This work studies the effect of chemical heterogeneities on bainitic phase transformation during continuous cooling in a systematic way. A high-throughput combinatorial method is used to map the austenite decomposition kinetics in composition space of substitutional solutes. The method relies on in-situ time- and space-resolved high-energy (synchrotron) X-ray diffraction and compositionally graded samples. Varying conditions of carbon content, austenitic grain size and cooling rate are applied.
The large amount of data collected are used to build and train a machine learning model predicting the transformation kinetics and final microstructure for different segregation levels and conditions. The results are then integrated into representations of actual pieces of nuclear vessel steel forgings presenting mesoscale chemical heterogeneities inherited from segregation of substitutional elements during ingot solidification. 
A thorough characterization of the final microstructures is carried out and correlated to kinetics sequence in order to gain physical understanding of the bainite formation sequence in a non-uniform composition field for different temperature ranges. Orientation relationship between parent and child grains are investigated with (high resolution) EBSD and the pairing and spatial distribution of crystallographic variants are quantified, allowing to identify two dominant transformation regimes. A physical model for mechanical stabilization of austenite during transformation is developed.
Finally, the effect of local composition on carbon redistribution during tempering and post-welding stress relief treatment is investigated. The mechanisms for carbon enrichment in segregated zones are presented, along with carbides population characterization and diffusion calculations.