Studio 1 / Deep Skins Faculty / Alexandre Dubor, Aldo Sollazzo IaaC, 19/12/2019
MRAC Master in Robotics and Advanced Construction
Tutors / Alexandre Dubor, Raimund Krenmueller
Localize, Collate and Design Towards Circular Building Environment
Anna Batallé İrem Yağmur Cebeci Matt Gordon Roberto Vargas
Studio 1 / Deep Skins IaaC, 19/12/2019
MRAC Master in Robotics and Advanced Construction
Tutors / Alexandre Dubor, Raimund Krenmueller
Content Framework The need to create a more sustainable and circular building environment
Proposal Research goals through the construction process
Research Definition of the challenges and opportunities and identification of the digital methods that can be applied to enhance existing workflows
Case Study Proof of concept of the new methodology.
Open Up Future Vision of the project
This research aims to understand how the digitalization of the construction sector is going to lead us to a more sustainable building environment. It presents a series of technological advancements that can be applied in the deconstruction, design and fabrication phase to transform and optimize material ows.
FRAMEWORK
Deconstruction
Demolition Building Linear Life Cycle
Optimize the workflow to sort and locate construction material Demolition
Site demolition produces a significant amount of waste. All material that is derived from from demolished sites ends up in landfills, without any attempt to be sorted on reusable and non-reusable pieces which can later be used in new structures.
Resources optimization Construction and demolition waste in the EU accounts for approximately 25% to 30% of all waste generated in the EU. Therefore, responsible management of waste is an essential aspect of sustainable building as well as rationalizing the use of new material. https://ec.europa.eu/environment/waste/construction _demolition.htm
Waste Generation
Deconstruction
Resources
Economic challenge and sustainable impact Carbon Footprint
In construction, four materials are commonly used: concrete, masonry, wood and metal, these will also be the focus throughout this project. The graph below showcases the current life cycle of the materials as well as how much of them are being wasted by land-ďŹ lling or down-cycling.
Material Uniqueness > Added value + 10% to 20%
Manufacturer Manufacture Transport
The aim is to increase the amount of material recycled and reused, which will also lead in decreasing the quantity that is being down-cycled or resulting in the landďŹ lls. Transport
Material Retrieval Raw Material
Waste Management
Raw Material
Recycled
Reused
PROPOSAL
Building Circular Life Cycle
tion
truc
ns eco
Stages
nst
ruc
tion
Pre
-de con
stru
ctio n
D
Co
Service Plan
Services
Demolition Waste
GHG Emission Production
Embodied Energy
Lifetim
e
Structure
Skin
Building Owner
Building Layers Services > 7 - 15 years Space Plan > 3 - 30 years Skin > 20 years Structure > 30 - 300 years
Material Shape and Data
03
Deconstruction Company
tion
Co nst
truc
Material Location and Data
ns eco
ruc
D
tion
Pre
01 | 02
-de con
stru
ctio n
Building Life Cycle Actors Involved Technologies
03 Technologies
02 Machine Learning
e Lifetim
01 Computer Vision
Structure
Design
Service Plan
Designer Architect Material Shape and Data
Services
03 Computational Design 04 Robotics
Skin
Building Owner
Proposed Products
Collaborative Design Tool
Assistive Design Workow
Digital Material Dataset
Designer - Generic User
Architect
Demolition Company
Social Awareness Material Resources
Responsible Design
Resources Management
RESEARCH
Siftsite The production of materials needed to satisfy the demand for new architectures generates a great impact on our environment. Meanwhile, every year resources are lost during building demolitions. Almost thirty percent of the waste generated in the European Union comes from the construction industry.
The term urban mining is born from the perspective of environmental concerns and aims for more eďŹƒcient use of the construction materials available in the shape of building stock. The purpose is to reduce the use of virgin sources, generating savings through improving secondary resource use and lowering the negative environmental impacts.
The sifting of buildings to obtain material is something that has recurrently happened in history but the tools used to perform these tasks haven’t evolved. At SIFTSITE we are developing a new workow to sort and locate material from pre-demolition sites to create a more reliable report of the resources available. http://www.iaacblog.com/programs/digitalizing-mate rial-collation-predomolition-sites-studio-ii/
Digitalize and Automate the process of creating the material dataset
Building Inspection
01
ClassiďŹ cation and Localization
02
Geometry Reconstruction
Material Report
03
04
Building Inspection
01
Manual Drone Flight
Capture Images of the Demolition
Classification and Localization
Brick
Concrete
Metal
Wood
None
Classification and Localization 02
Training Data
Images from the building Inspection
Analyze by Larger Sliding Kernel
Classification and Localization
Material Classification
Classification and Localization 02
Finding Features by Corners
Describing Regions Around Features
Clustering Descriptions into codewords
Comparing Histograms of codewords counts
Geometry Reconstruction
02
Material Localization
03
Dense Point Cloud
Geometry Reconstruction
ClassiďŹ cation and Localization
03
Segmentation
Geometry Extraction
Colored Mesh
ASSISTIVE DESIGN WORKFLOW
Material Language Best Fitting Function for Element based in Material Properties
Adobe wall made of soil + 10% granulated old bricks
Metal Beam
Brick A Transition from grey-orange
Assembly Rules
Construction System
Brick B Grey Brick
Fabrication Methods Concrete Beam
Brick C Concrete Brick Deconstruction System Stone
Design Exploration Case Study
Design Exploration Stacking System
Design Exploration Stacking System
The generation of the design is done through 3 steps. First a simple design space is created allowing to focus the research in the definition of the matching algorithm. The optimization solver takes places in order to find the best solution given various parameters (cost, material waste…) http://www.iaacblog.com/programs/post-collation-design-exploration/
Generative Design
Generat ive Design
Matching Algorithm
Optimization Solver
Design Exploration Stacking System In order to relate the material with the design, we need to build a language that understands the construction constraints. A grid is created and the length of the elements is the one that is taken to process in the matching algorithm. We have the length of the elements coming from the design and the material dataset as inputs for the matching algorithm as well as a tolerance of 10 percent. The outputs that we are getting from the matching algorithm are then used for the optimization and the shape reconstruction of the element. And the matching is working mainly through the lengths and ids of the pieces in design and the dataset. The algorithm is taking the length of the needed piece from the design and it is matching it with the length in the dataset to first use it in the design structure but also to delete the element from the left materials in the dataset. For the ones that is not exactly matching but can fit in the design, the algorithm will cut the batten and place the left part back to the material dataset. For the materials that there was no match, a new element, from outside the material dataset, is going to be used. With the optimization solver we are minimizing the amount of cuts and outer materials. First matching algorithm that we had can be seen below. There was material that it could potential be reused in the material dataset but the design was already asking for new material. We improved the algorithm by allowing to cut material in case none was founded. As we can see, the design asks for new material when the one from the dataset has finished. To improve its performance we could start relating the quality of the material to the structural dependencies. Meaning that new material will be place where higher forces appear and less quality material will be used as a filler.
Design Exploration Stacking System
There were three primary factors that we chose to judge the designs by. Firstly of course we want to make the deepest use of the available material, so we want to reduce the number of unmatched parts in the design. Secondly, we want to require as little additional labor at construction time as possible, so we want to reduce the total amount of material cuts needed. Finally, the construction should be mindful of its carbon impact, so we want to reduce the needed transporation distance for all the included elements. All of these factors are normalized, weighted, and combined as a ďŹ tness function to be minimized.
Design Exploration Stacking System
COLLABORATIVE DESIGN TOOL
Collaborative Design Tool Human-Machine Interaction
Visualization
Hand detection
Remote Controller Research
Ideal Outcome
Collaborative Design Tool Human-Machine Interaction
Collaborative Design Tool Human-Machine Interaction http://www.iaacblog.com/programs/courses/mrac/2019-2020-mrac/h-3-hardware-iii-seminar-mrac-2019-2020-3rt/
Collaborative Design Tool Human-Machine Interaction
CASE STUDY
Case Study
Material Dataset
Potential Process Transformation Through Digitalisation
SITE I DECONSTRUCTION SITE
Iaac IAAC Pujades 102
Inspection, Organization and, Fabrication Automation
Design Strategy
Deconstructing wood framing
SITE III CONSTRUCTION
Plug in Barcelona SITE I AUTOMATED WAREHOUSE Storing and modifying of elements
Constructing wood facade design
Potential Process Transformation Through Digitalisation
>>
>>
Site Scanning for Digital Material Dataset Creation
DeďŹ nition and Extraction of Material Properties for Construction and Reuse
Material Dataset
>>
Intuitive design of a large set of discrete elements
>>
Best ďŹ tting function for each element based on its properties
Design Strategy
Develop Fabrication Methods that support new construction solutions for repurposed materials Automation
Companies Interviews and Takeaways
Pre-Deconstruction Site
Deconstruction
Design
Fabrication
% 10-20
Cost & Carbon Footprint comparison
Load Testing for structural performance
Material Informed Design to minimize the labor and extra material
Avoiding Storage Time directly from deconstruction to construction
Adding New Elements around %10-20 for structural assurance
MATERIAL DATASET
Digital Material Dataset
Material Dataset Data acquisition | IaaC Main Building
1 IAAC Main Building Laser scan
2 Selected Area
3 Selected Area Segmentation by Dimensionality and Planes
Colored by ID
Digital Material Dataset
Material Dataset Data acquisition | IaaC Main Building
Canupo Classifier Classification Parameter: Dimensionality
Classifier 1: Roof board
Confidence Threshold: 0.95
Confidence Threshold: 0.95
Classifier 2: Rafters
Confidence Threshold: 0.6
Digital Material Dataset
Material Dataset Data acquisition | IaaC Main Building
Quality by texture | Variation on each piece Visual Assessment
Final Dataset 220 Elements
Properties DeďŹ nition
Properties DeďŹ nition Data acquisition | IaaC Main Building
Quality by texture | Variation on each piece Visual Assessment
Name
Iaac beams
Location
Category
Material type
Material specific
Quantity
Carrer de Pujades, 102
Harvest
Wood
Pine
189
Availability
Length
Width
Height
Quality
From June 2028
1000-3000mm
100-500mm
100-500mm
Good
Density
Fiber Stress
E
Compressive Strength
Shear Strength
470 kg/m3
79.5 MPa
11,1Gpa
41.15 Mpa
8.25 Mpa
Description
Several beams from an old warehouse transformed to a research facility. They were used on the ceiling. They are believed to be 60 years old. The parts are available in different sizes.
100%
0%
Properties DeďŹ nition
Initial Structural Testing
All Elements Tested to Max Calculated Load
BUILDING A
BUILDING B
BUILDING C
BUILDING D
Sampling From Each Site Tested to Failure
Computer vision detection for knots and splits Example Test Machine : Admat Expert 2600 Example Defect CV : A Multiple Systems Approach to Wood Defect Detection, Xiangyu Xiao, 1998
DESIGN STRATEGY
Design Strategy
Design workflow
GENERATIVE DESIGN
ELEMENT MATCHING ALGORITHM Selecting from MATERIAL DATASET via ranges
OPTIMIZATION SOLVER EVOLUTIONARY ALGORITHMS
Design Strategy
Material Dataset Design Shape Fitness Dimensions Material Structural Analysis Structural Data Cost Analysis Distance to material resource Characteristics Quality
CASE STUDY PARAMETERS Length Range: 100mm to 10000mm Width Range: 50mm Height Range: 100mm Minimal Quality: 0 - 1 Maximum Distance: 100km Number of Items: 800-1500
Design Strategy
Discrete Wood Elements Plug in Building Barcelona Current Building
http://www.miasarquitectes.com/port folio/plug-in-building-barcelona/ Mias Architects Project Size Area: 7.600 sqm Use: Offices
Proposal - Front
Proposal - Interior
Design Strategy
Design Strategy Solar Performance
mer
Sum 72ยบ Wi
nte
r2
6ยบ
Fitting Function
Construction System Tolerances and adaptability
Design for Disassembly. Dry joints allow for and easy disassembly of the structure
Tolerance x: 40 mm
Tolerance y: 40 mm
Tolerance z: 100mm
Adaptability. Depending on the material available the system can adapt (x,y,z direction) so that less transformations need to be done in the material
Fitting Function
Matching Algorithm Best Fitting
Design
Elements Needing Material Cuts
Fully Matched Elements
Material Dataset
Materials not matched Elements Needing New Material
Matched materials
Fitting Function
Matching Algorithm Sorted Elements
Scenario A
Scenario B
Scenario C
[...]
DATASET ITEM MATCHES WITHIN TOLERANCE
CUTOFF ADDED BACK TO DATASET
[...] ELEMENT CUT TO MATCH DESIGN
UNMATCHED ELEMENT WILL NEED NEW MATERIAL
Fitting Function
Optimization Solver Parameters Design Space
Controllers
Controls Openings Scale Move Distance of influence
Optimized Parameters Degree of Use of Reused Materials Unmatched Element Count Cut Element Count
Carbon Cost of Chosen Materials Average Element Carbon Cost
Controls Facade Density
Structural Analysis
Distance of influence
Structural Maximum Utilization Structural Total Deflection
Fitting Function
Design Analysis Structural Analysis via Karamba
Material Cost and Embodied Carbon
Recovered Material
New Material
+ Transport Cost for 100km Shipping : Truck : 4.0 kg CO2 / m3 Material Rail : 2.7 kg CO2 / m3 Material
- Sequestered Carbon: Softwood Average : 0.34 kg CO2/ kg Material Varies by Age
+ Scanning, Analysis and Data Storage Electrical Costs Supports at Building Face
Total Beam Deection and Worst-Case Utilization are Calculated
+ New Lumber Processing and Disruption of Forest System
Transport Data from : Acoya Timber Transport Calculator, Example Material Pine
Fitting Function
Design Optimization
Fitting Function
Design Optimization Design Results From Dataset Choice
Dataset Slice 1 (1000 Elements) Maximum Distance : 200km
Dataset Slice 2 (850 Elements) Maximum Distance : 100km
Dataset Slice 3 (650 Elements) Maximum Distance : 50km
Cut Elements : 265 / 391 Unmatched Elements : 0
Cut Elements : 303 / 443 Unmatched Elements : 0
Cut Elements : 311 / 409 Unmatched Elements : 1
AUTOMATION
Material Properties
Transformation
Unloading and Sorting
Milling and Assembly
Prototype Fabrication : Checking Dimensions and Sorting by Order of Construction
Prototype Fabrication : Cutting to Size, Milling for Joinery, and Position
Properties Definition
Warehouse
4
From Iaac IAAC Pujades 102 to Plug in Barcelona
3
ARRANGEMENT
2
INSPECTION AND MARKING
REPLACING automated gantry crane bots move bundles around the storage system
5
ORGANIZATION bundles are organized roughly by length again and the bots just fill out the closest positions first and the database stores the location.
pieces are arranged in bundles by site of origin and roughly by length (e.g. <3m, <6m, <9m)
6
dimensions and visual information is measured and database entries are created/qr codes are marked
7 1
ARRIVING OF ELEMENTS elements arrived without organization from the demolition site
STRUCTURAL TESTING according to order specifications, bundles are retrieved and opened, individual elements are structurally tested
LOADING FOR SHIPPING specific and structurally tested pieces are loaded to be shipped
Properties DeďŹ nition
Storage and Logistics
200615_6bae4d51-1ab7-4fe0-a0f0-ea1ef32ba7d9
PROCESSING DATE + ELEMENT UUID
OPEN UP
References
Iterative Aggregation
Spatial Frames
Robotic Timber Construction Complex Timber Structures
HG-A Live components_Part to Whole
Prostho Museum Research, Kengo Kuma
The Sequential Structure 2, ETH Zurich
https://www.archdaily.com/544023/part-to-whole-hg-a-live-components
http://www.archtalent.com/projects/gc-prostho-museum-research-c enter
https://gramaziokohler.arch.ethz.ch/web/e/lehre/187.html
Others MAS DFAB: Gradual Assemblies https://ethz.ch/en/news-and-events/eth-news/news/2018/08/progra mming-for-perfect-shade.html
Facade Possibilities Design Strategy Morphogenesis Unit Principle | Rules | Direction
Interlocking by Aggregation 3D Multiplicity of dimensions is encourage
Joining Segments into Larger Beams Across Structure
9 Unique Orientations Coming from Cube Face Strategy
Facade Possibilities Design Strategy
Recursive Subdivision based on Performance Criteria
Pattern Generation | Point Selection for Grow Algorithm
Grow Algorithm (Depth) | Wood Location
Structural VeriďŹ cation
Facade Possibilities Design Explorations
Non Planar Surfaces | Solar Radiation
Karamba Simulation with Supports at Wall and Gravity Load
Attractor Points Position | Visibility
Attractor Points Radius | Visibility
Pattern Generation | Density
Structural Optimization
Final Product
Facade Possibilities Simple Interlocking system | Performance & Adaptability Density Based on Solar Radiation Cluster | Iterative Assembly