Decision Making Framework For Design Exploration Parametric Design, Outside The Box, Generative Design, Outside The Black Box. Youngjune Lee Master of Science, Advanced Architectural Design Territories of Investigation : Architecture and Ecology(A+E) Cornell University The College of Art, Architecture and Planning December 2021 Copyright by Youngjune Lee All Rights Reserved. No part of this portfolio may be reproduced or used in any manner without written permission of the copyright owner except for the use of the quotations in a review. ©2021 Abstract Advancements in the utilization of computers in architectural design have opened new horizons in how design is approached. Not only have they enabled us to discover new types of design, but they have also helped us find optimal solutions. The evolution in the application of such tech- niques can been divided into three phases. First, architects used computational tools to discover interesting shapes. Second, architects started to use the computer as an apparatus to make more logical decisions within the design process. Lastly, architects are now designing the decision-making process itself to fully utilize the potential of com- puters. Each of these design paradigms has been strongly affected by engineering and computer science while slowly adapting to practice, discov- ering novel designs, or reducing time in optimizing models. However, such methods still have certain limitations in terms of intuitiveness in processing design requirements or designers’ intent. Now we are entering a new phase of evolution where ar- chitects and computer scientists are working to implement artificial intelligence to help reduce the gaps between the tool and designers. Contents Background 1 The Shift 3 - Fall 20 Design Studio Generative Design 7 Importance of Tools 9 - Fall 21 Elective - Independent Study - Fall 21 Design Studio Outside the Box 13 - Summer 21 Design Studio Outside the Blackbox 19 Bibliography 20 A Framework for Integrated design apporach A Framework for Integrated design apporach By the help of new digital design technology, we are now capable of combining multiple layers of information into design process. This information can range from materiality, structural performance, environmental impact or even lifecycle assessment. As Achim Menges points out, now our role has shifted to orchestrating each genotypes of the de- signed system. What is more interesting is that as many architects has built robust system to analyze de- sign schematics, architects have more freedom in choosing the forms that suits with the context or narrative which does not start from structural behavior. Sunlight Hour Structural Analysis Planarization Parametric Model Genetic Algorithm This concept shares the core idea of using an apparatus that has certain feedback like the hanging chain model of Gaudi or Frei otto, the only difference is that the result they try to Design pipeline for the framework simulate is not limited to structural behavior. For example, in Safra Neuron Center’s screen, designers of the Fosters and partners have used the formal logic of neuron to design a façade where it’s form has evolved for exchanging electrical stimulus between cells. Topology Optimization Figure. Structural analysis of façade following the formal logic of neuron with Genetic Algorithm Planarization Moreover, we can also use inspiration from nature or manmade material to use as a form finding device. The structure can follow the logic of human muscle or curvature of a textile but still retain the structural and programmatic needs of architecture. Skeleton Panelization Sunlight Hours Structural Analysis Joineries Planarize Planarize Planarize Planarize While cost of developing such framework can be expensive, since each algorithm would Cylinder - Radius iteration 0 iteration 1000 iteration 2000 iteration 3000 be most likely unique per project, the supporting tools such as environmental or structur- Cylinder - Height planar al analysis can be re-used to other projects and has become more accessible throughout Opening - Position U non-planar Opening - Position V recent years. Thus, as more and more designers develop the framework to develop their Opening - Radius design, all the other designers can also benefit form their contribution. • Klinger (eds.), Manufacturing Material Effects: Rethinking Design and Making in Architecture (New York: Routledge, 2008), 242P • Achim Menges, ‘Integral Formation and Materialisation: Computational Form and Material Gestalt’, in B. [1] Example of Computational Design Process Kolarevic and K. Klinger (eds.), Manufacturing Material Effects: Rethinking Design and Making in Archi- tecture (New York: Routledge, 2008), 199P By using multitude of evalution throught out • Barbash, S., D. Chorafas, H. Sompolinsky and A. Citri. “Safra Neuron Screen : Design and Fabrication.” the design process, the architect can find the (2016) optimal parameter that defines the final de- ARCH 6509 YL3472 ARCH 6509 sign. YL3472 Background Advancement of technology has enabled architects to design complex geometries with multitude of functions. The need for designs to meet more requirements lead to develop new tools to aid designers and reciprocally such new tools opened possibility to discover new design. Similar relationship of mutual dependency between design and machine or tool can also be found from cathedral design and drawing instruments from early nineteen and eighteen centuries. (Witt 2010) In the early eighteenth century, the need of complex geometries re- sulting from vaults and arches consequently lead to invention of instruments such as Ellip- sograph to easily draft conic sections multiple times. Moreover, such invention of tools allowed architects to experiment even more complex or unprecedented designs more which in turn once again advanced the geometry and de- sign. 1 Cap (PP) Label (PVC) 30 Bottle (PET) x2 Components Under Recycled Redundant Heating Post Deconstruction Lorem ipsum dolor sit amet, consectetur adipiscing elit, Most of the plastics doesn't get recycled because of the During the recycling process, the plastic needs to turn The problem with reusing the recycled plastic after sed do eiusmod tempor incididunt ut labore et dolore economic reasons. It is cheaper to use new crude oil, the into pallets. After that, it can be injection molded into deconstruction is critical problem not only because we magna aliqua. Nibh tortor id aliquet lectus proin nibh nisl. recycling machine can only handle specific types of resins, new plastic proudcts. So while recycling and reusing don't want to create another waste but also because the Purus faucibus ornare suspendisse sed nisi lacus sed and some plastics are in small quantitiy to be recycled. the plastic we need to redundantly heat it twice. plastic gets deteriorated after each recycle. viverra tellus. Feugiat in fermentum posuere urna nec Only 30% of plastic bottle is recycled eventhough they are Because of that plastic gets deteriorated after each tincidunt praesent. one of the most easiest product to recycle. recle and can only be recycled 3 to 7 times Quality Up-Cycling Sorting Rinsing Drying Filtering Pelletizing Injection Molding Product Regular Manufacturing Recycling [2] Paviltion to Upcycle Plastic Bottles 2 The Shift Although, it is debatable whether the tool is limiting the architect from designing beyond the boundaries of the software or is confined by the options provided, development of com- puter aided design software and as architects who utilize such software become mature, the approach in using the computer has also changed. When it comes to early generation in com- putational design, the architects merely used the parameters that define the outcome as an option, where there was no strong difference between each option. While there are different terms used by scholars, such design method is widely understanded as parametric design (Stasiuk 2018; Caetano, Santos, and Leitão 2020) a term vastly used and well known for Petric Shucmather’s Parametricism. However, in the next generation, architects had criticized on merely using the parame- ters to generate interesting forms but used it as a tool to make decisions based on desired criteria. Not only this helped the designers to give persuasive power to their design, but also brought development of new tools that evalu- ated the design options. 3 [3] Double-shot Casting Method Using double-shot casting method to create custom patterns with melted palstic bottles and caps. 4 [4] Mockup of the furniture made with plywood 5 [5] Example of Generative Design Frameworks From top left to bottom right, Project Discover (Nagy et al. 2017), Lady- bug Pollination (Mackey and Sadeghipour Roudsari 2018), StructureFit (C. T. Mueller and Ochsendorf 2015), Autodesk Insight®, Autodesk Tal- ly®, Timur Dogan(Bernett and Dogan 2019) 6 Generative Design The new generation of computational de- signers borrowed the concept of evolutionary optimization of iteratively narrowing down the possible options generative by the algorithm, which is call design space. BESO (Kicinger, Arciszewski, and de Jong 2005; Allaire et al. 2019; Xia and Breitkopf 2014) and Pareto-Op- timal (Deb and Deb 2014) has largely influ- enced to the concept of this new method. This new method finds the best selection of parameters that describes the design which satisfies all to compounded design require- ments. While there is no clear definition of such method, it is widely understood as generative design (Nagy et al. 2017). Using generative design methods, architects can find the optimal solution that has optimal space planning and environmental and struc- tural performance or any type of numerable quality that can be computed at the same time. In the recent years, various dashboard style generative design frameworks has been re- leased in market where architects can easily apply such technique at will. (Rolvink, Mueller, and Coenders 2014) 7 MatlabAPI matlab-engine [6] Thermal Analysis Using Matlab+Grasshopper Grasshopper Plugin developed using Hops, CPython and Matlab PDE to analyze ther- mal conduction in generic geometries 8 Importance of Tools As much as the generative design framework is important in decision making, the tools that evaluate the design option has same amount of impact in quality of design. Since a lot of design requirements used in architectural design is rarely quantifiable, it is mostly up to the architects to come up with unique solutions project by project. Therefore, the concept of master builder has raised again, to develop custom made software to analyze the qualities of design computation- ally. This has also helped reinforcing the com- putational designer community by democratiz- ing the software that has been inaccessible. 9 Youngjune Lee Minimax 7 Youngjune Lee Minimax 8 CLT + Gypsum [7] Housing Units Assembly Strategy Youngjune Lee Minimax 9 10 view to park view to view from street waterfront dynamic sketch multi-objective sunlight hour evaluation 1st Phase 2nd Phase Regular Unit - 50 regular units - 136 Flexible Unit - 35 flexible units - 128 Social Space - 11 Commercial - 6 Office - 8 Voxel with both views Voxel with water views Voxel with green views Voxel with no views 0 20 40 80 Site Plan Youngjune Lee [8] Orientation and allocation of units were driven by accumulat- Minimax 10 ing view factor and thermal comfort. Youngjune Lee [9] Perspective Renderings from the courtyard Minimax 12 11 The Weaver Project with Moshe Borouchov Yiran Wang [11] The Weaver Algorithm I Using the patterns found form Phys- arum Slime, we generated universally applicable space-filling curve inspired from nature 12 Outside the box While generative design has helped archi- tects to make decisions based on metrics in design process, another aspect of such para- digm is that it can be universally applied to any project. For that reason, number of designers even tried to apply this technique to the forms that has not been applicable in architecture before. By help of the generative design framework, architects can use form generator that has less to do with feasible structures such as patterns found in nature. For example, from the Safra Neuron Center by Foster+Partners, patterns of neuron was used as a generator and used the multi-objective optimization method to satisfy the mechanical requirements. (Musil Foster et al. 2016). Architects can now incorporate complex ge- ometries that hasn’t been sought as reason- able options before. Moreover, such trend also helped development of computational tools to ease access for the designers to esoteric algo- rithms. 13 [13] The Weaver Algorithm II Introducing multitude of food generat- ed interlocking shapes with aversion and attraction. 14 [12] Trajectories of the weaver in space 15 [15] The Weaver Algorithm III Benchmark of the Weaver Algorithm as a space filling curve. 16 [14] Accumulated trajectories of the two weavers Each weaver is only attracted to specific type of foods and avoid the other, which in result creates intricate patterns. 17 [16] Renderings of Pavilion Design Using Weaver Algorithm 18 Outside the Blackbox It is hard to deny that generative design has of- fered the architects more freedom in utilizing the computer as an assistant in making decision. How- ever, there is still certain amount of gap between the designer’s and how the software works. For the computer to interpret the qualities of the design, all design objectives must be quantifiable and by the limit of pareto-front, each number should have a dominance over other. However, not all the design requirements and qualities can be quantifi- able or has dominance over other. While the designers want to regulate the perfor- mance of the design and find out the parameter that satisfy with the design requirements, not all the soft- ware is capable to work in that way. For example, the designer might want to generate an origami pattern from any given shape, but it is difficult to create a software that behaves that way. Rather than that, it is more likely that. (Demaine and Tachi 2017) Therefore, in the recent studies, artificial intelli- gence is largely implemented to construct more in- tuitive relationship with the architects and the tools, which would be the next phase of computational de- sign framework.(Abdel-Rahman et al. 2019) 19 Bibliography Abdel-Rahman, A., M. Kosicki, P. Michalatos, and M. Tsigkari. 2019. “Design of Thermally Deformable Laminates Using Machine Learning.” In Advances in Engineering Materials, Structures and Systems: Innovations, Mechanics and Applications - Proceedings of the 7th International Conference on Structural Engineering, Mechanics and Computation, 2019, 1016–21. CRC Press/Balke- ma. Alfaris, Anas, and Riccardo Merello. 2008. “The Generative Multi-Performance Design System.” In . ACADIA. Allaire, Grégoire, Lorenzo Cavallina, Nobuhito Miyake, Tomoyuki Oka, and Toshiaki Yachimura. 2019. “The Homogenization Method for Topology Optimi- zation of Structures: Old and New,” January. Arrieta, Alejandro Barredo, Natalia Díaz-Rodríguez, Javier del Ser, Adrien Ben- netot, Siham Tabik, Alberto Barbado, Salvador García, et al. 2019. “Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Chal- lenges toward Responsible AI,” October. Bernett, Allison, and Timur Dogan. 2019. “Early Design Decision-Making Framework Based on Multi-Objective Building Performance Simulation Incor- porating Energy, Carbon Footprint and Cost.” In Building Simulation Confer- ence Proceedings, 3:1617–24. International Building Performance Simulation Association. Caetano, Inês, Luís Santos, and António Leitão. 2020. “Computational De- sign in Architecture: Defining Parametric, Generative, and Algorithmic Design.” Frontiers of Architectural Research. Higher Education Press Limited Company. Deb, Kalyan, and Kalyanmoy Deb. 2014. Multiobjective Optimization Using Evolutionary Algorithms. New York: Wiley. Demaine, Erik D., and Tomohiro Tachi. 2017. “Origamizer: A Practical Algorithm for Folding Any Polyhedron.” In Leibniz International Proceedings in Informat- ics, LIPIcs, 77:341–3416. Schloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing. Kicinger, Rafal, Tomasz Arciszewski, and Kenneth de Jong. 2005. “Evolution- ary Computation and Structural Design: A Survey of the State-of-the-Art.” Computers and Structures 83 (23–24): 1943–78. 20 Mackey, Chris, and Mostapha Sadeghipour Roudsari. 2018. “The Tool(s) Ver- sus The Toolkit.” In Humanizing Digital Reality, 93–101. Springer Singapore. Mueller, Caitlin, and John Ochsendorf. 2013. “From Analysis to Design: A New Computational Strategy for Structural Creativity.” Mueller, Caitlin T., and John A. Ochsendorf. 2015. “Combining Structural Per- formance and Designer Preferences in Evolutionary Design Space Explora- tion.” Automation in Construction 52: 70–82. Musil Foster, Josef, Hester Rd, An UK jmusil, Darron Haylock Foster, Matthew Hayhurst Foster, An UK mhayhurs, Eilon Vaadia, et al. 2016. “Safra Neuron Screen: Design and Fabrication.” Nagy, Danil, Damon Lau, John Locke, Jim Stoddart, Lorenzo Villaggi, Ray Wang, Dale Zhao, and David Benjamin. 2017. “Project Discover: An Application of Generative Design for Architectural Space Planning.” In . SimAUD. Rolvink, Anke, Caitlin Mueller, and Jeroen Coenders. 2014. “State on the Art of Computational Tools for Conceptual Structural Design.” Shea, Kristina, and Jonathan Cagan. 1999. “Languages and Semantics of Grammatical Discrete Structures.” Stasiuk, David. 2018. “Design Modeling Terminology.” https://archinate.files. wordpress.com/2018/06/dstasiuk-design-modeling-terminology1.pdf. Sutherland, Ivan Edward. 1980. “Sketchpad: A Man-Machine Graphical Com- munication System.” New York. Witt, Andrew J. 2010. “A Machine Epistemology in Architecture Encapsulated Knowledge and the Instrumentation of Design.” Xia, Liang, and Piotr Breitkopf. 2014. “Concurrent Topology Optimization Design of Material and Structure within FE2 Nonlinear Multiscale Analysis Framework.” Computer Methods in Applied Mechanics and Engineering 278 (August): 524–42 21