Material Fingerprinting

  • Funding: European Research Council (ERC) Grant 101141626 DISCOVER.
  • PIs and collaborators: Ellen Kuhl (Stanford University), Moritz Flaschel, Denisa Martonová (FAU Erlangen).

Material model discovery without solving an optimization problem? Yes!

The key idea: every material exhibits a unique response in an experiment, which can be interpreted as its fingerprint. If we build a database of these fingerprints linked to their corresponding mechanical models during an offline phase, we can then rapidly identify the model of a new, unseen material during an online phase using a pattern recognition algorithm. 🕵‍♀️🔍

  • fast model discovery
  • no optimization problem
  • no risk of getting stuck in local optima
  • physically admissible and interpretable material models

In our preprint, we demonstrate this concept for both homogeneous (supervised) and heterogeneous (unsupervised) experiments. The code is available on github.

Material Fingerprinting

Last updated: August 2025