Hidden deep within Alex Donkers’ dissertation is an image of the National Monument on Dam Square, in Amsterdam. Donkers dismisses it as a brief case study, but it nicely illustrates the necessity of his research. Because the 22-meter-tall pylon consists of blocks of natural stone that are prone to cracking. Therefore, (large-scale) restoration is needed from time to time, Donkers explains. To the frustration of many, this is still not happening efficiently enough.

“Someone is sent to Dam Square for an inspection, walks around the pylon, takes a few pictures and then draws up a report. But there are several different parties involved, each doing it in their own way. Then, a life-size image of the Monument’s side view is printed out and someone hand-draws in lines to distinguish the blocks, all of which are assigned a number. It’s reported that block 228 has a crack in it and needs to go to the construction site. On site, they remove the block in question – hopefully the right one, as some blocks have already been taken out; do they count? – and it’s sent to the restorer with a post-it note barely hanging on. Somewhere on a hard drive, there’s a photo, ‘228.jpg’.” Donkers shrugs almost apologetically. “It’s hard to believe but that’s still how things are done in construction.”

The right language

To make construction more efficient and transparent, Donkers delved into the world of ‘digital twins’ and developed a framework for collaborating with multiple parties in construction using dynamic data. Thanks to various technological advancements, such as the Internet of Things and Artificial Intelligence, the concept of ‘digital twinning’ is also gaining ground. In simple terms, this is a virtual representation of a physical object, process or system. But it’s more than just a 3D model, because it can continuously be updated with data from the ‘real’ world. The construction world is also experimenting with digital twins, but according to Donkers, there has been little success so far. “Actually, things go wrong in every step of the so-called digital twin cycle: collecting data, creating a usable representation, and extracting information from it to eventually be able to adapt the physical environment. So we went back to basics to design a model that does actually work.”

And it may seem rather obvious, but one of the most important conditions for a functional digital twin is that users understand each other’s language, says Donkers. “That’s where the problems start. Take a wall, for example. An architect sees that as an inner leaf, an insulation layer and an outer leaf. A structural engineer looks mainly at the center line of the inner leaf for the sake of the load-bearing structure; an urban planner thinks of the exterior of a wall and how it fits in with the surroundings, and a building physicist makes graphs of the wall’s material properties and thicknesses. And that’s just a wall. Our first step was to define a usable language. That starts with creating ontologies – dictionaries of sorts – with unambiguous definitions. Only then can you move forward.”

Building the system

Donkers used new methods developed at Computer Science to connect data in the right way. That can be a bit of a shock for hardcore architectural engineers, he smiles. “I define a wall as w3id.org/beo, with a whole website attached to it – try explaining that at the drafting table.” In addition to his work as a ‘translator’, Donkers has primarily aimed to make building users central in his models. Can we factor in people’s experiences in a building so we can learn how best to design for the end-user?

After developing multiple digital twin tools, Donkers set to work on a real-world experiment from his home base, Vertigo. He investigated how dynamic data could be used to optimize the indoor climate for employees. “That requires a lot of information about the building’s ‘shell’. For example, whether it’s nice and warm inside depends on how the exterior wall is constructed. We developed a smartwatch application that allowed us to collect real-time data from participants. An office space was covered with sensors so we could monitor the indoor climate. And adjust it. Because if too many people thought it was too dark in the office space, the digital twin could check if the lights were on and take action if necessary.”

A chatbot for your building

Of course, these are only experimental studies, Donkers emphasizes, and it’s not as if every TU/e employee will soon have a smart watch to optimize their workspace. “The use of a semantic web – a framework for sharing and (re)using data – is still relatively new in the world of construction. We’re at the forefront when it comes to developing methods to place the building user at the center. By making my tools publicly available, other researchers can now continue to build on this. This way, we will soon be able to interact even more with a building, especially with all the advancements in AI technology. Imagine having a chatbot for your building that you can actually ask questions! For this too, having the right dictionaries is key. Because in the end, it’s about speaking the same language while also having the courage to embrace that there are differences between people.”