At the first Molten Salts and Corrosion in Extreme Environments Workshop, held this March in Lanzarote, Spain, the scientists presented research results focused on the automated evaluation of images of aluminide-diffusion coatings taken with a Scanning Electron Microscope (SEM).
To obtain sufficient high-quality data for machine learning, the researchers use Blender 3D software to generate synthetic images to augment their data. These synthetic images, along with verified data, are used to train a Deep Learning model for SEM image segmentation based on the U-Net architecture, which allows for the recognition of specific coating parameters. The model is subsequently applied to the evaluation of SEM images of diffusion coatings produced from three different slurry compositions.
The research results are published in the article Segmentation and Metallographic Evaluation of Aluminium Slurry Coatings Using Machine Learning Techniques in the journal High Temperature Corrosion of Metals.
At the 51st International Conference on Metallurgical Coatings and Thin Films, ICMCTF 2025, held in May in San Diego, scientists presented the results of their collaborative research, which uses machine learning to evaluate the operational aging and service time of aluminide-diffusion high-temperature corrosion-resistant coatings. The study explores the potential of machine learning in predicting the time-dependent behaviour of key parameters of the coating system. A publication detailing these findings is currently being prepared for submission to a scientific journal.