Topological 6

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Additive Manufacturing 36 (2020) 101406

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Research Paper

Additive manufacturing-driven design optimization: Building direction and T


structural topology
Shaoying Lia, Shangqin Yuana,c,*, Jihong Zhua,b,*, Chuang Wanga, Jiang Lia, Weihong Zhanga,**
a
State IJR Center of Aerospace Design and Additive Manufacturing, School of Mechanical Engineering, Northwestern Polytechnical University, 710072, Xian, Shaanxi,
China
b
MIIT Lab of Metal Additive Manufacturing and Innovative Design, NPU-QMUL Joint Research Institute, Northwestern Polytechnical University, 710072, Xian, Shaanxi,
China
c
Unmanned System Research Institute, Northwestern Polytechnical University, 710072, Xian, Shaanxi, China

ARTICLE INFO ABSTRACT

Keywords: Additive manufacturing (AM) has been adopted by high-value-added industries to revolutionize product life
Additive manufacturing cycle performance, from advantageous design topology to functional advances. Compared with conventional
Stereolithography apparatus manufacturing methods such as tooling and molding, the design constraints induced by AM usually include
Topology optimization support structures, building direction (BD), feedstock material properties, and process parameters. These new
Material anisotropy
factors need to be considered in AM-driven product design and manufacturing. In this study, an AM-driven
Solid anisotropic material with penalization
Building direction
topology optimization method coupled with a transversely isotropic material model and solid anisotropic ma-
terial with penalization (SAMP) is proposed to establish a quantitative correlation between process-related
parameters and the mechanical properties of printed materials, further implementing such correlation for pro-
cess and topology optimization using a gradient-based algorithm. Specifically, the coordinate transformation
matrix is combined with the transversely isotropic stiffness and strength of the printed material using stereo-
lithography apparatus (SLA) to describe the elastic matrix under different BDs. Thereafter, case-dependent
product performances are investigated based on an integrated method considering the net-effect of structural
design and BD. The proposed approach can easily achieve AM-driven topology optimization of complex products
with desirable mechanical performance. Furthermore, the established topological model can be broadly applied
to complex functional part design and optimization, as well as case studies on AM-driven product evaluation.

1. Introduction selective laser melting (SLM), selective laser sintering (SLS), stereo-
lithography apparatus (SLA), and fused deposition molding (FDM),
Additive manufacturing (AM) is a revolutionary transformation of have been widely applied in the aerospace and automotive industries
design and manufacturing that is versatile and flexible for building [12–14].
three dimensional (3D) objects by digitally depositing and selectively SLA is one of the most significant AM technologies [15]. In SLA
solidifying materials through a layer-by-layer approach [1,2]. AM- process, the liquid photocurable resin converts into a solid polymer
driven topology optimization (TO) designs have been applied in aero- upon the selective ultraviolet (UV) light exposure in a two-dimensional
space engineering owing that they are lightweight and adaptable [3]. (2D) plane. The process of resin deposition and curing are repeated
TO design has made remarkable progress in practical engineering ap- layer-by-layer until the 3D components are completed [16]. The ‘green’
plications [4–6] since the pioneering works were proposed by M.P. components are usually post-cured by UV oven to enhance stiffness and
Bendsøe and N. Kikuchi [7]. Recently, the manufacturability of TO strength [17]. The high-resolution SLA is able to achieve complex
products has been strongly aided by the advanced development of AM structures such as lattice with superior dimensional accuracy [18]. In
technologies [8]. The integration of TO and AM broadens the freedom addition, an increasing number of commercially available printable
in structural design and manufacturing of end-use components [9–11]. resins are developed with a cost-effective advantage in prototyping and
Structures fabricated by AM processes, including, but not limited to, practical applications [19].

Corresponding authors at: State IJR Center of Aerospace Design and Additive Manufacturing, School of Mechanical Engineering, Northwestern Polytechnical

University, 710072, Xian, Shaanxi, China.


⁎⁎
Corresponding author.
E-mail addresses: shangqin.yuan@nwpu.edu.cn (S. Yuan), jh.zhu@nwpu.edu.cn (J. Zhu), zhangwh@nwpu.edu.cn (W. Zhang).

https://doi.org/10.1016/j.addma.2020.101406
Received 4 December 2019; Received in revised form 22 April 2020; Accepted 19 June 2020
Available online 24 June 2020
2214-8604/ © 2020 Elsevier B.V. All rights reserved.
S. Li, et al. Additive Manufacturing 36 (2020) 101406

However, different combinations of process parameters result in a anisotropic materials and structures [45]. It is confirmed that the in-
large variance of products in terms of geometric accuracy and me- creasing ratio of anisotropy and the deviation of BD have a significant
chanical performances, such as stiffness, strength, and toughness. These effect on the stiffness of the optimized structures [46]. Although this
instabilities and uncertainties of product performances usually bring research counts anisotropy of materials and inspired TO design, it lacks
new challenges to obtain reliable and high-performance products with systematic methods focusing on the quantitative influences of aniso-
customized functions for end-use applications [20]. The factors of layer tropic mechanical properties on TO designs. For specific 3D printed
thickness, post-curing period, and building direction (BD) exert sig- anisotropic materials, it is essential to establish a universal method for
nificant impacts on the microstructure and dimensional accuracy of quantifying the elastic matrix, further integrating BD into the TO
SLA-printed products [21]. A 24-day aging cycle makes SLA-printed method.
specimens stiffer and more brittle than the newly printed group due to The quantitative method of anisotropic mechanical properties is
the increased cross-linking within the polymeric matrix [22]. Scanning limited for layer formed materials, which obstructs the further appli-
strategy and material selection can be developed to reduce residual cation of AM in complex structures. In addition, the conventional TO
stress [23]. BD directly affects the surface quality and mechanical approach utilize the ‘ideal’ material property and it causes the final TO
property of the tensile samples [24,25]. Post-curing operations improve design cannot achieve anticipated performance due to anisotropic me-
the mechanical performance of printed structures by UV and heat chanical behaviors. To address the above research gaps, an AM-driven
source [26]. The AM process brings unique features to the final printed optimization design method is proposed to integrate structural design
components. and AM process, which quantify the relationship between process-re-
In AM processes, the anisotropy property induced by the layer-by- lated parameters and material property and introduce property para-
layer construction process is often observed [27], which significantly meters into structural TO design. Hence, the transversely isotropic
affects the strategy of TO for the design of load-bearing structures via elastic matrix is applied to describe the mechanical coefficients corre-
AM [28]. The ‘bottom-up’ manufacturing constructs a CAD model in a sponding to SLA-printed components. Through off-axis experiments, the
layer-by-layer manner to create discrete 3D CAD models, resulting in elastic matrix is obtained using a systematic data-driven approach to
different mechanical strengths along the vertical direction and trans- predict the orientation-dependent trend of anisotropic material prop-
verse plane [29–31]. D. Feldman and A. Barbalata [32] reported that erties. The accuracy of the anisotropic model is validated by experi-
the strength of the SLA-printed components depended on the in- mental results, and the orientation-dependent functions of material
tramolecular bonding strength of the polymer chain and then enhanced properties are further implemented in structural optimization to obtain
by the cross-linkers between these longchain-macromolecules. More- designs with desirable mechanical performance. Moreover, the AM-
over, the cross-linking reaction influenced the degree of curing and was driven TO method can be generally applied to other AM technologies in
affected by layer thickness, BD, and post-curing [33]. The SLA-printed a layer-by-layer building manner.
components typically exhibit an anisotropic mechanical behavior on
account of the inhomogeneous degree of curing which is usually caused 2. Modeling and experiment
by the uncured and partially cured resins trapped within the photo-
polymer [34]. To date, although intensive efforts have been devoted to 2.1. Modeling
eliminating anisotropy from the printed components, the anisotropic
properties of printed materials are still inevitable, adversely affecting The transversely isotropic material model is proposed to instruct the
structural performance [17,21,22]. Therefore, it is necessary to con- experimental design, which provides the rationale to obtain the elastic
sider anisotropic variables induced by the layer-by-layer building pro- matrix of SLA-printed components. According to the engineering re-
cess in the AM-driven TO design. quirements, two universal optimization models are selected to design
The transversely isotropic property is the preliminary observation, case-dependent products. One model is adopted to optimize BD for
and it is well-known in long fiber-reinforced composites [35]. Thus, it specific structures for the sake of promoting mechanical performance
follows that the anisotropic elastic model of composites can be applied by simply changing the posture of parts in the printing platform. The
to quantify the BD-dependent mechanical properties. In SLA process, other model integrates the optimization of BD and structural topology
feeding materials are superimposed layer-by-layer, inducing a uniform concurrently, which improves TO designs by implementing the or-
microstructure and delaminated features among inner layers. There- ientation-dependent mechanical stiffness to utilize the AM process-in-
fore, transverse isotropy is used to describe the anisotropic behaviors of duced anisotropic effects.
3D printed components. Transversely isotropic mechanical property
was applied to the SLA-printed parts, and the experimental results 2.1.1. Anisotropic elastic material model
confirmed that TO structural designs considering anisotropic mechan- As illustrated in Fig. 1, parts are built layer-by-layer along axis z in
ical property were much stronger than traditional TO designs [36]. For the SLA system and the UV laser scans in a straight line. According to
other AM technologies, M. Domingo-Espin et al. [37] determined the the curing law of photosensitive resin, the cured shape of a single
stiffness matrix of fused deposition molding (FDM) printed material by scanning line is a parabola in the micro-scale [15]. Therefore, as shown
tensile experiments. A. Amado-Becker et al. [38] characterized and in Fig. 1(b), the cured time and cured degree of resin are different in the
modeled engineering constants of sintered Nylon 12 through ultrasound multi-line scanning process, causing mechanical properties along BD to
propagation velocity in specific directions of selective laser sintering be quite different from those in the other two directions. Micro-
(SLS) printed material. It proves that composite mechanics covers the structures of the SLA-printed components and the zoomed diagram-
anisotropy of the single-material printing, which is induced by the matic sketch of layer features are exhibited in Fig. 1(c) and (d), re-
layer-by-layer process. spectively. The alternation of inter layer and inner layer generates the
BD is an important process parameter influencing the mechanical mode of ‘strong-weak-strong’ material due to the non-uniform curing of
property, surface quality, support structures, building time, and product photo-resin. It mainly results in the anisotropic properties of SLA-
cost [39–43]. Recently, AM-driven structural TO considering BD-in- printed components.
duced anisotropic properties has been investigated to improve me- Fig. 2 plots the SLA-printed components and long fiber-reinforced
chanical performance. For considering the material anisotropy in state- composites in their local coordinate system. The positive direction of
of-the-art AM-driven TO, the numerical models have drawn more at- axis 3 represents BD and fiber orientation in Fig. 2(a) and (b), respec-
tention than experimental validation. Directional variables were in- tively. The anisotropy caused by lamination of SLA-printed components
troduced into TO formulation without experimental validation of the shows transversely isotropic mechanical behaviors like long fiber-re-
optimized designs [44]. A strength-based TO method was presented for inforced composites. Thus, the SLA-printed components can be treated

2
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 1. (a) Diagrammatic sketch of SLA equipment; (b) The cured cross-section of single laser scanning and multiple laser scanning; (c) Microscopic image of SLA-
printed components; (d) Schematic representation of the layered features.

as a special kind of composite material. The transversely isotropic materials in the local coordinate system (BD is the positive direction of
elasticity is introduced to describe the SLA-printed components owing axis 3) can be applied to the TO algorithm and the specific constants are
to the uniform mechanical properties in the 1–2 coordinate plane. determined by statistics characterization. However, the local coordinate
Five engineering constants that can be obtained by off-axis tensile system is not consistent with the global coordinate system if BD is
experiments to determine the elastic matrix according to elasticity changed.
theory. The flexibility matrix is formulated as follows: Euler rotation was carried out in the main direction of the printed
components, and the elastic matrix under the global coordinate system
1
was obtained through the transformative matrix. The positive and ne-
E1 gative rotating angles followed the convention of the right-handed co-
v12 1
ordinate system, and the rotation sequence was defined as around axis 1
E1 E1 firstly and then around axis 2.
v13 v13 1 The rotating order is first around axis 1 and then around axis 2. Due
sym
E1 E1 E3 to the symmetry of the transversely isotropic model, the rotation
Cm =
1 around axis 3 in the global coordinate system (BD) does not affect
0 0 0
G13 mechanical properties. Therefore, two directional variables α and β
1 (which are both in the range of -90° to +90°) were applied to describe
0 0 0 0
G13 the anisotropic direction.
1 Based on linear elasticity [47] and the derivation of the rotating
0 0 0 0 0
G12 (1) matrix [44], the elastic matrix in the global coordinate system was
obtained by Eq. (3).
The elastic matrix is given by Eq. (2):

Dm = Cm1 (2) D = R ( ) Dm R ( )T (3)

The elastic matrices are Dm and D in the local and global coordinate Where,
systems, respectively. The elastic matrix of transversely isotropic

Fig. 2. (a) The SLA-printed components; (b) Long fiber-reinforced composites.

3
S. Li, et al. Additive Manufacturing 36 (2020) 101406

cos2 ( ) sin2 ( )sin2 ( ) sin2 ( )cos2 ( ) sin( )sin(2 ) sin(2 )sin2 ( ) sin(2 )cos( )
0 cos2 ( ) sin2 ( ) 0 sin(2 ) 0
sin2 ( ) sin2 ( )cos2 ( ) cos2 ( )cos2 ( ) sin( )sin(2 ) sin(2 )cos2 ( ) sin(2 )cos( )
sin(2 )sin( ) sin(2 )sin( )
R= 0 cos( )cos( ) sin( )cos(2 ) sin( )cos( )
2 2
sin(2 )cos( ) sin(2 )cos( )
0 sin( )cos( ) cos(2 )cos( ) sin( )sin( )
2 2
sin(2 ) sin2 ( )sin(2 ) sin(2 )cos2 ( ) sin(2 )sin(2 )
sin( )cos(2 ) cos( )cos(2 )
2 2 2 2 (4)

2.1.2. Optimization for BD E


=
1 T K
U U
SLA-printed components exhibit different mechanical properties in i 2 i (10)
various BDs due to the anisotropic features induced by the layer-by-
With the consideration of Eq. (5), the sensitivity of strain energy
layer process. Thus, it exists a direction that maximizes the stiffness of
concerning the rotating angle θi can be written as follows:
the printed components.
The current work was accomplished based on the cooperation of E 1 T BTR ( ) Dm RT ( ) Bd
Altair HyperWorks®, ANSYS®, and Boss-Quattro®, and the optimization = U U
i 2 i
model was addressed in detail here. With the support of HyperWorks®,
1 T D( )
the CAD models were imported to the finite element analysis (FEA), in = U BT Bd U
which each element was recorded, and the global coordinate system 2 i
N
was determined. Boundary conditions and loads were applied to the 1 T Di ( )
= ui BT Bd ui
structure. Then the FE models were analyzed in ANSYS® solver to cal-
i
i=1
2 i i (11)
culate the displacement and stress, obtaining the strain energy of each
discrete element. Finally, the printing angle, strain energy, and pseudo Where ui is the displacement vector of the i-th element.
density were introduced into Boss-Quatro® to generate the new pseudo Combining with Eq. (3), the derivative of Di with respect to the
density and BD by utilizing GCMMA. rotating angle can be expressed analytically as
The global stiffness matrix K was assembled by the element stiffness T
Di ( ) R( ) R( )
matrix function Ki. Combining with Eq. (3), it can be written as follows: = Dm RT ( ) + Dm RT ( )
i i i (12)
n
K= BTDBd = BTRDm RTBd = BTRDm RTBd i GCMMA is an appropriate algorithm to search the optimal BD in
i
i (5) that it is proven to be efficient and reliable [48]. For an optimized
model with fewer variables, two types of algorithms are mainly used.
Where B represents the strain matrix, Ω represents the global design
One is the gradient-based algorithm that demonstrates a fast con-
domain, n is the number of elements in the whole design domain, and
vergence rate, improving the computing efficiency while prone to ob-
Ωi denotes the design domain of the i-th element. The optimization
tain local minima, such as gradient descent, Newton's method, con-
model with BD was expressed mathematically as follows:
jugate gradient, and Lagrange multiplier. Another one is the heuristic
find: = { , }T algorithm with the advantages of being insensitive to initial values and
1 1 outstanding global searchability, such as the genetic algorithm and
min: E = F TU = U TKU particle swarm algorithm. These advantages are clarified based on the
2 2
subject to: KU = F time consuming of the FEA process. Future research could focus on
combining two kinds of algorithms to achieve better results with ac-
90 , 90 (6)
ceptable efficiency.
Where α and β are Euler angles controlling the rotation of parts. Fur-
ther, E is the strain energy, and the optimal objective is to minimize the 2.1.3. TO considering BD
global strain energy. F, K, and U are represented as the external force, Compared with Section 2.1.2, the structural topology and BD were
the global displacement vectors and the global stiffness matrix, re- optimized simultaneously in this section. It aims at coupling process-
spectively. dependent features and optimization algorithm of design to predict the
The sensitivity analysis is conducted by calculating the derivative of mechanical performances of orientation-dependent TO designs.
the strain energy, which can be expressed as SIMP (Solid Isotropic Material with Penalization) was generally
adopted for isotropic solid material to achieve clear topologies [49] and
E U 1 T K PAMP (Porous Anisotropic Material with Penalization) was employed
= U TK + U U = { , }T
i i 2 i (7) in structural TO with microstructures in the macro-scale [50]. Simi-
larly, the SLA-printed components could be optimized using SAMP
For the governing equation, the differentiation to the rotating angle
(Solid Anisotropic Material with Penalization).
θi can be written as
Design variables were defined as follows:
K U F
U+K = X = { 1, , n ; , }T (13)
i i i (8)
Where ρi denotes the pseudo-density value of the i-th element.
Since the external force F is independent of the rotating angle, the The elastic matrix of the i-th element can be written as follows:
derivative of F to θi is 0. Eq. (9) can be obtained.
p
Di ( , ) = i RDm RT (14)
U K
U TK = UT U Where p = 4 is the penalty factor.
i i (9)
The global stiffness matrix is similar to Eq. (14), which can be
Substituting Eq. (9) into Eq. (7) yields written as

4
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Table 1 to the local coordinate system. Firstly rotating -β around axis 2, and
Physical properties of the resin. then rotating -α around axis 1 in the local coordinate system, yields the
Resin parameters Magnitude final posture.
The sensitivities of material volume V with respect to the pseudo-
Critical exposure 11.5 mJ/cm2 density variable ρi and aspect directional variable θi are expressed as
Exposure 54 mJ/cm2
Penetration depth 0.16 mm V
Density 1.12 g/cm3 (25 ℃)
= vi
i (21)
Viscosity of resin 260 cps (30 ℃)
V
=0
n n i (22)
K= Ki = BTDi ( , ) Bd
i
i
(15) Throughout the program, GCMMA algorithm was applied to update
i i
the design variables and a sensitivity filtering technique was utilized to
TO considering BD can be expressed mathematically as follows: handle numerical problems such as checkerboard patterns and mesh
find: X = { 1, , n ; , }T dependency [51]. The mathematical formulations for the two types of
optimization models were established with sensitivity analysis.
1 1
min: E = F TU = U TKU
2 2
subject to: KU = F 2.2. Experimental design
V V
0 < min 1, i = 1, ,n Table 1 outlines the physical properties of the commercially avail-
i
able resin SOMOS11120 used in this study. The geometrical dimensions
-90 , 90 (16)
of tensile and compressive specimens were manufactured following the
Where V is the material volume of the whole design domain with an ISO 3167−2014 and ISO 604, respectively. The tensile specimen is il-
upper limit. ρmin is a small positive value as a lower bound of pseudo- lustrated in Fig. 3 and the compressive specimen is a cube with a di-
density, e.g., 0.001, which can avoid singularity of the global stiffness mension of 50 mm × 10 mm × 4 mm. Table 2 gives the parameters of
matrix during iteration. The rotating angles α and β distributes between the printing process. The post-curing machine equipped with the sur-
-90° and +90°. face light source and rotating platform were adopted to cure printed
Sensitivity analysis is similar to the method used in Section 2.1.2. components in an identical procedure by UV light with a power of 350
Derivatives of strain energy E are obtained in Eq. (17). The derivation is W for 2000s to ensure the uniformity of the curing degree. These parts
analogous to Eqs. (12)–(14). were printed under identical parameters except for BD. The tensile and
compressive rates were 2 mm/min.
E 1 T K
= U U As shown in Fig. 4, the tensile and compressive specimens rotate
i 2 i (17) around axis x from 0° to 90°, with the difference of 15° between ad-
Considering Eqs. (14) and (15), the sensitivity of global strain en- jacent groups. As a result, tensile and compressive experiments include
ergy for the pseudo-density variables ρi can be written as follows: seven groups, which are labeled with T and C, respectively. For ex-
ample, the tensile parts of 0° are T0°, and the corresponding compres-
N
E 1 Ki sive parts are C0°. Each batch contained six specimens.
= u iT ui
i 2 i=1 i
A VHX-6000 digital microscope was used to observe the surfaces of
BT p
Di ( ) Bd the printed components. The tensile and compressive specimens were
1 T i i
conducted by TestResources™ equipped with a 100 kN load cell to ac-
= ui ui
2 i quire the load-displacement curves. PMLAB DIC-3D system was utilized
to capture the horizontal and vertical strains of each specimen for
p
1 T i BT Di ( ) Bd i
i
= ui ui further calculating the elastic modulus and Poisson's ratio. The specific
2 i
p experimental results are shown in Section 3.1.
= Ei
i (18)
3. Result and discussion
Where Ei is the strain energy of the i-th element. The derivative of ro-
tating angles is obtained in Eqs. (19) and (20). 3.1. Tensile and compressive properties of the printed specimens
E 1 T K
= U U The surface undulation of printed specimens built in the BDs of 0°,
i 2 i
N 45°, and 90° are illustrated in Fig. 5. Three types of surfaces are ob-
1 T Ki served including top surface, side surface, and front surface, marked at
= ui ui
i=1
2 i 90°, 45°, and 0°, respectively. The top surface of 90° and the side surface
N
1 T BTDi ( , ) Bd of 0° have a similar surface roughness since they are both perpendicular
= ui ui to the BD. The top surface of 90° exhibits a ‘dot-by-dot’ undulating
i=1
2 i
N
1 T p Di ( )
= ui BT i Bd i ui
i=1
2 i i (19)
T
Di ( , ) T( ) T( )
= Dm T T ( ) + Dm T T ( )
i i i (20)

The derivative of Di (ρ, θ) is the sum of two matrices, which can be


deduced by the optimized α and β, and α and β represent the rotating
angles of the local coordinate system as well. For unmanufactured
structures on the worktable, the rotation is the inverse process relative Fig. 3. Geometrical dimensions of the tensile specimen.

5
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Table 2 layer planes and the crack path is affected by the inhomogeneous
Process parameters of SLA. bonding strength, exhibiting slight fluxion. Shear failure is observed in
Process parameters Magnitude T45°.
Figs. 6 and 7 plot the load-displacement curves of tensile and
Laser power 220 mW compressive specimens. One curve is selected to represent its corre-
Radius of the laser beam 0.12 mm
sponding batch. As listed in Table 4, elastic modulus and Poisson’s ratio
Laser beam wavelength 355 nm
Layer thickness 0.1 mm
are deduced from the tensile curves and DIC analysis. Poisson's ratio of
Scanning strategy XYSTA each specimen is around 0.4. The elastic modulus of T0° and T90° are
Scanning interval 0.1 mm approximately equal, and Poisson's ratio of T0° and T90° are both 0.37.
Scanning speed 6500 mm/s It proves the applicability of the transversely isotropic model for the
Time of UV post-curing 2000 s
printed material. The specimens in T0° and T90° possess similar elastic
modulus and Poisson's ratio, while T45° possess lower elastic modulus
pattern due to the gridded scanning path. The maximum height of the and higher Poisson's ratio than these in T0° and T90°. The specimens of
profiles (Rz) is 15.2 μm. The front surface of 0° and the side surface of T45° exhibited 6.6 % lower average elastic modulus compared to the
90° show clear layered features where Rz is 18.49 μm. The side surface specimens of T0°, which can be explained by the difference of shear
of 45° possesses a zigzag-shaped undulation. A sawtooth curve is dis- moduli of materials between the inner-layer and inter-layer space.
covered in the edges along the axis x, where Rz is 27.88 μm. For the side Furthermore, the specimens of T0° exhibited 23.8 % larger ultimate
surfaces of 0° and 45°, the changed angle between the surface normal tensile stress compared to the specimens of T90°.
direction and BD makes ‘dot-by-dot’ transition gradually to ‘line-by- The specimens in groups T0°–T45° exhibit similar strain at break,
line’ undulation, which is similar to observations by P. Delfs and M. however, less plastic deformation is observed in specimens of
Töws [52] for layer-by-layer AM processes. Surface quality that varies T60°–T90°. Strain at break is below 5% in T60°–T90° and nominal strain
with BD causes irregular undulations at different locations of the at break is above 10 % in T0°–T45°, which is at least a factor of 2 in
printed complex structures, especially for the lattice with tiny rods. It is T60°–T90°. As the strong ductility of inner-layer materials is observed,
possible to predict that the printed tensile specimens exhibit different specimens of T0°–T45° possess significant plastic mechanical behaviors
crack propagation modes due to irregular stress concentration. due to the inner layer being pulled. Specimens of T60°–T90° are mainly
After the off-axis experiments, the fracture shape and surface of pulled perpendicular to the inter layer, where the bonding strength is
specimens in T0°, T45°, and T90° were examined under the microscope. weak. Therefore, specimens of T0° and T90° exhibits the maximum and
Table 3 lists the macroscopic images. Vertical and horizontal building minimum ultimate tensile stress and strain, respectively.
lines are visible in specimens T0° and T90°, respectively. As shown in Table 5 lists the compressive strength. The compressive strength of
Fig. 6, for T0° and T45°, the break occurs after yielding whereas for each specimen is expressed as a negative value to distinguish tensile
T90° it occurs prior to yielding. strength from compressive strength. It can be concluded that the dif-
The fracture surface of T0° is uneven due to the plastic fracture with ference of the maximum and minimum compressive strength is 27.1 %
the nominal strain at break of 17 % ± 1.7 % and Rz of 179.82 μm. The and both the tensile and compressive strength are orientation-depen-
fracture surface of T90° is quite smooth compared to the one with the dent with a maximum gap of 20 % between 0° and 90°.
strain at break of 1 % ± 0.1 % and Rz of 10.27 μm. Moreover, T0° Fig. 8 illustrates that the tensile strength decreases with the increase
possesses a large plastic deformation and rough failure surface, and of building angle while the compressive strength shows the opposite
T90° exhibits a small strain at break and smooth failure surface, cor- trend. Since the compressive strength is negative, both trends represent
responding to ductile fracture and brittle fracture, respectively. T45° that the tensile and compressive capability of the printed material are
exhibits an inclined section, and the fracturing angle is 45°. In addition, declined owing that the inner-layer material possesses higher tensile
delamination is observed clearly in the rough fracture section. The and compressive capabilities than the inter-layer material. The inner-
layered feature and surface roughness dominate the fracture mode. In layer region plays the role of ‘skeleton’, enhancing mechanical strength.
the inter layer, the bonding strength is weak. The initiation point of The longitudinal tensile and compressive direction begins with a hor-
fracture exists at the defects inside the specimens and zigzag-shaped izontal direction, then transforming gradually to vertical. Therefore,
surface depressions of which ordinarily lead to the stress concentration. T0° and C0° exhibit the maximum tensile and compressive strength. For
The uniform in-plane fracture and smooth fracture surface of T90° are T90° and C90°, the inner-layer and inter-layer region bear the uni-
able to be explained by the crack propagation in the inter-layer plane. directional load respectively, thus the loading capacity is limited by the
However, the crack in T0° propagates through inner-layer and inter- inter-layer inferior material. The strengths of other groups within the 0°

Fig. 4. Specimens printed in different BDs.

6
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 5. Surface undulation of printed specimens.

and 90°, depend on the angle between the principal stress direction and
BD.
In this section, an empirical study is performed. The specimens were
printed in different BDs, and the surface roughness and layer features of
specimens were observed under a microscope after off-axis experi-
ments. The tensile specimens exhibit different failure modes and surface
morphology of fracture as the irregular surface undulation affects the
stress concentration distribution and crack propagation. The layered
feature and surface roughness dominate the stiffness and strength of the
SLA-printed components, inducing the transverse isotropy. A function is
established to describe the orientation-dependent properties for the
AM-driven TO.

3.2. Elastic matrix of the SLA-printed components

Fig. 6. Load-displacement curves of tensile specimens.


Based on the layer-by-layer forming mechanism and mechanical
properties of the SLA-printed specimens, the transverse isotropy is in-
vestigated to evaluate the SLA-printed components. The elastic matrix in Table 4, therefore, the shear modulus can be deduced from Eq. (23)
in the local coordinate system can be obtained according to Eq. (1) and [47] and Eq. (24). G12 and G13 are 686.3 MPa and 644.3 MPa, re-
Table 4. In addition, ν12 is 0.41. E1 (E0), E45, E3 (E90), and ν13 are listed spectively. Matrices Cm and Dm are obtained as below.

Table 3
Fracture characteristics of tensile parts.
Specimens Fracture shape Fracture surface

T0°

T45°

T90°

7
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 8. Tensile and compressive strength of each batch.

Fig. 7. Load-displacement curves of compressive specimens.

Table 4
Experimental data of tensile specimens.
Specimens Elastic Poisson’s Ultimate tensile (Nominal) Strain
modulus ± SD ratio ± SD stress ± SD at break ± SD
(E, MPa) (MPa)

T0° 1935.19 ± 21 0.37 ± 0.019 46.72 ± 0.02 17 % ± 1.7 %


T15° 1907.6 ± 23 0.40 ± 0.024 44.77 ± 0.55 12 % ± 0.5 %
T30° 1885.82 ± 24 0.41 ± 0.024 44.2 ± 0.62 11 % ± 1.4 %
T45° 1814.71 ± 43 0.41 ± 0.025 41.42 ± 1.37 14 % ± 0.8 %
T60° 1870.94 ± 10 0.41 ± 0.006 40.63 ± 0.25 4 % ± 1.8 %
T75° 1826.04 ± 38 0.38 ± 0.017 39.45 ± 0.83 3 % ± 0.3 %
T90° 1925.13 ± 39 0.37 ± 0.02 37.75 ± 1.45 1 % ± 0.1 %

Fig. 9. The predictive and experimental elastic modulus.


Table 5
Experimental data of compressive specimens.
Specimens Compressive strength (MPa) SD (MPa) elastic modulus at any printing angle can be calculated.
C0° −50.06 2.43
Fig. 9 indicates that the predictive elastic modulus is consistent with
C15° −49.25 2.46 experimental values basically with the maximum difference of 6.6 %
C30° −45.37 1.05 between T0° and T45°. The predictive curve is similar to a secant or
C45° −43.39 2.19 cosecant curve and it can be deduced from Eq. (25). The experimental
C60° −41.36 2.85
and predicted values exhibit deviations due to environmental variations
C75° −38.31 1.69
C90° −36.49 0.11 in temperature and humidity. However, as shown in Fig. 9, T75° is quite
different from the theoretical prediction, unlike the other cases. The air
bubbles in the resin and the force of the scraper cause uneven curing
G13 = G23 =
1 and defects during printing [19,25], resulting in the instability of the
4
E45
1
E1
1
E3
+
2 13
E1 (23) inter-layer materials. The theoretical prediction is based on the ex-
perimental results of T0°, T45°, and T90°. T0° has the lowest random
G12 =
E1 error while T90° possesses the highest random error. Therefore, when
2(1 + 12 ) (24) the printing angles are close to 0° and 90°, the experimental and pre-
dicted values exhibit better agreement and greater difference respec-
5.17 2.12 1.92 0 0 0 tively. As a result, the prediction of T15° is consistent with the ex-
2.12 5.17 1.92 0 0 0
1.92 1.92 5.19 0 0 0 periment while T75° possesses a large deviation between the
Cm = 10 4 ×
0 0 0 15.52 0 0 experiment and prediction. The 3.6 % error exists between numerical
0 0 0 0 15.52 0 prediction (1895.5 MPa) and experimental results (1826.1 MPa) is
0 0 0 0 0 14.57 tolerable in practical engineering design.
It is difficult to quantify the influence of process parameters on the
3754.4 2382.7 2270.4 0 0 0 mechanical properties of printed materials using theoretical prediction.
2382.7 3754.4 2270.4 0 0 0
2270.4 2270.4 3606.6 0 0 0 Therefore, an experimental data-driven investigation was conducted to
Dm = obtain elastic matrix and strength values. Anisotropic mechanical be-
0 0 0 644.3 0 0
0 0 0 0 644.3 0 haviors were revealed through the characterization of microstructures
0 0 0 0 0 686.3 and tensile experiments. The off-axis experiments demonstrate the
transverse isotropy of the SLA-printed material. The material evaluation
According to composite mechanics [47], the predictive formula of
illustrates that the difference in elastic properties reaches 6.6 %, while
biaxial modulus is
the strength differs by approximately 23.8 %.
1 After the determination of the parameters of the elastic matrix, the
E =
1
E1
cos4 + ( 1
G13
2 13
E1 ) sin 2 cos2 +
1
E3
sin4
(25)
strain energy of the printed components under certain loads can be
accurately predicted. The BD affects the direction of anisotropy and the
Where θ is the printing angle of the tensile bars. Hence, the predicted mechanical properties of the printed structures. Therefore, the

8
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 10. (a) Boundary condition and external force (b)The MBB beam.

transversely isotropic elastic matrix and directional parameters can be


further applied in TO design to utilize the advantages and avoid the
disadvantages of anisotropy induced by the AM process, improving the
mechanical performance of case-dependent objects.
Fig. 12. The strain energy of all BDs.
3.3. Case-dependent optimization
Table 6
For the special functional requirements, the end-use products (such
Strain energy of three typical BDs.
as complex lattices) need to ensure their configurations but the struc-
tural stiffness and strength can be improved by BD optimization. In Case Strain energy(mJ) α(°) β(°)

addition, process-dependent anisotropy needs to be taken into account Emin 61.53 0 0


in the design stage. In this section, two typical optimizations are im- Emiddle 63.59 0 90
plemented to satisfy the engineering requirements. BDs are achieved Emax 65.06 0 51.21
based on a 2D MBB beam and a 3D L-shape plate with sandwich pyr-
amid lattice in Section 3.3.1 and 3.3.2. In Section 3.3.3, structural to-
pology and BD are optimized simultaneously on a 3D L-shaped beam. Table 6. The Emin and the Emax present the maximum and the minimum
stiffness respectively, and Emiddle represents the BD with minimum re-
3.3.1. MBB beam sidual stress. In the circumstances of Emin with the printing angle of α =
As shown in Fig. 10, the MBB beam with the dimension of 200 mm 0° and β = 0°, the long rod of the MBB beam is printed horizontally, and
× 16 mm × 50 mm is preset at the global coordinate system. The the residual stress in the length direction is relatively large. Corre-
whole design domain was discretized to 46,752 elements. An external spondingly, when the condition changes into Emiddle with the printing
force, of 306 N, was applied at the center of the top surface. The BD is angle of α = 0° and β = 90°, the residual stress is small on account of
determined by α and β, which are the two design variables in optimi- avoiding a large gap in length and width. In the practical production
zation formulation. process, the MBB beam is printed in α = 0° and β = 90° to ensure the
Fig. 11 gives convergence curves of strain energy and building an- success of printing. In addition, it makes parts on the worktable as
gles. The BD is optimized with three different initial values, which many as possible to improve the printing economy. This case studies the
eventually converge to the same strain energy. When α and β are 0°, BD influence of BD on the structural stiffness and guides adjusting the BD
is the positive direction of the axis z, and the MBB beam possesses the of parts in practical production. The difference between Emin and Emax
maximum stiffness. The optimal solution is achieved before the 20th reaches 6%.
iteration, which displays a fast convergence rate using GCMMA. The structures with the largest difference in strain energy were
For the two design variables, both design intervals were traversed to obtained and validated to highlight the influence of BD. The specimens
observe the change of structural strain energy. Fig. 12 shows the surface were measured using TestResources™ with a 100 kN load cell under the
of the strain energy in all combinations of α and β with three special specific boundary condition. The loading rate of 2 mm/min was ap-
BDs. A total of 900 points were calculated to map the surface, which plied. As shown in Fig. 13(a), the MBB beams are printed in a specific
possesses two symmetrical planes due to the symmetry of the MBB direction in the worktable according to the optimal results and then
beam and the unidirectional pressure. Three typical BDs are listed in verified in Fig. 13(b). The rotation around axis z in the global

Fig. 11. Convergence curves:(a) Strain energy with different initial values (b) α and β.

9
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 13. (a) Printed parts in worktable (b) Static boundary condition.

Fig. 15. The L-shaped plates with sandwich pyramid lattice.

be observed that the optimal solution is achieved before the 20th


iteration in both numerical examples, revealing that the GCMMA ex-
hibits a fast optimal speed. Furthermore, the optimization improves
efficiency by a factor of 45 compared to the traversal method with 900
Fig. 14. Load-displacement curves of Emin and Emax. FE calculations. When α is −52° and β is −32°, the L-shaped plate
exhibits the maximum stiffness.
Fig. 17 illustrates the surface of the strain energy for the L-shaped
coordinate system does not affect the mechanical properties, allowing
plates. The surface is symmetrical about the center of the origin since
structures to be adjusted in the workspace.
the force is vertically downward. Five special BDs, marked as Case A −
Fig. 14 shows the load-displacement curves of two printed struc-
E, and their computed results are listed in Table 7. Case A is the initial
tures. The stiffness and strength of the MBB beams were evaluated by
solution. The Emin is Case D and the Emax is Case E. The difference be-
the slope and maximum load of the linear region, which differed by
tween Emin and Emax is 3.5 %. Cases C and B are the local maximum and
12.8 % and 20.3 %, respectively. It can be concluded that BD greatly
local minimum, respectively.
influences structural mechanical performance by evaluating the effec-
Two typical numerical examples were used to optimize the BD. The
tiveness of the optimization model based on strain energy. For the 2D
MBB beam is a 2D structure under an in-plane load. The printed parts
structures bearing in-plane force, such as the MBB beam, ensuring the
exhibit high performance compared with other directions when the
structure coincides with the 1–2 plane in the local coordinate system is
main plane of the structure is parallel to the working platform. It can be
able to improve the structural mechanical performance compared with
deduced that 2D structures with in-plane force can be placed flat on the
other BDs.
platform to achieve uniform mechanical performance. Moreover, the
fewest support structures are required and the printing efficiency is
high. In comparison, optimization for the 3D L-shaped plate achieves a
3.3.2. L-shaped plates with sandwich pyramid lattice
special printing angle that cannot be predicted by empirical decision-
The previous optimal solution achieves a special BD, and it re-
making.
presents that the MBB beam is placed ‘flat’ on the platform. However,
In the current engineering application of the layered AM, the de-
this optimal BD can be obtained by empirical decision-making due to
termination of printing orientation usually ensures the successful con-
the special structure and in-plane load. Additionally, to prove the ef-
struction of components. Residual stress and surface roughness are also
fectiveness of the optimal model, a 3D L-shaped plate with sandwich
significant factors. The building time and cost are significantly affected
pyramid lattice was selected since lattice structures possess highly
by the orientation of the parts. The support structure is also important
complex geometrical configurations. In Fig. 15, the outside length of
because it influences the energy costs, post-processing, and finishing
the L-shaped plate is 115 mm, and the inside length is 95 mm. The unit
steps. Operators need balance trade-offs between surface quality,
cell of the sandwich lattice is a pyramid with cylindrical trusses with a
building time, cost, and support structures in fabricating products
radius of 1 mm and the thickness of one plate is 2 mm. Two lattices are
[53,54]. Beyond these factors, this section provides a new perspective
placed in the width direction of the L-plate where two ends are solid
for designers to determine BD through material evaluation and opti-
with a length of 15 mm. The boundary condition and force are applied
mization methods.
to the solid ends. The force is 520 N, and it is perpendicular to the main
plane of the structure. The whole design domain was discretized to
545,232 elements. 3.3.3. L-shaped beam
The BD was optimized when initial values α and β are 60°. Fig. 16 The structural configuration and BD are optimized to consider the
plots the convergence curves of strain energy and printing angles. It can directional parameters in the AM-driven TO. This optimization model

10
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Fig. 16. Convergence curves (a) Strain energy with specific initial values (b) α and β.

Fig. 17. Strain energies of all BDs.

Table 7 Fig. 18. L-shaped beam with force and boundary conditions.
Five special cases.
Case Strain energy (J) α (°) β (°)
to ensure consistency of surface quality and volume of the seven L-
A 87.18 0 0
B 86.56 −90 −50.2 shaped beams.
C 88.12 −90 9.4 Table 8 summarizes the BD optimization results of six cases and the
D 85.98 −52 −32 concurrent optimization results are demonstrated in Fig. 19. The
E 88.95 −90 −83.4 maximum and minimum volumes of the seven structures are 48,193
mm3 and 47,746 mm3 respectively, which means the volume difference
is less than 1%. The concurrent optimized structure possesses the
aims to utilize the advantage of anisotropy in the product design stage.
minimum strain energy while Case D has the maximum value. The
The geometric dimension, force, and boundary conditions are
maximum Von Mises Stress of each case is recorded and it is located at
shown in Fig. 18. The unit of distance is millimeter, and the thickness
zone 2 due to the stress concentration. Case D possesses higher max-
along the axis y is 15 mm. The 3D model was meshed into 137,090
imum stress and lower stiffness than other cases since the change in BD
elements in Altair HyperWorks®, containing 128,090 elements in the
influences structural configuration and objective stiffness and stress. As
design domain. The non-design domain of the lower-left corner is fixed
shown in Fig. 19, zone 1 and 2 of the concurrent optimized structure
as the boundary condition and the L-shaped beam is loaded by a ver-
possess different patterns compared to Case A to F. For Case A and
tical surface force F = 490 N. The volume fraction is less than 25 % and
concurrent optimized structure, zone 1 and 2 are symmetric about x-z
then the FE model is analyzed in ANSYS®. The sensitivity information is
plane while they are asymmetric for Case B to F owing that the material
input to Boss-Quattro® [50]. Sensitivity filtering was used to avoid
property is asymmetric about the x-z plane. In addition, zone 1 and 2
mesh dependency and checkboard patterning, and the filter radius is
determine stiffness and failure mode for each model. For the concurrent
two times the element size.
optimized structures, the material aggregates in the middle of the beam
Structural TO was performed in a range of different BDs due to the
and the evolution of directional variables prefer to place the main
effects of BD on the stiffness of structures. The FEA models possess
bearing rod in the inner-layer plane, eliminating the tiny branch and
zigzag surfaces that cannot be printed directly. The optimized struc-
centralizing the side rods. For traditional optimization problems, the
tures were smoothed to 3D models that can be directly fabricated.
direction of anisotropy should be carefully pre-selected before TO.
Seven STL models are provided in the supplementary materials
Fig. 20 plots the convergence curves of the strain energy. As shown
STL_MODEL. The identical parameters are applied for the post-process
in the figure, the strain energy declines quickly before iteration 30 and

11
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Table 8
Optimization results of six cases.
Case α, β Structural topologies Strain Maximum Von Volume
energy Mises Stress (mm3)
(mJ) (MPa)

A 0,0 2691.6 52.96 48193

B 15,0 2669.2 52.61 48006

Fig. 19. Concurrent optimization results.

C 30,0 2710.6 52.16 48021

D 45,0 2731.8 58.20 47844

Fig. 20. Convergence curves of strain energy.

E 60,0 2689.4 52.09 47981 converges to the final value after iteration 60. The strain energy of
concurrent optimization is 2562.5 mJ, which is almost 35.9 % below
the initial design. In addition, it converges the minimum objective value
compared to Case A to F.
Fig. 21 illustrates the strain energy of Cases A–F and concurrent
optimization. Compared with Cases A–F, the strain energy obtained by
TO considering BD is small. The evidence indicates that the introduc-
tion of BD changes the structural topology and the direction of aniso-
tropy, fully exploiting material anisotropy and matching the design and
manufacturing stages. For empirical decision-making, the initial angle
F 75,0 2639.3 55.60 47904 may result in local minima when the mapping surface possesses several
peaks and troughs. The potential solution is arranging initial BD evenly.
In addition, it is important to investigate the consistency of theoretical
prediction and experimental validation.
In Fig. 22(a), The two models are printed with identical parameters
and post-processing as outlined in Section 2.2. The specimens were
validated under the specific boundary condition using TestResources™
equipped with a 5 kN load cell with a loading rate of 2 mm/min. Failure
modes, load, and displacement were recorded and plotted, as shown in
Fig. 22(b). Stiffness and strength can be deduced from the load-dis-
placement curves. The concurrent optimization result possesses better
mechanical properties than Case A. The stiffness increases by 8.9 %

12
S. Li, et al. Additive Manufacturing 36 (2020) 101406

stiffness and strength. The AM-driven TO method addresses both


structural design references and appropriate process parameters, brid-
ging the information gap between design and manufacturing, and
providing potential guidance for the investigation of process-property-
structure relationships. Transverse isotropy and a coordinate transfor-
mative matrix are inspired by the use of long fiber-reinforced compo-
sites. The material evaluation is proven effective through the char-
acterization of microscopic observation and macroscopic experiments.
TO formulation and rigorous sensitivity analysis were established to
design products with the help of GCMMA. The integrated optimization
method includes the design and manufacturing stages, capitalizing on
the advantages and avoiding the disadvantages of anisotropy induced
by the AM process.
The case-dependent optimization explores the influence of BD on
mechanical properties, integrating the function of orientation-depen-
Fig. 21. Strain energy of Cases A–F and concurrent optimization. dent material properties into the TO algorithm. Other process para-
meters, such as laser power and scanning speed can also be quantified
according to the slope and maximum load of the linear region. In ad- and applied in the TO model through this method. As an extension,
dition, the maximum load increases by 14.3 % due to the change in strength-based TO can also be developed through experimental data-
failure modes. The failure mode is fracture in case A, in which the left fitting approach. In addition, other design constraints, such as building
branch preliminary fractured because of the reducing strength. It is time, and warpage, can also be incorporated to establish correlations
similar to the specimens in T90°. Then the intersection of branches is and introduced into the TO model.
broken due to stress concentration. For the L-shaped beam in TO con-
sidering BD, the inner layer of the left branch acts as a ‘fiber’, which
enhances stiffness and strength. Thus, concurrent optimization exhibits 4. Conclusion
high mechanical performance and demonstrates the validity of the
proposed method. Moreover, Table 8 and Fig. 19 show the maximum In this study, an AM-driven TO method is proposed to quantify the
Von Mises Stress of Case A and concurrent optimization. They are at the correlative function between directional parameters and transverse
same level of stress concentration according to FEA. Nevertheless, isotropy based on SLA-printed materials. Case-dependent optimization
concurrent optimization Case exhibits an improvement of 14.3 % utilizes a gradient-based algorithm and SAMP to find the appropriate
compared with Case A. It implies that it is necessary to establish specific BD and structural configuration. The realization of this method includes
failure criteria for different AM printed materials. the evaluation of transversely isotropic properties through character-
In this section, structural topology and BD are optimized to promote ization of microscopic observation and macroscopic experiments, and
mechanical properties. For the BD optimization model, the results concurrent optimization of BD and structural topology. The material
clearly improve mechanical performance of the 2D MBB beams and 3D evaluation demonstrates that the difference in elastic properties varies
L-shaped plates with sandwich pyramid lattice. Thus, simply adjusting by 6.6 %, and the strength varies by approximately 23.8 %. In addition,
the BD can achieve high structural performance. In the optimization the correlative function of building orientation and mechanical prop-
model that considers BD and structural configuration simultaneously, erties of printed materials (stiffness and strength) is further applied in
the optimized products possess advanced stiffness and strength com- the optimization model to obtain optimized products with desirable
pared with randomly oriented products. In case-dependent products, performance. Experimental verification proves that the optimal design
the maximum loading capacity of the MBB beam and L-shaped beam proposed by the AM-driven TO approach improves 8.9 % in stiffness
are improved by 20 % and 14 %, respectively, due to the consistency of and 14.3 % in strength as a result of implementing an orientation-de-
stiffness and strength. Both calculations and experiments indicate that pendent function with directional variables. Future research may focus
introducing directional variables into optimization models enhances on extending the design variables, design constraints, and objective
the structural mechanical performance. functions of the TO algorithm, combined with the actual working state
Experimental validation demonstrates substantial improvements in in AM.

Fig. 22. (a). The printed structures (b). Load-displacement curves with failure modes.

13
S. Li, et al. Additive Manufacturing 36 (2020) 101406

Declaration of Competing Interest [23] J.W. Stansbury, M.J. Idacavage, 3D printing with polymers: challenges among ex-
panding options and opportunities, Dent. Mater. 32 (2016) 54–64, https://doi.org/
10.1016/j.dental.2015.09.018.
The authors declare that they have no known competing financial [24] W.M. Wang, C. Zanni, L. Kobbelt, Improved surface quality in 3D printing by op-
timizing the printing direction, Comput. Graph. Forum, Wiley Online Library, 2016,
interests or personal relationships that could have appeared to influ- pp. 59–70.
ence the work reported in this paper. [25] S. Guessasma, L. Tao, S. Belhabib, J. Zhu, W. Zhang, H. Nouri, Analysis of micro-
structure and mechanical performance of polymeric cellular structures designed
using stereolithography, Eur. Polym. J. 98 (2018) 72–82.
Acknowledgements [26] W. Gao, Y. Zhang, D. Ramanujan, K. Ramani, Y. Chen, C.B. Williams, C.C.L. Wang,
Y.C. Shin, S. Zhang, P.D. Zavattieri, The status, challenges, and future of additive
This work is supported by the National Key Research and manufacturing in engineering, Comput. Des. 69 (2015) 65–89.
[27] D.L. Naik, R. Kiran, On anisotropy, strain rate and size effects in vat photo-
Development Program (2017YFB1102800), NSFC for Excellent Young polymerization based specimens, Addit. Manuf. 23 (2018) 181–196, https://doi.
Scholars (11722219), and Key Project of NSFC (51790171, org/10.1016/j.addma.2018.08.021.
[28] P. Zhang, J. Liu, A.C. To, Role of anisotropic properties on topology optimization of
5171101743). additive manufactured load bearing structures, Scr. Mater. 135 (2017) 148–152,
https://doi.org/10.1016/j.scriptamat.2016.10.021.
Appendix A. Supplementary data [29] S.-H. Ahn, M. Montero, D. Odell, S. Roundy, P.K. Wright, Anisotropic material
properties of fused deposition modeling ABS, Rapid Prototyp. J. 8 (2002) 248–257.
[30] C.S. Lee, S.G. Kim, H.J. Kim, S.-H. Ahn, Measurement of anisotropic compressive
Supplementary material related to this article can be found, in the strength of rapid prototyping parts, J. Mater. Process. Technol. 187 (2007)
627–630.
online version, at doi:https://doi.org/10.1016/j.addma.2020.101406.
[31] S. Guessasma, S. Belhabib, H. Nouri, Significance of pore percolation to drive ani-
sotropic effects of 3D printed polymers revealed with X-ray μ-tomography and finite
References element computation, Polymer (Guildf). 81 (2015) 29–36.
[32] D. Feldman, A. Barbalata, Synthetic Polymers: Technology, Properties,
Applications, Springer Science & Business Media, 1996.
[1] S.A.M. Tofail, E.P. Koumoulos, A. Bandyopadhyay, S. Bose, L. O’Donoghue, [33] Y. Yang, L. Li, J. Zhao, Mechanical property modeling of photosensitive liquid resin
C. Charitidis, Additive manufacturing: scientific and technological challenges, in stereolithography additive manufacturing: bridging degree of cure with tensile
market uptake and opportunities, Mater. Today 21 (2018) 22–37. strength and hardness, Mater. Des. 162 (2019) 418–428, https://doi.org/10.1016/j.
[2] C. Wang, J.H. Zhu, W.H. Zhang, S.Y. Li, J. Kong, Concurrent topology optimization matdes.2018.12.009.
design of structures and non-uniform parameterized lattice microstructures, Struct. [34] J.Y.H. Fuh, L. Lu, C.C. Tan, Z.X. Shen, S. Chew, Processing and characterising
Multidiscipl. Optim. 58 (2018) 35–50. photo-sensitive polymer in the rapid prototyping process, J. Mater. Process.
[3] M. Tomlin, J. Meyer, Topology optimization of an additive layer manufactured Technol. 89–90 (1999) 211–217, https://doi.org/10.1016/S0924-0136(99)
(ALM) aerospace part, Proceeding 7th Altair CAE Technol. Conf. (2011) 1–9. 00073-4.
[4] T. Abballe, M. Albertelli, G. Allaire, A. Caron, P. Conraux, L. Dall’Olio, C. Dapogny, [35] Y. Xu, J. Zhu, Z. Wu, Y. Cao, Y. Zhao, W. Zhang, A review on the design of lami-
C. Dobrzynski, B. Jeannin, F. Jouve, RODIN project, Topology Optimization 2.0? nated composite structures: constant and variable stiffness design and topology
(2015). optimization, Adv. Compos. Hybrid Mater. 1 (2018) 460–477.
[5] M.P. Bendsøe, O. Sigmund, Material interpolation schemes in topology optimiza- [36] K.K. Yang, J.H. Zhu, C. Wang, D.S. Jia, L.L. Song, W.H. Zhang, Experimental vali-
tion, Arch. Appl. Mech. 69 (1999) 635–654. dation of 3D printed material behaviors and their influence on the structural to-
[6] X. Huang, M. Xie, Evolutionary Topology Optimization of Continuum Structures: pology design, Comput. Mech. 61 (2018) 581–598, https://doi.org/10.1007/
Methods and Applications, John Wiley & Sons, 2010. s00466-018-1537-1.
[7] M.P. Bendsøe, N. Kikuchi, Generating optimal topologies in structural design using [37] M. Domingo-Espin, J.M. Puigoriol-Forcada, A.-A. Garcia-Granada, J. Llumà,
a homogenization method, Comput. Methods Appl. Mech. Eng. 71 (1988) 197–224. S. Borros, G. Reyes, Mechanical property characterization and simulation of fused
[8] I. Gibson, D.W. Rosen, B. Stucker, Additive Manufacturing Technologies, Springer, deposition modeling Polycarbonate parts, Mater. Des. 83 (2015) 670–677.
2014. [38] A. Amado-Becker, J. Ramos-Grez, M. José Yañez, Y. Vargas, L. Gaete, Elastic tensor
[9] L. Meng, W. Zhang, D. Quan, G. Shi, L. Tang, Y. Hou, P. Breitkopf, J. Zhu, T. Gao, stiffness coefficients for SLS Nylon 12 under different degrees of densification as
From topology optimization design to additive manufacturing: today’s success and measured by ultrasonic technique, Rapid Prototyp. J. 14 (2008) 260–270.
tomorrow’s roadmap, Arch. Comput. Methods Eng. (2019) 1–26. [39] P. Delfs, M. T̈ows, H.J. Schmid, Optimized build orientation of additive manu-
[10] J.H. Zhu, W.H. Zhang, L. Xia, Topology optimization in aircraft and aerospace factured parts for improved surface quality and build time, Addit. Manuf. 12 (2016)
structures design, Arch. Comput. Methods Eng. 23 (2016) 595–622, https://doi. 314–320, https://doi.org/10.1016/j.addma.2016.06.003.
org/10.1007/s11831-015-9151-2. [40] S.H. Masood, W. Rattanawong, P. Iovenitti, A generic algorithm for a best part
[11] T. Zegard, G.H. Paulino, Bridging topology optimization and additive manu- orientation system for complex parts in rapid prototyping, J. Mater. Process.
facturing, Struct. Multidiscipl. Optim. 53 (2016) 175–192. Technol. 139 (2003) 110–116, https://doi.org/10.1016/S0924-0136(03)00190-0.
[12] J. Edgar, S. Tint, Additive manufacturing technologies: 3D printing, rapid proto- [41] K.P. Roysarkar, P.S. Banerjee, A. Sinha, M.K. Banerjee, Multi-objective optimization
typing, and direct digital manufacturing, Johnson Matthey Technol. Rev. 59 (2015) for part quality in stereolithography, 2009 Int. Conf. Comput. Ind. Eng. CIE 2009,
193–198. 2009, pp. 617–623, , https://doi.org/10.1109/iccie.2009.5223746.
[13] K.V. Wong, A. Hernandez, A review of additive manufacturing, ISRN Mech. Eng. [42] R. Paul, S. Anand, Optimization of layered manufacturing process for reducing form
2012 (2012) 1–10, https://doi.org/10.5402/2012/208760. errors with minimal support structures, J. Manuf. Syst. 36 (2015) 231–243, https://
[14] S. Yuan, J. Bai, C. Kai Chua, K. Zhou, J. Wei, Characterization of creeping and shape doi.org/10.1016/j.jmsy.2014.06.014.
memory effect in laser sintered thermoplastic polyurethane, J. Comput. Inf. Sci. [43] P.E. Reeves, R.C. Cobb, Reducing the surface deviation of stereolithography using
Eng. 16 (2016). in-process techniques, Rapid Prototyp. J. 3 (1997) 20–31, https://doi.org/10.1108/
[15] C. Hull, P.F. JACOBS, Rapid Prototyping and Manufacturing: Fundamentals of 13552549710169255.
Stereolithography, (1992). [44] H. Alm Grundström, Topology Optimization for Additive Manufacturing
[16] C.K. Chua, K.F. Leong, 3D Printing and Additive Manufacturing: Principles and Considering Stress and Anisotropy, (2017).
Applications (with Companion Media Pack) - Fourth Edition of Rapid Prototyping, [45] A.M. Mirzendehdel, B. Rankouhi, K. Suresh, Strength-based topology optimization
(2014), https://doi.org/10.1142/9008. for anisotropic parts, Addit. Manuf. 19 (2018) 104–113.
[17] G.V. Salmoria, C.H. Ahrens, M. Fredel, V. Soldi, A.T.N. Pires, Stereolithography [46] L.N.S. Chiu, B. Rolfe, X. Wu, W. Yan, Effect of stiffness anisotropy on topology
somos 7110 resin: mechanical behavior and fractography of parts post-cured by optimisation of additively manufactured structures, Eng. Struct. 171 (2018)
different methods, Polym. Test. 24 (2005) 157–162, https://doi.org/10.1016/j. 842–848.
polymertesting.2004.09.008. [47] R.M. Jones, Mechanics of Composite Materials, CRC press, 2014.
[18] J.R.C. Dizon, A.H. Espera, Q. Chen, R.C. Advincula, Mechanical characterization of [48] O. Sigmund, K. Maute, Topology optimization approaches, Struct. Multidiscipl.
3D-printed polymers, Addit. Manuf. 20 (2018) 44–67, https://doi.org/10.1016/j. Optim. 48 (2013) 1031–1055.
addma.2017.12.002. [49] O. Sigmund, A 99 line topology optimization code written in Matlab, Struct.
[19] T. Liu, S. Guessasma, J. Zhu, W. Zhang, H. Nouri, S. Belhabib, Microstructural Multidiscipl. Optim. 21 (2001) 120–127.
defects induced by stereolithography and related compressive behaviour of poly- [50] Y. Radovcic, A. Remouchamps, BOSS QUATTRO: an open system for parametric
mers, J. Mater. Process. Technol. 251 (2018) 37–46, https://doi.org/10.1016/j. design, Struct. Multidiscipl. Optim. 23 (2002) 140–152.
jmatprotec.2017.08.014. [51] O. Sigmund, J. Petersson, Numerical instabilities in topology optimization: a survey
[20] S.L. Campanelli, G. Cardano, R. Giannoccaro, A.D. Ludovico, E.L.J. Bohez, on procedures dealing with checkerboards, mesh-dependencies and local minima,
Statistical analysis of the stereolithographic process to improve the accuracy, CAD Struct. Multidiscipl. Optim. 16 (1998) 68–75.
Comput. Aided Des. 39 (2007) 80–86, https://doi.org/10.1016/j.cad.2006.10.003. [52] P. Delfs, M. Töws, H.J. Schmid, Surface roughness optimized alignment of parts for
[21] K. Chockalingam, N. Jawahar, U. Chandrasekar, K.N. Ramanathan, Establishment additive manufacturing processes, Proc. 25th Int. Solid Free. Fabr. Symp. (2015)
of process model for part strength in stereolithography, J. Mater. Process. Technol. 1334–1344.
208 (2008) 348–365, https://doi.org/10.1016/j.jmatprotec.2007.12.144. [53] P. Alexander, S. Allen, D. Dutta, Part orientation and build cost determination in
[22] K. Puebla, K. Arcaute, R. Quintana, R.B. Wicker, Effects of environmental condi- layered manufacturing, Comput. Aided Des. 30 (5) (1998) 343–356.
tions, aging, and build orientations on the mechanical properties of ASTM type I [54] D. Frank, G. Fadel, Expert system-based selection of the preferred direction of build
specimens manufactured via stereolithography, Rapid Prototyp. J. 18 (2012) for rapid prototyping processes, J. Intell. Manuf. 6 (5) (1995) 339–345.
374–388.

14

You might also like