INTELLIGENT FABRICATION -DIGITAL BRIDGES

10.24904/footbridge2017.10515 INTELLIGENT FABRICATION - DIGITAL BRIDGES Matthew TAM Louis BERGIS Dragos NAICU Architect Bollinger + Grohmann Vienna, Austria Civil Engineer-Architect Bollinger + Grohmann Paris, France Structural Engineer Bollinger + Grohmann Berlin, Germany [email protected] [email protected] [email protected] Klaas DE RYCKE Adam ORLINSKI Ewa JANKOWSKA Civil engineer-Architect Professor, ENSA-Versailles Partner, Bollinger + Grohmann Managing Partner, Bollinger Grohmann Sarl Paris, France Architect Bollinger + Grohmann Vienna, Austria Architect Bollinger + Grohmann Paris, France [email protected] [email protected] [email protected] Summary Students of the Ecole nationale supérieure d’architecture de Versailles participated in the workshop ‘Intelligent Fabrication – Digital Bridges’ where they investigated bridge designs using parametric simulation tools and numerical fabrication techniques. Working in the environment of CAD software (Rhino), extended with parametric plug-ins (Grasshopper and Karamba) the workshop focused on combining complex parametric modelling techniques with FEM simulation. Students developed an intuitive design approach through the analysis of digital and physical models. This paper explains the processes and outcomes of the two week workshop where three designs were constructed at 1:1 and others at 1:5 and 1:10. The proposals displayed a broad range of design and structural concepts, with some investigating topological and geometric optimization and others, the process of fabrication using 3d printing and laser-cutting. Keywords: Karamba; optimisation; computation; timber; digital fabrication; bridge construction; physical modelling; parametric design; structural design Fig. 1. Bridge designs developed over the two week workshop 1. Introduction The bridge, a structure that spans a physical object, is one of the oldest built typologies in the world. Be it a simple log to cross a river or the ancient Roman aqueducts, they meticulously combine the expertise of architects, engineers, artists and consultants. Over time, several bridge typologies have emerged that conform to their specific structural behaviour. By distinguishing the elements that compose these typologies, and combining them with extensive material experimentation and research as well as current computation methods we are able to discover new and innovative forms. The design process of such a footbridge was the focus of a two week workshop at the Ecole nationale supérieure d’architecture de Versailles where students of the third year structural design course collaborated with engineers, architects and experts. Using parametric tools of Grasshopper and Karamba paired with computation techniques in topological and geometric optimisation, students were able to develop comprehensive bridge proposals constructed at 1:1 scale using timber profiles to demonstrate the research of the two weeks. 2. 2.1 Methodology Study of typologies Students were first introduced to the basic structural systems and types of bridge design: beam, arch, truss, suspension and cable stayed bridge (see Fig. 2). Core to the workshop was to precisely examine the structural properties (supports, compression, tension, bending moments) of these well-known typologies and draw inspiration to develop their own concepts for their designs. It was also important to study the weaknesses in specific bridge designs and what causes them to fail. The task was to design a single timber footbridge that would span ten metres and could accommodate two people at one time. The goal was to establish a clear structural system combined with a compelling design proposal. Fig. 2. Optimisation of a beam on two supports towards compression and tension as a guide for bridge design (Image by Fritz Leonhardt [1]) 2.2 Technology In order to translate their ideas from the notebook to the computer, the students took part in a series of tutorials to fully grasp the principles of parametric design in Rhino and Grasshopper. Karamba [2] was introduced as a tool to analyse the structural performance of their designs and to obtain real time feedback which could be later translated into their designs. It allows for quick simulation of the impact of external loads, support conditions, horizontal and vertical stability, bending moments and the analysis of the structures displacement response. These tools can be paired with optimisation algorithms to assess a structure's viability and efficiency. Real world structures need to satisfy many, often contradicting requirements, and such tools allow informed design decisions to be made which prevent a design from becoming arbitrary. 2.3 Material and Fabrication The students could choose whether to start sketches by text, models or digital modelling. For some, it was interesting to test ideas on physical models and then to transfer this to the computer. For others, the reverse proved more efficient. Pinewood strips in various dimensions (3mm, 6mm and 8mm square profiles) were provided at the beginning in order for students to experiment and conceptualise their bridge schemes. It was vital for the development of the designs that students explored physical study models to grasp the materiality and behaviour. Supplement to the traditional low-tech frame construction of timber bridges, various groups also investigated fabrication using the laser cutter and 3d printer to achieve diverse surface treatments and innovative timber connections between members (see Fig. 3). Fig 3. Standard frame construction, laser cut surfaces being assembled, 3d printed connections 3. Output ‘Intelligent Fabrication – Digital Bridges’ was a two week long intensive workshop with over 140 students from the third year Bachelor's course. Under the direction of Professor Klaas De Rycke, the students investigated bridge designs in collaboration with architects, engineers and fabrication experts using parametric tools such as Grasshopper and the Finite Element plugin Karamba as well as digital fabrication techniques. In the first week, the students split up into groups of four to develop bridge designs through physical model studies and digital 3d modelling in Rhino and Grasshopper. The 35 groups were asked to come up with a digitally designed bridge. Bridge typologies, their design, the possibilities of digital form-finding and fabrication and general structural approaches was presented in an introductory lecture. The students received half day training on the tools, half day of designing and modelling and finally some further lectures by external experts. After the initial week of experimenting with various forms and concepts, it was evident that several of the designs retained their overall structural systems, while others chose to investigate different ideas. The collection of bridge designs were divided into the following categories which either described the bridge type or construction technique: truss bridge, arch bridge, twisted frame, reciprocal (Da Vinci) bridge [3] or a modular system (see Fig. 4). Fig.4. The various different bridge types - truss, arch, twisted frame, reciprocal, modular Three of the bridge designs were selected to be constructed at full scale in the second week, while remaining students further developed their proposals at 1:10 and 1:5 scale using computational tools like genetic algorithms, topological optimisation and node design, paired with 3d printing and lasercut components. The work accumulated with a final exhibition where the three final constructed bridges were tested and proven, and exhibited alongside the scale models, vast array of study models, animations and drawings. 3.1 Full Scale Bridges The three bridges that were finally selected to be built in full scale demonstrated a clear design concept, structural system and constructability (see Fig. 5). They also showcased contrasting design ideas in response to the requirements set out at the beginning. The ‘Triangle Reciprocal Structure’ (see Fig. 12) is a play on the standard Da Vinci bridge using triangular module and with duplication and rotation to create an interesting modular system that is self-supporting. The project ‘Domestical Wildness’ (see Fig. 9) interprets a conventional truss bridge by displacing the diagonal members to create two irregular trusses. Paired with optimisation algorithms, a vast array of design options were generated to meet a specific set of criteria, from which an informed design decision was made. The third bridge ‘Hyperbolic Paraboloid’ (see Fig. 8) applies the geometry as the basic form of the bridge, presenting a unique interpretation of the standard bridge. Fig. 5. Final construction of the three bridges (left); and during the construction phase (right) 3.1.1 Hyperbolic Paraboloid Inspired to create a structure that integrated both convex and concave curves, the geometry of the hyperbolic paraboloid became a source of inspiration. This shape is a double curved, grid shell based on a ruled surface geometry imitating compression and tension forces in a beam on two supports (see Fig. 6). Through experimentation with study models the students were able to establish a construction method by assembling a grid in plan, and then lifting up the edges, thus creating an arch. Each intersection would then be bolted loosely, allowing twisting of the beams during the process of the transforming it to its final shape (see Fig. 8). The bolts would afterwards be tightened and the structure would be kept in the final position by cables connecting the bottom extremities of the bridge and other cables blocking the deformation of the frames along the top curve. Fig. 6. Compression and Tension forces in a simple beam (Image by Fritz Leonhardt [1]) Fig. 7. Analysis in Karamba showing the utilisation (Red -100% to Blue +100%) Fig. 8. Construction detail; Timber connections (left); Completed Structure (right) The students had planned to use a double ruled surface - a hyperbolic paraboloid - for construction issues and deformed it by lifting in order to increase its inertia. In order to stiffen the rhombic cells tessellation, cables were placed for triangulation. Analysis was performed using Karamba with a particular focus on the asymmetry of the hypar in elevation. It was found that the structural performance varied significantly with the geometric parameters of the shape, the height and degree of asymmetry away from the centreline. Through iteration, a successful balance was achieved between design intent and structural response. 3.1.2 Domestical Wildness During the initial experiments using study models and digital modelling, several themes arose: superposition, the frame, and regularity (see Fig. 9). There was a desire to bring complexity to the project, so a frame was built to highlight irregularity through a rational system also providing structural viability. The main load bearing elements of the bridge are two irregular trusses. The diagonals of each truss are composed of two groups of elements which lean towards each other. Their superposition offers a random appearance of the project. With the help of Karamba, the students created a parametric setup to obtain a visually random design but with an optimal structure. The project evolved all week long thanks to the simulations of the bridge deformations and also due to the creation of 1:10 models. The deck of the bridge is independent from the structure of the side trusses and varies its heights at specific points. These factors allowed the students to create a new space easily appropriable for everyone. Fig. 9. Original concepts - bridge profile, lattice frame, fishnet mesh, superposition; final constructed bridge Under the guidance of the instructors, the students developed a process using Karamba paired with Octopus, a multiobjective optimisation tool [4], to optimise their design. The fitness criteria to meet included minimizing the maximum displacement of the structure and the length of timber struts required. Furthermore, the Cross Section Optimisation algorithm embedded in Karamba was also implemented to define cross section sizes for the beam elements. As no single best solution exists when performing such optimisation processes, many solutions which are optimal in one respect are produced and one can decide from the various members of the pareto set produced by the optimisation engine (see Fig. 10). Of particular interest was the students’ desire to pick a solution that had a rather compelling scheme which had two diagonal members released from the top chord. Fig. 10. Optimised Form; Iterations from the Optimisation Process Fig. 11. Joint Detail The final option chosen had a maximum displacement of 2.73 cm, requiring 105.9 m of timber struts. The two side trusses were constructed out of various layers with the diagonals in one direction in the first layer, the bottom and top chords in the second layer and the diagonals in the opposite direction in the third layer (see Fig. 11). An additional layer for the top chord was used to stiffen the entire structure. 3.1.3 Triangle Reciprocal Structure During the first week, experiments around the triangle and the Leonardo da Vinci bridge were made, with different shapes, sizes, thicknesses, and different assemblies such as notches, nails, joints, etc. (see Fig. 12) The purpose of the first bridge was to create two arches composed of isosceles triangles and to mirror them to make the bridge more stable. In order to construct the bridge, the structure was optimised by connecting two reciprocal structures making the bridge more stable and resist horizontal movement. The bridge deck is created by the space between the two triangular structures and it is placed on wooden beams that transfer the loads to the structure (see Fig. 13). Each triangular element rests on the one before and the one after. This disposition generates an arch structure system that transfers loads through interconnected components. A wooden shim is placed to fill the empty space between two triangles. The isosceles triangle elements are made of two four meter beams and a one-meter beam. They are fixed together using a wooden element placed on each corner. Fig. 12. Design process & initial studies of triangulated reciprocal structures The bridge is composed of two arches of five interlocked triangular modules. The principal edges of those modules are composed of 4-meter beams, submitted to bending and compression. They are supported at their ends and held at their centre. The shortest edges of the modules are connecting those beams and generate the bridge curvature and they are submitted to shear force and bending. The bridge was built without scaffolding, by simply lifting it and placing the modules one after another. Fig 13. The structural analysis in Karamba showing the utilisation of elements; Attaching the triangular caps Fig. 14. Constructed Bridge 3.2 Bridges in scale 1:5/1:10 The remaining groups were tasked in the second week to develop their initial bridge concepts with topological and geometrical optimisation, further calculations or detailed construction techniques. They consolidated their ideas within their corresponding topics in 1:5 or 1:10 scale. Some focused on designing unique joint connections between the timber elements with 3d printing, whilst others transformed their originally frame bridge designs into surfaces. Groups that were primarily engaged with modular systems looked towards optimising their structures according to structural performance and following bidirectional evolutionary structural optimisation (BESO) [5] techniques to reduce material. 3.2.1 Design of Joints Groups which had developed truss or rotated frame schemes encountered that although the typologies of such bridges is simple in its form, they often have many members (up to eight members) which intersect at one point. This has very important practical implications to the feasibility of construction. Through study models, students were able to quickly establish difficulties in achieving precise connections when assembled. The triangulation, often with different angles, meant that members would not meet as intended. To overcome this issue, 3d printing in ABS filament was employed to design highly customised joints in essence referring to the techniques nowadays relatively available on the market for steel 3D printing. The groups looked at identifying the main types of joints within the structures and introduced a basic joint typology which moulded according to the different orientations of the beam (see Fig. 15). Others decided to create unique articulated knots that could adapt itself to every case in situations where each joint within the structure was different (see Fig. 16). Fig 15. Joint typology for multiple joint designs Fig 16. Adaptable Joints using one type 3.2.2 Topological Optimisation of Surfaces The initial exercise was to design bridges that were pure framed structures, however it soon became apparent many structures had an inherent surface quality. The structures were reinterpreted as surfaces which could then be analysed in Karamba for their structural performance. The goal was to dissolve the surfaces by creating perforations or removing areas deemed less integral to the overall structure through topological optimisation after the initial geometrical optimisation (Fig. 17). Through applications of patterns and grid divisions the students developed bridge designs that were highly dynamic and reacted to the stresses within the surfaces. Fig. 17. Topological Optimisation in Karamba; Perforation of the surface; Lasercut model 3.2.3 Topological Optimisation of Frame Systems Straying away from the typical arch or truss bridge typologies, several groups looked to creating modular systems that not only generate the deck of the bridge, but also integrate seating and other elements into their designs. This design starts rather from a detail and through additive/subtractive design methods get to a bridge typology. Common issues with such systems are that they demand a high quantity of material and the objective was to explore how material (or modules) could be reduced by analysing structural performance. By running a Bidirectional Evolutionary Structural Optimisation (BESO for Beams) algorithm in Karamba [3], less strained and effective elements are removed to achieve a structurally optimal design while maintaining the overall appearance of the structure (see Fig. 18). Fig. 18. Topological Optimisation through BESO in Karamba; translated into physical models 4. Discussion The overall idea of the workshop environment is to show students that several ways of design approach are possible which all can add value to design. All the design steps of physical modelling, physical construction constraints, and structural aspects, digital modelling or sketching are considered non-hierarchical in this approach and have mutual beneficial aspects and overlaps. 4.1 Collaboration A pillar of good and efficient design in the built environment is the communication and collaboration that brings together various skill sets. Given the group work nature of the workshop as well as the constant presence of experts providing guidance and input, the students were able to design in an environment that aimed to encourage these two aspects. Additionally, the challenge and aim of full scale construction meant that this was actually required of them over the course of the two weeks. The outcome of their work, showcased in the previous sections, stands as proof of this. Furthermore, throughout the duration of the workshop, guests were invited from various disciplines to offer their expertise to the students. This gave insight into different aspects of the design, architecture and engineering industry, and was vital for the students in understanding the current practices and tools. Architects Jacques Anglade and Marie Zawistowski shared their experiences in working with timber in environmentally sensitive projects; Sylvain Usai of Design to Production and Ewa Jankowska both showed innovations in current complex 3d geometry development and CAD planning; while Didier Valroff from Simonin, a manufacturer of glulam wood products, talked about the process of construction and planning of timber projects. 4.2 The Notion of Parametrics One valuable insight with respect to parametric workflows was that, on the whole, the creativity of the students was significantly aided by learning to set up parametric geometric models for the bridge designs. Furthermore, clever use can allow the user to identify what are the driving parameters that most significantly affect a design and what are the limits, the extreme possibilities for a particular design. This quick design evaluate - iterate process, coupled with physical model testing, proved to be a very productive way to push forward towards the final bridge concepts. This is paramount to the current working environment and communication between all parties to understand each other on the same level in a design; every design member needs to understand the rules driving the design. As such, the experiment of getting the parametric approach adapted to a real challenge proved very interesting but not easy. 4.3 Structural Assessment in the Parametric Design Environment Fig.19. Karamba – a basic setup The primary goal of the workshop was to develop schemes that demonstrated a strong understanding of the structural concepts of bridge design. Fully embedded in the parametric environment of Grasshopper, Karamba allows for a quick and intuitive approach for designing and analysing structures, with the intention to bring the architect and the engineer closer together (see Fig. 19). Karamba was developed with the aim to significantly reduce the computational time for optimisation processes of structures. Some of the proposed bridge designs contained over 500 elements and without such tools, one would not be able to realise such concepts in such a short amount of time. These tools present a perfect pedagogical tool given the possibility to explore a large number of structural and topological possibilities. It is important that students see what tools are available to them. Even applications on cell phones permit simple calculation of beams and are usually very visual and intuitive. Having a sense and ability of conceiving structure by using visual techniques with tools such as Karamba, with Graphic Statics [6] (see Fig. 20) or with Push me Pull me [7] and entirely integrating this into a real design issue can outweigh the pure teaching of formulas. Fig.10. Graphic Statics in eQUILIBRIUM by Block Research Group 4.4 Digital Fabrication Students learned and experienced the actual advances in digital manufacturing and the innovation brought forth through expertise and creativity. This design and build approach could alter the current perception of conception of solutions. In essence, fabrication is being brought back to the forefront of the debate and should have an impact on early design decisions. As Violet Le Duc would have probably loved to see is that new materials, fabrication or logics can be used for the approach of the design task. This will inherently affect the design task in itself. 4.5 Student Feedback ‘As students, we aim to experiment with forms and concepts without limits. This was for me the very first studio associating computing and fabrication. The workshop was really challenging in that the footbridge topic was full of constraints that we needed to solve. Additionally we also had to be very pragmatic in our design by taking into account the economic aspect of the project. Limited by the amount of material, money and time, the workflow between digital and physical model allowed us to stay focus on the concrete fabrication process. I personally worked on a 1:5 scale model which was very close to the 1:1 scale. Therefore we needed to carefully detail all the connections and choose the right timber sections without using any glue. ‘ ‘Another interesting aspect of the studio was to meet foreign designers and architects and by working in close collaboration with them we learned how to share our ideas and communicated them by using animated figures of our structural digital model or by exchanging grasshopper scripts and tips. To summarise, I really discovered a domain in which I would like to further in my studies. Digital fabrication is a brand new world that will totally reshape our profession and open new perspectives. Therefore, I think this kind of studio that must be developed in more architectural or engineering schools, to maintain a high degree of innovation from the youngest generation.’ Cyriac Levet, student 5. Conclusion The exploration of a design brief within an intense workshop learning environment is a well-known educational method that has been applied in this context. By carefully orchestrating a steep learning curve not only with respect to software and design tools, but also to structural and design driven concepts, participants gained an immense amount of experience in a short amount of time. The tools and techniques that we have today coupled with the knowledge of an expert team ensured a constant dialogue and input within the students projects. The added challenge of realising full-scale examples provides both a challenge and a reward upon completion. 6. Acknowledgements Bollinger + Grohmann Karamba Ecole nationale supérieure d’architecture de Versailles (ENSA-Versailles) Professor: Klaas de Rycke Instructors: Louis Bergis, Clément Duroselle, Adam Orlinski, Matthew Tam, Thomas Charil, Dragos Naicu, Ewa Jankowska Student input: Cyriac Levet Student Assistants: Stephen Jacquens, Florian Bourguignon Lecturers: Jacques Anglade, Marie et Keith Zawitowsky, Didier Valroff, Sylvain Usai, Ewa Jankowska Students: Third Year Structural Design students of ENSA-Versailles (T31) 7. [1] References Leonhardt F., Vorlesungen über Massivbau, Springer-Verlag, Berlin, 1984 [2] PREISINGER C., “Linking Structure and Parametric Geometry”, Architectural Design, vol. 83, March/April 2013, pp. 110–113 [3] LAW, B., Self Supporting Bridge, http://www.leonardo-da-vinci-models.com/selfsupportingbridge.html, visited April 2017 [4] VIERLINGER R., “A Framework for Flexible Search and Optimization in Parametric Design”, Rethinking Prototyping - Proceedings of the Design Modelling Symposium Berlin, 2013. [5] XIE Y., and STEVEN G., Evolutionary Structural Optimisation, Springer-Verlag, London, June, 1997 [6] BLOCK RESEARCH GROUP, eQUILIBRIUM, http://block.arch.ethz.ch/equilibrium/, visited April 2017 [7] SENATORE G., Push Me Pull Me, http://expeditionworkshed.org/workshed/push-me-pull-me/, visited April 2017