Steel properties depend strongly on grain size, which has so far been evaluated by visual inspection. A model developed by Schaeffler and Fraunhofer IWM now promises a more reliable classification.
Deep learning classifies microstructure images into different grain size ranges.
Fraunhofer IWM
Researchers at the Fraunhofer Institute for Mechanics of Materials IWM, in collaboration with Schaeffler Technologies, have developed a deep learning model for grain size determination in martensitic and bainitic steels. It is intended to supplement or replace the previous time-consuming visual inspection by trained metallographers, whose results are too subjective, inaccurate and thus not sufficiently reliable, especially for safety-relevant applications, as a ring study conducted by the partners has shown.
According to the partners, the developed model can evaluate component areas of any size, and in doing so can consistently evaluate grain sizes in an objective, automated and reproducible manner. To this end, it was trained in advance with classified image data. In addition to grain size, the model can distinguish between martensitic and bainitic states as well as between different steel alloys of the 100Cr6 and C56 families. The model is currently being implemented at Schaeffler.