Machine learning powered multi-method imaging workflows for advanced material and failure analysis

Machine learning powered imaging workflows for advanced material and failure analysis Imaging-based methods represent highly valuable tools for material and failure analysis. In particular the combination of experimental methods with artificial intelligence (AI) provides enhanced possibilities in terms of efficient and fast quantification of the microstructure as well as to gain an improved understanding of the underlying structure-property relationship. Here, we provide an overview of our recent studies [1-6] and beyond. In detail, we show how advanced two- and three-dimensional imaging methods like scanning electron microscopy (SEM), focused ion beam (FIB)-SEM, X-ray computed tomography and scanning acoustic microscopy can be combined with artificial intelligence-based image analysis workflows to gain improved insights. We illustrate how image enhancement algorithm based on deep learning and generative AI as well as semantic segmentation and object detection foster the extraction of microstructure features. Enhanced segmentation with an accuracy of about 90% and above based on an efficient annotation technique provides the basis for accurate microstructure quantification, prediction and understanding of the correlation of extracted microstructure features and material properties. Further, we show how generative AI-based diffusion probabilistic models provide possibilities for the synthetic reconstruction of microstructure images. This introduced framework enables the high-fidelity extrapolation of morphological changes for unmeasured experimental points, essentially filling the gaps in sparse physical data. It displays an essential step towards “Digital Twin” material science, beyond common experimental workflows. [1] C. Cui, R. Brunner et al. npj materials degradation https://doi.org/10.1038/s41529-024-00456-8 [2] C. Cui, R. Brunner et al. npj materials degradation, https://doi.org/10.1038/s41529-025-00603-9 [3] A. Wijaya, R. Brunner et al. communications materials https://doi.org/10.1038/s43246-024-00493-5 [4] R. Wilhelmer, R. Brunner et al. Scientific Reports, https://doi.org/10.1038/s41598-025-08308-4 [5] C. Cui, R. Sinojiya, R. Brunner et al. Materials Advances https://doi.org/10.1039/D5MA01192B [6] R. Sinojiya, R. Brunner et al. Communications Materials, https://doi.org/10.1038/s43246-023-00339-6