Set-Based Approaches to MEP Systems Using Generative Design

The architecture, engineering, and construction (AEC) industry is at an exciting crossroads. I want to describe how lean construction is going to be an integral part of a significant new trend. The use of building information modeling (BIM) is mature. Cloud computing is removing computing barriers in lots of different ways. And there's a slew of burgeoning technology on the horizon that is exciting and innovative. Several principles of lean construction will become very important in the near future and lean practitioners have an exciting opportunity to add value to new AEC processes.

 

You may have heard Autodesk CTO Jeff Kowalski talk about generative design during the 2015 Autodesk University keynote. Generative design is a design method where computational power and processes are combined with intelligent automation to create sets of design options. This approach is different from past approaches because using computers to produce outputs increases the number of solutions. Depending on the scenario, we can produce hundreds, thousands, or even millions of different solutions.

To demonstrate these numbers, let me provide a simple generative design scenario that we at BuildingSP have been working on and understand well. BuildingSP automatically routes and models mechanical, electrical, and plumbing (MEP) systems in 3D and without clashes. You can think of our work like Google Maps – if you provide a start point, an end point, and a series of parameters (go by car, bus, or bike, etc.), you get a routed solution. In our practice, we can pick the preferred elevation of a route and at least five other parameters. For a given electrical conduit route, a very simple set of "inputs" might look like this:

Route an electrical conduit from the main switchboard to subpanel "S1," use a preferred elevation of 10'-0," use only standard angle conduit bends and bias to the right side.

For an elevation of 10'-0", there are six different possible combinations of solutions for that input set, e.g., use standard angle conduit bends or do not use standard conduit angle bends and bias to the left, right, or without bias. If I pick any elevation between 10'-0" and 12'-0" and using a resolution of ¼", there are now 576 (96 x 6) combinations for one route. If we have four routes that have the same types of parameter choices, then the possible number of route solutions is 5764, or 110 billion combinations.

The math gets a bit more complicated. In real-world applications, the following conditions exist:

  1. Buildings contain more than four conduit routes. They contain several system types and each system has many different routes.
  2. The number of parameters is higher than just three and the parameters change depending if we're routing a piping route versus a conduit run.
  3. The order in which we route systems also matters because that determines which system gets priority.
  4. Not all possible solutions are valid solutions. Many combinations won't result in a scenario that is valid.

Based on this, the math gets complicated and will be variable, but it's fair to say that generative design and computation tools produce lots of different result sets. How does generative design, with the huge number of possible solutions, intersect with the field of lean construction?

The answer is set-based design. Set-based design is a lean construction methodology where multiple viable alternatives are generated.   This methodology has been used by Toyota in its work, but also by David Mar in his approach to structural engineering. Set-based design creates a scoring system, customized on a given project, by which we can evaluate individual routes and complete models. This is important because, given the huge number of models that can be generated, we need a way of comparing models so we can determine which is better.

Let's get back to our MEP system example. In MEP coordination, we can think of several different types of values that we may find important. These values may include:

  1. Total material cost
  2. Total material and labor installation cost
  3. Life cycle cost
  4. Maintainability
  5. Constructability
  6. Feasibility of pre-fabrication
  7. Aesthetics

Set-based design reconciles the comparison between these different values. While material cost is a straightforward, technical value to be defined, aesthetics are qualitative. At BuildingSP, we're creating measures to evaluate these and other properties. For example, we have a ratio that defines the constructability of a MEP solution in a defined space.

A lean practitioner is needed to work with a client to confirm how these ratios align with their goals. In a given building, there would be baseline values based on initial interviews or examination of existing buildings or models. On a room-by-room basis, values may differ to compensate for the function of that room. The aesthetic values of a reception area will be higher than the maintainability of that space. A utility corridor outside an electrical room would be completely opposite. The value system created through set-based design then eliminates the weaker solutions created through generative design and causes a convergence of solutions that match the values of a given project.

Generative design produces a huge number of possible solutions. Set-based design within the lean methodology creates a value system, broadly considering sets of possible solutions and gradually narrowing the set of possibilities to converge on a final solution.

The nature of our work is changing, and we have to adapt to the new ways of creating value for our clients. In the practice of AEC, there's a revolution occurring that will dramatically increase the productivity of creating 3D models. Lean construction focuses on the values created by deliverables, which provides the framework that allows computational methods to converge on the answers that will build projects.

Tags: Generative Design Lean Thinking

Drop us your email address and stay connected with us!