In this paper, various kinds of applications are presented, in which tomographic image data depicting microstructures of materials are semantically segmented by combining machine learning methods and conventional image processing steps. The main focus of this paper is the grain-wise segmentation of time-resolved CT data of an AlCu specimen which was obtained in between several Ostwald ripening steps. The poorly visible grain boundaries in 3D CT data were enhanced using convolutional neural networks (CNNs). The CNN architectures considered in this paper are a 2D U-Net, a multichannel 2D U-Net and a 3D U-Net where the latter was trained at a lower resolution due to memory limitations. For training the CNNs, ground truth information was derived from 3D X-ray diffraction (3DXRD) measurements. The grain boundary images enhanced by the CNNs were then segmented using a marker-based watershed algorithm with an additional postprocessing step for reducing oversegmentation. The segmentation results obtained by this procedure were quantitatively compared to ground truth information derived by the 3DXRD measurements. A quantitative comparison between segmentation results indicates that the 3D U-Net performs best among the considered U-Net architectures. Additionally, a scenario, in which "ground truth" data is only available in one time step, is considered. Therefore, a CNN was trained only with CT and 3DXRD data from the last measured time step. The trained network and the image processing steps were then applied to the entire series of CT scans. The resulting segmentations exhibited a similar quality compared to those obtained by the network which was trained with the entire series of CT scans.
It is an open question how the particle microstructure of a lithium-ion electrode influences a potential thermal runaway. In order to investigate this, information on the structural changes, in particular cracked particles, caused by the failure are desirable. For a reliable analysis of these changes a reasonably large amount of data is necessary, which necessitates automatic extraction of particle cracks from tomographic 3D image data. In this paper, a classification model is proposed which is able to decide whether a pair of particles is the result of breakage, of the image segmentation, or neither. The classifier is developed using simulated data based on a 3D stochastic particle model. Its validity is tested by applying the methodology to hand-labelled data from a real electrode. For this dataset, an overall accuracy of 73% is achieved.
Far-field three-dimensional X-ray diffraction microscopy allows for quick measurement of the centers of mass and volumes of a large number of grains in a polycrystalline material, along with their crystal lattice orientations and internal stresses. However, the grain boundaries—and, therefore, individual grain shapes—are not observed directly. The present paper aims to overcome this shortcoming by reconstructing grain shapes based only on the incomplete morphological data described above. To this end, cross-entropy (CE) optimization is employed to find a Laguerre tessellation that minimizes the discrepancy between its centers of mass and cell sizes and those of the measured grain data. The proposed algorithm is highly parallel and is thus capable of handling many grains (>8,000). The validity and stability of the CE approach are verified on simulated and experimental datasets.
Random tessellations are well suited for the probabilistic modeling of three-dimensional (3D) grain microstructure of polycrystalline metals. The present paper deals with so-called Gibbs-Laguerre tessellations where the generators of a Laguerre tessellation form a Gibbs point process. The goal is to construct an energy function of the Gibbs point process from a suitable set of potentials, such that the resulting Gibbs-Laguerre tessellation matches some desired geometrical properties. Since the model is analytically hardly tractable, our main tool of analysis are stochastic simulations based on Markov chain Monte Carlo. These enable us to investigate the properties of the models, and, in the next step, to apply the thus gained knowledge to do a statistical reconstruction of an aluminum alloy based on 3D tomographic image data.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.