Content uploaded by Salomé Galjaard
Author content
All content in this area was uploaded by Salomé Galjaard on Nov 10, 2015
Content may be subject to copyright.
New Opportunities to Optimize Structural
Designs in Metal by Using Additive Manufacturing
Salomé Galjaard
Arup
Sander Hofman
Arup
Shibo Ren
Arup
Abstract. An initial research has been carried out, exploring the opportunities of
using the design freedom created by Additive Manufacturing in metal on structural
elements. The application of Additive Manufacturing (AM) bears the potential of
increasing efficiency and shortening lead time by reducing processing steps,
material use and labor intensity.
The aim of this research is the exploration of AM technology potentials as a
feasible and robust design and manufacturing solution for structural building
elements. This research is based on the redesign of an existing node, applying the
new production opportunities of AM. This resulted in insights into the different
design steps involved and knowledge on process, costs and structural
characteristics. The start was a topology optimization on an existing design for a
structural node, followed by a further rationalized design for production. A
comparison in aesthetics, structural behavior and costs has been carried out
between the new design and the conventional design. The first results look very
promising and the work will be continued to enable the use of these new
opportunities in structural designs in the near future.
1 Introduction
From 2008 till 2013 Arup was involved in the design of the tensegrity
structures for the Grote Marktstraat in The Hague, the Netherlands, which were to
be used as street lighting. Figure 1 shows a rendering of one of the lighting
elements.
Tensegrity - tensional integrity - is a structural principle based on the use of
isolated components in compression inside a net of continuous tension, in such a
way that the compressed members do not directly connect and the tension cables
define the system spatially.
S. Galjaard, S. Hofman and S. Ren
Figure 1: Rendering of one of the original tensegrity structures in The Hague. This design
will not be built as such. Architect: ELV Architecten © Studio i2
Due to the irregular shape of the structure, the structural nodes, connecting the
cables to the struts within the tensegrity, all have slightly different shapes. There
are 1200 variations in angle and position of the attached cables. Figure 2 shows
one of the nodes from the original design.
Figure 2: Model of the structural node and light fixture, linking six cables to a strut.
Conventional production of the nodes includes high labor intensity as it is
literally a ‘one of’ production a 1200 times over with very little rationalization.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Each node will be made from six to seven unique machined steel plates, welded on
a central tube in varying directions.
Funded by Invest in Arup we were able to explore novel methods of production
in order to research opportunities to produce the nodes in a faster and smarter way.
This research project has been executed separately from the original project and
only after our work on it was finished.
1.1 Additive Manufacturing
The term Additive Manufacturing refers to a whole set of different production
techniques, able to process different groups of materials like plastics, metals and
ceramics. The American Society for Testing and Materials (ASTM) defines it as a
process of joining materials to make objects from 3D model data, usually layer
upon layer, as opposed to subtractive manufacturing methodologies, such as
traditional machining.
During this research project we focused specifically on Selective Laser
Sintering (SLS) for metals, also called Direct Metal Laser Sintering or Selective
Laser Melting. During the production process, powdered metal material is
selectively melted layer by layer via lasers. Layers typically have a thickness of
20-100 µm. The metal is fully melted into a solid homogeneous mass. There is a
wide variety of alloys available and new ones are continuously being developed.
Products produced by SLS can have thin walls, deep cavities or hidden
channels, but sometimes need support structures due to the weight of the solidified
material and heat dissipation. The technique is mostly used in industries that
benefit from weight reduction and where products are complex and produced in
low quantities, like the aviation, aerospace and automotive industry.
2 Method
The set-up of our research consisted of the following three steps:
1. Topology optimization
2. Production of the optimized design
3. Detailing of the original design for comparison
2.1 Topology optimization
With the rapid development of computer aided design technique, topology
optimization as a design approach has been widely implemented to optimize
material layout when designing load-bearing structures. In the classical problem of
a Michell truss [Michell 1904], the optimal design concerns a minimum-weight for
the planar truss was generated for a given load and boundaries. In the earlier 60s, a
modern optimization theory was further developed based on mathematical
programming and sensitivity analysis [Schmit 1960]. This has triggered the
intensive development of various mathematical methods for topology optimization
and numerous commercial products to solve the design problem with remarkable
S. Galjaard, S. Hofman and S. Ren
complexity [Bendsϕe and Sigmund 2004]. By assuming a design domain subject to
a set of boundary condition, the optimal form can be predicted subject to the
prescribed performance targets.
The definition of the design domain is based on the traditional node design,
following a set of simplified starting points regarding loads, material and other
boundary conditions for one of the nodes. The starting points for analysis were as
follows:
1. 3D geometry of the original node.
2. To create an isolated model, limited to the part to which the cables are
connected. The connection to the strut and light fixture and pin & fork of the cable
are not included.
3. The node itself will be connected to a strut on one end and a lighting
fixture will be attached on the other end.
4. The node has six cables attached. Loads per cable have a maximum of
109 kN in tension. Design to load path.
5. The cables have a fork and pin that need enough space to be connected.
This requires the positions of the connection openings to be some distance apart.
6. The diameter and finishing of the node to be similar to the connecting
strut for aesthetical reasons, no design space constraints.
7. Reference material is a suitable steel/stainless steel or aluminum alloy.
8. Minimalize material, weight reduction.
The software package Within Enhance has been used by WithinLab to perform
the topology optimization. It features an optimization engine and an integrated
FEA solver. Based on the given starting points and boundary conditions
WithinLab designed a geometrically optimized node using the state-of-the-art
freedom of AM. Figure 3 shows the original node with the direction of the loads
from the connecting cables.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Figure 3: Direction of loads from connecting cables.
The topology optimization method is used to generate an optimal form. The
initial condition is modified to make the node self-supporting given the
manufacturing process. Based on that, a design space is defined as the maximum
volume within which the optimization can occur by removing material where it is
less necessary.
A single-target optimization process is then performed with a predetermined
target volume percentage and the design objective of minimizing the total material
weight. The Von Mises stress is analyzed during the optimization process as one
of the main design parameters.
2.2 Production of the optimized design
The AM node was produced by CRDM/3D Systems to gain insight in
production requirements and costs. For this, possible improvements of the design
for production are considered in more detail. Issues to consider were:
1. Size of the node in relation to the laser printer compartment.
2. Number of parts to fit in printing chamber.
3. Layer thickness in relation to printer capacity (Wattage) and production
time.
4. Printing direction.
5. Amount of support structure or form changes to prevent support structure.
6. Wall thickness and gaps.
7. Heat dissipation.
8. Availability of printing material (powder).
9. Post-processing, heat treatment, polishing, surface treatment, &c.
Figure 4 illustrates some of the above mentioned issues.
S. Galjaard, S. Hofman and S. Ren
Figure 4: Key aspects related to design for AM production, left: front view of the AM node,
right: back view of the AM node.
Given the printing direction and geometrical definition as indicated in Figure 5,
a hangover analysis has been performed to analyze the self-supporting feature of
the node. A general design rule is that overhangs with an angle less than 45° to the
horizontal need support structure. This can however vary depending on the
production method and machine brand used.
Support structures use up material, increase production time and require
several post-processing steps. Eliminating support structure therefore has a
positive influence on the cost and quality of the end product.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Figure 5: Printing direction and hangover angle.
The product that resulted from the topology optimization has been redesigned
to include the AM features so as to speed up the production process and reduce the
post-processing, such as support removal or machining.
The original node was designed for stainless steel but was in the project for
budget reasons later changed to galvanized S355 steel. In consultation with
CRDM/3D Systems, Maraging steel 1.2709 of ultra-high strength (yield stress of
1000 N/mm2) in fine powder is selected for the AM node.
2.3 Detailing of the original design for comparison
The original design has been further engineered and shop drawings were made
in order to produce it in a traditional way and to verify and compare the structural
characteristics, costs and aesthetics.
3 Results
3.1 Topology optimization
With the given boundary conditions, the optimized design features an efficient
organic form which follows the flow of the internal forces within the structure. It is
noted that the constraints that have been used for the optimization are set at the top
of the node, which is opposite to a realistic situation where the steel node is fixed
with the tube at its bottom. Figure 6 shows several steps in the optimization
process as done with the WithinLab software.
S. Galjaard, S. Hofman and S. Ren
Figure 6: Optimization process with Within Enhance, by WithinLab.
3.2 Production of the optimized design
In order to produce the AM node, the design of the node is further analyzed
and optimized by taking into account the production features of the AM process.
Areas with hangover angles smaller than the selected angle of 45° are
highlighted in red in Figure 7 to evaluate the node’s self-supporting features.
Figure 7: Hangover analysis with in red areas that exceed a 45° angle.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
With this knowledge the node has been further optimized for production. Some
of the changes are only meant as an expample of the possibilities as they would
normally require to be implemented earlier in the design proces. Figure 8 shows
the design after topology optimization (left) and after design for production (right).
Adaptations to the design made after the topology optimization are:
Move holes for connecting cables closer to the center.
Reduce the height of the element from 420 mm to 370 mm.
Change the support condition at the bottom.
Replace the top plate for lighting support with a ring to reduce the weight
and support structure.
Prevent support structure by including results from overhang analysis.
Reshape the constant radius arches to gothic arches and add chamfers to
eliminate the support structure.
Add branch elements near top plate to have the geometry itself become the
support structure.
Further remove material where stress level is low based on the results of
the structural analysis as shown in the next paragraph.
Replace flat and horizontal members to reduce the support structure
needed.
Figure 8: The optimized node after topology optimization (left) and the further optimized-
for-production model (right).
S. Galjaard, S. Hofman and S. Ren
3.3 Detailing of the original design for comparison
Both the traditional node and the node designed for AM have been produced on a
40% scale due to financial and production considerations. The comparison on
aesthetics, structural behavior and costs is presented in this paragraph and is based
on the original size of the node.
Aesthetics. Figures 9 to 12 show different versions of the original and optimized
node for reasons of comparison.
Figure 9: Comparison of the traditional design and the design after the topology
optimization. Left to right: traditional design, design after the topology optimization,
comparison with two models overlapping at the same position.
Figure 10: Comparison of the traditional and the topology optimized design in the structure.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Figure 11: A model of the traditional node, scaled to 40%, in galvanized steel. It is
produced in the traditional way by cutting and welding (left). A model of the node, scaled to
40%, produced by Additive Manufacturing in Maraging steel (right).
Figure 12: Two close-up images illustrating the complex structure in the middle of the AM
node (left) and showing the integrated support structure of the upper ring (right).
S. Galjaard, S. Hofman and S. Ren
Structural comparison. Both structures are able to take the design loads. Figure 13
shows the calculated stresses in the two designs. The Von Mises stresses are lower
than the yield stress with the exception of local peak values which would be
acceptable given the plastic behavior of the steel and the ultimate tensile stress
(510 MPa) for S355. Note that the design load forms an unrealistic combination
that is in equilibrium only when a shear force at the support at the bottom plate is
accepted. More realistic combinations would likely lead to lower stresses.
Figure 13: Plot of Von Mises Stresses [MPa] in the original design (left) and the design
after topology optimization (right). Highest stresses can be observed in red, lowest in blue.
Cost comparison. The costs of several parameters are estimated for the original
node and the AM node. Figure 14 shows an overview of the costs for the original
node, the AM node when produced in 2-5 years time and the predicted costs for a
further optimized node, based on the lessons learnt during this design process.
The costs are based on the original size of the nodes, not the scaled version.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Figure 14: Cost comparison for traditional fabrication versus AM in the near future, also for
a further optimized node.
The printing costs of the current design for AM are high at this point in time.
This is mainly due to the relatively large size of the elements. Printing with the
currently available machines might take up to 15 days with a 200W laser, printing
in 14 micron thick layers. Next generation lasers will have a much higher wattage
(1000W), printing 10 times faster using fewer layers. Also the size of the building
chambers of the AM machines is increasing. In larger printers, more pieces will fit
the printing chamber further reducing the cost per piece.These bigger and faster
machines can be available after testing of the results as soon as the coming year.
In the figure above a cost comparison is also added for a further optimized
node, on which the research team is working at the moment. The node can be
reduced in size by redesigning it including the connection points to the cables and
light fixture. Also a possible 60% further reduction of material by making use of
the higher strength of the available Maraging steel powder is assumed.
4 Discussion
The topology optimization resulted in an organic form with less material while
the original functions as cable connectors are still ensured. During the process of
the research, we realized that the initial boundary conditions for the topology
optimization of the node restricted a more optimal outcome:
Because of the almost three times higher strength of Maraging steel a
substantial reduction of the material can still be obtained, further
reducing the total weight of the element. This will be in addition to the
15% weight reduction in the current optimization.
The structural comparison between the original node and the AM
node shows us that although both structures can take the design loads,
S. Galjaard, S. Hofman and S. Ren
the AM structure does not yet utilize its ultra-high strength. Some
simple tensile tests on printed test samples from the same material
confirm the potential applicability of this additional strength.
1
Additional testing on the printed steel parts will be necessary to safely
incorporate this production method in the building industry.
The AM design can be optimized to reduce the peak values thus
making more efficient use of the material. Eliminating these stress
concentrations would also be highly beneficial for structures with
fatigue issues.
Lowering the weight would have a significant effect on the forces in
and size of the cables in the tensegrity structure as a whole, attributing
to further reductions.
The loads and the connection of the cables can be optimized to make a
more compact node
Incorporating the connection of the lighting fixture could lead to
additional advantages.
We also expect that a more integrated design approach, in which the
topology optimization and optimization for production are executed in
parallel, will result in a more optimal solution.
The studied node would today not be an economic alternative for the
traditionally machined and welded node. The large size would require other
parameters as maintenance or aesthetical motives to govern, even when printed in
a few years. However for a further optimized version reducing size and material
use, the printing costs may soon be less than the high labor cost of the traditional
node.
Even though there is still a lot of work and testing to be done, we believe the
initial results are very promising. Our work gets close to the essence of form
follows function and besides the different aesthetics, which should be judged
subjectively, the design freedom, material savings, limited storage and transport
costs and possible installation benefits should be enough of a stimulus to continue
the research to see whether we can add Additive Manufacturing to our list of
possible production techniques in the Building Industry.
If we manage to get further proof of the safe applicability of this production
technique, we believe that it could change the way we design. Not only on a
structures level, but for buildings and architecture as a whole. We believe this
technique should be used next to the existing ones that have proven their value
over many decades of construction. But from the moment that we believe Additive
Manufacturing can safely be used, our imagination will be the only limiting factor
in designing for the future.
1
No strength-enhancing post-processing operations were applied to the test samples after printing.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Future steps
Our follow-up research will be two-fold: to process the insights mentioned in
chapter 4 involves a complete re-evaluation of the starting points and redesign of
the node, which will be the next step in our work. We will also try to integrate the
functionality of other products in the tensegrity structure in the node – increasing
the complexity of the node, but limiting the amount of products in the structure
and making installation easier.
The second part of our research will focus on material testing. We are setting
up a complete list of tests, necessary to get the right input and feedback on
reliability and potential limitations of the powdered materials in combination with
the production technique.
Acknowledgements
We are grateful for the support of Arno Held from EOS e-Manufacturing
Solutions in the early stages of the research project. Dr. Siavash H. Mahdavi and
his software company Within have provided invaluable input for the topology
optimization. Daniel Kirk and Simon Hammond have been very supportive during
the AM production phase. We learned a lot from their knowledge and expertise
and hope to continue to work with them in future AM related projects.
References
ASTM International, Designation: F2792-12a, Standard Terminology for Additive
Manufacturing Technologies, page 2.
BENDSΦE, M.P. AND SIGMUND, O. 2004. Topology Optimization; Theory,
methods and Applications. Springer, 2 edition.
Direct Manufacturing Research Center 2013. Thinking ahead the Future of
Additive Manufacturing – Innovation Roadmapping of Required
Advancements, Paderborn.
GALJAARD, S. 2013. Tensegrity Structures with Integrated Street Lighting.
IABSE Future of Design, IABSE UK newsletter, edition 35.
MICHELL, A.G.M. 1904. The limits of economy of material in framed structures,
Phil. Mag. 6, 589-597.
SCHMIT, L.A. 1960. Structural design by systematic synthesis, Proc. 2nd ASCE
Conf. Electronic Computation (ASCE, New York).
S. Galjaard, S. Hofman and S. Ren
SNELSON, K. 2012. The Art of Tensegrity, International Journal of
Space Structures, Volume 27, Number 2 & 3.
STRAUSS, H. 2013. AM Envelope – The potential of Additive Manufacturing
for façade construction. Delft: Delft University of Technology, Faculty of
Architecture, Architectural Engineering + Technology department.
New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing
Authors’ address:
Salomé Galjaard (salome.galjaard@arup.com)
Sander Hofman (sander.hofman@arup.com)
Shibo Ren (shibo.ren@arup.com)
Arup
Naritaweg 118
1043 CA Amsterdam
The Netherlands
