Accepted Manuscript

Title: Micromechanical analysis of the effective properties of

lattice structures in additive manufacturing

Author: José Souza Alexander Großmann Christian

Mittelstedt

PII: S2214-8604(18)30270-7
DOI: https://doi.org/doi:10.1016/j.addma.2018.07.007
Reference: ADDMA 449

To appear in:

Received date: 23-4-2018

Revised date: 22-6-2018
Accepted date: 14-7-2018

Please cite this article as: José Souza, Alexander Grossmann, Christian
Mittelstedt, Micromechanical analysis of the effective properties of lattice
structures in additive manufacturing, <![CDATA[Additive Manufacturing]]> (2018),
https://doi.org/10.1016/j.addma.2018.07.007

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 Micromechanical analysis of the eective properties of
lattice structures in additive manufacturing

t
José Souzaa , Alexander Groÿmanna,∗, Christian Mittelstedta

ip
a Technische Universität Darmstadt, Konstruktiver Leichtbau und Bauweisen,
Otto-Berndt-Straÿe 2, 64287 Darmstadt, Germany

cr
us
Abstract
Lightweight design is an area of mechanical engineering that becomes increas-

an
ingly important in many industries, as they pursue reduced mass and more e-
cient parts. A special class of materials for load-bearing structures are metallic
cellular materials with cubic unit cells, which can be manufactured conveniently
M
through laser beam melting (LBM). Such materials exhibit a rather complex
microstructure and can be analysed using analytical and numerical methods
wherein the determination of properties such as relative density, eective elas-
d

tic and yield strength properties is of special interest. This paper addresses
closed-form analytical methods based on beam theories for the determination
te

of the eective properties of additively manufactured microstructures such as

lattices, and a comparison with experimental results [1, 2] which leads to excel-
p

lent agreements for relative densities lower than 40%, although results reveal a
ce

great dependency on the manufacturing strategy. Lastly, a classication con-

cerning the topology of the cellular units is presented as well in order to help
the engineer choose appropriate geometries for specic application purposes. In
Ac

conclusion, this structural concept may be applied in many elds such as bio-
engineering and in functional graded materials as they are applied in lightweight
engineering.
Keywords: Additive Manufacturing, Cellular Solids, Lattice Structures,
Lightweight Design

∗ Corresponding author

A
Preprint submitted to Journal of L TEX Templates June 22, 2018

Page 1 of 43
 1. Introduction

t
Modern industrial manufacturing techniques are geared towards optimized

ip
components concerning production time, structural mass reduction and general
design exibility. Rapid prototyping techniques have become increasingly rel-

cr
5 evant during the last few decades due to the possibility of producing complex
geometries in small quantities or one-of-a-kind production with aordable re-

us
sources. This gives rise to the possibility of creating new structural concepts
inspired for example by biological structures, which would not be feasible with

an
conventional means of manufacturing. Naturally, structural designs as they ap-
10 pear in nature in many forms are the main source of inspiration, so that natural
growth processes are imitated in order to apply them to engineering problems,
M
with a mammals' trabecular bone being a good example. Depending on the
applied load, mass balance shifts [3] result in a self-optimized structure with
graded mechanical properties.
d

15 In the sense of structural mechanics, such structures are classied as open-

te

celled cellular solids [2] which can be interpreted to be consisting of so-called unit
cells, i.e. representative cells which are assumed to be distributed over the entire
structure and which are simple enough so that their structural analysis can be
p

performed with acceptable computational eort. Such unit cells of cellular solids
ce

20 are analysed on a micro to meso-scale, thus enabling the estimation of the impact
on macroscopic properties (mechanical, electrical and thermal) of the solid when
the topology of the unit cell is altered [4]. In the past, such cellular solids could
Ac

only be manufactured as stochastic materials with very limited means of actively

manipulating their mechanical properties to a desirable degree.
25 However, modern additive manufacturing processes such as selective laser
melting in the case of metallic structures enables the creation of cellular ma-
terials with a dened strictly periodic unit cell distribution, resulting in more
complex and rened applications such as osseointegrative implants [5]. Some au-
thors include a substructure into the lattice structure to increase the static and

Page 2 of 43
30 dynamic mechanical propierties of structures, which has been done for polymer
honeycombs [6] and ceramic octet nanolattices [7].
Although a literature survey shows that there is an impressive body of work

t
ip
concerning the analysis and experimental study of lattice structures [1, 5, 8,
9, 10], studies concerning their mechanical behavior by closed-form analytical

cr
35 methods are not available.
However, it is of paramount importance to perform detailed parameter stud-
ies of such cellular solids in order to understand their behavior, hence the avail-

us
ability of analytical equations connecting geometry and mechanical properties
is a crucial point. The objective of the current work is thus the investigation

an
40 of mechanical properties of open-celled cellular solids geared towards the ap-
plication in additive manufacturing design. We will show that the resulting
closed-form analytical models that will be developed in this paper enable an
M
estimation of the unit cell stinesses as well as strength properties, and it will
be shown that these properties can be correlated with variations of the amount
45 of applied material, i.e. the relative density.
d

Depending on the component geometry, according to [11] manufacturing

te

technologies are divided into three main groups: subtractive, forming and ad-
ditive manufacturing (AM) technologies. The rst one is characterized by the
removal of material aiming to obtain the nal geometry, for instance milling
p

50 and turning. The manufacturing of a part through the application of external

ce

force or heat denes a forming technology, such as deep drawing, forging and
casting. Dierently from the latter described conventional methods, additive
manufacturing produces the desired shape through the addition of material.
Ac

Based on the principle that all geometries may be dened by a superposition

55 of layers, the additive manufacturing process enables the generation of complex
geometries through a layerwise technology. The process is divided in two main
subprocesses: the manufacturing of each section and the joining of all layers
[12]. In other words, after virtually slicing the geometry into thin layers, its
structure is subdivided into a nite number of sections that are printed over
60 another. The nal result is an approximation of the CAD geometry formed by

Page 3 of 43
 a superposition of all layers. In the German guidelines VDI 3405 [13] eleven
dierent manufacturing techniques are listed.
However, the current work focusses on the analysis of metal components

t
ip
with minimum dimensions as small as 0.2 mm [8] and thus requires experimental
65 data for the validation of theoretical models. Powder bed techniques such as

cr
Laser Beam Melting (LBM, a specic subclass of selective laser melting) are
therefore chosen to be the standard manufacturing procedure for this study
[1, 8]. In Figure 1 this process is briey described. It allows the production of

us
components consisting of virtually any metal that allows welding, for instance
70 aluminium, titanium, or steel alloys, among others. Mechanical properties of the

an
nal materials are classied as transversely-isotropic [8]. Experiments conducted
by the same author used 290 dierent stainless steel 1.4404 test specimens, built
in 145 dierent directions. As a result, it was concluded that elastic properties
M
do not depend on the manufacturing direction of the specimen, whereas strength
75 properties vary linearly with the leverage angle.
d
te

Wait next
Slicing and step End of
STL �ile
Process
p

Lower powder Melt current

bed section
ce

Apply powder
layer
CAD File Physical Part
Inside Machine
Ac

Virtual Level Physical Level

Figure 1: Simplied manufacturing process for powder bed manufacturing process e.g. Laser
Beam Melting (LBM)

The focus of the present work is to evaluate mechanical properties of cellular

materials fabricated by selective laser melting as well as discussing potential
engineering applications of such materials with mesostructure. In order to full

Page 4 of 43
 this objective, the rst step is to perform a bibliographic research on current
80 fabrication methods as well as the manufacturing of lattice structures.
The scope of this paper is to investigate the mechanical behavior of truss-

t
ip
based lattice structures, which has been the focus of investigation by authors
before [1, 8]. However, most publications focus on numerical and experimental

cr
work, which is inappropriate in order to establish general design rules for engi-
85 neers. For that reason, this paper uses a strain energy based homogenization
method and Castagliano's 2nd theorem in order to generate analytical solu-

us
tions for the mechanical behavior of the structures shown. The ndings will be
compared with current experimental ndings from Rehme [8] and Merkt [1].

2. Cellular Design in Additive Manufacturing

2.1. Cellular Materials an

M
90

Cellular solids are made up of an interconnected network of solid struts or

plates which form the edges and faces of cells [2]. They consist of cells, which are
d

then repeated throughout the space, thus forming the lattice structure. Gen-
erally, cellular solids are categorized in open and closed celled such as regular
te

95 and stochastic structures. In nature, they are found in structures such as cork,
wood, bones and sponges, which possess dierent functionalities in their respec-
p

tive organisms' physiologies. Because such materials exhibit a low density along
ce

with advantageous mechanical properties, their use in lightweight structures is

suitable in pursuance of high stiness-to-weight-ratios.
100 Additive manufacturing research mainly focusses on open celled regular cel-
Ac

lular solids, like hierachical honeycomb structures [6], truss-based lattices [1, 8],
octet and rhombicuboctahedron structures [14] and surface based lattices like
gyroid or diamond structures [15]. Closed celled cellular solids are undesirable
for selective laser melting because the metal powder cannot be removed after
105 manufacturing these structures.
Lattice structures can be divided according to their mechanical properties:
bending or stretch dominated, as illustrated in Figure 2. In the rst case struts

Page 5 of 43
 carry their loads mainly through bending, so that a lower stiness and strength
are expected when compared to mainly stretch oriented structures. However,
110 they possess a stress plateau in the plastic region after a stress peak. On the

t
ip
other hand, stretch dominated lattice structres exhibit a higher stiness and
strength, as well as a prominent cold hardening in the plastic region. While

cr
bending dominated structures are ideal for energy absorbers due to the large
amount of absorbed energy for lower stresses, stretch dominant lattice structures
are well suited for e.g. lightweight applications, where operating conditions are

us
115

mostly within the linear-elastic region. Being able to assess properties of the
cellular material based on its relative density is extremely important, in order

an
to both understand its behavior and to be able to compare dierent geometries.
The base of such an analysis resides in the so-called Percolation Theory [16],
120 which describes the formation of clusters through random occupation of lattices.
M
stretch-dominated
bending-dominated

Yield stress Densification

(stretch-dominated)
d Stress σ
te

bending-dominated Stress peak

Plateau stress
p

Densification
Settling of Yield stress
(bending-dominated) strain
cell edges
ce

stretch-dominated Strain ɛ
Ac

Figure 2: Diagrams of bending and stretch dominant behavior (left) as well as their stress-
strain curves (right) based on [8]

A cellular structure is in this context an innite cluster of unit cells that

together form the foam [4]. Considering that the probability of a cluster being
formed is substituted through the relative density, the mechanical properties of

Page 6 of 43
 an open-celled cellular solid may be described by

te
K∗ ρ∗

=C (1)

t
KS ρS

ip
125 where (.)∗ denotes the material properties of the cellular structure and (.)S
the one of the solid material. The quantity K is the considered mechanical

cr
property (e.g. a stiness or strength), ρ is the density, and C and te are con-
stants determined by analytical models or by experiments. If the relative density

us
attains a unit value, the eective property should be the same as for the consti-
130 tutive material, hence C must always be equal to 1. As for the exponent te , it
determines how the property will vary as the relative density changes. Because

an
of this, when higher stiness-to-weight ratios are pursued, lower te values are
advisable.
M
2.2. Manufacturing Strategies for Lattices Structures
135 There are two manufacturing strategies to be considered: contour-hatch
(CH) and point exposure (PE), as depicted in Figure 3. The rst one is the
d

machine standard and uses the conventional slicing technique, in which the
te

geometry's contour is rst melted, followed by the melting of the internal area
following hatch vectors. PE on the other hand, as described by [8], consists of
p

140 a cloud of points, where the laser exposure occurs for a certain time, forming
molten pools. Although PE needs the development of a proper slicing code and
ce

its struts' surfaces are rough, manufacturing time is decreased and thinner struts
(>150 µm) are possible opposed to machine standard (>250 µm). CH is on the
Ac

other hand capable of achieving better surface quality, mechanical properties

145 and less stress concentration [1]. However, scan delay leads to higher process
times and therefore costs. The dierentiation of both strategies is important for
understanding the limitations of the generated models in the current work.

2.3. Design of Additively Manufactured Lattice Structures

The manufacturing of lattice structures built inside closed components is
150 one promising possible application for these geometries, which enables the cus-

Page 7 of 43
 t
ip
cr
us
Figure 3: Diagrams of point exposure (a) and contour hatch scan strategies (b), based on [1]

an
tomization of mechanical properties as well as a suitable material distribution
throughout the considered structural element.

f2bcc f2bccz
M
bcc bccz
d
te

fcc fccz f2cc f2ccz

p
ce

Figure 4: Tessellated patterns with cubic cells. The abbreviations can be explained as follows:
fcc: face centered diagonal struts, bcc body centered diagonal struts, z: vertical struts
Ac

The design of the geometric parameters of lattice cells faces both an upper
bound and a lower bound concerning the ratio between cell size and strut diame-
155 ter. The rst one corresponds to manufacturing limits such as the balling eect
and deformation whilst the latter pertains to geometric restrictions. Higher
slenderness ratios lead to thinner struts, which may deform under load applied
by the coater. Considering this limit, Rehme [8] applied a simple Bernoulli

Page 8 of 43
 beam model to limit the deformation and providing a simple formula for the
160 maximum cell size:

t
r
15EIdmin

ip
(2)
3
amax =
8

cr
us
an
Figure 5: Diagram of an overlap between two struts with x depicted as its length, based on
Rehme [8]
M
Therein, amax is the maximum allowed cell size, EI is the strut's bending
stiness and dmin the smallest strut diameter possible with the available AM
system. A simple equation has been developed [8] to asses the minimum strut
d

diameter as a function of the laser beam focus d0,86.5 and the layer thickness lz .
te

dmin = d0,86.5 + 2lz (3)

p

165 The lower slenderness ratio bound is achieved on the other hand when the
lattice transits to a solid with pores and stop being a framework. This happens
ce

when two neighbor struts are so thick that their overlapping region takes up
more than half of its length [8] which is illustrated in Figure 5. Therein the
Ac

overlap length x is thus assessed as follows

d d
x= + (4)
2 sin α 2 tan α
The critical ratio a
is then obtained, depending on the cell unit's ge-


170
d crit

ometry, equating l with a multiple of x. Regarding the considered cellular

structures shown in Figure 4, Rehme [8] produces the equations listed in Table
1.

Page 9 of 43
 Cell type α [◦ ] l [mm] a


d crit
√ √
f cc 60 2a = 2x 1.5
√ √ √
45

t
f ccz 2a = 2x (1 + 2)/ 2
√ √ √

ip
f2 cc 60 2a = 2x + d (1 + 3)/ 2
√ √ √
f2 ccz 45 2a = 2x + d (2 + 2)/ 2
√ √ √

cr
bcc 70.5 3a = 2x (2 2)/ 3
√ √ √ √
bccz 54.7 3a = 2x + d ( 1.5 + 0.5 + 1)/ 3
√ √ √ √

us
f2 bcc 35.3 2a = 2x + d ( 3 + 2 + 1)/ 2
√ √ √ √
f2 bccz 35.3 2a = 2x + d ( 3 + 2 + 1)/ 2

Table 1: Estimation of critical aspect ratios for lattice unit cells [8]

an
3. Geometry Parameterization and Homogenization Scheme
M
175 The space for the considered cell geometries is tessellated according to crys-
tallographic geometries, as suggested by Rehme [8] and depicted in Figure 4.
The slenderness ratio [17] is chosen to be the geometric parameter:
d

a
SR = (5)
te

t
where a is the cell size and t the strut thickness. This value is rst correlated
p

to each conguration's relative density.

180 Because only one constitutive material is used, this value may be obtained
ce

through the ratio between struts' and total volume. As described in an earlier
work [18], this may be done through a simple model disregarding intersections
or numerically considering all intersections. The latter was done using Monte
Ac

Carlo density integration with deviation to previous iteration less than 0.01 %.
185 Although this is more complex and time intensive, the results prove to be better
related to experiments across a large relative density spectrum and equal to the
data obtained through CAD. The scaling laws presented in the current work
are therefore plotted using this method. Basically, Monte Carlo integration
generates a random number of points both in the corresponding lattice, as shown

10

Page 10 of 43
190 in Figure 6, and in the total volume. Thereafter the relative is obtained, dividing
one by the other

t
ρ∗ nlattice

ip
= (6)
ρS ntotal

cr
us
an
a M
t a
a
d

Figure 6: Monte Carlo integration of a f2 ccz lattice structure

te

Figure 7 shows the inuence of the intersection for the relative density when
neglected and when considered, for example through a Monte Carlo integration.
p

Especially for low slenderness ratios yielding high relative densities the inter-
ce

195 sections have a signicant inuence on the structure's relative density. Further-
more, the plot shows experimental results that have been conducted by Merkt
[1].
Ac

In order to compare dierent cellular geometries, the material macroscopic

properties should be assessed for all structures. This is done by applying a sim-
200 ilar homogenization process as employed by Demiray [10] and Iltchev [19]. The
microstructure of a representative element from the cellular material is replaced
by a macroscopically equivalent medium, which is considered to be ideally ho-
mogeneous. The selected space is named representative volume element (RVE),
which is a repeating unit of the structure due to the assumed structural period-

11

Page 11 of 43
 t
ip
cr
f2cc z

us
an
Figure 7: Relative density calculation and measurement for a f2 ccz structure
M
205 icity. The chosen homogenisation scheme is a strain energy based concept. The
strain energy for a Timoshenko beam can be calculated according to Equation
B.1, where N denotes the normal force, V the lateral force and M the corre-
d

sponding moments [20]. For the same macroscopic deformation state both RVE
te

possess the same stored amount of strain energy. Macroscopic stresses (σM )ij
210 and strains (εM )ij are dened as
p

Z
ce

1
(εM )ij = (ε)ij dVΩRV E
VΩRV E VΩRV E
Z
1
(σM )ij = (σ)ij dVΩRV E (7)
VΩRV E VΩRV E
Ac

where (εij ) and (σij ) are microscopic strains and stresses respectively, which
are integrated throughout the cell domain VΩRV E . The macroscopic strain en-
ergy density stored in the representative volume element RVE should be the
same as in the homogenized medium, thus:

12

Page 12 of 43
 E ∗
ΠRV
M
E
= (ΠRV
M )

t
1 1 ∗
)ij (ε∗M )ij (8)

ip
(σM )ij (εM )ij = (σM
2 2

215 Considering the behavior of both the framework and equivalent medium to

cr
be linear-elastic, Hooke's Law applies as constitutive equation. After manipu-
lating the equations, it may be proven that the compliance matrix (sM )ij may

us
be obtained through

∗
(sM )ij = 2[(σM ∗
)i (σM )j ]−1 ΠRV E ∗
(σM,i ) (9)

an
M

where (σM
∗
)i is an applied unitary load and ΠRV
M
E ∗
(σM,i ) the strain energy
220 stored during the application of this load. Considering an orthotropic behavior
M
of the cellular material, 6 load cases (3 normal stresses and 3 shear stresses) are
needed to determine the diagonal (sM )ij . The remaining 6 Poisson's ratios vij
∗

are obtained through the strains, which follow from (σM

∗
)i :
d

−(εM )j
∗
νij = (10)
te

(ε)i
with i being the principal loading direction and j the perpendicular one.
p

225 Figure 8 visualizes this approach.

ce

4. Analytical Model

4.1. Assessing Eective Elastic Constants

Ac

The current model considers cells with high slenderness ratios, so that a
beam model may be used. To better represent these results and be able to
230 evaluate lower slenderness ratios, a shear elastic beam is considered. Because
the printed metal possesses anisotropic strength properties, three coordinate
systems are dened: material (m) aligned with the machine axis, part (p) with
the part, cell (c) with the cell and strut (s) with xs parallel to the strut. They
are illustrated in Figure 9, which also reveals the simplication made due to

13

Page 13 of 43
 t
ip
cr
us
Cellular Solid Homogenized Body
ti ti

an
*
Ω Ω
Macro-Scale

M
Y Y
0
X u i X u0i
d
Z Z
Meso-Scale

te

y y
p

x x
z RVE z RVE *
ce

Figure 8: Homogenization schematic for cellular lattice structures in order to obtain eective
elastic constants
Ac

14

Page 14 of 43
235 symmetry. In Figure 10 the angles θs and ϕs are used to describe the rotation
of the strut's coordinate system with respect to the cell. The struts, which share
material with other cells, have their stiness multiplied by a depreciation factor

t
ip
inversely proportional to the number of cells. For instance, strut 4 has its axial
stiness multiplied by 0.25.

cr
yc yc
xc zc xc zc

us
1

an
Sym.

zp zm 4
x m yp 2
3
xp ym
M
θh RVE RVE sym

Figure 9: Simplication of the f2 bccz RVE structure due to its symmetry. Also shown the
d
coordinate systems material (m), part (p) and cell (c) and the hatch angle θh
te

240 Boundary conditions are applied so that opposite planes of the RVE remain
parallel to each other and opposite nodes do not rotate relatively to each other,
p

similar to [10], thus maintaining the structure's periodicity. They are applied
to nodes O, P1 to P4 , whereas the load is applied only at point O, as depicted
ce

in Figure 11 (left). With the displacements of this node alone, it is possible

245 to obtain all required load cases. The boundary conditions can be found in
Appendix 1.
Ac

Because for some load cases the boundary conditions engender a statically
indeterminate RVE, an energy method should be used applying uniform loads.
This way the problem is simplied to a statically determinate system and the
250 loads may be calculated. Castigliano's 2nd theorem is thus chosen, because all
boundary conditions found on the nodes can be considered adequately. The

15

Page 15 of 43
 ys xs

t
ip
zs
ys

cr
li
zs

us
yc φs
xc
zc
t i, A i , I i
θs
an
Figure 10: Geometric parametrization of a generic i-strut and its length li (left). A strut's
M
section with respectively thickness ti , sectional area Ai and momentum of inertia Ii (right)
d
te

t,n
fi fi
O
O
p
ce

n
4 1 n Vy
3 Nx
2 n
Vz
P1
Ac

P4 yc
P3 xc zc
P2 Pn
RVE sym Strut n

Figure 11: Load and boundary conditions of RV Esym (left) and a n-strut (right)

16

Page 16 of 43
 displacement experienced by node O is assessed as follows:

∂Πi,n
uci,n = = scij,n fj,n
c
(11)

t
c
∂fi,n

ip
where scij,n is the compliance matrix in cell coordinates of an n-strut. Be-
cause all nodes are connected through node O:

cr
uci = uci,n (12)

us
and

X
fic = c

an
(fj,n )
n
! !
X X
fic = (scij,n )−1 uci = ccij,n uci (13)
n n
M
with ccij,n as the stiness matrix of a n-strut. As a result, the components
255 of the global stiness matrix are obtained
d
!
X
ccij = ccij,n = (scij )−1 (14)
te

n
The corresponding equations behind the abbreviated notation above can be
obtained from Appendix 2 where the schematic for Figure 11 (right) is employed.
p

In order to obtain the deformation for the RVEsym depicted in Figure 11 (left)
ce

one must apply the mentioned equations on the struts 1 through 4 and based on
260 the described geometry, macroscopic strains and stresses may be obtained from
the displacement uci and the concentrated load fjc as follows for axial loading:
Ac

uci
(εM )i = 2 (15)
a
fic
(σM )i = 4 2 (16)
a
The stored strain energy density is next described as

1 4
ΠRV
M
E ∗
(σM,i )= (σM )i (εM )i = 3 uci fic (17)
2 a

17

Page 17 of 43
 Substituting in the homogenization theory equation:

a c c −1 c c c
(sM )ij = (f f ) fi sij fj (18)

t
4 i j

ip
Considering unitary load for axial load cases, the rst three diagonal ele-
265 ments may be obtained, as well as the Poisson's ratio

cr
1 4
Ei∗ = = c (19)
(sM )ii asii

us
and

scij
∗
(20)

an
νij =−
scii
for i, j = 1, 2, 3. The shear properties are similarly determined with unitary
load, however, with a dierent formulation for the shear strain, so that the shear
M
moduli may be obtained as follows:

   
2 2
d

G∗12,23 = arctan + arctan

asc asc
 11   22 
te

2 2
G∗13 = arctan c + arctan (21)
as11 asc33

270 4.2. Assessing Anisotropic Yield Strength Properties

p

For the strength analysis, two possible failure mechanisms may occur: mate-
ce

rial yielding and strut buckling. The most critical one determines the strength
property. The loads are rst distributed through all struts connected to O and
the stresses (normal and shear), described as function of the strut coordinates.
Ac

275 The shear matrix is next transformed to the material coordinate system. Hill's
criterion [21] is chosen to identify material yielding. As for strut buckling, the
critical buckling load is estimated using Euler's buckling equations [20]. The
yield strength according to Rehme [8] may be assessed according to
 
σ ∗ , |σ ∗
| ≤ |σ ∗
| 
buckle buckle yield
∗
Rp0.2 = (22)
 σ ∗ , |σ ∗ | > |σ ∗ | 
yield buckle yield

18

Page 18 of 43
 The magnitude for the yield stress can be found in Appendix 3 and the one
280 for the buckling stress in Appendix 4.

t
5. Numerical Model

ip
Although the analytical modeling is very ecient for the study of simple load

cr
cases as well as giving a better understanding of the mechanical properties of
lattice structres, a nite element (FE) model has more exibility and allows for

us
285 more complex simulations. Because there are few experimental results found in
literature to the studied cellular geometries, one single method does not suce to
obtain reliable data. The analysis developed in the current section is thus used

an
both as a verication of the analytical results and to verify the use of beam
elements for the study of open celled cellular structures. Timoshenko beam
elements with second order interpolation functions are hence used, which are
M
290

compatible to the analysis previously described. The same boundary conditions

as in the analytical model were considered, so that all opposite planes remain
d

parallel to each other. However, the whole RVE is analyzed with the FE model,
so that the verication gains a greater signicance. The depreciation factor
te

295 for stiness values is also applied to the struts with shared cells, as previously
described. The forces are applied as concentrated loads as depicted in Figure
p

10.
ce

In an analogous procedure as in the closed-form analytical case, the elastic

properties are obtained as follows:
Ac

19

Page 19 of 43
 Fy
5 8 Fx
y y

t
x x
6 7

ip
y
Fxy Fxz

cr
x
z 4
1 y x
x z

us
2 3

Figure 12: Geometry Parametrisation of RVE (left) and depiction of load cases (right)

E2∗ =
2 F an
M
a (u2,max )F y
∗ 2 F
E1,3 =
a (u1,max )F x
 
u(1,3),max
d
∗
ν21,23 =−
u2,max Fy
te

 
∗ u2,max
ν12,32 =−
u1,max F x
 
∗ u3,max
=−
p

ν13,31
u1,max F x
    −1
4F (u1,max )F xy (u2,max )F xy
ce

G∗12,23 = 2 arctan + arctan

a a a
    −1
4F (u1,max )F xz (u3,max )F xz
G∗13 = 2 arctan + arctan (23)
a a a
Ac

300 Therein, F is the modulus of the applied forces1 and ui,max the maximum
displacement. The load cases are represented by Fx , Fy , Fxy and Fxz , as de-
picted in Figure 12. This way, only four of them are needed to obtain all 12
elastic constants.

1 Here we use F as the spring constant to relate force and displacement according to
F orcei = F · ui

20

Page 20 of 43
 6. Results and Analysis

305 Both analytical and numerical models are used, as well as experimental data

t
from the open literature. This way it is possible to perform analysis of all

ip
studied geometries and verify as well as validate the developed model.

cr
6.1. Comparison of Models
Due to the fact that no experiments were performed in the current work, a

us
310 careful analysis of the obtained results and their comparison with other works
is mandatory to better comprehend the object of study. Cellular structures
were studied and exponential scaling laws were obtained from experiments in

an
works presented by Merkt [1] and Rehme [8]. These results are thus compared
to the models developed in the current work, so that a relative density interval
315 is determined for which the structure may be modeled with beams. Both works
M
used stainless steel 1.4404 as constitutive material, however, with dierent ap-
proaches. Merkt [1] devised scaling laws for the conguration f2 ccz . Using
ultrasound, he was able to determine the modulus of elasticity as E =190 GPa
d

and Poisson's ratio ν as ν = 0.35, which is also used in the current model. The
te

320 cell size has been maintained constant at 2 mm while the slenderness ratio has
been varied through the strut thickness. The manufacturing strategy chosen for
p

the LBM process was the contour-hatch procedure, in order to obtain parts with
higher quality. The results of both models are compared to Merkt's experimen-
ce

tal equation in Figure 13. It is rstly important to notice that both the results
325 by the FEM as well as the analytical model coincide, which veries the solution.
Ac

About the accuracy of the beam model, it is possible to infer that the analyzed
structural behavior can be predicted with relatively low error (<20 %) for slen-
derness ratios higher than 3.8, which corresponds to 25.5 % relative density.
This shows that although nal geometries may dier from the ideal ones, beam
330 models are able to predict the structure's behavior for a relatively broad spec-
trum. The experienced overestimation for lower slenderness ratios may occur
due to both inaccurate shear models and geometric imperfections.

21

Page 21 of 43
 Comparison between E*2 estimation methods for f cc
2 z

t
f2cc z

ip
2
*
2
*

cr
us
an
Figure 13: Comparison between developed models and experimental results from Merkt [1]

Rehme [8] experimented cell patterns f2 ccz , f cc and f2 bccz , obtaining scal-
M
ing laws for each one. Instead of maintaining the cell size constant, he xed
335 the strut thickness at 0.5mm. His results together with the approached models
are plotted in Figure 14. Like in the last comparison, both FEM and analytical
d

results match, verifying the latter. All geometries present the same behavior.
te

For lower slenderness ratios, the models have great deviations as high as 50%
for f cc at AR = 10, whereas at higher relative densities errors are lower than
20%. Possible reasons for this deviation at lower relative densities are mainly
p

340

constructive. The cellular solid tested in experiments was produced using point-
ce

exposure, that as seen in Section 2.3 produces struts with higher rugosity than
the contour-exposure method. Because the strut thickness (500 µm) is in the
same order of magnitude, the cross-sectional variation along its length becomes
Ac

345 more signicant. As the cell size varies with the slenderness ratio, these dier-
ences become increasingly more relevant. The values at lower relative densities
are thus the most aected, resulting in lower stiness as the idealized model.

6.2. Comparison of Topology Performances

The next step is to apply beam models for all geometries and load cases.
350 Because the variation of strut thickness is of greater relevance to local optimiza-

22

Page 22 of 43
 Comparison between E2* estimation methods

f2cc z fcc f2bcc z

t
ip
2

2
*

cr
us
2

2
*

an *
f2cc z fcc f2bcc z
M
Figure 14: Comparison between models and experimental equation obtained by Rehme [8]
d
tion than the variation of cell size, the diagrams shown in the current section
are established using parameters seen in Merkt's case [1] from Section 6.1 and
te

thus varying the strut thickness while maintaining a cell size of 2mm. In Figures
Figure 15 through 18, the engineering constants for each lattice geometry are
p

355 plotted. Eective elastic moduli in the three principal directions are displayed
ce

in Figure 15. The diagrams are plotted with both relative density and slender-
ness ratio, so that the impact of mass distribution in the cell may be seen. If
the designer wishes to select a cell with higher stiness-to-weight ratio, a lower
Ac

decay should be preferred in the relative density plot. The diagrams reveal that
360 the use of vertical struts reduce the devaluation of elasticity modulus caused by
mass decrease signicantly. This leads to a stiness increase of e.g. as high as 21
times between f cc and f ccz at relative density 2%. As a result, geometries f ccz ,
f2 ccz , bccz and f2 bccz possess the highest elasticity modulus in the y -direction.
Under transverse loading (xz ), although the presence of vertical bars increases
365 results for geometries with face-centered struts, geometry bccz does not increase

23

Page 23 of 43
 the elastic modulus signicantly. This happens because in the rst case struts
are connected so that vertical displacement is strongly coupled with transverse
displacements. However, the second case presents strut connection at the cell's

t
ip
center, allowing deformation in transverse direction with a smaller inuence of
370 coupling with displacements along the y -axis. As a result, this cell has stretch

cr
dominant behavior in the y -direction and bending dominant behavior in the
transverse direction.

us
an
M
d
p te
ce
Ac

Figure 15: Comparison of eective elasticity modulus in direction y and x/z between cellular
geometries

Diagrams containing shear moduli are depicted in Figure 16. It is possible

to conclude in the rst case G∗12,23 that all geometries possess the same magni-
375 tude considering relative density. Nevertheless, geometries f cc, f2 cc and f2 bcc
exhibit the highest values due to the presence of face-centered struts, which are
aligned to the applied load. Consequently, they are loaded mainly axially, hav-

24

Page 24 of 43
 ing therefore lower deformation. Dierently from this load case, the geometries
may be divided into two categories based on G∗13 . The ones with body-centered
380 struts show higher shear stiness due to the alignment with the load path, thus

t
ip
reducing displacements in this direction. Geometries f cc, f ccz , f2 cc and f2 ccz
have in contrast no struts aligned to applied loads, leading to magnitudes 100

cr
times lower.

us
an
M
d
p te
ce
Ac

Figure 16: Comparison of eective shear moduli in plane xy/yz and xz between cellular
geometries

Using the methodology described in Section 4 for obtaining strength data and
385 the results obtained in Rehme's study [8] for constitutive material properties,
the axial and shear strength data were assessed for all geometries. The results
are plotted in Figure 17 and Figure 18. It can be inferred that geometries with

25

Page 25 of 43
 vertical struts present the best Rp0.2y
∗
results, characterized by less inclined
slopes and higher values. In the transverse direction (Rp0.2x
∗
), however, f ccz
390 and f2 ccz presented the best results, while the other geometries had more or

t
ip
less similar slopes. For conguration bccz this happens because most of the load
is transfered to strut 2, so that it yields similarly to conguration bcc. In the

cr
case of conguration f2 bccz , this happens due to the number of struts inside the
cell, leading to thinner cross-sections and lower strengths. The distributions of
∗
reveal that body-centered struts presented lower values than the ones

us
395 Sp0.2xy
with face-centered and/or vertical struts. This happens because the latter has
struts loaded mainly through axial force, whereas geometries with bcc-struts

an
have it as critical due to bending. Shear strength Sp0.2xz
∗
has on the other hand
approximately the same slope for all geometries because all struts are loaded
400 through bending. Structures bcc and bccz present the highest values thanks to
M
the fact that they have body-centered struts, which has a shorter lever arm than
the others for this load case.
After characterizing elastic and strength properties of all studied cell ge-
d

ometries, the next step is to use these values to classify lattice structures and
te

405 thus discuss their application. The engineer should manage to choose the most
appropriate cell type for each individual application depending on mechanical,
process and geometrical properties. Possible tools to achieve this goal is to an-
p

alyze all scaling laws and identify consequences in the utilization of each strut.
ce

This is summarized in Table 2.

410 Aiming to better evaluate each cell geometry based on important criteria,
new parameters are chosen and represented in a radar plot (see Figure 19).
Ac

Complementing elastic and yield properties, stored energy at the yield point is
assessed. All these mechanical properties are given grades between 0 and 1 based
on their percolation exponent and sheer value. Anisotropy is also given a grade
415 based on the ratio between elasticity moduli in y and x/z directions, so that
higher anisotropies have higher grades. The remaining criteria are operational:
slenderness ratio range and build time. Geometries presenting lower slenderness
ratio transition from framework to solid with pores were better evaluated. As

26

Page 26 of 43
 t
ip
cr
us
an
M
d
p te
ce

Figure 17: Comparison of eective yield strength in y and x direction between cellular geome-
tries
Ac

27

Page 27 of 43
 t
ip
cr
us
an
M
d
p te
ce

Figure 18: Comparison of eective shear strength in plane xy/yz and xz between cellular
geometries
Ac

28

Page 28 of 43
 t
ip
Strut-Type Pro Contra
Face-

cr
centered • Auxetic behavior • Reduction of G∗13 and
Sp0.2xz∗
• Superior ∗
E1,3 and

us
∗
Sp0.2xy

• Slightly higher G∗12,23

an
and Rp0.2x
∗

Body-
M
centered • Isotropic when alone • Reduction of E1,3
∗
, E2∗
and Sp0.2xy
∗
• Superior G∗13 and
d
∗
Sp0.2xz • Slightly lower Rp0.2x
∗

and Rp0.2y
∗
te

Vertical
p

• Induce anisotrpy • Reduction of G∗13 ,

ce

∗
Sp0.2xy and Sp0.2xz
∗
• Increased E2∗ and
∗
Rp0.2y
Ac

• Higher E1,3
∗
and Rp0.2x
∗

with f cc-struts

Table 2: Evaluation of strut types based on mechanical properties

29

Page 29 of 43
 for the build time, it was estimated through the cross-sectional area of each cell.
420 Higher values would mean more hatch vectors and therefore longer build time,
meaning a lower grade. Analyzing the diagram in Figure 19, it is possible to infer

t
ip
that geometry f2 ccz has the best evaluation concerning mechanical properties,
which conrms ndings from Rehme [8].

cr
7. Summary and Conclusions

us
425 The main objective of the present work was to investigate the mechanical
properties of lattice structures, so that their use in additive manufacturing de-
sign may be better understood. The used analysis tools are closed-form analyt-

an
ical as well as nite element models, which are then compared to experimental
results extracted from the literature. As a result, relative density distributions
and mechanical properties scaling laws are assessed for open-celled cellular solids
M
430

with cubic unit cells. With these data, struts are evaluated according to their
eect on the overall structure and radar plots are generated to support engineers
d

in the cellular solid's design.

Because of the alignment to shear loading, geometries f cc, f2 cc and f 2ccz
te

435 present higher shear moduli and strength on planes xy and yz . Vertical struts
induce anisotropy by increasing stiness and strength in vertical direction, so
p

that the solid presents stretch-dominant behavior. If combined to face struts,

ce

transversal mechanical properties are also enhanced due to coupling between

displacements in x-, y - and z -Direction. Body-centered bars are on the other
440 hand the only one presenting quasi-isotropic mechanical properties when applied
Ac

in conguration bcc. It also increases shear moduli and strength in the xz -plane
due to the better alignment to applied forces whereas other geometries have
orders of magnitude as low as about 100 times. However, it has inferior Young's
moduli and yield strength in axial directions, so that its use together with other
445 strut types (f2 bcc and f2 bccz ) leads to decrease in mechanical properties in axial
directions when comparing solids with regard to relative density. The current
work was able to approach the determined objectives in the introduction, so that

30

Page 30 of 43
 t
ip
U U
Sp0.2,xz p0.2,xy Up0.2,xz Sp0.2,xz p0.2,xy Up0.2,xz
Sp0.2,xy Aniso Sp0.2,xy Aniso

cr
BT BT
Gxz Gxz
Range Range
Gxy Gxy
Ey Ey

us
Up0.2,x Up0.2,x
Ex Ex
Up0.2,y Up0.2,y
Rp0.2,xRp0.2,y Rp0.2,xRp0.2,y

U U
Sp0.2,xz p0.2,xy Up0.2,xz Sp0.2,xz p0.2,xy Up0.2,xz

an
Sp0.2,xy Aniso Sp0.2,xy Aniso

BT BT
Gxz Gxz
Range Range
Gxy Gxy
M
Ey Ey
Up0.2,x Up0.2,x
Ex Ex
Up0.2,y Up0.2,y
Rp0.2,xRp0.2,y Rp0.2,xRp0.2,y

U U
Sp0.2,xz p0.2,xy Up0.2,xz Sp0.2,xz p0.2,xy Up0.2,xz
d

Sp0.2,xy Aniso Sp0.2,xy Aniso

BT BT
te

Gxz Gxz
Range Range
Gxy Gxy
Ey Ey
p

Up0.2,x Up0.2,x
Ex Ex
Up0.2,y Up0.2,y
Rp0.2,xRp0.2,y Rp0.2,xRp0.2,y
ce

U U
Sp0.2,xz p0.2,xy Up0.2,xz Sp0.2,xz p0.2,xy Up0.2,xz
Sp0.2,xy Aniso Sp0.2,xy Aniso

BT BT
Gxz Gxz
Ac

Range Range
Gxy Gxy
Ey Ey
Up0.2,x Up0.2,x
Ex Ex
Up0.2,y Up0.2,y
Rp0.2,xRp0.2,y Rp0.2,xRp0.2,y

Figure 19: Radar diagrams with comparative notes between lattice cell geometries

31

Page 31 of 43
 unit cell geometries were classied according to their mechanical properties. The
engineer is therefore able to choose appropriate geometries for each application
450 through e.g. radar plots, which show in a diagram advantages and disadvantages

t
ip
of each cell type.
In future works, testing fabrication methods with LBM and experimenting

cr
the studied geometries is crucial in order to validate the used models and con-
clusions that were made in this work. Because many simplications were made,
this step is extremely important in order to identify errors. After obtaining

us
455

stress-strain curves for printed metals and better describing plastic properties
for the used LBM machine, post-yield analysis considering buckling are also

an
recommended to be run for all relevant load cases (especially for axial loads).
Energy absorption may be then attributed for cellular units and therefore con-
460 sidered in the engineer's design of lattice structures. A more complete study on
M
the impact of the staircase eect and rugosity on mechanical properties is also
important for adjusting manufacturing parameters as well as performing better
component designs. The transition between lattices and solid parts represent
d

stress concentration regions and should therefore be better optimized. These

te

465 studies should be supported by experimental data.

Appendix 1. Boundary Conditions

p
ce

Tables A.3 through A.5 show the boundary conditions for the corresponding
load cases where `0` means restriction of the degree of freedom (DOF) and `-`
means unconstrained.
Ac

32

Page 32 of 43
 DOF O P1 P2 P3 P4
u1 - 0 0 - -
- 0 0 0 0

t
u2

ip
u3 - - 0 0 -
Φi 0 0 0 0 -

cr
Table A.3: Assignment of boundary conditions for axial load case in cell's coordinate system
for the RV Esym depicted on Figure 11 (left)

us
DOF O P1 P2 P3 P4
u1 - 0 0 0 0
0 0 0 - -

an
u2
u3 0 0 0 0 0
Φ1 0 0 0 0 0
M
Φ2 0 0 0 0 0
Φ3 0 0 0 - -

Table A.4: Assignment of boundary conditions for shear load case xy in cell's coordinate
d

system for the RV Esym depicted on Figure 11 (left)

te

DOF O P1 P2 P3 P4
u1 - - 0 0 -
p

u2 0 0 0 0 0
- 0 0 - -
ce

u3
Φ1 0 0 0 0 -
Φ2 0 - 0 - -
Ac

Φ3 0 0 0 0 -

Table A.5: Assignment of boundary conditions for shear load case xz in cell's coordinate
system for the RV Esym depicted on Figure 11 (left)

33

Page 33 of 43
470 Appendix 2. Eective Elastic Constants
Appendix 2.1. Deformation Energy (General)

t
The deformation energy which is used as a basis for Castigliano's 2nd theo-

ip
rem can be determined as follows

cr
!
l
Nx2 M2 My2 V2 Mz2 Vy2
Z
1
Πi = + x + + z + + dx (B.1)
2 0 EA GIT EIyy kGA EIzz kGA

us
Appendix 2.2. Transformation Matrices
475 In order to obtain the normal and lateral force distribution in the strut

an
according to Figure 11 one must apply a force transformation using the posed
transformation matrices.
M
 
cos θz − sin θz 0
 
Ts1 =  sin θz (B.2)
 
cos θz 0
 
d
0 0 1
 
cos θy 0 sin θy
te

 
Tc1 = 0 (B.3)
 
1 0 
 
− sin θy 0 cos θy
p

 
cos θy cos θz sin θz − sin θy cos θz
ce

 
Tsc = − cos θy sin θz cos θz (B.4)
 
sin θy sin θz 
 
sin θy 0 cos θy

Appendix 2.3. Forces Transformation

Ac

The following equations show the corresponding normal and lateral forces
480 such as the moment's distribution in the struts.

   
P F
   x
Vy  = T sc · Fy  (B.5)
   
   
Vz Fz

34

Page 34 of 43
 P = cos θy cos θz Fx + sin θz Fy − sin θy cos θz Fz (B.6)

Vy = − cos θy sin θz Fx + cos θz Fy + sin θy sin θz Fz (B.7)

t
Vz = sin θy Fx + cos θy Fz (B.8)

ip
My = My0 + Vz x

(B.9)

cr
= (sin θy Fx + cos θy Fz )x + My0

Mz = Mz0 + Vy x

us
= (− cos θy sin θz Fx + cos θz Fy + sin θy sin θz Fz )x + Mz0 (B.10)

Appendix 2.4. Deformation Energy (Here)

an
Applying the forces of Appendix 2.3 to Equation B.1 yields

1 1
Πi = {l[−IAFx2 Gks l2 − IAFz2 Gks l2
6 EAks G
M
−3IAFz GMz0 sin θy sin θz ks l + 3IAFx GMz0 sin θz cos θy ks l

−IAFy2 G cos2 θz ks l2 + IAFz2 G cos2 θz ks l2 − IAFz2 G cos2 θz cos2 θy ks l2

−3IAFx GMy0 sin θy ks l − 3IAFy GMz0 cos θz ks l

d

−3IAFz GMy0 cos θy ks l + IAFx2 G cos2 θz cos2 θy ks l2

te

−3EFx2 cos2 θz cos2 θy + 3EFz2 cos2 θz cos2 θy

−3Fy2 G cos2 θz ks + 3Fz2 G cos2 θz ks − 3IMy0

2
Aks G
p

2
−3IMz0 Aks G − 3Fz2 G cos2 θz cos2 θy ks
ce

+6EFx Fz sin θy cos2 θz cos θy + 3EFz2 cos2 θz

+3Fx2 G cos2 θz cos2 θy ks + 3EFz2 + 3Fy2 Gks + 3EFx2 + 3EFy2 cos2 θz

Ac

−6Fx Fz G sin θy cos2 θz cos θy ks + 6Fx Fy G cos θz sin θz cos θy ks

−6Fy Fz G sin θy cos θz sin θz ks − 2IAFx Fz G sin θy cos2 θz cos θy ks l2

+2IAFx Fy G cos θz sin θz cos θy ks l2 − 2IAFy Fz G sin θy cos θz sin θz ks l2

−6EFx Fy cos θz sin θz cos θy + 6EFy Fz sin θy cos θz sin θz ]} (B.11)

with

35

Page 35 of 43
 ∂Πi
=0
∂My0

t
1 1

ip
⇒ My0 = − Fx sin θy l − Fz cos θy l (B.12)
2 2
∂Πi
=0

cr
∂Mz0
1 1 1
⇒ Mz0 = Fx sin θz cos θy l − Fz sin θy sin θz l − Fy cos θz l (B.13)
2 2 2

us
yields

an
1 1
Πi = {[{(Fx − Fz )(Fx + Fz )(G(IAl2 + 3)ks − 3E) cos2 θy
6 EAks G
−2(G(IAl2 + 3)ks − 3E) sin θy Fz Fx cos θy
1
M
− G((−12 − 4IAl2 )Fz2 + (IAl2 + 12)Fy2 )ks + 3EFy2 − 3EFz2 } cos2 θz
4
1
+ Fy sin θz (cos θy Fx − sin θy Fz )((IAl2 + 12)Gks − 12E) cos θz
2
3
+ Iks (Fx2 sin2 θz + Fz2 )Gl2 A cos2 θy
d

4
3
− Iks (sin θz − 1)G sin θy (sin θz + 1)Fz Fx l2 A cos θy
te

2
3
+ Iks G(Fz2 sin2 θz + Fx2 )l2 A sin2 θy
4
p

−(IAFx2 l2 + IAFz2 l2 − 3Fy2 )Gks + 3E(Fx2 + Fz2 )]l} (B.14)

ce

Appendix 2.5. Deformations According to Castigliano's 2nd Theorem

485 Using Castiglianos 2nd theorem delivers the deformations posed below. These
Ac

are used to calculate the strain of a lattice structure.

36

Page 36 of 43
 ∂Πi
ux =
Fx

t
(2 G(IAl2 + 3)ks − 3E cos2 θy cos2 θz + 23 Iks sin2 θz Gl2 A cos2 θy

ip
1 + 32 Iks Gl2 A sin2 θy − 2Il2 Aks G + 6E)lFx
ux =
6 EAks G

cr


1 sin θz cos θy (IAl2 + 12)Gks − 12E cos θz lFy
+
12 EAks G
(−2(G(IAl2 + 3)ks − 3E) sin θy cos2 θz

us
1 − 23 Iks (sin θz − 1)G sin θy (sin θz + 1)l2 A cos θy )lFz
+ (B.15)
6 EAks G

∂Πi

an
uy =
Fy
1 sin θz cos θy ((IAl2 + 12)Gks − 12E) cos θz lFx
uy =
12 EAks G
1 − 12 (IAl2 + 12)Gks + 6E cos2 θz + 6ks G lFy



M
+
6 EAks G
1 sin θy sin θz Fz ((IAl2 + 12)Gks − 12E) cos θz l
− (B.16)
12 EAks G
d

∂Πi
uz =
te

∂Fz
(−2(G(IAl2 + 3)ks − 3E) sin θy cos θy cos2 θz
1 − 23 Iks (sin θz − 1)G sin θy (sin θz + 1)l2 A cos θy )lFx
uz =
p

6 EAks G
2
1 sin θy sin θz ((IAl + 12)Gks − 12E) cos θz lFy
ce

−
12 EAks G
((−2(G(IAl + 3)ks − 3E) cos2 θy − 21 G(−12 − 4IAl2 )ks − 6E) cos2 θz
2

1 + 32 Iks Gl2 A cos2 θy + 23 Iks G sin2 θz l2 A sin2 θy − 2Il2 Aks G + 6E)lF z

+
Ac

6 EAks G
(B.17)

Appendix 3. Strut Plastication Analysis

Appendix 3.1. Transformation Matrices for the Material, Part, Cell and Strut
Coordinate System
490 Figures 9 and 10 show a RVE for a f2 bccz structure. The subsequent trans-
formation matrices are necessary to provide the interchange between the coor-

37

Page 37 of 43
 dinate systems. The indices are dened as follows: material (m), part (p), cell
(c), strut (s).

t
ip
 
cos θh − sin θh 0
 
(tmp )ij =  sin θh (C.1)
 
cos θh 0

cr
 
0 0 1
    
1 0 0 cos(− π2 ) 0 sin(− π2 ) 0 0 −1

us
    
(tpc )ij = 0 cos π2 − sin π2    = −1
     
0 1 0 0 0
    
π π π π
0 sin 2 cos 2 − sin(− 2 ) 0 cos(− 2 ) 0 1 0
(C.2)

an
 
cos θs cos ϕs − cos θs sin ϕs sin θs
 
(tcs )ij =  (C.3)
 
sin ϕs cos ϕs 0 

M 
− sin θs cos ϕs sin θs sin ϕs cos θs

Appendix 3.2. Stress State Transformation

Assuming a circular cross section of the strut the normal stress distribution
d
495

is given by
te

Nx Mzz y Myy z
p

σx,s (x, y, z, F ) = + +
A Izz Iyy
(C.4)
ce

σy,s = σz,s = 0

where F is the external load intensity, Nx , My and Mz the internal forces

and moments distribution. Shear stresses are caused by shear forces Vy and Vz
Ac

and the torsion moment Tx . This yields the shear stresses

2
64Vy ( d4 − y 2 ) Tx z
τxy,s (y, z, F ) = −
3πd4 Ip
2
64Vz ( d4 − z 2 ) Tx y
τzx,s (y, z, F ) = +
3πd4 Ip
τyz,s = 0 (C.5)

38

Page 38 of 43
500 This yields the stress state
 
σx (x, y, z, F ) τxy (y, z, F ) τxz (y, z, F )

t
 
σij,s =  τyx (y, z, F ) (C.6)
 
0 0

ip

 
τzx (y, z, F ) 0 0
In order to obtain the stress state from the material coordinate system one

cr
must apply the transformation matrices presented in Appendix 3.1 as follows:

us
σij,m (x, y, z, F ) = (tmp )ik (tpc )kl (tcs )lm σmj,s (x, y, z, F ) (C.7)

Appendix 3.3. Plastication Criterion

an
Considering the anisotropic material behavior Hill [21] presented a criterion
505 able to account for anisotropic behavior through a generalized Mises criterion:
M
Ca (σ11 −σ22 )2 +Cb (σ22 −σ33 )2 +Cc (σ11 −σ22 )2 +2Cd σ23
2 2
+2Ce σ31 2
+2Cf σ12 =1
(C.8)
d

where
te

   
1 1 1 1 1 1 1 1
Ca = + − , C = + − ,
p

2 2 2 b 2 2 2
2 σ2T σ3T σ1T 2 σ3T σ1T σ2T
 
1 1 1 1 1 1 1
ce

Cc = 2 + 2 − 2 , Cd = S )2
, Ce = S )2
, Cf = S )2
2 σ1T σ2T σ3T 2(τ23 2(τ31 2(τ12
(C.9)

Applying the stress state transformation presented in Equation C.7 nally

Ac

yields the Hill criterion in the relevant material coordinates

F S(x, y, z, F ) = Ca (σ11,m − σ22,m )2 + Cb (σ22,m − σ33,m )2 + Cc (σ11,m − σ22,m )2

2
+ 2Cd σ23,m 2
+ 2Ce σ31,m 2
+ 2Cf σ12,m =1 (C.10)

39

Page 39 of 43
 with F S being the factor of safety. When it is less than 1 the material is
510 in the elastic regime. Since all internal loads are linearly proportional to the
external force magnitude F it can be extracted that

t
ip
s
1
F (x, y, z) = (C.11)
F S(x, y, z, F = 1)

cr
In order to obtain yielding in the most critical point of the strut, the param-
eters x, y and z must be found who maximize this function. It can be inferred

us
from the fact of linear bending moment distribution along the strut and the
515 no-rotation boundary condition in the point O that the highest normal stress is
located at x = 0. To determine the other two dimensions a MATLAB function

an
has been written because they cannot be obtaines analytically. Substituting
this into Equation 16 nally yields
M
4{F (x, y, z)}min
σyield = (C.12)
a2

Appendix 4. Strut Buckling Analysis

d
te

520 To evaluate Equation 22 the load necessary to obtain strut buckling must
be determined. According to the boundary conditions discussed in Appendix 1
p

two buckling cases must be considered: column with xed ends σbuckling,f f and
xed at the base and pinned at the top σbuckling,f p . Substituting the equations
ce

presented by Gere [22] for critical load in Equation 16 yields

Ac

16πEI
σbuckling,f f =
a2 l2
8.184π 2 EI
σbuckling,f p = (D.1)
a2 l2

525 Rehme [8] mentions that the inuence of shear deformation in thicker struts
should be considered which can be done through

σbuckling
σbuckling,shear = kbs σbuckling
(D.2)
1+ AG

40

Page 40 of 43
 with kbs being a shear correction factor that equals 1.11 for a circular cross
section [20].

t
References

ip
530 [1] S. J. Merkt, Qualizierung von generativ gefertigten gitterstrukturen für

cr
maÿgeschneiderte bauteilfunktionen, Dissertation, RWTH Aachen, Aachen
(2015).

us
[2] L. J. Gibson, Cellular solids: Structure and properties, second edition Edi-
tion, Cambridge solid state science series, Cambridge University Press,

an
535 Cambridge, 1997.

[3] I. Fogelman, G. Gnanasegaran, H. van der Wall (Eds.), Radionuclide and

hybrid bone imaging, Springer, Heidelberg, 2012.
M
[4] J. Kovacik, F. Simancik, Aluminium foammodulus of elasticity and
electrical conductivity according to percolation theory, Scripta Materialia
d

540 39 (2) (1998) 239246. doi:10.1016/S1359-6462(98)00151-1.

te

[5] C. Emmelmann, P. Scheinemann, M. Munsch, V. Seyda, Laser additive

manufacturing of modied implant surfaces with osseointegrative character-
p

istics, Physics Procedia 12 (2011) 375384. doi:10.1016/j.phpro.2011.

03.048.
ce

545 [6] Y. Chen, T. Li, Z. Jia, F. Scarpa, C.-W. Yao, L. Wang, 3d printed hierar-
chical honeycombs with shape integrity under large compressive deforma-
Ac

tions, Materials & Design 137 (2018) 226234. doi:10.1016/j.matdes.

2017.10.028.

[7] L. R. Meza, S. Das, J. R. Greer, Strong, lightweight, and recoverable three-

550 dimensional ceramic nanolattices, Science 345 (6202) (2014) 13221326.

[8] O. Rehme, Cellular Design for Laser Freeform Fabrication, 1st Edition, Vol.
v.4 of Schriftenreihe Lasertechnik, Cuvillier Verlag, Göttingen, 2010.

41

Page 41 of 43
 URL https://ebookcentral.proquest.com/lib/gbv/detail.action?
docID=5020518

t
555 [9] S. L. Campanelli, N. Contuzzi, A. D. Ludovico, F. Caiazzo, F. Cardaropoli,

ip
V. Sergi, Manufacturing and characterization of ti6al4v lattice components
manufactured by selective laser melting, Materials (Basel, Switzerland)

cr
7 (6) (2014) 48034822. doi:10.3390/ma7064803.

[10] S. Demiray, W. Becker, J. Hohe, Strain-energy based homogenisation of

us
560 two- and three-dimensional hyperelastic solid foams, Journal of Materials
Science 40 (22) (2005) 58395844. doi:10.1007/s10853-005-5017-6.

an
[11] M. Burns, Automated fabrication: Improving productivity in manufactur-
ing, PTR Prentice Hall, Englewood Clis, N.J., 1993.

[12] A. Gebhardt, J.-S. Hötter, Additive Manufacturing: 3D Printing for Pro-

M
565 totyping and Manufacturing, 2016.
URL http://dx.doi.org/10.3139/9781569905838
d

[13] Verein Deutscher Ingenieure, Vdi 3405, additive fertigungsverfahren: grund-

lagen, begrie, verfahrensbeschreibungen.
te

[14] L. Liu, P. Kamm, F. García-Moreno, J. Banhart, D. Pasini, Elastic and

p

570 failure response of imperfect three-dimensional metallic lattices: The role

of geometric defects induced by selective laser melting, Journal of the Me-
ce

chanics and Physics of Solids 107 (2017) 160184. doi:10.1016/j.jmps.

2017.07.003.
Ac

[15] I. Maskery, A. O. Aremu, L. Parry, R. D. Wildman, C. J. Tuck, I. A.

575 Ashcroft, Eective design and simulation of surface-based lattice structures
featuring volume fraction and cell type grading, Materials & Design 155
(2018) 220232. doi:10.1016/j.matdes.2018.05.058.

[16] A. Aharony, D. Stauer, Introduction To Percolation Theory, 2nd Edition,

Taylor and Francis, Hoboken, 2003.
580 URL http://gbv.eblib.com/patron/FullRecord.aspx?p=167644

42

Page 42 of 43
 [17] V. Silva, Mechanics and strength of materials, Springer, Berlin and New
York, 2006.

t
[18] A. Groÿmann, J. Souza, C. Mittelstedt, Mechanics of cellular solids and

ip
lattices and their lightweight potential in additive manufacturing, SAMPE
585 Conference Proceedings 14-16 Nov 2017.

cr
[19] A. Iltchev, V. Marcadon, S. Kruch, S. Forest, Computational homogenisa-
tion of periodic cellular materials: Application to structural modelling,

us
International Journal of Mechanical Sciences 93 (2015) 240255. doi:
10.1016/j.ijmecsci.2015.02.007.

an
590 [20] S. P. Timoshenko, J. M. Gere, Theory of Elastic Stability, Dover Civil and
Mechanical Engineering, Dover Publications, Newburyport, 2012.
URL http://gbv.eblib.com/patron/FullRecord.aspx?p=1894787
M
[21] R. Hill, A theory of the yielding and plastic ow of anisotropic metals, Pro-
ceedings of the Royal Society A: Mathematical, Physical and Engineering
d

595 Sciences 193 (1033) (1948) 281297. doi:10.1098/rspa.1948.0045.

te

[22] J. M. Gere, S. P. Timoshenko, Mechanics of materials, 3rd Edition, Chap-

man & Hall, London, 1991.
p
ce
Ac

43

Page 43 of 43