Generative Design and the Architecture/Engineering Professional

Can Computers Become the Next Frank Lloyd Wright?

Machine Learning and AI have existed in principal since the 1950’s, and pioneered by Arthur Samuel (1901 – 1959) through the programing of computers so that they may learn from past experience. Since Samuel’s early experiences of having computers play and learn from games of checkers, machine learning has integrated into modern manufacturing, analytics, and now into the profession of Architectural Design.  The ultimate question is how this application will impact the role of the Architect and Engineering professional as it relates to building design.

Autodesk – The Industry Standard

Autodesk “builds software that helps people imagine, design, and create a better world.” [1] The company’s drafting and design software AutoCAD was the standard program in the creation of computer generated construction documents for many years until their Revit 3D Building Information Modeling (BIM) platform arrived in the early 2000’s.  This software gave Architects and Engineers the ability to create three dimensional computer models of their building designs and coordinate all of the various systems such as Mechanical, Electrical, Fire Protection, and Structural elements.  It has become the new standard for the design professional.

Images courtesy of Autodesk, https://www.autodesk.com/products/revit-family/overview [2]

Images courtesy of Autodesk, https://www.autodesk.com/products/revit-family/overview [2]

Rise of Machine Learning

At the Data Center Dynamics (DCD) conference in San Francisco in June of 2017, which I attended, Lloyd Taylor, VP of Cloud Infrastructure Technology for Autodesk stated that, “Machine Learning has no preconceptions, and that is what makes it so beneficial.” During his keynote address to the conference he discussed the intersection of Architecture and product creation with machine learning, noting the aviation manufacturer Airbus’s application of machine learning in the design of panels for the a320 aircraft.

Image courtesy of Autodesk, https://www.autodesk.com/customer-stories/airbus [3]

This application is an example of Generative Design which “is a technology that mimics nature’s evolutionary approach to design. It starts with your design goals and then explores all of the possible permutations of a solution to find the best option. Using cloud computing, generative design software quickly cycles through thousands—or even millions—of design choices, testing configurations and learning from each iteration what works and what doesn’t.” [2]  Given certain constraints input by the Architect, Engineer, or Designer the software leverages Cloud Computing to develop tens of thousands of design iterations in the same time that humans can develop only a few.  The software is able to optimize the parameters, structural integrity, weight, and shape in the example of Airbus a320 panel, in ways that human intelligence can’t.  Mr. Taylor of Autodesk did state in his address that the danger of machine learning is the level of accuracy of the system training set of data from which the design algorithms draw from.  If this set of data is not accurate, then various flaws in the design can occur.

Generative Design in Building Architecture

“Jordan Brandt, Autodesk’s resident futurist, makes a clear distinction between explicit design and generative design. ‘Generative design is when you state the goals of your problem and have the computer create design iterations for you.’ ” [4]   

Autodesk put Generative Design to the test and allowed for their own offices in Toronto to be configured using design algorithms to show how the concept of Generative Design can be utilized in the Architecture industry. [5]   During the programming phase of the tenant fit out design Autodesk’s internal research team solicited input from those employees who would be housed and working in the space.  The goal was to locate each employee in the optimum location on the floor plate while balancing constraints vs. needs/wants of the employee.  Some of the criteria entered into the design algorithm including adjacency to team members, daylight, lack of visual or audio distractions, and proximity to office amenities against the physical constraints of the building floor plate.  Autodesk noted in their article that the “general algorithm harmonized all the equations to find solutions and took into consideration the following objectives: Generation of data (geometries, position of utilities, spaces), Evaluation of each objective, Production of all solutions meeting the requirements, Analysis and choice of the best solutions.”  [5]

The process generated hundreds of potential layouts and the design team categorized each in relation to the ability to maximize the requirements, shown graphically in a spider diagram shown below.

Autodesk Office Generative Design, Image Courtesy of Autodesk [5]

Engineering the “Water Cube” for Beijing 2008 Olympics

The aquatics center known as “The Cube” was a key feature of the 2008 Beijing Olympics, known for both its unique architecture and the platform for Michael Phelps record setting performances.  Developed by the team of China State Construction and Engineering Co (CSCEC), PTW Architects (Australia) and Arup Engineering (Australia) the story of how the building took shape through Generative Design was highlighted in a story written in Architecture Australia magazine. [6] PTW and Arup Engineering drew inspiration from the physics of soap bubbles for the exterior building skin, which in turn made the structural systems extremely complex.  The building contained thousands of exterior “soap bubble” panels, supported by 22,000 structural elements.  Arup noted that the task of engineering the structure though a manual calculation process would be impossible, so they turned to software and generative design to optimize the structural members given the inputs and constraints of the architectural form. [6]  Utilizing Autodesk add in design programs such as “Rhino3D” and “Grasshopper”, shown in the image below [7], allowed for PTW and Arup to perform the calculations and automate the process which they note became very efficient.

Generative Design Model of Water Cube, Image Courtesy of PTW Architects, Arup Engineering, Autodesk, Grasshopper [7]

The Role of the Design Professional

The practice of Generative Design is not new however the advent of new tools will make this practice far more prevalent in design architecture and engineering of the built environment. Autodesk is an example of how the Architecture/Engineering industry is utilizing machine learning to push designs to the edge, but verify structural integrity.  In addition, construction costs can be reduced through optimization of building structures.  As mentioned, the danger of Generative Design and machine learning is establishing the proper set of inputs and constraints for which the computer applications can draw from to generate the design.  The role of the design professional will be to understand and accept the liability for the outputs of the program.  This responsibility will rely solely on the understanding of best professional practice and good old fashion experience.

References:

[1] https://www.autodesk.com/company

[2] https://www.autodesk.com/products/revit-family/overview

[3] https://www.autodesk.com/customer-stories/airbus

[4] https://www.wired.com/2015/09/bizarre-bony-looking-future-algorithmic-design/ 

[5] http://autodesk.typepad.com/bimtoolbox/2017/06/generative-design-applied-on-buildings.html

[6] http://architectureau.com/articles/practice-23/

[7] https://moreaedesign.files.wordpress.com/2010/09/10082020watercube20in20grasshopper.jpg

 

 

 

Users who have LIKED this post:

•