You are on page 1of 10

J Mater Sci (2013) 48:1413–1422

DOI 10.1007/s10853-012-6892-2

Modeling of uniaxial compression in a 3D periodic re-entrant


lattice structure
Li Yang • Ola Harrysson • Harvey West •

Denis Cormier

Received: 23 February 2012 / Accepted: 11 September 2012 / Published online: 25 September 2012
Ó Springer Science+Business Media, LLC 2012

Abstract In this study, the behavior of a parametric 3D great deal of attention. Auxetic structures are expected to
re-entrant dodecahedron lattice structure with negative exhibit excellent shear stiffness [1–4] and indentation
Poisson’s ratio was studied. Four geometrical configura- resistance [1, 5, 6], high fracture toughness [3, 7], good
tions for the re-entrant dodecahedron were designed, and energy dissipation ability [8–11], and unique acoustic
the relationship between the mechanical properties and the absorption abilities [12–14]. Therefore, auxetic structures
design parameters was determined through beam theory. possess significant potential to improve performance in a
Samples were fabricated successfully via electron beam variety of applications such as sandwich panel cores,
melting. Compressive tests as well as finite element anal- energy and sound damping structures, radome frames [13],
ysis (FEA) were performed, and the results were compared aerospace filler foams, and biomedical implants.
with theoretical predictions. The modeling yielded explicit Several researchers have established models for auxetic
analytical equations of various mechanical properties structures. Almgren [15] proposed a 3D truss structure with
including Poisson’s ratios, modulus and strength, and the springs and hinge joints in which Poisson’s ratio was -1.
compressive strength and the modulus from the prediction Lakes [1] proposed a process to produce auxetic structures,
match well with the experiments, as well as the FEA and idealized the structure into a 3D representation to
results. The methodology used by this study also demon- explain the material’s behavior. Following the lead of
strated a feasible approach to design 3D auxetic cellular Lakes’ seminal work, various models have been created to
structure for various applications. explain the auxetic behavior and to verify the experimental
observations [16–23]. According to the auxetic mechanism,
the most models can be categorized into two types: rigid
Introduction nodule/flexible fibril and rigid nodule/hinge joint. The rigid
nodule/flexible fibril model is used to describe the auxetic
Auxetic structures are unlike conventional cellular mate- behavior of structures made from rapid drawing of sintered
rials in the sense that they exhibit negative Poisson’s ratios PTFE, because of the formations of large stiff nodules and
in one or more directions. Since auxetic structures were coalescing fibrils in the microstructure during the process
first described, their unique properties have attracted a [21, 24].
The rigid nodule/hinge joint model seeks to explain the
auxetic behavior of structures made from the traditional
L. Yang  O. Harrysson (&)  H. West heating and isotropic pressing process [1]. Multiple
Edward P. Fitts Department of Industrial & System Engineering,
research groups presented both 2D and 3D theoretical
North Carolina State University, 400 Daniels Hall, 111 Lampe
Dr., Raleigh, NC 27695, USA models of hinge connected nodules involving triangles or
e-mail: harrysson@ncsu.edu squares, and auxetic behavior is predicted accordingly [17,
18]. In a more recent study, researchers also experimented
D. Cormier
with an ‘‘egg-rack’’ single-layer auxetic lattice structure
Department of Industrial & System Engineering, Rochester
Institute of Technology, James E. Gleanson Building, 81 Lomb [19]. Through experimentation, it was shown that the egg-
Memorial Dr., Rochester, NY 14623, USA rack structure exhibited auxetic behavior.

123
1414 J Mater Sci (2013) 48:1413–1422

Some researchers have taken an approach of construct- structure. Experimental based property analysis was per-
ing truss structures with rigid joints. For instance, Theoc- formed and showed promising results for this type of
aris et al. [22] proposed a mathematical model in which the structure.
auxetic behavior is realized through a star-shaped inclusion This study was based upon a 3D array of the re-entrant
in the structure. Similarly, Larsen et al. [23] proposed a dodecahedron unit cell structure described elsewhere in the
double-arrow model in which the auxetic behavior is literature [15, 26, 28]. In a single isolated unit cell, indi-
realized through inward rotation of re-entrant ribs upon vidual struts are not shared by an adjacent unit cell. This is
loading. Wan et al. focus on applications in which the rib in contrast with a 3D array of unit cells in which interior
joints are rigid. Under this condition, they proposed a 2D vertical struts are shared with adjacent unit cells, whereas
analytical re-entrant lattice model based on a large exterior (or edge) vertical struts are not. The mechanical
deflection assumption [25]. Their work established a set of properties of an isolated unit cell are therefore known to be
equations that reveals the relationship between structural quite different from those of an array of unit cells. This is
parameters and the resulting Poisson’s ratio which is highly the so-called ‘‘edge effect’’ [29]. Much of the literature
relevant to this study and will be further discussed in the devoted to the study of auxetic structures has focused on
‘‘theory’’ section. Evans et al. [26] estimated the perfor- either 2D representations or 3D analysis of individual unit
mance of a 3D dodecahedron auxetic structure using finite cells. It is also the case that until recently, it was not
element analysis (FEA). Through FEA, they examined the possible to experimentally validate these models, as there
modulus and Poisson’s ratio of several designs and was no way to produce alternative auxetic geometries for
observed significant changes in mechanical properties of testing in which the strut sizes and angles could be pre-
the 3D auxetic structure with changes in geometrical cisely varied.
design parameters. With that said the aim of this article was to model and
The majority of research involving auxetic structures to experimentally validate the properties of 3D arrays of
date has been focused on 2D structures. There has been metallic re-entrant dodecahedrons in which very specific
relatively little analytical modeling of 3D arrays of auxetic design alternatives are produced. Validated analytical 3D
unit cells. One practical reason for this has been the fact lattice models can be used to design mechanical compo-
that 3D auxetic structures with specifically engineered nents of which strength or stiffness, for example, are tuned
geometries have been difficult to manufacture until very to meet the needs of the engineering application. In this
recently. The most prevalent method for producing auxetic article, the relationships between design parameters and
structures to date was developed by Lakes in 1987. Lake’s mechanical properties of the structure have been developed
method [1] includes three basic steps: as sets of equations. Through physical tests and FEA, the
validity of the models has been established.
1. Place a conventional (typically polymer) foam struc-
ture into a mold, and heat it up to a ‘‘softening’’
temperature;
Theory
2. Isotropically compress the structure at the elevated
temperature;
Unit cell design
3. Allow the structure to cool down while in the
compressed state, and then release it from the mold
The re-entrant lattice structure considered in this article has
after it has cooled down.
the symmetrical anisotropic geometry shown in Fig. 1a.
The procedure involves the use of a mold that can be This structure was selected because it can be readily
isotropically compressed. The technique is relatively extended into 3D. Figure 1b illustrates the 3D re-entrant
straightforward, although it is difficult to produce samples unit cell, and Fig. 1c shows a 3D lattice block array based
with more complex geometries. Furthermore, the use of a on the unit cell. From Fig. 1b, it is apparent that with the
regular stochastic foam structure as the preform greatly cubic geometry bounding box, the unit cell could be readily
restricts the degree of spatial control over physical prop- patterned in all the three primary directions and thus fill up
erties of parts fabricated via this process. Recently, the volume.
Schwerdtfeger et al. [27] approached this issue using the For this study, there are four primary design parameters
electron beam melting (EBM) process and successfully for the re-entrant lattice structure: length of the vertical
produced Ti6Al4V samples that exhibit negative Poisson’s struts (H), the length of the re-entrant struts (L), the
ratios. In their work, a 3D lattice structure is developed re-entrant angle (h), and the square strut cross section
from the 2D re-entrant structure originally proposed by thickness (t). Although struts having a square cross section
Almgren [15]. The resulting 3D structure exhibits struc- are assumed throughout the calculations, results using other
tural anisotropy similar to that of a hexagonal honeycomb cross sectional geometries are readily obtained using the

123
J Mater Sci (2013) 48:1413–1422 1415

(a) 2D re-entrant lattice (b) Unit cell of 3D re-entrant lattice (c) 3D re-entrant lattice
Fig. 1 Re-entrant lattice structure

same procedures. Figure 2 illustrates these design param- remote compressive stress. Practically speaking, Eq. (1) is
eters for the structure. only valid when the re-entrant angle of the beam structure
Consider a remote compressive stress applied in the is sufficiently large ([10° depending on the L/t ratio).
y-axis direction on an infinite 3D re-entrant lattice structure. When h becomes very small, all of the struts in the cell are
Due to the symmetry of the structure, each vertical strut at either vertical or nearly vertical. This type of lattice has
the corner of a unit cell is shared by the three neighboring almost no room to collapse inward when subjected to a
unit cells. The compressive force applied to the vertical edge compressive load and will therefore behave more like a
struts is therefore distributed across the four adjacent unit bulk solid than a lattice material. In practice, there would
cells. The force on the central vertical strut is not shared by be little reason to design a lattice material with such acute
any other unit cells. As a result, the force components on strut angles, hence the analysis presented here is suitable
each vertical strut due to the compressive stress could be for the vast majority of applications.
shown as illustrated in Fig. 1b, where we have Due to the symmetry of the structure, an analysis of the
1 unit cell mechanics can be illustrated as shown in Fig. 3a,
F ¼ rð2L sin hÞ2 ¼ 2rL2 sin2 h; ð1Þ b, where T and P corresponded to the normal and shear
2
force on the re-entrant strut, and M corresponded to the
where r is the remote compressive stress. Due to the moment on the re-entrant strut. Again, all external torsion
in-plane symmetry of the unit cell, all external torsional and bending moments balance out. Therefore, the analysis
and bending moments balance out. Therefore, Fig. 1b fully of the structure’s loading is reasonably straightforward.
demonstrates the loading situation of a unit cell under
Deformation of the structure

The primary deformation mode of open-cell structures is


by elastic bending of the struts [30]. The force components
of the simplified structure are shown in Fig. 3b upon the
decomposition of the structure. The relationships between
these force components are
F
P¼ sin h ð2Þ
4
F
T ¼ cos h ð3Þ
4
PL FL
M¼ ¼ sin h: ð4Þ
2 8
It is obvious that the vertical struts are only subject to
Fig. 2 Design parameters for re-entrant lattice structure normal compressive stress, whereas the re-entrant struts are

123
1416 J Mater Sci (2013) 48:1413–1422

(a) Simplified structure (b) Decomposition of force components


Fig. 3 Analysis of the simplified structure

subject to both normal and shear stress. According to


classic beam theory, the axial compression of re-entrant
struts is negligible due to the bending. On the other hand,
the shear strain introduced by the shear stress on the
re-entrant strut might need to be considered when the strut
length is short. Therefore, the deformation of the simplified
structure is the linear superposition of three components:
(1) Compressive deformation of the vertical struts, Fig. 4 Deflection angles of the re-entrant strut
(2) Bending-induced deflection from the re-entrant strut,
and
(3) Shear-induced deflection from the re-entrant strut.
The compressive deformation of the vertical struts (Dy1)
can be readily shown to be
2rHL2 sin2 h
Dy1 ¼ ð5Þ
Et2
For the re-entrant strut, consider the boundary conditions
introduced by joints at both ends. The deflection angles h1
and h2 shown in Fig. 4 are equal under these conditions.
Fig. 5 Demonstration of the size change
From Timoshenko beam theory, the angle of deflection of
the re-entrant strut h0 could be expressed as
dx
h0 ¼ þ c; ð6Þ Coupling the result that h1 = h2 with Eqs. (6–8), the
dx maximum deflection angle could be expressed as
where x is the deflection of the strut and c is the shear ML 6P
strain, which in turn have the following relationships: h1 ¼ h2 ¼ þ ; ð9Þ
6EI 5GA
dx ML
¼ ð7Þ where I is the second moment of inertia of the cross
dx EI
section.
P
c¼ ; ð8Þ Since the joint is assumed to be rigid, the re-entrant
jGA angle should remain unchanged following deformation.
where j is the geometrical factor, G is the shear modulus of Following the similar approach used by Onck et al. [31],
the material, and A is the cross sectional area of the strut. In the size change of the structure caused by the deflection of
the case of a rectangular cross section, j = 5/6. the re-entrant strut as shown in Fig. 5 can be expressed as

123
J Mater Sci (2013) 48:1413–1422 1417

Dx ¼ L sinðh þ h1 Þ  L sin h  Lh1 cos h ð10Þ


Dy2 ¼ L cos h  L cosðh þ h1 Þ  Lh1 sin h ð11Þ
Note that the analysis does not distinguish between the x1
and x2 directions, as the deflection along both axes should
be identical due to the symmetry.

Poisson’s ratio

Since the equations for forces acting upon a unit cell can
be used to analyze the behavior of an entire re-entrant
Fig. 6 Poisson’s ratio with different design parameters t = 1 mm,
lattice structure, it is possible to estimate the Poisson’s
L = 5 mm
ratio for a particular lattice design. Due to symmetry of
the structure, Poisson’s ratio for both myx1 and myx2 can be
approximately -6 can be achieved with parameters of
expressed as
t = 1 mm and L = 5 mm, as shown in Fig. 6. Through
ex DxðH  L cos hÞ proper selection of the geometric parameters, it is possible
myx1 ¼ myx2 ¼  ¼ : ð12Þ
ey ðDy1 þ Dy2 ÞðL sin hÞ to tailor the Poisson’s ratio for a 3D re-entrant lattice
structure to suit the needs of a particular application.
Substituting Eq. (1) through Eq. (11) into Eq. (12), the
Poisson’s ratio for both directions can be expressed as
 2  Effective modulus
L 3
4Et4 þ 10Gt2 cos hða  cos hÞ
myx1 ¼ myx2 ¼  a
 L2 3
 2 ; ð13Þ Based on Eqs. (5) and (11), the effective modulus of a 3D
Et2 þ 4Et4 þ 10Gt2 sin h re-entrant lattice structure is obtained as
where a = H/L. r rðH  L cos hÞ ða  cos hÞ
In cases where the cross sectional strut size (t) is much Ey ¼ ¼ ¼ 2 2  4 2 
;
ey Dy1 þ Dy2 2aL sin
2
h
þ L 4 þ 3L 2 sin4 h
Et 2Et 5Gt
smaller than the length of the re-entrant strut (L), the strut
could be simplified as an Euler–Bernoulli beam, where the ð15Þ
shear-induced deflection can be ignored. Under this case, where Ey is the effective modulus and E is the modulus of
there also exists 4at2  L2sin2h indicating that the com- the solid material. As previously described, when
pression of the vertical strut is negligible. Hence, Eq. (13) 4at2  L2sin2h (i.e., compression of vertical struts is
could be further simplified into negligible), Eq. (15) simplifies to
cos hða  cos hÞ  t 4
myx1 ¼ myx2 ¼ ð14Þ Ey ¼ 2ða  cos hÞ E ð16Þ
sin2 h L sin h
Equation (14) has the same form as the equation proposed It could be seen from Eq. (16) that when the compression
previously [25] for a 2D re-entrant model. This coincidence of the vertical struts is not negligible, the modulus of the
is readily explained by the symmetry of the structure. If the structure is greatly compromised.
contribution of the vertical strut compression is ignored,
the deformation of the structure is completely determined
by the in-plane deflection of the re-entrant struts. There-
fore, the design parameters H, L, and h could fully describe
the behavior of both the 2D and 3D re-entrant lattice
structure considered in the article.
Figure 6 plots the Poisson’s ratio as a function of a and
h according to Eq. (13). The material property values of
Ti–6Al–4V are used for the plot, since the same type of
material was used for the experimental verification work.
Therefore, the values used for the elastic modulus (E) and
the shear modulus (G) are 114 and 43 GPa, respectively. It
is apparent that the Poisson’s ratio becomes increasingly
negative as the H/L ratio increases and the re-entrant Fig. 7 Normalized effective modulus versus lattice geometry
angle h decreases. A negative Poisson’s ratio value of t = 1 mm, L = 5 mm

123
1418 J Mater Sci (2013) 48:1413–1422

The effective modulus as a function of the lattice


geometry is graphed in Fig. 7, using Eq. (15). The nor-
malized effective modulus value is defined as Ey/Es. It is
readily seen that the effective modulus increases as the
re-entrant angle decreases and the H/L ratio increases.
According to Gibson’s theory, the modulus of an open-
cell structure is expressed as [30]
 n
E q
¼ C1 ¼ C1 qnr ; ð17Þ
Es qs
where Es and qs are the modulus and density of the solid
bulk material and qr is the relative density of the cellular Fig. 8 Structural constant C1 versus lattice geometry t = 1 mm,
L = 5 mm
structure. C1 is a structure specific constant that describes
the sensitivity of the modulus to changes in relative density.
For regular stochastic open-cell cellular structures C1 & 1. buckling and plastic failure. Elastic buckling is most likely
The relative density of the 3D re-entrant lattice structure to occur on the vertical struts, while plastic failure is more
is written as likely to occur on the re-entrant struts.
According to the elastic theory, beams can lose stability,
t2 2 2
4  4H þ t2 H þ t2  4H þ t2  8L þ t2  16L  20t3 buckle, and elastically fail under axial compressive stres-
qr ¼
2ðH  L cos hÞ  ð2L sin hÞ2 ses. Consider the case of a vertical strut in the 3D re-entrant
H þ 4L  5t  t 2 lattice structure. Since both ends of the strut are fixed, the
¼ 2
ð18Þ critical compressive force necessary to initiate buckling
2 sin hðH  L cos hÞ L
(Fcr) is given by Euler’s equation as
In Eq. (18), the last term in the numerator takes the volume of
4p2 EI
the strut intersections into account and is a generalized Fcr ¼ ; ð21Þ
approximation. When the strut cross sectional dimension t is l2
much smaller than H and L, Eq. (18) can be further simplified as where Fcr is determined by the material and the beam
aþ4  t 2 dimensions. I indicates the smallest moment of inertia in
qr ¼ 2
ð19Þ the cross sectional plane. In the case of a square cross
2 sin hða  cos hÞ L
section, I = t2/12. From Fig. 3, the maximum total force
From Eqs. (16) and (19), it is obvious that when for elastic buckling (Felastic) is therefore
4at2  L2sin2h, Eq. (17) also holds true for the 3D re-entrant p2 Et4
lattice structure. Simple rearrangement of terms gives Felastic ¼ : ð22Þ
3H 2
Ey 8ða  cos hÞ3 2 The corresponding compressive stress is therefore
¼ C1 q2r ¼ qr ð20Þ
Es ða þ 4Þ2
Felastic p2 Et4
2 2 2 relastic ¼ ¼ ð23Þ
Note that when 4at  L sin h in Eq. (20), the relative 2L sin h 6H L2 sin2 h
2 2 2

density of the structure is quite small. The resulting modulus


Since the re-entrant strut is subject to the combination of
will likewise be small. Therefore, the simplification condi-
in-plane normal and shear stresses, the principal stresses
tion is regarded as a low density simplification condition.
can be written as
The relationship of the structural constant C1 with the H/ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L ratio and h is shown in Fig. 8. With increasing H/L ratio ry r2y
and/or re-entrant angle h, the sensitivity of the modulus to r1 ¼ þ þ s2y : ð24Þ
2 4
the relative density increases. This indicates that with
proper design parameters, the modulus of the re-entrant Considering that the re-entrant strut has non-zero axial
lattice can be tailored over a wide range by changes in stress, the distribution of the normal stress on the cross
relative density. One simple way of achieving this is by section is illustrated in Fig. 9. The shift of the neutral plane
changing the strut cross sectional dimension t. position is determined by establishing force equilibrium in
the axial direction.
Yield strength When the re-entrant strut starts to yield, failure first
occurs at the upper surface as marked by location A in
Under compressive stress, the 3D re-entrant lattice struc- Fig. 9. It then propagates through the whole cross section
ture is subject to two possible failure modes: elastic as the compressive force increases. When the stress reaches

123
J Mater Sci (2013) 48:1413–1422 1419

Fig. 9 Distribution of normal stress on the cross section

the yield point, the structure fails. This sequence is shown (a) Elastic buckling
in Fig. 10. Therefore, at the moment of complete yielding,
the equilibrium equations are given as
rm tðt1  t2 Þ ¼ T ð25Þ
t1 þ t2 ¼ t ð26Þ
Zt=2 Zt=2 t1Zt=2

M¼ rm tsds þ rm tsds  rm tsds ð27Þ


t1 t=2 0 0

Here, rm is the maximum normal stress. From Eq. (24), the


maximum principal stress then becomes
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (b) Plastic failure
rm r2m
r1 ¼ þ þ s2y : ð28Þ Fig. 11 Maximum normalized stress level for the failure of the
2 4
When the yield point is reached, r1 = rY, and rm can be structure t = 1 mm, L = 5 mm
written in terms of rY as
r2Y  s2m
rm ¼ ð29Þ equation, the force needed to induce plastic failure (Fplastic)
rY
can be obtained. The corresponding maximum compressive
The maximum shear stress on the square cross section at a stress is then
given shear force T is
Fplastic
3P 3P 3F rplastic ¼ ð32Þ
sm ¼ ¼ ¼ sin h ð30Þ 2L2 sin2 h
2S 2t2 8t2
From Eqs. (23), (31), and (32), the relationship between the
Substituting Eqs. (29) and (30) into Eq. (27), and equating maximum stress level and design parameters H/L and h are
the expression for M with Eq. (4), we have shown in Fig. 11. In Fig. 11b, the maximum stress levels
 
2 2
r2Y  9F64tsin4 h t3 have been normalized by the bulk solid yield strength
F 2 cos2 hrY FL values in order to eliminate the effect of material type. For
M¼   ¼ sin h
4rY 2
64r2Y  9F tsin
4
2
h
t 8 simplicity of comparison, the stress value for elastic
buckling is also normalized by the bulk solid yield
ð31Þ
strength. Again Ti–6Al–4V was used as the material, where
Equation (31) is an implicit equation involving F, and can E and rY have been taken as 114 GPa and 1050 MPa,
be solved by simple iteration calculation. By solving the respectively, in the theoretical calculation.

Fig. 10 Progression of yielding


across the re-entrant strut’s
cross section

123
1420 J Mater Sci (2013) 48:1413–1422

Figure 12 compares the ratio of maximum elastic The yield strength of a regular open-cell cellular struc-
buckling stress to yield stress versus the lattice geometric ture can be written as [30]
parameters. As can be seen, elastic buckling is the domi- r
¼ C2 qnr ¼ C2 q3=2 ð33Þ
nant failure mode when the H/L ratio a is large and the rs r

re-entrant angle h is small. Combining elastic and plastic


where rs is the yield strength of the bulk solid material, and
failure modes, the failure surface for a given parameter
C2 is a constant that is determined by the structure’s H/
combination can be seen in Fig. 13.
L ratio, t and h. For regular stochastic open-cell structures,
C2 & 0.23 [30]. For the 3D re-entrant lattice structure
considered in this article, Eq. (31) is in implicit form. There
is no analytical expression to evaluate n and C2 directly.
Since Eq. (33) is a generic equation for open-cell struc-
tures, the n value of a 3D re-entrant lattice structure is
taken as 1.5 [30].
The instantaneous C2 values for different lattice design
parameters are shown in Fig. 14. From Fig. 14, it is obvi-
ous that the constant C2 becomes larger as the H/L ratio
increases and/or the re-entrant angle h decreases. C2 is
typically significantly larger than 0.23 [30], which clearly
indicates that the normalized yield strength of the
Fig. 12 Comparison of two failure modes

Fig. 13 Failure surface of 3D re-entrant lattice structure t = 1 mm, Fig. 14 Values of constant C2 with different design parameters
L = 5 mm t = 1 mm, L = 5 mm

Table 1 Parameters of different designs for compressive test


Designed sample dimensions
Design H (mm) L (mm) h (°) H/L t (mm) RD (%) PR value

1 5.5 4.2567 70 1.29 0.8 12.77 -0.367


2 4.125 3.1925 70 1.29 0.8 19.96 -0.367
3 7.737 3.7802 45 2.05 0.8 23.30 -1.899
4 6.457 3.1537 45 2.05 0.8 31.09 -1.899

Measured Ti–6Al–4V sample dimensions (average)

1 5.423 ± 0.007 4.267 ± 0.002 70 1.271 ± 0.002 0.79 (0.87) 13.65 ± 0.25
2 4.052 ± 0.005 3.213 ± 0.003 70 1.261 ± 0.001 0.79 (0.87) 22.93 ± 0.07
3 7.620 ± 0.008 3.807 ± 0.003 45 2.002 ± 0.010 0.79 (0.87) 22.23 ± 0.43
4 6.381 ± 0.036 3.187 ± 0.005 45 2.001 ± 0.004 0.79 (0.87) 31.22 ± 0.54

123
J Mater Sci (2013) 48:1413–1422 1421

re-entrant lattice structure could be tailored across a wide in this case. Four different designs were used with geo-
range by changing the relative densities. metric parameter values as shown in Table 1. Four repli-
cations of each design configuration were built. Designs 1
and 2 consisted of 4 9 4 9 4 unit cell arrays, while
Experimental methods and results designs 3 and 4 had 4 9 4 9 3 unit cell arrays due to their
large relative dimension in the third direction.
Ti–6Al–4V compressive samples were fabricated using the Design variations 1 and 2 in Table 1 were structured to
Arcam EBM process in order to evaluate the validity of the possess the same Poisson’s ratio. Design variations 3 and 4
models. The Ti–6Al–4V parts made from EBM are quite were structured to possess the same Poisson’s ratio as
stiff, and therefore the small deflection assumption applies shown in Table 1. Design variations 2 and 3 have similar
relative densities, thus allowing comparisons to be made
among different groups in order to estimate mechanical
properties. The measured dimensions of fabricated samples
varied slightly as shown in Table 1, but showed good
consistency overall. Therefore, the discussion of compari-
son among groups would still be valid. Note that there are
two average values for the strut dimensions, which indi-
cates that the dimensions for the vertical and re-entrant
struts were not identical. In Table 1, the strut sizes for the
vertical struts are shown in parenthesis in order to distin-
guish them from the values for the re-entrant struts.
The compressive tests were performed using an Applied
Test Systems 1620C. The samples were placed between
(a) Yield strength two tool steel plates and compressed at a constant strain
rate of 1.27 mm/min. The strain during the compression
was measured with an extensometer attached to the two
platens, and the stress was obtained by dividing the applied
force by the original geometry bounding areas.
The test results for both yield strength and effective
modulus are shown in Fig. 15. The theoretical predictions
match the experimental results quite well, particularly
considering the variability in strut size and surface quality
produced in the manufacturing process. As shown in
Fig. 16, the strut surface produced by the EBM process is
not smooth. To accurately determine the actual strut sizes
(b) Effective modulus is therefore a challenge. The approach employed here is to
find the maximum cross section distance (d) between par-
Fig. 15 The comparison between different results t = 0.97 mm allel lines that fall entirely within the strut as shown in

(a) Surface of a strut (b) Evaluation of the effective strut size


Fig. 16 Strut size of the parts made by EBM

123
1422 J Mater Sci (2013) 48:1413–1422

Fig. 16b. The intent of this approach was to prevent surface References
roughness from making the strut diameter appear larger
than it actually is. 1. Lakes R (1987) Science 235:1038
2. Scarpa F, Tomlin PJ (2000) Fatigue Fract Eng Mater Struct
FEA was also performed for the four designs. Solid- 23:717
Works was used to create the CAD model of the different 3. Lakes RS (1993) J Mech Des 115:696
designs, and COSMOS was used to simulate the mechan- 4. Scarpa F, Tomlinson G (2000) J Sound Vib 230:45
ical behavior in compression. The results of the FEA cal- 5. Lakes RS, Elms K (1993) J Compos Mater 27:1193
6. Alderson KL, Simkins VR, Coenen VL, Davies PJ, Alderson A,
culation are also shown in Fig. 15. Comparison between Evans KE (2009) Phys Status Solidi B 57:1865
the analytical models, the FEA, and the actual experimental 7. Bianchi M, Scarpa FL, Smith CW (2008) J Mater Sci 43:5851.
results indicates that the analytical models presented in this doi:10.1007/s10853-008-2841-5
article provides a good estimation of the actual experi- 8. Bezazi A, Scarpa F (2007) Int J Fatigue 29:922
9. Bezazi A, Scarpa F (2009) Int J Fatigue 31:488
mental results, and in most cases comparable with the FEA 10. Scarpa F, Pastorino P, Garelli A, Patsias S, Ruzzene M (2005)
calculations. Since the amount of calculations required for Phys Status Solidi B 242:681
FEA of an actual cellular structure becomes intractable for 11. Scarpa F, Ciffo LG, Yates JR (2004) Smart Mater Struct 13:49
real world lattice geometries involving thousands of unit 12. Howell B, Prendergast P, Hansen L (1996) Appl Acoust 43:141
13. Scarpa F, Smith FC (2004) J Intell Mater Syst Struct 15:973
cells, the analytical model provides a useful tool for design 14. Scarpa F, Bullough WA, Lumley P (2004) Proc Inst Mech Eng
purposes. Furthermore, the equations provide useful Part C 218:241
quantitative insight regarding the relationships between 15. Almgren RF (1985) J Elast 15:427
geometric design parameters and the resulting mechanical 16. Prall D, Lakes RS (1996) Int J Mech Sci 39:305
17. Grima JN (2006) J Mater Sci 41:3193. doi:10.1007/s10853-
properties. They are therefore quite useful as a tool for 006-6339-8
tuning mechanical properties for the needs of specific 18. Gaspar N, Smith CW, Alderson A, Grima JN, Evans KE (2011)
design applications. J Mater Sci 46:372. doi:10.1007/s10853-010-4846-0
19. Grima JN, Ravirala N, Galea R, Ellul B, Attard D, Gatt R,
Alderson A, Rasburn J, Evans KE (2011) Phys Status Solidi B
248:117
Conclusions 20. Williams JJ, Smith CW, Evans KE, Lethbridge ZAD, Walton RI
(2007) Acta Mater 55:5697
A uniaxial compressive analytical model for a 3D 21. Caddock BD, Evans KE (1989) J Phys D Appl Phys 22:1877
22. Theocaris PS, Stavroulakis GE, Panagiotopoulos PD (1997) Arch
re-entrant lattice auxetic structure has been established as a Appl Mech 67:274
function of four geometric design parameters that dictate 23. Larsen UD, Sigmund O, Bouwstra S (1997) Int J Mech Sci 6:99
the size and shape of each unit cell. Various properties of 24. Evans KE, Caddock BD (1989) J Phys D Appl Phys 22:1883
the lattice structure, including Poisson’s ratio, the failure 25. Wan H, Ohtaki H, Kotosaka S, Hu G (2004) Eur J Mech A/Solids
23:95
mode, the failure strength, and the effective modulus, were 26. Evans KE, Nkansah MA, Hutchinson IJ (1994) Acta Metall et
determined analytically as a function of the geometric Mater 42:1289
design parameters. The analysis showed that the properties 27. Schwerdtfeger J, Heinl P, Singer RF, Korner C (2010) Phys
of the 3D re-entrant lattice structure were highly tailorable Status Solidi B 247:269
28. Lakes RS, Witt R (2002) Int J Mech Eng Educ 30:50
by changes of the geometric design parameters. 29. Brezny R, Green DJ (1990) J Mater Sci 25:4571. doi:
Comparison of the predicted performance of the ana- 10.1007/BF01129908
lytical models with FEA modeling as well as experiments 30. Gibson LJ, Ashby MF (1997) Cellular solids: structure and
with electron beam melted Ti–6Al–4V lattice blocks properties, 2nd edn. Cambridge University Press, Cambridge
31. Onck PR, Andrews EW, Gibson LJ (2001) Int J Mech Sci 43:681
indicated that the analytical model is accurate enough for
this type of high modulus materials. The practical signifi-
cance is that the analytical models can be used as a tool for
initial phase tuning of mechanical properties when FEA
analysis is computationally intractable.

123

You might also like