Progress in Additive Manufacturing

https://doi.org/10.1007/s40964-018-0061-3

FULL RESEARCH ARTICLE

Topology optimization and additive manufacturing for aerospace

components
Laura Berrocal1   · Rosario Fernández1 · Sergio González1 · Antonio Periñán1 · Santos Tudela1 · Jorge Vilanova2 ·
Luis Rubio3 · Jose Manuel Martín Márquez4 · Javier Guerrero3 · Fernando Lasagni1

Received: 8 February 2016 / Accepted: 8 August 2018

© Springer Nature Switzerland AG 2018

Abstract
One of the main challenges for the aerospace industry nowadays lies in weight reduction of aircraft components without
compromising their structural functionalities. With that goal, structural and topology optimization show up as a combination
of design and modelling techniques. Based on the finite element method (FEM), component optimization consists of removing
material which is dispensable, keeping proper functioning of the modelled part. The result is an optimized geometry, usually
with a complex shape, which is possible to manufacture thanks to additive manufacturing (AM) technologies. In this paper,
the topological optimization methodology has been used to redesign the following components: (1) a connector support of
the VEGA space launcher, (2) a typical lever component from civil aircrafts and (3) housing part from fan cowl structures.
In all of the cases, a significant weight reduction has been reached without major impact on their mechanical behaviour.
Finally, components (1) and (2) were manufactured by laser beam melting (LBM) technology to demonstrate the possibility
of the couple, optimization and AM concepts, as a way to improve the future aerospace structures.

Keywords  Topology optimization · Additive manufacturing · Aerospace components design · Laser beam melting · Laser-
based powder bed fusion of metals · Weight reduction

1 Introduction methods apply mathematical computation and structural

analysis on the pursuit of the optimal geometry: structural
In the last decades, topology optimization has risen as a pow- parameters are set up as design variable and mechanical con-
erful tool for designers and mechanical engineers to enhance straints are required to develop the problem [1]. In this way,
and lighten aerospace components. Structural optimization parameters such as bar length, wall thickness, cross-section

* Laura Berrocal Javier Guerrero

[email protected] [email protected]
Rosario Fernández Fernando Lasagni
[email protected] [email protected]
Sergio González 1
CATEC-Advanced Center for Aerospace Technologies, C/
[email protected]
Wilbur y Orville Wright 19, La Rinconada, 41309 Seville,
Antonio Periñán Spain
[email protected] 2
Airbus Defence and Space—Space Systems, Av. De Aragón,
Santos Tudela 404, 28022 Madrid, Spain
[email protected] 3
Airbus Defence and Space—Centro Bahía de Cádiz (CBC),
Jorge Vilanova Carretera Puerto Santa María‑Sanlúcar, Km. 5.5, El Puerto
[email protected] de Santa María, 11500 Cádiz, Spain
4
Luis Rubio Airbus Defence and Space—Tablada Factory, Av. García
[email protected] Morato s/n, 41011 Seville, Spain
Jose Manuel Martín Márquez
[email protected]

13
Vol.:(0123456789)
 Progress in Additive Manufacturing

area or hole’s size may rule the mathematical problem in siz- slightly greater than zero to avoid mathematical singulari-
ing or shape optimization, allowing the geometry modifica- ties [13].
tion to achieve the optimal shape that fulfils the mechanical Relative density intermediate values are penalized to pro-
requirements [2]. In both cases, structure typology remains mote a solution with fully dense and completely void zones.
unchanged after optimization. On the other hand, topology To include this consideration, the objective function must be
optimization allows creating complex designs from basic penalized as formulated in the following equation:
geometries by distributing mass in the global domain con- Ne
sidered and in the most efficient way. Admittedly, topology
∫
∑ 1
̄ =
F(𝜌) (𝜌e ) p 𝛾mat dΩ , (2)
optimization offers wider freedom of design in contrast with e=1
Ωe
other methods.
The theoretical background of topology optimization where Ωe is the number of element e, γmat is the density
was set by Bendsøe and Kikuchi in 1988 [3]. Through the material and p ≥ 1 is the penalization parameter for interme-
homogenization method they defined the physical domain diate densities. Penalization parameter should be included
under study as a truss-like microstructure composed by an within a range depending on the FE discretization. It must
infinite number of infinitesimal voids which modify size and be noticed that material properties depend on the value of ρ,
shape to adopt the optimum mass distribution. Henceforth, which means stiffness will also adopt relative values associ-
several microstructural models and associated computational ated to density element.
methods were settled to address the problem, to be named, Therefore, an iterative process of successive relative
the level set method (LSM) [4–6] or isogeometric analysis density distribution (RDD) profiles will be evaluated to
[7], among others [8–10]. In particular, practical topology achieve the objective function, by fulfilling the mechani-
optimization examples presented in this paper have been cal constraints. In the optimal solution, the RDD involves
developed based in the solid isotropic material with penali- the minimum use of material, satisfying design constraints.
zation model (SIMP) [11] approach, or density method, and This mass distribution should be interpreted and processed
computations have been carried out through finite element through design, and finally, the optimized components must
analysis (FEA). The SIMP microstructural model states the be validated to ensure an acceptable mechanical response.
material domain under study as a continuous medium, thus it Topology optimization results are complex geometries,
can be discretized through the finite element method (FEM). unfeasible or just not affordable to obtain through conven-
The design variable is called relative density (ρ) and it is tional manufacturing methods (material subtractive tech-
assumed constant within each element of the domain. Rela- niques). Nevertheless, nowadays the industry has a wide
tive density takes values from zero to one, one meaning full offer of novel manufacturing technologies with enough
dense element, it has the real density of the material, and potential to be applied for optimized parts fabrication: addi-
zero refers to no-material or void. tive manufacturing (AM) processes [14]. Since AM is based
The implementation of the topology optimization prob- on the selected addition of material, almost any geometry
lem is subject to mechanical requirements settled in the could be reproduced. Additionally, the reduction of waste
defined domain: displacement limits or dynamic constraints of material due to the recycle possibility is one of the main
must be established. Stress constrained problems need dif- advantages of this technology. Post-processing operations
ferent formulated approaches and several studies have been are also less required than in traditional manufacturing tech-
undertaken in this field [12]. Additionally, the objective niques, reducing costs in terms of time and resources.
must be set, in most of the cases, it will be the minimization Although in the last years AM technologies can use a
of mass. Thus, the topology optimization problem through wide range of materials on its processes (polymers, metals
the SIMP may be formulated as presented in the following or even hybrid materials), metal-based systems are probably
equation: the most relevant ones for aerospace industry. Aluminium
alloys, stainless steel, titanium alloys, and materials such
{ }
Find 𝜌̄ = 𝜌e , e = 1, … , Ne
minimizing F(𝜌) ̄ as ­Inconel® are now available to work with this technology.
(1) Laser beam melting (LBM) has to be underscored regarding
0 < 𝜌min ≤ 𝜌e ≤ 1, e = 1, … , Ne
to guarantee gj (𝜌),
̄ j = 1, … , m
its maturity status and the daily on-going research work with
industrial application. Finally, heat and surface finish treat-
where the design variable ρe represents the relative density ments are commonly added to the manufacturing process to
of the element e (constant within the element), and Ne is the enhance mechanical properties of AM parts [15].
total number of elements in the discretized domain. F(𝜌̄ ) is Nowadays, topology optimization and AM concepts have
the objective function and gj(𝜌̄ ) represents m constraints. been introduced into the aerospace industry [16–18], with
Lower relative density limit ρmin should adopt a value some parts already on board [19]. This methodology is also

13
Progress in Additive Manufacturing

being applied in other fields as the automotive industry [20, distribution is obtained that fulfils the optimization prob-
21]. lem constraints.
In this paper, the redesign of commercial metallic parts • Final design The final design is an interpretation from the
for aerospace application is studied. Components have been topology optimization result. This step must consider: (1)
manufactured by LBM technology, validating the developed applicable design rules from the original component, (2)
optimized design for manufacturability issues. Three differ- applicable manufacturing rules (machining and/or addi-
ent components belonging to an aerospace launcher, civil tive manufacturing limitations), and (3) post-manufac-
and military aircrafts are presented. turing processes (for instance, surface treatments).
• Final validation A validation by FEA is required to guar-
antee that the optimized component accomplishes with
2 Optimization and manufacturing: case the mechanical requirements established by the original
studies FEM and/or design office.
• Manufacturing The last step is the component manufac-
Parts redesigning by applying the topology optimization turing. In this paper, Additive Manufacturing has been
methodology involves a set of steps that includes not only selected as the fabrication solution to be able to obtain
the FEM calculations, but the geometry adaptation of the the complex geometries that topology optimization
component previous to and after topology optimization involves.
computation itself. The different steps (Fig. 1) are described
hereafter: According to this procedure, three different aerospace
components have been optimized as it is presented below.
• Reference FEM A calculation at the beginning of the
optimization process is recommended for reference 2.1 Connector support
results. By this calculation, the component mechanical
behaviour is analysed. The results will be used for the First studied part is a structural component for VEGA
final FEM validation of the optimized component. In this launcher, attached to the Adaptateur Charge Utile (ACU),
step, the part digital model, a DMU (digital mock-up) highlighted in Fig. 2. Its main function is the supporting of
adaptation, is analysed without geometry modification an electrical connector. The original support is an assembly
(for instance, auxiliary lines removal). This action allows of four parts: a top conduct (Fig. 2, detail B) rests on a main,
obtaining an optimal meshing process.
• Preform design In many cases, a simpler geometry is
defined to perform the mass reduction of the component.
In general, it is generated a model using the envelope
volume dimensions of the original part. Interfaces with
other components are respected and geometries interfer-
ing with other pieces are cleared away.
• Topology optimization In this step, the mathematical
optimization process is carried out. Thus, it must be
established: (1) the design variable, (2) function to be
optimized and (3) the problem restrictions (including
non-design space). It involves an iterative process until
reach the feasible solution (or until run out with all the
set iterations). Method parameters may be changed to
Fig. 2  First case study: connector support: view of connector support
obtain an acceptable solution. After computation, a mass assembled to the adaptateur of charge utile (ACU), and detailed struc-
ture

Fig. 1  Diagram of the topol-

ogy optimization and redesign
process

13
 Progress in Additive Manufacturing

bent central plate; two additional plates are joined at both is proposed for an extra optimization: the upper part keeps
sides of the central plate (Fig. 2, detail C). The complete the same configuration, but the base of the component is
structure involves a volume envelope of 297 × 175 × 145 remodelled into a cylinder-like shape. This second initial
­mm3. All parts are joined together by means of rivets, and configuration is presented in Fig. 3b.
finally attached to the ACU main structure by six fixed points Finally, for the component optimization, an AlSi10Mg
(Fig. 2, detail D). It is made of aluminium alloy Al7075 alloy (ρ = 2680 kg/m3, E = 71 GPa, ν = 0.33, σy = 165 MPa)
(ρ = 2800 kg/m3, E = 70.3 GPa, ν = 0.33, σy = 352 MPa) and was selected due to its commercial availability for LBM.
the assembly weight is 453 g, without considering junction
elements (rivets). 2.2 Lever
The main mechanical limitations refer to the dynamic
behaviour. For this launcher part, the category is set by its The second application case consists on a typical door lever
natural frequencies, as well as its mechanical requirements. of civil aircraft. The original geometry is presented in Fig. 4,
Applicable standard stipulates: (1) first natural frequency on with overall dimensions of 187 × 189 × 22 m ­ m3. It is origi-
launchers longitudinal axis (in this case, longitudinal axis nally manufactured in one piece of aluminium alloy 2653T6
of upper tube) must be over 100 Hz; and (2) 50 Hz as the (ρ = 2670 kg/m3, E = 72.4 GPa, ν = 0.33, σy = 240 MPa). The
minimum value allowed on other directions. component weight is 176 g. Due to its commercial avail-
The mechanical response of the original part is estab- ability, ALM AlSi10Mg aluminium alloy is considered for
lished through the following considerations. The electrical optimization.
connector is assumed as a point mass (300 g) located at the Mechanical functionalities are associated to a static load-
centre of the top interface support connector (detail A in ing case. The lever is considered joined to the door along the
Fig. 2). Displacement constraints are established as six fixed passing lateral (detail A in Fig. 4) and clamped in this point.
points in the interface ACU support. A force of 100 N, applied against the corrugated surface in
Dynamic behaviour is studied by Laga [22], the two the handle zone of the component (detail B in Fig. 4), is
first eigenfrequencies (corresponding to eigenmodes on simulated as a distributed load in Z direction, as is shown
lateral axes) are 73.03 and 85.38 Hz, respectively. On the in Fig. 5a.
other hand, successive natural frequencies have values over In this case, a reference FEA is performed. The geometry
100  Hz. Hence, in the original component, mechanical is modelled with 3D elements, further described in Sect. 3.2
requirements are satisfied. of this paper. From the static analysis, the point of maximum
Before topology optimization setup, a redesign of the Von Mises stress (with a value of 23.61 MPa) is located in
original geometry is considered: all parts are unified into a a stress concentration zone, as shown in detail in Fig. 5b.
single component and lateral holes are closed to make pos- Maximum displacement takes place in the extreme of the
sible new truss-like design. The result is a lightener, stiff- handle zone, with a value of 0.310 mm.
ener piece, which can be observed in Fig. 3a, named as first
initial configuration. Furthermore, one additional preform 2.3 Housing

The last case considered in this paper is the formulation of an

alternative design for a component located in a civil aircraft.

Fig. 3  Preform geometries of ACU supports for topology optimiza-

tion: a first initial configuration, and b second initial configuration Fig. 4  Second case study: typical lever. Original geometry

13
Progress in Additive Manufacturing

Fig. 5  Lever mechanical analy-

sis: a loading diagram, and b
Von Mises stress distribution
(MPa)

m3, E = 109.86 GPa, ν = 0.31, σy = 1013 MPa), available for

LBM technology. The original component weight is 745 g.
The housing is fixed on the fan cowl inner surface by means
of 16 rivets. Frontal holes (detail A in Fig. 6b) are designed
to lodge the connection elements with the opposite CLP sub-
system side (right hand-side component in Fig. 6a). A bolt,
part of the CLP actuator main system, is joined through lat-
eral holes (detail B in Fig. 6b) and communicates the actua-
tion force F (82,200 N) to the housing.
For FE analysis, the connection with the opposite hous-
ing is modelled as a sliding displacement condition, with
direction of the cylindrical connecting elements. The 16
base points are constrained to zero displacement. The CLP
actuator bolt assembly is configured by means of rigid ele-
ment nets in both lateral holes (detail B in Fig. 6b), con-
nected to each other with another rigid element. The force
F is applied in the middle point of this major rigid element
with parallel to cylindrical housings’ connectors. Geometry
is discretized with 3D lineal tetrahedral elements (1 mm
average size). After the linear static analysis, the maximum
displacement value is calculated in the centre of the bridge
of frontal panel (0.125 mm), see point 1 in Fig. 7a. Von
Mises stresses are concentrated in fixing hole points and the
maximum is located in the inner surface left-hand lateral
hole (detail A in Fig. 7b) reaching 476 MPa (not exceeding
the material yield stress).
Fig. 6  Third case study: housing. a CLP system, and b original
geometry and loading diagram
3 Topology optimization process
The piece under study is a housing, part of the new cowl loss
prevention system (CLP) of the fan cowl structure. The acti- Topology optimization has been performed using a FEM-
vation system is presented in Fig. 6a and it is used to keep based software, to lighten the components proposed. The
closed both fan cowl structures. The analysed component software solves the optimization problem using the SIMP
(Fig. 6b) is enclosed in a volume envelope of 207 × 171 × 83 method. The design variables are the relative densities asso-
­mm3, approximately. The considered material for calcula- ciated to each of the finite elements composing the domain.
tions and topology optimization is TiAl6V4 (ρ = 4430 kg/ The objective function and constraints can be defined in

13
 Progress in Additive Manufacturing

Fig. 7  Housing static analysis

results: a displacement distribu-
tion (mm), and b Von Mises
stress distribution (MPa)

terms of mass, static displacement, natural frequency, 2. Once the previous step is finished, the convergence test
weighted compliance, etc. The optimization computational has to be performed. The feasible design is achieved
method is an automatically iterative process in which the if the change in the objective function is less than the
following steps are executed (from i to iii in a loop, until objective tolerance (a value of 0.005 has been consid-
convergence or limit of iterations are reached): ered in this paper), for two consecutive iterations, and
constrains violation are less than 1%.
1. Mechanical analysis through FEM on an initial design 3. In case convergence is not reached, compute the update
(for instance, homogenous distribution of material). of density variable. In this paper, the Method of Feasi-
From this step, the stress and displacement fields are ble Directions has been used for this step. This method
known. was implemented for structural optimization problems

13
Progress in Additive Manufacturing

by Garret N. and is described in [23]. This method is in 3.1 Connector support

fact a generalization of the method first developed by
Zoutendijk in [24]. For the first initial configuration optimization, tetrahedral
linear element (2 mm average size) and a point mass element
Once FE model has been created, it must be differenti- have been utilized to discretize the support’s geometry and
ated into design and non-design spaces. This division is to model the electrical connector’s mass, respectively. A net
conducted by the assignment of specific properties to the of rigid elements is used to model the connection between
elements. It is adopted a relative density value equal to point mass and top support interface, it means all degrees of
one to the non-design space, for every step of the process. freedom (d.o.f.) of dependent nodes (support interface) are
With this action design variables associated to non-design associated to the central independent node, where the point
space will not change during the process and the elements mass is located.
will be fully dense in the optimum. Although intermediate FE model is set and the whole 3D domain is considered
values of relative density are penalized in the mathemati- as design space, with the exception of the top interface con-
cal problem, they might appear in the optimal solution. nector support and the join points with ACU surface. In
They will be interpreted as solid or void zones during the Fig. 8a, dark regions belong to design space and grey areas
redesign process. correspond to non-design space. Though static behaviour
In this paper, mechanical requirements associated to is not critical for this component, a linear static case is per-
specific structural analyses are considered and minimiza- formed. The electrical connector weight is considered as a
tion of mass is formulated as objective function in every force ­FA applied with perpendicular direction to top interface
study case. Then, the iterative optimization is carried surface, pushing the interface area. Von Mises stresses are
out to obtain the optimal relative density distribution. controlled under material yield strength. Stress restrictions
Every FEM calculation has been performed by means are not limiting compared to those associated to modal anal-
of a computer with the following characteristics: 2xIntel ysis, which leads the topology optimization redesign. First
Xeon E560 (2.4 GHz and 4 Cores) microprocessor with natural frequency is constrained with a minimum value of
12 GB RAM. The optimal RDD is used to redesign origi- 50 Hz. Mass fraction minimization is formulated and both
nal parts. A Computer Aided Design (CAD) software has static and modal analyses are performed.
been used for the components’ final geometry design. Figure 9a shows the relative density distribution map
obtained (elapsed time 25 min), which is used as guideline
for the component redesign. As it can be seen, despite the

Fig. 8  Geometry and loading

diagram for connector support
optimization: a first configura-
tion, and b second configuration

13
 Progress in Additive Manufacturing

Fig. 9  Topology optimization
process of the first configuration
for ACU connector support. a
Result from optimization: rela-
tive density map, with relative
density scale from 0 to 1, and b
optimized geometry

penalized formulation of the problem, several areas of inter- response, the first natural frequency has a value of 78.9 Hz
mediate density values appear, they may be evaluated dur- (non-longitudinal axis direction). Second and subsequent
ing the redesign step. The result of remodelling is shown eigenfrequencies are over 100 Hz, thus mechanical require-
in Fig. 9b. This first optimized configuration has a weight ments are fulfilled.
of 197 g which means a mass reduction of 57% compared For the topology optimization of the second original
to original assembled component. Final FEM validation is configuration, an identical FE model formulation is exe-
undertaken to guarantee mechanical functionality. Static and cuted, the load case is shown in Fig. 8b. The relative den-
dynamic analyses are carried out. From static validation, a sity distribution map obtained in the process (elapsed time:
maximum Von Mises stress value of 16.98 MPa is obtained 22 min.) is displayed in Fig. 11a where it can be observed
(see Fig. 10a). Maximum displacement is calculated in the multiple areas of intermediate densities. By interpret-
top interface, with a value of 0.647 mm. Related to dynamic ing this diagram, the geometry is redesigned obtaining

Fig. 10  Results from optimized connector supports static analysis: a Von Mises stress (MPa) distribution of first optimized configuration, and b
Von Mises stress (MPa) distribution of second optimized configuration

13
Progress in Additive Manufacturing

Fig. 11  Topology optimization
process of the second configura-
tion for ACU connector support.
a Result from optimization:
relative density map, with rela-
tive density scale from 0 to 1,
and b optimized geometry

a piece of 168 g weight (see Fig. 11b), which supposes a 3.2 Lever

weight reduction of 63%, compared with original assem-
bled component. Finally, from static analysis, a maximum Lever domain has been discretized with tetrahedral linear
VM stress of 16.95 MPa (Fig. 10b) and a maximum dis- elements (1 mm average size). Load and boundary condi-
placement of 0.792 mm have been obtained. From modal tions are described in Sect. 2.2. FE mesh is classified in
analysis, a first frequency of 70.9 Hz (non-longitudinal design and non-design spaces as shown in Fig. 12a. Mass
axis) has been calculated. Second natural frequency has a reduction is defined as the objective function and Von Mises
value of 83.0 Hz (non-longitudinal axis) and subsequent stresses are controlled at any point of the geometry, not
eigenfrequencies are over 100 Hz. Therefore, mechanical exceeding the material yield strength. With these consid-
functionality is guaranteed. erations, optimization computation is implemented (elapsed

Fig. 12  Topology optimization process of lever: a optimization setup, loading diagram, b relative density map, with relative density scale from 0
to 1, and c optimized geometry

13
 Progress in Additive Manufacturing

time 31 min.). Optimal relative density diagram is shown in Table 1  Displacement constraints (mm) for topology optimization of
Fig. 12b. The result of the edition is a component with bar case study 3 (housing) and displacement results from validation of
optimized geometry
structure typology, see Fig. 12c. The final geometry, with a
weight of 121 g, supposes saving third part of the original Point 1 2 3 4 5
mass. The last step before manufacturing consists on the
Displacement constraint (mm) 0.130 0.110 0.110 0.055 0.055
validation through FEM: the static analysis defined for opti-
Displacement result (mm) in 0.1293 0.105 0.104 0.043 0.042
mization is undertaken and a maximum Von Mises stress opt. geometry
value of 64.3 MPa is obtained (Fig. 13).

3.3 Housing 4 Components manufacturing

The original component FE discretization, presented in Two of the optimized components have been manufactured
Sect. 2.3, is used for the topology optimization computation by LBM technology using a RENISHAW ALM250 system
and the linear static analysis previously described is imple- [25]. The manufacturing results are depicted below:
mented. Optimization constraints are linked to the elements
located in five points: in lateral holes (points 2 and 3 in 4.1 Connector support
Fig. 7), on top edge of the frontal surface (points 4 and 5 in
Fig. 7, respectively), and finally in the frontal bridge (point Three different configurations have been manufactured
1 in Fig. 7). Displacement restrictions are listed in Table 1. (scale 1:2) to show the evolution in the topology optimi-
As in previous cases, Von Mises stresses are controlled zation of this component. Figure 16a shows the connector
under yield strength and mass fraction minimization is the
objective of the problem. With those definitions, the opti-
mization calculation is executed (elapsed time: 1 h 21 min.),
the density distribution map shown in Fig. 14a is obtained.
For safety reasons, all riveted interfaces are kept in the new
design, although in relative density diagram they appear as
removable parts. The geometry displayed in Fig. 14b has
a weight of 600 g, a 19% less mass than the original one.
Finally, the optimized geometry is validated by means of
FEA. Maximum stress of 489 MPa is under the material
yield strength (Fig. 15) and displacements at controlled
points are listed in Table 1.

Fig. 14  Topology optimization process of housing: a relative density

Fig. 13  Von Mises stress (MPa) distribution obtained from optimized map, with relative density scale from 0 to 1, and b optimized geom-
lever static analysis etry

13
Progress in Additive Manufacturing

Fig. 15  Von Mises stress (MPa) distribution obtained from optimized

housing static analysis

support generation placed on the manufacturing plate. It can

be seen the supports used during the fabrication process as
well as the selected positioning of the pieces. It is important
to highlight the strategy adopted for the building of opti-
mized designs, including grid structures in holes resulting
Fig. 16  Manufacturing results of the connector support generation
from optimization as a way of manufacturability improve- (from left to right): unified original design, first optimized design
ment. Furthermore, the use of the grid structure reduces with gridded structure and second optimized design with gridded
the number of supports needed, without major penalization structure. a Connector support designs on manufacturing plate, and b
after post-processing operations
in component weight (see Table 2). This strategy can be
observed in Fig. 16a, and after post-processing in Fig. 16b.
• For the connector support, the modal requirement is sat-
isfied in both optimized configurations, being the first
4.2 Lever eigenfrequency (the most restrictive limit value from a
design point of view) of 78.9 and 70.9 Hz for the first and
The selected orientation of lever component inside the man- second configurations, respectively. On the other hand,
ufacturing chamber can be observed in Fig. 17. The image the maximum Von Mises stress in both cases is 16.98 and
shows a view of the pre-processing software used to define 16.95 MPa (for first- and second-optimized configura-
the component location and the supports needed for its cor- tions) lower than the material yield strength (165 MPa).
rect fabrication. Figure 18 shows the lever as fabricated, In both cases, the weight reduction reached is 57 and 63%
after the process of elimination of supports. for first and second designs, respectively.
• Regarding the lever optimization, a maximum stress
of 64.3 MPa is observed at stress concentrator, almost
one order magnitude lower than material yield stress
5 Conclusions (165 MPa). It means that the optimized component, with
a weight reduction of 32%, fulfills the stress requirement.
By combining concepts of topology optimization and AM, • Finally, the housing optimized design reaches an impor-
it has been possible to reduce considerably the weight of the tant weight reduction (19%) keeping all fixing points,
studied components. Table 2 provides an overview of the with a maximum stress value of 489.8 MPa, meaning
results obtained in this work. It must be also highlighted the a large safety margin until the material yield stress
accomplishment of mechanical criteria in every case, with (1013 MPa). It is important to notice that a larger weight
the following additional comments: reduction can be achieved by reducing joining (rivets)

13
 Progress in Additive Manufacturing

Table 2  Topology optimization Component Original Number Opmizaon Weight Number of La
ce

results of three case studies: weight of parts weight reducon parts of design
connector support, lever and (g) of (g) (%) opmized weight/
housing original design weight
design reducon
(g/%]
Connector 453 4 197 57 1 251/45
support

168 63 1 196/57

Lever 176 1 121 32 1 N/A

Housing 745 1 600 19 1 N/A

Fig. 18  Optimized aircraft door lever manufactured by LBM after

Fig. 17  Software view of lever positioning inside manufacturing
post-processing operations
chamber

13
Progress in Additive Manufacturing

interfaces from 16 to 12 with similar maximum of stress 9. Bertsch C, Cisilino AP, Calvo N (2010) Topology optimization
value (489.3 MPa). of three-dimensional load-bearing structures using boundary ele-
ments. Adv Eng Softw 41:694–704
10. Sigmund O, Clausen PM (2007) Topology optimization using
Using the topology optimization methodology and a mixed formulation: an alternative way to solve pressure load
additive manufacturing technology, it is possible the problems. Comput Methods Appl Mech Eng 196:1874–1889
development of highly optimized aerospace parts. With 11. Bendsøe MP, Sigmund O (2003) Topology optimization: theory,
methods and applications. Springer, Berlin
the combination of both concepts, weight reduction with- 12. París J (2007) Restricciones en tensión y minimización del peso.
out compromising parts mechanical requirements can be Una metodología general para la optimización topológica. Uni-
achieved. This coupled strategy may be really attractive and versidade da Coruña, Spain
useful in first stages of design, creating light and optimal 13. París J, Martínez S, Navarrina F, Colominas I, Casteleiro M (2010)
Topology optimization of aeronautical structures with stress con-
components from inception. straints: general methodology and applications. Universidade Da
It must be also considered the need of smoothing opti- Coruña, Spain
mized geometries to minimize stress concentrations. Designs 14. Singh S, Ramakrishna S, Singh R (2016) Material issues in addi-
must be adapted at the same time according to AM manufac- tive manufacturing: a review. J Manuf Process 25:185–200
15. Vrancken B, Thijs L, Kruth J-P, Van Humbeeck J (2012) Heat
turing restrictions (for instance, minimization of supporting treatment of Ti6Al4V produced by selective laser melting: micro-
elements). structure and mechanical properties. J Alloy Compd 541:177–185
Future developments can involve other AM concepts such 16. Jones R (2011) The design future. http://altai ​ r enli ​ g hten​
as the inclusion of grids and lattice structures, and consider .com/2011/12/desig​n-futur​e-blog/. Accessed 14 Dec 2011
17. Tomlin M, Meyer J (2011) Topology optimization of an additive
intermediate values of relative density effect. In this inves- layer manufactured (ALM) aerospace part. In: The 7th altair CAE
tigation, lattice structures have been used for manufacturing technology conference, pp 1–9
reasons only. The applicability of grids to components opti- 18. Machunze W, Lehmann T, Hein P (2013) Topology design of a
mization, by considering them as mechanically relevant, is metallic load introduction bracket manufactured by ALM. http://
docsl​ide.us/techn​ology​/topol​ogy-desig​n-of-a-metal​lic-load-intro​
one of the potential research fields nowadays. ducti​on-brack​et-manuf​actur​ed-by-alm.html. Accessed 23 Apr
2013
19. GE Aviation (2015) GE aviation’s first additive manufactured
References part takes off on a GE90 engine. http://www.geavia​ tion.​ com/press​
-relea​se/ge90-engin​e-famil​y/ge-aviat​ion%E2%80%99s-first​-addit​
ive-manuf​actur​ed-part-takes​-ge90-engin​e. Accessed 14 Apr 2015
1. Schmit LA (1960) Structural design and structural optimizations. 20. Merkt S (2013) DPP—optimization potentials by laser based
In: Proceedings on the 2nd conference on electronic computation, manufacturing. Altair Conf 2013:6–13
American Society of Civil Engineering, New York, pp 105–122 21. Carello M, Airale A (2013) Design and building of the prototype
2. Haslinger J, Mäkinen RAE (2003) Introduction to shape optimiza- IDRApegasus. http://es.slide​share​.net/Altai​rHTC/desig​n-and-
tion: theory, approximation and computation. SIAM, Philadelphia build​ing-of-the-proto​type-idrap​egasu​s. Accessed 9 May 2013
3. Bendsøe MP, Kikuchi N (1988) Generating optimal topologies in 22. Laga R, González Requena I, Vilanova Calvo J (2013) Additive
structural design using a homogenization method. Comput Meth- manufacturing. Trade-off de tecnologías, Universidad Politécnica
ods Appl Mech Eng 71:197–224 de Madrid, Spain
4. Osher S, Santosa S (2001) Level set methods for optimization 23. Vanderplaats G (1973) Structural optimization by methods of fea-
problems involving geometry and constraints I. Frequencies of a sible directions. Comput Struct 3:739–755
two-density inhomogeneous drum. J Comput Phys 171:272–288 24. Zoutendijk G (1960) Methods of feasible directions: a study in
5. Wang MY, Wang X, Guo D (2003) A level set method for struc- linear and non-linear programming. Elsevier, Amsterdam
tural topology optimization. Comput Methods Appl Mech Eng 25. Additive Manufacturing Systems. http://www.renis​haw.com/en/
192:227–246 addit​ive-manuf​actur​ing-syste​ms--15239​. Accessed 21 May 2014
6. Shojaee S, Mohammadian M (2012) Structural topology optimiza-
tion using an enhanced level set method. Sci Iran 19:1157–1167
Publisher’s Note Springer Nature remains neutral with regard to
7. Dede L, Borden MJ, Hughes TJR (2011) Isogeometric Analysis
jurisdictional claims in published maps and institutional affiliations.
for topology optimization with a phase field model. Arch Comput
Methods Eng 19:427–465
8. Sigmund O (2001) A 99 line topology optimization code written
in Matlab. Struct Multidiscip Optim 21:120–127

13