Large-area, high-resolution characterisation and classification of damage mechanisms in dual-phase steel using deep learning

Abstract

High performance materials, from natural bone over ancient damascene steel to modern superalloys, typically possess a complex structure at the microscale. Their properties exceed those of the individual components and their knowledge-based improvement therefore requires understanding beyond that of the components’ individual behaviour. Electron microscopy has been instrumental in unravelling the most important mechanisms of co-deformation and in-situ deformation experiments have emerged as a popular and accessible technique. However, a challenge remains: to achieve high spatial resolution and statistical relevance in combination. Here, we overcome this limitation by using panoramic imaging and machine learning to study damage in a dual-phase steel. This high-throughput approach now gives us strain and microstructure dependent insights into the prevalent damage mechanisms and allows the separation of inclusions and deformation–induced damage across a large area of this heterogeneous material. Aiming for the first time at automated classification of the majority of damage sites rather than only the most distinct, the new method also encourages us to expand current research past interpretation of exemplary cases of distinct damage sites towards the less clear-cut reality.