According to the Norwegian Statistics Office, an average of around 1,100 single-family homes have been demolished per year in Norway over the last decade. However, only 7 percent of the wood from these buildings was recycled.
Georgios Triantafyllidis, a doctoral student at the Norwegian University of Science and Technology (NTNU) in Gjøvik, believes that much more could be recycled.
“The construction sector currently uses a lot of raw materials and also generates large amounts of waste. This means that the industry is not only a large consumer of new natural resources, but also one of the largest causes of global CO₂ emissions,” explained the researcher.
If we continue on this path, the climate and environmental impact is expected to increase as more and more buildings are renovated and modernized.
“However, if we improve our reuse of building materials, these numbers could look very different,” Triantafyllidis said.
Extremely limited knowledge
So, together with Professor Lizhen Huang and other colleagues at NTNU, he developed a model to make it easier to categorize and calculate the amount of building materials available for reuse in our homes.
“If we want to move to a more circular economy, we first need to get an overview of how much material is actually available and what quality it is. We currently know very little about this, especially when it comes to buildings,” Huang said.
Don't judge a book by its cover
One way to acquire this knowledge is to analyze the composition of materials and calculate the quantities manually. However, it is a very expensive and time-consuming process – not least when mapping entire cities.
Therefore, one proposed solution was to use 3D scanning technology and machine learning. It significantly automates the process, but also brings its own challenges.
“There are already commercially available technologies that can scan a building and create a 3D representation of it. The problem is that such representations only reveal what is visible from the outside. They do not reveal anything about the materials hidden behind them, and certainly not about their condition or quantity,” explains Huang.
It is also the case that these tools often misinterpret the data, adds Triantafyllidis:
“Because the approach uses visual data, problems often arise when something appears to be made of one material but is actually made of something completely different.”
In practice, this means there is no reliable way to map the materials present in existing buildings at scale.”
That is, until now.
Fragments become an advanced model
The method developed by NTNU researchers is based on Building Information Modeling, or BIM for short. This is a process in which digital 3D models of buildings are created. These models often contain many additional details, such as the dimensions and quality of the materials used and associated technical installations.
The advantages of a common model that everyone involved can identify with have made BIM an established standard in large construction projects. However, because BIM is relatively new, such models do not yet exist for the majority of the building stock.
Therefore, researchers wondered whether it would be possible to use architectural and floor plans, technical specifications, land records, regulations, photographic materials and other available information to automate the creation process.
If so, it would be a solution that saves both time and money, explains Triantafyllidis:
“Rather than relying on expensive equipment and a range of experts and specialists, our method is based on information that already exists. On its own, that information may not be very valuable – a building code here, an architectural drawing there – but when we put the pieces of the puzzle together, a picture that is suddenly far more comprehensive emerges.”
95% accurate
Initially, researchers focused on whether the method was able to calculate the amount of material in a building. To investigate this, they based their study on a fairly typical Norwegian single-family home from the 1980s with a floor area of 140 square meters.
First, they manually created a calibration model of the building type. They then fed the model all the available information they could find. The BIM model then automatically took shape based on the data, providing the researchers with a much more detailed model of the house.
The experiment showed that the method was able to calculate the amounts of material present in the exterior walls and roof with 95 percent accuracy. However, they emphasize that further experiments are needed to confidently confirm the model's accuracy. However, one thing is clear: the results are promising.
And that's not all:
“Because it is easy to customize and based on data available for almost all buildings, the method can also be easily scaled up if necessary,” said Triantafyllidis.
Renovation updates
Despite the promising results, researchers are still looking for ways to make the method even better and more applicable. One option could be to include information directly from homeowners.
Many Norwegian houses have undergone significant changes since they were built, but not all changes have been recorded. Since their method is largely based on publicly available documentation, inaccuracies can occur.
“A renovation can involve anything from small cosmetic changes to major structural interventions. In many cases it will not make a difference. But let's say that large interior walls are demolished – the material amounts could then change without our method necessarily detecting this,” said Triantafyllidis.
The researchers therefore imagine a solution in which homeowners self-register all major changes they make to their home.
“By incorporating this type of information from homeowners, the models would always be up to date and would provide a very accurate view of what materials may be available for reuse,” he said.
The researchers emphasize that this is an important prerequisite for achieving a well-functioning circular economy.
Reference: Triantafyllidis, G., Müller, DB, Wellinger, S. & Huang, L. (2025). Accelerating circular cities through semi-automated building information modeling for existing buildings. Journal of Cleaner Production, 145783. https://doi.org/10.1016/j.jclepro.2025.145783