Passer au contenu

Research project selected under the 2022 call for proposals

Principal Investigator : Frédéric MOMPIOU

Involved Teams :

  • CEMES / Physics of Plasticity and Metallurgy, PPM

Type of project : Disruptive Project

Date (start/end) : 2021 – 2024

Dislocation segmentation and tracking using deep learning and computer vision

Mechanical properties of metallic alloys are largely governed by the motion of nano-scale linear defects called dislocations and their interactions with the microstructure.

Hence, understanding dislocation dynamics is of fundamental interest to predict material strength. At CEMES, moving dislocations are directly observed during in-situ TEM straining experiments. Up to date, the dynamics analysis is performed manually, which limits statistical treatments, although a large database of observations is available. Moreover, this approach misses a large amount of information by sampling observations and averaging quantities.

The overall objective of this project is to take benefit of computer vision coupled to deep learning methods to exploit databases in order to construct numerical twins of in-situ observations. We expect from this to retrieve quantitative information that could be further implemented in meso-scale simulations.