ArticlePDF Available

Smart Nodes Pavilion -Bi-Directional Evolutionary Structural Optimization and Additive Manufacturing

Authors:

Abstract and Figures

SmartNodes is a research project which addresses innovation in construction through the application of additive manufacturing (AM) techniques and topology optimisation algorithms used to design efficient and elegant nodal connections of large scale spatial structures. The prototypes discussed in this paper demonstrate a design process and the ambitions of the research group through a scale model of a single layer lattice, and an example full-scale metal node. The focus of this paper is the use of the BESO (Bi-Evolutionary Structural Optimisation) technique to optimise connections in building structures and to ultimately replace welded, forged and cast connections by 3D printed connections.
Content may be subject to copyright.
Proceedings of the International Association for Shell and Spatial Structures (IASS)
Symposium 2015, Amsterdam
Future Visions
17 - 20 August 2015, Amsterdam, The Netherlands
Smart Nodes Pavilion Bi-Directional Evolutionary
Structural Optimization and Additive Manufacturing
James O’DONNELL 1, Hamed SEFI 2, Ben SITLER 3, Nicholas WILLIAMS 4,
Kristof CROLLA, Yi-Min (Mike) XIE2
1, 2 RMIT University, Melbourne, Australia
Nicholas.Williams@rmit.edu.au, s3494880@student.rmit.edu.au, mike.xie@rmit.edu.au
3, 4 Arup, Melbourne, Australia
Ben.Sitler@arup.com, James.ODonnell@arup.com
Abstract
SmartNodes is a research project which addresses innovation in construction through the application
of additive manufacturing (AM) techniques and topology optimisation algorithms used to design
efficient and elegant nodal connections of large scale spatial structures. The prototypes discussed in
this paper demonstrate a design process and the ambitions of the research group through a scale model
of a single layer lattice, and an example full-scale metal node. The focus of this paper is the use of the
BESO (Bi-Evolutionary Structural Optimisation) technique to optimise connections in building
structures and to ultimately replace welded, forged and cast connections by 3D printed connections.
Keywords: Additive manufacturing, Optimisation, 3D Printing, Evolutionary, Connections
1. Introduction
The Smart Nodes project has been a successful collaboration between RMIT, ARUP and LEAD.
Initially conceived in 2013, the project team saw significant potential benefits in using additive
manufacturing to produce geometrically complex connections in large span or pavilion style
structures.
For many years, engineers and architects have worked collaboratively to simplify and rationalize
complex freeform geometry into practical and cost-acceptable solutions. With ingenuity and
innovation, it has been shown that many architectural forms can be found using just a handful of
intersecting geometries such as the torus, sphere and cone. Through geometrical rationalisation, such
forms can make use of simple processes and standard connections which can be manufactured in large
volumes (Whitehead and Peters).
With the invention of additive manufacturing, the manufacturing imperative for repetition has turned.
Instead of requiring the rationalization of geometry to provide repetitive production-simplified
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
connections, we have the ability to 3D print each and every bespoke connection created by a given
freeform geometry. Furthermore, with the use of innovative algorithms to augment a human designer,
we are able to optimize parts for specific load cases and criteria. In this case, Bi-directional
Evolutionary Structural Optimisation techniques have been used to minimise the weight and printing
time required for each connection.
For long span roofs (sports stadium roof for example), some 15-40% of the self-weight is attributed to
the connections alone. The ability to optimize these connections for weight based on the exact load
cases at each connection provides a saving in the overall weight of the structure, leading to a saving in
the material cost of the structure and a saving in the extent of foundations required to support the
structure.
Additive manufacturing technology in metals is novel and relatively new, with further maturation
required for large scale commercial manufacture.
2. Process
The design team collaborated closely to create a workflow that would best suit the additive
manufacturing process to produce the most optimised solution. Since this area of design and research
is in its infancy, the project met with multiple challenges, which have been highlighted where
appropriate in the following stages.
The design process can be summarized in the flow chart below (Error! Reference source not
found.).
The various stages are discussed below with focus on the two stages highlighted in red.
Pavilion Set out and Form Generation
The top four stages in are discussed in detail by a separate paper which has been submitted by our
design team. This paper covers the form generation, structural analysis using Karamba, Optimisation
using millipede and the node manufacture using Stratsys Fortus printers.
Figure 1 - Process flow. This paper covers the red highlighted stages
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Structural analysis using GSA
The first two stages tended to be an iterative process with the initial structural analysis feeding back
into the architectural design. A parametric process was adopted, with all structural properties, such as
boundary conditions, loads, load combinations and member groups for sizing and pattern loading,
defined by the engineers in McNeel Grasshopper. All geometric changes were made in McNeel
Rhinoceros, with each iteration being pushed through as an axis model to GSA (OASYS General
Structural Analysis).
This parametric design process is increasingly adopted on major public structures during the early
schematic phases by leading architect and engineering consultants. Structural engineering firms with
computational design groups will run stress based steel member optimization scripts to automatically
size the members, enabling rapid feedback of the cost, and is typically used in parallel with structural
optimization routines, which iteratively size steel members based on code combined stress and
buckling requirements.
Given the focus on the 3D printed nodes, the design concept for the pavilion was a 1/5 scale bare
sculptural model to be located indoors. As a proof of concept, the loading requirements were
significantly less than that required for a full scale external, public pavilion. For the purpose of the
scale model the design team adopted the self-weight of the pavilion as the only design load, with a
nominal lateral load to prevent collapse during erection and due to accidental contact with the
completed structure.
The GSA nodal forces (Fx, Fy, Fz, Mx, My, Mz) were exported to the design team at each node
location for analysis and optimization of the connections using the BESO process.
BESO process
Structural optimisation seeks to achieve the best performance for a structure whilst satisfying various
constraints[1]. Over the last three decades the availability of high-speed computers and the significant
improvements in algorithms used for design optimisation have transformed the topic of structural
optimisation from the previous narrowness of mostly academic interest to the current stage, where a
growing number of engineers and architects have started to experiment with and benefited from the
Figure 2 - Nodal design zone for incoming member actions
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
optimisation techniques[1]. Among various numerical methods for topology optimisation,
Evolutionary Structural Optimisation (ESO) and Bi-directional Evolutionary Structural Optimisation
(BESO) have been extensively investigated by many researchers around the world. Bi-directional
evolutionary structural optimisation (BESO) method is based on the idea that by gradually removing
inefficient materials and adding material to the most needed locations simultaneously, the structure
evolves towards an optimum[1].
The BESO code which is used in this project is linked to ABAQUS as a structural analysis engine. A
valid structural model in ABAQUS is required to be used as the design domain. The structural solid
element type C3D8R is used. Each node in the Smart Nodes structure is unique and different from the
other nodes as a result of the different number and direction of elements which are being connected
through that node. Therefore the structural model of the node should be compatible with the
geometrical and topological conditions of the node. Obviously, the best starting shape which is
capable of fitting any direction of connected beams is a sphere, but a design domain consisting of
spherical shapes and eye-connection plates which are aligned to the direction of connecting beams is
unnecessarily complicated and can be simplified to a polygonal shape instead (Figure 3).
Figure 3 - Different node shapes as the initial design domain (a) spherical initial design domain
(b) transitional shape for design domain (c) mesh for transitional shape for design domain
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
In the Smart Nodes project, more than 20 different nodes have been designed. The nodes are
optimized by using BESO for real topological and loading conditions which have resulted from
structural analysis of the whole structure as described in the previous section. For the purpose of
illustrating the process, two case studies for a three beam connected node is demonstrated here. The
first case is subjected to two equal bending moments in two branches and fixed in the third branch.
The second case is subjected to two equal vertical loads at the end of the connection plate at two
branches and fixed at the third branch. Figure 4 shows the details of beam to node connections.
The structural model consists of three parts: The central design “space” is connected to three
connection locations each with two plates (labelled PL-A in Figure 4), Part-1-1. The BESO process
will design this entity. Elements which belong to one of the six PL-A’s are defined as NON-DESIGN
and are set to be avoided by the BESO process. The second part consists of six bolts and the third part
is the three PL-B’s (standard cleat/connector plates from beam member). The control parameter is the
maximum volume removal ratio, which is set to 98% to have the widest range of solutions. The
maximum number of iterations is set to 200 which is enough to reach the maximum volume removal
ratio.
Figure 5 shows the boundary and load conditions of case 1. Bending moments in two of the branches
are imposed as two positive and negative pressures which are applied to distinct areas of dimension
10mm x 10mm on PL-B at two branches. The third branch is fixed at the opposite face.
Figure 4 - Details of beam to node connections
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Figure 6 shows the boundary and load conditions of case 2. Vertical loads in two of the branches are
imposed as surface tractions at end face of PL-B on two branches. The third branch is fixed at the
opposite face.
Figure 5 - Boundary and load conditions of case 1, three symmetrical beams with pure bending
are connecting through this node
Figure 6 - Boundary and load conditions of case 2, three symmetrical beams are connecting
through this node, two of them impose pure shear at the end of connection plates with the third
branch fixed
Push/Pull moments in two branches
Third branch fixed
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Figure 7 shows some of the iterations of the BESO procedure for case 1. As it is clear from the
figures, the result consists of three distinct areas of material in each layer and it is approximately
symmetric as the loads and geometry are symmetric. It’s not surprising that the natural load path tends
to follow a triangular truss like geometry.
Figure 8 shows some of the iterations of the BESO process for case 2. The resultant design has formed
shear braces in addition to case 1 in order to resist the shear components of the imposed loads.
Figure 7 Some iterations of BESO procedure for case 1, three symmetrical beams are connecting through this node, two
of them impose pure bending at the end of connection plates and the third one is fixed
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Deciding which iteration is suitable as the final solution is dependent on two factors. The first is that
the stress level of the final solution should not exceed the allowable stress level.
The second factor is that the slenderness of elements should meet both buckling and 3D printing
requirements. Thus selection of a suitable iteration is required to be carried out manually at this stage.
In case 1, iteration 57 is selected as the final solution. In case 2, Iteration 41 is selected as the final
solution. The rough surface of final results of BESO can easily be smoothed through existing
Laplacian smoothing algorithm for printing purposes (Figure 9).
Figure 8 - Some iterations of BESO procedure for case 2, three symmetrical beams are
connecting through this node, two of them impose vertical loads at the end of connection plates
and the third one is fixed
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Figure 10 shows stress contours in structural analysis results for both cases. In both cases, the stress is
distributed uniformly in the designed domain.
Figure 9 - Smoothing by Laplacian algorithm.
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Verification of optimised node geometry
In order to verify the BESO process and the resultant nodal geometry as described in section 2.3, the
design team post-analysed the resultant geometry using Strand7. The two example cases described in
stage 2.3 were imported into Strand7 to confirm the maximum stresses.
Given the nature of the material it was appropriate to use the Von Mises criterion to predict the yield
envelope. The maximum stress criteria was set at 0.9fy for grade 300 steel. In practice, and because of
the finite element analysis process, it is inevitable that discreet localised areas may be subject to
higher stresses. It is the engineer’s judgment to decide what level of overstress should be tolerated. if
any, understanding that peak stresses will be redistributed to other adjacent low stress material.
However, this process would require a more time intensive non-linear material analysis.
Figure 10 - Stress contours in structural analysis results of optimised node under bending and
shear loads.
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
The printing process
Once the final node geometry has been chosen the additive manufacturing process can begin. The
process can be summarised in the following steps:
Preparation of support material
Printing
Post processing of the print
o Removal of support material
o Cleaning/polishing/applying finishes
Testing
Preparation
The preparation of the support material can be time consuming and can cause a digitally optimised
node to be very difficult to physically print if it has not been considered. The printing process as
described in the next section works from top to bottom laying up molten metal. It is capable of
creating extraordinary forms but it cannot defy the laws of physics! It must be laid up on a layer
below. As such, if the optimised form includes negative angles, large irregular openings or hanging
structure, additional support material will be required for a successful print.
Whilst commercially available software exists which is able to include the support material within the
print model, there is still a lot of manual time required from a competently trained AM printer
technician. Minimising support structure speeds up printing and post processing time which can
significantly increase the economy of the solution.
Figure 11 - Von Mises stress contours
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Future Visions
Printing
RMIT University have access to their Advanced Manufacturing Precinct (AMP) which used a
selective laser melting machine to print the metal alloy powders. The process is well documented
elsewhere, but briefly, the metal powder is deposited in fine layers, a high intensity laser selectively
melts the powder which forms the piece being printed, and leaves the rest of the powder untouched.
The process repeats, layer by layer, until the final piece is “printed”.
Post processing
The final piece needs to be cut from the base plate and the support structure needs to be cut and round
smooth. This adds time and cost to the process, and so the optimisation of the support structure is an
important component of achieving a cost effective process.
Further work
The following high level topics have been identified as key items for priority review:
Testing of prints
Multiple load case optimisation
Comparative cost analysis with traditional connections
References
Huang, X. and M. Xie, Evolutionary topology optimization of continuum structures: method and
applications. 2010: John Wiley & Sons.
ASTM F2792 - 12a, Standard Terminology for Additive Manufacturing Technologies,” West
Conshohocken, 2012.
Aremu, A. et al, Suitability of SIMP and BESO Topology Optimization Algorithms for Additive
Manufacture, 21st Annual International Solid Freeform Fabrication Symposium (SFF) An Additive
Manufacturing Conference, Austin, Texas, 679-692. 2010
Galjaard, S., Hoffman, S. Ren, S. (2015): New Opportunities to Optimize Structural Designs in Metal
by Using Additive Manufacturing, in P. Block et al. (eds.), Advances in Architectural Geometry 2014,
Springer, Switzerland, pp. 79-92
Gibson, I., Rosen, D., and Stucker, B. (2010): Chapter 11 - Additive Manufacturing Technologies in
Additive Manufacturing Technologies Rapid Prototyping to Direct Digital Manufacturing, Boston,
MA: Springer US.
Whitehead, H. and Peters, P. (2008): Geometry Form and Complexity, in David Littlefield, Space
Craft: Developments in Architectural Computing London: RIBA.
Zhai, Y., Lados, D., and LaGoy, J. (2014): Additive Manufacturing: Making Imagination the Major
Limitation, JOM, vol. 66, no. 5, pp. 808816.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Additive manufacturing (AM) is expanding the range of designable geometries, but to exploit this evolving design space new methods are required to find optimum solutions. Finite element based topology optimisation (TO) is a powerful method of structural optimization, however the results obtained tend to be dependent on the algorithm used, the algorithm parameters and the finite element mesh. This paper will discuss these issues as it relates to the SIMP and BESO algorithms. An example of the application of topological optimization to the design of improved structures is given.
Book
Additive Manufacturing Technologies: Rapid Prototyping to Direct Digital Manufacturing deals with various aspects of joining materials to form parts. Additive Manufacturing (AM) is an automated technique for direct conversion of 3D CAD data into physical objects using a variety of approaches. Manufacturers have been using these technologies in order to reduce development cycle times and get their products to the market quicker, more cost effectively, and with added value due to the incorporation of customizable features. Realizing the potential of AM applications, a large number of processes have been developed allowing the use of various materials ranging from plastics to metals for product development. Authors Ian Gibson, David W. Rosen and Brent Stucker explain these issues, as well as: Providing a comprehensive overview of AM technologies plus descriptions of support technologies like software systems and post-processing approaches Discussing the wide variety of new and emerging applications like micro-scale AM, medical applications, direct write electronics and Direct Digital Manufacturing of end-use components Introducing systematic solutions for process selection and design for AM Additive Manufacturing Technologies: Rapid Prototyping to Direct Digital Manufacturing is the perfect book for researchers, students, practicing engineers, entrepreneurs, and manufacturing industry professionals interested in additive manufacturing.
  • Y Zhai
  • D Lados
  • J Lagoy
Zhai, Y., Lados, D., and LaGoy, J. (2014): Additive Manufacturing: Making Imagination the Major Limitation, JOM, vol. 66, no. 5, pp. 808-816.