Navigation and service

Third “TiM” workshop of BASE: Strengthening trust in computer-based models

News

Stand: 2025.02.25

Third “TiM” workshop of BASE: Strengthening trust in computer-based models

On 19 and 20 February, the third and, for the time being, final BASE workshop on the topic Trust in Models (TiM) took place. A total of 21 researchers from ten institutions participated: 

  • Helmholtz Centre Dresden-Rossendorf (HZDR)
  • RWTH Aachen University
  • Technical University of Munich (TUM)
  • Clausthal University of Technology
  • Freiberg University of Mining and Technology
  • GFZ German Research Centre for Geosciences, Potsdam
  • Helmholtz Centre for Environmental Research – UFZ, Leipzig
  • GRS – Gesellschaft for Plant and Reactor Safety, Braunschweig
  • Martin Luther University Halle-Wittenberg
  • BASE

Computer-based models and repository safety research

In many areas of nuclear waste disposal safety research, computer models and their evaluation play a central role. In particular, assessing repository safety over a timespan of one million years is not possible without numerical models. 

Key factors for building trust in numerical models identified

With regard to the topic of repository search and nuclear waste disposal, the following aspects were considered essential for building trust in computer-based models: 

  • Complexity: Models must be able to represent complex processes in a repository while also capturing fundamental processes.
  • Transparency: This ranges from publishing the code to providing detailed descriptions of boundary conditions, parameters and the underlying physics.
  • Benchmarks: Codes must be compared with one another, and models cross-checked – ideally by independent working groups.
  • Experiments: These can close calibration gaps in numerical models and validate the models.
  • Human factor: Work steps must be standardised, and automation and error detection should be introduced wherever the “operator at the device” is a variable.
  • Artificial intelligence: Can help train datasets, detect programming errors and standardise modelling processes.

These aspects will now be taken up and addressed using specific examples in follow-up formats to the “TiM” workshops.

Stand: 2025.02.25