Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 60 (2017) 205 – 210

27t h CIRP Design 2017

Design and development of an additive manufactured component by

topology optimisation
Dan Waltona*, Hadi Moztarzadehb
a
University of Warwick, Coventry, CV4 7AL, United Kingdom
b
WMG, University of Warwick, Coventry, CV4 7AL, United Kingdom

* Corresponding author. Tel.: +44944 427 361; E-mail address: [email protected]

Abstract
This paper investigates a design and development process for Electron Beam Melting (EBM) which incorporates a simulation-driven design
process called topology optimisation. Research consists of a review of EBM design principles and validation of mechanical properties for
Ti-6Al-4V ELI. Findings are applied to a case study whereby a pair of suspension uprights are redesigned and manufactured by EBM with the
objective of mass reduction. Previous studies indicate that optimisation shape controls can potentially minimise the number of supports required
for EBM. Meanwhile, a parametric solid/surface modelling approach can allow for greater control of design intent when designing for larger
assemblies or structures. Application of the proposed strategy resulted in the case study having a 36 % reduction in mass in comparison to a
CNC aluminium design. Whilst the EBM alternative design also yields an 86 % reduction in raw material use, there is a sevenfold increase in
cost for manufacture alone. This work is an example of topology optimisation being a suitable approach when Designing for AM (DfAM). But,
the cost and time constraints associated with EBM limits application of the process to high-performance industries such as motorsport, aerospace,
or tooling solutions.
© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2017 The Authors. Published by Elsevier B.V.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 27th CIRP Design Conference.
Peer-review under responsibility of the scientific committee of the 27th CIRP Design Conference

Keywords: Design for Additive Manufacturing (DFAM); Topology Optimisation; Electron Beam Melting; Powder Bed Manufacturing.

1. Introduction Importantly, EBM is one of the few metal-AM processes

capable of creating products to a near net shape form.
Topology optimisation is a simulation-driven, design This study has been organised to investigate how topology
technique which is used to create conceptual structures. It is an optimisation can be included within the design and
incredibly powerful tool, and can be used with the objective of development process for EBM. This will be achieved by the
mass reduction or thermal control. Resulting topologies are following objectives:
often organic, but usually cannot be manufactured without x Understand the design principles of EBM;
simplification for traditional manufacturing processes [1]. x Define an EBM-based approach to topology optimisation;
Fortunately, the emergence of Additive Manufacturing (AM) x Develop an EBM case study by redesigning a set of rear
enables the potential of topology optimisation to be further suspension uprights for a Formula Student racing vehicle;
utilised. This is due to the layer-by-layer process of AM which x Validate the mechanical properties of EBM manufactured
allows almost unrestricted design limitations. material;
Of the many AM processes available, Electron Beam x Manufacture, post-process and evaluate the case study.
Melting (EBM) arguably has the greatest potential for
producing end-use components. As with all AM processes, 2. Additive Manufacturing
EBM parts are directly manufactured by inputting CAD data
into a machine. Material choice includes titanium alloys AM describes a family of technologies which are accruing a
(Ti-6Al-4V) or Cobalt-Chrome alloys as standard. However, great deal of interest from the automotive, aerospace, and
some research institutions are exploring other materials [2]. medical sectors [3]. The primary driving factor for

2212-8271 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 27th CIRP Design Conference
doi:10.1016/j.procir.2017.03.027
206 Dan Walton and Hadi Moztarzadeh / Procedia CIRP 60 (2017) 205 – 210

development is AM’s ability to create almost limitless limit of materials [6]. The choice of post-AM processing is
geometries without the need for tooling. Recent research has entirely driven by the customer’s time and financial
also opened applications for a range of polymers, metals and requirements, and each choice must be carefully evaluated to
ceramics. However, it is of the upmost important to understand understand its effect on the process chain
that AM technologies discussed are not a replacement for
traditional processes due to AM’s comparatively slow 2.2. Design Rules for EBM
manufacturing rate [1].
The British Standards Institution (BSI) and American There are a small number of design rules which are
Society for Testing Materials (ASTM) have classified AM unavoidable for EBM, and are similar across majority of AM
processes into seven groups [4]. The suitability of each technologies.The following rules are necessary in order to
technology is largely dependent upon the designer’s choice of design a successful EBM workpiece:
material, surface finish, component size, cost and intended
production volume. However, the scope of this report is x Do not exceed the size limits of the equipment. The largest
interested EBM, a powder bed fusion process [2]. EBM is the EBM machine, the Arcam Q20plus, has a maximum build
only powder bed fusion process which utilises an electron beam size of 350 mm in diameter and 380 mm in height [2];
to selectively melt regions of material within the powder bed x Overhanging surfaces which create an acute angle to the
(see Figure 1). It is also capable of creating several melt pools powder bed may require supports to prevent molten material
simultaneously and so typically has faster build speeds in from sinking into powder. Typically, angles of greater than
comparison with other powder bed fusion processes [4]. 45° will not need supports but this is material dependent;
Manufactured components are also reported to be free of both x Do not create fully-enclosed voids or hollows as powder
residual stress and martensitic structures due to the powder bed cannot be removed [5];
being at an elevated temperature throughout the process [2]. x Tooling access is required when removing supports [5]
x Provide material overstock to high tolerance faces (e.g.
bearings). The overstock can then be removed by a post-
EBM processes such as CNC machining or EDM [7];
x Be aware of the minimum feature size for a given material
or AM system [7].

2.3. Design Guidelines for EBM

Figure 1: Illustration of the EBM process. Design guidelines are not compulsory and will depend on
aspects such as component geometry, intended use, or
2.1. Process Chain production volume. However, consideration of these factors
can lead to designs which are more suitable for production. The
The EBM process begins with a CAD model of the intended following non-exhaustive guidelines are suggested:
part which is then approximated by a triangular mesh format.
This mesh is formatted as a STereoLithography (STL) file, but x Minimise the requirement for supports by considering
is commonly referred as Surface Tessellation Language to component orientation at the start of the design process [8];
avoid confusion with the similarly named AM process. The x Supports are likely to damage the surface of the material
mesh is then digitally sliced into a large number of co-planer which they are removed. So be aware that overhanging
cross sections which represent each build layer. In processing, surfaces or anchoring point will have a poor surface finish;
a cross-section is created by spreading a layer of metallic x Powder removal from internal geometries can be difficult
powder across the build chamber by using a rake. The due to the flow behavior of powder or partially sintered
cross-section is then traced by a focused electron beam(s) material. During design, consider methods to ease powder
which selectively melts regions of material. Once a single layer removal, such as tooling access [5];
is completed, another layer of powder is spread over the x Avoid thin, vertical structures as they are prone to breaking
previous and the next cross section is processed. The process is if knocked by the powder rake [7]. If they cannot be avoided,
repeated until the entire form has been generated [1]. consider re-orientating the component within the build
A number of basic post-AM processes are recommended, or chamber or increasing the footprint area of the feature;
even required, after completion of a build. First, the lightly x Avoid sharp edges or corners. They act as stress raisers and
bound powder which surrounds a consolidated part must be can cause distortion and peeling from the build plate [7];
removed. This is best completed by a powder recovery system x Consolidate assemblies and manufacture in-situ to eliminate
which recycles unprocessed material [5]. Second, the support assembly time and simplify the supply chain [7];
structures which prevent thermal or mechanical distortion x Provide line of sight access to all faces if surface finishing
during processing are removed by hand tools, post-machining, components by processes such as shot peening [7];
or electrical discharge machining. Although components are x Minimise variations in section thickness to prevent warping
usable by this stage, they can be improved by further due to differing thermal gradients around the melt pool [9].
processing. For example, Hot Isostatic Pressing (HIP) can be x Validate the mechanical properties of a material for a given
utilised to reduce porosity and drastically improve the fatigue AM system and process parameters.
 Dan Walton and Hadi Moztarzadeh / Procedia CIRP 60 (2017) 205 – 210 207

3. Topology Optimisation 3.1. Considering AM Rules and Guidelines in Pre-Processing

Topology optimisation is a simulation-driven, design The design domain describes the maximum volume which
technique which is used to optimise structures to give a can contain part geometry. It is good practice to initially
conceptual design [10]. This report focusses on the use of the maximise this volume to improve the optimisation outcome by
SIMP topology optimisation approach, however there are other starting with the build volume of the AM machine and then
approaches such as BESO which will not be explored here [11]. subtracting the interaction of other components in an assembly.
First, a Finite Element (FE) model is established under set of The initial design space can then be refined to minimise
boundary conditions (Figure 2a). The optimisation is run for a computation time by eliminating low stress areas from
specified objective, and elements are prioritised a pseudo- preliminary optimisations. Regions of the domain must also be
density coefficient which indicates importance (Figure 2b) as a specified as design and non-design space, whereby the non-
function of Young’s elastic modulus such that: design space is fixed material in which to apply boundary
conditions (see Figure 2a).
‫ܧ‬ு ൌ ߩ௣ ‫ܧ‬଴ Optimisation objectives and constraints are used to define
the purpose of a topology optimisation. A single objective is
In this form, ‫ܧ‬௛ is the calculated Young’s modulus of an set, and then a number of complimentary constraints can be
optimized element, ߩ is the pseudo-density coefficient, ‫ ݌‬is a assigned. For example, a minimise compliance objective can
penalization factor used to discriminate between void and solid be constrained by a mass value and stress value. There are a
material, and ‫ܧ‬଴ is the Young’s modulus of the bulk material great number of optimisation objectives, and they depend
[10]. The pseudo-densities are on a continuous scale whereby entirely upon the application of the focus part.
a value of 0 represents a void and 1 represents solid material. Shape controls are used to fine-tune an optimisation towards
Low density elements are filtered out, and the optimal topology a design which follows set design rules. Existing software
for the applied scenario is revealed (Figure 2c). Finally, the FE (such as HyperMesh) can consider aspects such as draw
mesh is interpreted using CAD in order to adapt a component direction, symmetry, or extruded shapes for conventional
for a given manufacturing process (Figure 2d). The result is an processes. Unfortunately, this is not the case for AM and so
organic shaped component which has converged towards a existing shape controls must repurposed. The following shape
state of homogeneity with regards to internal stress. controls are suggested; however, their application will depend
on the situation:

x Limiting the maximum allowable member thickness is

likely to increase the number of structural members and
consequently decrease the distance of overhanging surfaces
[1]. As a result, the designed component would require a
reduced number of supports;
x Increase the minimum allowable member thickness close to
the maximum. This will prevent large changes in cross-
section which can cause warping during the EBM process.
This value should be greater than the feature resolution of
the AM system. However, this value will be largely driven
by the software restrictions which dictate the difference
between minimum and maximum member thickness;
x Single draw shape controls can reduce the need for supports
by eliminating overhanging surfaces. However, this will
impede the optimality of designed components and the user
must have a knowledge of build orientation prior to the
optimisation stage. The use of an extrusion shape control is
a similar approach, but will significantly damage component
optimality;
Figure 2: Topology Optimisation work flow, with the example of a 75 % x Lattice optimisation is a recent AM-specific shape control.
mass reduction for a cube under a compressive load on the top face. It is a continuation of topology optimisation which assigns
intermediary regions with a lattice. However, it introduces
The design rules of a conventional manufacturing process issues with powder removal if the lattice is too fine.
limit the benefits of topology optimisation. However, AM has
almost unlimited design freedoms and so the potential benefits 3.2. Topology Interpretation
of topology optimisation can be further utilised. This section of
the paper will investigate methods of considering the EBM Topology interpretation follows the optimisation stage and
design rules, and guidelines, of EBM during topology is the process of converting an optimised 3D mesh into a
optimisation. manufacturable form. There are two distinct approaches which
can be taken, each depending on the needs of the business.
208 Dan Walton and Hadi Moztarzadeh / Procedia CIRP 60 (2017) 205 – 210

The first approach includes tracing optimised topologies The results are summarised in Table 1. The loading conditions
within any CAD package. This makes use of readily available of the upright were determined using a mixture of physical
skills within the engineering community, such as solid, surface, testing and hand calculations (Figure 3). The suspension
or NURBS modelling. This can be time-consuming process and upright is simulated under a static loading. However, the failure
the result will depend on user CAD proficiency and process criteria is defined as the fatigue limit of EBM Ti-6Al-4V which
knowledge. However, the result is a fully parameterised design is approximately 450 MPa at 106 cycles [6].
which can easily be manipulated to account for AM build
orientation, support architecture, or heat dissipation. This 4.2. Simulation
approach is best suited for applications whereby cycle time,
material cost, or high performance are a priority. The objective of the optimisation is to minimise weighted
The second interpretation approach is to include a compliance. The design was constrained to a maximum mass
post-optimisation stage which uses surface-fitting software or of 0.6 kg and a maximum stress of 450 MPa. The resulting
a further shape optimisation. In unique situations, it may even topology is shown in Figure 4.
be possible to manufacture directly from the optimised mesh
format to eliminate the interpretation stage entirely. Whilst Table 1: Properties of as-built Ti-6Al-4V ELI. (SD = Standard Deviation)
Property Value
time efficient, these approaches are likely to create designs
which require more supports or post-processing which is only Young’s Modulus 106 GPa (SD: 9)
acceptable if the benefits of reduced design time outweigh the Ultimate Tensile Strength 956 MPa (SD: 38)
cost of manufacture. Strain at Break 12.0 % (SD: 1.0)
Fatigue limit at 106 cycles [6] 450 MPa
4. Case Study
Poisson’s Ratio [12] 0.30

To demonstrate the topic of topology optimisation for EBM,

a case study has been organised with the University of
Warwick’s Formula Student Team. The focus component is a
pair of rear uprights which are the main structural member of
the suspension assembly. The design alternative must follow
the following design specification:

x The design must be a lightweight alternative to the

original, machined aluminium design;
x The design must be produced in Ti-6Al-4V ELI by EBM;
x The two rear uprights must be identical;
x The material utilization of the design must be improved.

4.1. Simulation Pre-Processing

Pre-processing includes creating the design domain,

assigning boundary conditions, applying shape control, and
setting the objectives and constraints of the optimisation. This
case study uses Altair HyperMesh as a pre-processor and
Optistruct as a solver for the optimisation. The upright design Figure 3: The design space and loading conditions on the rear upright.
space is constrained by the maximum build space of the Arcam
A2X, and the neighboring components in the suspension
assembly. The component was described by a volume
tetramesh with an average element size of 2.0 mm and a total
of 122,018 elements. The following shape controls were
applied to the design space:

x A minimum member sizing of 6 mm. This is large enough

to encourage member connectivity and prevent process
defects from significantly effecting the component.
x A maximum member sizing of 12 mm. HyperMesh requires
that this value is at least twice the minimum member size.
x A symmetry constraint in the y-z plane shown in Figure 3.

Tensile properties of as-built EBM Ti-6Al-4V ELI were

validated by in-house mechanical testing on rectilinear samples
which were built vertically with respect to the build chamber. Figure 4: Post optimisation result, showing element psuedo-densities >0.25.
 Dan Walton and Hadi Moztarzadeh / Procedia CIRP 60 (2017) 205 – 210 209

4.3. CAD Interpretation

Remembering that a pseudo-density of 1 is solid, and 0 is a

void - the model shown in Figure 4 illustrates elements with a
density greater than 0.25. Ideally, the “optimal” design would
filter out elements with a density of less than 0.505. However,
a number of member connectivity issues arise when this is
done. There are methods to further discriminate between solid
or void elements, but they will not be discussed here [10].
Topology interpretation has been completed by tracing the
optimised mesh in SolidWorks 2014. This has allowed for
various modifications to the geometry, such as: tooling access
for support removal; rounding of edges; and providing
machining overstock for the bearing housing. The traced
geometry is then validated using Optistruct, and reuses the Figure 5: The as-built upright before support removal.
same boundary conditions used in optimisation. At this stage,
the parametric model allowed for rapid design iterations to
remove any stress concentrations which remained in the design.
The final design has a mass of 0.536 kg, a reduction of 36 %.
FE analysis indicates that the component has a maximum
von-Mises stress of 265 MPa, which equals a factor of safety
of 1.70. The maximum displacement of the component is
0.32 mm, a 28 % increase on the previous design. The topology
interpretation stage was completed in approximately 20 hours.

4.4. Manufacture

The final upright design was manufactured by an ARCAM

A2X EBM machine using Ti-6Al-4V ELI. Both parts were
manufactured in a single build. The as-built component and
support architecture is shown in Figure 5. Faces of the bearing
housing were CNC machined, but no additional surface
finishing or heat treatments were used. The final component,
ready for assembly with the vehicle, is shown in Figure 6.
Figure 6: The post-machined EBM upright.
4.5. X-ray CT

The component was inspected using X-ray Computed

Tomography (CT). No significant voids or layer shifts were
found which could compromise the integrity of the vehicle,
However it is likely that there is porosity which is smaller than
the resolution of equipment. An inclusion was found within the
bearing housing as shown in Figure 7, but its effect is expected
to be minimal as it is located in a low-stress area of the design.

4.6. Final Comments

There are a number of post-AM processes which could be

completed to improve the structural integrity of the upright. Figure 7: X-Ray CT of as-built EBM upright. Inclusion is circled in red.
First of all, recent studies indicate that HIP could be used to
increase the fatigue limit of Ti-6Al-4V ELI from 450 MPa to 5. Evaluation
700 MPa for 106 cycles [6]. This could allow for a further mass
reduction to the component; but would increase process cost The case study was optimised under a single combination of
and is beyond the scope of this research project. A second topology optimisation shape controls in order to achieve the
consideration is to investigate methods of improving the objectives of the research project. However, this approach is
surface finish of the component. This could also lead to somewhat limited and a true engineering application must
improvements in component fatigue behavior and aesthetics. consider alternative shape controls (as discussed in section 3.1)
and even alternative materials in order to fit the requirements
of the specification and customer. Nonetheless, even the
210 Dan Walton and Hadi Moztarzadeh / Procedia CIRP 60 (2017) 205 – 210

concise design optimisation process presented demonstrates Acknowledgements

how topology optimisation and EBM can be combined to create
lightweight components with. improvements in component This project forms part of a Mechanical Engineering degree
strength. A comparison of the original design and the EBM at the University of Warwick. Research was enabled by WMG
titanium design is summarized within Table 2. Costs include who provided supervision, and resources for mechanical
materials and labour for machining/building a single part. testing. In addition, the Manufacturing Technology Centre
provided technical expertise, prepared and manufactured the
Table 2: A comparison of the original upright to the EBM alternative. component by EBM, and provided HIP and X-ray CT services.
Machined EBM
Assessment Criteria % Change
Aluminium Titanium
References
Mass 832g 536g - 36 %
Factor of Safety 0.95* 1.70** + 79 % [1] D. Brackett, I. Ashcroft and R. Hague, "Topology
Maximum Displacement 0.25 mm 0.32 mm + 28 % Optimization For Additive Manufacturing," in SFF
Raw Material Use 7.0 kg 1.0 kg - 86 % Symposium 2011, Austin, TX, 2014.
Cost of Manufacture £342.00 £ 2980.97 + 772 % [2] Arcam AB, "Electron Beam Melting," 2016. [Online].
Available: http://www.arcam.com/technology/electron-
*w.r.t yield strength. **w.r.t fatigue limit at 106 cycles.
beam-melting/. [Accessed 2 September 2016].
There are, however, a number of considerations which are [3] W. Gao, Y. Zhang, D. Ramanujan, K. Ramini, Y. Chen,
raised about the topology optimisation and EBM design C. B. Williams, C. C. W. Wang, Y. C. Shin and S. Zhang,
process. First of all, the compliance of designs can be "Status, challenges, and future of AM in engineering,"
compromised due to a reduction of member sizes, and therefore Elsevier, 2015.
a reduction in the second moment of area which affects the [4] The British Standards Institution 2016, "Additive
bending stiffness of the component. This issue is largely Manufacturing - General Principles - Terminology," BSI
dependent on the initial design space and the choice of material, Standards Limited 2016, London, 2016.
and further indicates the need for a comparative analysis [5] B. Vayre, F. Vignat and F. Villeneuve, "Identification on
between designs optimised under different parameters. The some design key parameters for additive manufacturing:
cost implications of EBM are also significant as it is difficult application on Electron Beam Melting," in Forty Sixth
to justify more than a sevenfold increase in cost in comparison CIRP Conference on Manufacturing Systems 2013,
to CNC machining for low production volumes. Grenoble, 2013.
It is therefore recommended that EBM be limited to high [6] X. Zhao, S. Li, M. Zhang, Y. Liu, T. B. Sercombe, S.
performance applications where the function of a component is Wang, Y. Hao, R. Yang and L. E. Murr, "Comparison of
of greater influence than the manufacturing cost. For example: the Microstructures and mechanical properties of Ti-6Al-
to achieve high performance vehicles for motorsport; to 4V fabricated by SLM and EBM," Materials and Design
minimise fuel consumption over the lifetime of aircraft; or to 95, pp. 21-31, 2016.
reduce cycle times by using optimised tooling solutions. [7] Crucible, "3D printing metal - technical guide," 2012.
[8] B. Leutenecker-Twelsiek, C. Klahn and M. Meboldt,
6. Conclusion "Considering Part Orientation in Design for Additive
Manufacturing," in 26th CIRP Design Conference,
This case study on the design and development of additive Stockholm, 2016.
manufactured components by topology optimisation has [9] ASM International, "Cooling Rates & Microstructure," in
resulted in the following learning outcomes: Casting Design and Performance, Ohio, The Materials
Information Society, 2009, pp. 21-22.
1. EBM can help exploit the benefits of topology [10] M. P. Bendsøe and O. Sigmund, Topology Optimisation:
optimisation to produce functionally superior components Theory, Methods, and Application, Berlin Heidelberg:
owed to almost unlimited geometry capabilities; Springer-Verlag, 2004.
2. There are few design rules to be adhered to when designing [11] X. Huang and Y. M. Xie, Evolutionary topology
for EBM. However, there are a wealth of guidelines which optimisation of continuum structures: methods and
improve the viability of designs depending on the applications, Chicester: Wiley, 2010.
requirements of the customer and the application. [12] J. Karlsson, A. Snis, H. Engqvist and J. Lausmaa,
3. Shape controls during topology optimisation can minimise "Characterization and comparison of materials produced
the requirement of support structures and consequently by EBM of two different Ti-6Al-4V powder fractions,"
reduce time and cost of manufacture. However, the use of Journal of Material Processing Technology 213, pp.
shape controls will impede the optimality of results and the 2109-2118, 2013.
ideal solution must be found by a comparative analysis;
4. A fully parametrised model allows for a simpler redesign
process when applying EBM design rules and guidelines;
5. AM is a costly process which should be applied to
businesses which are driven by high performance. For
example, motorsport, aerospace, or tooling solutions.