New research initiative on microstructural phase identification using Vision Transformers

Within the framework of the DIGEST project, researchers at Universidad Politécnica de Madrid have launched a new line of investigation under Work Package 2 of Subproject 1, aiming to enhance the microstructural characterization of metallic materials.

Understanding the microstructural phases of metallic alloys plays a crucial role in predicting their wear behaviour and service degradation, both of which are key to improving the accuracy of predictive maintenance models. However, current methods for identifying phases in metallographic images often rely on manual analysis or traditional machine vision techniques, which may lack generalization across complex structures and industrially relevant datasets.

To address this challenge, the research team will explore the use of Vision Transformers and Deep Transfer Learning to automatically identify and classify phases in metallographic micrographs. This approach is expected to improve recognition accuracy, reduce annotation requirements, and enable scalable deployment in digital inspection pipelines.

This activity is being developed in close collaboration with the AGH University of Krakow (Poland), through the participation of a visiting researcher from the team led by Professor Dorota Wilk-Kołodziejczyk, whose group has long-standing expertise in metallurgical analysis and materials informatics.

This joint effort exemplifies DIGEST’s commitment to advancing smart maintenance by integrating AI technologies with deep materials knowledge.