ETAFOR(M) Research Cluster 4, 2016-2017 M.Arch Architectural Design UCL, The Bartlett School of Architecture
RESEARCH CLUSTER 4, GILLES RETSIN, MANUEL JIMENEZ, VICENTE SOLER MetaFor(M):: Vallie Alamanou, Ahmad Eltoutngi, Miguel Garcia, Virginie Guillaume.
The Bartlett School of Architecture UCL
01 INTRODUCTION
05 COMPUTATION
1.1. Overview
5.1. Computation
1.1.1 Project Overview
1.2. Research Context
1.2.1. Questioning 3D Metal printing 1.2.2. Comparison with traditional methods 1.2.3. Proceeding Research and Development 1.2.4. Rethinking discrete assembly
02 DESIGN 2.1. Design I
2.1.1.3D Spatial Lines 2.1.2. Spatial Lines. Combiantorics 2.1.3. Combinatorics 2.1.4. Large Scale Aggregation
2.2. Design II
2.2.1. Design Adaptation to Material Research 2.2.2. Design of two elements 2.2.3. Combinatorics of Face Types 2.2.4. Combinatorics of Elements 2.2.5. Building Blocks
06 PROTOTYPES 6.1. Prototype I
3.1. Fabrication
3.1.1. Material Research 3.1.2. Analog Bending 3.1.3. Digital Bending 3.1.4. Tools Assembly
04 AUTOMATION 4.1. Robot Fabrication
6.1.1. Research Context 6.1.2. Automation of Assembly v.01 6.1.3. Automation of Assembly v.02 6.1.4. Automation of Assembly v.03 6.1.5. End Effector and Set up 6.1.6. Robot Choreography
6.1.1. Design and Fabrication Process
6.2. Prototype II
6.2.1. Design and Fabrication Process
6.3. Prototype III
03 FABRICATION
5.1.1. Stress Analysis 2D 5.1.2. Stress Lines 5.1.3. Stress Analysis 3D 5.1.4. Connections Rules. Blue Line 5.1.5. Catalog of Connections 5.1.6. Computing Behaviours of the Elements I 5.1.7. Connections Rules. Stickers 5.1.8. Computing Behavioursof the Elements II 5.1.9. Catalog 5.1.10. Large Scale Aggregation 5.1.11. Structural Mereology 5.1.12. Structural Combinatiorics 5.1.13. B.E.S.O Optimization 5.1.14. Large Scale Iterations
6.3.1. Design and Fabrication Process
07 ARCHITECTURAL SPECULATION 7.1. Architectural Development
7.1.1. Design Process 7.1.2. Iterarion 01 7.1.3. Iterarion 02 7.1.4. Iterarion 03 7.1.5. Iterarion 04 7.1.6. Iterarion 05 7.1.7. Iterarion 05. Assembly Process
08 PANELLING SYSTEM 8.1. Panelling System
81.1. Tile Design 81.2. Attachment to the Unit. Typologies
01
Introduction
Context and thesis
Overview
Project Overview
The research is based in the automation of steel structures assembly. Since the project is part of a discrete design agenda, it is focused in the assembly of inexpensive, standardized and discrete units. In this case, they units are composed of steel bars of two different cross sections, that are fabricated in a serialised way, with specific bends and holes in limited places. The larger aggregation is organised in a building block logic, with predefined rules and limited connections, where the pieces’ geometry defines the course for the aggregation. The building block generates linear aggregations into surfaces, but can also be combined three dimensionally to form wider volumes. The components’ organisation creates patterns and densities enlightening the structural intensities that can be assembled without the need of managing several unique mass customized pieces. Instead, the elements are created through the addition of several steel bars assembled using automatic riveting with robots.
Steel Rods Cut in 500, 1000 and 1200 mm
Two Sections Flat (16x3mm) and Square (16mm)
Computation Stress Lines Growth following stress lines
The outcome of this approach is an intricate heterogeneous whole, with indeterminate differentiated spaces that can be recombined into a large number of different configurations. The intriguing entity has imprecise boundaries so that it can always be expanded with the addition of more blocks. The aesthetic outcome can be described as a hairy fibre structure with fuzzy borders, a steel cloud, that is characterised at the same time by hierarchy between the elements.
Fully Automated Fiber Steel Structure Assembly Multi-scale aggregation
The design process performs in a unified way towards defined objectives. The elements are a product of design, based on connection possibilities, rotation capabilities and structural characteristics. The way the units aggregate, on the other hand, is partially driven by a process of pre-established rules through algorithmic logic, namely the direction of the unit is oriented respectively with the vectors obtained through the traces of stress cloud, nevertheless, they can be set to meet specific goals, such as creating something functional and spatial.
Hierarchically and Scalable Material Pieces with Different Behaviours On Site Robotically Automated Fiber Steel Structures Computation and Structural Mereology
8
Combinatorial Rules Continuous Blue Line Stacking Stickers
90m3
nufacturer
Discrete Factory
Shipping Transported to the site Where discrete
ehouse
Shipping
Logistics Efficiency
Shipping
facturer
Discrete Factory
ouse
Where discrete elements Automation of the fabrication and assembly are created Assembly of Elements 2 Elements
Automated Riveting Rivet gun
Shipping
Disassemblage and reconfiguration
Shipping for fabrication
ory
srehouse are created
Serialized Bends Lines with 45ยบ bendings
Reassemblage On Site
Fabrica Shipping
Flexible and configurable Architecture
Discrete Factory On Site
anufacturer
On Site
Fabrication and assemblage
Fabrication and assemblage
Disassemblage
Shipping for fabrication Shipping and shipping
Disassemblage and reconfiguration Other Discrete Factories and On Site Fabrication and assemblage
Reversible
Reassembla
Flexible and configu
Robotically Disassembled on Site Taking out the rivets
Reduction of Material Waste
Discrete Factory On Site
ufacturer
Fabrication and assemblage
house
mblage figuration
On Site
Robotically Assembled on Site
Bending andand riveting Fabrication assemblage
elements are created
Reassemblage On Site
Flexible and configurable Architecture Disassemblage and shipping
Reversible
9
Other Discrete Fac
Fabrication an
Research Context
Questioning 3D Metal printing
MetaFor(M) RC4 Bartlett 2017
As the project tries to materialize lines in space, the research comparison needs to be done with other projects that deal with steel structures. Most of their research focuses on 3D printing steel, because steel is a structural material that can be used for bridging great spans, visually light and cost efficient when used in manufacturing. However, the counterpart of this method is that is that is extremely expensive compared to traditional methods, such as hot extruding the steel profiles. More specifically, one of the projects that
On site 3D Wire Bending and Assembly of Discrete Elements
Fast discrete fabrication Discrete design and assembly Riversible
deals with 3D printing steel is MX3D. The project deals with metal 3D printing as a tool to achieve freedom in form design, yet it remains cost and time inefficient, while the result is continuous and non-reversible. On the other hand, another interesting project is Metal Mesh Mould from Gramazio Kohler Architects, which develops a spatial robotic extrusion process, mimicking 3D printing with robotic welding and bending wires. However, the process is still continuous, expensive and not reversible. 10
Metal 3D Printing -Freedom in form design -Very expensive -Continuous fabrication -Non-reversible
MX3D Metal 3D Printing -Freedom in form design - Cost effective -Continuous fabrication on site -Non-reversible
Metal Mesh Mould -2D Wire bending -Very expensive -Fabrication on site -Non-reversible
11
Research Context
Comparison with traditional methods
Founded on topological optimization, MetaForm’s algorithm uses the structural stress data obtained from a force field in the shape of stress lines and drives the generation of different types of patterns through these lines. Therefore, the assembly mimics, in the most efficient way, the concept of organizing matter around these principle stress lines and aggregates the elements around them basing the connections accordingly to its mechanical behaviour. Modernist approaches deal with the concept of space with portico
systems, a concatenation of well-defined elements such as the steel deck or the space frame. However, the outcome of these approaches always results in a closed repetitive system. Compared with the digital syntax of structural mereologies, there is only a limited set of possible relations between the elements, where slabs and columns are the same structural element. This method uses two different types of steel bars achieving an intricate network result, more flexible and volumetric. 12
Steel Deck Serialized, standarized, optimized, portico structure, fast assemblage. Different elements and different joints, simple and repetitive.
Space Frames Serialized, standarized, optimized, portico structure, expensive drawings. Different elements and different joints but still very repetitive.
Fully Automated Fiber Steel Structure Assembly Serialized, standarized, optimized, volumetric, cheap and fast. Two different types of bars with heterogeneous result, more flexible.
13
Research Context
Proceeding Research and Development
Continuous
In the discussion concerning analogous versus discrete manufacture, we place ourselves towards a discrete paradigm. Continuous systems seem to face several challenges and show limited performance. The main problems are identified in the speed of the total fabrication process, the constraint of using multiple materials simultaneously, the nonreversibility and limited adaptation capability of what is manufactured. On the other hand, digital materials are units “assembled from a discrete set of parts, reversibly joined in a discrete set of relative positions and orientations” (Gerschenfeld et al 2015). This way, not only the problems presented are overcome, but most importantly robotic assembly of these parts is simplified and much faster as the parts in fact are indicators of the assembly steps and process. The unit defines a simple system by being serially repeated and intelligently combined with other elements and algorithmically assembled, resulting in heterogeneous and differentiated morphologies.
Problems in material transitions Slow fabrication and computation Less adaptive Non-reversible
MX3D Bridge MX3D 2015 Continuous Slow process In place fabrication
Bearing these in mind, combined with the need for scaling up to manufacturing architectural elements, the focus of the project is directed towards the use of steel. Comparing other projects, our focus is directed towards a fully automated assembly of steel structures.
Discrete Wires
“WireVoxels”, “SpaceStream”, “The Clouds of Venice”, “Mesh Mould” are a few research projects that have used similar material and fabrication processes as the ones Metaform is interested in. Specifically, last year’s RC4 project “WireVoxels” automates the bending of the steel bars, while the rest of the assembly is done manually, something that can be a major drawback in the effort of scaling up. “SpaceStream”, another Bartlett project from 2014, that used steel wire for defining architectural elements and space, used fully customized and handmade process, making it difficult to manufacture a larger structure in an efficient way. “The Clouds of Venice” shows an automated way of bending steel and constructing a whole installation of several customized parts in an effective way, however the result still looks homogeneous. “Mesh Mould”, on the other hand, is structural and efficient in the way of making, as it tries to resemble 3D printing with bended wires, but the process is continuous and the result homogeneous, while the process can only be used for fabricating surfaces.
Mesh Mould Metal Gramazio and Kohler 2017 Homogeneous Used for reinforcement In place Robotic Bending
Discrete Simple serialized assembled units Fast aggregations of complex structures Higher resolution Reversible
14
Analog Fabrication
Space Stream RC6 Bartlett 2015 Handcraft Casting Molds Different Form Languages
MetaFor(M) RC4 Bartlett 2017
Interlace WireVoxels.RC4 Bartlett 2016
Hierarchically and Scalable Material Pieces with Different Behaviours On Site Robotically Automated Fiber Steel Structures Computation and Structural Mereology
Welding Assembly 2D Robotic Bending Voxelisation and Stress analysis computation
Venice Cloud Supermanouvre 2012 Highly Customized Homogeneous Result 3D Robotic Bending
Automated Fabrication
15
Research Context Rethinking discrete assembly
Material Manufacturer Material Manufacturer Material Manufacturer Wire Warehouse
Discrete Factory Factory Factory Discrete
On Site
On SiteOn Site
Where discrete elements are created Whereelements discrete elements are created Where are created
WireWarehouse Warehouse Wire
Fabrication and assemblage Fabrication and assemblage Fabrication and assemblage
Shipping Shipping
Shipping Shipping
Shipping
Shipping
Shipping for fabrication Shipping for fabrication
Disassemblage Disassemblage and reconfiguration and reconfiguration
Reassemblage On Site Reassemblage On Site
Flexible and configurable Architecture Flexible and configurable Architecture
Discrete Factory On Site Discrete Factory On Site Fabrication and assemblage
Material Manufacturer Material Manufacturer Wire Warehouse
Fabrication and assemblageDisassemblage
Wire Warehouse
Shipping for fabrication
Reassemblage On Site Flexible and configurable Other Discrete Factories andArchitecture On Site Other Discrete Factories and On Site
and reconfiguration
Disassemblage Disassemblage and shipping and shipping
Fabrication and assemblage Fabrication and assemblage
Reversible Reversible
Material Manufacturer Wire Warehouse
Discrete Factories On Site Fabrication and assemblage
Disassemblage and Shipping Reversible
The project challenges the production chain process by removing the intermediate agent, which is the factory. Factories are where the elements are created, however, in this case prefabricating the element in a controlled environment is not efficient since it reduces the final volume that can be transported and it becomes complex in terms of logistic assembly. Moving the factory directly to the site, fabrication and assembly becomes the same process. The manufacturer ships the material directly to the site. This increases the
volume of transportation up to 9 times, being the most efficient way to deal at the same time with shipping, logistics and fabrication. Theoretically speaking, 450,000 bars could fit inside a 90 m3 truck and could generate 9 volumetric aggregations that fit in a bounding box of 18 x 14 x 6 m. On the other hand, by prefabricating the elements, it would be possible to transport enough material to generate 40% of the volumetric aggregation only. 16
Other Discrete Factories On Site Fabrication and assemblage
1 bar = 0.0002 m3
1 element = 0.02 m3 90m3
=
Truck
Volume of 90 m3
10368 m3
Maison Domino 1152 m3
5 bars Raw Material
=
50000 bars Raw Material
=
1 Element Assemblage of Bars
10000 Elements Assemblage of Bars
90m3
9 Volumetric Agreggations 9 x (18 x 14 x 6 m)
=
Volumetric SpaceFrame (18 x 14 x 6 m)
90m3
etiS nO
Truck 90m Volume of Transportation 3
450 000 Bars Raw Material
90 000 Elements Assemblage of Bars
yrotcaF etercsiD
egalbmessa dna noitacirbaF
9 Volumetric Agreggations 9 x (18 x 14 x 6 m) gnippihS
egalbmessasiD noitarugfinocer dna
etiS nO egalbmessaeR
90m
erutcetihcrA elbarugfinoc dna elbixelF
3
etiS nO
egalbmessa dna noitacirbaF
Truck (90m3) Volume of Transportation
detaerc era stnemele etercsid erehW
noitacir
yrotcaF etercsiD deOtayerro ctecraFstenteemreclseiD etercsid erehW etiS n egalbmessa dna noitacirbaF
4500 Elements Assemblage of bars
0.4 Volumetric Agreggations gnippihS 0.9 x (18 x 14 x 6 m) etiS nO dna seirotcaF etercsiD rehtO egalbmessa dna noitacirbaF
egalbmessasiD gnippihs dna el
etiS nO egalbmessaeR
erutcetihcrA elbarugfinoc dna elbixelF
egalbmessasiD noitarugfinocer dna
17
etiS nO yrotcaF etercsiD egalbmessa dna noitacirbaF
noitacir
Research Context
):
Rethinking discrete assembly
1. Transport
Shipping the equipment and material to the site
2. Bending
The Robot bends the bars in specific angles
3. Assembly
The robot rivets the bars into elements and later into bigger aggregations
Fully Automated Fiber Steel Structure Assembly
18
The truck carries all the material and the equipment necessary for the fabrication and assembly (robots with specific end effectors). The first process, the bending, starts after unloading the material and setting the robots. The robot bends the bars in specific angles. Following this process, another robot or potentially the same, assembles the bars into elements, and later into larger aggregations. Finally these larger aggregations are assembled together resulting in a large fiber steel structure assembly. 19
02
Design
Initial Approaches
Design
3D Spatial Lines
Type 1 Straight Line
Type 2 Diagonal Line
Type 3 L-Shape Line
Piece 1 Straight Unit
Piece 2 Diagonal Unit
Piece 3 L-Shape Unit
The initialisation of the design starts with the generation of spatial lines, meaning that the 45 degree angle is created in a volumetric space, shifting directions in different planes.
volumes with particular behaviours. These volumes are contained in a grid of bounding boxes. The three elements respond to the change of direction needed for creating a heterogeneous aggregation, namely one straight, one L-shape and one diagonal.
The project, in this stage, makes use of three different steel rods with round profiles, one straight and the other two bent, in one place and in two places respectively, always in 45ยบ angle. These rods are assembled in a specific way to create three different 22
+ Same face connection and orientation Type 1
2 Type 1 Possible overlapping arrangement
+ Same face connection and different orientations 2 Type 2
Type 2
Possible overlapping arrangement
23
Design
Spatial Lines. Combiantorics
Face Type 2 3 Point Connections Same face connection and different orientations
Face Type 1 2 Point Connections
Face Type 1 2 Point Connections
Face Type 1 2 Point Connections
Same face connection and orientation
Same face connection and orientation
Same face connection and orientation
Piece 1 Straight Unit
Piece 2 Diagonal Unit
For the connections between units, the faces of the bounding box are used as a guide for establishing point to point connections. The orientation of the pieces is based on the connections possibilities, creating heterogeneous aggregations in each step. This way, the connections are limited and controlled, leading to a discrete set of rules for connection between the three units. This is how the combinatorial logic is established, which is implemented in order for the aggregation process to begin. 24
Piece 3 L-Shape Unit
Piece 1 Straight Unit
Piece 1 Straight Unit
Piece 1 Straight Unit
Piece 1 Straight Unit
Piece 1 Straight Unit
25
Design
Spatial Lines.Combinatorics (Rules)
1 connection
2 connections
rotation for connection
2 connections
2 connections
Elongation
Elongation
26
The establishment of the combinatorial logic is based on the point to point connections between units. In order for the assembly to grow and be structural, only two or three point connections are allowed. In order to achieve this, the units are rotated in the three axis accordingly, so that the rules are met every time. Whenever there are loose ends, the length of the line designed is elongated so that it makes a more stable and continuous whole. 27
The approach for the architectural scale, is directed towards the creation of larger aggregations of the units, that later can form even larger structures and be considered as building blocks. Instead of starting with a simple two-floor slab structure, a more complex, multi-layered building system is developed. This building system is non-typological: it is a mere set of part to whole relations, that can be deployed into different buildings.
Design
Large Scale Aggregation
Design
Design Adaptation to Material Research
Square Section 16 x 16 x 2mm 2 Type of Bends
a. Line 2b b. Line 2b’ 2 x 1000
45º
Flat Section 16 x 3mm 4 Types of Bends
a. Line 1 b. Line 2a c. Line 2a’ d. Line 3 1 x 500 2 x 1000 1 x 1300
Round Section Multiple Possible Connections
Line 1 12 x 3 x 400 mm
Square/Flat Section 4 Discrete Connections
Line 2a 12 x 3 x 1000 mm
Bending 6 Types of Bends
Line 2b 12 x 12 x 1000 mm
As mentioned before, the aggregations are defined by building blocks. These metal tiles are composed by round section steel bars. However, the connections through overlapping become almost countless, making the connection logic not discrete.
three different lengths, the result is five different bar typologies, bent in specific places in 45º angle in 2D plane, leaving the option of only four possible connections. This way discreteness is achieved not only in a design level, but also material wise.
As a second iteration, taking into account the fabrication feedback, the design is shifted to the use of two specific cross sections, one hollow square bar (16 x 1.5mm) and the other one flat (16 x 3mm). By having
With these bars, two units are created. The first one, a straight one, is used for elongation and surface behaviour, while the other one, an L-shape unit, is used for changing direction in 3D space, in the three axes. 30
Line 3 12 x 3 x 1200 mm
Conector
45º
Conector
45º
Conector
Conector
45º Conector 45º Conector
Conector
Line 1 12 x 3 x 400 mm
45 º Bending 2 x 45º Bends
Connections 3 Holes for riveting
Line 2b 12 x 12 x 1000 mm
45 º Bending 2 x 45º Bends
Connections 5 Holes for riveting
Conector
45º
Conector
Conector Conector
45º
45º
45º
Line 2a’ 12 x 3 x 500 mm
45 º Bending 2 x 45º Bends
Conector
Conector
Connections 3 Holes for riveting
Line 2a 12 x 3 x 500 mm
45 º Bending 2 x 45º Bends
Conector
Conector
45º 45º
Conector Conector Conector
Line 3 12 x 3 x 1200 mm
45 º Bending 2 x 45º Bends
31
Connections 5 Holes for riveting
Connections 3 Holes for riveting
Design
Design of two elements
Type 1
Type 2
2 Type 2b
2 Type 3
Face Combinatorics
32
Type 1
Type 2
2 Type 1
2 Type 2b
1 Type 2a’
Face Combinatorics
33
Design
Building Blocks
Top Part Aggregation
Lower Part Aggregation
34
35
03
Fabrication
Material Research
Fabrication Material Research
As the round section steel rods, allow for limitless connection possibilities between the units, square profile rods are used, with two different profiles, one square and one flat, so that the connections between them are limited and controlled, allowing for a discrete set of connections. As described before in the design part, the two units created, the straight and the L-shape, are generated by the two different types of rods, cut in four different lengths and bent in five different positions, in 45-degree angle. The restriction caused by the shift from the circular to the square and rectangular section allowed us to resolve the issue of linear, non-volumetric and homogeneous circular rods’ designs.
Apart from the fabrication flexibility that the shift to the flat and square profile bars offer to the project, design wise there are different visual readings based on the orientation of the bars in space in relationship to the viewers. By combining these two thicknesses together, we could break out of the homogeneity trap of steel structures which gives the system the ability to form complex and more articulated geometries based on the combinations. Moreover, following the idea of digital materials, these two units have limited ways to connect through the designed holes impeded into the steel bars. With the same units, we were able to use the space efficiently, create more volumetric aggregations, and generate higher resolution spaces. 38
39
Fabrication Analog Bending
Some general outcomes can be reached from the fabrication research. Thicker material is preferred and slower forming speed, as the amount of springback is less, but always accordingly to the design limitations. The grain direction is also taken into account for the metal rolling process. As far as the friction is concerned, during the bending process, steel rods are forced between the lower die section and the forming punch, so if the clearance between these two sections is less than the metal thickness, intense friction is created. 40
41
Fabrication Digital Bending
42
The bending machine is characterised by one block body and specific mechanical characteristics. The bending disk can be rotated in two torques, clockwise and anti-clockwise. By using the bending machine, optimal performance can be achieved, while at the same time utilizing low power. In order to create the aluminium piece, a 3D printed mock-up was first designed and fabricated, where the tool set for stirrup bending speed variator, double foot 43
pedal, selector panel and special tooling for spirals can be identified. As for the joint system, we are using the riveting system which results into a seamless connection yet highly structural that would allow us shift to the architectural scale. It is highly cheap and strong connection technique which in on one hand cuts down the cost of the system and on the other hand allows the system be reversible.
Fabrication
Tools for Assembly
Bending Machine
Air Pressure Polisher
Air Pressure Rivet Gun
44
Rivets 4.8 * 25mm
Central Punch
45
06
Automation
Robot Fabrication
Automation Research Context
Analog Fabrication
Nowadays, a few architects are involving robotic arms in the fabrication process, along with computational methods trying to achieve heterogeneous spaces, structurally differentiated, and yet highly efficient in using steel as a material. For example, an international architectural and innovation practice called Supermanoeuvre, worked on the International Venice Architecture Biennale in 2012, on a project called “Clouds of Venice.” This project combines robotic fabrication with algorithmic design strategies. It challenges the cost of construction, but still adds some spatial quality to the architecture through robotic fabrication. The final installation was made of over a 1,000 distinct parts that are digitally generated then robotically fabricated by developing a customized end effector. Then these pieces are “welded together manually to generate a new spatial experience of highly diffused and gradient spatial readings” (Supermanoeuvre 2012). The use of the robotic arm to generate the steel bars speeded up the process and added another level of accuracy to the project. However, from an aesthetic point of view, the installation still looks quite homogeneous. Additionally, it is not reversible, since the parts are customized and fit in a specific place, which is a goal our project is trying to achieve. Also, the whole structure was assembled by hand, which is not efficient in logistics throughout the assembly process.
Space Stream RC6 Bartlett 2015
Interlace WireVoxels.RC4 Bartlett 2016
Venice Cloud Supermanouvre 2012
Bendilicious Maria Smigielska 2016
Another example of robotic fabrication project is Mesh Mould Metal (research project in ETH Zurich), by Gramazio Kohler Research. The research investigates the generation of a single robotically fabricated metal mesh that combines formwork and reinforcement. “The Mesh Mould is developed to be a fully loadbearing construction system by automating the process of bending and welding 3mm steel wires” (Gramazio Kohler Research 2017). This was achieved by a sophisticated end-effector, automating both the bending and welding processes. However, the final outcome could be characterised as a repetitive “homogeneous” mesh of extruded layered metal wires. Moreover, the concrete was applied by hand and it was not part of the construction process which defeats the idea of automating the construction. On the other hand, the reversibility and the scalability of the system was called into question, since welding is used. The idea of combining both the bending and the joining of the steel in one process is the breakthrough of this research.
Mesh Mould Metal Gramazio and Kohler 2017
Automated Fabrication
48
Serialized Discrete Parts
Robotic Assembly
MetaForM RC4 Bartlett 2017
Hetrogeneous Aggregation
To sum up, both previous projects were successful in using industrial robots as part of the fabrication and/ or construction process to reduce the cost, increase accuracy and speed up the process. They used discrete circular steel rods as the design element, which is very hard to control while connecting, due to the unlimited face connections. Moreover, welding was used to connect the steel rods together which 49
makes both systems irreversible, yet it reduces the cost of the expensive customized joints that space frames use. However, we cannot categorize any of the projects as a digital material system. The circular profile of the steel bars allows for unlimited possible connections between the elements; hence, both projects are completely analogue systems.
Robot Fabrication
Automation of Bending and Assembly v.01
Bending End Effector
Bridging the gap between fabrication and robotic assembly process has always been a key goal in the project. Since we are developing a digital material with a set of repetitive limited connections, the concept of using a robotic arm is highly feasible. Therefore, we started developing an end effector that combines both functions in one process where the robot not only could pick these discrete elements and place it in the right place but also fixes these prefabricated steel bars together. With this method, the process of the digital
Picking and Placing Steel Rods
design is the same as the fabrication system through these discrete digital steel bars. Thus, we will be able to automate the entire process which will speed up the construction time and decreases the human labor required on site. In smaller scale prototypes, we prefabricate steel bars since the cost is not high. The robotic arm picks the bent bars, places and joins them together.
50
Bending Steel Rods
Robot Fabrication Automation of Assembly v.02
Rivet Gun Springs 2 Parallel Holder
Pneumatic Solenoid Valve
Gripper Holder Connector to Robot
We developed an end-effector that is attached to a robotic arm which will allow us to combine both the fabrication and assembly process. Furthermore, due to the four possible connections resulting from the material profile, the robotic part of the fabrication and assembly becomes easier since it only allows for four possible gripping faces. 52
Robot Fabrication Automation of Assembly v.03
Flange Pneumatic Vacuum Generator
Rivet Gun Vacuum Generator Holder
Pneumatic Bulkhead Adapter Suction Cup Holder Suction Cup
Pneumatic Vacuum Generator
The second iteration of the end effector combines two pneumatic systems. The first one is used for picking and placing the bars through a pneumatic vacuum generator that is connected to suction cups. The other pneumatic system is for the automatic riveting gun that is used to fix these bars together. In addition, integrating the pneumatic suction system gave us the
flexibility in terms of integrating and adding different type of materials to our design which is helpful when we start thinking about architectural scale. Thus, it will result in reducing both the time and the money specifically during construction process which will increase the efficiency of the system.
54
Feeder
End Effector
Robotic Arm
Aggregation
Aggregation Base
55
Robot Fabrication End Effector and Set Up
Rivet Gun Holder Suction Cup Holder Pneumatic Vacuum Generator Suction Cup Rivet Gun
Robot ABB 1600
56
57
Robot Fabrication Robot Choreography
The mechanism of the assembly process comes from the combination of the two sections (flat and square). There are two lines of production, the positioning of the square bars is based on a picking and placing procedure, linked to the base, while on the other hand, the flat bars links the square bars together through riveting, being attached to the already placed bars and stiffening the aggregation.
The sequence followed, which is crucial to avoid collision between the robot and the aggregation, is initialised from the top right corner, ending in the bottom left corner of the base.
59
60
61
05
Computation Aggregation Logics
Computation
The MetaForm’s algorithm is running within a preconceived design space, which was previously explored, under some constraints and conditions to generate the space. The results of the simulation are a series of continuous traces that display two sets of structural results, compression and tension. The stress field is represented by a layout of discrete vectors forces. These forces express and represent the flow of the stress through the structure, to generate forms that integrate the structure, that are materially efficient and aesthetically appealing.
Unit Direction
Stress Analysis 2D
To approach this methodology, the first step needed was the discretization of the field, which had to be developed firstly in 2D and then applied in a threedimensional space.
Shift Direction
Unit Directions
Color Gradient
Vertical Growth
Different Scales
In order to form the computational logic, stress analysis is used in a simple slab example, which is closer to a 2D surface, rather than a 3D space. The stress analysis is computed based on variable support and load conditions, generating a diagram of different tensions, directions and colour gradient. The stress field generated is discretized and placed into the grid. The discretization of the tangent vectors are replaced into to its nearest predefined 45º angle. The grid is then used for the pieces’ orientation in space, the colour for vertical growth and the distance of the lines are used for the decision of the different scales.
[5, 8, -10]
Scale x 1
Scale x 0.5
Distance
64
Units
Force Z axis
Force Y, Z axis
Force X, Y, Z axis
1.a [0, 0, -10]
1.b [0, 0, -15]
1.c [0, 0, -20]
2.a [0, -5, -5]
2.b [0, -10, -10]
2.c [0, -20, -20]
3.a [5, 5, -10]
3.b [2, 2, -5]
3.c [2, -4, -5]
65
Computation Stress Analysis 2D
Grid
Max Stress Values Bar
Principal Stress Lines
Aggregation based on the Stress Analysis
Computa tion Stress Analysis 2D
The stress analysis may be used, as it is a logical step towards an outcome that needs to resemble space and be structurally achievable. However, the form generated is not something structural or spatial yet. The actual aggregation of the units happens in a second level of editing, following a bottom-up logic, based on the specific connections between the units that are possible. These logics are interconnected and finally crucial to the forming
of a whole structure, one that follows both local connections and stress lines, where units create an interesting whole by local interrelations, but meet a global purpose set from the designer. However, the actual outcome can be characterised as a more of a 2.5D structure, since it is derived from the analysis of a surface, namely the slab, lacking both in structural integrity and aesthetics complexity, even though the aggregation is to some extent controlled both from the designer and the stress data.
Computation Stress Lines
70
Piece Type
Rotation
x y z
One approach that is used in MetaFor(M)’s algorithm is the aggregation through the stress lines. In a similar way in how topology optimization works, by discretizing the domain of the design in discrete components or finite elements. Proceeding later in the organisation and placement of the material around the lines of high stress values, the algorithm runs around these lines aggregating the elements in the space.
71
The discretized lines are registered in two different types of forces, tension and compression. The strategy applied around compression lines, is to stack the elements in a compact way following as most strict as possible the original trajectory of the traces. In these points, the structure becomes thicker simulating a compact mass to transmit the compression forces through the aggregation.
Computation Stress Analysis 3D
Loads and Supports
Voxelized Space
Continuous Stress Lines
Discretized Vector Field
Adding up the third dimension, the square grid becomes a voxel grid, the grid points become a three -dimensional vector field and the simulation allows to evaluate the stress values in a volume. Discretizing once more the field results in obtaining the tangent vector for directionally in each point of the discrete grid. In this case the starting point is settled as a bounding box with specific conditions (loads, supports and dense areas). A topological optimization analysis is ran resulting in the generation of 3D stress lines,
that are divided and discretized into 45ยบ angle based lines. These angles act as a guide for the orientation. After obtaining the principal lines of stress, these run through the previously optimized spaces, and the field is discretized into the regular grid, the algorithm then proceeds to add a series of serialized and standardized elements through a procedural combinatoric logic with limited connections and specific rules along the curves, producing a structural optimized aggregation.
Computation
Connections Rules. Blue Line
Point 06 (1000,240,0)
Point 08 (540,540,0)
As a third iteration, the project focuses on the geometry of the units and combinatorics. Since the amount of possible connections is countless, due to the limitless points in which each unit can connect to the others, more rules and constrains need to be introduced, so that the project can be characterised as discrete. More specifically, a “Blue line� is established
Point 13 (0,0,0)
Point 03 (0,0,0)
in each unit, namely we individualize one of the thick steel rods in each unit and the point connections are done only through these rods. By this constraint, the possible connections are decreased to a great degree, which makes it easier for the designer to be in charge of the aggregation process.
74
Piece 2 Connection through Flat-Square
Piece 2 Connection through Points
Piece 1 Connection through Flat-Square
Piece 1 Connection through Points
Piece 2 Connection through Points
Piece 2 Connection through Stacking
Straight Element Connections The Blue Line
Piece 1 Connection through Points
L- Shape Element Connections The Blue Line
75
Piece 1 Connection through Stacking
Computation
Catalog of Connections
Connection 01 2 Point connection = Alignment
Connection 02 2 Point connection 180 Y axis
Connection 03 2 Point connection 90 Z axis
Connection 04 1 Point connection 180 X axis
Connection 05 1 Point connection 90 Z axis
Connection 06 1 Point connection 180 X axis + Shifted Z axis
Connection 07 1 Point connection 180 Z axis
Connection 08 1 Point connection 180 X axis & 90 Z axis
Connection 09 1 Point connection 270 Z axis
Connection 10 1 Point connection 180 X axis & 90 Z axis
76
Connection 11 1 Point connection = Aligment
Connection 12 2 Point connection 90 Z axis
Connection 13 1 Point connection Shifted Aligment Z axis
Connection 14 2 Point connection 180 Z axis
Connection 15 1 Point connection =Aligment Z axis
Connection 16 2 Point connection 180 Z axis
Connection 17 1 Point connection 90 Z axis
Connection 18 1 Point connection 180 Z axis
Connection 19 1 Point connection 90 Z axis
Connection 20 2 Point connection 270 Z axis
77
Computation
Catalog of Connections
Connection 01’ 1 Point connection 270 X axis A -C
Connection 07’ 2 Point connection 180 X axis C - C, D - D
Connection 02’ 1 Point connection 90 X axis B-A
Connection 08’ 2 Point connection 180 X axis C - C, D - D
Connection 03’ 2 Point connection 180 Y axis B-A
Connection 09’ 1 Point connection 90 X axis D-B
Connection 4’ 1 Point connection 180 Y axis and 90 Z axis A-A
Connection 10’ 2 Point connection 90 Z axis and 90 X axis B- C
Connection 05’ 2 Point connection 180 Y axisand 90 Z axis B-A
Connection 11’ 2 Point connection 180 X axis C - C, D - D
Connection 06’ 3 Point connection 180 Y axis and 270 Z axis B-B
Connection 12’ 1 Point connection 270 Z axis D -D
78
Connection 13’ 1 Point connection =Aligment
Connection 19’ 2 Point connection 270 X axis
Connection 14’ 2 Point connection = Aligment
Connection 20’ 2 Point connection 270 X axis
Connection 15’ 1 Point connection = Aligment
Connection 21’ 2 Point connection 270 X axis
Connection 16’ 1 Point connection 90 X axis
Connection 22’ 2 Point connection 180 Z axis
Connection 17’ 1 Point connection 90 X axis
Connection 23’ 2 Point connection 180 Y axis
Connection 18’ 1 Point connection 90 X axis
Connection 24’ 2 Point connection 180 Y axis
79
Computation
Computing Behaviours of the Elements
Computation
Connections Rules. Stickers
Point 06 (1000,240,0)
Point 13 (0,0,0)
Point 08 (540,540,0)
Point 03 (0,0,0)
06 05
04
03 09
01
04
02
01
12
02
13
03
05 06
00 11
07 08
Since the aggregations made can be characterised still as quite unorganised, the next step for controlling the behaviour of the aggregations made is to establish the “Stickers” rule. More specifically, the establishment of specific points in each rod and in each unit, where the connections mentioned above take place. By these two constraints, the “Blue Line” and the “Stickers”, the possible connections are decreased. At this stage, the code is informed by the new rules. In the bounding box set by the designer, through
10
every iteration, a new unit is created, either a straight one or an L-shape. When the unit is created, another unit that already exists, is picked randomly from the aggregation formed, to establish where the new unit is going to be connected. After that, a connection point is picked randomly in the existing unit and according to the rules set, the unit generated is connected to that point. However, this is not always possible, as the code keeps running and the aggregation grows, less and less points remain available in each unit aggregated for further connections. That is why, 82
00
every time, before connecting a new unit to an existing one, the algorithm firstly checks if the point in the existing unit selected is empty and available for connections. If it is, then the aggregation proceeds as described. If not, then the new unit is discarded and another unit is generated, picking randomly a connection point in the existing unit selected in the aggregation. The algorithm stops, either when there are no more possible connections or when the units generated reach the counter set by the designer.
Connection 10’’ 1 Point connection =Aligment Square - Flat
Connection 11’’ 2 Point connection 180 Z axis Square - Square
Connection 12’’ 2 Point connection 90 X axis Square - Flat
Connection 13’’ 2 Point connection 270 Z axis Square - Square
Connection 14’’ 2 Point connection 180 X axis Square - Square
Connection 15’’ 1 Point connection 90 Z axis and 90 X axis Square - Square
Connection 16’’ 1 Point connection 270 X axis Square - Flat
Connection 17’’ 2 Point connection 90 X axis 270 Z axisSquare - Square
Connection 18’’ 1 Point connection 90 Z axis Square - Square
Connection 19’’ 2 Point connection 90 X axis 270 Z axis Square - Square
83
Computation
Computing Behaviours of the Elements
numConn = 5 numUnits = 118 rotationDeg = 0.45.90.135.270.360
numConn = 4 numUnits = 50 rotationDeg = 0.90.270.360
numConn = 5 numUnits = 183 rotationDeg = 0.90.270.360
numConn = 3 numUnits = 35 rotationDeg = 0.45.90.135.270.360
numConn = 5 numUnits = 109 rotationDeg = 0.90.270.360
numConn = 5 numUnits = 93 rotationDeg = 0.90
numConn = 3 numUnits = 62 rotationDeg = 0.45.90.135.270.360
numConn = 5 numUnits = 49 rotationDeg = 0.90
numConn = 1 numUnits = 9 rotationDeg = 0.90.270.360
Computation Catalog
86
Column Iteration 01
Column Iteration 02
Column Iteration 03
Slab Iteration 01
Slab Iteration 02
Slab Iteration 03
Stairs Iteration 01
Stairs Iteration 02
Stairs Iteration 03
As the rules are informing the performance of the code, the next computational iteration focuses on the specific rules set in order to achieve specific behaviours such as slabs, columns, steps. A great population of generated aggregations, an “army� of different pieces in varying sizes is created, that can form new aggregatations by their in between combinations, or whose boundary can be the starting point for new aggregations around that core. Each 87
solution produced is evaluated afterwards by the designer based on matters of structural ability, spatial qualities and aesthetics. Based on that, the aggregations are discarded or kept in order to form a catalogue of possible aggregations, larger chunks formed out of the simple units.
Computation
Large Scale Aggregation
In this approach, top down decisions are not used at all. The designer makes decisions about the parameters of the aggregation at an early stage but sets the rules without knowing or predicting the final result, which can be different in each iteration. Through this methodology, there is a great population of generated aggregations, based on the number of units connected, set from the designer. Each solution produced is evaluated by the
designer based on matters of structural ability, spatial qualities and aesthetics. Based on that, the aggregations are discarded or kept in order to form a catalogue of possible aggregations, larger chunks formed out of the simple units. However, while a generative algorithm continues the procedure each time trying to improve the results through the fitness function applied, in the project the results produced are random and assessed 90
each time by the designer afterwards, narrowing each time the final choices to more optimal aggregations as the procedure keeps evolving. The aggregations produced have spatial qualities, but in the methodology followed, the designer’s control over the process is limited, while there is no way to have a clear perception beforehand of how the aggregations will look like or behave.
91
Computation Structural Mereology
The effectiveness of this approach relies in “cracking� the optimal orientation of the pieces per its position in space and neighbouring connections. Eventually, with this method any tile could be used with no need of a preconceived structural behaviour. MetaForm pieces are unstable by themselves, with a very flexible and wobbly behaviour, they are not even self-bearing pieces. However, after assembling some of them together, the whole generated becomes increasingly stiffer.
Therefore, fully geometrically understanding the piece becomes the priority. In the interest of obtaining the optimal orientation in every position, it is important to define the centre of mass for calculating the inertia of the pieces. The inertia is calculated in every orientation of the piece and the higher values are recorded. This analysis will describe the different orientation that the elements need to follow to work in favour of its mechanical behaviour.
92
93
Computation
Structural Combinatorics
L Shape with L Shape Connections
Straight Shape with Straight Shape Connections
L Shape with Straight Connections
The two MetaForm tiles have different behaviours, a direction behaviour that drives the aggregation and an orientation behaviour according to its mechanical properties. For example, the L-shape unit in addition to being used for changing the direction of the aggregation, it is also oriented perpendicular to the stress lines. This creates singularities in which sometimes the orientation must prevail over the directionality, otherwise it will compromise the stability of the 95
aggregation. The singularities are solved by the addition of more pieces to redirect the aggregation again in the stress lines path. Once more, the system works around the stress lines, it takes the discretized lines registered in the two different types of forces and aggregates through stacking for compression and point to point connection for tension. Furthermore, the orientation is taken into consideration, prevailing the strong connections between elements and the optimal orientation of the pieces.
Computation
Structural Combinatorics
Once more, the system works around the stress lines, it takes the discretized lines registered in the two different types of forces and aggregates through stacking for compression and point to point connection for tension. This is the same methodology as the previous approach, however in this case, the orientation is taken into consideration, prevailing the strong connections between elements and the optimal orientation of the pieces. Furthermore, in contrast to the previous approach, another stiffening methodology is applied. Instead of using traditional bracing, one of the two tiles are placed in specific places so that they shew the aggregation and lock possible movements in all directions. Once a whole aggregation is finished, some lateral forces are applied to the system to determine where the bracing in needed. This will create some deflections in the x and y axis that will stablish where the bracing is needed.
96
97
Computation
Structural Combinatorics
Stiff Aggregation. Iteration 01
Stiff Aggregation. Iteration 02
Stiff Aggregation. Iteration 03
Stiff Aggregation. Iteration 04
98
Stiff Aggregation. Wall Test
99
Computation B.E.S.O. Optimization
0% 240 Elements
25% 180 Elements
50% 120 Elements
50% 120 Elements
75% 180 Elements
75 % 180 Elements
Total Length 4410mm
Max. Allowed Disp. 13mm
Model Disp. 8mm
Constantly aggregating elements sometimes leads to redundancy. A system with redundant elements Is clearly not an optimized structure. Consequently, MetaForm develops a third procedure to delete the unnecessary pieces. This operation is based in the bi-directional evolutionary structural optimization (B.E.S.O.). This tool is an important branch of topology optimization, it uses a mesh independent and bidirectional evolutionary solver, which allows both material removal and addition.
The tool is applied in the beam system to check which bars are working under stress. Based on a range of utilization percentage we check which bars have minimum or non-stress. If more than two bars are not working in the system, since each element is composed by five steel bars, the whole element is removed. Finally, the stability of the system is checked one last time.
Computation
Large Scale Iterations
102
06
Prototypes
Material Realization
Prototype I
Design and Fabrication Process
Mild Square Bars length: 6m thickness: 16x5mm weigth: 2.1kg/m The first large prototype focused on highlightening the capability of the units to shift directions and form aggregations in all of the three axes, creating volumetric aggregations in space. Additionally, the need for a building system was expressed through the corner prototype, which can be translated into a slab, a column and generally a structure that changes the direction of the agggregation in architectural space, if aggregated with other pieces with specific behaviour.
Cutting
Drilling
Bending
Assembly
106
Mild Steel Flat Bars length: 3m thickness: 16x5mm weigth: 0.63kg/m
107
Prototype I
Design and Fabrication Process
108
109
All the pieces are firstly fabricated based on the design and then assembled together following a specific sequence. The crucial part for the structural ability is the triangular base in the middle, generated from three L-shape units, holding the whole structure together. It acts as a core, both in the design, but in the physical world as well, a starting point from where the rest of the aggregation expands to the outer borders.
110
111
Prototype II
Design and Fabrication Process
Mild Square Bars length: 6m thickness: 16x5mm weigth: 2.1kg/m The second prototype follows the same strategy but the aim is the fabrication of a structure that could be related and compared with a traditional building system. A truss is designed, based on the discrete combinations the two units allow. Comparing with real life applications of building systems, the system produced through the project’s prototype is more efficient in matters of material used, cost and fabrication time, while at the same time possessing the structural qualities and load bearing characteristics of the similar traditional systems used nowadays. The structural ability of the truss was tested digitally, through stress analysis, which was giving feedback to the design through a continuous feedback loop, leading to the final prototype design and assembly sequence. The prototype was physically tested, after it was fabricated for bearing 240 kg of weight.
Cutting
Drilling
Bending
Assembly
112
Mild Steel Flat Bars length: 2 x 1m thickness: 3 mm weigth: 0.63kg/m
113
114
115
Prototype III (B-Pro Show) Design and Fabrication Process
Mild Square Bars length: 6m thickness: 16x5mm weigth: 2.1kg/m For the prototype built for the B-Pro Show, the logic remains the same, but the units change to be fabricated only from square bars of 16*16mm hollow profile, helping in the structural ability and load bearing characteristics of a heavier whole. The particular prototype consists part of the architectural speculation for a scale-up approach in the design process of the discrete assemblies. It consists part of the main column that supports the whole structure. In this particular moment, the lower part of the column branches out in different directions, due to the flexibility of rotation that the discrete pieces allow, through their combinatorics. This way, the column turns to become a slab and the other way around. The same system is used for the design of steps, columns, corners, slabs, openings, resulting in a similar aesthetic effect on the top plan, as well as in the elevation, forming the bones, the structural system of a larger aggregation with architectural qualities.
Cutting
Drilling
Bending
Assembly
Prototype III (B-Pro Show) Design and Fabrication Process
An important aspect of the Metaform project, that aligns with the idea of discreteness, is the recyclability of pieces or material. Specifically, the second prototype, the truss, was fabricated by reusing 10% of the material used in the first prototype. The reason for the low amount is that the design changed in some technical details, so the new units consist of rods with different lengths than the previous ones. However, in the third prototype, exhibited in the BPro Show, we managed to recycle more than 30%of the material used in the truss by disassembling it and reconfiguring the material for new pieces that form the final aggregation. This way, the projects´ physical realisation becomes efficient in matters of cost and fabrication time.
118
Corner: 22 pieces 0% Recycled pieces
Truss: 31 pieces 10% Recycled pieces
Bpro: 62 pieces 34% Recycled pieces
119
07
Architectural Speculation
Architectural Development Design Process. Iteration 01
Initial Conditions
Stress Lines
Stiffness Diagram
Deflection Diagram
Principal Stress
Von Mises Diagram
The same logic that was followed and explained before in a 2D grid is now applied to a 3D volume in order to extract a more volumetric whole, that expands in space. The guidelines for the aggregation are given through the stress analysis data, but at the same time detailed spatial qualities consist an aim to be accomplished through combinatorics. Through a series of diagrams, the main stress and compression lines are used as a driving force of the design, leading to a general diagram of differentiated spaces. A
simplified version of the stress field is used to define the unit needed, and its direction, orientation and order, in the place that is going to be placed. That way, a general plan of the whole aggregation is created. However, the form is not something structural or spatial yet. The actual aggregation of the units happens in a second level of editing, following a bottom-up logic, based on the specific connections between the units that are possible. These logics are interconnected and finally crucial to the forming of a whole structure, one that 122
follows both local connections and stress lines, where units create an interesting whole by local interrelations, but meet a global purpose set from the designer. The aesthetic outcome is a hairy, fuzzy aggregation of rods, aesthetically complex, but still lacking in structural integrity and hierarchy of space.
123
Architectural Development Iteration 01
124
125
Architectural Development Iteration 02
126
127
Architectural Development Iteration v.01
128
129
Architectural Development Iteration 03
A similar computational logic is followed in this iteration as described before. The difference is that instead of aggregating unit by unit, based on the space diagram that is created through the stress analysis data, the aggregation is performed in a higher level, from larger chunks generated through the code explained before. Thus, the outcome remains hairy and quite messy, but some architectural qualities are evident, for example portico systems, enclosure of space, steps or corners that extend to be a slab or a roof, changing direction in the three axes. 130
131
132
133
Architectural Development Iteration 04
Since in this iteration, a building system can be identified, even though the redundancy of pieces is apparent, a step further is researched, by testing the placement of discrete panels. These panels are designed as surfaces that are attached to the discrete units in specific positions, so by the aggregation of the units, the panels form different patterns as well. These panels give to the whole design a different reading, a more articulated view of the whole design and a rather detailed space, where eventually a slab or a staircase can be easily identified. 134
135
136
137
Architectural Development Iteration 05. Design Process
Apart from the previous iterations, the desired outcome remains an architectural speculation inside a bounding box of 18*14*6m. In order to achieve that, we followed a series of steps, including both top-down and bottom-up decisions, with interrelated feedback loops throughout the process, in different scales and levels of the design. At that point, robotic fabrication constraints play a major role for the actual realisation of such a project, since we are taking into account the maximum volume that can be built in one go with the robot, in order to start dividing the whole bounding box based on that, for matters of efficiency, as described below. The starting point for the articulation of the plan is the vertical circulation, namely the staircase needed to bridge two different levels, in the middle of the plan. Then vertical supports are placed, based on the previous decision when needed. From the fabrication feedback, we know that the largest bounding box we can fill with an aggregation is 12m*2m*1m, due to the range of the robot. Trying to make the whole process more efficient, we try to use voxelised volumes, with the dimensions mentioned, in order to fill the larger slabs in the different levels. A general diagram is created, like a top plan, showing the space articulation and circulation, but also the sequence of the fabrication steps. After that, we zoom in a smaller scale, in a more detailed way in order to fill the voxelised volumes with the actual unit aggregation. Each one of these volumes contains specific data about deflection, tension, compression, direction of the stress flow. Based on the sequence established earlier, the rules set for the connection of the units and the different behaviours and the data from the stress analysis, the code is used, starting from the connection points between the volumes, where stronger connections are needed and proceeding from the inside out, filling the space.
Connection Area
Isolation of the Voxels
Identification of Neighbouring Connections
Connection Area
Connection Area
Connecting Units
Connecting Units
The interlevel feedback loops and multiscale analysis lead to a result that is both intriguing aesthetically but at the same time contains spatial qualities and could be characterised as an architectural space. The unit combinatorics offer the opportunity for a differentiated effect, yet since the whole aggregation is following rules, the aggregation is controlled by the designer as for the general space it is going to fill. The stress analysis that
was rejected before as a generalised way of forming the basis of the design, is used in this iteration as a tool of clarifying which connections between the units are stronger and which ones are weaker. This way there is a categorisation between stronger connections needed for support and weaker ones, used for detailing, ornamentation, transparency or density formation, or even for the enhancement of support in some places.
Connection Area
Connecting Units
Aggregation Following the Lines
As a final step, the Bi-directional Evolutionary Structural Optimization (BESO) is used in the final outcome. In Metaform´s case, BESO is used in the whole final aggregation to check for redundant pieces. By removing redundant pieces, without affecting the structural integrity of the structure, the efficiency of the system is increased, since the material, the weight of the structure and the fabrication time needed are decreased.
Architectural Development Iteration 05. Assembly Process
Deployment of the material and the track
Base, where the chunks will be placed
Bending with the robot
Pick and place of the rods to the temporary base by the robot
The larger aggregation starts to form
Larger aggregation, building chunk
140
The general set up takes into account the procedure needed to move from steel bars to an architectural space, from smaller complex chunks that are robotically assembled. The elements are serialized and standardized then shipped to the site along with the robotic arm, robotic track, bending bed and the end effector. After that, the robot starts with the picking, bending if needed, placing and riveting generating a bigger chunk of units. The
chunks are fabricated to 12 meter in length, 1 meter in depth, and 2 in height. The process of assembling the bars follows specific rules and order to avoid collision between the element itself and the elements and the robot. Consequently, these chunks are lifted by a crane and placed in specific sequence where scaffolding is not required. This results in accomplishing an intriguing whole, minimizing the whole time of the process. The advantages of such approach are time and logistics 141
efficiency, low cost in material and fabrication, resulting into aesthetically various architectural spaces. Considering the reversibility of the system, we can disassemble the building into chunks and reconfigure them, ship them to another site. In addition, we can disassemble all of pieces and reuse the material which will result into reducing the material wastage.
Architectural Development Iteration 05. Design Process II
Slab Voxelisation
Generated stress field through specific loads and supports applied
Discretization of the stress field
Further voxelisation based on the fabrication constraints, in three different types of discrete larger voxels.
Generation of the lines inside the voxels, that guide the aggregation of the units.
The final outcome could be characterised as the bones of the structure, a building system
Architectural Development Iteration 05
The outcome of this process represents a building system, a skeleton that consists of articulation throughout two levels. A large column starts from the ground and branches out in different directions, in order to retain the stability of the whole. Apparently, the structure is not habitable at this point, but shows how discrete design and fabrication could work in a larger scale, in architectural space in a quick and efficiwent way, through robotic automation.
The next step deals with enclosure. The final design outcome changes in the facade to have places to hold glass panels. Metal sheets are also put on the floor , based on the pattern generated in the previous iteration, in order to make the slabs walkable.
144
145
146
147
Architectural Development Iteration 05. Assembly Process II
Set up for the assembly line of 12*2*1m
Assembling inside the voxelised boundng boxes
Generating aggregations in a sequence
The robot generates aggregations inisde the three different voxelised bounding boxes
The aggregation grows inside the predefined bounding boxes
Now the aggregations are ready to be picked and placed by the crane in a specific spot
148
Starting by the base, the crane picks and places the chunks already made by the assembly line with the robot in specific places to form the larger architectural aggregation
At the same time, the robot keeps aggregating and forming the larger chunks, making the whole fabrication line quick and efficient. 149
Architectural Development Iteration 05
150
151
152
153
ACKNOWLEDGEMENTS: We would like to express our gratitude to our tutors Gilles Retsin, Manuel Jimenez Garcia and Vicente Soler (Design Computation Lab, Research Cluster 4 tutors at The Bartlett School of Architecture, UCL) for their advice and support in both the research and design projects throughout the year. Particular gratitude is also due to Sherif Tarabishy and Charitini Skaltsari (MSc-Architectural Computation) for their help on computational strategies and technical matters.
The Bartlett School of Architecture Bartlett Prospective MArch Architectural Design Research Cluster 4
Design Computation Lab UCL Co-Founders| Mollie Claypol Manuel Jimenez Garcia Gilles Retsin Vicente Soler
MetaFor(M) | Vasiliki Alamanou Ahmad Eltoutngi Miguel Garcia Jimenez Virginie Guillaume Tutors | Gilles Retsin Manuel Jimenez Garcia Vicente Soler