Bild

Konferens: Datadriven mekanik och materialfysik
Naturvetenskap & IT
Euromech Colloquium 656
Euromech Colloquium 656
Combining classical principles of mechanics and physics of solids with cutting-edge data science techniques has resulted in very accurate and efficient data-driven approaches. Data-driven models have a very high capacity to harness vast volumes of data generated in material science, engineering, and physics to uncover hidden patterns, relationships, and insights. During the last decade, the following have seen key developments, and further enhancements are ongoing:
Despite the great progress, challenges persist, including (i) the need for robust data sets, (ii) the interpretability of complex models, and (iii) the integration of physics-based constraints. Overcoming these challenges will pave the way for a deeper understanding of material behavior and the realization of more efficient and sustainable engineering solutions.
In this colloquium, topics of interest include (not limited to):
● Traditional and physics-enhanced machine learning for surrogate modelling;
● Learning and exploiting latent representations of materials behavior;
● Image-based machine learning methods;
● Data-driven process modelling of materials;
● Generative learning for material optimization;
● Micromechanics-based data-driven methods for designing materials;
● Bayesian machine learning for uncertainty quantification, data assimilation and inverse modeling;
● Manifold learning for reduced-order modeling of material behavior;
● Efficient data assimilation and active learning, among others.