Professional Documents
Culture Documents
Yedong Wang1, Shikai Jing1, Yonghui Liu2, Guohua Song1, Longfei Qie1
and Hao Xing1
Abstract
Additive manufacturing technology can make products of arbitrary shapes, greatly expanding the design space of lattice struc-
tures. Compared with traditional solid lattice structures, the lattice structures with hollow struts have higher flexural
strength, which arouses interests of designers recently. However, owing to the more complex shapes of structures, the
model generation and design are facing more challenges. In this article, a novel generative design method for the creation of
lattice structures with hollow struts is proposed. This method consists of three stages: initialization, analysis, and optimiza-
tion. First of all, a ground structure is generated automatically based on initial conditions. Then, the finite element analysis is
used to get the stress and coordinate information of finite element nodes as well as deformation information of the ground
structure. At last, a rapid optimization method is presented based on the idea of mapping the strut equivalent stress to the
strut wall thickness to optimize materials distribution. The proposed method is validated through a case study, demonstrat-
ing that this method can enhance performance of products while reducing the complexity of the optimization problem.
Keywords
Generative design, hollow struts, lattice structure, optimization, additive manufacturing
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering
in the place where they are needed.’’8 Traditionally, the pyramid-shaped arrangement as the sandwich plate
optimization of lattice structures is completed by topol- and found that the strength was mainly determined by
ogy methods, mainly including the homogenization the relative density. When the relative density was
method and ground structure method.9 Bendsøe and greater than 3%, the structure would not bend.
Kikuchi10 proposed the homogenization method first in Although lattice structures with hollow struts have
1988. This method extends the concept of composite better performance, the difficulties of the structures
media from the domain of porous materials to topology design are also greatly increased due to their complex
optimization by means of introducing microstructure in shapes. In order to improve the design efficiency of lat-
the macroscopic material analysis, which transforms tice structure, a new generative design method is pro-
the complex topology optimization problem into a rela- posed in this article. There are two outstanding
tively simple size optimization problem. Thus, the per- characteristics of this method: (1) the lattice structures
formance of the macrostructure can be obtained from can be generated automatically under the consideration
different microstructures. The ground structure method of directions, which can save much modeling time and
is development for discrete structures. It is designed to (2) the wall thickness of struts are variable to obtain the
adjust the initial ground structure, the design variables optimal materials distribution by means of a mapping
of which are the shape and size of the cross section.11 method, which avoids complex topology optimization.
Lattice structures are typical discrete structures, and
therefore, designers usually adopt the ground structure
method to improve its performance. Basic concepts
However, owing to the large number of struts in the Hollow struts
lattice structures, difficulties occurred during the design
process. Graf12 reported that there would be 2n kinds Because hollow struts have higher second area moments
of designs when the structure had n struts, with the than solid struts, the lattice structures constructed with
hypothesis that the strut diameter could only be 0 or 1. hollow struts are found to be stronger than their solid
It takes too much time to look for the optimal solution strut counterparts.21
using traditional optimization methods. In order to The shapes of the cross sections can influence the
reduce the computational complexity, researches tend mechanical properties of the struts. Compared to sec-
to take the following two kinds of solutions. The first tions of other shapes, hollow struts with circular sec-
kind is to improve the optimization methods using tions have higher second moment of area and better
heuristic methods. Chang and Rosen13,14 proposed the flexural behavior.22 The hollow strut model is shown in
size matching and scaling method (SMS), which com- Figure 1. The symbol L represents the length of the
bined a solid-body finite element analysis and a library strut. The symbols R and r represent the outer radius
of pre-defined lattice configurations to generate a struc- and inner radius, respectively. Then the wall thickness
ture’s lattice topology. Zhang et al.15 eliminated the of the strut can be expressed as
need for time-consuming optimization by mapping the
cell density information from structure topology opti- t=R r ð1Þ
mization results to struts diameter with a weighted
Therefore, the key to determine the wall thickness of
average method. Obviously, the selection strategy of
the strut is to determine its outer radius and inner
different weighting factors will have a significant influ-
radius. The full representation of a hollow strut is
ence on the design results. The second kind is to opti-
shown below
mize the direction of lattice structures. The strain
energy of lattice structures in different directions was Strut = (SNode ID, ENode ID, R, ri , L) ð2Þ
calculated by Tang et al.,16 which showed that the
directions could dramatically affect the mechanical where SNode ID and ENode ID represent the ID of
properties of structures. In order to avoid the bending start node and end node, respectively. The definition of
and shearing loads of struts, Reinhart and Teufelhart17
reported that the optimal arrangement of the skin–
lattice structures was along the force direction.
Additionally, some researchers have made a break-
through from the structure itself recently.18 They
replaced the original solid struts with hollow struts,
which were the shape of tubes. Queheillalt and
Wadley19 designed a kind of light sandwich truss struc-
ture formed of hollow struts, the strength of which was
much larger than that of the solid truss structure.
Pingle et al.20 adopted the tubular structures of Figure 1. Hollow strut model.
Wang et al. 3
Figure 2. Comparison of optimized lattice unit cell with solid struts and hollow struts: (a) solid lattice unit cell, (b) hollow lattice
unit cell, and (c) hollow lattice unit cell with two struts removed for visual display. The struts in (b) and (c) have identical outer
radius and different inner radius.
Ground structure
Figure 3. The relationship between: (a) a lattice frame and
Ground structure is the initial structure of the topology (b) a lattice structure.
optimization.23 In topology optimization theories, the
optimum topology is a subset of a ground structure.
The performance is optimized by adding, deleting, and difficult to generate, this article will make it in the fol-
modifying parts of the ground structure. In this article, lowing section.
the ground structure is the initial structure of identical
outer radius and inner radius which need to be further
adjusted. The cross sections of the struts become design Design method
variables of the optimization problem. The ground
structure is generated according to the given condi- The proposed design method aims at optimizing the
tions. Since hollow structure of arbitrary directions is materials distribution of lattice structures while meeting
4 Advances in Mechanical Engineering
Figure 4. Schematic illustration of the IAO stages during the design process of an AM lattice structure: (a) define the design space
and give the constraints, (b) the initial ground structure with hollow struts of identical wall thickness is obtained with the automatic
generation method, (c) results of finite element analysis are obtained, and (d) the explicit lattice structure is obtained from the
mapping method.
V = V T ð3Þ
The 3D lattice frame is generated after two rotation
steps from the base lattice frame: rotating a degrees
around the origin of in the XOY plane first, and then
rotating g degrees around some axis (the y-axis is cho-
sen in this article). The rotation matrix can be con-
cluded as
0 1 0 1
cos a sin a 0 cos g 0 sin g
T = @ sin a cos a 0A @ 0 1 0 A
0 0 1 sin g 0 cos g
ð4Þ
Figure 6. Hollow lattice structures of different directions.
The nodes coordinates of the lattice frame can be
obtained with the help of equation (5)
where the symbol Q represents the finite element nodes
set; the symbol n represents the number of finite element
xi = zi l sin g + (xi l cos a yi sin a) cos g
nodes; the symbols x, y, and z represent the coordinates
yj = xi l sin a yi l cos a ð5Þ of the finite element node i; the symbol si represents the
zk = zi l cos g + (xi l cos a yi sin a) sin g equivalent stress of the finite element node i.
In addition, the deformation of the ground structure
Second, the ground structure is generated from the under the action of outer load should also be recorded,
lattice frame when given the outer radius and inner which is used to measure the stiffness of the structure.
radius of struts. The wall thickness of each strut is iden-
tical in the ground structure. Assuming the outer
radius, the number of struts, the target volume, and the Wall thickness optimization
length of struts are R, i, V, and L, respectively; the
The struts of the ground structure have the same wall
inner radius of struts can be obtained by
thickness, which needs to be optimized to get better per-
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi formance. ‘‘Optimal performance’’ means that struc-
V
r = R2 ð6Þ tures can withstand the maximum external load in the
piL case of a certain volume or that structures have the min-
In this way, we can get all the parameters needed for imum deformation in the case of a certain volume and
the ground structures modeling. According to the external load. The formulation of the proposed optimi-
requirements of designers, the ground structures of any zation problem for lattice structure can be approached
direction can be generated automatically based on UG as a size optimization problem as well as a multiple
NX9.0 secondary development technology. A strut objective design problem. Typically, this problem is for-
model can be obtained through Boolean operation with mulated with the compromise decision support problem
two cylinders of different radius. Some lattice struc- (cDSP).24 The mathematical representation for the opti-
tures of different a and g values are shown in Figure 6. mization problem is expressed in Table 1.
In Table 1, the symbols GS, Vt, and si represent the position relation between the finite element nodes and
ground structure, target volume, and equivalent stress struts, which can be expressed in the following form
value of strut i, respectively. The outer radius and inner
radius of strut i can range from the lower radius bound, Given: Q = (e1 , e2 , . . . , en ), S = fS1 , S2 , . . . , Sm g
rmin, to the upper radius bound, Rmax. V and d are the Find: Q = fQ1 , Q2 , . . . , Qm g ð9Þ
actual volume and deformation of the structure, respec-
Q1 2 S1 , Q2 2 S2 , . . . , Qm 2 Sm
tively. The symbol v1 and v2 are set manually by the
designer as the weighting variables for d and V to char- where the symbol Q represents the finite element nodes
acterize the relative importance of the volume and set; the symbol S represents the struts set. The symbol
deflection design goals during optimization. n represents the number of finite element nodes; the
In the case of a certain volume of materials, the symbol m represents the number of struts. The symbol
materials required for each strut are determined by the 2 represents the meaning of ‘‘located on.’’ The prob-
size of the strut’s equivalent stress value. More materi- lem is to find out a subset of Q that located on the cor-
als should be distributed on struts with greater stress, responding strut according to the above information.
which can improve the utilization ratio of materials. A strut model can be obtained through Boolean
Therefore, this article presents the idea of mapping the operation with two cylinders of different radius. This
strut equivalent stress to the strut wall thickness. problem is equivalent to judging the position relation-
The stress distribution space can be represented by ship between the finite element nodes and the outer
Ps, and the wall thickness geometry can be represented cylinders of each strut, which is easier to work out. So
by Pt. Then, the mapping idea can be expressed in the this problem can be further transformed into a mathe-
following form matical model
Ps ! F ! Pt ð8Þ
Given : O1 (x1 , y1 , z1 ), O2 (x2 , y2 , z2 ), outr(Si ) = R
ð10Þ
where the symbol F represents the mapping function. Find : Qi Q, Qi 2 Si
According to its stress value, the strut’s wall thickness
can be obtained based on the mapping function, as is The centers of two end surfaces of the strut Si are O1
shown in Figure 7. The ring represents the cross section and O2. The symbol outr represents the outer radius of
of a strut; the points in the ring represent finite element the strut Si. The goal of the mapping method is to find
nodes; the size of a point represents the equivalent stress out a subset of the finite element nodes set which is
value of the corresponding node. located on the strut Si at the same time.
The struts have the same outer radius instead of inner As is shown in Figure 8, there are three kinds of posi-
radius. The strut of greater stress has the smaller inner tion relations between finite element nodes and struts.
radius. In order to obtain exact wall thickness values, The finite element node A belongs to the strut Si only
there are two problems remaining to be solved: (1) aver- when A and Si are in the relation shown in Figure 8(c),
age stress of each strut and (2) the mapping function. the mathematical expression is
The stress value of a strut is obtained by calculating
the average stress value of the finite element nodes dist(A, L) R
ð11Þ
located on the strut. So it is necessary to judge the proj(A, O1 O2 ) 2 O1 O2
Wang et al. 7
where the symbol L represents the axis of the strut; the Figure 9. The algorithm to determine the position relationship
symbol dist(A, L) represents the distance from A to L; between finite element nodes and struts.
the symbol proj(A, O1 O2 ) represents the projection point
of A on the line segment O1 O2 .
The position relationship between finite element The outer radius is determined by the strut with the
nodes and struts is determined using an automatic maximum volume, which is regarded as a solid strut.
judgment algorithm developed by authors shown in The outer radius is the same as the radius of the solid
Figure 9. strut.
The finite element nodes on the strut Si can be Considering the precision of the 3D printer and
expressed as fA1 , A2 , . . . , Ak g, where k is the number of computer-aided design (CAD) software, the wall thick-
nodes. The equivalent stress for each node can be ness and the inner radius of struts shall not be too
expressed as fseqv(A1 ), seqv(A2 ), . . . , seqv(Ak )g. Then, small. The range of these parameters in this article is as
an average stress value of the strut Si can be obtained follows
P
k
0:1 mm r (R d) mm
seqv(Ai ) ð14Þ
i=1 d = 0:05 mm
s(S i ) = ð12Þ
k
where the symbol d represents the minimum wall thick-
The outer radius and inner radius of the strut can be ness. This may lead to the change of the total volume
obtained based on the mapping formula of the required materials. Empirical studies show that
the change is generally less than 0.1%, which can be
s(Si )
V (Si ) = P
m Vt negligible.
s(Si )
j=1
rffiffiffiffiffiffiffiffiffiffi
Vmax ð13Þ
R= Case study
pL
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V (Si ) In this section, a design optimization case for a 3D can-
r(Si ) = R2 tilever beam, shown in Figure 10, is given to validate
pL
the proposed design method of lattice structures with
8 Advances in Mechanical Engineering
Parameter Value
Length (mm) 10
Load (N) 100
Elastic modulus (N/mm) 2 3 1011
Target volume (mm3) 1600
Unit cells along length 5
Unit cells along width 1
Unit cells along height 2
Figure 11. Ground structure model: (a) axonometric drawing and (b) sectional drawing.
Wang et al. 9
Figure 12. Finite element analysis: (a) finite element nodes distribution and (b) structure deformation analysis.
r
rv = ð15Þ
rs
16. Tang Y, Kurtz A and Zhao YF. Bidirectional evolution- 21. Queheillalt DT and Wadley H. Pyramidal lattice truss
ary structural optimization (BESO) based design method structures with hollow trusses. Mat Sci Eng A 2005; 397:
for lattice structure to be fabricated by additive manufac- 132–137.
turing. Comput Aided Design 2015; 69: 91–101. 22. Barrass CB and Derrett CDR. Chapter 11. Second
17. Reinhart G and Teufelhart S. Optimization of mechanical moments of area—moments of inertia. In: Barrass CB
loaded lattice structures by orientating their struts along and Derrett CDR (eds) Ship stability for masters and
the flux of force. Proc CIRP 2013; 12: 175–180. mates, 2006, pp.94–102. Butterworth-Heinemann Ltd.
18. St-Pierre L, Fleck NA and Deshpande VS. The predicted 23. Chu J, Engelbrecht S and Graf G. A comparison of
compressive strength of a pyramidal lattice made from synthesis methods for cellular structures with application
case hardened steel tubes. Int J Solids Struct 2014; 51: to additive manufacturing. Rapid Prototyping J 2010; 16:
41–52. 275–283.
19. Queheillalt DT and Wadley H. Hollow pyramidal lattice 24. Seeperasd CC, Allen JK and Mcdowell DL. Robust design
truss structures. Int J Mater Res 2011; 102: 389–400. of cellular materials with topological and dimensional
20. Pingle SM, Fleck NA and Deshpande VS. Collapse imperfections. J Mech Design 2006; 128: 1285–1297.
mechanism maps for a hollow pyramidal lattice. P Roy 25. Alzahrani M, Choi SK and Rosen DW. Design of truss-
Soc Lond A Mat 2011; 467: 985–1011. like cellular structures using relative density mapping
method. Mater Design 2015; 85: 349–360.