Parametric Structural and beyond Anke Rolvink, Roel van de Straat and Jeroen Coenders

international journal of architectural computing issue 03, volume 08 319 Parametric Structural Design and beyond Anke Rolvink, Roel van de Straat and Jeroen Coenders

Abstract In order to directly make insightful which implications follow from structural design changes and to be able to adapt a structural design quickly to geometrical design changes made by the , the structural engineer may embed a parametric and associative design approach in the structural design process.This approach focuses on parametric modelling and the development of parametric tools which serve specific needs in the structural design process, allowing for instance to quickly communicate and discuss alternatives or to inform design team members of structural results of changing design parameters. The paper presents multiple projects within these categories of parametric approaches.They are concentrated on design and analysis with the goal of presenting practical examples of these approaches in structural design which were integrated in the full design process in order to benefit from the qualities of a multi-disciplinary parametric and associative design process.

320 1. Introduction Design of the built environment requires the collaboration of a team of different roles and disciplines:The client, the architect, the structural engineer, the MEP consultant, etc. However, most parametric and associative design systems and research do not focus on a multi-disciplinary approach, but mainly part of the architectural domain: the geometry. However a parametric approach from the architectural perspective alone does not serve the collaborative possibilities of a parametric and associative design process.Working within a multi-disciplinary design team, the project can benefit if the structural engineer adopts a structural parametric and associative design approach that follows the general design intentions of the project and that provides insight, shows possibilities and presents boundary conditions which have to be taken into account by other members of the design team. This paper presents a number of research and development projects as well as case studies of real and structures, within Arup around the globe which all exemplify the influence of structural parameters in the design process.The experiences by the engineers and computational designers will be discussed alongside some of the technical details of the approaches. The paper will be subdivided in two main components: design projects which are based on parametric modelling and research and development projects that employ and enhance the possibilities of parametric technology. All projects made use of Bentley’s GenerativeComponents [1] and/or McNeel’s Grasshopper plug-in for Rhinoceros [2] as parametric and associative modelling system or as a base for custom tool development.

2. Structural Parametric Modelling Over the past years,Arup has more and more experienced the need to have full control over the complete design to improve the behaviour and performance of the design by design variations and optimisation. Especially in the case of adaptation of the structural design to the geometric architectural design (usually with a complex geometry) the project and designers can greatly benefit from parametric control. If construction models and analysis model have to be remodelled manually, especially in early stages of the design where the architectural design is usually liable to many influential design changes, the parametric approach delivers further benefits of easier change and exploration.The projects below show how parametric modelling approaches are embedded in the structural design process to quickly set up structural alternatives, generate construction models and to reanalyse the structure.

Parametric Structural Design and beyond 321 2.1. Competition entry for the Austrian Pavilion of the EXPO2010 Viennese , SPAN and Zeytinoglu Architects designed a geometrically complex surface model for the competition entry of the Austrian Pavilion for the EXPO2010 in Shanghai, China. For the structural engineers of Arup the main challenge was, next to proving a sound structural concept, to convince the jury of the buildability of the project within a limited timeframe and within a tight budget. Based on initial hand sketches, the structural design was set up parametrically in GenerativeComponents in a way that a number of key structural elements could be analysed individually, but also were associated into an overall parametric model that could be communicated with the architects, Figure 1.

The parametric model proofed its value in allowing for quick design ᭡ Figure 1. Left:The parametric model updates when the geometric surface model was edited by the architects or of the structural elements. Middle:The rendered structural model including when structural alternatives had to be examined. For example for the the profile sections. Right:The design of a cantilevering , a parametric setup was essential in generating architectural image render of the quick construction and analysis models.The truss, with a height of 10m competition entry. Image (c) SPAN and arranged for the 18m long cantilever at the south-west side of the . Zeytinoglu Architects. The complexity in the design of the truss was related to the bad soil conditions, urging the designers to avoid tensile forces in the foundation.As a result, the tuning of the downward counter loads from the first floor and roof structure with the upward loading from the rotational moment of the cantilever was matter of constantly changing the number and location of the floor beams as well as their support locations which determined the floor loads that were transferred to the backside of the truss. Employing this parametric approach to model and analyse structural elements based on structural parameters allowed for a quick setup of the structural design and a proposed building sequence, Figure 2.The main benefit however was that the parametric model could easily demonstrate that a complex could be simplified to a fairly straightforward structure, consisting of mainly standard elements which could be easily assembled, convincing the jury of the buildability of the project within the given boundary conditions.

322 Anke Rolvink, Roel van de Straat and Jeroen Coenders ᭡ Figure 2.The proposed building 2.2. Scheme Design for Coastal Canopies sequence of the main structural elements of the Austrian Pavilion The second project aimed to design a series of coastal canopies with a complex geometry. Having considered typical structural systems for a series of freeform canopies, the limitations (long spans, material constraints, tight budget, ease of construction and specific architectural details) gave rise to a single acceptable solution: a steel structure following a rationalised approximation of the original geometry.Together with the architect a system was set out in Grasshopper to interpret the architectural geometry using simple geometrical surfaces, such as spheres and cones and settled upon a system of interconnected tori, Figure 3. Four patches of four tori would be connected tangentially together, all meeting at a single point, forming the basis of the geometry of each canopy. These tori were generated parametrically so that their base radii, their

᭤ Figure 3. Geometry logic based on four tori

Parametric Structural Design and beyond 323 inclinations and translations relative to global coordinates could all be controlled based on parameters.The idea being that the engineer would define the rule-set and the architect would determine which geometry they preferred based on the rules was agreed upon. The base geometry permitted the development of a parametrically defined structural grid upon the surfaces of the tori, so that the maximum length of any element could be fixed.This created structural elements based on arc geometry, with much repetition in the structural nodes and elements, with only a handful of different node types per canopy, creating a cost-effective solution, Figure 4. Considering that there were upwards of 2000 elements per canopy, this would facilitate prefabrication of the steel arc members. Additionally, the base geometry has been used to generate the cladding panelisation system, which inherits the repetitious quality of the toroidal geometry.This creates a set of panel types that only vary where drainage is

᭢ Figure 4.Arc based structural grid system for one of the canopies

required.These panels also vary at the perimeter of the surfaces where the thickness between the top of bottom cladding surfaces taper to give the illusion of a very slender volume. The final geometry is that of the tori with the original architectural perimeter (in plan) slicing through the base geometry giving the edges a fluid “random” flow, thus presenting to the naked eye what appears to be a freeform surface, but is in fact a highly rationalised surface.

324 Anke Rolvink, Roel van de Straat and Jeroen Coenders 2.3. NSP Arnhem transfer hall The NSP Arnhem transfer hall project, designed by UNStudio includes a large freeform concrete shell with a complex geometry, Figure 5.The geometry of the shell has been defined by the architect in Rhinoceros as two free form surfaces, consisting of NURBS surfaces. However, the architectural geometry was not directly usable for structural analysis , since it only comprised geometrical surfaces and for purposes an analysis model was required.

᭤ Figure 5.The NSP Arnhem transfer hall. Image (c) UNStudio

Since Rhinoceros’ meshing tools did not provide the opportunity to generate a centre or offset mesh following the demands of the team and to directly transfer the mesh data to a FEM software application, a toolbox for the modelling of complex concrete shells from free form surfaces has been developed by the Arup team.The custom developed Grasshopper plug-in supported the engineers in generating FEM models which allowed for the analysis of the design. The workflow of the toolbox asks for single surfaces on which mesh points and face edges are tensioned based on user input, such as the number of elements in U- and V-direction of the surface. Subsequently, the user can for instance pull the vertices to a surfaces edge or any other location on the surface or add or delete vertices and point connections, Figure 6.

Parametric Structural Design and beyond 325 Finally, the generated mesh geometry can be exported to the FE analysis ᭡ Figure 6. Left:The original two software application Infograph [3] via the toolbox’ interoperability interface, surfaces from the architect. Middle: Figure 7.As such, the toolbox provides for the connection between the The meshes modelled on the surfaces. analytical power of the FE analysis software application and the modelling Right: Generation of a centre mesh capabilities of Rhinoceros. from the two meshes constrained to the original surfaces.

᭡ Figure 7. Screenshot from the FEM 2.4.The Kurilpa tensegrity analysis software with the centre mesh of the NSP Arnhem transfer hall Another project which has been based on a parametric and associative modelling approach is the Kurilpa tensegrity bridge in Brisbane,Australia. The competition design brief called for an architecturally striking river crossing to link Brisbane’s central business district with the newly developed arts and cultural precinct on the city’s South Bank and a regenerated and rapidly growing West End.The concept of the design, a multi-mast, cable-stay structure, based on the principles of tensegrity has resulted in a bridge that

326 Anke Rolvink, Roel van de Straat and Jeroen Coenders is both lightweight and incredibly strong. In dimensions the bridge is 470m long with a main span of 120m and features two large viewing and relaxation platforms, two rest areas, and a continuous all-weather canopy for the entire length of the bridge, Figure 8.

᭡ Figure 8.A picture of the Kurilpa tensegrity bridge, taken from the During the design process various tasks have been performed Bicentennial bikeway simultaneously.The use of GenerativeComponents provided the ability to start modelling the complex tensegrity superstructure, even whilst the important bridge centreline geometry was still being finalised.This allowed a compression of the critical path, by enabling simultaneously working on linked design tasks, Figure 9. Different challenges appeared during the design process, such as the site geometry, existing structure and the complex sculptural tensegrity superstructure. Bentley’s MX [4] design package has been used to model site geometry, including the horizontal and vertical alignments, while balancing with functional requirements and property boundaries.After the site geometry was finalised it was imported into GenerativeComponents. Using the software’s flexible, associative modelling technology, key set-out points from the final MX centreline drove the model of the superstructure geometry that had been prepared with project specific components along the centreline driven by those key points.These components had built in ‘solvers’ to meet the design criteria.The final geometry was the resultant of the relationships between the components and the allowed clearances.The geometrical model was subsequently imported from GenerativeComponents into analysis software for structural analysis and optimisation. Bentley’s extraction technology has been used to complete the pre-assembled Bentley Triforma model with documentation and steelwork drawings.The final 3D model has also been used to create a 4D model in [5].

Parametric Structural Design and beyond 327 ᭡ Figure 9.The workflow of information from concept model to construction model

᭢ Figure 10.The parametric setup of the bridge’s superstructure in GenerativeComponents

328 Anke Rolvink, Roel van de Straat and Jeroen Coenders The parametric setup of the project has been proven successful when at the end of the project the steel detailer came up with some issues concerning the connection geometry that resulted as an artefact of the superstructure geometry.The GenerativeComponents model could then be utilised again to quickly resolve the geometry in an aesthetic solution. It involved reversing the orientation of four masts to rectify the detail at the base of the mast to achieve a more aesthetic solution.This actually meant changing the orientation of 17 of the 20 masts due to the arrangement of cables and clearances, but was completed in a matter of hours due to the parametric setup of the project.

2.5. Cable stay bridge option study A parametric approach was adopted for the design option study of a curved highway bridge in the UK.The general form of the bridge was a single pylon and cable plane on the inside of the highway curve.The GenerativeComponents model that was used to parametrically asses the multiple cable arrangements and tower location proposals was set up based on:

• The 3D inroads alignment curve generated by the highway alignment design team. • The existing topography model

These were referenced into GenerativeComponents as spline curves and formed the basis of the parametric model. Pylon and abutment locations were defined by free points along the alignment which allowed the locations to be “dragged” along the alignment and provided a high degree of real-time interaction with the structure. Background mapping was also referenced into the model to assist in locating the structure.Additional free points defined the top of the pylon and upper and lower cable points which allowed the locations to be moved on screen. Graph variables were used to define cable spacing and structural widths. As part of the model, a simplified set of solids were added to indicate the approximate size of the deck, pylon, cables and parapet.At a design workshop the model was then displayed on a large screen and used by the architect and structural engineers to investigate various arrangements in real-time. Multiple views were set up in Bentley’s Microstation to quickly show the anticipated form of the structure from several viewpoints. Key aspects like headroom at abutments and vertical clearances to the river were easily reviewed.Text labels were added to display key dimensions such as span and pylon height in real-time without needing to refer to the dimension tool.

Parametric Structural Design and beyond 329 ᭡ Figure 11.The final result of the This approach enabled the general arrangement to be agreed in a single cable stay bridge option study workshop rather than through prolonged discussions involving the issuing of sketches and responding to review comments.At the end of the workshop a .DGN file of the agreed model was exported from GenerativeComponents and issued to the architect for preparation of the detailed visualisations and rendered images, Figure 11.Additional (hidden) lines were added to the model which were subsequently exported to structural analysis software Oasys GSA [6] to enable the initial loading and design checks.

Once the GenerativeComponents model had been set up, the ᭡ Figure 12.The parametric model of parametric approach led to considerable time savings being achieved in the the cable stay bridge investigation of bridge options, Figure 12.The ability to sit around the table

330 Anke Rolvink, Roel van de Straat and Jeroen Coenders and review options in real-time enabled a holistic approach to be utilised that accommodated the requirements of varying parties at the same time and maximised value for the client.Any ongoing revisions to the highway alignment were simply re-imported into the GenerativeComponents model and a revised .DGN and analysis model generated with a few clicks of the mouse.The fully parametric nature of the model and use of alignment curves as direct references allow the .GCT file to be re-used on similar future projects with only minimal changes required.

3. Structural Parametric Tool Development The previous chapter showed examples of design projects where structural principles were embedded in the design process by creating parametric models.The following three research projects focus on an approach where specific design systems or tools have been build for the design development during early stage structural design and structural analysis.These tools serve specific needs for both engineers as architects, but are generically applicable, mainly in early stages of the multi-disciplinary design process.

3.1. Salamander for Rhinoceros Salamander is a plug-in for Rhinoceros and Grasshopper, designed to add new features to the program in order to help engineers with creating and modifying structural analysis models within Rhino.The plug-in extends the functionality of Rhino with the ability to store structural data (section/material properties, loads etc.) linked to the geometry and manipulate it via a customised interface for viewing and editing this data. The model can then be exported to a finite element package such as GSA for analysis. In other words, Salamander links geometry in the 3D Rhino model to structural data and keeps them in-sync. As an example, points in Rhino represent nodes, lines represent 1D elements and so on. Structural modelling logic is imposed, such that if a node is moved any elements attached to that node will also have their geometry updated.This allows the engineer to edit the model in Rhino as they would in an analysis package, with the added advantage of the availability of Rhino’s geometric toolkit and user-friendly modelling environment. Data can be displayed in a graphical form in real-time in the Rhino viewport where extruded section profiles can be rendered in 3D, points of support and releases can be labelled etc.The structural data can also be browsed through a GSA-like ‘data tree’ and edited via pop-up windows. Salamander also includes a set of tools to eliminate time- consuming and repetitive manual work such as algorithms to align elements to surface normals, set releases between differing sections and tools to check the viability of the data before it is exported for analysis.

Parametric Structural Design and beyond 331 ᭡ Figure 13.Two different 3D Salamander has its own built in manual interface, but it also has the engineering models modelled in capability to be controlled through Grasshopper allowing the structural Salamander.The right model is model to be created parametrically, see Figure 13. Once the model’s nodes parametrically controlled entirely and elements have been created via a Grasshopper component, they remain through Grasshopper bound to the Grasshopper model and will automatically update themselves to match any change, with structural data remaining intact.This makes it possible to change the geometry of the structural model very rapidly merely by adjusting a slider or moving some control geometry, allowing engineers to investigate a range of options very rapidly. Furthermore, the generation of the structural model can be integrated with the generation of the architectural model, meaning that the structural engineer does not have to spend time recreating a structural model from scratch every time the architect makes a small change to the design and the structural implications of these changes can be more quickly and easily understood.

3.2.Tall Building Simulation Tool The Tall Building Simulation Tool is an example of a multi-disciplinary virtual design environment.This project was a collaborative effort between architects, structural engineers, mechanical engineers and cost consultants. The tool provides a dashboard interface for parametric design of standardised cases of high- rise buildings and measures various key design drivers, such as cost, environmental performance, energy, etc, Figure 14.The user can design a building by adjusting a large number of parameters on the system. The tool provides the ability for the multi-disciplinary optioneering of a high-rise building from 15 to 60 storeys.The structural plug-in for this project is based on assumptions and simplified calculations and is essentially capable of modelling standard stability cores with concrete floors and concrete columns or standard stability cores with composite floors, steel beams and steel columns.The tool serves as a tool for early discussions on key design drivers and allows for improved communication between different project participants working cooperatively in an integrated high- rise project.

332 Anke Rolvink, Roel van de Straat and Jeroen Coenders ᭡ Figure 14.The interface of the Tall To improve performances, the design team needs to understand Building Simulation Tool showing the influence of parameter changes to the design constraints and performances. building outline, core structural Furthermore, when more disciplines are involved, knowledge of how design elements and the performance variables and performances interact becomes increasingly important. Parametric studies can be carried out to explore the influence of structural limits, allowing the design team to make better informed decisions about which design performance is governing.

3.3. StructuralComponents Another development towards (parametric) which supports the designing structural engineer is StructuralComponents.This toolbox focuses on employing the parametric and associative approach in the stages of a building; when the design concept of a building is conceived and studied.The current design process incorporated in the toolbox allows the engineer to quickly compose various concepts on a dashboard, resulting in structural design models, which can be judged based on various structural performances.The toolbox allows for concepts to be adjusted and analysed relatively quickly to be able to study the influence of parameter changes and alternative concepts.

Parametric Structural Design and beyond 333 One of the challenges of StructuralComponents is to support and augment the creativity of the engineer.This is especially important during the early design stages since the impact of choices made during these stages is often high and of high influence during the rest of the design process. Little information is known at these stages to base decisions on, making it desirable to access a wide range of options.Another challenge is the unique nature of design projects, since each project brings its own hurdles to overcome.The aim of StructuralComponents is therefore not to present a single solution for design problems or workflow, but to provide the engineer a toolbox that provides parts of solutions that can be easily be composed to a total solution and adapted to the design challenges of each unique situation in design. A prototype of the toolbox has been developed which is based on the structural design of tall buildings.The structural engineer can use the prototype to compose different concepts by interrelating predefined and custom components.These components can be loaded into a parametric software application. Figure 15 shows the interface of StructuralComponents for Grasshopper.A structural model comprises structural components, which are predefined elements, such as cores, outriggers, columns and frames.These elements are pre-programmed blocks ᭢ Figure 15.A model in the current of differential equations and have the ability to perform real-time structural version of StructuralComponents, analysis. modelled in Grasshopper

334 Anke Rolvink, Roel van de Straat and Jeroen Coenders ᭡ Figure 16.A structural model visualised in Rhinoceros The information on the building models performance is presented on a dashboard and visible to the engineer in a single view, Figure 16.These results are given in the form of dials and graphs.When a model is adapted, the output changes accordingly.The graphic output of the structural behaviour of the building structure give a clear impression not only to the structural engineers, but it also allows the architect to get a feeling of the working methodology of the engineers.

4. Discussion The various projects described in this paper showed some of the possibilities for a parametric and associative design approach in the structural engineering practice of Arup. However, it is important to note that adopting a full parametric and associative design approach requires a change in for all the parties involved in the design process. Quick design changes imposed solely on for instance the architectural surface geometry may not be beneficial to the parametric and associative design process when structural, environmental or financial implications cannot be interrelated directly to these design changes.The paper showcases a number of possibilities of which the authors think that they serve the full parametric and associative design process as these parametric approaches are able to follow the iterative process in the early stages of the design and take into account relationships with other disciplines.

5. Conclusion With the growing number of developments in parametric and associative design for structural modelling and analysis, new possibilities arise allowing a

Parametric Structural Design and beyond 335 more process, where amongst others the architect and engineer can communicate via parametric models.Assessing an extensive variety of design options in a short time supports the creative process that is often bounded by time limitations and greatly increases the design flexibility throughout the entire process. The presented projects showed that integrating structural design intelligence based on the parametric and associative design approach enables the engineer to make better-informed decisions and to better communicate them.Additionally, it allows the structural to get in- sync with the constant changing geometric definitions and variable design requirements.

Acknowledgements The authors of the paper would like to thank the many contributors to this paper, who provided information on the various projects; Jan-Peter Koppitz and Kayin Dawoodi (Scheme design for coastal canopies), Christopher Pynn and Ken Enright (Kurilpa tensegrity bridge),Antony Schofield (Cable stay bridge option study), and Paul Jeffries (Salamander).

References 1. Aish, R., Introduction to GenerativeComponents, a parametric and associative design system for , building engineering and digital fabrication, white paper, http://www.bentley.com [15-05-2010]. 2. McNeel, Grasshopper - generative modelling for Rhino, http://www.grasshopper3d.com [15-05-2010]. 3. Infograph, InfoGraph GmbH - Software for structural design, http://www.infograph.eu/ [15-05-2010]. 4. , Bentley Microstation, http://www.bentley.com [15-05-2010]. 5. , Autodesk Navisworks Products, http://www.autodesk.com [15-05-2010]. 6. Oasys, Structural software, http://www.oasys-software.com [15-05-2010].

Anke Rolvink, Roel van de Straat and Jeroen Coenders Arup Netherlands Email: [email protected]; [email protected]; [email protected]

336 Anke Rolvink, Roel van de Straat and Jeroen Coenders