Development of advanced industrial steels with enhanced mechanical properties, requires to understand the kinetics of phase transformations, particularly that of austenite-to-ferrite. A better knowledge of the role of substitutional alloying elements on ferrite growth, coupled with the effect of interstitial carbon, and their interactions with the migrating ferrite/austenite interface is required.
Aiming at addressing this objective in a systematic way, we have developed a complete combinatorial high-throughput methodology consisting in combining compositionally graded samples with time- and space-resolved in situ X- ray diffraction. First, diffusion couples containing millimeter-scale solute gradients were produced. Secondly, these compositionally graded materials were continuously scanned by a synchrotron high-energy X-ray beam during in-situ heat treatments to obtain phase transformation kinetic maps in composition space for ferrite growth in both isothermal and non-isothermal paths. In this way, large amounts of kinetic data were collected for different compositions and thermal paths.
Finally, this large dataset was used to validate & optimize a modified version of the three-jump solute drag model. This model has been successfully used to predict the kinetics for multiple substitutional solutes, compositions and temperatures in ternary and quaternary systems. One limitation of modelling such large datasets with multiple compositions and non-isothermal conditions is the computing time required especially for precise thermodynamic calculations. Model acceleration was reached by using multiple approaches of machine learning. These approaches proved to be effective in finding a model for predicting the austenite-to-ferrite phase transformation with much lower computing time.