Susan Smith has worked as an editor and writer in the technology industry for over 16 years. As an editor she has been responsible for the launch of a number of technology trade publications, both in print and online. Currently, Susan is the Editor of GISCafe and AECCafe, as well as those sites’ … More »
From Manufacturing to Architecture and Back: The Journey of Generative Design
July 6th, 2017 by Susan Smith
Generative Design has been around for several years. It came on the scene as the product “Generative Components (GC)” from Bentley Systems, used in the group formed for the express purpose of exploring architectural design “what if” with advanced associative and parametric tools, Smart Geometry, that allows users to go far beyond the limitations of traditional CAD software.
Traditional CAD software at one time was mostly for computer aided documentation – 2D plans and drawings that are also the lifeblood of the CAD industry and process. Generative design allows users to generate conceptual design for specific goals, whether they be challenging shapes of buildings or challenging use cases.
For years, Generative Design has been a research project as well as active technology, a use case for the very few who were working on out of the box types of buildings, buildings that stretched the limitations of ordinary computer aided design software and asked for more. GenerativeComponents to this day, graphically presents both design components and abstract relationships between them. This capability lets GenerativeComponents go beyond making geometry explicit; it makes design intent explicit as well. Although designers are working graphically, based on intuition and experience in architectural design, their work is captured in logical form.
In 2013, the last year of Bentley’s hosting of SmartGeometry (sg) before Autodesk took it over, sg teams used data in design, working with the environment, information-centric, recognizing that there are “only a subset of relevant factors that can be modeled in a traditional design CAD package.” The built environment must last for generations.
As one of many examples of the use of sg, Park House Oxford Street combined workflow and embodied the evolution of parametric modeling and how they use it. The original shape was generated by its textual constraints, and they generated the curve through 3D modeling with a good understanding of geometry.
The 3D model then progressed through smart models and the next evolution of Bentley models. It has 39 apartments, office buildings, and shops. The challenge was to make it all come together as it is a city block and integrate it into the townscape.
“Dimension driven design came in looking at how we play with curves to give us final form of a main entrance, with tangential curves ,” said Paul Rogers, architect.
“The roof is angled, and how you start to bring elements together is interesting using parametric modeling and GC and to start to understand how you put flat paneling on tight geometry.”
Without gc, John Ball, architect said, this building wouldn’t be built because “people would refuse to take the risk. We did this straight out of the box with some fair amount of help from Bentley, a little guidance to make a building possible.”
Clearly, parametric technology was lifted from the manufacturing industry to apply to architectural challenges. Users used this software along with software from other companies such as Gehry Technologies, Rhinoceros, SketchUp, BricsCAD, trying to find ways of implementation and thereby shaping generative design to their individual project needs.
As time has gone on, the research has grown to include materials, and not just materials for buildings but for things too. Materials for future buildings, materials for manufactured items.
Autodesk has shifted the focus of generative design back to manufacturing. Take a look at Netfabb 2018, its generative design/additive manufacturing solution, where machine learning comes into play. Autodesk Generative Design is a service that allows users to generate multiple options from which engineers and architects can choose. It is derived from their generative design product, Project Dreamcatcher.
This vision bears little resemblance to the Bentley GenerativeComponents’ architectural vision.
The part that is similar is the ability to form “what if” scenarios. What is enormous about generative design is that it will take the specs plugged in by the designer, such as plans, size, site, style, materials, cost, etc, and then be able to create a large number of design solutions from that information. The result is intelligent algorithms based on machine learning and advanced simulation, that could lead to design solutions that could then lead to a savings in cost, time, materials and an altogether more efficient, pleasing design.
Generative design now can take advantage of recent advancements in artificial intelligence and simulation so that software can play a more participatory role in the invention of the form. So now Dreamcatcher becomes the experimental design platform with “focused research probes into generative design systems.”
It’s interesting to see this, as generative design has employed parametric modeling, once the sole province of manufacturing, in its quest to be able to design buildings with more challenging parts and solve a lot of architectural engineering problems that have plagued designers for a long time. Now as per Autodesk, it is moving back to mechanical design, and generative design with additive manufacturing technologies is addressing product design as it has done all along.
Here’s how Autodesk Generative Design Service works: the service itself examines thousands of materials and processes in various combinations, and then generates multiple options that the engineers can then pick from to decide what course they want to take. They know they’ve explored every option at that point. They are not limited to the usual few options provided by traditional CAD tools.
CTO Jeff Kowalski calls generative design the “true embodiment of CAD.”
Ultimately, perhaps generative design will be able to solve other problems not yet solved in architecture, such as building a building that is not only energy efficient, but that behaves as it was intended. Because buildings are complex and require many teams to build, they don’t generally behave as intended.
So generative design is more than a tool for architectural purposes, but also a tool that can span industry segments fluidly, being used to manufacture products and then use those products for building.
The system kicks back to the user design alternatives including the performance data for each solution, which designers can evaluate in real time. When working with design space, whether it be building footprint or a product, designers can make necessary adjustments according to the data and choose a design, then output it to the next level, fabrication or beyond.
Dreamcatcher is able to generate and evaluate complex solution sets with the help of purpose-built, scalable and parallelized cloud computing framework that the company has code-named Saturn. Saturn’s infrastructure makes it possible to do high performance computing with computationally adept optimization and analysis engines, that include multi-physics simulations.
Who is the Dreamcatcher team? This team comes from the Computational Science and Design Research groups of Autodesk Research in the Office of the CTO, including collaborators throughout Autodesk Corporation, industrial partners and academic partners. Specialists from the professions of mathematical optimization, geometry, machine learning, mechanical engineering, material science, structural mechanics, user experience research, software design and development based in located in London, San Francisco and Toronto make up the global team.