From mechanism-based to data-driven approaches in materials science

Abstract

Abstract A time-honored approach in theoretical materials science revolves around the search for basic mechanisms that should incorporate key feature of the phenomenon under investigation. Recent years have witnessed an explosion across areas of science of a data-driven approach fueled by recent advances in machine learning. Here we provide a brief perspective on the strengths and weaknesses of mechanism based and data-driven approaches in the context of the mechanics of materials. We discuss recent literature on dislocation dynamics, atomistic plasticity in glasses focusing on the empirical discovery of governing equations through artificial intelligence. We conclude highlighting the main open issues and suggesting possible improvements and future trajectories in the fields.

Keywords

• Machine learning
• Perspective
• Dislocation dynamics
• Plasticity in glasses
• Equation discovery