model to 𝑐BCC = 35 𝜌, while SC combined with Octet gives 𝑐Octet = 45 𝜌.
strut diameters for each individual strut orientation, using DIC cameras Besides the different unit cell types, further objectives are to ex-
for optical estimations of the actually printed strut diameters. amine different volume fractions, unit cell sizes and load orientations.
The chosen geometries aim at exploring a wide range of lattice features,
2. Experimental procedure while fulfilling the manufacturing constraints among which there is a
minimum truss diameter of 0.4 mm. SC-BCC lattice specimens were
2.1. Design of lattice tensile specimens printed with 10%, 20%, 30% volume fraction at the central part and
a 4.5 mm cell size (Fig. 2a–c), all of them satisfying the limit of the
Lattice tensile testing inherently requires more complex specimen recommended minimum printed diameter. SC-Octet specimens were
geometries with higher production cost than compression testing. Mas- possible to print at sufficiently good quality only for 20% and 30%
sive solid blocks are necessary for gripping. Furthermore, a gradual nominal volume fractions (Fig. 2d–e). In addition, several variants of
change of strut diameters is also essential to ensure close to uniaxial the SC-BCC lattice specimen with 20% volume fraction were produced,
loading conditions at the centre of the specimen and prevent fracture shown in Fig. 2f–i. These include the different orientations of [110]
at the boundaries between the gripping blocks and the lattices (Köhnen and [111] (Fig. 2h and i), in addition to different cell sizes of 3 mm
et al., 2018). Therefore, each specimen consists of a central region of and 6 mm (Fig. 2f and g).

2
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 2. LPBF 316L lattice structures with the corresponding unit cell models.

Table 1 powder material. Furthermore, a random, fine, speckle pattern was

Most relevant build parameters. painted on top of a white base layer for each lattice (see Fig. 3b), in
Speed [mm/s] Power [W] Hatch distance [mm] Layer height [mm] order to detect and track the surface by DIC.
700 235 0.12 0.06
3. Numerical modelling

In the present section the numerical modelling is outlined, which

2.2. Materials and manufacturing
has been performed to investigate the previously introduced lattice
types (Fig. 2). Firstly, the employed material model and its parameters
The lattice specimens were printed using the SLM280 LPBF system
are introduced, that are specific for LPBF 316L. Secondly, an estimation
and the AISI 316L powder material, supplied by SLM Solutions Group
of the actually printed strut diameters is presented. These refined
AG. The chemical composition of the powder was in the range specified dimensions form the basis of the models, noted as ‘‘as-fabricated’’,
by the supplier, with a mean diameter of 34 μm. Further relevant details which are used for studying the impact of manufacturing deviations.
can be found in the work of Funch et al. (2020). The scanning strategy Finally, the FEM models are described, including the boundary condi-
used parallel stripes with a 67° rotation between subsequent layers. tions, loads and further simulation environment settings. Note that both
Further relevant build parameters are summarised in Table 1. After homogenised unit cell simulations with periodic boundary conditions
printing, stress relief was performed at 550 °C for 2 h, to prevent the and simulations of full-scale lattice specimens are also performed.
specimens from warping upon removal from the build plate. The spec-
imens were subsequently glass-blasted. In total nine different lattice 3.1. Material model
structures were printed (see Fig. 2), and each of them was printed for
two repetitions of tensile testing, except the SC-BCC with 30% volume The applied material model parameters have been determined by
fraction (Fig. 2c), for which only one specimen was produced. uniaxial tensile testing of the bulk material. These experiments were
carried out in a previous work (Somlo et al., 2022), investigating
2.3. Tensile testing with DIC the anisotropy of LPBF 316L. Although the tensile test results have
indicated a few percentage of elastic anisotropy, this minor effect is
The tensile experiments were carried out on an MTS 312.21 100 kN neglected in the present work, and elastic isotropy is assumed with a
servohydraulic testing machine, in displacement control mode with a Young’s modulus of 𝐸 = 195 GPa, which is the average uniaxial stiffness
displacement rate of 0.05 mm s−1 , at room temperature. The specimens of bulk specimens printed in different orientations. Plastic anisotropy is
were clamped with an MTS 647 side-loading hydraulic machine, using accounted for by the Hill-48 yield criterion. The yield function is given
100 bar gripper actuator pressure. by
In order to monitor the deformation and failure behaviour of the
𝛷(𝝈) ≡ 𝜑Hill-48 (𝝈) − 𝜎𝑦 = 0 (3)
specimens, DIC measurements were performed using the 12 megapixel
Aramis 3D camera system supplied by GOM International AG, shown where 𝜎𝑦 is the yield stress and 𝜑Hill-48 (𝝈) is the equivalent Hill-48 stress
in Fig. 3. Additionally, the DIC measurements were also used to de- defined by
termine the actually printed strut dimensions of different lattices in √
( )2 ( )2 ( )2
the unloaded state. The 50 mm lenses provided a good compromise 𝜑Hill-48 (𝝈) = 2
𝐹 𝜎22 − 𝜎33 + 𝐺 𝜎33 − 𝜎11 + 𝐻 𝜎11 − 𝜎22 + 2𝐿𝜎23 2
+ 2𝑀𝜎31 2
+ 2𝑁𝜎12
between the image resolution and focal depth, facilitating not only 2D,
(4)
but 3D measurements. The image rate was set to 1 Hz and the data was
processed with the Aramis Professional Software. Prior to testing, the where 𝐹 , 𝐺, 𝐻, 𝐿, 𝑀 and 𝑁 are material parameters and 𝜎𝑖𝑗 are
specimens were subjected to an ultrasound bath to remove unmelted the components of the Cauchy stress tensor. Assuming that index 3

3
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 3. Uniaxial tensile experimental set up: (a) Gripped SC-BCC [111] specimen with DIC system, (b) High resolution image of the specimen obtained by DIC.

Table 2
Applied parameters of isotropic hardening law.
𝜎0 [MPa] R0 [MPa] R∞ [MPa] b
480 700 120 80

corresponds to the BD, transverse isotropy in the build plane dictates

𝐹 = 𝐺 and 𝐿 = 𝑀, resulting in four independent Hill-48 parameters.
In the work of Somlo et al. (2022), a polycrystal RVE model was
used in addition to experimental data for fitting all four parameters.
The present work makes only use of the experimental data reported
by Somlo et al. (2022), which showed that the yield stress in the
direction perpendicular to BD is approx. 6% higher than parallel to BD.
This anisotropy can be captured with 𝐹 = 𝐺 = 0.5 and 𝐻 = 0.44, while
the remaining Hill-48 parameters are kept at their standard value for
the von Mises criterion, i.e. 𝐿 = 𝑀 = 𝑁 = 1.5.
The hardening behaviour is modelled by the Voce-type isotropic
hardening law, defined by
( 𝑝𝑙
)
𝜎𝑦 = 𝜎0 + 𝑅0 𝜀̂ 𝑝𝑙 + 𝑅∞ 1 − 𝑒−𝑏𝜀̂ (5)

where 𝜎0 is the initial yield stress, 𝑏 is the exponential saturation

parameter, 𝜀̂ 𝑝𝑙 is the equivalent plastic strain, 𝑅0 and 𝑅∞ are the linear
and exponential coefficients, respectively.
The plasticity parameters, shown in Table 2, have been determined Fig. 4. Fitting of Voce-type model to experimental stress strain curve obtained parallel
by fitting Eq. (5) to an averaged experimental stress strain curve, to the BD.
obtained by uniaxial tensile testing of the bulk specimens along the
BD. The obtained fitting, illustrated in Fig. 4, has been carried out in
terms of engineering stress strain curves, thus the numerical true stress truss members. It has been observed that truss members with identical
from Eq. (5) had to be transformed into its engineering counterpart, orientation in a specific lattice structure have similar deviations from
accounting for a finite transverse strain. A compromise had to be made the nominal value. Hence, in addition to the ideal lattice models with
at low strains to be able to fit the entire range from 0 to 30% strain nominal dimensions, refined versions have been created, aiming to
appropriately, which is the range of strains observed during the lattice reproduce the actually printed diameters of the struts. While some
simulations. deviations from ideal geometries are easily explained, others are not.
Due to the large bridge distance for the largest 6 mm unit cell size too
3.2. As-fabricated models accounting for manufacturing deviations thin or even broken trusses were observed parallel to the build plane,
as shown in Fig. 5b. On the other hand, some of the repetitions of
Inspection of the printed specimens revealed a significant difference the SC-BCC lattice with [110] cell orientation exhibited an unexpected
between the targeted nominal and actually printed dimensions of the

4
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 5. Manufacturing defects of (a) SC-BCC [110] and (b) SC-BCC, 6 mm lattice specimens captured by DIC.

Fig. 6. FE mesh for SC-BCC 10% specimen: (a) 1/8th model with symmetry planes and (b) illustration of the mesh resolution for one unit cell at the junction of the symmetry
planes.

Fig. 7. Normalised value of printed strut diameters grouped by strut orientation.

variation in diameters between the SC struts oriented at +45◦ and those For each lattice specimen variant, an as-fabricated specimen model
oriented at −45◦ with respect to the BD, as shown in Fig. 5a. was created based on the measured strut diameters in the central
uniform region of one of the two printed specimen repetitions. In order

5
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 8. Anisotropic stress strain curves of the applied (a) material model and (b) SC-BCC 20% unit cell with periodic boundary condition.

Fig. 9. Effect of dimensional errors of the manufacturing on anisotropy and average Young’s modulus, E. Superscripts ‘min’ and ‘max’ refer to minimum and maximum homogenised
Young’s moduli among all possible loading directions. The subscripts ‘nominal’ and ‘as-fabricated’ refer to unit cells built with the theoretical or measured strut dimensions,
respectively.

to mitigate the uncertainties of the measurements and the manufactur- the literature, since specimen specific edge effects are excluded. The
ing, strut diameters in each characteristic orientation were estimated homogenisation method is computationally very efficient, but their re-
based on the average of 12 measurements. All measurements were sults are usually not easy to reproduce experimentally, as experimental
performed in the central region, where strut diameters were supposed specimens require a very large number of unit cells in order to avoid
to be constant. The measurements exhibited stochastic variations with boundary effects. This work presents unit cell simulation results only
standard deviations, depending on the specimen, in the range of 10– for the baseline SC-BCC cell (4.5 mm, 20%) with different orientations,
100 μm, but only average values were used for the modelling. Note that but further similar results can be found in the work of Tancogne-Dejean
the manufacturing deviations near the grips were not accounted for in and Mohr (2018).
the modelling, since their effect on the overall response was assumed to In the case of the lattice specimen models, depending on the unit cell
be negligible, due to the much higher volume fraction in these regions. orientation, a few symmetry conditions could be applied to decrease
computational time. For the lattices with [100] unit cell orientation,
3.3. FE modelling symmetry is maintained in all three Cartesian coordinate directions,
thus 1/8th models of these structures have been employed. However,
The FE simulations of the present work were performed within the for the lattices with [110] and [111] cell orientation half and full
commercial software ANSYS Workbench. All simulations account for models were required, respectively.
large deformations and an implicit solver was employed. The same Non-uniform meshes, with quadratic tetrahedral elements were gen-
setup was used both for unit cells with periodic boundary conditions erated for all lattice specimen models with the nTopology software,
and for full-scale lattice tensile specimens. For each lattice specimen and they were subsequently imported to ANSYS Workbench. A detailed
type, two versions of the model were created; one with the nominal mesh convergence study was performed for individual lattices to de-
strut diameters, and one using the actual printed strut diameters, termine a suitable mesh size, see Appendix A. The study showed that
estimated as described in Section 3.2. using a feature size of 0.25 mm in the nTopology mesh settings, results
The homogenised unit cell models are ideal for validating the in sufficiently fine meshes with a minimum element size of ca. 0.1 mm
targeted mechanical properties of the cells by comparing them with in the thinnest struts, as shown in Fig. 6. The generated meshes for the

6
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 10. Force–displacement curves of lattice models obtained by FE simulations with nominal and as-fabricated dimensions: (a) SC-BCC and SC-Octet lattices with different volume
fractions, (b) all SC-BCC lattices with 20% volume fraction. Results of SC-BCC lattice with 20% volume fraction and 4.5 mm cell size (green curves) are repeated in both plots.

Fig. 11. Force–displacement curves from tensile tests and FE simulations, using the models with as-fabricated truss diameters. The results are presented for different: (a) volume
fractions of SC-BCC 4.5 mm unit cells, (b) unit cell sizes of SC-BCC with 20% volume fraction, (c), unit cell orientations of SC-BCC, 20%, 4.5 mm and (d) volume fractions of
SC-Octet unit cells.

7
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 12. Equivalent total strain maps of (a) SC-Octet 20%, (b) SC-BCC [111], (c) SC-BCC [110] and (d) SC-BCC 6 mm lattice structures obtained by DIC and FE models at 5 mm
imposed displacement.

Fig. 13. Fracture surfaces of the following lattices: (a, b) SC-BCC 20%, (c, d) SC-BCC 3 mm, (e) SC-BCC 6 mm, (f, g) SC-BCC [110], (h, i) SC-Octet 20%.

full, half and 1/8th specimen models have approximately 4 million, 2 side of the specimen and fixed support on the other side. While for
million and 750 000 elements, respectively, and all of them are depicted the symmetric specimens, frictionless support was used at the planes of
in Appendix A. A full-model simulation takes ca. 23 h on 48 nodes symmetries.
using 416 GB RAM and for the half and the 1/8th models the required
resources are reduced significantly, ca. proportionally to the element 4. Results and discussion
number. Regarding the periodic unit cell models, the computational
time is not substantial, as they included two orders of magnitude less
This section presents and compares the experimental and different
elements than the lattice specimen models.
numerical results, primarily by means of force–displacement curves
The lattice specimen FE models, employed in this study, allowed and strain distribution maps. First, the measured strut diameter inac-
to compare simulation results with experiments, and thereby validate curacies are reported, then the homogenised unit cell simulations are
the FE modelling choices, such as material model and approximation presented, followed by the comparison of the specimen models using
of the lattice specimen geometry. The applied boundary conditions for the nominal and measured strut dimensions. The nominal models are
a full specimen model, include a uniaxial displacement loading on one used to compare with literature, while as-fabricated models, accounting

8
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. 14. Case study of SC-BCC 6 mm lattice: (a) Observed manufacturing defects, (b) Corresponding model with missing trusses and (c) Numerical and experimental
force–displacement curves with strain distribution maps.

for manufacturing inaccuracies, are compared with experiments. A Mohr (2018). Despite the more specific material model of the present
comparison between the two models allows to study the impact of work, Fig. 8b is very close to the respective result from Ref. Tancogne-
manufacturing imperfections. Finally, the failure behaviour of some Dejean and Mohr (2018). Therefore, any further differences in terms of
specific specimens is investigated in detail. plastic anisotropy observed in the tensile specimens can be attributed to
other factors, such as boundary effects and manufacturing inaccuracies,
4.1. Measured strut diameters than the material model itself.
Since the actually printed truss member diameters differ substan-
Using the optical method described in Section 3.2, strut diameters tially from the targeted nominal values (Fig. 7), the effect of these
were estimated in the central uniform region of the specimens. The deviations on the elastic properties has also been investigated numer-
average measured values for each specimen type and each charac- ically, and the results are shown in Fig. 9. These results are obtained
teristic strut orientation are reported in Fig. 7, normalised against with purely elastic unit cell models, using the computational homogeni-
the corresponding nominal value. In general, the printed struts were sation module of the nTopology software (nTopology, 2022; Nemat-
oversized in average, with only very few exceptions. The results from Nasser and Hori, 1993). The homogenisation results, after including
Fig. 7 provided the basis for creating the as-fabricated variants of the the measured strut diameters from Fig. 7, show that manufacturing
unit cell and lattice specimen models of the present work. inaccuracies lead to violation of the elastic isotropy. Interestingly, the
elastic isotropy of the SC-Octet cells is much less sensitive to the
4.2. Unit cell simulation results dimensional inaccuracies than SC-BCC cells, which stems from the
higher complexity and lower anisotropy of the Octet unit cell compared
to the BCC. However, for SC-Octet cells the high positive bias in all
The unit cell simulation results are presented in Fig. 8a by means of
diameters, seen in Fig. 7, leads to considerably higher Young’s moduli
engineering stress strain curves. Despite the more specific anisotropic
compared the targeted values.
plasticity model employed in this work, the results are in a very
good agreement with the literature (Tancogne-Dejean and Mohr, 2018)
employing isotropic material model. Elastic isotropy is fulfilled within
4.3. Numerical results with nominal and as-fabricated lattice specimen
few percent, as expected based on the results of Tancogne-Dejean and
models
Mohr (2018). The observed plastic behaviour also supports the results
of Tancogne-Dejean and Mohr (2018). Specifically, the [110] orienta-
tion exhibits a substantially higher yield strength than the [111] and In order to demonstrate the effect of manufacturing errors on the
[100] orientations, due to a larger numbers of struts being relatively mechanical properties at the component level, uniaxial tension simula-
aligned with the load. Note that these unit cell results serve the purpose tions were performed with lattice specimen models using the nominal
of studying the impact of the slightly more complex material model, and measured strut dimensions. The force–displacement curves, pre-
in terms of plastic anisotropy, compared to Ref. Tancogne-Dejean and sented in Fig. 10, show that for all cell types the models with the

9
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

measured strut dimensions are stronger than their nominal counter- lattices are compared in Fig. 12, obtained by DIC and FEM, and their
parts, although the differences vary substantially depending on the unit results exhibit excellent agreement. The illustrated lattices aim to show
cell. It should be noted that the numerical simulations did not include some representative cases at an advanced deformation state. Strain
any failure model, hence most of the curves do not predict failure, maps for all investigated lattice types at different deformation states are
except for few one that indicate some local necking. included in Appendix B. Note that the DIC images shown in Fig. 12b–
For most lattice types, the simulations with nominal and d utilise the 3D reconstruction capability of the DIC software. Hence,
as-fabricated models are in a reasonable agreement, i.e. their yield Fig. 12d shows even the deformation of some BCC struts with an out-
strength values are within 10% range. However, for the SC-Octet lattice of-plane inclination, which are in general difficult to capture. The inner
with 20% nominal volume fraction and the SC-BCC lattice with [110] struts in Fig. 12a–c are missing, since the DIC system could not detect
orientation, the differences are substantial between the nominal and those surfaces. As the deformation progresses, the amount of non-
as-fabricated models. These large deviations stem from the significantly detected surfaces, namely the number of unidentified struts increases,
larger measured strut dimensions compared to the nominal values. Note especially for the lattices with [110] and [111] unit cells orientations.
that high deviations between the force–displacement curves of lattices
Fig. 12c presents the same SC-BCC [110] lattice used for illustrating
with the same nominal volume fractions can be attributed to differences
the manufacturing errors in Fig. 5a. It is clearly visible that the differ-
in the effective cross section areas as well as boundary effects. For a
ence between the diameters of the diagonal SC struts at ±45◦ , reported
specimen with a finite number of unit cells, any struts at the external
in Fig. 7, leads to an asymmetric strain map both from DIC and FEM.
boundaries are not shared with any other cells, and they have therefore
Both strain maps in Fig. 12c indicate a shear band deformation that
an increased contribution to the strength, compared to interior struts.
results in the fracture pattern of this specimen shown in Fig. 13g. This
This means an increased effective cross section area depending on the
unexplained manufacturing error occurred only in one of the two prints
relative amount of struts lying at the external boundaries. This leads
of this specimen. As a consequence, the second replication of this test
to different effective cross sections for the same unit cell depending on
exhibited a less dominant shear deformation (see Fig. 13f).
the orientation and the total number of cells.
Fig. 13 shows the fracture pattern for some of the tested specimens,
4.4. Comparison between experimental and numerical tensile test results including repetitions of the same specimen types. The figure shows the
fracture patterns depend somewhat on the unit cell types. For instance
At this stage, the uniaxial tensile experiments comprising two test SC-BCC 3 mm specimens (Fig. 13c–d) show quite flat fracture surfaces,
repetitions are compared to FE simulations of the as-fabricated mod- while shear bands and zig-zag patterns are more typical for unit cells
els accounting for the printed geometry. Fig. 11 shows that the ex- with larger sizes and rotated orientations.
periments exhibit excellent repeatability with regard to strength and
A detailed study has been carried out to investigate the effect of
ductility. Furthermore, for most of the cases the simulations are in a
printing defects in individual transverse struts in the SC-BCC 6 mm
good agreement with the experimental tensile test results. Interestingly,
lattice shown in Fig. 14. Based on a visual investigation of the printed
the numerical results for both cell types with 30% volume fractions
specimen, two transverse struts of the lattice model have been removed,
underestimate the experimental results; despite the fact that the model
and additional FE analyses (Fig. 14a–b) were performed. The objectives
strut sizes were in general increased based on real measured strut
were to determine the impact of missing struts on strength and examine
diameters. This indicates that there are possibly other effects, such
potential shear band or zig-zag deformation patterns prior to failure.
as additional melted regions at the intersections of the struts, not
accounted for in the model that have a significant impact for these rela- As Fig. 14c shows, up to the yielding point the initial nominal
tively high volume fraction lattices. Comparing the force–displacement model fits the experimental results very well. This apparent agreement
curves of the experiments and the models, substantial overestimations is due to opposing effects of strengthening due to larger printed strut
by the FE models can only be observed for SC-BCC [110] and SC-Octet diameters and weakening due to broken trusses. Considering the force–
20% lattices, shown in Fig. 11c and d respectively. This difference displacement curve of the most complex model with measured strut
could stem from a possible overestimation of the actual strut diameter dimensions and removed struts, it also exhibits an outstanding match
due to the limitation of the visual measurement method described in with the experiment in terms of yielding and failure. Furthermore,
Section 3.2. at the maximum load, the few struts experiencing the highest strain
Regarding the numerically underestimated performance of SC-BCC localisation, indicate the imminent fracture path. Although the exact
[111] lattice, it should be noted that this specific specimen has very path depends on the assumed defects, this result provides a strong
few, highly loaded BCC struts (Fig. 12b), which leads to strength indication of a non-planar fracture pattern, as illustrated in Fig. 14c.
properties that are very sensitive to the estimated strut diameters.
Note that even though this case study is carried out for one specific
Besides, the measurement of the truss diameter especially for this lattice
printed specimen, these general manufacturing defects are of high
is difficult, due to the out-of-plane orientation of all truss members
importance for lattice structures in AM. Among a few other related
relative to the surface plane of the tensile specimen.
studies (Červinek et al., 2021; Zhang et al., 2019), the present work
Excellent numerical validations are obtained for SC-BCC lattices
justifies that the manufacturing inaccuracies should be accounted for,
with cell size of 3 mm and 4.5 mm, shown in Fig. 11b. Also de-
and it demonstrates a possible method for that.
cent agreements are achieved for SC-BCC 10% (Fig. 11a) and 6 mm
(Fig. 11b) lattices, and the overestimation of the simulation results for
the 10% lattice might be explained by some neglected defects, i.e. fully
or partially broken truss members.
The analyses presented lead to the conclusion that the unit cell
results from Section 4.2 obtained with the same strut dimensions, 5. Conclusion
provide realistic estimations of the effect of actual printing inaccuracies
on the homogenised mechanical properties.

4.5. Failure behaviour Elastically isotropic truss lattice structures produced by LPBF using
316L stainless steel were investigated both experimentally and numer-
The validity of the specimen models was also evaluated by com- ically. The uniaxial tensile behaviour of different unit cell types, sizes,
parison with DIC measurements. The strain distribution maps of four orientations and volume fractions were studied.

10
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

The applied material and geometric models were demonstrated and fraction or eliminating horizontally aligned struts, i.e. using [110] or
verified against the literature by means of homogenised unit cell simu- [111] orientations.
lations using periodic boundary conditions. The dimensional deviations
of the printed specimens were estimated by visual measurements, in
order to refine the nominal models. Based on these modelling concepts, Declaration of competing interest
FE analyses of the full-scale lattice specimens were performed for both
the nominal and the actual printed geometries.
The authors declare that they have no known competing finan-
The numerical and experimental results were compared in terms of
cial interests or personal relationships that could have appeared to
force–displacement curves and strain distribution maps, and they ex-
influence the work reported in this paper.
hibited a good match. The results obtained indicate to what extent the
proposed lattices can be produced with LPBF process, and they quantify
both the geometrical deviations and their impact on anisotropy and
strength. The main conclusions of this study are summarised as follows: Acknowledgements

• With the investigated unit cell sizes, SC-BCC and SC-Octet lattice The present research has been conducted in the framework of the
structures can be adequately printed by LPBF in 316L, with AM-LINE 4.0 project, funded by the Innovation Fund Denmark under
10% and 20% volume fractions, respectively. Nevertheless, a few the grant number 7076-00074B. All specimens were manufactured by
printing defects and significant dimensional inaccuracies occur. one of the project partners, the Danish Technological Institute.
• Strut diameter inaccuracies were estimated for the printed lat-
tices, and their effect on anisotropy, stiffness and yielding was
quantified by unit-cell simulations. SC-Octet lattices exhibited
Appendix A. Mesh convergence study and meshes for all lattice
the largest deviations, leading to considerably higher average
types
stiffness (due to larger volume fraction), but their elastic isotropy
was barely affected.
• FE models of entire lattice specimens, accounting for as-fabricated
See Figs. A.15 and A.16.
strut diameters, could satisfactorily reproduce the experimen-
tal tensile force–displacement curves and strain distributions ob-
tained with DIC. Accounting for asymmetric dimensional inac-
curacies and characteristic printing defects in the FE model al- Appendix B. Different deformation states for all lattice types
lowed to reproduce the corresponding experimentally observed
deformations.
See Fig. B.17.
This work provides guidance regarding which unit cells are manufac-
turable with an adequate quality by LPBF technology and which should
be further improved. For instance, the weaker response for the 4.5 mm Appendix C. Images of fractured lattices
unit cell at 10% volume fraction and the 6 mm unit cell at 20% volume
fraction in the experiment compared to the simulation, could be an
indication of a too large bridge distance. The respective printing defect See Figs. C.18, C.19 and C.20.
can be eliminated by reducing the unit cell size, increasing the volume

Fig. A.15. Mesh convergence study with 0.25 and 0.5 min feature size.

11
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. A.16. Meshed FE models for all lattice types with the employed 0.25 min feature size.

12
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. B.17. Equivalent total strain maps obtained by DIC at different displacements, (a) SC-BCC 10%, (b) SC-BCC 20%, (c) SC-BCC 30%, (d) SC-BCC 3 mm, (e) SC-BCC 6 mm, (f)
SC-BCC [111], (g) SC-BCC [110], (h) SC-Octet 20% and (i) SC-Octet 30%,.

13
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. C.18. Fracture surfaces of the SC-Octet lattices: (a, b) 20%, (c, d) 30%.

Fig. C.19. Fracture surfaces of the SC-BCC lattices: (a) 10%, (b) 20% (c) 6 mm, (d) 3 mm, (e) [110] and (f) [111].

14
K. Somlo et al. International Journal of Solids and Structures 246-247 (2022) 111599

Fig. C.20. Fracture surfaces of the SC-BCC lattices: (a) 10%, (b) 20%, (c) 20% (d) 6 mm, (e) 3 mm, (f) [110] and (g) [111].

Köhnen, P., Haase, C., Bültmann, J., Ziegler, S., Schleifenbaum, J.H., Bleck, W., 2018.
Mechanical properties and deformation behavior of additively manufactured lattice
structures of stainless steel. Mater. Des. 145, 205–217. http://dx.doi.org/10.1016/
References
j.matdes.2018.02.062.
Leary, M., Mazur, M., Elambasseril, J., McMillan, M., Chirent, T., Sun, Y., Qian, M.,
Andersen, M.N., Wang, F., Sigmund, O., 2021. On the competition for ultimately stiff Easton, M., Brandt, M., 2016. Selective laser melting (SLM) of AlSi12Mg lattice
and strong architected materials. Mater. Des. 198, 109356. http://dx.doi.org/10. structures. Mater. Des. 98, 344–357. http://dx.doi.org/10.1016/j.matdes.2016.02.
1016/j.matdes.2020.109356. 127.
Bonatti, C., Mohr, D., 2019. Mechanical performance of additively-manufactured Liverani, E., Toschi, S., Ceschini, L., Fortunato, A., 2017. Effect of selective laser
anisotropic and isotropic smooth shell-lattice materials: Simulations & experiments. melting (SLM) process parameters on microstructure and mechanical properties
J. Mech. Phys. Solids 122, 1–26. http://dx.doi.org/10.1016/j.jmps.2018.08.022. of 316L austenitic stainless steel. J. Mater Process. Technol. 249, 255–263. http:
Červinek, O., Werner, B., Koutný, D., Vaverka, O., Pantělejev, L., Paloušek, D., 2021. //dx.doi.org/10.1016/j.jmatprotec.2017.05.042.
Computational approaches of quasi-static compression loading of SS316L lattice Maconachie, T., Leary, M., Lozanovski, B., Zhang, X., Qian, M., Faruque, O., Brandt, M.,
structures made by selective laser melting. Materials 14 (9), http://dx.doi.org/10. 2019. SLM lattice structures: Properties, performance, applications and challenges.
3390/ma14092462. Mater. Des. 183, 108137. http://dx.doi.org/10.1016/j.matdes.2019.108137.
Charmi, A., Falkenberg, R., Ávila, L., Mohr, G., Sommer, K., Ulbricht, A., Sprengel, M., Nemat-Nasser, S., Hori, M., 1993. Micromechanics : Overall Properties of Heterogeneous
Saliwan Neumann, R., Skrotzki, B., Evans, A., 2021. Mechanical anisotropy of Materials, Vol. 37. North-Holland.
additively manufactured stainless steel 316L: An experimental and numerical study. 2022. nTopology. https://ntopology.com/, Accessed: 22-02-2022.
Mater. Sci. Eng. A 799, 140154. http://dx.doi.org/10.1016/j.msea.2020.140154. Somlo, K., Frodal, B., Funch, C., Poulios, K., Winther, G., Hopperstad, O., Børvik, T.,
Drücker, S., Schulze, M., Ipsen, H., Bandegani, L., Hoch, H., Kluge, M., Fiedler, B., Niordson, C., 2022. Anisotropic yield surfaces of additively manufactured metals
2021. Experimental and numerical mechanical characterization of additively man- simulated with crystal plasticity. Eur. J. Mech. A http://dx.doi.org/10.1016/j.
ufactured Ti6Al4V lattice structures considering progressive damage. Int. J. Mech. euromechsol.2022.104506.
Sci. 189, 105986. http://dx.doi.org/10.1016/j.ijmecsci.2020.105986. Tancogne-Dejean, T., Diamantopoulou, M., Gorji, M.B., Bonatti, C., Mohr, D., 2018.
du Plessis, A., Yadroitsava, I., Yadroitsev, I., le Roux, S., Blaine, D., 2018. Numerical 3D plate-lattices: An emerging class of low-density metamaterial exhibiting optimal
comparison of lattice unit cell designs for medical implants by additive manufactur- isotropic stiffness. Adv. Mater. 30 (45), 1803334. http://dx.doi.org/10.1002/adma.
ing. Virtual Phys. Prototyp. 13 (4), 266–281. http://dx.doi.org/10.1080/17452759. 201803334.
2018.1491713. Tancogne-Dejean, T., Mohr, D., 2018. Elastically-isotropic truss lattice materials of
Flores, I., Kretzschmar, N., Azman, A.H., Chekurov, S., Pedersen, D.B., Chaudhuri, A., reduced plastic anisotropy. Int. J. Solids Struct. 138, 24–39. http://dx.doi.org/10.
2020. Implications of lattice structures on economics and productivity of metal 1016/j.ijsolstr.2017.12.025.
powder bed fusion. Addit. Manuf. 31, 100947. http://dx.doi.org/10.1016/j.addma. Tancogne-Dejean, T., Spierings, A.B., Mohr, D., 2016. Additively-manufactured metallic
2019.100947. micro-lattice materials for high specific energy absorption under static and dynamic
Funch, C.V., Christiansen, T.L., Somers, M.A.J., 2020. Effect of edge print parameters loading. Acta Mater. 116, 14–28. http://dx.doi.org/10.1016/j.actamat.2016.05.054.
on microstructure and high temperature solution nitriding response of additively Xu, Y., Li, T., Cao, X., Tan, Y., Luo, P., 2020. Compressive properties of 316L stainless
manufactured austenitic stainless steel. Surf. Coat. Technol. 403, 126385. http: steel topology-optimized lattice structures fabricated by selective laser melting. Adv.
//dx.doi.org/10.1016/j.surfcoat.2020.126385. Energy Mater. http://dx.doi.org/10.1002/adem.202000957.
Hanzl, P., Zetková, I., Daňa, M., 2019. Uniaxial tensile load of lattice structures Zhang, X.-Y., et al., 2019. Topological design, permeability and mechanical behavior
produced by metal additive manufacturing. Manuf. Technol. 19 (2), 228–231. of additively manufactured functionally graded porous metallic biomaterials. Acta
http://dx.doi.org/10.21062/ujep/274.2019/a/1213-2489/mt/19/2/228. Biomater. 84, 437–452. http://dx.doi.org/10.1016/j.actbio.2018.12.013.
Herzog, D., Seyda, V., Wycisk, E., Emmelmann, C., 2016. Additive manufacturing of
metals. Acta Mater. 117, 371–392. http://dx.doi.org/10.1016/j.actamat.2016.07.
019.

15