Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - MuDiLingo (A Multiscale Dislocation Language for Data-Driven Materials Science)

Teaser

A major achievement within the past 18 months was the successful application of the Discrete-To-Continuum (D2C) method and the way it is used to characterize dislocation microstructures within a machine learning context. Using this method we have been able to classify the...

Summary

A major achievement within the past 18 months was the successful application of the Discrete-To-Continuum (D2C) method and the way it is used to characterize dislocation microstructures within a machine learning context. Using this method we have been able to classify the relaxed microstructure of different sample sizes between 30 and 90 nm. In other words: we can now differentiate between different microstructures, which is a prerequisite for being able to search a data base or to compare different samples.

A second objective was the manual digitization of TEM images. This has been successfully achieved and furthermore, the resulting stress fields could be analyzed in great details. This makes information available that so far have not been accessible for TEM microscopy. A publication in under way.

We furthermore set the foundation for automatic analysis of TEM images using maching learning. This is work in progress and will become one of the strong point of the project.

Work performed

\"- learning python/C++
- development of an elastic eigenstrain solver
- creating of initial dislocation structures from DDD and MD
- coarse graining and analysis of dislocation structures
- first steps towards \"\"machine learning dislocations\"\"
\"

Final results

We were the first to use machine learning to statistically analyse and characterize dislocation microstructures

Website & more info

More info: https://tu-freiberg.de/fakult4/imfd/mimm.