The combination of correlative microscopy methods and machine learning (ML)-based micrograph segmentation allows an objective, reproducible and automated microstructure analysis. Large datasets based on systematic variation of processing parameters can be evaluated in high-throughput microstructure quantification models, laying the foundation for correlation of said processing parameters with microstructure characteristics, including phases and phase fractions as well as morphology. Correlative microscopy techniques combine light optical and scanning electron microscopy with EBSD measurements, ensuring a precise and scale-bridging microstructure quantification.

For a systematic analysis of correlations between microstructure features, processing parameters, and material properties, ML, feature engineering, and data science based methods are powerful tools allowing investigative assessment and modelling of relationships for complex steel microstructures where conventional deterministic or thermodynamic models attain their limits. This approach seeks to gain a deep understanding of the underlying phase transformation processes by analyzing correlations between microstructure characteristics before and after phase transformations. These microstructure-based ML models that can predict phase transformations as well as resulting phases and morphologies are a cornerstone of understanding phase transformation behavior with an unprecedented level of detail.