119092 research outputs found
Sort by
Combining grain boundary diffusion and segmentation: a novel production route for resource-efficient Nd–Fe–B magnets
No full textTo enhance coercivity in high-performance Nd-Fe-B magnets, resource critical heavy rare earth elements Dy and Tb are used in the grain boundary diffusion process (GBDP). However, GBDP is limited to thin magnets, typically around 4 mm which restricts their applications. Here, we report a novel processing route by combining the GBDP process with a magnet stacking architecture using a < 10 μm thin low-melting multi-element Tb₁₀Pr₆₀Al₁₀Cu₁₀Ga₁₀ alloy which has now a dual function: it acts as a binder between the magnet segments and is at the same time an efficient source for core-shell formation on the individual crystallite level. This not only enables production of magnets with any thickness without losing performance, but also paves the way for resource efficiency using cheap and abundant light rare earth Cerium in hybrid magnets by strategically stacking different grades with tailored chemical compositions resulting in a unique macroscopic magnetic hardening
Optimizing α′′ -Fe16N2 as a permanent magnet via alloying
No full textBased on first-principles calculations, we investigate the effects of 27 alloying elements on the intrinsic magnetic properties of a′′-Fe16N2, in order to further optimize its properties for permanent magnet applications. Analysis on the thermodynamic stabilities based on formation energy and distance to the convex hull reveals that 20 elements can be substituted into Fe16N2, where there is no strong site-preference upon doping. It is observed that all alloying elements can essentially reduce the saturation magnetization, whereas the magnetic anisotropy can be significantly modified. In terms of the Boltzmann- average intrinsic properties, we identify 8 elements as interesting candidates, with Co, Mo, and W as the most promising cases for further experimental validations
Advancing the quantification of land-use intensity in forests: the ForMIX index combining tree species composition, tree removal, deadwood availability, and stand maturity
No full textMany forests have a long history of human land use, which shapes species communities and ecosystem processes, making robust and quantitative measures of land-use intensity in forests desirable. We here introduce the ForMIX (Forest Management IndeX), a compound index combining altered tree species composition, tree removal, deadwood availability and stand maturity, which are each calculated as the deviation from expectations in an unmanaged old-growth forest reference. The index and its components allow for mechanistic inference on the consequences of land use in forests as they are based on biotic resources and niches directly affected by forest land use. Using basic forest inventory data from 150 sites distributed over three regions of Germany, we demonstrate the properties of ForMIX, which differentiates well among forest types and silvicultural systems and is robust to decisions regarding reference values and components. Reference values used in ForMIX are dynamic, could be adapted to ongoing climate change and may require refinement for different geographic regions. ForMIX advances the quantification of land-use intensity in forests by being biologically meaningful, usable and comparable across forest types, derivable from standard forest inventory data, and easy to apply, understand and interpret
Knowledge engineering with large language models: accelerating fuzzy rule bases development for energy-aware expert systems
No full textExpert systems offer a promising way to automatically identify energy efficiency potentials in industry and thereby contribute to energy cost savings and decarbonization. In these systems, domain-specific knowledge is embedded and linked to automated analyses of measurement data. Until now, knowledge engineers have extracted, structured, and represented the necessary domain-specific knowledge in a form usable by expert systems, which is time-consuming and costly. This article presents a hybrid approach that couples expert systems with large language models to support the work of knowledge engineers. Energy performance indicators, selected by the energy manager, serve to quantify changes in energy performance and reproduce the heuristic decision-making of human experts on a quantitative basis. These indicators then form the basis for a rule set that targets areas with the highest potential energy savings. For practical implementation, a fuzzy rule base is applied because it captures decisions made under imprecise information and allows conditions and conclusions that can be partially true or false. Building the fuzzy rule base involves assigning membership functions to input and output variables and defining their linguistic partitioning, since these choices shape both sensitivity and interpretability. The rule base is implemented as generally understandable IF–THEN rules. The premise consists of energy performance indicators that are associated with linguistic variables and combined using logical operators. The conclusion contains priority numbers, which are also associated with linguistic variables and express the energy efficiency potential. In the hybrid setup presented in this article, large language models formalize given energy performance indicators and fuzzy rules, propose membership functions to populate the fuzzy rule base, and generate visualization scripts in Python. This leads to accelerated development while preserving transparent, comprehensible, and reproducible decision logic characteristic of expert systems. The approach is demonstrated using a foam panel production line in the chemical industry
Optimizing autonomous multi-view stereo scans using AI based image masking within cultural heritage digitization
No full textImage masking is essential in the field of 3D reconstruction for cultural heritage objects. It is used to accelerate the reconstruction process by removing background noise and accurately reconstructing the object only. The autonomous iterative Multi-View Stereo 3D scanner from the Fraunhofer Institute for Computer Graphics Research, within the Cultural Heritage Digitization department, requires binary masks to scan objects efficiently, regardless of the surrounding environment, geometry or color of the object, background color, and stabilizing mount. However, conventional masking methods can produce incorrect masks, leading to an inefficient or even abortive scan. Until now, these cases have been solved by parameter optimizations of the conventional masking method or changes in the scanning environment. This does not align with the principles of automation, since non-technical users in museums, archives, etc. should be able to use the autonomous iterative scanning workflow without additional effort. In addition to the real scan data used for training the presented networks, an automated Blender pipeline is also introduced, which generates additional synthetic data for training. Therefore, we evaluate if the latest stateof-the-art artificial intelligence segmentation methods can be used for these challenging cases without compromising their performance in simpler scenarios. This paper shows that with the proper network and datasets, masks of difficult objects or scenarios can be generated that can be used within the autonomous iterative scanning workflow. Thus, parameter and environment optimizations are no longer necessary
Time dependent effects of concrete decks on the cyclic soil-structure interaction of integral bridges
No full text
Patches of Nonlinearity: Instruction Vectors in Large Language Models
No full textDespite the recent success of instruction-tuned language models and their ubiquitous usage, very little is known of how models process instructions internally. In this work, we address this gap from a mechanistic point of view by investigating how instruction-specific representations are constructed and utilized in different stages of post-training: Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO). Via causal mediation, we identify that instruction representation is fairly localized in models. These representations, which we call Instruction Vectors (IVs), demonstrate a curious juxtaposition of linear separability along with non-linear causal interaction, broadly questioning the scope of the linear representation hypothesis commonplace in mechanistic interpretability. To disentangle the non-linear causal interaction, we propose a novel method to localize information processing in language models that is free from the implicit linear assumptions of patching-based techniques. We find that, conditioned on the task representations formed in the early layers, different information pathways are selected in the later layers to solve that task, i.e., IVs act as circuit selectors
Evaluating spin–orbit effects on the thermochemistry of proton-coupled electron transfer
No full textImplementation of inter-factory connectivity for demand-oriented data analysis and visualization
No full textMulti-factory production boosts the global competitiveness of manufacturing companies. However, incompatible IT systems and limited information sharing hinder inter-factory communication. This paper presents case study on developing a technical solution for data exchange and analysis along a multi-factory production following a process model based on a value stream-oriented analysis of information needs. As a result, a demonstrator across three factory facilities is built, encompassing a traceability terminal, 3D printing, milling, and assembly processes, seamlessly integrated with an Internet of Things (IoT) platform offering diverse demand-oriented services. Subsequently, the multi-factory demonstrator is evaluated in production