Frank Mücklich

Current conventional methods of evaluating microstructures are characterized by a high degree of subjectivity and a lack of reproducibility. Modern machine learning (ML) approaches have already shown great potential in overcoming these challenges. Once trained with representative data in combination with objective ground truth, the ML model is able to perform a task properly in a reproducible and automated manner. However, in highly complex use cases, it is often not possible to create a definite ground truth. This study addresses this problem using the underlying showcase of microstructures of highly complex quenched and quenched and tempered (Q/QT) steels. A patch-wise classification approach combined with a sliding window technique provides a solution for segmenting entire microphotographs where pixel-wise segmentation is not applicable since it is hardly feasible to create reproducible training masks. Using correlative microscopy, consisting of light optical microscope (LOM) and scanning electron microscope (SEM) micrographs, as well as corresponding data from electron backscatter diffraction (EBSD), a training dataset of reference states that covers a wide range of microstructures was acquired in order to train accurate and robust ML models in order to classify LOM or SEM images. Despite the enormous complexity associated with the steels treated here, classification accuracies of 88.8% in the case of LOM images and 93.7% for high-resolution SEM images were achieved. These high accuracies are close to super-human performance, especially in consideration of the reproducibility of the automated ML approaches compared to conventional methods based on subjective evaluations through experts.

The self-propagating reaction in binary Al/Ru multilayers with two different bilayer thicknesses (89 and 178 nm, respectively) forming single-phase AlRu intermetallic compound is investigated experimentally and by means of phase-field simulations. Experimentally, the time-temperature evolution in free-standing films was recorded with a high-speed pyrometer, and the resulting microstructure was determined from electron backscatter diffraction measurements. The phase-field model is constructed based on the minimization of the grand potentials for which the required thermodynamic data are taken from the Calphad database. The simulation of the reaction and subsequent AlRu grain growth starts from Al-rich and Ru-rich layer fillings. After the formation of the AlRu phase is complete, the grain growth during cooling is simulated based on the experimentally recorded time-temperature curves. Finally, the resulting grain sizes obtained from the simulation are found to be in good agreement with the experimental results. Furthermore, it is shown that the final grain sizes in both simulations and experiments depend on the initial bilayer thicknesses.