materials
Article
Research on 3D-Print Design Method of Spatial Node Topology
Optimization Based on Improved Material Interpolation
Xianjie Wang 1,2,3 , Fan Zhang 2, * , Yang Zhao 3 , Zhaoyi Wang 2 and Guangen Zhou 4
Citation: Wang, X.; Zhang, F.; Zhao,
Y.; Wang, Z.; Zhou, G. Research on
3D-Print Design Method of Spatial
Node Topology Optimization Based
on Improved Material Interpolation.
Materials 2022, 15, 3874. https://
doi.org/10.3390/ma15113874
Academic Editor: Dimitrios Tzetzis
Received: 8 April 2022
Accepted: 24 May 2022
Published: 29 May 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Materials 2022, 15, 3874. https://doi.org/10.3390/ma15113874
1 Key Laboratory of Civil Engineering Structure and Mechanics, Inner Mongolia University of Technology,
Hohhot 010051, China; xianjiewang@ynu.edu.cn
2 School of Architecture and Planning, Yunnan University, Kunming 650106, China; 18608581557@163.com
3 School of Civil Engineering, Zhejiang University, Hangzhou 310058, China; ceyzhao@zju.edu.cn
4 Zhejiang Southest Space Frame Co., Ltd., Hangzhou 310058, China; zgg1967@163.com
* Correspondence: 12019202517@mail.ynu.edu.cn
Abstract: Designing a high-strength node is significant for space structures. Topological optimization
can optimally allocate the material distribution of components to meet performance requirements.
Although the material distribution after topology optimization is optimum, the structure becomes
complicated to manufacture. By using additive manufacturing technology, this problem can be well
solved. At present, both topology optimization technology and additive manufacturing technology
are quite mature, but their application in the design of spatial nodes is very recent and less researched.
This paper involves the study and improvement of the node optimization design–manufacturing
integrated method. This study used the BESO optimization algorithm as the research algorithm.
Through a reasonable improvement of the material interpolation method, the algorithm’s dependence
on the experience of selecting the material penalty index P was reduced. On this basis, the secondary
development was carried out, and a multisoftware integration was carried out for optimization and
manufacturing. The spatial node was taken as the research object, and the calculation results of the
commercial finite element software were compared. The comparison showed that the algorithm
used in this paper was better. Not only was it not trapped in a local optimum, but the maximum
stress was also lower. In addition, this paper proposed a practical finite element geometric model
extraction method and smoothing of the optimized nodes, completing the experiment of the additive
manufacturing forming of the nodes. It provides ideas for processing jagged edges brought by
the BESO algorithm. This paper verified the feasibility of the multisoftware integration method of
optimized manufacturing.
Keywords: improved BESO algorithm; optimal design of space nodes; material interpolation method;
additive manufacturing; model postprocessing
1. Introduction
The evolutionary structural optimization (ESO) method and its improved method
(BESO) were first proposed by Xie [1,2] et al. The method gradually removes the ineffi-
cient materials in a structure so that the structure can gradually reach the optimal state
in terms of performance. At the same time, this method has an extensive prospect for
secondary development. With the continuous advancement of computer technology and
3D-printing technology, topology optimization is no longer satisfied with just finding the
shape, but has turned its attention to new technologies that are integrated with additive
manufacturing. Regarding topology optimization techniques for additive manufacturing,
there has been a large amount of research beneficial to practical applications. First, in the
study of manufacturing molding, the researchers considered the process constraints of
additive manufacturing technology, including self-supporting constraints [3], maximum
and minimum constraints [4,5], and anisotropic constraints [6,7]. This type of research is
https://www.mdpi.com/journal/materials
Materials 2022, 15, 3874 2 of 17

key to the fusion of topology optimization and additive manufacturing. The second is
the defects and properties of additive manufacturing materials and components [8]. The
researchers studied the anisotropy of additive manufacturing components. It turned out
that this anisotropy is ubiquitous, not only for metal materials [9], but also for polymer
materials [10,11]. Although additive manufacturing encounters common manufacturing
precision errors, the mechanical properties of additively manufactured products can be well
matched in experiments and finite element analysis [12], which is beneficial to obtaining
accurate and stable material properties for topology optimization analysis. According to
statistics [13], there are currently more than 200 additive manufacturing technologies world-
wide. Among the many fields of additive manufacturing services, the research on additive
manufacturing in the construction field began late. The reason is that, firstly, the design of
various structures has specific specifications, and secondly, the current commercial topol-
ogy optimization software is not open enough to meet the customization requirements. In
space-structure design, because space nodes require a high strength and a light weight,
they have great potential in optimized manufacturing, which has aroused the high interest
of researchers. Existing studies include bifurcated nodes [14], rod-tie nodes [15], cable
dome nodes [16], and aluminum alloy nodes for exterior wall facades [17]. The fusion
of topology optimization and additive manufacturing makes the design of space nodes
more accessible and convenient, especially complex models after topology optimization,
which can also be easily manufactured. These studies were all based on using the SIMP
method, and the current BESO method still has some problems in practical application:
less software and the existence of rough edges after optimization. Based on MATLAB and
Abaqus, this paper carries out the secondary development of the BESO method, discusses
the further improvement of the material method with an aim to establish a convenient
and free integrated method, and then provides ideas for the optimal manufacturing of
space nodes.
This study took space nodes with complex forces as the research object. The following
work was done using the combination of 3D printing and topology optimization. First,
the traditional material interpolation method was improved. The improved material
interpolation formula could automatically penalize according to the material’s properties.
Secondly, based on the secondary development of ABAQUS, a platform for the integration
process of structure from optimization to manufacturing was designed by using MATLAB,
and the molding experiment was carried out. Finally, the optimization results of this study
were compared with the results of the Tosca module that is supplied with ABAQUS. By
comparison, it was found that the optimization results reported this paper were better
than the results of Tosca in terms of the mechanical properties. When the percentage of
volume optimization was low, the shape of the optimization results found in this study
were more reasonable.

2. BESO Topology Optimization Method

2.1. Mathematical Description of Topology Optimization
This article only discusses the topology optimization of space nodes under static loads.
Taking the minimum structural flexibility as the objective function, the strain energy density
Xi of the element as the design variable, and the volume fraction as the constraint condition,
the optimization problem can usually be described as:
(
min C = 21 FT U
(1)
s.t. V ∗ − ∑in Vi = 0

In the formula, F T represents the transposition of the load matrix, U represents the
displacement matrix, V* is the target volume fraction, Vi is the volume of the ith unit, Xi is
Materials 2022, 15, 3874 3 of 17

the strain energy density of the ith unit, and n is the total number of structural units. It can
be equivalent to: (
min C = ∑in=1 Vi Xi
(2)
s.t. V ∗ − ∑in=1 Vi = 0
Under the condition of satisfying the maximum structural rigidity, the volume fraction
will be continuously reduced until it is reduced to the target volume fraction.

2.2. Improved Material Interpolation Method

The current commonly used material interpolation method is an interpolation method
widely used in the variable density method [18,19]. This method can tend the density of
the unit to 0/1 and make the density division of the unit continuous. When the BESO
method uses the material interpolation model [20,21], in order to make the division of the
element clearer, it is often necessary to define a very small soft element parameter Xmin to
make it infinitely close to 0 to replace the void area after the element is deleted [22]. This
is to prevent the sensitivity of the soft cell from being directly equal to 0 and causing a
matrix singularity.
In the past, the penalty index value that needed to be selected was 1.8 in order to
produce practical effects. At the same time, it was affected by the penalty index and the
artificially selected minimum value, as well as the two material properties. Not only was
the efficiency low, but it also had many parameters and was greatly affected by humans.
In order to improve this problem and improve the adaptability of the algorithm, we
made necessary modifications to it:

E( Xi ) = ( E1 − E2 )∧ (1 − p) + E2 (3)

where p is the penalty index of the material, E1 is the elastic modulus of the solid element
material, and E2 is the elastic modulus of the soft element material. Then, the material
interpolation method is:
  
 12 1 − EE2 UT KU
 for solid element material1
1
ai =  ∧  (4)
( E1 − E2 ) (1− P)
 12

∧ UT KU for soft element material2
( E1 − E2 ) (1− P)+ E2

where ai is the sensitivity of the element, K is the stiffness matrix calculated from the elastic
moduli E1 and E2 , and U is the displacement matrix.
When E1 > E2 and the difference is large, the above formula can be equivalent to:
  
 12 1 − EE2 UT KU for solid element material1
1
ai =   (5)
1 0
U T KU for soft element material2
2 0+ E2

which is:   
1 E2

2 1− E1 UT KU for solid element material1
ai = (6)
 0 for soft element material2
As shown in Figure 1, through the adjustment formula, when the penalty index p
started from 0, the soft cell sensitivity was in a reduced state.
Materials 2022,15,
Materials2022, 15,3874
x FOR PEER REVIEW 44 of
of 17
18

Figure1.1.The
Figure Thevariation
variationininsensitivity
sensitivitywith
withthe
thepenalty
penalty index
index p before
p before thethe improvement
improvement in the
in the mate-
material
rial interpolation method formula (black) and after improvement
interpolation method formula (black) and after improvement (red). (red).

Asthe
As thepenalty
penaltyindex
indexincreased,
increased,the thenumerator
numeratorof ofthe
thematerial
materialinterpolation
interpolationformula
formula
shrank, so
shrank, so the result ofof the
theformula
formulaalso alsoshrank.
shrank.ByBy multiplying
multiplying thethe
calculated
calculated value of the
value of
softsoft
the element
elementobtained
obtained bybythethe
finite
finiteelement
elementcalculation,
calculation,the thesensitivity
sensitivity value
value of thethe soft
soft
elementafter
element afterthe
thepenalty
penaltywaswasobtained,
obtained,which
whichgave
gavepractical
practicalmeaning
meaningto tothe
thesoft
softelement
element
asthe
as thedeleted
deletedpartpartin
inthe
thecalculation.
calculation. InIn addition,
addition, in in the
the following
following sensitivity
sensitivityfiltering,
filtering,thethe
use of this method could also reduce the contribution of the soft
use of this method could also reduce the contribution of the soft unit sensitivity. unit sensitivity.
Therefore,the
Therefore, theimproved
improvedmethodmethodplayed
playedthe
thefollowing
followingroles:
roles:
Toaa certain
To certain extent, thetheoptimization
optimizationinstability
instabilitycaused
caused bybythethe
artificial selection
artificial selectionof the
of
the p-value
p-value waswas weakened.
weakened. At the
At the samesame
time,time,
the the enhanced
enhanced penalty
penalty effect
effect can can be seen
be seen in
in Fig-
Figure 1; when
ure 1; when p was
p was greater
greater thanthan
0.5,0.5,
thethe sensitivity
sensitivity waswas reduced,
reduced, andandthe the penalty
penalty effect
effect had
had
begunbegun to occur.
to occur.
The artificially selected minimum
The artificially selected minimumvalue valueXX min
min was
was deleted,
deleted, andandthethe formula
formula result
result was
was changed to be automatically adjusted according to the elastic modulus
changed to be automatically adjusted according to the elastic modulus of the solid element of the solid
element
and soft and soft materials,
element element materials, which the
which increased increased the adaptability
adaptability of the material of the material
interpolation
interpolation
formula. formula.

2.3.
2.3. Optimization
Optimization Criteria
Criteria
The
The deletionand
deletion andretention
retentioncriteria
criteria of of
thetheunit generally
unit adopt
generally the the
adopt dichotomy
dichotomyto set
tothe
set
sensitivity threshold of the unit division, which can
the sensitivity threshold of the unit division, which can be defined as:be defined as:

1
Vth =V 2 ( a=max(a+ a+ )
 1 min
 amin )
1 th 2 max
Vth =2 ( amax +th) Vadd > V ∗ (7)
  1
 Vth =Vth(th 1 
∗*
= + a
(a + th) V < V (7)

2 nin
max Vadd >V
del
 2
 1
In the formula, Vth is the sensitivity
Vth = (th + threshold <V *determines whether the cell is
amin ) Vdelthat
 2
deleted or retained, amax is the maximum sensitivity, amin is the minimum sensitivity, Vadd is
the volume
In the of the solid
formula, Vthunit,
is theVdel is the volume
sensitivity of thethat
threshold unit, and V*
softdetermines is the target
whether volume.
the cell is de-
In order to realize the evolution process of the improved algorithm,
leted or retained, amax is the maximum sensitivity, amin is the minimum sensitivity, Vadd is the initial sensitiv-
ity
thethreshold
volume ofisthedefined
solid as VthV=del(aismax
unit, amin )/2.of the soft unit, and V* is the target volume.
the+volume
During the iteration process, we
In order to realize the evolution process made theoffollowing mathematical
the improved algorithm, description:
the initial sensi-
tivityWhen Vadd >isV*,
threshold then aas
defined minV=
th (a= (a
min max++Vath )/2.Vth = (amin +amax )/2, when Vadd < V*, then
)/2;
min
amax =During
(Vth + athe )/2; Vth =process,
maxiteration (amin + awe )/2. the following mathematical description:
maxmade
Materials 2022, 15, 3874 5 of 17

This continued until the goal was achieved; that is, when Vadd = V*, the iteration ended.

2.4. Optimization of the Process

When performing the analysis of the topology optimization solution of the 3D model,
the steps were as follows:
(1) Firstly, establish the ABAQUS finite element model that needs to be optimized, and
output the finite element calculation model INP file.
(2) ABAQUS is called through the MATLAB background to perform the finite element
analysis. After the calculation and analysis are completed, the ODB database func-
tion of Python is used to extract all the element design variables Xi of the solved
design domain.
(3) Then, the sensitivity analysis based on the improved material interpolation method is
performed on the extracted data in MATLAB, and then the optimization analysis is
performed based on the optimization criterion.
(4) The elements that need to be deleted and retained are divided into different element
sets, written as a new finite element model INP file, and iterated again until the target
volume fraction is reached and the optimization ends.
(5) After the optimization, extract the 3D model format file of the finite element model
composed of the retained element set, and perform the additive manufacturing.
Throughout the process, RHINO, ABAQUS, MATLAB, and Python were used for
Materials 2022, 15, x FOR PEER REVIEW 6 of 18
optimization, smoothing, and then 3D printing using the printer’s supporting software.
The process is shown in Figure 2 below.

Start
Python extracts the
calculated data in
Modeling through the .ODB file
Python
RHINO 3D software

Import the model Load into matlab for

into Abaqus for sensitivity analysis
ABAQUS
mesh division and based on material
load addition interpolation
method

Matlab background
calls Abaqus solver
MATLAB
for finite element
analysis Sensitivity filtering

Optimization
Analysis
Geometric model
Based extraction of finite
on the Unit division, write
element model
new new inp file
.inp
file

Whether to Model smoothing

reach the
No objective
function
3D printing
Yes

End of
End of optimization
manufacturing

Figure 2.
Figure 2. Optimization-manufacturing
Optimization-manufacturing software integration
software method
integration flow chart.
method flow chart.

3. Space Node Optimization-Manufacturing Example

3.1. Model Parameters and Modeling
There are various forms of space nodes, and the combination of node design and
topology optimization makes the diversity of node shapes fully developed. The designer
can optimize the established model according to the connection method of the nodes, load
Materials 2022, 15, 3874 6 of 17

3. Space Node Optimization-Manufacturing Example

3.1. Model Parameters and Modeling
There are various forms of space nodes, and the combination of node design and
topology optimization makes the diversity of node shapes fully developed. The designer
can optimize the established model according to the connection method of the nodes,
load application, and boundary conditions, etc., and obtain the topological results freely
generated based on actual needs after optimization. Therefore, in this design area, a
sufficient design space needs to be provided.
Materials 2022, 15, x FOR PEER REVIEW 7 of 18
In the 3D modeling software Rhino, we built a square node model [23] with ear plates
as shown in Figure 3.

Figure 3. The node model established using the 3D design software Rhino, including the design
domain
Figure and thenode
3. The part of the lug
model for applying
established thethe
using load.
3D design software Rhino, including the design
domain and the part of the lug for applying the load.
The model parameters are given in Table 1.
We imported the established model into the finite element software in a 3D model
Table 1. Structural parameters.
file format, added attributes to it, and divided the mesh. Figure 4 shows the finite element
model with meshes. The blue part in the figure is the design domain, and the red part is
Structure Size (mm) Density (kg/m3 ) Elastic Modulus (MPa) Poisson’s Ratio
the connecting part of the ear plate. The grid size of the design domain was 2 mm and the
Design domain × 80 was
80 unit × 80C3D8R, with a total
7850 of 64,000 units. 2.1 × 105 0.28
Ear plate thickness 10 7850 2.1 × 10 5 0.28
Ear plate height 40 7850 2.1 × 105 0.28

We imported the established model into the finite element software in a 3D model file
format, added attributes to it, and divided the mesh. Figure 4 shows the finite element
model with meshes. The blue part in the figure is the design domain, and the red part is
the connecting part of the ear plate. The grid size of the design domain was 2 mm and the
unit was C3D8R, with a total of 64,000 units.
Considering that the bending moment had little influence on the actual project, the
pure shear force, the pure axial force, and the combined force of the combination of the
two were mainly considered in the optimization. Six degrees of freedom constraints were
imposed on the end face of the connecting part to place one end in a fixed state. Figure 5 is
a schematic diagram of the working conditions of applying a shear load, axial load, and
combined load.

Figure 4. The node model of the finite element mesh division in ABAQUS (the red part is the load
application area, and the green part is the design domain).
 domain and the part of the lug for applying the load.

We imported the established model into the finite element software in a 3D model
file format, added attributes to it, and divided the mesh. Figure 4 shows the finite element
model with meshes. The blue part in the figure is the design domain, and the red part is
Materials 2022, 15, 3874 7 of 17
the connecting part of the ear plate. The grid size of the design domain was 2 mm and the
unit was C3D8R, with a total of 64,000 units.

Materials 2022, 15, x FOR PEER REVIEW

Figure 4. The node model of the finite element mesh division in ABAQUS (the red part is8the
of load
18
Figure 4. The node model of the finite element mesh division in ABAQUS (the red part is the load
application area,
application and
area, thethe
and green part
green is the
part design
is the domain).
design domain).

Considering that the bending moment had little influence on the actual project, the
pure shear force, the pure axial force, and the combined force of the combination of the
two were mainly considered in the optimization. Six degrees of freedom constraints were
imposed on the end face of the connecting part to place one end in a fixed state. Figure 5
is a schematic diagram of the working conditions of applying a shear load, axial load, and
combined load.

(a) (b) (c)

Figure 5. Schematic diagram of finite element application of different load cases (pink arrows are
Figure 5. Schematic diagram of finite element application of different load cases (pink arrows are
axial forces, yellow arrows are shear forces, orange and blue ends are where the boundary condi-
axial forces, yellow arrows are shear forces, orange and blue ends are where the boundary conditions
tions are applied, and the green part is design domain): (a) shear load; (b) axial load; (c) combined
are applied, and the green part is design domain): (a) shear load; (b) axial load; (c) combined load.
load.

InInthe
thefigure,
figure,the
theblue
blue area
area on
on the end face
the end face is
is the
thefixed
fixedend
endofofthe
theboundary;
boundary; the
the pink
pink
load
load is the axial force, which was uniformly applied to the end face of the ear plate in thethe
is the axial force, which was uniformly applied to the end face of the ear plate in
form
formofofsurface
surfaceload;
load;and
andthetheyellow
yellowarrow
arrowis is
the shear
the force,
shear which
force, whichwas
wasapplied to to
applied thethe
end
face of the ear plate in the form of a point load.
end face of the ear plate in the form of a point load.
3.2.
3.2.Optimization
OptimizationEffect
Effectbefore
before and
and after
after Improvement
Improvement
The
Theoptimization
optimizationresults
resultsneed
needto toconsider
consider the
the influence
influence ofof boundary burrs and
boundary burrs andcheck-
checker-
boards in actual applications. The improved material interpolation
erboards in actual applications. The improved material interpolation method greatly method greatly weak-
ened impact of the penalty index p,
weakened the impact of the penalty index p, and the penalty effect was more dependentthe
the and the penalty effect was more dependent on
gap
on between the material
the gap between properties.
the material WhenWhen
properties. the material properties
the material of the
properties softsoft
of the element
ele-
and the solid element were quite different in magnitude, the sensitivity
ment and the solid element were quite different in magnitude, the sensitivity was very was very adaptive,
and it had and
adaptive, a good
it hadperformance in the control
a good performance of the checkerboard.
in the control The sensitivity
of the checkerboard. could
The sensitivity
becould
adjusted to the infinitesimal range according to the difference between
be adjusted to the infinitesimal range according to the difference between the ma- the materials;
e.g., the e.g.,
terials; combined load case.
the combined loadAs shown
case. in Figure
As shown 6, the6,optimized
in Figure structure
the optimized had had
structure a good
a
printable shape.
good printable shape.
 on the gap between the material properties. When the material properties of the soft
ment and the solid element were quite different in magnitude, the sensitivity was v
adaptive, and it had a good performance in the control of the checkerboard. The sensiti
could be adjusted to the infinitesimal range according to the difference between the
Materials 2022, 15, 3874 8 of 17
terials; e.g., the combined load case. As shown in Figure 6, the optimized structure h
good printable shape.

Figure 6. Optimized molding effect after improvement with different angles, showing a smoot
Figure 6. Optimized molding effect after improvement with different angles, showing a smoother
profile (the red part is the load application area, and the green part is the design domain).
profile (the red part is the load application area, and the green part is the design domain).

TheThe material’s elastic

material’s elasticmodulus
modulus and and
targettarget
volumevolume
fraction were kept constant
fraction were keptwithconstant
no w
no checkerboard
checkerboard control. As
control. shown
As in
shownFigure
in 7, because
Figure 7,the unit
becausestress
thebecame
unit concentrated
stress became con
Materials 2022, 15, x FOR PEER REVIEW and complicated near the connection position of the ear plate, it was prone to unclear unit 9
trated and complicated near the connection position of the ear plate, it was prone
selection. Therefore, it can be seen that there was a local checkerboard effect at the position to
clear unitthe
where selection. Therefore,
ear plate was it to
connected can
thebe seen that there was a local checkerboard effe
node.
the position where the ear plate was connected to the node.
Compared with the optimization results in Figure 7, the improved optimization
sults in Figure 6 was obviously more printable.

Figure 7. Optimized
Figure 7. Optimizedmolding effect
molding effect before
before improvement
improvement with different
with different anglesisland
angles showing showing
units island u
(the red part is the load application area, and the green part is the design domain).
(the red part is the load application area, and the green part is the design domain).

Compared with the optimization results in Figure 7, the improved optimization results
3.3. in
Analysis
Figure 6 of
wasOptimization Results
obviously more under Different Working Conditions
printable.
3.3.1. Axial Force Acting Conditions
3.3. Analysis of Optimization Results under Different Working Conditions
In order
3.3.1. to make
Axial Force theConditions
Acting optimization results clearer, the target volume fraction wa
very low. Taking
In order a volume
to make fraction ofresults
the optimization 20% clearer,
as an example, the evolution
the target volume rateset
fraction was ER was 0
Thevery
objective function
low. Taking of the
a volume structural
fraction of 20% optimization was
as an example, the the maximum
evolution rate ER wasstiffness.
0.08.
The objective function of the structural optimization was the maximum stiffness.
Figure 8a,b show graphs of the optimization results under pure axial force co
tions. ItFigure
can be8a,b show graphs of the optimization results under pure axial force conditions.
seen that the optimized structure topology was clear, and the unit
It can be seen that the optimized structure topology was clear, and the unit area between
between
adjacentadjacent
lugs waslugs wasunder
curved curved underofthe
the action action
axial force,of axialconformed
which force, which
to its conformed
stress t
stress condition.
condition. ThereThere was
was no no checkerboard
checkerboard phenomenon
phenomenon as a whole,asand
a whole, and the conne
the connectivity
ity was good.
was good.
As shown in Figure 8b, there was a uniform higher stress area between adjacent
plates, while the middle part of the node had a lower stress value, and the overall f
was uniform. This meant that the structure as a whole was in a state of high-efficie
stress, which was in line with the actual stress situation. Figure 9 shows the change in t
compliance versus the volume fraction for the overall optimization process. With the
tinuous reduction in the volume fraction, the total compliance increased, indicating
 plates, while the middle part of the node had a lower stress value, and the overall force
was uniform. This meant that the structure as a whole was in a state of high-efficiency
stress, which was in line with the actual stress situation. Figure 9 shows the change in total
compliance versus the volume fraction for the overall optimization process. With the con-
tinuous reduction in the volume fraction, the total compliance increased, indicating that
Materials 2022, 15, 3874 9 of 17
the remaining material was used efficiently. When the volume fraction reached the target
value, the total compliance fluctuated continuously, and eventually converged.

(a) (b)
Figure
Figure 8.
8. Schematic
Schematicdiagram
diagramof
ofoptimization
optimizationresults
resultsunder
underaxial
axialload
loadconditions:
conditions:(a)
(a)cutaway
cutawayview;
view;
(b) stress nephogram.
(b) stress nephogram.

As shown in Figure 8b, there was a uniform higher stress area between adjacent ear
plates, while the middle part of the node had a lower stress value, and the overall force
was uniform. This meant that the structure as a whole was in a state of high-efficiency
stress, which was in line with the actual stress situation. Figure 9 shows the change in
total compliance versus the volume fraction for the overall optimization process. With the
continuous reduction in the volume fraction, the total compliance increased, indicating that
Materials 2022, 15, x FOR PEER REVIEW 10 of 18
the remaining material was used efficiently. When the volume fraction reached the target
value, the total compliance fluctuated continuously, and eventually converged.

Figure 9. Optimization process under the condition of axial force (the black dotted line is the change
Figure 9. Optimization process under the condition of axial force (the black dotted line is the change
curve of compliance C, and the red dotted line is the volume-reduction curve).
curve of compliance C, and the red dotted line is the volume-reduction curve).
3.3.2.Shear
3.3.2. ShearConditions
Conditions
Similarly,when
Similarly, whenthethetarget
targetvolume
volumefraction
fractionwas
wassetsettoto20%,
20%,the
theevolution
evolutionrate
ratewas
was
0.08, the optimization target was the maximum stiffness, and the shear condition
0.08, the optimization target was the maximum stiffness, and the shear condition analysis analysis
wasperformed.
was performed.
Figure10a,b
Figure 10a,b show
show thethe stress
stress nephograms
nephograms of the
of the optimization
optimization results.
results. UnderUnder the ac-
the action
oftion of shear,
shear, thecan
the node node can be approximately
be approximately regarded regarded as theoffusion
as the fusion three of three cantilever
cantilever beams.
Asbeams.
shown Asinshown
Figurein10a,
Figure
an 10a, an X-shaped
X-shaped supportsupport structure
structure appearedappeared
in the in the middle
middle of theof
the node,
node, whichwhich conformed
conformed to topology
to the the topology of the
of the cantilever
cantilever beambeam under
under shear
shear force.The
force. The
optimization results had no checkerboard phenomenon
optimization results had no checkerboard phenomenon as a whole. as a whole.
 0.08, the optimization target was the maximum stiffness, and the shear condition analysis
was performed.
Figure 10a,b show the stress nephograms of the optimization results. Under the ac-
tion of shear, the node can be approximately regarded as the fusion of three cantilever
beams. As shown in Figure 10a, an X-shaped support structure appeared in the middle of
Materials 2022, 15, 3874 10 of 17
the node, which conformed to the topology of the cantilever beam under shear force. The
optimization results had no checkerboard phenomenon as a whole.

(a) (b)
Figure 10. Schematic diagram of optimization results under shear conditions: (a) lateral view; (b)
Figure 10. Schematic diagram of optimization results under shear conditions: (a) lateral view;
stress nephogram.
(b) stress nephogram.

FromFrom
thethe point
point of of view
view of of
thethe stressdistribution,
stress distribution,thethemaximum
maximum valuevalue appeared
appeared atat the
the connection between the node and the ear plate. After optimization, the overall stressdis-
connection between the node and the ear plate. After optimization, the overall stress
tribution of
distribution of the
the structure
structurewas wasuniform,
uniform,indicating
indicating that
thatthethe
overall structure
overall waswas
structure in ain
highly
a
efficient
highly state
efficient of of
state stress.
stress.InInFigure
Figure11,11, as the volume
as the volumefraction
fractioncontinued
continued to to decrease,
decrease, the the
average
average flexibility
flexibility increased
increased untiluntil the target
the target volumevolume
fractionfraction was reached,
was reached, and theand the flexi-
flexibility
bility value began to converge. This purpose of this stage was
value began to converge. This purpose of this stage was to continuously optimize to continuously optimize
the
Materials 2022, 15, x FOR PEER REVIEW the element distribution without changing the target volume. The flexibility value 11 ofgradu-
18
element distribution without changing the target volume. The flexibility value gradually
ally converged
converged accurately,
accurately, andthe
and finally, finally, the optimized
optimized optimization
optimization results
results were were obtained.
obtained.

Figure 11. Optimization process under shearing conditions (the black dotted line is the change curve
Figure 11. Optimization process under shearing conditions (the black dotted line is the change curve
of compliance C, and the red dotted line is the volume-reduction curve).
of compliance C, and the red dotted line is the volume-reduction curve).

3.3.3. Combined Load Conditions

In the same way, the combined
way, the combined working
working condition
condition analysis
analysis was
was carried
carried out.
out.
Figures 12 and 13 show the optimization results and the optimization process under
the combined
combinedload,load,respectively.
respectively. Figure
Figure 12a12a shows
shows thatthat the force
the force transfer
transfer paths paths
of the of the
whole
whole structure
structure were effectively
were effectively utilized,
utilized, and Figure
and Figure 12b shows
12b shows thatstress
that the the stress distribution
distribution was
was uniform.
uniform. The modeling
The modeling conformed
conformed to thetocombination
the combination
of theof the optimization
optimization resultsresults
under un-
the
twothe
der single
twoworking conditions
single working of axialofforce
conditions axialand shear
force and force.
shear The upper
force. part ofpart
The upper theof
node
the
was
nodeaffected by thebyaxial
was affected forceforce
the axial to maintain a closed
to maintain diamond
a closed top surface,
diamond andand
top surface, under the
under
action of the shear force, the energy of the upper element was strengthened
the action of the shear force, the energy of the upper element was strengthened and the and the energy
of the lower
energy of theelement was weakened,
lower element so theso
was weakened, upward arc-shaped
the upward element
arc-shaped accumulation
element accumu-
appeared
lation at the end
appeared ofend
at the the boundary constraint.
of the boundary constraint.
 whole structure were effectively utilized, and Figure 12b shows that the stress distribution
was uniform. The modeling conformed to the combination of the optimization results un-
der the two single working conditions of axial force and shear force. The upper part of the
node was affected by the axial force to maintain a closed diamond top surface, and under
the action of the shear force, the energy of the upper element was strengthened and the
Materials 2022, 15, 3874 11 of 17
energy of the lower element was weakened, so the upward arc-shaped element accumu-
lation appeared at the end of the boundary constraint.

Materials 2022, 15, x FOR PEER REVIEW (a) (b) 12 of 18

Figure
Figure 12.12.Schematic
Schematicdiagram
diagramofofoptimization
optimizationresults
resultsunder
undercombined
combinedconditions:
conditions:(a)(a)cutaway
cutawayview;
view;
(b) stress nephogram.
(b) stress nephogram.

Figure
Figure 13.13. Optimization
Optimization processunder
process undercombined
combinedworking
working conditions
conditions (the
(the black
blackdotted
dottedline
lineisisthe
the
change
change curve
curve of of compliance
compliance C,C,and
andthe
thered
reddotted
dottedline
line is
is the volume-reduction
volume-reductioncurve).
curve).

3.4. Algorithm Comparison

3.4. Algorithm Comparison
In order to compare and discuss the advantages and disadvantages of the BESO
In order to compare and discuss the advantages and disadvantages of the BESO al-
algorithm, the optimized result of the SIMP algorithm was selected as the comparison
gorithm, the optimized
object here. As one of theresult of themainstream
current SIMP algorithm was selected
algorithms, the SIMPasalgorithm
the comparison object
is embedded
here.
in itAsasone of the current
an optimization mainstream
tool algorithms,
in many finite the SIMPprograms.
element software algorithmThe is embedded
Tosca module in it
as an
that comes with ABAQUS uses the SIMP algorithm as an optimization solution tool. that
optimization tool in many finite element software programs. The Tosca module In
comes
order with ABAQUS
to ensure uses the SIMP
the reliability of the algorithm
comparison, as this
an optimization solution
section substituted thetool.
entireInnode
order
to ensure
model abovethe reliability
into ABAQUS of thetocomparison,
solve. this section substituted the entire node model
above intoThe ABAQUS
comparisontoused solve.
nodes under shear conditions and complex combined conditions
asThe
the comparison
comparisonobject. The objective
used nodes function
under shear set in ABAQUS
conditions was still
and complex the minimum
combined condi-
strain
tions energy,
as the the volume
comparison was The
object. usedobjective
as the constraint
functioncondition, and the value
set in ABAQUS wasthe
was still 20%.
mini-
mum strainAs shown
energy, in the
Figure 14, when
volume the SIMP
was used as the algorithm
constraint was used for and
condition, optimization,
the value the
was
stress
20%. of the overall structure under the shear condition was uniform, the area of maximum
stress was concentrated
As shown in Figure 14, in the
whenmiddle of the algorithm
the SIMP thin rod, and wasthere
usedwasforaoptimization,
clear difference. the
As shown in Table 2, the maximum stress of the SIMP algorithm results is 10.12% larger
stress of the overall structure under the shear condition was uniform, the area of maxi-
than the BESO algorithm results, and the minimum stress is smaller than the BESO results.
mum stress was concentrated in the middle of the thin rod, and there was a clear differ-
It should be noted that when using SIMP to optimize, the volume in the first step was
ence. As shown in Table 2, the maximum stress of the SIMP algorithm results is 10.12%
reduced to 20%. The stress distribution of the BESO optimization results was more uniform,
larger than the
indicating BESO
that algorithm
the whole results,
structure was and
in athe minimum
highly stress
efficient loadisstate,
smaller
whichthan the BESO
effectively
results. It should be noted that when using SIMP to optimize, the volume in the first step
was reduced to 20%. The stress distribution of the BESO optimization results was more
uniform, indicating that the whole structure was in a highly efficient load state, which
effectively reduced the damage caused by the stress concentration in the structure. The
Materials 2022, 15, 3874 12 of 17

reduced the damage caused by the stress concentration in the structure. The circle in
Figure 15 is the unit that was deleted by mistake in the first step, and the unit was not
supplemented even in subsequent iterations. In the calculation of BESO, compared with
Materials 2022, 15, x FOR PEER REVIEW 13 of 18
SIMP’s volume-shrinkage method, its method of gradually reducing the volume ensured
that the weak force-transmitting members of the structure could be retained.

Materials 2022, 15, x FOR PEER REVIEW 13 of 18

(a) (b)
Figure 14.
Figure 14. The
Thestress
stressnephograms
nephogramsofof
the design
the domain
design of the
domain optimized
of the jointjoint
optimized under the shear
under con-
the shear
dition: (a) BESO; (b) SIMP.
condition: (a) BESO; (b) SIMP.

Table 2. Comparison of optimization results of shear conditions.

Algorithm Maximum Stress (MPa) Minimum Stress (MPa)

(a) (b)
BESO 10,180 11.34
Figure 14. The stress nephograms of the design domain of the optimized joint under the shear con-
SIMP
dition: (a) BESO; (b) SIMP.
11,210 4.1 × 10−6

Figure 15. Stress nephogram of SIMP optimization results under shear conditions.

As shown in Table 3, the overall stress of the results obtained by the SIMP algorithm
appeared to be lower, but the optimization results of the SIMP algorithm were still far
from the
Figure 15. BESO
Stress algorithm,
nephogram and
of
Figure 15. StressSIMPthe optimization
maximum
nephogram of SIMPstress
resultswas 25.48%
under
optimization shearhigher.
results Comparing
conditions.
under shear conditions.Figure
16 and Figure 17, it can be seen that SIMP described the skeleton of the structure more
As shown in Table
three-dimensionally, As 3,shown
and the
theoverall
in Tablestress
appearance ofwas
3, the the smoother.
overallresults
stressobtained
of the by the
Theresults
BESO SIMP algorithm
obtained by the
algorithm SIMP
used ap-
algorithm
the
small squares of each unit of the structure as the carrier of design variables, so the optimi- still far
peared to be appeared
lower, but the to be lower,
optimization but the
results optimization
of the SIMP results of
algorithm the SIMP
were algorithm
still far from were
the
BESO from the BESO algorithm,
stressand the maximum stress was 25.48%Figures
higher. 16 Comparing
and 17, Figure
zationalgorithm,
results alsoand the
used maximum
the small squares was 25.48%
as the higher.
structural Comparing
characterization unit. Although
it can be seen that16 and
SIMP Figure 17,
described it can be seen
the skeleton that SIMP described the skeleton of the structure more
BESO showed a three-dimensionally,
more excellent performance,
and the
it of the
still
appearance
structure
needs
was
be more
tosmoother. three-dimensionally,
smoothedThe
before
BESO
printingused the
algorithm
and
and the appearance was smoother. The BESO algorithm used the small squares of each
manufacturing. small squares of each unit of the structure as the carrier of design variables, so the optimi-
zation results also used the small squares as the structural characterization unit. Although
BESO showed a more excellent performance, it still needs to be smoothed before printing
and manufacturing.
Materials 2022, 15, 3874 13 of 17

unit of the structure as the carrier of design variables, so the optimization results also used
the small squares as the structural characterization unit. Although BESO showed a more
excellent performance, it still needs to be smoothed before printing and manufacturing.

Table 3. Comparison of optimization results of combined working conditions.

Algorithm Maximum Stress (MPa) Minimum Stress (MPa)

Materials
Materials2022,
2022,15,
15,xxFOR
FORPEER
PEERREVIEW
REVIEW 1414ofof1818
BESO 4421 58
SIMP 5600 2.2 × 10−6

(a)
(a) (b)
(b)
Figure
Figure 16.
Figure16. Stress
16.Stress cloud
Stresscloud
clouddiagramininin
diagram
diagram the
thethedesign
design
design domain
domain
domain ofofoptimized
optimized
of optimized joints
joints
joints under
underundercombined
combined
combined condi-
condi-
conditions:
tions:
tions:(a)
(a)BESO;
BESO; (b)
(b)SIMP.
SIMP.
(a) BESO; (b) SIMP.

Figure 17.Stress
Figure17. Stresscloud
clouddiagram
diagramofofSIMP
SIMPoptimization
optimizationresults
resultsunder
results undershear
under shearconditions.
shear conditions.
conditions.
4. Optimizing Results
Table for Additive Manufacturing
Table3.3.Comparison
Comparisonofofoptimization
optimizationresults
resultsofofcombined
combinedworking
workingconditions.
conditions.
4.1. Model Postprocessing
Algorithm Maximum Stress (MPa) Minimum Stress
Stress(MPa)
WhenAlgorithm Maximum
using ABAQUS finite element Stress
analysis (MPa) under
software, Minimum (MPa)
normal circumstances,
BESO
BESO
it is impossible to modify the geometric features 4421 58
4421 of the model in the generated
58 INP file.
SIMP
SIMP 5600
5600 2.2
2.2××10
−6
10−6

4.4.Optimizing
OptimizingResults
ResultsforforAdditive
AdditiveManufacturing
Manufacturing
4.1. Model Postprocessing
4.1. Model Postprocessing
When using ABAQUS finite element analysis software, under normal circumstances,
 Materials 2022, 15, 3874 14 of 17
Materials 2022,
Materials 15,15,
2022, x FOR PEER
x FOR PEERREVIEW
REVIEW 15 15
of of
18 18

In this paper, a geometric model extraction method based on the finite element model of
WhenWhen thethe BESO
BESO algorithm
algorithm waswas combined
combined withwith thethefinite
finite element
element software,
software, in inorder
order
the element set partition was innovated, and the optimized element set geometric model
to to
ensure
ensure thethe two-way
two-way addition
addition andand deletion
deletion of ofelements,
elements, thethe deleted
deleted andand retained
retained ele-
ele-
extracted the OBJ format.
ments were divided into two material sets: set-1 and set-2, respectively.
ments were divided into two material sets: set-1 and set-2, respectively. Therefore, the part Therefore, the part
When the BESO algorithm was combined with the finite element software, in order to
of of
thetheunitunitsetset that
thatneeded
needed to to
bebe deleted
deleted waswas still retained
still retained in in thethe final
final optimization
optimization result
result
ensure the two-way addition and deletion of elements, the deleted and retained elements
model.
model. The The method
method usedused in in
this paper
this paper needed
needed to to
extract
extract thethe optimized
optimized results
results fromfrom thethe
were divided into two material sets: set-1 and set-2, respectively. Therefore, the part of the
INPINP filefile
after
after optimization.
optimization.
unit set that needed to be deleted was still retained in the final optimization result model.
The The model
model extraction
extraction method
method is is
defined
defined asasfollows:
follows:
The method used in this paper needed to extract the optimized results from the INP file
after optimization.  M M=M = 1M1 M
 eM e
The model extraction method   is defined as follows:
 N =
N N=1 N 1 N
2N 2 (8)(8)
 ' 
 N N= N=  =
'
NMN 2N 2M1 ∩ Me

N = N1 ∩ N2 (8)
InIn thethe formula,
formula, MM 1 is thethe
1 is element
element number  matrix
number matrix
0 of ofthethe first material
first material setsetset-1.
set-1. MeMise is
N = N ∩ N2
thethe
matrix
matrix of ofallall
elements
elements andand their
theirassociated
associated nodenode number.
number. Through
Through thethe intersection
intersection of of
MM 1 and
1 and MM e, the Inmatrix
e, the matrix
the MM
formula,containing
M 1 is the
containing the unit
the number
unit
element number
number of of
thethe first
matrix material
first material
of setset
the first and andthethe
material num-num-
set set-1. M e
berber
of of
itsitsassociated
is associated
the matrix nodes
nodes
of all is elements
obtained.
is obtained. NN
and istheir
the
is thematrix
matrix
associatedof of
thethe intersection
node number.ofThrough
intersection thethe
of nodenode num-
the num-
intersection
berberof of
thethe
offirst
M material
first
1 and M e ,element
material element
the and
matrix and
M the node
the node
containing number
number
the unitof ofthe the
numbersecond
second ofmaterial firstelement,
material
the element,
material Nset
1N1and the

is is
thethenodenode number
number of the
of the(set-1)
(set-1)solid
solidelement
element unit,
unit,and andN Nis the
is
number of its associated nodes is obtained. N is the matrix of the intersection of the node
2 2 thenode
node number
number of the
of soft
the soft
element
element (set-2).
(set-2).
number N′ofisthe
N′ thethe
is sequence
first sequence
material number
number
element matrix
matrix
and the of ofall nodes
all
node nodes
number thatthatareare
of thenotnot related
second related to to
material thetheelement,
entity
entity unit.
unit.
N 1 is the node number of the (set-1) solid element unit, and N 2 is the node number of the
The The BESO
soft BESO method
element method
(set-2).was N 0realized
was isrealized through
the sequence through thethecollaboration
number collaboration
matrix of ofMATLAB
of all nodes MATLAB
that are not and andrelated to
ABAQUS.
ABAQUS. theHowever,
However,
entity unit. thethe optimized
optimized model
model obtained
obtained bybythethe BESO BESO method
method at at
this time
this time was was
a model
a model composed
composed
The BESO of ofunitunit
methodblocks,
blocks,
was andand thethe
realized model
model
through had had
thea jagged
acollaboration
jagged boundary,
boundary, which
of MATLAB which isand
not
is not
ABAQUS.
friendly
friendly to printing.
to printing.
However, InIn
the order
order
optimized to to
obtain
obtain
model a smooth
aobtained
smooth boundary,
boundary,
by the BESO thethe smoothing
smoothing
method atfunction
function
this time of was
thethea model
of
three-dimensional
composedmodeling
three-dimensional ofmodeling software
unit blocks, andRhino
software the was
model
Rhino was used
hadusedatojagged
toadjust
adjust it, it,
and
boundary, and the boundary
which
the is not friendly
boundary
smoothing
smoothing toprocess
printing.
process was Inperformed
was order to obtain
performed byby a smooth
adjusting
adjusting boundary,such
parameters
parameters the
such smoothing
as thethe
as smoothingfunction
smoothing coeffi-of the three-
coeffi-
cient
cientand anddimensional
thethe amount
amount modeling
of smoothing.
of software
smoothing. ExamplesRhinoofwas
Examples thethe
of used to adjust results
optimization
optimization it,results
and under
theunder
boundary
thethe com- smoothing
com-
bined
bined load process
load condition waswith
condition performed
with by adjusting
checkerboard
checkerboard control
control parameters
areare shown
shown such
in as the
Figures
in Figures 18smoothing
and
18 and19.19. coefficient and
AllAll the
the the amount
jagged
jagged of smoothing.
boundaries
boundaries of of Examples
thethe
model
model were of the
were optimization
smoothed,
smoothed, and andtheresults
model
the modelunder
waswas the
in combined
a print- load
a print-
in
able
ablestate condition
stateasas a whole. with
a whole. checkerboard control are shown in Figures 18 and 19.

(a)(a) (b)(b)
Figure 18.18.
Figure Smooth
Figure model
Smooth
18. of of
model
Smoothdifferentofangles
different
model under
angles
different shear
under conditions:
shear
angles (a)(a)
conditions:
under front; (b)
front;
shear conditions: side.
(b)
(a) side.(b) side.
front;

(a)(a) (b)(b)
Figure 19.19.
Figure Smooth
Figure model
Smooth
19. of of
model
Smooth different
model of angles
different under
angles
different undercombined
angles combined
under working
working
combined conditions: (a)(a)
conditions:
working top;
conditions: (b)(b)
top;
(a) bot-
top;bot-
(b) bottom.
tom.
tom.
All the jagged boundaries of the model were smoothed, and the model was in a
printable state as a whole.
Materials 2022, 15, x FOR PEER REVIEW 16 of 18
Materials 2022, 15, x FOR PEER REVIEW 16 of 18
Materials 2022, 15, 3874 15 of 17

4.2.
4.2. 3D-Printing
3D-Printing Effect Effect
4.2. 3D-Printing
Regarding Effect
Regarding topological results,
topological results, Yang
Yang Zhao
Zhao etet al.
al. [24]
[24] printed
printed andand manufactured
manufactured cable cable
and Regarding topologicalWang results, Yang Zhao et al. and
[24] manufactured
printed and manufactured cable
and rod nodes, and Longxuan Wang et al. [25] printed and manufactured four-forked
rod nodes, and Longxuan et al. [25] printed four-forked
steel and rod
casting nodes,
nodes. Bothand Longxuan
used a method Wang
of et al. [25]lost
combining printed
wax and manufactured
casting, which was afour-forked
more
steel casting nodes. Both used a method of combining lost wax casting, which was a more
steel casting nodes.
economical, Both manufacturing.
used a method of this combining lost wax casting, which was a
economical, practical
practical method
method of of manufacturing. In In this method,
method, aa node
node moldmold was was manu-
manu-
factured more by economical,
printing a practical
wax mold, method
and then of manufacturing.
the node was In this
manufactured method,
by a node
casting. mold was
factured by printing a wax mold, and then the node was manufactured by casting.
Inmanufactured by printing a wax mold, and then the node was manufactured by casting.
In this
this study,
study, related
related forming
forming experiments
experiments were were carried
carried out
out using
using the the optimized
optimized re- re-
sults. In this study, related forming experiments were carried out using the optimized results.
sults. TheThe STLSTL format
format filefile of
of the
the smoothed
smoothed model model waswas exported
exported fromfrom Rhino
Rhino and and loaded
loaded
into The STL format file of the smoothed model was exported from Rhino and loaded into the
into the
the software
software
software
supporting
supporting
supporting
the
the
theThe
additive
additive
additive
manufacturing
manufacturing
manufacturing
machine
machine
machine for
for
for adjustment
adjustment
adjustment of
of
of the
the
the printing
printing
printing process
process parameters.
parameters. The support
support methods
methods were
were usually
usually column-shaped
column-shaped support
support
and process parameters. The support methods were usually column-shaped support and tree-
and tree-shaped
tree-shaped
shaped
support.
support.
support.
Compared
Compared
Compared with
with
with column-shaped
column-shaped
column-shaped
support,
support,
support, the
the
the tree-shaped
tree-shaped
tree-shaped support
sup-
sup-
can save
port
port can
can save
save material
material and
and provide
provide stronger
stronger support.
support. In
In this study,
thissome
study, some
some 44 mm
mm diameter
diameter
tree material and provide stronger support. In this study, 4 mm diameter tree supports
tree supports
supports
were
were
weretoadded
added added
support
to
to support
support
thethe
the
the printing
printing printing
of some
of
of some
some suspended
suspended suspended
parts. The
parts.
parts. The
The printing
printing printing
process used
process
process used
used in
in this
this study
study was
was the FDM
FDM fused
fused deposition
deposition manufacturing
manufacturing printing
printing tech-
tech-In the
nology. in this
In study was the FDM fused deposition manufacturing printing technology.
nology. In the
researchthe research
research
on the
on
on the
the properties
properties properties
of
of
of PLA-printed
PLA-printedPLA-printed
parts,
parts,
parts, Sohrabian
Sohrabian Sohrabian
and
and
and Vaseghi
Vaseghi Vaseghi
[12] et
[12]
[12]
al. carried
et
et al.
al. carried
carried out
out detailed
detailed experiments
experiments and
and finite
finite element
element analysis
analysis of
of aa scaffold
scaffold architecture
architecture
out detailed experiments and finite element analysis of a scaffold architecture printed in
printed
printed in
in PLA
PLA [12]. The
The mechanical properties of the
the PLA materials were good, and
and could
PLA [12]. [12]. mechanical
The mechanical properties
properties of theofPLA PLA materials
materials werewere good,good,and could could
guarantee
guarantee
guarantee good
good manufacturing
manufacturing accuracy.
accuracy. Therefore,
Therefore, it
it was
was very
very suitable
suitable as
as the
the manufac-
manufac-
good manufacturing accuracy. Therefore, it was very suitable as the manufacturing material
turing
turingfor material
material for
for this
this model.
model. The
The parameters
parameters are shown in Table 4.
this model. The parameters are shownare shown
in Table 4.in Table 4.
Table 4. Process parameter table for the printing molding experiment.
Table Table
4. Process parameter
4. Process table for
parameter thefor
table printing molding
the printing experiment.
molding experiment.
Speed
Speed
Print Type
PrintPrint
Type Type Material
Material
Material Speed (mm/s)Fill
Fill Rate
Fill Rate Temperature
Rate (°C)
Temperature
Temperature
◦
(°C) ( C)
(mm/s)
(mm/s)
FDMFDM PLA
PLA 60 60 15% 15% 220 220
FDM PLA 60 15% 220

Figures
Figures 20
20 and
Figures
and2021
21 show
and the
the optimization
21 show
show results
the optimization
optimization after
after the
results
results the improved
after material
the improved
improved interpo-
material
material interpola-
interpo-
lation.
lation. The results were postprocessed and the overall structure was clear. There was no
The
tion. results
The were
results postprocessed
were postprocessedand the
and overall
the structure
overall was
structure clear.
was There
clear. was
There was nono fine
fine treelike,
treelike,difficult-to-print
difficult-to-printarea,
area,and
andit could
it couldbe
beapplied
applied to
toadditive
additivemanufacturing.
manufacturing.
fine treelike, difficult-to-print area, and it could be applied to additive manufacturing.

FigureFigure
20. Print
20.effect
Printpicture after improvement
effect picture (Under(Under
after improvement shear conditions).
shear conditions).
Figure 20. Print effect picture after improvement (Under shear conditions).

Figure 21. Print effect picture after improvement (combined working condition).
FigureFigure
21. Print
21.effect
Print picture after improvement
effect picture (combined
after improvement working
(combined condition).
working condition).
Materials 2022, 15, 3874 16 of 17

5. Conclusions
Based on the self-developed optimization-manufacturing integrated approach, this
paper proposed an improved material interpolation sensitivity calculation method. The
modeling comparison experiment on the space nodes verified that the integrated forming
topology method had good feasibility. The conclusions are given below:
(1) The material interpolation method of the optimization algorithm was improved, the
penalty effect of the material method was strengthened, and the adaptability was
increased. Compared with the SIMP method, the secondary developed BESO method
did not easily fall into the optimal local solution, and the performance of the obtained
optimization results was better.
(2) Based on the multisoftware collaborative computing platform independently devel-
oped by the research group and the model extraction method proposed in this paper,
there was no need to modify the model grid and edit it again, so the method could
easily and quickly realize the entire process of design-manufacturing.
This method combined the algorithm and the finite element analysis software in the
form of a third-party plug-in, and completed the additive molding of the space node very
well. In the future, its openness can provide a good framework platform for the addition of
various constraints in the further refinement of topology technology.

Author Contributions: Data curation, G.Z. and Z.W.; Methodology, F.Z., X.W. and Y.Z.; Software, F.Z.
and G.Z.; Validation, Z.W., X.W. and Y.Z. All authors have read and agreed to the published version
of the manuscript.
Funding: The project was supported by the Key Laboratory of Civil Engineering Structure and
Mechanics, Inner Mongolia University of Technology (No. 2021KY002).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Xie, Y.M.; Steven, G. A simple evolutionary procedure for structural optimization. Comput. Struct. 1993, 49, 885–896. [CrossRef]
2. Yang, X.Y.; Xie, Y.M.; Steven, G.P.; Querin, O.M. Bidirectional evolutionary method for stiffness optimization. AIAA J. 1999, 37,
1483–1488. [CrossRef]
3. Langelaar, M. Topology optimization of 3D self-supporting structures for additive manufacturing. Addit. Manuf. 2016, 12, 60–70.
[CrossRef]
4. Zhang, W.; Li, D.; Zhang, J.; Guo, X. Minimum length scale control in structural topology optimization based on the Moving
Morphable Components (MMC) approach. Comput. Methods Appl. Mech. Eng. 2016, 311, 327–355. [CrossRef]
5. Zhou, M.; Lazarov, B.S.; Wang, F.; Sigmund, O. Minimum length scale in topology optimization by geometric constraints. Comput.
Methods Appl. Mech. Eng. 2015, 293, 266–282. [CrossRef]
6. Dapogny, C.; Estevez, R.; Faure, A.; Michailidis, G. Shape and topology optimization considering anisotropic features induced by
additive manufacturing processes. Comput. Methods Appl. Mech. Eng. 2018, 344, 626–665. [CrossRef]
7. Li, S.; Yuan, S.; Zhu, J.; Wang, C.; Li, J.; Zhang, W. Additive Manufacturing-driven Design Optimization: Building Direction and
Structural Topology. Addit. Manuf. 2020, 36, 101406. [CrossRef]
8. Lian, Y.P.; Wang, P.D.; Gao, J.; Liu, J.K.; Li, Q.H.; Liu, C.M.; He, X.F.; Gao, L.; Li, H.; Lei, H.S.; et al. Research progress on some key
mechanical problems in metal additive manufacturing. Adv. Mech. 2021, 51, 648–701.
9. Alsalla, H.H.; Smith, C.; Liang, H. Effect of build orientation on the surface quality, microstructure and mechanical properties of
selective laser melting 316L stainless steel. Rapid Prototyp. J. 2018, 24, 9–17. [CrossRef]
10. Rezayat, H.; Zhou, W.; Siriruk, A.; Penumadu, D.; Babu, S.S. Structure-mechanical property Relationship in fused deposition
modelling. Mater. Sci. Technol. 2015, 31, 895–903. [CrossRef]
11. Carneiro, O.S.; Silva, A.F.; Gomes, R. Fused deposition modeling with polypropylene. Mater. Des. 2015, 83, 768–776. [CrossRef]
12. Sohrabian, M.; Vaseghi, M.; Khaleghi, H.; Dehrooyeh, S.; Kohan, M.S.A. Structural Investigation of Delicate-Geometry Fused
Deposition Modeling Additive Manufacturing Scaffolds: Experiment and Analytics. J. Mater. Eng. Perform. 2021, 30, 6529–6541.
[CrossRef]
13. Li, X.F. Classification and Analysis of Foreign Additive Manufacturing Standards. Electr. Mach. Tool. 2020, 5, 56–59.
Materials 2022, 15, 3874 17 of 17

14. Han, L.Y.; Du, W.F.; Xia, Z.; Ye, J.; Gao, B.Q. Research on generative intelligent design of four-forked cast steel joints. J. Build.
Struct. 2022, 29, 1–13.
15. Ji, P.; Chen, Y.L.; Jang, T.; Banon, C. Topology Optimization of Airmesh Metal Additive Manufacturing Nodes. Space Struct. 2021,
27, 71–77.
16. Zhou, Y.C.; Zhao, Y.; Xiang, X.A. Research on topology optimization of cable dome nodes in Ya’an Tianquan Gymnasium. Space
Struct. 2021, 27, 74–80.
17. Strauss, H. AM Envelope: The Potential of Additive Manufacturing for Facade Constructions; TU Delft: Delft, The Netherlands, 2013.
18. Sigmund, O. A 99 line topology optimization code written in Matlab. Struct. Multidiscip. Optim. 2001, 21, 120–127. [CrossRef]
19. Liu, K.; Tovar, A. An efficient 3D topology optimization code written in Matlab. Struct. Multidiscip. Optim. 2014, 50, 1175–1196.
[CrossRef]
20. Huang, X.; Xie, Y.M. Bi-directional evolutionary topology optimization of continuum structures with one or multiple materials.
Comput. Mech. 2009, 43, 393–401. [CrossRef]
21. Bendsøe, M.P.; Sigmund, O. Material interpolation schemes in topology optimization. Ingenieur-Archiv 1999, 69, 635–654.
[CrossRef]
22. Wang, X.J.; Zhang, X.A. Multiphase material layout and material/structure integrated design. Chin. J. Solid Mech. 2014, 35,
341–346.
23. Chen, M.C.; Zhao, Y.; Xie, Y.M. Topology optimization and additive manufacturing of spatial structure nodes. China Civ. Eng. J.
2019, 52, 1–10.
24. Zhao, Y.; Chen, M.C.; Wang, Z. Topology optimization design of cable and rod structure nodes for additive manufacturing.
J. Build. Struct. 2019, 40, 58–68.
25. Wang, L.X.; Du, W.F.; Zhang, F.; Zhang, H.; Gao, B.Q.; Dong, S.L. Topology optimization and 3D printing manufacturing of
four-branch cast steel joints. J. Build. Struct. 2021, 42, 37–49.