Machine learning for plasticity
Markus Stricker Ruhr-Universität Bochum, Germany
The understanding and design of the microstructure of metals is one of the cornerstones of materials science. Microstructural design necessitates atomistic and mesoscale simulations methods in order to resolve defect dynamics at their relevant length and time scales, often not directly accessible in experiments. Here, we present two approaches for understanding dislocation behavior: a neural network interatomic potential potential for pure Magnesium and a study of data fusion from experiment and discrete dislocation dynamics. Both approaches provide new possibilities assess dislocation behavior which can be used for guiding the discovery of materials and processing routes. The approaches are subsequently put into a larger perspective of microstructure and alloy design in current and future developments.