Machine Learning Advances Building Material Stock Studies for Circular Economy

Strategic Insights -

Building material stock studies are crucial for progressing the circular economy in construction, but current models often lack accuracy and scalability, hindering their adoption. To address this issue, researchers at the Swiss Federal Institute of Technology developed a novel methodology leveraging large language models to extract data from building energy performance certificates. This approach enabled the creation of a dataset of over 20,000 buildings, significantly larger than those used in previous studies. By leveraging this dataset, the researchers developed a machine learning model to predict material composition based on building characteristics, such as construction year, use, and location.

Key Takeaways:

  • The study introduced a novel methodology using a large language model to extract previously untapped building material data from building energy performance certificates.
  • The research created a dataset of over 20,000 buildings, significantly larger than those used in previous studies, enabling the development of a machine learning model to predict material composition.
  • The machine learning model integrated knowledge of construction history to estimate the material stock of walls in terms of volume, mass, and associated CO2 emissions for each building in the dataset.
  • The analysis revealed significant regional variations in material use patterns, emphasizing the critical role of location, often overlooked in existing building material stock models.
  • The findings provide valuable insights for improving building stock modeling and highlight the importance of regionally tailored policies in advancing the circular economy in the construction sector.
  • The research was funded by various institutions, including Eidgenoessische Technische Hochschule Zurich, Future Cities Lab Global, National Research Foundation, Singapore, and Innosuisse-Schweizerische Agentur fur Innovationsforderung.
  • The study's data was extracted from building energy performance certificates, highlighting the potential of this approach for improving building material stock studies.

Statistics:

  • The dataset created consisted of over 20,000 buildings.
  • The machine learning model predicted material composition based on building characteristics, such as construction year, use, and location.
  • The analysis revealed significant regional variations in material use patterns.
  • The estimated material stock of walls in terms of volume, mass, and associated CO2 emissions for each building in the dataset revealed a complex and varied picture.

Sources:

  • Spatiotemporal Mapping of Swiss Exterior Wall Material Stock Using a Large Language Model and Architectural History. Journal of Industrial Ecology, 2025;29(4):1350-1363.