We've talked in the last several posts about artificial intelligence (AI); machine learning; and the intersection of these topics with the mechanical, electrical, and plumbing (MEP) industry. Jeff Kowalski, Autodesk CTO, talked about forthcoming innovations and technologies in his 2016 Autodesk University keynote. We think it's critical that MEP firms start to think about how these innovations will change our industry.
The last topic we wanted to cover from the 2016 keynote is generative design. At this point, we've seen enough examples of generative design from Autodesk that we can imagine the current "look" of generative design. However, these examples may not feel relevant to our work in MEP systems and building infrastructure. We wanted to provide a definition of generative design and use MEP examples to demonstrate how it will affect our industry.
The very recognizable "look" of generative design. (Source: Autodesk)
The look of generative design is currently defined by the work of Autodesk's Office of the Chief Technology Officer and their work on Project Dreamcatcher and Within. These programs use algorithms to optimize isotropic materials (materials that have the same properties in all directions) into structural shapes that achieve optimized goals. Using these programs, we've seen examples of car chassis, bows for archery, footwear, and aircraft bulkheads.
Here are some pretty cool kicks. (Source: Autodesk)
These examples don't often feel relevant to our work in MEP. Most are 3D printed, and they're structural, not system-based like MEP. But the methods used to achieve these optimized solutions are absolutely relevant to our work. Generative design is a methodology and Project Dreamcatcher and Within are just two implementations of these methods. Let's define generative design in a more general way and then relate it to MEP systems.
Here's a 3D-printed car chassis. (Source: Autodesk)
Generative design is a technology that uses three components to produce solutions to a design scenario: goals and constraints, computational form synthesis, and optimization. Goals and constraints dictate the overall function being sought. Our best example of a goals and constraint scenario is Google Maps. Google Maps asks for a start address, end address, and constraints for your transportation type. Goals and constraints are the high-level definition of your desired scenario. Computational form synthesis is the automation that creates solutions that satisfy the goals and constraints. In Project Dreamcatcher, this is an algorithm that creates forms by intelligently removing material from a structural solid. Form synthesis is always an automation that creates a 3D object. Finally, optimization is a method of selecting outcomes that are better than other workable outcomes. Optimization scores and selects scenarios. For Dreamcatcher, that optimization is often centered around the total weight of the object.
How does this definition of generative design relate to MEP systems?
At BuildingSP, the first step to creating a clash-free route through a project in Revit is to define start and end points for a system, along with parameters. For a piping system, this could be a start point at a variable air valve, going to a zone selector box in a central location, with parameters of only using 90-degree elbows. The start and end points with parameters are the goals and constraints component of generative design.
The computational form synthesis is handled through a combination of GenMEP and Autodesk Revit. GenMEP determines the clear path through a building, if one is available. Once a solution is found for a route, it's passed back to Revit where the path is modeled using the appropriate system. The combination of GenMEP and Revit provides computational form synthesis.
Optimization is the final component of generative design, and probably the most interesting for MEP systems. For a given route, GenMEP optimizes the route along its length as it traverses the maze of the building environment. For each unique set of goals and constraints, we get one solution which is optimized for that route. But if we change the goals and constraints, we get a new solution, which may or may not be better. We'll write a future post describing some of the various strategies that we use to achieve optimization, which include iterations on parameters using Dynamo and cost-based 5D BIM approaches.
One example of a 5D BIM approach using GenMEP and Dynamo particularly excites us. Alejandro Mata of MOE in Copenhagen, Denmark, is using a combination of cost data, GenMEP, Dynamo, DynamoBrowser, and Revit to do optimized 5D approaches for ductwork and piping. Alejandro has promised us a video soon so we'll be sure to post it for you to see.
In summary, generative design currently has a defined look that is being driven by structural shapes created through 3D printing. But generative design is a technology rather than a defined implantation. GenMEP, especially when combined with Dynamo in the Autodesk Revit environment, is a generative design solution for MEP systems that may not look like how we perceive generative design, but which is highly relevant to the practice of BIM. To see GenMEP in action, be sure to take a look at our demo page.
As you may have noticed, we've been doing a lot of writing about the impact of computational BIM, generative design, and the future of how we work. One thing we haven't done is turn some of this thinking into more rigorous analysis. But now we're ready to do that!
If you work for a general contractor, subcontractor, or design firm and want to collaborate on whitepapers that quantify how computational methods of working will change our industry, reach out to us and let’s talk about what we can do. We're open to collaborations worldwide and have lots of ways of measuring performance indicators to gain insight into change in our industry. Contact Brett Young at firstname.lastname@example.org.