3D imaging techniques are a valuable method in material sciences for analysis and design of materials. Using μCT and FIB-SEM devices, it is possible and easy to obtain images at a resolution well suited for micro scale simulations and to investigate the impact of microstructural morphology on material properties.
Even before starting to run simulations (for example, flow or mechanics simulations) on these images, valuable insights can be obtained from the analysis of the geometry of the scanned materials. In GeoDict 2022, the capabilities for geometry analysis have been improved further. For the segmentation of μCT and FIB-SEM images, we extended the trainable AI segmentation of the ImportGeo-VOL module. Now, more data can be labeled for the deep learning-based methods, still keeping the training times at a reasonable range. Besides this, we added the GeoDict-AI module for easy training of customized neural network from GeoDict generated material models. The training can be aimed to the segmentation of material phases or the labeling of individual objects (e.g., fibers) in μCT-images.
In the recent years, advances in synchrotron and μCT imaging have pushed the times required to take these images down to a time scale where 4D imaging is becoming more and more available. These kinds of image series can be also analyzed with GeoDict, and objects can be tracked over time, with the possibility of close comparison of in-situ experiments and simulations.