Article published in
Current Chinese Science, 2021, 1, 151-159
Lessons learnt from a national competition on
structural optimization and additive manufacturing
Yulin XIONG1, Dingwen BAO1, Xin YAN1,2, Tao XU1, Yi Min XIE1,*
1 Centre for Innovative Structures and Materials, School of Engineering, RMIT University, Melbourne 3001,
2 Centre of Architecture Research and Design, University of Chinese Academy of Sciences, Beijing 100190,
* Corresponding author. E-mail address: email@example.com (Y.M. Xie).
In a recent national competition on structural optimization and additive manufacturing in China, we won the first
prize with a highly efficient and elegant structural design. This paper describes important lessons learnt from the
competition. We used the bi-directional evolutionary structural optimization (BESO) technique for the
conceptional design and a smoothing technique to work out the details of the structural boundaries. This approach
was applied to the design of a hinge arm which must satisfy all the conditions set by the competition committee.
We carried out a nonlinear quasi-static loading simulation before the competition committee fabricated the design
using a stereolithography 3D printer. In the actual mechanical testing conducted by the national competition
committee, our design displayed superior performance to other entries. The insights gained from this award-
winning design will be useful for engineers and architects to employ topology optimization techniques to create
efficient and elegant structures in the future.
Keywords: Topology optimization; Bi-directional evolutionary structural optimization (BESO); Structural
optimization; Additive manufacturing; Load-bearing structural design; National competition;
As an advanced design technique, topology optimization has received much attention over the past three decades.
The aim of topology optimization is to find an optimal material layout to maximize the structural performance
while satisfying certain constraints. It is a useful tool for the conceptional design. The development of topology
optimization can be traced back to Michell’s work to minimize the weight of the truss layout . Later researchers
proposed several methods for topology optimization of continuum structures, e.g., the solid isotropic material with
penalization (SIMP) method , the bi-directional evolutionary structural optimization (BESO) method , 
and the level set method (LSM) , , etc. By changing the prescribed objective, these techniques can be applied
to various disciplines, including biomechanics , thermodynamics ,  acoustics  and optics , .
Conventional stiffness-based topology optimization is well developed and increasingly used for structural designs
in many fields, e.g., architecture , automotive  and aerospace industries . It provides conceptional
designs in the load-bearing system to maximize structural stiffness while meeting the volume constraint. Some
optimized designs for engineering applications such as railway vehicles  and aircraft components 
demonstrate the effectiveness of topology optimization. However, there are still some difficulties in achieving
optimized designs in practice. Firstly, for the element-based topology optimization techniques, the results contain
zig-zag boundaries . Thus, post-processing is necessary to determine the actual structural boundary , .
Secondly, topology-optimized designs are often accompanied by geometric complexity, which are hard to
manufacture by traditional machining. Additive manufacturing (AM) techniques provide a solution to this
problem . Benefiting from its characteristics of the layer-by-layer formation of 3D components , AM can
be integrated with topology optimization to create highly efficient products with complex geometries in practical
In this work, we describe an efficient methodology for realizing optimized designs in practice, which makes the
topology optimization technique a usable and accessible technology. The BESO technique provides the
conceptional design, and the topology-optimized result is post-processed to obtain smooth structural boundaries.
The case study that we present considers the load-bearing structural design of a hinge arm. The design satisfies
all the conditions set by the third national competition on structural optimization and additive manufacturing,
which was organized by China Aerospace Science and Industry Corporation in Beijing in 2019. This competition
is one of the most prestigious structural optimization contests in China, attracting entries from a large number of
experienced aerospace engineers and researchers. In total there were 135 entries in the preliminary competition,
of which only 18 teams got into the final. All the designs must be verified through numerical simulations by
designers and confirmed experimentally by the competition committee. During the actual loading test in the final
competition, our design carried the highest load and won the first prize, which validated the effectiveness of our
design methodology. The insights gained from this award-winning design will be useful for engineers and
architects to use BESO to create elegant and efficient structures in the future .
The paper is organized as follows. In Section 2, the BESO technique for designing load-bearing structures is
introduced. In Section 3, the post-process technique for reconstructing the optimized result is proposed. In Section
4, the methodology is applied to the design of a hinge arm. In Section 5, the numerical simulation and loading test
are used for investigating the mechanical behaviour of the design, and the advantages of the topology-optimized
design are discussed. In Section 6, the main conclusions of this study are summarized.
2. Methodology for designing lightweight and load-bearing structure
2.1. BESO method
In this work, we use the BESO technique to create efficient load-bearing structures. As an important branch of
topology optimization, the BESO technique is based on the simple concept of gradually removing inefficient
material from a structure and adding material to the most needed locations at the same time . It is widely
recognized owing to its high-quality topology solutions , ease of understanding and implementation , and
excellent computational efficiency .
Considering the design domain where the material will be distributed. The whole domain is modelled as
elements that fill in the domain. We denote as the design variable of element . Since the material has only
linear elastic behaviour, the problem of maximizing the load-bearing capacity of structures is equivalent to
maximizing structural stiffness. With a given volume constraint , the structural stiffness is maximized when the
compliance is minimized . The mathematical form of the optimization model can be expressed as:
F u u Ku
where is the structural compliance, and are the global force vector, global displacement vector, and
stiffness matrix, respectively. Different from other topology optimization methods, the design variable of the
BESO technique is of discrete values. and indicate the solid and void element , where
to avoid singularities during finite element analysis (FEA) . Using the adjoint method, the
compliance with regard to the change in the design variable can be found as 
( ) 1 ,
C x p x
= − = −
i i i i
u u u K u
where , and denote the displacement vector of element , solid elemental stiffness matrix, and the penalty
exponent of material interpolation scheme , respectively. The sensitivity of element in the BESO technique
is defined as
1 ( ) 1 1 ,
i i i
p x x x
= − = =
i 0 i
u K u
where denote the stiffness matrix of element . It should be noted that the term
in the elemental
sensitivity is the elemental strain energy that carried out from FEA.
The raw sensitivity is processed to avoid mesh-dependent solutions and checkerboard patterns . The and
are the filter radius and the distance between the elements and , respectively. An elemental sensitivity filter
scheme is used in each iteration, which is defined as follows.
where is the weight factor of the filter scheme, which is formulated as
ij min ij
w r r=−
The weight is independent of the elemental sensitivity and is calculated before the optimization process. To
improve the convergence of the BESO technique, the filtered sensitivity is further averaged with its historical
information . Here, the sensitivity of the current iteration is simply averaged with that of the previous
k k k
i i i
BESO updates design from the initial full design and gradually reduces the structural volume in each iteration.
The volume of next iteration is determined by an evolutionary ratio and the current structural volume
min( , (1 )).
V V V er
It is noted that once reaches objective volume , it will remain constant. The evolutionary ratio is 1% in
this work. A threshold is employed for updating the design variables according to the and their
sensitivities. The update scheme of BESO is defined as
1 if ,
min i th
The present scheme indicates that the design variables of solid elements are switched from 1 to if their
sensitivities are lower than the threshold, and the design variables of void elements are switched from to 1
if the sensitivities are higher than the threshold. The optimization algorithm is implemented by Python and linked
to Abaqus. The validity and convergence of the basic BESO computational framework have been analyzed and
verified in .
2.2. Smoothing technique for post-processing
The optimized results from BESO technique contain zig-zag boundaries, which may lead to the reduction of
structural performance and difficulty to manufacture. In order to obtain accurate structural boundaries, a
smoothing technique is proposed for reconstructing the element-based model [35-36].
As a well-known method for extraction of iso-surface, the marching cubes (MC) algorithm is integrated into the
smoothing technique to determine structural contour surface in the first step. The structural contour surface is
defined as a surface that represents points of a constant design variable (normally ) within a volume of
space. The algorithm traverses the design domain, taking eight adjacent points as an imaginary cube, then
determines the polygons that represent the part of the iso-surface within the cube. According to the states of eight
points, a total 256 () possible polygon configurations are defined in the algorithm. The polygon configurations
can be reduced to 23 patterns considering rotational symmetry. For computational efficiency reason, these patterns
are typically encoded and stored in a lookup table according to the states of the points in the cube. The vertex of
the generated polygons is placed along the cube’s edge by linearly interpolating the value of the points that are
connected by the edge . These individual polygons are then merged into the iso-surface. From the perspective
of computer graphics, the MC algorithm transformed the element-based model into a polygon mesh model (i.e.
the iso-surface) without changing the structural topologies. This model type is widely used in computer graphics
to generate polygon mesh from element-based mesh, for its advantages of small storage space and easy
modification. In computational graphics, there are many modified versions of the MC algorithm for certain
requirements (such as the type of polygons). In this work, the basic MC algorithm is found to be quite effective.
In this work, the Laplacian surface editing technique  is employed to smooth the structure for the
determination of the actuary structural boundaries. The technique encodes vertices of the iso-surface from
Euclidean coordinates to Lagrange coordinates. It provides a representation of the surface mesh, where the
reconstruction of global coordinates always preserves the geometric details. Thus, the structural topology would
not be changed during surface editing. The Laplacian coordinate of a vertex is defined as
j N i i
where and represent the neighbours of a vertex and neighbour set, respectively. And is the number
of the neighbour of . Surface editing operations can be efficiently and robustly applied to surface mesh with
Laplacian coordinates . When the surface is edited, the Laplacian coordinates of vertices are decoded into
Euclidean coordinates. Laplacian smoothing is one of the most common algorithms for mesh denoising. It
repeatedly and simultaneously adjusts the coordinate of each vertex in the mesh to the geometric centre of its
i j i
j N i i
Although the Laplacian smoothing algorithm is simple and efficient, it may produce an over smoothed result.
Some small structural features such as right angles and sharp boundaries might be lost during Laplacian smoothing.
An optional way is to convert the triangular mesh to a quadrangular mesh before Laplacian smoothing.
Quadrangular mesh could snap the sharp details, which is more suitable for mechanical models than triangular
3. Application of the methodology to design a hinge arm
The case study that we consider the design of a hinge arm. The design conditions are set by the competition
committee. As shown in Fig. 1a, the hinge arm is fixed on the base by four connecting bolts, with an upward load
on its cantilever. The optimization goal is to maximize the load-bearing capacity of the structure while meeting
volume constraint (40 ml).
The structure is fabricated by the stereolithography apparatus printer RS4500. The material used in 3D printing is
photosensitive resin, and its Young’s modulus and Poisson’s ratio is 2510 MPa and 0.41, respectively. Since the
material has only linear elastic behaviour, the optimization objective can be equivalent to minimizing structural
compliance. There are some limitations of additive manufacturing that need to be considered in topology
optimization. It is required that the structural member size is larger than 1 mm, and the wall thickness is no smaller
than 0.5 mm. Besides, there should be no enclosed void in the structure.
Fig. 1b shows the requirements of the design domain, including non-design domain, bolts positions and loading
position provided by the organization. It is noted that the height and length of the design domain are not restricted
(red dot lines). Therefore, determining the appropriate design domain size for topology optimization is the key to
achieving the structural design with high performance.
Fig. 1 (a) Boundary conditions and (b) requirements of the design domain.
With uncertain height and length of the design domain, several initial designs (Fig. 2a) were tested for finding an
appropriate design domain size. The design domain size of these models is changed according to the unreasonable
part in their results. Four tunnels are dug out in the initial design for installation of the bolts, and the left part is
devised for the installation of the loading tool and the connecting pin. The heights and length of these initial
designs are shown in Fig. 2a. The coarse mesh is used for determining the size of the design domain quickly.
Elements size is about . In order to simplify the finite element calculation, z-direction of
the upper surfaces of four bolt holes are fixed, and the inner surface of them are pinned to simulate the structure
being bolted to the base. The bottom face of the front of the structure is fixed in the z-direction as well to simulate
the supporting force provided by the base. The optimized results are shown in Fig. 2b. The top and left sides of
the optimized results are flat, which is unreasonable. These flat parts are caused by the limitation of the design
domain and indicate that the design domain should be increased in height and length.
Fig. 2 (a) Test design domains and (b) the optimized results.
The height, width and length of the final initial design is 95 mm, 60 mm and 165 mm, respectively. The height of
tunnels is 36 mm. A finer finite element mesh is employed to obtain an accurate optimal solution. Elements size
is about . It is noted that topology optimization based on a fine mesh might produce a
better solution, but more thin members. These thin members in the structure are easily broken during quasi-static
loading test. In order to improve the robustness and bearing capacity of the structure, the filter radius is set to 8
mm to obtain an optimized design with thicker members.
As shown in Fig. 3a, the BESO result is an arch shape supported by three cylinders. These cylinders extend into
six branches that connect to the two bolts holes in the middle, which are bolted to the base. External forces cause
an additional torsional load with the bolts in the middle as a rotational fulcrum. To balance the torsional load, the
BESO technique naturally produces a vertical cylinder at the front part of the structure. In this part, a balanced
torque can be generated by using the supporting force provided by the base. Besides, the front bolts are connected
to the cylinder, and they give tension to limit structural deformation in the length direction. The evolution history
of structural compliance and volume can be seen in Fig. 3b.
Fig. 3 (a) The BESO result and (b) evolution histories of the compliance and the volume fraction.
Because the mesh result is composed of finite elements, such as triangles or quadrangles for shell and cubes or
tetrahedrons for solid, the BESO result is always a mesh with a coarse, irregular surface. The Grasshopper plugin
Ameba (http://ameba.xieym.com) developed by XIE Technologies has been used to solve the problem. The
smooth technique in Ameba has a strong ability to smooth the irregular surfaces. To place the loading bar and
bolts tightly and stably, some solid elements are fixed to retain the accurate geometry details during the smoothing
process. Figure 4 illustrates the process of smoothing. Once the form of design is finalized, it has been imported
into Abaqus for finite element analysis to get the accurate structural performance feedback which helps to re-test
and fine-tune the form to fix some structural defects (e.g. stress concentration caused by sharp boundaries) and
ensure the design has a better structural performance than unmodified one based on keeping the basic generated
Fig. 4 (a) The original model, (b) the MC result and (c) the smoothed model.
4. Experimental setup
In the paper, the smoothed BESO result is manufactured in photosensitive resin by a stereolithography apparatus
printer. The load-bearing capacity of the specimen was measured experimentally through a quasi-static loading
test. The competition committee provides both the 3D printer and the experimental machine. The experiment was
carried out on the compression test machine, as shown in Fig. 5. The base is made of 10 mm thick steel plate and
is fixed to the supporting frame by bolts on both sides. There are four bolt holes with a diameter of 6.5 mm in the
middle. These bolt holes are spaced 30 mm apart in width and 45 mm apart in length. Structures are fixed
underneath the base by four bolts. To avoid the torsion, the load is transferred from the loading cell to
the cantilever of the specimen through a connecting pin.
Fig. 5 The setup of the loading test.
During the test, the displacement and force are measured with a sampling frequency of 20 Hz. The force signals
are detected by the loading cell and collected by an oscilloscope. The test machine loaded downwards; the
specimen is subjected to an upward load correspondingly. The loading process will continue until the specimen
is destroyed. The competition committee provides these machines. It should be noted that in this study, only the
vertical reaction force of the structure is considered during loading. Because the loading time of the specimen
cannot be accurately estimated, the oscilloscope has a total duration of 100 seconds. It records the force curve
during the loading test, and the highest value of the curve is the maximum load that the specimen can bear (i.e.,
load-bearing capacity). The designs from 17 teams were tested on-site during the competition. The measurement
was obtained for ambient temperatures of 20 degrees.
5. Results and discussion
The experimental loading history of the award-winning design in the paper is detected by the loading cell on the
top is shown in Fig. 6a. The loading apparatus consists of a pin and a clamp connected to it. The testing force
curve and the displacement curve are blue and red, respectively. It should be noted that since the data is recorded
after the connecting pin has contacted the structure and started loading, a positive force can be observed in the
initial stage. The speed of the loading cell is 10 mm/min. As the loading process continues, the force gradually
rises until the structure is destroyed. The peak force is 1.920 kN, which appears at about 90 s. Then the force drops
rapidly, which indicates that the structure is completely broken and detached from the base. As shown in Fig. 6a,
the horizontal section of the force curve (from 15s to 30s) may be caused by a slippage of the pin and clamp,
resulting in an underestimation of the structural load-bearing capacity.
Fig. 6 (a) Testing force and displacement curve and (b) simulation force and displacement curve of BESO
A quasi-static nonlinear FEA on the same configuration of the quasi-static loading test is employed to verify the
proposed methodology. The material model used in the simulation is plastic rather than linear elastic. In this
case, the maximum stress of the material is set to 37 MPa at the strain of 3%. Then the material yields and
breaks at the strain of 9%. The smoothed model is meshed and simulated to investigate the response of the
structure. The connecting pin is loaded with the speed of 10 mm/min. Z-direction of the upper surfaces of four
bolt holes are fixed, and the inner surface of them are pinned. The bottom face of the front of the structure is
fixed in the z-direction to simulate the support provided by the base. The simulation loading history and
displacement of the BESO design is shown in Fig. 6b. The trend can be seen in the simulation that the force
increases until the structure broke. The peak force is 2.11 kN, which is larger than the test result. There is no
force platform during the simulation. These phenomena confirm that the structural load-bearing capacity is
underestimated in the test. In addition, In the loading test, when the structure is broken, the testing machine
stops recording the force, so the curve after 90s is a horizontal line. In the simulation, however, after the
structure is broken, the remaining part still bears the load, so the force slightly increases.
The comparison of the broken structure and the unbroken structure is shown in Fig. 7. It can be observed that four
parts of the structure are destroyed at the same time. It indicates that the structure is of equal strength to the design.
It should be noticed that only one side of the middle bolts was broken, which is caused by the structural buckling.
It is recommended to introduce structural buckling constraint into the optimization in the future.
Fig. 7 The comparison between the unbroken and broken models.
In this paper, we present an effective topology optimization approach to maximizing the structural load-bearing
capacity and establish a procedure to achieve optimized structural designs. The bi-directional evolutionary
structural optimization (BESO) technique provides the conceptional design, and the smoothing technique is used
for determining the final structural boundary. This methodology is applied to the design of a hinge arm which
satisfy all the conditions set by the national competition on structural optimization and additive manufacturing.
The design is fabricated by a stereolithography 3D printer and tested by the competition committee. The
mechanical test validated the advantages of the methodology. In the loading test of the final competition, our
design carried the highest loading and won the first prize of the competition, clearly demonstrating the ability of
BESO to create efficient and elegant design. From the competition, we learnt important lessons for designing
structures with superior performance, e.g., the importance of defining an appropriate initial design domain, the
usage of a relatively large filter radius to create a robust design, and the crucial role played by the smoothing
technique to reduce stress concentrations in some regions. These insights will be useful for engineers and
architects to employ topology optimization techniques to create elegant and efficient structures.
We wish to thank the committee from China Aerospace Science and Industry Corporation for organizing the 3rd
national competition on structural optimization and additive manufacturing. We would like to thank team
members of the Centre for Innovative Structures and Materials at RMIT University for sharing ideas and providing
help. We also wish to acknowledge the technical support from Ameba software development team of Nanjing
Ameba Engineering Structure Optimization Research Institute. This project was supported by the National Natural
Science Foundation of China (51778283).
1. Michell AGM. The limits of economy of material in frame-structures. The London, Edinburgh, and Dublin
Philosophical Magazine and Journal of Science 1904; 8(47):589–97.
2. Bendsøe M. Optimal shape design as a material distribution problem. Structural Optimization 1989; 1:193–
3. Xie YM, Steven GP. Xie YM, Steven GP. A simple evolutionary procedure for structural optimization.
Computers and Structures 1993; 49(5): 885-896.
4. Huang X, Xie YM, Burry MC. A new algorithm for bi-directional evolutionary structural optimization. JSME
International Journal Series C 2006; 49(4):1091–1099.
5. Wang MY, Wang X, Guo D. A level set method for structural topology optimization. Computer methods in
applied mechanics and engineering 2003; 192(1–2):227–246.
6. Allaire G, Jouve F, Toader AM. Structural optimization using sensitivity analysis and a level-set method.
Journal of Computational Physics 2004; 194(1): 363-393.
7. Zhao Z-L, Zhou S, Feng XQ, Xie YM. On the internal architecture of emergent plants. Journal of the
Mechanics and Physics of Solids 2018; 119: 224-239.
8. Haertel JHK, Engelbrecht K, Lazarov BS, Sigmund O. Topology optimization of a pseudo 3D thermofluid
heat sink model. International Journal of Heat and Mass Transfer 2018; 121: 1073-1088.
9. Li Q, Steven GP, Querin OM, Xie YM. Shape and topology design for heat conduction by Evolutionary
Structural Optimization. International Journal of Heat and Mass Transfer 1999; 42(17): 3361-3371.
10. Azevedo FM, Moura MS, Vicente WM, Picelli R, Pavanello R. Topology optimization of reactive acoustic
mufflers using a bi-directional evolutionary optimization method. Structural and Multidisciplinary
Optimization 2018; 58(5): 2239-2252.
11. Li W, Meng F, Chen Y, Li Y fan, Huang X. Topology Optimization of Photonic and Phononic Crystals and
Metamaterials: A Review. Advanced Theory and Simulations 2019; 2(7): 1900017.
12. Christiansen RE, Wang F, Sigmund O, Stobbe S. Designing photonic topological insulators with quantum-
spin-Hall edge states using topology optimization. Nanophotonics 2019; 8(8): 1363-1369.
13. Tang J, Xie YM, Felicetti P. Conceptual design of buildings subjected to wind load by using topology
optimization. Wind and Structures 2014; 18(1): 21-35.
14. Cavazzuti M, Baldini A, Bertocchi E, Costi D, Torricelli E, Moruzzi P. High performance automotive chassis
design: a topology optimization based approach. Structural and Multidisciplinary Optimization 2011; 44(1):
15. Zhu JH, Zhang WH, Xia L. Topology optimization in aircraft and aerospace structures design. Archives of
Computational Methods in Engineering 2016; 23(4): 595-622.
16. Das R, Jones R, Xie YM. Optimal topology design of industrial structures using an evolutionary algorithm.
Optimization and Engineering 2011; 12(4): 681-717.
17. Krog L, Tucker A, Rollema G. Application of topology, sizing and shape optimization methods to optimal
design of aircraft components. 3rd Altair UK HyperWorks users conference, UK, 2002.
18. Liu H, Tian Y, Zong H, Ma Q, Wang MY, Zhang L. Fully parallel level set method for large-scale structural
topology optimization. Computers & Structures 2019; 221: 13-27.
19. Huang X. Smooth topological design of structures using the floating projection. Engineering Structures 2020;
20. Liu B, Jiang C, Li G, Huang X. Topology optimization of structures considering local material uncertainties
in additive manufacturing. Computer Methods in Applied Mechanics and Engineering 2020; 360: 112786.
21. Xiong Y, Yao S, Zhao Z-L, Xie YM. A new approach to eliminating enclosed voids in topology optimization
for additive manufacturing. Additive Manufacturing 2020; 32: 101006.
22. Pham DT, Gault RS. A comparison of rapid prototyping technologies. International Journal of machine tools
and manufacture 1998; 38(10-11): 1257-1287.
23. Xie YM, Yang K, He Y, et al. How to obtain diverse and efficient structural designs through topology
optimization. In: Proceedings of IASS Annual Symposia. International Association for Shell and Spatial
Structures (IASS), Barcelona, Spain, October 7-10, 2019; pp. 1-8.
24. Huang X, Xie YM. Evolutionary Topology Optimization of Continuum Structures: Methods and Applications.
John Wiley & Sons 2010.
25. Zhao Z-L, Zhou S, Feng XQ, Xie YM. Morphological optimization of scorpion telson. Journal of the
Mechanics and Physics of Solids 2020; 135: 103773.
26. Zhao Z-L, Zhou S, Cai K, Xie YM. A direct approach to controlling the topology in structural optimization.
Computers & Structures 2020; 227: 106141.
27. Xia L, Xia Q, Huang X, Xie YM. Bi-directional evolutionary structural optimization on advanced structures
and materials: a comprehensive review. Archives of Computational Methods in Engineering 2018; 25(2):
28. Zhu J-H, Yang K-K, Zhang W-H. Backbone cup – a structure design competition based on topology
optimization and 3D printing. International Journal for Simulation and Multidisciplinary Design Optimization
2016; 7: A1.
29. Chen A, Cai K, Zhao Z-L, et al. Controlling the maximum first principal stress in topology optimization.
Structural and Multidisciplinary Optimization 2020; 1-13.
30. Huang X, Xie YM. A further review of ESO type methods for topology optimization. Structural and
Multidisciplinary Optimization 2010; 41(5): 671-683.
31. Bendsøe MP, Sigmund O. Material interpolation schemes in topology optimization. Archive of applied
mechanics 1999; 69(9-10): 635-654.
32. Sigmund O, Petersson J. Numerical instabilities in topology optimization: a survey on procedures dealing
with checkerboards, mesh-dependencies and local minima. Structural optimization 1998; 16(1): 68-75.
33. Huang X, Xie YM. Convergent and mesh-independent solutions for the bi-directional evolutionary structural
optimization method. Finite Elements in Analysis and Design 2007; 43(14): 1039-1049.
34. Zuo ZH, Xie YM. A simple and compact Python code for complex 3D topology optimization. Advances in
Engineering Software 2015; 85: 1-11.
35. He Y, Cai K, Zhao Z-L, et al. Stochastic approaches to generating diverse and competitive structural designs
in topology optimization. Finite Elements in Analysis and Design 2020; 173: 103399.
36. Yang K, Zhao Z-L, He Y, et al. Simple and effective strategies for achieving diverse and competitive
structural designs. Extreme Mechanics Letters 2019; 30: 100481.
37. Ju T, Losasso F, Schaefer S, Warren J. Dual contouring of hermite data. Proceedings of the 29th annual
conference on Computer graphics and interactive techniques 2002; 339-346.
38. Sorkine O, Cohen-Or D, Lipman Y, Alexa M, Rössl C, Seidel HP. Laplacian surface editing. Proceedings of
the 2004 Eurographics/ACM SIGGRAPH symposium on Geometry processing 2004; 175-184.
39. Zhang Y, Hamza A Ben. Vertex-based diffusion for 3-D mesh denoising. IEEE Transactions on image
processing 2007; 16(4): 1036-1045.
40. Jakob W, Tarini M, Panozzo D, Sorkine-Hornung O. Instant field-aligned meshes. ACM Trans Graph 2015;
The material used in 3D printing is photosensitive resin, and its Young’s modulus is 2510 MPa and Poisson’s
ratio is 0.41. The material model used in the loading simulation is plastic, and the maximum stress of the
material is set to 37 MPa at the strain of 3%. The material breaks at the strain of 9%.
List of Abbreviations
SIMP: solid isotropic material with penalization
BESO: bi-directional evolutionary structural optimization
LSM: level set method
MC: marching cubes
This project was supported by the National Natural Science Foundation of China (51778283).
ETHICS APPROVAL AND CONSENT TO PARTICIPATE
AVAILABILITY OF DATA AND MATERIALS
The data supporting the findings of the article is available from the authors.
CONSENT FOR PUBLICATION
There is no individuals’ data in this study.
CONFLICT OF INTEREST
There is no conflict of interest in the study.