Composite Structures 235 (2020) 111821

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Mechanical performance and fatigue life prediction of lattice structures: T

Parametric computational approach
⁎
Chenxi Penga, Phuong Tranb,a, , H. Nguyen-Xuanb, A.J.M. Ferreirac
a
Department of Civil & Infrastructure Engineering, RMIT University, VIC, Australia
b
CIRTECH Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Viet Nam
c
Departamento de Engenharia Mecanica, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, Porto 4200-465, Portugal

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

Keywords: Additive manufacturing (AM) highly complex lattice structures with exceptional engineering properties are of
Lattice structure special interests for a wide range of engineering applications including biomedical (implant, scaffold), auto-
Finite element analysis motive (shock-absorbing, load sensors) and civil engineering (protective layers). This research focuses on de-
Elastic modulus veloping a numerical framework to predict the mechanical and fatigue properties of lattice structures with
Yield strength
various relative densities and architectures. The relationship between geometry parameters and relative density
Fatigue strength
Additive Manufacturing
of four lattice structures is investigated. Finite Element Analysis (FEA) is employed to simulate uniaxial com-
pression test so that the elastic modulus and yield strength of lattices are evaluated. Parametric studies on
hundred designs for each lattice are conducted to obtain associated design maps. Among the lattice structures
investigated, the Simple Cubic (SC) unit cell shows the highest elastic modulus and yield strength at any relative
densities, while these properties of Body Centred Cubic (BCC) unit cell is the lowest. Numerical results also
reveal the nonuniform distributions of strain & plastic dissipation energy between layers, when multilayer lattice
structures are subjected to compression. The fatigue property of lattice structures is also predicted numerically
showing the dependence on both relative density and unit cell topology. The normalised fatigue behaviour of SC
and Simple Cubic Body Centred Cubic (SC-BCC) are independent of relative density, while Face Centred Cubic
(FCC) and BCC are sensitive to its relative density.

1. Introduction lattice structures have been suggested and investigated in the literature.
Similar to CMS, it is known that LS is suitable to be applied as light-
Cellular materials and structures (CMS) attract significant interest weight core materials in sandwich structure subjected to shear and
due to its low weight-to-thermomechanical performance, such as large compression loads. Research showed that lattice structure could im-
densification strains, high specific strengths, and extensive plastic de- prove compressive and shear strengths compared to traditional hon-
formation. Thus, they have been extensively analysed for energy ab- eycombs [13,14]. In particular, lattice structure could also be applied
sorption applications [1–10]. Moreover, their performance and beha- for energy absorption purpose for protection from impact. Schaedler,
viours could be optimised and adjusted through the choices of topology et al. [15] have found that lattice structure can provide a better re-
(e.g., foam, honeycomb, and lattice), cell size and build material. sponse to impulsive loads than that of conventional materials by quasi-
Calladine and English [11] described two different types of structures. static compression test. Imbalzano, et al. [16, 65] have investigated
Type I cellular structures present a nearly constant (if not increasing) lattice structure under extreme loads. It is found that the LS absorbed
strength under loading, while type II structures show a softening caused double the amount of impulsive energy via. plastic deformation com-
by the buckling of the cellular material and are more sensitive to in- pared with monolithic ones. Since lattice structure has high surface area
ertial effects. to volume ratio, it is also applied for thermal management [17,18].
Lattice structure (LS), a type of CMS, has truss-like structures with Additionally, a significant body of research has demonstrated that lat-
interconnected struts and nodes in three-dimensional (3D) space [12]. tice structure could potentially be used as biocompatible material for
By changing the geometry, lattice structures could be designed to meet design of implants.
specific functional requirements. Various engineering applications of Conventional powder sintering technique has been used to

⁎
Corresponding author at: Department of Civil & Infrastructure Engineering, RMIT University, VIC, Australia.
E-mail addresses: [email protected], [email protected] (P. Tran).

https://doi.org/10.1016/j.compstruct.2019.111821
Received 27 October 2019; Received in revised form 15 December 2019; Accepted 17 December 2019
Available online 19 December 2019
0263-8223/ © 2019 Published by Elsevier Ltd.
C. Peng, et al. Composite Structures 235 (2020) 111821

manufacture lattice structures for biomedical applications [19,20]. stretching governed, the Maxwell criteria is employed based on calcu-
However, they are extremely brittle and have an affinity to crack pro- lation Maxwell number:
pagation during low-stress cycles. Moreover, the geometry of unit cell is
M = s − 2n + 3(two − dimensional) (1)
difficult to modify and manipulate. With the maturity of metal 3D
printing technology such as Selective Laser Melting (SLM) and Electron M = s − 3n + 6(three − dimensional) (2)
Beam Melting (EBM), it is now possible to precisely print porous
structures allowing the development of complex geometries, with good where s is the number of struts and n is the number of nodes in each
uniformity and repeatability of the response [21–23], when compared lattice unit cell. When lattice structures with M < 0 (under-stiff) is
to conventional fabrication technique. At this stage, the aim is to design loaded, it will behave as bending-dominated structure. It is found that
lattice structures that meet the requirements of different applications. the elastic modulus and yield strength of the bending-dominated cel-
Researchers have employed different methods to model the mechanical lular material are inferior to stretching governed ones [40]. Meanwhile,
properties of lattice structure. One most direct way to investigate the it has low load-bearing capacity since the strut would be easier to fail
performance of lattice structure is through experiments. In last decade, when subjected to bending than tension or compression. If M equal to 0
various experiments such as tensile [24], compression [25], bending (just-stiff), the lattice structure behaves as stretching-dominated. The
[26], fatigue [27], impact [28] and dynamic [29] tests have been ap- individual strut will be loaded under tension or compression. When
plied to study the mechanical properties of lattice structures. Mean- M > 0 (over-stiff), the lattice structure is statically-indeterminate. It
while, numerical modelling is employed to study the mechanical will exhibit predominantly stretching-dominated behaviour [39]. The
properties of lattice structures by researchers [30–35]. Moreover, ap- increasing of Maxwell’s number, M, indicates an increase in the ratio
proaches based on machine learning are proposed to reduce the com- between the number of struts to that of joint, which will lead to higher
putational cost and run such numerical simulation efficiently [36–38]. cell rigidity. Thus, the Maxwell criteria could be employed to identify
However, most research focus on the mechanical performance of lattice cell topology to whether to maximize load-bearing capacity or max-
structure at certain relative densities, while a comprehensive database imize compliance.
is not available yet.
In this paper, we aim to develop a numerical framework to in- 2.2. Analytical analysis of lattice structures
vestigate the mechanical properties and fatigue life prediction of dif-
ferent lattice structures. Detailed unit cell designs and relationships There are two analytical models developed to analyze the me-
between their geometrical parameters and relative density will be dis- chanical properties of CMS. The Ashby-Gibson model is the most fa-
cussed in Section 2. The finite element model to simulate the uniaxial mous model in predicting the properties of cellular materials. Gibson
compression test is introduced in Section 3 followed by numerical re- and Ashby state that relative density (ρ∗ / ρS ) is the most important
sults. Then, the numerical investigation of fatigue performances of characteristic in cellular materials [41]. Although the cell shape, size,
different lattice structures will be implemented in Section 4 based on edge and face connectivity can affect the mechanical properties, they
results from previous section. The last section will summarize conclu- are neglected in this model, only considering ligament bending. Fig. 2
sions obtained by this research. presents the Ashby-Gibson model (a) and its deformation in bending (b)
and buckling (c).
Based on these assumptions, an equation to predict the relationship
2. Unit cell design
between elastic modulus and relative density is given by,
2
2.1. A brief classification of cellular structures E∗ ρ∗
= C1 ⎛⎜ ⎞⎟
ES ⎝ ρS ⎠ (3)
The cellular materials & structures (CMS) could be divided into two
types based on their behaviours under loading. The deformation be- where E ∗ is the elastic modulus of cellular material, ES is the elastic
haviours include stretching and bending, which affects the failure me- modulus of solid material, ρ∗ is the density of cellular material, ρS is the
chanism of unit cells. Fig. 1 illustrates these two types. The response density of solid material, C1 is a dimensionless constant and it is de-
type of a lattice structure is determined by the number of nodes and termined by experiment as: C1 ≈ 1.
struts [39]. To understand whether the lattice structure is bending or However, this model has been regarded as unrealistic and over-

Fig. 1. Schematically represented bending (a) and stretching (b, c) governed structures.

2
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 2. Schematic representation of the Ashby-Gibson model [41].

simplified due to omittance of inhomogeneous material distribution, are shown in Fig. 3a. For geometry parameters, the unit cell size is
torsion in the edges and effect of arbitrary loading directions [41–47]. It defined as the edge length, and the strut size is defined as strut cross-
has been often used to predict at the beginning of the bulking event in section diameter. The baseline value of the unit cell size and strut
the linear-elastic region. Zhu, et al. [47] has analysed the treatment of diameter is set to 4.0 mm and 1.5 mm respectively. As presented in
Kevin cells in high strains. They found that buckling occurs although all Fig. 3b, the modelling of unit cell is started with a three-dimensional
the ligaments are not aligned with the principle compression axes. In sketch, where the nodes are created and cell size is defined. Then, lines
their analysis, edge shortening caused by axial compression is ignored in the stretch are assigned with thickness and rendered to beams. Fi-
while the quadratic relation of bending compliance is reversed [42]. nally, lattice with multiple unit cells is assembled automatically by
Different from Ashby- Gibson model, bending, stretching and twisting linear patterning.
of ligaments are coved in it. Hence, a more accurate analysis is pro- The Maxwell number of these unit cells are shown in Table 1. It
posed by them. could be seen that SC, BCC and SC-BCC unit cells are theoretically
Zhu model is given by, under-stiff, while FCC is over-stiff according to Maxwell criteria.
−1
Next step, the effect of cell size and strut size on relative density is
2
E∗ 2 ρ∗ ⎛ ρ∗ ⎞ investigated, and relative density maps for proposed unit cell structures
= C ⎛⎜ ⎞⎟ ⎜1 + C2 ⎜⎛ ⎟⎞ ⎟
ES 3 2 ⎝ ρS ⎠
⎝
ρ
⎝ S ⎠⎠ (4) are constructed by measuring the geometrical files, as shown in Fig. 4.
The results show that the relative density of FCC unit cell is very sen-
8 2I
where C2 = while I and A is the moment of inertia and the second
A2
sitive to the change of cell size and strut size. For example, when fixing
moment of area, respectively. the cell size at 2.0 mm the increasing strut size from 0.5 mm to 1.0 mm,
the relative density of FCC doubles from 50% to nearly 100%. By
2.3. Unit cell and lattice structure design comparison, the relative density of SC, BCC and SC-BCC unit cell show
less sensitivity to changes in cell size and strut size. Especially, if the
In our research, four different unit cells, including Simple Cubic cell size is fixed at 2.0 mm and the strut size increases from 0.5 mm to
(SC), Body Centred Cubic (BCC), Simple Cubic Body Centred Cubic (SC- 1.0 mm, the relative density of SC unit cell increases from 10% to 30%.
BCC) and Face Centred Cubic (FCC) are investigated. Their geometries This is due to the fact that there are more struts in FCC unit cell than

Fig. 3. (a) Geometries of SC, BCC-1, FCC, SC-BCC lattice unit cells; (b) Modelling of the unit cell and assembling of lattice structure.

3
C. Peng, et al. Composite Structures 235 (2020) 111821

Table 1 Table 2
Maxwell number for the proposed unit cell. Properties of Ti-6Al-4V in FE model [50,51].
Cell type SC BCC SC-BCC FCC Properties Value

Strut, s 12 8 20 44 Density ρ (kg/m3) 4430

Joints, n 8 9 9 16 Elastic Modulus E(GPa) 113.8
Maxwell number, M −6 −13 −1 2 Poisson’s ratio ν 0.342
A (MPa) 1098
B (MPa) 1092
that in other ones. The number of struts in a unit cell clearly affects the n 0.93
M 1.1
sensitivity of relative density to changes in cell size and strut size. C 0.014
However, with the same range of cell size and struct size, FCC unit cell Referential strain rate ξ (s−1) 1
has the highest relative densities among all structures investigated. Tm(K) 1878
Relative density from 10% to nearly 100% (completely solid) could be Tt(K) 715
obtained by FCC unit cell, while other unit cells only cover relative
density from 10% up to 50%. n
σeq = [A + Bεeq ][1 + C lnε ∗̇ ][1 − T ∗m] (5)

3. Finite element analysis where σeq is equivalent stress, εeq is the equivalent plastic strain andε ∗̇ is
strain rate. A is the Johnson-Cook yield stress, and the strain hardening
3.1. Material properties behavior is determined by B and n.
The dimensionless T ∗m is given by,
This study will focus on lattice structures fabricated by Ti-6Al-4V. It T − Tr
is known as a material with high tensile strength which has many ap- T ∗m =
Tm − Tr (6)
plications such as aircraft, medical devices and structural components
[48]. In order to capture both elastic and plastic behaviours of material, where T is the experiment temperature, Tr is the ambient temperature
the Johnson-Cook strength model [49], which is a numerical model and Tm is the material melting temperature. Overall, the properties of
based on various experiments, has been employed. Material constitutive Ti-6Al-4V for finite element analysis are collected from literature and
behavior is given by, summarized in Table 2.

Fig. 4. Relative density map for different types of lattice structures.

4
C. Peng, et al. Composite Structures 235 (2020) 111821

3.2. Description of the numerical model 4. Numerical results

The mechanical responses of lattice structures under uniaxial com- 4.1. Effect of the number of unit cells
pression are numerically investigated by finite element modelling im-
plemented in ABAQUS/Standard 2018 (Simulia, Providence, RI). In The differences between the behaviours of a single unit cell and that
order to simulate uniaxial compression, two rigid plates are created. A of lattices with multiple unit cells are evaluated. As presented in Fig. 6,
normal contact behaviour is defined between the specimen and rigid lattice structure with 1, 8, 27 SC unit cells are modelled and evaluated.
plates by hard contact formulation, while penalty friction formulation The stress–strain curves for different lattices are plotted in Fig. 6a. It
with a friction coefficient of 0.3 is applied to describe the tangential could be seen that the number of unit cells has a limited effect on the
behaviours. A uniaxial quasi-static displacement loading equivalent to behaviours of lattice structures within the elastic region. The elastic
engineering strain of 5% is applied on the top plate, while all the modulus calculated from 1 unit cell, 8 unit cells and 27 unit cells
translations and rotations of the bottom plate are fixed. The unit cell is models is 24.8 GPa, 25.0 GPa and 25.2 GPa, respectively, while asso-
meshed by second-order tetrahedral C3D10 element. A mesh con- ciated yield strength for 1 unit cell, 8 unit cells and 27 unit cells are
vergence study is conducted by increasing the number of elements 173.6 MPa, 175.3 MPa and 175.5 MPa, respectively. The stress–strain
(reducing the global seed size) on different unit cells. To ensure the curves of different lattice structures diverge after yielding points. The
results are independent of mesh density, the unit cell will mesh by more engineering stress on lattice structure with 27 unit cells reaches
than 4462 elements through the study. 265 MPa, while that on a single unit cell is 240 MPa. If we focus on the
distribution of engineering stress at ε = 0.005, ε = 0.01 and ε = 0.02 on
3.3. Post-processing of the results lattices, as shown in Fig. 6 (b-d), the stress distributions are identical at
ε = 0.005. Under compression, a high stress zones start to form on
The effective engineering stress and strain of lattice structure are vertical struts, where cells are connected to each other for lattices with
defined as, 8 and 27 unit cells.
Furthermore, changes in plastic dissipation energy per unit cell and
F
σ= strain energy per unit cell with respect to the increase in engineering
A0 (7)
strain are extracted and shown in Fig. 7a and Fig. 7b. As presented the
Δl Fig. 7a, plastically dissipated energy is observed for all structures at an
ε=
l0 (8) engineering strain of ε = 0.008, which indicates that they have close
yielding point. Similar trend is observed for strain energy per unit cell
where F is the reaction force measured on the bottom plate, and Δl is
(Fig. 7b), while results for different lattices diverge after yielding point.
the displacement captured on the upper plate. l 0 is the unit cell size and
Lattice with more unit cells results in lager plastic dissipation and strain
A0 is the cross-section area. Once the simulation finished, the dis-
energy per unit cell, which could be attributed to a high stress zone
placements of the upper plate and reaction forces on the lower plate are
formed on vertical struts.
extracted from the output database. Then, a stress–strain curve could be
Similarly, BCC lattices with 1, 8, 27 unit cells are evaluated in a
generated, as shown in Fig. 5b. The gradient between every two points
similar manner and presented in Fig. 8. It could be seen that lattice
on the stress–strain curve is calculated. A linear elastic region and strain
structures behave similarly within the elastic region. Then, the en-
hardening behaviour are observed on the stress–strain curve. The ef-
gineering stress–strain curves for different lattices diverge after yielding
fective elastic modulus (or simply elastic modulus, in short) of the
point. Compared to lattices with SC unit cells, the engineering stress
lattice structure is evaluated based on the linear region, while the ef-
drops with the increasing of the number of the unit cell. If we focus on
fective yield strength (or simply yield strength, in short) of the lattice
the distribution of engineering stress at ε = 0.005, ε = 0.01 and ε = 0.02
structure is captured at where the modulus changed more than 0.5%.

Fig. 5. (a) FE model to simulate the uniaxial compression test; (b) Typical stress–strain curve of SC unit cell with baseline parameters.

5
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 6. (a) Stress–strain curves obtained by different unit cell numbers; (b) Stress distribution on different lattices at ε = 0.005; (c) Stress distribution on different
lattices at ε = 0.02 ; (d) Stress distribution on different lattices at ε = 0.02 .

on lattices, as shown in Fig. 8 (b-d), the stress distribution on lattices dissipation energy per unit cell and strain energy per unit cell changes
are identical at ε = 0.005. Then, at ε = 0.01 and ε = 0.02 , large zones with the increase in engineering strain, as shown in Fig. 9 a&b, plastic
with nearly no stress on middle top of the lattice structures with 8 and dissipation energy starts to occur on all lattices at ε = 0.008 which in-
27 unit cells are observed. If we further investigate the plastic dicates that they have close yielding point. For strain energy per unit

Fig. 7. (a) Plastic dissipation energy per unit cell of lattices with 1, 8, 27 SC unit cells; (b) Strain energy per unit cell of lattices with 1, 8, 27 SC unit cells.

6
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 8. (a) Stress–strain curves obtained by different unit cell numbers; (b) Stress distribution on different lattices at ε = 0.005; (c) Stress distribution on different
lattices at ε = 0.02 ; (d) Stress distribution on different lattices at ε = 0.02 .

Fig. 9. (a) Plastic dissipation energy per unit cell of lattices with 1, 8, 27 BCC unit cells; (b) Strain energy per unit cell of lattices with 1, 8, 27 BCC unit cells.

cell, results for different lattices also diverge after yielding point. Lower higher than that of other layers. At ε = 0.05, the plastic dissipation
plastic dissipation energy and strain energy per unit cell has been ob- energy of first, second and third layer on the lattice with 27 SC unit cells
served for lattices with more unit cells, which could attribute to losing are 5587, 5665 and 5594 mJ, respectively. For strain energy of different
contact on the top surface of the lattice structure. layers, there is no difference between layers within the elastic region as
The energy absorbed by different layers on lattice structures with 27 shown in Fig. 10b. A similar trend has been observed after yielding,
SC and BCC unit cells is further investigated. As shown in Fig. 10a, the while the differences between strain energy of the second layer and
plastic dissipation energy of the second layer on SC lattice is slightly other layers are more significant. At ε = 0.05, the strain energy of first,

7
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 10. Plastic dissipation energy (a) and strain energy (b) at different strain levels for different layers of SC lattice consisted of 27 unit cells; Plastic dissipation (c)
and strain (d) energy at different strain levels for different layers of BCC lattice consisted of 27 unit cells.

second and third layer on lattice with 27 cube unit cells are 727, 754, and the SC-BCC unit cells. As for FCC and BCC unit cells, where vertical
727 mJ respectively. The plastic dissipation energy and strain energy struts are not presented, the highest engineering stress could be seen at
have significant non-uniform distributions between layers on BCC lat- the joint areas, where struts are connected.
tice as shown in Fig. 10 c&d. The plastic dissipation energy of the Further study is conducted to evaluate the effect of relative density
second layer is significantly higher than that of the first and third layer. on mechanical properties of unit cells. For all unit cell topologies, the
At ε = 0.05, the plastic dissipation energy of first, second and third layer cell size has been fixed at 4 mm, and the strut size is increased gradually
on BCC lattice are 9486, 17,716 and 9431 mJ respectively, and strain to adjust the relative density of lattice structures. The result of elastic
energy are 2326, 2529 and 2324 mJ. It could be concluded that, for modulus and yield strength of these structures with relative density
lattice structure with 27 unit cells, lager deformation is observed on the from 1% to 40% are presented in Fig. 12a and Fig. 12b respectively. It
middle layer compared to other layers. This could be attributed to the could be seen that for different unit cell structures, the elastic properties
fact that during the compression, the first and third layers lose partial are different for the same relative density. The elastic modulus in-
contact with the compression plate due to the lateral deformation of the creases with the increase in relative density as predicted by the ana-
lattice, while the middle layer is compressed by other layers con- lytical model. The SC unit cell has the highest elastic modulus at any
sistently. relative densities among unit cells investigated. By comparison, BCC
has a much lower elastic modulus compared to that of other unit cells.
SC-BCC and FCC show similar elastic modulus, especially at low relative
4.2. Effect of relative densities density. For example, at relative density of 6% the elastic modulus of
SC-BCC and FCC are 1.3 and 1.2 GPa, respectively. In order to analyse
Firstly, 4 different unit cell topologies at the same relative density of the dependence of elastic modulus of lattice structure on relative den-
0.45 have been investigated. As shown in Fig. 11a, SC unit cell shows sity systematically, Ashby-Gibson model is employed to fit the results.
the highest elastic modulus, while that of BCC unit cell is the lowest The result shows that SC unit cell has the calculated constant parameter
among all unit cells. It is also noted that SC unit cell is still within the C1 of 1.24. Value of C1 for SC-BCC and FCC are quite close to that
elastic region while yielding has occurred in other unit cells. In parti- proposed by Ashby-Gibson model, which is 1. BCC has relatively low C1
cular, the BCC unit cell yields first when compressed. SC-BCC and FCC compared to other unit cells investigated.
unit cells behave similarly in the elastic region but SC-BCC has higher Similar to elastic modulus (Fig. 12a), the SC unit cell has the highest
engineering stress in the plastic region. If we further investigate the yield strength (Fig. 12b) at any relative densities among unit cells in-
stress distribution on different unit cell topologies, as shown in Fig. 11b, vestigated, while BCC has a much lower yield strength compared to that
the highest engineering stress occurs on the vertical struts for the cube

8
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 11. (a) Stress–strain curves of different unit cell topologies with relative density of 0.45; (b) Stress distribution on different unit cell topologies at ε = 0.01.

of other unit cells. FCC unit cell has a lower yield strength than that of SC-BCC unit cell diverge with the increase of relative density. The
SC-BCC unit cell at low and high relative density (Fig. 12b). However, plastic behaviours of them are similar to that of SC unit cell.
the yield strength of the FCC and SC-BCC unit cells are very close at
relative density of 17% to 27%.
The engineering stress–strain curves of unit cells at different relative 4.3. Effect of geometrical imperfection
densities of 0.15, 0.25, 0.35 and 0.45 are normalised by its yield
strength, to analyse the deformation behaviour at different relative As discussed in the previous section, additive manufacturing process
densities as shown in Fig. 13. It could be seen that the elastic region of could introduce geometrical imperfections on the lattice structures.
SC and BCC unit cell are independent of relative density, which in- Such defects could also include porosity, semi-solidified metal particles,
dicates that the relationship between relative density and elastic mod- variations in strut’s thickness, residual stress and localised buckling due
ulus is similar to the relationship between relative density and yield to powder bed’s instabilities etc. These defects will affect the desired
strength. As stated by the Ashby-Gibson model, both of them are topology of the unit cell, leading to localised buckling under com-
quadratic functions. However, the plastic deformation behaviours of SC pression. Researchers have evaluated how the mechanical perfor-
and BCC unit cell are different. The normalised stress is increasing with mances of lattice structures would be affected by using uncertainty
the increase of relative density of SC unit cell, while that of BCC unit analysis [52,53] and imperfection modelling [16]. In this work, geo-
cell is decreasing. This could be associated with the low yield strength metrical imperfections are indirectly considered and incorporated in
of BCC compared to other unit cells, as shown in Fig. 12b. On the other the numerical simulation. Defects on the unit cell are modelled by a
hand, the normalised stress–strain curves of FCC and SC-BCC unit cells linear superposition of buckling eigenmodes with a scale factor for each
are more complicated in the elastic region. The curves of FCC unit cell mode, as given by,
converge to the same slope at high relative density, while the curves of

Fig. 12. (a) Elastic Modulus versus relative density; (b) Yield Strength versus relative density.

9
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 13. Normalised stress–strain curve of (a)SC, (b)BCC, (c)FCC and (d)SC-BCC unit cells at different relative densities.

m
of vertical struts along the compression direction in these unit cells, and
Δx = ∑ ωi φi , therefore structures tend to be stretching-dominated after imperfections
i=1 (9)
are introduced. Overall, the mechanical performance of SC and SC-BCC
where ωi is the mode shape and φi is the scale factor. The buckling effect unit cells are more sensitive to geometrical imperfections compared to
of all unit cells at a relative density of 0.45 is analysed. The first 10 that of FCC and BCC unit cells.
eigenmodes are determined by the analysis of the buckling behaviour
and applied imperfections to the perfect unit cell. The first 3 modes are 5. Fatigue study
multiplied by a scale factor of 5% of the strut size of each unit cell,
while the rest modes are assigned with 1% of strut size. Note that the 5.1. Prediction of fatigue failure
defect scale factors could be calibrated with experimental results and in
this work, the defect scaling factors are chosen for method demon- Although engineering structures are subjected to stresses and
stration purposes. The unit cells with imperfections are shown in strains, which are lower than the elastic limit under normal conditions,
Fig. 14a. stress concentration could cause the material to yield locally when
The uniaxial compressions of these unit cells are simulated, and data subjected to cyclic loads. Thus, apart from yield strength, fatigue
are extracted accoordingly. Fig. 14b presents the comparison between strength is also an important property that needs to be taken into
the stress–strain curves from perfect and imperfect unit cells. Results consideration. Performing a fatigue test could be expensive and time-
show that for FCC and BCC unit cells the introductions of geometrical consuming. Many parameters may affect the fatigue behaviour of lattice
imperfections have limited effect on the mechanical performance as structures such as relative density, cell size and strut size. Thus,
their stress–strain curves are very similar. By contrast, a drop in the studying the effect of these parameters before manufacturing is sig-
stress–strain curve is observed for SC and SC-BCC unit cell with im- nificant. Many researchers have applied numerical method to study the
perfections, which suggests that SC and SC-BCC unit cells are more fatigue properties [54–56]. In this research, we employed numerical
likely to buckle. This phenomenon could be attributed to the apperance simulation to study the effect of relative density and unit cell designs on

10
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 14. (a) Unit cells with imperfections; (a) Effect of geometrical imperfections on the stress–strain curves.

the fatigue behaviour of lattice structures. of crack initiation and the result is written to the output. These steps are
There are many different theories to predict fatigue life for mate- repeated for each node in the FE model, and the associated fatigue life is
rials. Brown-Miller theory, which shows the most realistic prediction of calculated.
fatigue life for ductile metals and reasonable prediction for brittle The FEA results (stress–strain curves) obtained from Section 3 are
metals, states that the maximum fatigue damage happens on the plane imported to Fe-Safe. The fatigue properties of Ti-6V-4Al are obtained
which experiences the maximum shear strain amplitude [57]. Thus, the from the experiments conducted by Edwards et al. [62,63]. The applied
damage of material will be a function of both shear strain and the strain stress is chosen between 0.2σy and σy with an interval of 0.05σy in order
normal to this plane. The Brown-Miller strain-life equation is given by, to capture the entire S-N curve. The frequency of the load is fixed at
15 Hz, and a constant minimum to maximum compressive loading ratio
Δγmax Δεn σ f' of 0.1 with sinusoidal wave shape is applied following a previously
+ = 1.65 (2Nf )b + 1.75εf' (2Nf )c
2 2 E (10) published fatigue test protocol on lattice structure [64]. Fig. 15 presents
the procedure to determine the S-N curve of lattice structure graphi-
where Δγmax is the maximum range of the shear strain, Δεn is the
cally.
maximum range of the normal strain, σ f' is the fatigue strength coeffi-
cient, εf' is the fatigue ductility coefficient, b is the fatigue strength
exponent, c is fatigue ductility exponent and 2Nf is the endurance in 5.2. Comparison of fatigue property between different lattice structures
reversals.
The prediction of fatigue life in this work is based on the Brown- In order to understand how cell topology will affect the fatigue
Miller theory and achieved by Fe-Safe 2018 (Simulia, Providence, RI). It property, relative density of all structures has been set as 0.45. The S-N
can be implemented to study fatigue property, such as where fatigue curves of all structures are investigated following the methodology
cracks occur and when fatigue cracks initiate [58–60], based on FEA discussed in the previous section. As shown in Fig. 15a, SC unit cell has
results. The general procedure for fatigue analysis in Fe-Safe starts with the highest fatigue strength among all the structures at any given load
reading the calculated FEA nodal stress tensor. Then, all the compo- cycle, while BCC unit cell shows the worst fatigue resistance. The effect
nents of the stress tensor are multiplied by the defined time history of of unit cell topology on fatigue properties follows a similar trend as that
applied load, so that a time history of stress tensor could be constructed. observed on elastic properties. Then, we normalise the S-N curves by
Based on the time history of the stress tensor, the time evolutions of in- the corresponding yield strength of the unit cell. As results presented in
plane principal stresses are determined following by the calculation of Fig. 15b, SC unit cell still has the highest normalised fatigue strength at
the principal strains. After that, the time history of shear strain, normal any relative densities. However, the normalised fatigue strength of BCC
strain and the associated normal stress are calculated. These data are unit cell becomes higher than that of FCC and SC-BCC unit cell. If the
used to compute the fatigue damages on each plane based on the unit cell is loaded with stress equal to its yield strength, SC-BCC will fail
Brown-Miller equation. Individual fatigue cycles for each plane are at 1500 load cycles while SC won’t fail until 105 load cycles. With the
identified by the Rainflow Counting Algorithm [61]. The fatigue da- increase of the load cycle, the normalised fatigue strength of all unit
mage in each cycle is calculated and summed to the total damage. Fi- cells converges to a certain value. The endurance limits have been
nally, the plane, which shows the shortest life is defined to be the plane found as 0.6σy for SC, 0.35σy for SC-BCC, 0.4σy for FCC and 0.45σy for BCC

11
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 15. Procedures to determine the S-N curve.

at relative density of 0.45. densities show a slight difference at high stress range. When increasing
Then, a further investigation of the effect of relative density on the the load cycle, all curves converge to 0.35σy .
fatigue behaviours is conducted following the same approach. The re- By contrast, the results of FCC and BCC unit cell seem to be irregular
sults are presented in Fig. 16. The left column is S-N curves for SC, SC- (Fig. 17). There are significant differences between normalised S-N
BCC, FCC and BCC unit cells at relative densities of 0.15, 0.25, 0.35 and curves obtained by different relative densities. For example, FCC unit
0.45, while the right column is the normalised S-N curves of these unit cell with relative density of 0.15 and 0.35 have similar normalised fa-
cells. It could be seen that the S-N curves are highly dependent on the tigue strength at different load cycles, while the similar behaviour has
relative density of lattice structures. The fatigue life improves when the been observed for BCC unit cell with relative densities of 0.25 and 0.45.
relative density increases for all lattice structures investigated. Fatigue However, the normalised S-N curves for these unit cells didn’t converge
strength of SC and BCC unit cell drops quicker at the high relative to the same trend. As found in the compression test, this phenomenon
density than that at low the relative density. By contrast, there isn’t the might be caused by losing contact on the surface of unit cell.
same effect observed for SC-BCC and FCC unit cell (Fig. 16).
After normalised the S-N curves by the corresponding yield strength
6. Conclusions
of unit cell, the behaviours of unit cells can be sorted into two groups
(Fig. 17). It could also be seen that relative density has less effect on
Four different lattice structures are studied by this research. The
normalised fatigue strength of SC and SC-BCC unit cell. For SC unit cell,
results show that the mechanical properties of lattice structure depend
when the stress is increased from 0.2σy to 0.7σy the maximum load cycle
on both relative density and unit cell topology. Both yield strength and
to fail has slight differences at various relative densities. However, with
elastic modulus increase with the increase in relative density for all
the increase of the load cycle, all curves converge to 0.6σy . Similarly, the
lattice structures are investigated. Among the structures studied by this
normalised S-N curves of SC-BCC unit cell with different relative
research, SC unit cell shows the highest elastic modulus and yield

Fig. 16. (a) S-N curve for different lattice structures with relative density of 0.45; (b) Normalised S-N curve for different lattice structures with relative density of
0.45.

12
C. Peng, et al. Composite Structures 235 (2020) 111821

Fig. 17. Effect of relative density on S-N curve and normalised S-N curve of SC, SC-BCC, FCC and BCC unit cells.

13
C. Peng, et al. Composite Structures 235 (2020) 111821

strength, while BCC unit cell has the lowest elastic modulus and yield [15] Schaedler TA, et al. Designing metallic microlattices for energy absorber applica-
strength at any given relative density. The relationship between elastic tions. Adv Eng Mater 2014;16(3):276–83. https://doi.org/10.1002/adem.
201300206.
properties and relative density of SC-BCC unit cell and FCC unit cell is [16] Imbalzano G, Tran P, Ngo TD, Lee PVS. A numerical study of auxetic composite
close to that predicted by Ashby-Gibson model. The number of unit cells panels under blast loadings. Compos Struct 2016;135:339–52. https://doi.org/10.
in lattice structure has limited effect on elastic properties, while its 1016/j.compstruct.2015.09.038.
[17] Lu T, Valdevit L, Evans A. Active cooling by metallic sandwich structures with
influence on the plastic behaviours is significant. A non-uniform dis- periodic cores. Progress Mater Sci 2005;50(7):789–815. https://doi.org/10.1016/j.
tribution of plastic dissipation energy and strain energy between layers pmatsci.2005.03.001.
for lattice structure under compression is observed. The middle layer is [18] HN Wadley, DT Queheillalt, Thermal applications of cellular lattice structures, In
Materials science forum, 2007, vol. 539: Trans Tech Publ, pp. 242-247.
subjected to more compression compared to other layers. This effect is [19] Asaoka K, Kuwayama N, Okuno O, Miura I. Mechanical properties and biomecha-
observed to be more substantial for BCC lattice, which could be at- nical compatibility of porous titanium for dental implants. J Biomed Mater Res
tributed to the lower elastic modulus and losing contact of the top 1985;19(6):699–713. https://doi.org/10.1002/jbm.820190609.
[20] Oh I-H, Nomura N, Masahashi N, Hanada S. Mechanical properties of porous tita-
surface under compression. The mechanical performance of SC and SC-
nium compacts prepared by powder sintering. Scr Mater 2003;49(12):1197–202.
BCC unit cells are more sensitive to geometrical imperfections com- https://doi.org/10.1016/j.scriptamat.2003.08.018.
pared that of FCC and BCC unit cells. [21] Fousova M, Vojtech D, Kubasek J, Jablonska E, Fojt J. Promising characteristics of
The fatigue property of lattice structures is also evaluated. For all gradient porosity Ti-6Al-4V alloy prepared by SLM process. J Mech Behav Biomed
Mater May 2017;69:368–76. https://doi.org/10.1016/j.jmbbm.2017.01.043.
the lattice structures investigated by this research, the fatigue resistance [22] Nguyen KC, Tran P, Nguyen HX. Multi-material topology optimization for additive
improves with the increase of relative density. SC unit cell has the manufacturing using polytree-based adaptive polygonal finite elements.
highest fatigue strength among all the structures, while BCC unit cell Automation Construct 2019;99:79–90. https://doi.org/10.1016/j.autcon.2018.12.
005.
shows the worst fatigue resistance. The effect of unit cell topology on [23] Tran P, Ngo TD, Ghazlan A, Hui D. Bimaterial 3D printing and numerical analysis of
fatigue properties follows a similar trend as observed on elastic prop- bio-inspired composite structures under in-plane and transverse loadings. Compos
erties. After normalising the fatigue strength by yield strength, relative Part B: Eng 2017;108:210–23. https://doi.org/10.1016/j.compositesb.2016.09.
083.
density has limited effect on fatigue resistance of SC and SC-BCC unit [24] Alsalla H, Hao L, Smith C. Fracture toughness and tensile strength of 316L stainless
cell. However, normalised S-N curves of FCC and BCC unit cell are still steel cellular lattice structures manufactured using the selective laser melting
sensitive to relative density. technique. Mater Sci Eng, A 2016;669:1–6. https://doi.org/10.1016/j.msea.2016.
05.075.
[25] Labeas G, Sunaric M. Investigation on the static response and failure process of
Declaration of Competing Interest metallic open lattice cellular structures. Strain 2010;46(2):195–204.
[26] Yang L, Cormier D, West H, Harrysson O, Knowlson K. Non-stochastic Ti–6Al–4V
foam structures with negative Poisson's ratio. Mater Sci Eng, A 2012;558:579–85.
The authors declare that they have no known competing financial
[27] Jamshidinia M, Wang L, Tong W, Ajlouni R, Kovacevic R. Fatigue properties of a
interests or personal relationships that could have appeared to influ- dental implant produced by electron beam melting®(EBM). J Mater Process Technol
ence the work reported in this paper. 2015;226:255–63.
[28] Winter R, et al. Plate-impact loading of cellular structures formed by selective laser
melting. Modell Simul Mater Sci Eng 2014;22(2):025021.
Acknowledgments [29] Salehian A, Inman DJ. Dynamic analysis of a lattice structure by homogenization:
Experimental validation. J Sound Vib 2008;316(1–5):180–97.
Research is supported by Vingroup Innovation Foundation (VINIF) [30] Ozdemir Z, et al. Energy absorption in lattice structures in dynamics: Experiments.
Int J Impact Eng 2016;89:49–61.
in project code VINIF.2019.DA04. [31] X. Ren, J. Shen, P. Tran, T. D. Ngo, and Y. M. Xie, Auxetic nail: Design and ex-
perimental study, Composite Structures, vol. 184, pp. 288-298, 2018/01/15/ 2018,
References doi: 10.1016/j.compstruct.2017.10.013.
[32] Harris J, Winter R, McShane G. Impact response of additively manufactured me-
tallic hybrid lattice materials. Int J Impact Eng 2017;104:177–91.
[1] Gibson LJ, Ashby MF. Cellular solids: structure and properties. Cambridge [33] Vu-Bac N, et al. A NURBS-based inverse analysis for reconstruction of nonlinear
University Press; 1999. deformations of thin shell structures. Comput Methods Appl Mech Eng
[2] Dharmasena K, et al. Dynamic compression of metallic sandwich structures during 2018;331:427–55. https://doi.org/10.1016/j.cma.2017.09.034.
planar impulsive loading in water. Eur J Mech-A/Solids 2010;29(1):56–67. [34] Talebi H, Silani M, Bordas SPA, Kerfriden P, Rabczuk T. A computational library for
[3] Fleck N, Deshpande V. The resistance of clamped sandwich beams to shock loading. multiscale modeling of material failure. Comput Mech 2013;53(5):1047–71.
J Appl Mech 2004;71(3):386–401. https://doi.org/10.1007/s00466-013-0948-2.
[4] Deshpande V, Ashby M, Fleck N. Foam topology: bending versus stretching domi- [35] Budarapu PR, Gracie R, Yang S-W, Zhuang X, Rabczuk T. Efficient coarse graining
nated architectures. Acta Mater 2001;49(6):1035–40. in multiscale modeling of fracture. Theor Appl Fract Mech 2014;69:126–43.
[5] Smith M, et al. The quasi-static and blast response of steel lattice structures. J https://doi.org/10.1016/j.tafmec.2013.12.004.
Sandwich Struct Mater 2011;13(4):479–501. [36] Rabczuk T, Ren H, Zhuang X. A nonlocal operator method for partial differential
[6] Radford D, McShane G, Deshpande V, Fleck N. The response of clamped sandwich equations with application to electromagnetic waveguide problem. Comput Mater
plates with metallic foam cores to simulated blast loading. Int J Solids Struct Continua 2019;59(1):31–55. https://doi.org/10.32604/cmc.2019.04567.
2006;43(7):2243–59. [37] Guo H, Zhuang X, Rabczuk T. A deep collocation method for the bending analysis of
[7] McShane G, Radford D, Deshpande V, Fleck N. The response of clamped sandwich kirchhoff plate. Comput Mater Continua 2019;59(2):433–56. https://doi.org/10.
plates with lattice cores subjected to shock loading. Eur J Mech-A/Solids 32604/cmc.2019.06660.
2006;25(2):215–29. [38] Anitescu C, Atroshchenko E, Alajlan N, Rabczuk T. Artificial neural network
[8] Ren X, Das R, Tran P, Ngo TD, Xie YM. Auxetic metamaterials and structures: a methods for the solution of second order boundary value problems. Comput Mater
review. Smart Mater Struct 2018;27(2):023001https://doi.org/10.1088/1361- Continua 2019;59(1):345–59. https://doi.org/10.32604/cmc.2019.06641.
665X/aaa61c. [39] Ashby MF. The properties of foams and lattices. Philos Trans A Math Phys Eng Sci
[9] Ren X, Shen J, Tran P, Ngo TD, Xie YM. Design and characterisation of a tuneable 2006;364(1838):15–30. https://doi.org/10.1098/rsta.2005.1678.
3D buckling-induced auxetic metamaterial. Materials & Design 2018;139:336–42. [40] Tan XP, Tan YJ, Chow CSL, Tor SB, Yeong WY. Metallic powder-bed based 3D
https://doi.org/10.1016/j.matdes.2017.11.025. printing of cellular scaffolds for orthopaedic implants: A state-of-the-art review on
[10] Tran P, Linforth S, Ngo TD, Lumantarna R, Nguyen TQ. Design analysis of hybrid manufacturing, topological design, mechanical properties and biocompatibility.
composite anti-ram bollard subjected to impulsive loadings. Composite Structures Mater Sci Eng C Mater Biol Appl 2017;76:1328–43. https://doi.org/10.1016/j.
2018;189:598–613. https://doi.org/10.1016/j.compstruct.2018.01.093. msec.2017.02.094.
[11] Calladine C, English R. Strain-rate and inertia effects in the collapse of two types of [41] Gibson LJ, Ashby MF. Cellular solids: structure & properties. 1st ed. Oxford
energy-absorbing structure. Int J Mech Sci 1984;26(11–12):689–701. [Oxfordshire]; New York: Pergamon Press; 1988.
[12] Mahmoud D, Elbestawi M. Lattice structures and functionally graded materials [42] Zhu HX, Knott JF, Mills NJ. Analysis of the elastic properties of open-cell foams with
applications in additive manufacturing of orthopedic implants: A review. J Manuf tetrakaidecahedral cells. J Mech Phys Solids 1997;45(3):319–43. https://doi.org/
Mater Process 2017;1(2):13 [Online]. Available: http://www.mdpi.com/2504- 10.1016/s0022-5096(96)00090-7.
4494/1/2/13. [43] Kwon YW, Cooke RE, Park C. Representative unit-cell models for open-cell metal
[13] Queheillalt DT, Wadley HNG. Cellular metal lattices with hollow trusses. Acta foams with or without elastic filler. Mater Sci Eng A 2003;343(1-2):63–70. https://
Materialia 2005;53(2):303–13. https://doi.org/10.1016/j.actamat.2004.09.024. doi.org/10.1016/s0921-5093(02)00360-x.
[14] Clough EC, Ensberg J, Eckel ZC, Ro CJ, Schaedler TA. Mechanical performance of [44] Buffel B, Desplentere F, Bracke K, Verpoest I. Modelling open cell-foams based on
hollow tetrahedral truss cores. Int J Solids Struct 2016;91. https://doi.org/10. the Weaire-Phelan unit cell with a minimal surface energy approach. Int J Solids
1016/j.ijsolstr.2016.04.006. Struct 2014,;51(19–20):3461–70. https://doi.org/10.1016/j.ijsolstr.2014.06.017.

14
C. Peng, et al. Composite Structures 235 (2020) 111821

[45] Barbier C, Michaud PM, Baillis D, Randrianalisoa J, Combescure A. New laws for lattice materials via computational mechanics: Application to lattices with smooth
the tension/compression properties of Voronoi closed-cell polymer foams in rela- transitions in cell geometry, International Journal of Fatigue, vol. 47, pp. 126-136,
tion to their microstructure. Eur J Mech A Solids 2014;45:110–22. https://doi.org/ 2013/02/01/ 2013, doi: 10.1016/j.ijfatigue.2012.08.003.
10.1016/j.euromechsol.2013.12.001. [56] S. Arabnejad Khanoki and D. Pasini, Fatigue design of a mechanically biocompa-
[46] Gong L, Kyriakides S. Compressive response of open cell foams Part II: Initiation and tible lattice for a proof-of-concept femoral stem, J Mech Behav Biomed Mater, vol.
evolution of crushing. Int J Solids Struct 2005;42(5–6):1381–99. https://doi.org/ 22, pp. 65-83, Jun 2013, doi: 10.1016/j.jmbbm.2013.03.002.
10.1016/j.ijsolstr.2004.07.024. [57] M. W. Brown and K. J. Miller, A Theory for Fatigue Failure under Multiaxial Stress-
[47] Zhu HX, Mills NJ, Knott JF. Analysis of the high strain compression of open-cell Strain Conditions, Proceedings of the Institution of Mechanical Engineers, vol. 187,
foams. J Mech Phys Solids 1997;45(11–12):1875–904. https://doi.org/10.1016/ no. 1, pp. 745-755, 2016, doi: 10.1243/pime_proc_1973_187_161_02.
s0022-5096(97)00027-6. [58] Baek S, Cho S, Joo W. Fatigue life prediction based on the rainflow cycle counting
[48] Gurrappa I. Characterization of titanium alloy Ti-6Al-4V for chemical, marine and method for the end beam of a freight car bogie. Int J Automot Technol
industrial applications. Materials Character 2003;51(2-3):131–9. https://doi.org/ 2008;9(1):95–101.
10.1016/j.matchar.2003.10.006. [59] Ås SK, Skallerud B, Tveiten BW. Surface roughness characterization for fatigue life
[49] G. R. Johnson and W. H. Cook, A constitutive model and data for metals subjected predictions using finite element analysis. Int J Fatigue 2008;30(12):2200–9.
to large strains, high strain rates and high temperatures, A Constitutive Model and [60] R. Fertig and D. Kenik, Predicting composite fatigue life using constituent-level
Data for Metals Subjected to Large Strains, High Strain Rates and High physics, in 52nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Temperatures, pp. 541-547, 1983. Materials Conference 19th AIAA/ASME/AHS Adaptive Structures Conference 13t,
[50] Lesuer D. Experimental investigation of material models for Ti-6Al-4V and 2011, p. 1991.
2024–T3. Livermore: University of California, Lawrence Livermore National [61] S. Downing and D. Socie, Simple rainflow counting algorithms, International
Laboratory; 1999. p. 1–36. Journal of Fatigue, vol. 4, no. 1, pp. 31-40, 1982/01/01/ 1982, doi: 10.1016/0142-
[51] R. Boyer, G. Welsch, and E. Collings, Materials Properties Handbook: Titanium 1123(82)90018-4.
Alloys (Materials Park, OH: ASM International, 1994), Google Scholar, pp. 3-31. [62] Edwards P, Ramulu M. Fatigue performance evaluation of selective laser melted
[52] Vu-Bac N, Lahmer T, Zhuang X, Nguyen-Thoi T, Rabczuk T. A software framework Ti–6Al–4V. Mater Sci Eng, A 2014;598:327–37. https://doi.org/10.1016/j.msea.
for probabilistic sensitivity analysis for computationally expensive models. Adv Eng 2014.01.041.
Softw 2016;100:19–31. https://doi.org/10.1016/j.advengsoft.2016.06.005. [63] P. Edwards, M. Ramulu, and A. O'Conner, Electron Beam Additive Manufacturing of
[53] Hamdia KM, Ghasemi H, Zhuang X, Alajlan N, Rabczuk T. Sensitivity and un- Titanium Components: Properties and Performance. 2013, pp. 061016-1.
certainty analysis for flexoelectric nanostructures. Comput Methods Appl Mech Eng [64] Yavari SA, et al. Fatigue behavior of porous biomaterials manufactured using se-
2018;337:95–109. https://doi.org/10.1016/j.cma.2018.03.016. lective laser melting. Mater Sci Eng C Mater Biol Appl 2013;33(8):4849–58. https://
[54] Zargarian A, Esfahanian M, Kadkhodapour J, Ziaei-Rad S. Numerical simulation of doi.org/10.1016/j.msec.2013.08.006.
the fatigue behavior of additive manufactured titanium porous lattice structures. [65] Imbalzano G, Tran P, et al. Three-dimensional modelling of auxetic sandwich panels
Mater Sci Eng C Mater Biol Appl Mar 2016;60:339–47. https://doi.org/10.1016/j. for localised impact resistance. J. Sand. Str. Mater. 2015. https://doi.org/10.1177/
msec.2015.11.054. 1099636215618539. In press https://journals.sagepub.com/doi/abs/10.1177/
[55] E. Masoumi Khalil Abad, S. Arabnejad Khanoki, and D. Pasini, Fatigue design of 1099636215618539.

15