The National Institute of Building Sciences (NIBS) held its annual Building Innovation conference this past week in Washington DC. For those not familiar with the organization, NIBS is responsible for the development and management of the National Building Information Model (NBIM-US™) standards. BIM, of course, is all about construction data in a digital format. This includes location, dimensions, material, equipment, costs, schedule and more. We attended this event to learn more about possible efforts being made for use of AI for access and analysis of the data in a BIM model.

Review of BIM

On many large projects, one or more BIM models may be created. Typically, on the design side, the model will contain all the data that you would find on a set of construction drawings - architectural, structural, mechanical, electrical, plumbing and miscellaneous. At a minimum, a BIM model can be used to create 2D drawings for use by contractors, subcontractors, suppliers and fabricators. Many specialty firms can fabricate items directly from data in the model, or by reading the data into specialized production equipment, a CNC machine or an additive-printer (3D printing). 

On the construction side, quite a few general contractors will use either the BIM model developed by the design team, or they will create their own model for the purpose of scheduling and/or clash detection. For scheduling, the contractor may add data such as construction equipment (cranes, loaders or a batch plant, for example), lifts, scaffolds, formwork and temporary structures. 

Using 4D schedule software, the contractor can create and coordinate workflows to best optimize construction by running visual simulations with the construction team. The clash detection helps determine how the MEP (mechanical, electrical and plumbing) will best fit in a space as well as helping coordinate the trades.

However, there is still a large segment of the industry that either does not know what BIM is, how they might use it or how to access it. This may be, in part, due to the costs, and skills, required to learn and implement its use. 

Let’s back up a minute and look at the business use case for any given participant. First, any contractor considering its use must decide what BIM use will do for them. Do they want to improve construction productivity? Will it help them reduce rework or better coordinate subcontractors? Will it help improve quality? Will it provide a competitive advantage? Or do they simply want to work on larger jobs where they know the owner requires the use of BIM?

AI Can Help Develop Better Design Data for the Model

In a session entitled “The Intersection of AI and Building Sciences: A New Era of Intelligent Infrastructure,” Kimon Onuma started off explaining AI to the audience. First, Onuma explained that artificial intelligence is essentially intelligence created by teaching a model (a robot or a computer) to learn things, just like a teaching a child. If you repeatedly show images of different kinds and colors of cats and tell it that those are cats, it will over time learn what a cat is. If you then show it a picture of a dog, it will know it’s not a cat. It won’t know it’s a dog until you show it a picture of a dog and say this is a dog. 

Models can be taught to do many things, from basic to complex, from recognizing cats to playing chess to learning to drive a car, if you teach it what road signs and stoplights are, explained Onuma. 

The second panelist, Roger Grant, described how AI is being used by a company from Norway called Consigli to make MEP design as efficient as possible. By reading all the rules that drive MEP and coding them into a system along with attendant mathematics, the application can then do all the calculations and iterations to generate a very accurate ceiling plenum or mechanical room space. 

Rules such as return air at 45 degrees instead of 90 degrees or placing ducts a certain distance from the main line, will yield improvements in efficiency. The same is true with cable trays and piping. So, you can reduce the amount of material used. By automating the parts of engineering that humans really don’t do as well, AI gives faster, more accurate and efficient results and it cuts material, construction and operating costs. That ultimately reduces carbon consumption, Grant notes.

Facility Design

Next up in the AI session, Onuma expanded the concept to design a complete facility, such as a port or a housing complex. You give it all the rules of the business process of how the facility will function, how the cranes and cargo ships will be controlled, what types of rooms are needed in the terminal building as well as information on the location. 

Then the AI can help produce initial designs and layouts, tailored to the location. The designer can work with this information until it is sufficiently developed for a particular site, then feed the information into a BIM model such as Revit. 

For example, designing a residential tower in Japan will have entirely different room layouts and furniture requirements than a tower in Kansas. Onuma emphasized how his firm has used AI tools to feed rules, data and feedback from customers into a model and have it generate successive useful designs. Onuma likens it to having an intern to do the grunt work, leaving more time to be creative. 

In addition to the engineering aspects, Onuma demonstrated the use of Midjourney to generate design ideas and what the space and views can look like from inside the structure. Onuma explained how architects can work with the imagery together with the BIM model and digital twins to implement the final design.

Onuma presented another example. The LA City College community college system, with over 5,000 buildings spread over 72 districts, pulled all their data out of a locked-up database system into a web service model accessible by APIs, allowing the college to drill down into any piece of equipment. He further described how AI could be used to sift through the 40,000 work requests a year to uncover issues that could be acted upon.

Applied AI

In closing, Onuma emphasized AI is not a single application. AI, BIM models and digital twins have many different users with different types of tools. The important thing is the ability to collaborate, unplugging or plugging in tools as needed.

In the next issue of AI Construction News, we’ll continue with our coverage of the conference and what we learned about digital twins along with the just-release Version 4 of the BIM standards.