Speakers
Description
Thermomechanical processing (TMP) of steels is strongly influenced by chemical composition through its effect on softening, precipitation response and deformation resistance. Modern metallurgical models therefore aim to describe material behavior directly as a function of composition and process conditions, moving away from predefined steel classes.
In this work, a comprehensive dataset of industrial steel chemistries is analyzed in order to explore composition-derived metallurgical descriptor spaces relevant for TMP. Starting from the chemical composition, several physically meaningful descriptors are calculated, including empirical hot-deformation parameters, transformation temperatures (A1 and A3), carbon equivalent indicators and micro alloy stoichiometry ratios.
While the metallurgical models themselves remain composition-based, these derived descriptors provide an alternative representation of the metallurgical state space that is directly related to thermomechanical processing behavior. Unsupervised machine learning methods such as clustering and exploratory visualization methods are used to analyze the structure of this descriptor space and to identify regions associated with distinct processing characteristics, for example differences in rolling resistance, transformation behavior or precipitation potential.
The results show that the descriptor-based representation reveals clear structures within large industrial composition datasets and can support the interpretation and calibration of metallurgical models used in digital process simulations.
The proposed framework provides a practical tool for analyzing extensive steel chemistry datasets and for linking chemical composition to thermomechanical processing behavior in a systematic and data-assisted way.