Open Access
Refine
Document Type
- Article (10) (remove)
Has Fulltext
- yes (10) (remove)
Is part of the Bibliography
- no (10) (remove)
Keywords
Institute
Based on a data-driven approach, a computer-assisted workflow for the quantitative analysis of optical Kerr microscopy images of sintered FeNdB-type permanent magnets was developed. By analyzing the domain patterns visible in the Kerr image with data-driven approaches such as traditional machine learning and advanced deep learning, we can quantify grain orientation and size with a better trade-off between accuracy and higher throughput than electron backscatter diffraction (EBSD). The key distinction between traditional machine learning and advanced deep learning lies in feature extraction. Traditional methods require manual, user-dependent feature extraction from input data, while advanced deep learning achieves this automatically. The predictions from the trained models were compared to the measurements from EBSD for performance evaluation. The proposed data-driven model is trained on the dataset created from the correlative microscopy technique, which requires the images of grains extracted from the Kerr microscopy and corresponding EBSD grain orientation data (Euler angles). The fine-tuned deep learning model shows better generalization ability than the traditional machine learning models trained on the manually extracted features and resulted in a mean absolute error of less than 5° for grain orientation of the anisotropic magnet samples when evaluated against the measured EBSD values. The developed approach has reduced the measurement effort for grain orientation by 5 times and have sufficient accuracy when compared to the EBSD.
In this study, we investigate the use of artificial neural networks as a potentially efficient method to determine the rate capability of electrodes for lithium-ion batteries with different porosities. The performance of a lithium-ion battery is, to a large extent, determined by the microstructure (i.e., layer thickness and porosity) of its electrodes. Tailoring the microstructure to a specific application is a crucial process in battery development. However, unravelling the complex correlations between microstructure and rate performance using either experiments or simulations is time-consuming and costly. Our approach provides a swift method for predicting the rate capability of battery electrodes by using machine learning on microstructural images of electrode cross-sections. We train multiple models in order to predict the specific capacity based on the batteries’ microstructure and investigate the decisive parts of the microstructure through the use of explainable artificial intelligence (XAI) methods. Our study shows that even comparably small neural network architectures are capable of providing state-of-the-art prediction results. In addition to this, our XAI studies demonstrate that the models are using understandable human features while ignoring present artefacts.
Die Begriffe künstliche Intelligenz und digitaler Zwilling sind prägende Themen der heutigen Zeit. Zwar beeinflussen beide Themen die Entwicklung der Gesellschaft auf verschiedene Arten und Weisen. Dennoch sind sie als Untersuchungsobjekt schlecht definiert und es herrschen etliche unterschiedliche Verständnisse dieser Begriffe.
Der vorliegende Aufsatz nähert sich dem Begriffen künstliche Intelligenz und digitaler Zwilling durch eine Literaturanalyse, um das Verständnis dieser Begriffe in ihrer Nutzung zu schärfen. Im Anschluss der Begriffsbetrachtung wird für jeden der zwei Begriffe betriebliche Anwendungsbereiche aufgezeigt. Folgend den Einzelbetrachtungen werden mögliche Forschungsrichtungen aufgezeigt und diskutiert.
Laser melting manufacturing of large elements of lunar regolith simulant for paving on the Moon
(2023)
The next steps for the expansion of the human presence in the solar system will be taken on the Moon. However, due to the low lunar gravity, the suspended dust generated when lunar rovers move across the lunar soil is a significant risk for lunar missions as it can affect the systems of the exploration vehicles. One solution to mitigate this problem is the construction of roads and landing pads on the Moon. In addition, to increase the sustainability of future lunar missions, in-situ resource utilization (ISRU) techniques must be developed. In this paper, the use of concentrated light for paving on the Moon by melting the lunar regolith is investigated. As a substitute of the concentrated sunlight, a high-power CO2 laser is used in the experiments. With this set-up, a maximum laser spot diameter of 100 mm can be achieved, which translates in high thicknesses of the consolidated layers. Furthermore, the lunar regolith simulant EAC-1A is used as a substitute of the actual lunar soil. At the end of the study, large samples (approximately 250 × 250 mm) with interlocking capabilities were fabricated by melting the lunar simulant with the laser directly on the powder bed. Large areas of lunar soil can be covered with these samples and serve as roads and landing pads, decreasing the propagation of lunar dust. These manufactured samples were analysed regarding their mineralogical composition, internal structure and mechanical properties.
Future lunar exploration will be based on in-situ resource utilization (ISRU) techniques. The most abundant raw material on the Moon is lunar regolith, which, however, is very scarce on Earth, making the study of simulants a necessity. The objective of this study is to characterize and investigate the sintering behavior of EAC-1A lunar regolith simulant. The characterization of the simulant included the determination of the phase assemblage, characteristic temperatures determination and water content analysis. The results are discussed in the context of sintering experiments of EAC-1A simulant, which showed that the material can be sintered to a relative density close to 90%, but only within a very narrow range of temperatures (20–30 °C). Sintering experiments were performed for sieved and unsieved, as well as for dried and non-dried specimens of EAC-1A. In addition, an analysis of the densification and mechanical properties of the sintered specimens was done. The sintering experiments at different temperatures showed that the finest fraction of sieved simulant can reach a higher maximum sintering temperature, and consequently a higher densification and biaxial strength. The non-dried powder exhibited higher densification and biaxial strength after sintering compared to the dried specimen. This difference was explained with a higher green density of the non-dried powder during pressing, rather than due to an actual influence on the sintering mechanism. Nevertheless, drying the powder prior to sintering is important to avoid the overestimation of the strength of specimens to be fabricated on the Moon.
Soft magnetic Fe-Al alloys have been a subject of research in the past. However, they never saw the same reception in technical applications as the Fe-Si or Fe-Ni alloys, which is, to some extent, due to a low ductility level and difficulties in manufacturing. Additive manufacturing (AM) technology could be a way to avoid issues in conventional manufacturing and produce soft magnetic components from these alloys, as has already been shown with similarly brittle Fe-Si alloys. While AM has already been applied to certain Fe-Al alloys, no magnetic properties of AM Fe-Al alloys have been reported in the literature so far. Therefore, in this work, a Fe-12Al alloy was additively manufactured through laser powder bed fusion (L-PBF) and characterized regarding its microstructure and magnetic properties. A comparison was made with the materials produced by casting and rolling, prepared from melts with an identical chemical composition. In order to improve the magnetic properties, a heat treatment at a higher temperature (1300 °C) than typically applied for conventionally manufactured materials (850–1150 °C) is proposed for the AM material. The specially heat-treated AM material reached values (HC: 11.3 A/m; µmax: 13.1 × 103) that were close to the heat-treated cast material (HC: 12.4 A/m; µmax: 20.3 × 103). While the DC magnetic values of hot- and cold-rolled materials (HC: 3.2 to 4.1 A/m; µmax: 36.6 to 40.4 × 103) were not met, the AM material actually showed fewer losses than the rolled material under AC conditions. One explanation for this effect can be domain refinement effects. This study shows that it is possible to additively manufacture Fe-Al alloys with good soft magnetic behavior. With optimized manufacturing and post-processing, further improvements of the magnetic properties of AM L-PBF Fe-12Al may still be possible.
In the fast-growing but also highly competitive market of battery-powered power tools, cell-pack-cooling systems are of high importance, as they guarantee safety and short charging times. A simulation model of an 18 V power tool battery pack was developed to be able to evaluate four different pack-cooling systems (two heat-conductive polymers, one phase change material, and non-convective air as reference) in an application scenario of practical relevance (the intensive use of a power tool followed by cooling down and charging steps). The simulation comprises battery models of 21700 cells that are commercially available as well as heat transfer models. The study highlights the performance of the different cooling materials and their effect on the maximum pack temperature and total charging cycle time. Key material parameters and their influence on the battery pack temperature and temperature homogeneity are discussed. Using phase change materials and heat-conductive polymers, a significantly lower maximum temperature during discharge (up to 26%) and a high shortening potential of the use/charging cycle (up to 32%) were shown. In addition to the cooling material sweep, a parameter sweep was performed, varying the external temperature and air movement. The high importance of the conditions of use on the cooling system’s performance was illustrated.
The surface topography of biodegradable polymer foils is modified by mechanical imprinting on a submillimeter length scale. The created patterns strongly influence the wetting behavior and allow the preparation of hydrophobic surfaces with controlled solid-liquid interaction. A detailed analysis of anisotropic surface patterns reveals that the observed effect arises from a combination of topographical and compositional changes that are introduced to the surface. As a main result it is found that an individual combination of material and structure is required for the production of water-repellent biopolymer foils that are highly attractive for packaging applications.
Transformations in the work–nonwork interface highlight the importance of effectively managing the boundaries between life domains. However, do the ways individuals manage the boundaries between work and nonwork life change from one day to the next? If so, which antecedents may explain these intra-individual fluctuations in boundary management? Drawing on boundary management, spillover, and resource theories, we investigate daily changes in segmentation preferences and integration enactments as a function of experiencing strain in work and nonwork life. Assuming that changes in segmentation preferences reflect an individual’s strategy to regulate negative cross-role spillover, we suppose that strain increases individuals’ segmentation preferences; at the same time, however, it could force individuals to enact more integration.
Identification and quantitative segmentation of individual blood vessels in mice visualized with preclinical imaging techniques is a tedious, manual or semiautomated task that can require weeks of reviewing hundreds of levels of individual data sets. Preclinical imaging, such as micro-magnetic resonance imaging (μMRI) can produce tomographic datasets of murine vasculature across length scales and organs, which is of outmost importance to study tumor progression, angiogenesis, or vascular risk factors for diseases such as Alzheimer’s. Training a neural network capable of accurate segmentation results requires a sufficiently large amount of labelled data, which takes a long time to compile. Recently, several reasonably automated approaches have emerged in the preclinical context but still require significant manual input and are less accurate than the deep learning approach presented in this paper—quantified by the Dice score. In this work, the implementation of a shallow, three-dimensional U-Net architecture for the segmentation of vessels in murine brains is presented, which is (1) open-source, (2) can be achieved with a small dataset (in this work only 8 μMRI imaging stacks of mouse brains were available), and (3) requires only a small subset of labelled training data. The presented model is evaluated together with two post-processing methodologies using a cross-validation, which results in an average Dice score of 61.34% in its best setup. The results show, that the methodology is able to detect blood vessels faster and more reliably compared to state-of-the-art vesselness filters with an average Dice score of 43.88% for the used dataset.