Professional Documents
Culture Documents
Composite Structures
journal homepage: www.elsevier.com/locate/compstruct
A R T I C L E I N F O A B S T R A C T
Keywords: In this work, we design and test a parameter optimization method by Python script to meet the urgent demand
Honeycomb metamaterial for lightweight honeycomb metamaterial. The method mainly focuses on the selection of parameters according
Auxetic effect to the mass and Poisson’s ratio of the honeycomb metamaterial. The bidirectional re‐entrant honeycomb is pro-
Parameter optimization posed as an objective to be optimized and the formula of Poisson's ratio is deduced theoretically to establish the
Multi‐island genetic algorithm
internal relation. Besides, the accuracy of the Python script results is verified by static compression experimen-
tal results and theoretical results. Combined the Python script programming model with the genetic algorithm
optimization method, the optimal honeycomb metamaterial solutions are obtained. Results show that the
parameter optimization method using multi‐island genetic algorithm (GA) can avoid a local solution’s appear-
ance, and both the shell model and the solid model can obtain the ideal optimal solution. Furthermore, the 3D
honeycomb has an admirable auxetic effect according to the optimized parameters, which provides a piece of
strong evidence for the continuous application of optimization algorithms to improve the mechanical proper-
ties of honeycomb metamaterial.
1. Introduction et al. [16] and Ma et al. [17] combined the design of the metamaterial
with deep learning. Bessa et al. [18] studied the fragile problem with
Many natural cellular materials are structured regularly and have the help of Bayesian machine learning. He et al. [24] utilized the
fascinating mechanical properties and functions for natural selec- machine learning method to predict the DNN model. Choi et al. [25]
tion/adaptation [1–4]. In recent decades, metamaterials have been adopted the inverse design method to achieve bistability by reducing
carefully designed to meet many applications including acoustics the size of the hinges. Wu et al. [26] introduced a machine learning‐
[5–7], optics [8–11], mechanics [12–19], electromagnetism [20,21], based method to design modular metamaterials with specific
and so on. The various properties of metamaterials can guide the struc- properties.
ture design and obtain excellent performance. The theoretical formula In addition, mechanical metamaterials are divided into two aspects
is deduced by establishing the mechanical model, and its correctness is from the structural dimension. The first category is planar mechanical
verified by abundant experiments. However, it is undeniable that the metamaterials, their deformation has nothing to do with the properties
analysis difficulty caused by the complex structure, and the result devi- outside the plane. The second category is 3D mechanical metamateri-
ation caused by the experimental conditions need to spend a lot of als, which take into account more complex forces and deformation. It
time to solve. is significantly simpler to derive a 3D structure [28–40] from planar
With the development of optimization methods, traditional topol- structures than to obtain it directly. The properties of planar structures
ogy optimization [12,22] and morphology optimization [13,23] are often play a leading role in the study of 3D structural properties. For
no longer the exclusive methods. The technology of machine learning example, Wang et al. [28,29] studied 3D re‐entrant hexagonal struc-
[8,16–18,24–27] presents a practical and effective method for solving ture and double arrow‐head structure on the basis of previous studies
complex optimization problems. In recent years, the genetic algorithm of a planar structure. Imbalzano [30] extended a 3D auxetic from 2D
has been widely concerned in the field of structural optimization, with re‐entrant auxetic structure and 3D re‐entrant elongated dodecahe-
its simplicity and powerful robustness. Jafar et al. [8] presented the dron to obtain the auxetic behavior also in the transverse planes. Wang
adaptive genetic algorithm for optical metasurfaces design. Hou and Liu [31,32] developed a 3D compression‐torsion structure from
⇑ Corresponding author.
E-mail address: htliu@hebut.edu.cn (H.-T. Liu).
https://doi.org/10.1016/j.compstruct.2021.113915
Received 27 June 2020; Revised 22 February 2021; Accepted 24 March 2021
Available online 30 March 2021
0263-8223/© 2021 Elsevier Ltd. All rights reserved.
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
two similar 2D structures. Furthermore, the U‐shaped structure [33], expressed by defining total length L, oblique wall length b, wall thick-
cross‐rods structure [34], axisymmetric rotation structure [35], and ness t, edge length d, wall length c and re‐entrant angle θ, as displayed
other 3D structures are all developed based on 2D structure properties. in Fig. 2 (a). The gauge thickness of the honeycomb panels is indicated
The research ideas of the above structures are all based on the applica- as h. As the method in literature [23], the geometry of the unit cell is
tion of 2D structure expansion to obtain a 3D structure, which has also defined by non‐dimensional parameters like α, β and γ (α ¼ d=b,
obvious reference value. To a certain extent, using machine learning β ¼ t=b, γ ¼ h=b).
and genetic algorithm to optimize the structure can improve the design
efficiency and calculation cost. However, to the authors’ knowledge, 2.2. Theoretical analysis
little or even no paper was reported by Python language and multi‐
island GA for mechanical metamaterials. The free‐body diagrams for loading in Fig. 2(b). Periodic boundary
In this paper, 2D auxetic honeycomb metamaterials are analyzed conditions [41,42] are used, the oblique walls on the left and right
by Castigliano's theorem, and the accuracy of the multi‐island GA sides are fixed. Thus, the structure is only subjected to the Y‐
results is provided for reference. The theoretical solution makes the direction load. By the moment equilibrium condition, the torque of
influence of various geometric parameters of the honeycomb on Pois- arbitrary cross‐section is given as
son's ratio more intuitive. Owing to the complexity of multiple vari-
M s ¼ Fssinθ M ð1Þ
ables, it is difficult to capture the optimal solution of Poisson's ratio.
Nevertheless, the multi‐island GA saves time cost by taking advantage where F is the external force, and it can be written as
of global traversal, and the parameterized model and the precise con- F ¼ σbhcosθ ð2Þ
straint conditions make the structure more efficient in the process of
parameter optimization. This optimization method plays a positive The deflection angle can be obtained by integrating the bending‐
role in determining the optimal parameters of the structure and deter- moment equation
mines a series of parameter array structures. Z
Ms
ΔðsÞ ¼ ds þ C ð3Þ
Es I
2. Structural design and theoretical analysis
where C is a constant of integration, E s is the Young’s modulus of
the initial material, and I is second moment of area of the wall.
2.1. Structural design
Considering the symmetry of the honeycomb, the deflection angle
at the turning point of the oblique wall is zero. Therefore, the bound-
The geometries of the quadrangular star honeycomb (QSH) and the
ary conditions are obtained as
bidirectional re‐entrant honeycomb (BRH) are displayed in Fig. 1(a)
and (b). Although the QSH is investigated by many researchers, the Δð0Þ ¼ 0; ΔðbÞ ¼ 0 ð4Þ
strength of the structure and the difficulty of manufacturing are two Submitting Eq. (4) to Eq. (3), we can obtain
fatal problems that cannot be ignored. In order to solve the existing (
problems in terms of structural characteristics, the sharp angle of the M ¼ Fb12sinθ
ð5Þ
structure is replaced by a right angle while the expansion mode of C ¼ 0
the structure remains unchanged. The unit cell of BRH can be
Fig. 1. Conceptual illustration of the amplified nodal design for the bidirectional re-entrant honeycomb. (a) QSH; (b) BRH.
2
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
Fig. 2. Design and measurement of BRH. (a) structure and unit cell; (b) periodic boundary condition and free-body diagrams; (c) experimental specimens and
compression model; (d) script results (SR), theoretical results (TR) and experimental results (ER) about νyx with varying angles.
The horizontal displacement and the vertical displacement of the ð2d þ bcosθÞ b3 þ bt 2 sinθcosθ þ 2d3
νyx ¼ 3 ð14Þ
oblique wall in the X‐direction include axial and flexural deformations, ð2d þ bcosθ bsinθÞ b sinθ2 þ bt 2 cosθ2 þ 2dt 2
as shown in Fig. 2(c). Based on Castigliano's second theorem, they are
obtained as and normalize the equivalent Poisson's ratio can be written as
8
R R ð2α þ cosθÞ 1 þ β2 tanθ þ 2α3 ðtanθ2 þ 1Þ
< δx ¼ bþ2d Ms M1 ds þ b F s F 1 ds νyx ¼ ð15Þ
0 Es I 0 Es A
ð6Þ ð2α þ cosθ sinθÞ tanθ þ β þ 2αβ ðtanθ þ 1Þ
2 2 2 2
: δy ¼ R b M s M 1 ds þ R bþ2d F s F 1 ds
0 Es I 0 Es A
Due to the size limitation of the honeycomb, it is necessary to avoid
where M s and F s are actual loads; M 1 and F 1 are unit loads (mo- contacts and intersections of the walls. Therefore, the geometric size of
ments and forces) in the X‐direction. They can be expressed as BRH meets the following constraints
8
Fbsinθ > L ¼ t þ 4d þ 2bcosθ
M s ¼ Fssinθ ð7Þ >
>
2 < c ¼ 2d
ð16Þ
>
> L=2 ⩾ bcosθ
F s ¼ Fcosθ ð8Þ >
: pffiffiffi
2ðd t Þ=2b ⩾ cos½ð3π=4 θÞ
bsinθ
M 1 ¼ ssinθ ð9Þ It is worth noting that the relationship of QSH and BRH can be
2 linked by setting the width to zero, and the BRH can transform the
F 1 ¼ cosθ ð10Þ QSH. Accordingly, Eqs. (14) and (15) are simplified as
2
Taking Eqs. (7)‐(10) into Eq. (6), the horizontal displacement and b þ t 2 tanθ
νyx ¼ 2 ð17Þ
the vertical displacement are calculated as b tanθ2 þ t 2 ð1 tanθÞ
( 3
2Fd3
δx ¼ Fb 12E
sinθcosθ
þ Fbsinθcosθ þ 12E Normalized the Poisson's ratio by
sI Es A sI
ð11Þ
1 þ β2 tanθ
3
sinθ2 Fbcosθ2
δy ¼ Fb12E þ þ 2Fd
s I E s A E s A νyx ¼ ð18Þ
tanθ2 þ β2 ð1 tanθÞ
The strain expressions in the X‐direction and the Y‐direction are
obtained as
( δx 3. Reliability of script results
ɛ x ¼ 2dþbðcosθsinθ Þ
δy
ð12Þ
ɛy ¼ 2dþbcosθ In this section, the accuracy of the script results is verified by the-
The equivalent Poisson's ratio are shown as oretical results and experimental results to ensure the reliability of the
optimization results in the process of geometric parameter optimiza-
ɛx
νyx ¼ ð13Þ tion. A series of specimens are printed by industrial‐grade SLA 3D prin-
ɛy
ter Lite800HD, and these specimens are all made of resin material
According to the Eqs. (12) and (13), the equivalent Poisson's ratio similar to ABS. The Young's modulus and Poisson's ratio of resin mate-
are expressed as rial are 2360 MPa and 0.4, respectively. The size of the specimen is
3
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
Table 1
The multi-island GA parameters setting.
4
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
Fig. 6. The structural parameter optimization by the multi-island GA and Python script model.
Table 2
The bounds of control parameters.
Variable θ α β
Therefore, the BRH can obtain a larger Poisson's ratio effect on the
basis of guaranteeing the star‐shaped structure.
5
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
The steps of the total optimization process include selecting opti- Fig. 10. Construction diagram of 3D BRH. (a)(b) The plates#1 and plates#2;
mization variables, creating Python script, testing and debugging (c)-(f) Transverse and longitudinal of plates; (g)-(h) 3D BRH#1 and 3D
Python script results, establish and solving optimization model. Con- BRH#2; (i) 3D BRH#3.
sidering the elastic buckling of the cell wall and geometry conflicts,
Fig. 9. (a) Initial model; (b) Optimized model; (c)(d) The deformation of Python script programming model before and after optimization; (e) Mises stress
nephograms of honeycombs under different strain conditions.
6
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
Fig. 11. Displacement nephograms of 3D BRH in X-Y-Z direction. (a) 3D BRH#1; (b) 3D BRH#2; (c) 3D BRH#3.
the bounds of control parameters are shown in Table 2. The Python of the Poisson’s ratio for BRH is modified from −5 to −4.5 and from
script is testing by running ABAQUS and debugging the script by mod- −0.5 to −2, while the mass of BRH gradually converges to 900 g and
ifying the corresponding parameters. cannot reach the previous result. This optimization method can save
time and economic cost to the greatest extent while ensuring structural
5. Results and discussion performance.
Similarly, the maximum Poisson's ratio is obtained by regarding the
5.1. Optimization result analysis mass as a constraint condition, as shown in Fig. 8. Although unit cells
are completely symmetrical in structure, the Poisson's ratio in the two
The previous work has proven the reliability of the Python script orthogonal directions is different due to the form of staggered connec-
results. Therefore, this section mainly explores the optimization prob- tion. The deformation of BRH for loading in X‐direction is presented by
lem of multi‐variables seeking the lightest mass within a certain Pois- setting νxy as the target. The large variation on Poisson's ratio or mass
son's ratio range. The multi‐island GA can avoid the local optimal reflects the process of algorithms. While the Poisson's ratio keeps
solution because it can traverse the parameter range. Fig. 7 (a) and increasing, the change of relevant parameters causes the sudden
(b) show the optimization iteration history of mass is targeted as an change of mass. In the previous structural design, owing to the law
optimization goal in the script. In this case, Poisson’s ratio of the hon- of a variable is not monotonous, the interactions between multiple
eycomb should be kept between 0.5 and 5. During the optimization variables are considered. In the present work, we take mass as the con-
process, the upper and lower limits on each variable are to ensure straint variable and Poisson's ratio as the optimization objective. The
the performance of the honeycomb. After changing the range of Pois- global optimal solutions are obtained directly by applying the opti-
son's ratio, the optimization result is changed accordingly. The range mization method of the multi‐island GA. Of course, add a stiffness of
7
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
the overall structure as a constraint is also an effective method. This is generally applicable to the target design of array‐type cube‐shaped
work is our future research direction by the multi‐island GA, and it structures.
is solved in the future.
Fig. 9 (a) and (b) show that the two configurations of the honey- Declaration of Competing Interest
comb before and after optimization have great changes in thickness,
angle and width, which are difficult to achieve by optimizing single‐ The authors declare that they have no known competing financial
variable. There are also obvious differences in the optimization results. interests or personal relationships that could have appeared to influ-
Fig. 9 (c) is shown the deformation of the Python script programming ence the work reported in this paper.
model before optimization. When the optimized honeycomb is sub-
jected to axial compression load, the lateral contraction displacement Acknowledgements
increases significantly, and the specific deformation is shown in Fig. 9
(d). The honeycombs are capable of achieving an increased shrinkage This work was supported by the National Natural Science Founda-
effect under increasing axial strain as shown in Fig. 9 (e). It means that tion of China (11702079).
greater deformation can be achieved by using more ductile materials.
References
5.2. 3D BRH
[1] Lees C, Vincent JFV, Hillerton JE. Poisson's ratio in skin. Biomed Mater Eng 1991;1
(1):19–23.
Based on the optimized unit cell of BRH, the corresponding 3D [2] Mardling P, Alderson A, Jordan-Mahy N, Le Maitre CL. The use of auxetic materials
configurations are designed. Fig. 10 (a) and (b) show two‐periodic in tissue engineering. Biomater Sci 2020;8(8):2074–83.
honeycomb plates in horizontal and vertical, one is composed of [3] Williams JL, Lewis JL. Properties and an anisotropic model of cancellous bone
from the proximal tibial epiphysis. J Biomech Eng 1982;104(1):50.
86 cells and is named plates#1, the other is composed of 85 cells [4] Veronda DR, Westmann RA. Mechanical characterization of skin-finite
and is named plates#2. Five plates#1 are arranged parallel and deformations. J Biomech 1970;3(1):111–24.
equidistant, as shown in Fig. 10 (c) and (d). Put them with the stag- [5] Tian Z, Shen C, Li J, Reit E, Gu Y, Fu H, et al. Programmable acoustic metasurfaces.
Adv Funct Mater 2019;29(13):1808489. https://doi.org/10.1002/adfm.
gered arrangement, the 3D BRH#1 is obtained. Similarly, four v29.1310.1002/adfm.201808489.
plates#2 meet the staggered arrangement, the 3D BRH#2 is also [6] Lee D, Nguyen DM, Rho J. Acoustic wave science realized by metamaterials. Nano
obtained. The two honeycombs can maintain the Poisson's ratio effect Converg 2017;4(1):3.
[7] Xie Y, Fu Y, Jia Z, Li J, Shen C, Xu Y, et al. Acoustic imaging with metamaterial
independently, but the void ratio is large and the space utilization is
luneburg lenses. Sci Rep 2018;8(1). https://doi.org/10.1038/s41598-018-34581-
insufficient in Fig. 10 (g) and (h). Therefore, the combination of the 7.
3D BRH#1 and the 3D BRH#2 is obtained and called 3D BRH#3, as [8] Samad JZ, Sandeep I, Hossein M. Adaptive genetic algorithm for optical
metasurfaces design. Sci Rep 2018;8(1):11040.
shown in Fig. 10 (i).
[9] Lv J, Zhou M, Gu Q, Jiang X, Ying Y, Si G. Metamaterial lensing devices. Molecules
The 3D BRH#1 and the 3D BRH#2 are analyzed by numerical sim- 2019;24(13):2460.
ulation, and the displacement nephograms are depicted in Fig. 11 (a) [10] Ma X, Pu M, Li X, Guo Y, Gao P, Luo X. Meta-chirality: fundamentals, construction
and (b). When uniform distributed axial load is applied in the Y‐ and applications. Nanomaterials-Basel 2017;7(5):116.
[11] Pitchappa P, Kumar A, Prakash S, Jani H, Venkatesan T, Singh R. Chalcogenide
direction, the 3D BRH#1 and the 3D BRH#2 maintain the same con- phase change material for active terahertz photonics. Adv Mater 2019;31
traction in the X‐direction and Z‐direction. To put it in a nutshell, νyx (12):1808157. https://doi.org/10.1002/adma.v31.1210.1002/adma.201808157.
and νyz are always equal in deformation. However, compared to [12] Yang H, Wang B, Ma Li. Designing hierarchical metamaterials by topology analysis
with tailored Poisson’s ratio and Young’s modulus. Compos Struct
BRH#1 and BRH#2, BRH#3 has a more compact space and shrinkage 2019;214:359–78.
effect. While the axial displacement decreases by 16.7% and the same [13] De Tommasi D, Marano GC, Puglisi G, Trentadue F. Morphological optimization of
shrinkage effect are also maintained, as shown in Fig. 11 (c). tensegrity-type metamaterials. Compos Part B-ENG 2017;115:182–7.
[14] Yang H, Ma Li. 1D to 3D multi-stable architected materials with zero Poisson's
Before the design work for the 3D structure, the Poisson’s ratio ratio and controllable thermal expansion. Mater Design 2020;188:108430.
behavior is tuned and optimized through the Python script model https://doi.org/10.1016/j.matdes.2019.108430.
and the multi‐island GA. Undoubtedly, the method of parameter opti- [15] Tan X, Zhu S, Wang B, Yao K, Chen S, Xu P, et al. Mechanical response of negative
stiffness truncated-conical shell systems: experiment, numerical simulation and
mization saves calculation cost and improves calculation efficiency to empirical model. Compos Part B-ENG 2020;188:107898. https://doi.org/
a great extent. 10.1016/j.compositesb:2020.107898.
[16] Hou Z, Tang T, Shen J, Li C, Li F. Prediction network of metamaterial with split
ring resonator based on deep learning. Nanoscale Res Lett 2020;15(1):83.
6. Conclusions [17] Ma W, Cheng F, Liu Y. Deep-learning-enabled on-demand design of chiral
metamaterials. ACS Nano 2018;12(6):6326–34.
In the present paper, a novel programmable optimization strategy [18] Bessa MA, Glowacki P, Houlder M. Bayesian machine learning in metamaterial
design: fragile becomes supercompressible. Adv Mater 2019;31(48):1904845.
honeycomb metamaterial is put forward for the first time. The main https://doi.org/10.1002/adma.v31.4810.1002/adma.201904845.
feature is to integrate modeling, operation and post‐processing by [19] Wang Y-B, Liu H-T, Zhang Z-Y. Rotation spring: Rotation symmetric compression-
using Python language. The multi‐island GA is used to improve opti- torsion conversion structure with high space utilization. Compos Struct
2020;245:112341. https://doi.org/10.1016/j.compstruct.2020.112341.
mization autonomy. [20] Watts CM, Liu X, Padilla WJ. Metamaterial electromagnetic wave absorbers. Adv
Firstly, theoretical models of cell array honeycomb are proposed Mater 2012;24(23):OP98–OP120.
and their theoretical solutions are derived based on Castigliano's sec- [21] Cong L, Savinov V, Srivastava YK, Han S, Singh R. A metamaterial analog of the
Ising model. Adv Mater 2018;30(40):1804210. https://doi.org/10.1002/adma.
ond theorem. Secondly, the reliability of the script results is verified v30.4010.1002/adma.201804210.
by the compression experiment and theoretical derivation, and the [22] Clausen A, Wang F, Jensen JS, Sigmund O, Lewis JA. Topology optimized
error is acceptable. Due to a small number of experimental samples architectures with programmable Poisson’s ratio over large deformations. Adv
Mater 2015;27(37):5523–7.
are not enough to explain the universality of the script, the idea of con-
[23] Gong X, Huang J, Scarpa F, Liu Y, Leng J. Zero Poisson's ratio cellular structure for
trol variable method is used to verify the reliability of the script. The two-dimensional morphing applications. Compos Struct 2015;134:384–92.
errors of the fitting result are very small and the script is very reliable. [24] He J, He C, Zheng C, Wang Q, Ye J. Plasmonic nanoparticle simulations and
inverse design using machine learning. Nanoscale 2019;11(37):17444–59.
Finally, the Python script programming model is combined with the
[25] Choi GPT, Dudte LH, Mahadevan L. Programming shape using kirigami
multi‐island GA to carry out optimization modes about minimum mass tessellations. Nat Mater 2019;18(9):999–1004.
and maximum Poisson's ratio. The results show that the honeycomb’s [26] Wu L, Liu L, Wang Y, Zhai Z, Zhuang H, Krishnaraju D, et al. A machine learning-
optimization efficiency is high, and the specific performance of the based method to design modular metamaterials. Extreme Mech Lett
2020;36:100657. https://doi.org/10.1016/j.eml.2020.100657.
optimized honeycomb is more significant. This optimization method
8
L. Wang, H.-T. Liu Composite Structures 267 (2021) 113915
[27] Hougne P, Imani MF, Diebold AV, Horstmeyer R, Smith DR. Learned integrated [34] Lu Z, Wang Q, Li X, Yang Z. Elastic properties of two novel auxetic 3D cellular
sensing pipeline: reconfigurable metasurface transceivers as trainable physical structures. Int J Solids Struct 2017;124:46–56.
layer in an artificial neural network. Adv Sci 2020;7(3):1901913. https://doi.org/ [35] Carneiro VH, Puga H. Axisymmetric auxetics. Compos Struct 2018;204:438–44.
10.1002/advs.v7.310.1002/advs.201901913. [36] Lee W, Jeong Y, Yoo J, Huh H, Park S-J, Park SH, et al. Effect of auxetic structures
[28] Wang X-T, Wang B, Li X-W, Ma Li. Mechanical properties of 3D re- on crash behavior of cylindrical tube. Compos Struct 2019;208:836–46.
entrant auxetic cellular structures. Int J Mech Sci 2017;131- [37] Carneiro VH, Puga H, Meireles J. Positive, zero and negative Poisson’s ratio non-
132:396–407. stochastic metallic cellular solids: Dependence between static and dynamic
[29] Wang X-T, Wang B, Wen Z-H, Ma Li. Fabrication and mechanical properties of mechanical properties. Compos Struct 2019;226:111239. https://doi.org/
CFRP composite three-dimensional double-arrow-head auxetic structures. Compos 10.1016/j.compstruct.2019.111239.
Sci Technol 2018;164:92–102. [38] Jeong S, Yoo HH. Shape optimization of bowtie-shaped auxetic structures using
[30] Imbalzano G, Tran P, Ngo TD, Lee PVS. A numerical study of auxetic composite beam theory. Compos Struct 2019;224:111020. https://doi.org/10.1016/
panels under blast loadings. Compos Struct 2016;135:339–52. j.compstruct.2019.111020.
[31] Wang L, Liu H-T. 3D compression–torsion cubic mechanical metamaterial with [39] Ai L, Gao X-L. Metamaterials with negative Poisson’s ratio and non-positive
double inclined rods. Extreme Mech Lett 2020;37:100706. https://doi.org/ thermal expansion. Compos Struct 2017;162:70–84.
10.1016/j.eml.2020.100706. [40] Mun J, Yun B-G, Ju J, Chang B-M. Indirect additive manufacturing based casting of
[32] Wang L, An M-R, Liu H-T. Compression spin bio-inspired arm: A conceptual model a periodic 3D cellular metal-Flow simulation of molten aluminum alloy. J Manuf
based on compression–torsion cubic mechanical metamaterials with variable Process 2015;17:28–40.
cross-section. Extreme Mech Lett 2020;41:101069. https://doi.org/10.1016/j. [41] Wang F, Sigmund O, Jensen JS. Design of materials with prescribed nonlinear
eml.2020.101069. properties. J Mech Phys Solid 2014;69:156–74.
[33] Yang H, Wang B, Ma Li. Mechanical properties of 3D double-U auxetic structures. [42] Harkati E, Daoudi N, Bezazi A, Haddad A, Scarpa F. In-plane elasticity of a multi
Int J Solids Struct 2019;180-181:13–29. re-entrant auxetic honeycomb. Compos Struct 2017;180:130–9.