VP Vizepräsident
Filtern
Dokumenttyp
- Zeitschriftenartikel (52)
- Vortrag (36)
- Forschungsdatensatz (18)
- Posterpräsentation (16)
- Beitrag zu einem Tagungsband (5)
- Sonstiges (5)
- Preprint (4)
- Buchkapitel (1)
- Video (1)
Schlagworte
- Machine Learning (15)
- Mass spectrometry (15)
- Bioinformatics (14)
- Automation (13)
- Database (12)
- Bonding analysis (11)
- Computational Chemistry (11)
- Machine learning (11)
- Materials Informatics (11)
- Metaproteomics (9)
Organisationseinheit der BAM
- VP Vizepräsident (138)
- VP.1 eScience (120)
- 6 Materialchemie (31)
- 6.0 Abteilungsleitung und andere (21)
- 1 Analytische Chemie; Referenzmaterialien (17)
- VP.0 Vizepräsident und andere (14)
- 8 Zerstörungsfreie Prüfung (12)
- 5 Werkstofftechnik (10)
- 5.2 Metallische Hochtemperaturwerkstoffe (8)
- 7 Bauwerkssicherheit (8)
Paper des Monats
- ja (2)
A high-resolution spatiotemporal wildfire propagation dataset for the Mediterranean and Europe
(2026)
Wildfires are becoming more frequent and severe under the influence of climate change, posing increasing risks to ecosystems, human health, and infrastructure. Accurate spatiotemporal data on wildfire propagation is essential for advancing fire behavior modeling, improving management strategies, and mitigating future impacts. However, existing datasets with both high spatial and temporal resolution are rare, costly, and time-consuming to produce. To address this gap, we present FireSpread_MedEU, a dataset comprising 320 consecutive burned area maps from 103 wildfire events across the Mediterranean and Europe between 2017 and 2023. Burned areas were derived from high-resolution Planet optical satellite imagery (~3 m spatial, mostly daily temporal resolution) using a semi-automated workflow, followed by manual refinement to ensure highest accuracy. Each dataset entry is enriched with detailed metadata and a subjective quality assessment. With its high level of spatiotemporal precision, FireSpread_MedEU provides essential data for the development and validation of machine learning models or wildfire simulation models. It opens new research opportunities in wildfire behavior analysis, risk assessment, and predictive modeling.
In this contribution, we present our Self-Driving Lab (SDL) for Nano and Advanced Materials, that integrates robotics for batched autonomous synthesis – from molecular precursors to fully purified nanomaterials – with automated characterization and data analysis, for a complete and reliable nanomaterial synthesis workflow. By fully automating the processing steps for seven different materials from five representative, completely different classes of nano- and advanced materials (metal, metal oxide, silica, metal organic framework, and core–shell particles) that follow different reaction mechanisms, we demonstrate the great versatility and flexibility of the platform. The system also exhibits high modularity and adaptability in terms of reaction scales and incorporates in-line characterization measurement of hydrodynamic diameter, zeta potential, and optical properties (absorbance, fluorescence). We discuss the excellent reproducibility of the various materials synthesized on the platform in terms of particle size and size distribution, and the adaptability and modularity that allows access to a diverse set of nanomaterial classes.
We also present several key aspects of the central backend that orchestrates the (parallelized) syntheses workflows. One key feature is the resource management or “traffic control” for scheduling and executing parallel reactions in a multi-threaded environment. Another is the interface with data analysis algorithms from in-line, at-line, and off-line measurements. Here, we will give examples of how automatic image segmentation of electron microscopy images with the help of AI can be used for reducing the “data analysis bottleneck” from an off-line measurement. We will also discuss various machine learning (ML) algorithms that are currently implemented in the backend and can be used for ML-guided, closed-loop material optimization in our SDL. Lastly, we will show our recent efforts in making the workflow generation on SDLs more user-friendly by using large language models to generate executable workflows automatically from synthesis procedures given in natural language and user-friendly graphical user interfaces based on node editors that also allow for knowledge graph extraction from the workflows. In this context, we are currently also working on a common description or ontology for representing the process steps and parameters of the workflows, which will greatly facilitate the semantic description and interoperability of workflows between different SDL hardware and software platforms.
Most machine learning models for materials science rely on descriptors based on materials compositions and structures, even though the chemical bond has been proven to be a valuable concept for predicting materials properties. Over the years, various theoretical frameworks have been developed to characterize bonding in solid-state materials. However, integrating bonding information from these frameworks into machine learning pipelines at scale has been limited by the lack of a systematically generated and validated database. Recent advances in high-throughput bonding analysis workflows have addressed this issue, and our previously computed Quantum-Chemical Bonding Database for Solid-State Materials was extended to include approximately 13,000 materials. This database is then used to derive a new set of quantum-chemical bonding descriptors. A systematic assessment is performed using statistical significance tests to evaluate how the inclusion of these descriptors influences the performance of machine-learning models that otherwise rely solely on structure- and composition-derived features. Models are built to predict elastic, vibrational, and thermodynamic properties typically associated with chemical bonding in materials. The results demonstrate that incorporating quantum-chemical bonding descriptors not only improves predictive performance but also helps identify intuitive expressions for
properties such as the projected force constant and lattice thermal conductivity via symbolic regression.
Machine learning in materials science and engineering – best practice, perspectives and pitfalls
(2026)
Machine learning (ML) is increasingly utilized to support the data driven analysis of relationships in multidimensional parameter spaces, ideally as an entry point for a more general phenomenological or physics-based model development. Applications include both forward and inverse problems as well as forward problems, for example parameter identification or modeling of structure-property relationships.
The talk will give an overview over a variety of solutions that benefit from the capability of artificial neural networks to approximate and interpolate complex relationships that are represented by a set of sparse data. The reason behind is that numerical simulations as well as experiments do often not allow to generate enough data such that the data set is not sufficient for a deep-learning approach in connection with the complexity of the problem at hand.
After a short introduction to artificial neural networks along with recommendations for data generation and feature engineering, the talk will cover a range of examples from nanoindentation and material parameter identification, the improvement of characterization techniques by ML correction methods towards recent problems in the prediction of structure-property relationships for materials with complex microstructure. All these examples have in common that a successful ML model typically requires a comprehensive understanding of existing knowledge, expertise in translating this knowledge into meaningful input features, a compact ML architecture, and robust validation of the trained model. The talk will conclude with the example of nanoporous metals that demonstrates the importance of high-quality and bias-free data for the applicability and trustworthiness of the trained model, also emphasizing the need for a culture of open data, specifically towards curated data sets for training and validation of ML models.
Microstructure and orientation effects on microcompression-induced plasticity in nanoporous gold
(2026)
Understanding the plastic deformation of nanoporous metals requires a detailed examination of their small-scale microstructural features. In this work, we present a computational study of micropillar compression in single crystal nanoporous gold (NPG) using crystal plasticity. This approach enables a systematic investigation of three key microstructural effects, including ligament size (50 ≤ 𝑙 ≤ 400 nm), solid fraction (0.2 ≤ 𝜑 ≤ 0.3), and initial crystal orientation ([001] and [111] ̄ ), on the plastic response far beyond yielding. After validation against experimental data, the study reveals that, in line with the ’smaller is stronger’ trend, besides the yield strength, the strain hardening rate also increases as ligament size decreases. Moreover, the strain hardening rate follows a power-law scaling with solid fraction, similar to the yield strength. The analysis of two distinct crystal orientations presents findings contrasting with previous assumptions. While the yielding onset remains orientation-independent, as expected, an increase in the strain hardening rate emerges for the harder [11-1] orientation with continued compression. An effect that becomes more pronounced with increasing solid fraction and decreasing ligament size. Under these conditions, harder orientations also amplify local stress heterogeneity. Notably, the stress distribution in NPG is nearly twice as wide as that observed in the single crystal bulk material (𝜑 = 1.0). Compared to the crystal plasticity approach, traditional isotropic plasticity predicts more uniform local stress fields.
This paper presents digital quality infrastructure methods for hydrogen refueling stations using the Asset Administration Shell as a standardized digital twin. Implemented at BAM’s test platform, it integrates real-time sensor data, calibration certificates, and compliance documents to support traceable, interoperable asset management. In combination with AI and semantic tools, the system will enable predictive maintenance, remote audits, and improved safety. This approach reduces downtime, enhances transparency, and offers a scalable model demonstrating the potential of digital twins in advancing metrological traceability and operational efficiency in hydrogen technologies.
Data-driven nanomechanical study of filled fluoroelastomer aged in air and hydrogen atmosphere
(2025)
Fluoroelastomer (FKM) composites are typically used as sealing materials in challenging non-ambient environments. Depending on the environment, two main aging mechanisms, chemical aging, and physical aging, can be identified. Chemical aging, the degradation of the elastomer, is present for example in thermal-oxidative conditions and can be directly observed as it affects the bulk. Physical aging, relaxation and rearrangement of the elastomers segmental conformation is commonly observed at elevated temperatures and effects predominantly the elastomer interphase. As a highly localized nanoscopic effect it is usually observed indirectly by phenomological approaches and not systematically understood. In this study, as a typical example for chemical aging, filled FKM was aged in air (150°C, 100 days). Physical aging of FKM was realized by exposure to chemically inert H2 (150°C, 50 bar, 100 days), since temperature and gas-induced swelling is known to promote physical aging. The effects of both conditions are directly compared with the initial unaged material. We use atomic force microscopy (AFM) force spectroscopy as a method to resolve nanoscopic heterogeneous FKM. With this method the effect of aging on the spatially distinguishable material phases was directly observed. In thermal oxidative aged FKM the matrix shows a decrease in van der Waals interactions and stiffness, indicating dehydrofluorination and chain scission. In H2 aged FKM, the development of an immobilized amorphous interphase (IAP) was observed, indicating physical aging. By additionally evaluating a larger data set with supervised machine learning, these observations were validated for a larger, statistically representative sample area, allowing conclusions to be drawn about the macroscopic behaviour of the material.
Environmental and scarcity issues of common functional magnetic materials for, e.g., permanent magnets have intensified the search for rare-earth-free alternatives. This challenge is increasingly met by machine learning of magnetic properties of transition-metal compounds. Surprisingly, bond-angle-derived features were not found to be relevant for magnetic structure prediction in previous studies using DFT-computed labels. This contrasts with the Kanamori-Goodenough-Anderson (KGA) rules of superexchange, present in every magnetism textbook.
These semiempirical rules predict whether a nearest-neighbor magnetic interaction in insulators is FM or AFM based on the bond angle, orbital symmetry, and orbital occupancy. For some cases, the rules can be simplified further to only consider the bond angle of neighboring magnetic sites (KGA rules of thumb).
We review magnetism—bond angle trends within the MAGNDATA database, the largest collection of experimentally determined magnetic structures. Observed trends follow the KGA rules of thumb, and exceptions can be rationalized. In contrast, bond angles in a popular theoretical DFT database show very different trends and do not depend on the magnetic ordering.
Building on our analysis, we engineer heuristic-derived features for the machine learning of magnetic structures. We introduce a new, informative label for predicting magnetic structures that can be extended to magnetic sites and structures of arbitrary complexity. We show that features derived from the heuristic are of high importance for this machine learning task. Beyond this, our model enables the prediction of non-collinear magnetic structures.
Atomistic simulations of matter, especially those that leverage first-principles (ab initio) electronic structure theory, provide a microscopic view of the world, underpinning much of our understanding of chemistry and materials science. Over the last decade or so, machine-learned force fields have transformed atomistic modeling by enabling simulations of ab initio quality over unprecedented time and length scales. However, early machine-learning (ML) force fields have largely been limited by (i) the substantial computational and human effort required to develop and validate potentials for each particular system of interest and (ii) a general lack of transferability from one chemical system to the next. Here, we show that it is possible to create a general-purpose atomistic ML model, trained on a public dataset of moderate size, that is capable of running stable molecular dynamics for a wide range of molecules and materials. We demonstrate the power of the MACE-MP-0 model—and its qualitative and at times quantitative accuracy—on a diverse set of problems in the physical sciences, including properties of solids, liquids, gases, chemical reactions, interfaces, and even the dynamics of a small protein. The model can be applied out of the box as a starting or “foundation” model for any atomistic system of interest and, when desired, can be fine-tuned on just a handful of application-specific data points to reach ab initio accuracy. Establishing that a stable force-field model can cover almost all materials changes atomistic modeling in a fundamental way: experienced users obtain reliable results much faster, and beginners face a lower barrier to entry. Foundation models thus represent a step toward democratizing the revolution in atomic-scale modeling that has been brought about by ML force fields.
FireSpread_MedEU
(2025)
Wildfires are becoming more frequent and severe under the influence of climate change, posing increasing risks to ecosystems, human health, and infrastructure. Accurate spatiotemporal data on wildfire propagation is essential for advancing fire behavior modeling, improving management strategies, and mitigating future impacts. However, existing datasets with both high spatial and temporal resolution are rare, costly, and time-consuming to produce. To address this gap, we present FireSpread_MedEU, a dataset comprising 313 consecutive burned area maps from 102 wildfire events across the Mediterranean and Europe between 2017 and 2023. Burned areas were derived from high-resolution Planet optical satellite imagery (~3 m spatial, mostly daily temporal resolution) using a semi-automated workflow, followed by manual refinement to ensure highest accuracy. Each dataset entry is enriched with detailed metadata and a subjective quality assessment. With its high level of spatiotemporal precision, FireSpread_MedEU provides essential data for the development and validation of machine learning models or wildfire simulation models. It opens new research opportunities in wildfire behavior analysis, risk assessment, and predictive modeling.