Graue Literatur
Filtern
Erscheinungsjahr
Dokumenttyp
- Beitrag zu einem Tagungsband (3980)
- Zeitschriftenartikel (768)
- Forschungsbericht (277)
- Forschungsdatensatz (200)
- Sonstiges (172)
- Dissertation (115)
- Beitrag zu einem Sammelband (99)
- Preprint (60)
- Posterpräsentation (17)
- Tagungsband (Herausgeberschaft für den kompletten Band) (11)
Sprache
- Englisch (3281)
- Deutsch (2391)
- Mehrsprachig (22)
- Französisch (10)
- Tschechisch (6)
- Russisch (6)
- Spanisch (5)
- Italienisch (3)
- Mongolisch (3)
- Portugiesisch (2)
Schlagworte
- Concrete (73)
- Corrosion (66)
- Simulation (59)
- Korrosion (47)
- Zerstörungsfreie Prüfung (47)
- Structural health monitoring (44)
- Monitoring (42)
- Non-destructive testing (42)
- NDT (40)
- Ultrasound (39)
Organisationseinheit der BAM
- 8 Zerstörungsfreie Prüfung (347)
- 7 Bauwerkssicherheit (212)
- 6 Materialchemie (175)
- 9 Komponentensicherheit (139)
- 3 Gefahrgutumschließungen; Energiespeicher (123)
- 1 Analytische Chemie; Referenzmaterialien (93)
- 4 Material und Umwelt (88)
- 5 Werkstofftechnik (86)
- 8.4 Akustische und elektromagnetische Verfahren (85)
- 8.2 Zerstörungsfreie Prüfmethoden für das Bauwesen (84)
Eingeladener Vortrag (wissenschaftliche Konferenzen)
- nein (8)
Fahrzeugbrände in Verkehrsinfrastrukturen wie Depots, Parkhäusern, Tunneln und Stationen sind ein wiederkehrendes Phänomen. In den vergangenen Jahren wurde eine alarmierende Zunahme schwerwiegender Brandereignisse beobachtet, insbesondere in Fahrzeugdepots. Diese Brände führen häufig zum Totalverlust sowohl der Fahrzeuge als auch der Gebäude und weisen bemerkenswerte Gemeinsamkeiten im Brandverlauf auf. Die Analyse jüngerer Brandereignisse, unter anderem im Rahmen einer Studie für den deutschen Versicherer Provinzial, zeigt, dass sich Brände in Busdepots, militärischen Einrichtungen und Parkhäusern oftmals schnell und mit hoher Zerstörungskraft ausbreiten. Während die baurechtlichen Vorschriften je nach Sektor und Land variieren, sind die Brandschutzanforderungen an Innenraummaterialien von Straßenfahrzeugen weltweit harmonisiert und basieren auf veralteten Regelwerken wie der FMVSS 302, die lediglich ein minimales Schutzniveau bieten. Demgegenüber unterliegen Materialien in Schienenfahrzeugen, Schiffen und Luftfahrzeugen deutlich strengeren Brandschutzanforderungen, was die vergleichsweise geringere Brandhäufigkeit in diesen Bereichen teilweise erklären könnte.
Machine-learning interatomic potentials are widely used as computationally efficient surrogates for density functional theory in atomistic simulations, enabling large-scale, long-time modeling of materials systems. We investigate how different fine-tuning strategies influence the prediction of harmonic phonon band structures, thermal properties, and the potential energy surface along imaginary phonon modes. We achieve substantial accuracy improvements with minimal additional data, with as few as 10 additional training structures already yielding significant gains. In addition to existing approaches, we introduce Equitrain, a finetuning framework that implements LoRA-based adaptation. Across 53 materials systems, we show that fine-tuned models consistently outperform both the underlying pretrained model and models trained from scratch. Equitrain achieves the best overall performance, and our results demonstrate that fine-tuning enables accurate phonon predictions.
Artificial intelligence is transforming molecular and materials science, but its growing computational and data demands raise critical sustainability challenges. In this Perspective, we examine resource considerations across the AI-driven discovery pipeline--from quantum-mechanical (QM) data generation and model training to automated, self-driving research workflows--building on discussions from the ``SusML workshop: Towards sustainable exploration of chemical spaces with machine learning'' held in Dresden, Germany. In this context, the availability of large quantum datasets has enabled rigorous benchmarking and rapid methodological progress, while also incurring substantial energy and infrastructure costs. We highlight emerging strategies to enhance efficiency, including general-purpose machine learning (ML) models, multi-fidelity approaches, model distillation, and active learning. Moreover, incorporating physics-based constraints within hierarchical workflows, where fast ML surrogates are applied broadly and high-accuracy QM methods are used selectively, can further optimize resource use without compromising reliability. Equally important is bridging the gap between idealized computational predictions and real-world conditions by accounting for synthesizability and multi-objective design criteria, which is essential for practical impact. Finally, we argue that sustainable progress will rely on open data and models, reusable workflows, and domain-specific AI systems that maximize scientific value per unit of computation, enabling efficient and responsible discovery of technological materials and therapeutics.
Modeling complex physical systems such as they arise in civil engineering applications requires finding a trade-off between physical fidelity and practicality. Consequently, deviations of simulation from measurements are ubiquitous even after model calibration due to the model discrepancy, which may result from deliberate modeling decisions, ignorance, or lack of knowledge. If the mismatch between simulation and measurements are deemed unacceptable, the model has to be improved. Targeted model improvement is challenging due to a non-local impact of model discrepancies on measurements and the dependence on sensor configurations. Many approaches to model improvement, such as Bayesian calibration with additive mismatch terms, gray-box models, symbolic regression, or stochastic model updating, often lack interpretability, generalizability, physical consistency, or practical applicability. This paper introduces a non-intrusive approach to model discrepancy analysis using mixture models. Instead of directly modifying the model structure, the method maps sensor readings to clusters of physically meaningful parameters, automatically assigning sensor readings to parameter vector clusters. This mapping can reveal systematic discrepancies and model biases, guiding targeted, physics-based refinements by the modeler. The approach is formulated within a Bayesian framework, enabling the identification of parameter clusters and their assignments via the Expectation-Maximization (EM) algorithm. The methodology is demonstrated through numerical experiments, including an illustrative example and a real-world case study of heat transfer in a concrete bridge.
In road construction, the reuse of reclaimed asphalt depends on the addition of large quantities of fresh binder. Bio-based rejuvenators can represent a sustainable alternative to this, but their performance concerning multiple reuses has yet to be investigated. For this study, fresh 50/70 bitumen was subjected to up to four ageing and rejuvenation cycles using eight commercially available rejuvenators (five bio- and three petroleum-based). The rheological and chemical evaluation of the raw materials and resulting products was carried out using a dynamic shear rheometer and Fourier transform infrared spectrometer. Through examina-tion of the infrared spectra, the rejuvenators were divided into three groups based on similar chemical compo-sition. Further, evaluated optimum dosage quantities required to rejuvenate aged binder indicate that bio-based rejuvenators are more resource-efficient than their petrol-based counterparts. Finally, a most promising bio-based rejuvenator candidate was found, allowing up to four reuse cycles.
The safety assessment of packages for the transport of radioactive material is a highly regulated and knowledge-intensive process. Regulatory authorities must evaluate complex safety reports that integrate mechanical, thermal, shielding, criticality, and operational analyses, supported by extensive heterogeneous documentation such as drawings, certificates, test results, and inspection records. Although these documents are increasingly available in digital form, they remain largely unstructured and weakly interconnected, requiring manual cross-checking of dependencies and assumptions.
This paper explores the potential of AI-driven documentation analysis to support regulatory safety assessments in the context of IAEA-regulated transport of radioactive materials. It examines the limitations of conventional digital approaches and introduces a multi-layered architecture based on Retrieval-Augmented Generation (RAG), multimodal document processing, and structured knowledge representations. In particular, the paper argues that standard RAG systems are insufficient to capture the deep interdependencies across safety documentation and proposes the integration of Knowledge Graphs and Graph-RAG techniques to enable traceable, multi-step reasoning.
Beyond technical feasibility, the paper emphasizes the importance of human-centered design, explainability, and trustworthiness in safety-critical and regulated domains. Concepts such as Explainable AI and Human-in-the-Loop operation are discussed as essential prerequisites for regulatory acceptance and long-term resilience. Finally, implementation challenges and future developments are outlined, including machine-readable standards and continuous compliance validation. The study demonstrates that AI-supported documentation analysis can significantly enhance efficiency, transparency, and robustness of safety assessments, provided that technical innovation is carefully aligned with regulatory, organizational, and human factors.
Packages for the transport of spent nuclear fuel and high-level radioactive waste must demonstrate their integrity under severe accident conditions to comply with the international transport regulations defined in International Atomic Energy Agency SSR-6. A key component of the approval procedure is the thermal fire test, which requires a fully engulfing 800 °C fire over a duration of 30 minutes. At the Federal Institute of Materials Research and Testing (BAM), such tests are currently conducted using propane gas fires. However, in the context of climate policy objectives, resource availability, and rising costs of fossil fuels, alternative and more sustainable energy sources for fire testing are being investigated.
Hydrogen represents a promising candidate due to its carbon-free combustion and alignment with BAM’s hydrogen strategy. Nevertheless, hydrogen flames exhibit fundamentally different physical and thermal characteristics compared to hydrocarbon flames, most notably a significantly lower radiative emissivity caused by the absence of soot formation. This reduced radiative heat transfer poses a challenge for replicating the boundary conditions required by SSR-6. One potential mitigation strategy is the use of hydrogen–methane blends, where methane serves as a carbon source to enhance flame emissivity while maintaining the possibility of a sustainable fuel pathway.
This paper presents an experimental investigation of hydrogen–methane jet flames with respect to their suitability for thermal fire testing of radioactive material transport packages. A modular experimental test rig was developed and installed at the BAM Test Site for Technical Safety, enabling controlled variation of burner geometry, thermal power, and fuel composition. A Design of Experiments approach based on a Central Composite Design was applied to systematically explore the three-dimensional parameter space. Flame geometry, radiative heat flux, and characteristic flame temperatures were evaluated using thermographic imaging, Gardon gauges, and thermocouples.
The results demonstrate that increasing the methane fraction significantly enhances flame radiation and geometry, while pure hydrogen flames exhibit higher average temperatures but substantially lower radiative heat flux. Quadratic response surface models reveal clear dependencies of flame characteristics on power, nozzle cross section, and methane ratio. Overall, the study confirms that hydrogen–methane blends are a viable option for tailoring flame properties toward the requirements of regulatory fire testing and provides a foundation for the design of future fully engulfing hydrogen-based fire test setups
.
Packages for the transport of high-level radioactive material are designed to withstand severe accident conditions. To obtain regulatory approval, such transport packages must comply with the specification-based requirements defined in the IAEA SSR-6 [1]. Demonstrating compliance could require the performance of specific mechanical and thermal tests, depending on the package type. Typically, IAEA SSR-6 [1] mandates a sequence of cumulative tests consisting of mechanical tests followed by a thermal fire test.
For approval of the fire test, the Bundesanstalt für Materialforschung und -prüfung (BAM, engl. Federal Institute for Materials Research and Testing) employs a reference package that reproduces the outer geometry of the original package to characterize the fire conditions and their effects on the package. This approach serves two purposes: first, it enables precise adjustment of the experimental parameters for the package design under approval; second, it provides input data for thermomechanical simulations (cf. [2]). Using this methodology to characterize the package boundary conditions, temperature evolutions within the fire reference package can be analyzed using finite element analysis. This allows direct comparison between experimental results and numerical simulations for the fire reference package and simultaneously supports preliminary simulations of the package design to be approved.
The thermal test of the SSR-6 [1] includes a fully engulfing 800 °C pool fire with a duration of 30 minutes, or an equally severe fire scenario, such as a propane gas fire. The fire reference test is conducted prior to the regulatory fire test of the package design under approval. In the case described here, the fire reference package is a closed cylindrical shell made of stainless steel, with a wall thickness of 10 mm, a length of 4,860 mm, and a diameter of 2,024 mm. The package was instrumented with thermocouples and filled with heat-resistant insulating material. On the lid side of the cylindrical body, a similarly designed metal sheet–encapsulated structure with insulation was used to replicate the external dimensions of the original impact limiter. Its diameter is 3,200 mm and its height is 1,680 mm.
The reconstruction of the thermal history of anthropogenic materials is crucial for understanding historical manufacturing techniques. Preparatory parameters such as firing temperature, heating and cooling rates, soaking time, and kiln atmosphere significantly affect the chemical and structural properties of the final product. Comparing historical materials with replicas produced under well-defined laboratory conditions helps identify indicators for these parameters. This comparative approach is greatly enhanced by spectroscopic analyses. Raman spectroscopy has proven to be a powerful tool in this field due to its high sensitivity to crystal-chemical alterations and high spatial resolution.
The results of thermal experiments with gypsum and carbonate raw materials at burning temperatures up to 1000 °C are presented. Precise measurements of Raman peak positions and Raman band widths enable the differentiation of chemically similar phases. Changes in the Raman band parameters are evident even after the subsequent hydration-hardening process of the fired samples, allowing the spectral discrimination of samples treated at different temperatures steps. These findings from the thermal experiments are further applied to Raman micro-spectroscopic mappings of medieval and reenacted mortars. The extracted Raman band parameters show comparable values between the experimental and real-life samples, proving Raman spectroscopy as a suitable tool for estimating the burning temperature and thus elucidating the manufacturing procedures of anthropogenic materials.
The transport of radioactive material requires regulatory approval based on the package type, as defined by the regulations of the International Atomic Energy Agency (IAEA). These approvals rely on comprehensive Package Design Safety Reports that evaluate mechanical, thermal, shielding, criticality and transport requirements, supported by specifications, inspections, certificates, drawings, and other technical documentations. Such safety reports contain numerous interconnected documents, and even minor changes, such as component modifications, updated material properties or revised regulations, may affect multiple sections. Although all reports follow the same regulatory framework, each package has unique design features, making every safety assessment distinct. Most documentation exists in digital form but remains largely non–machine-interpretable, limiting automated analysis of dependencies across documents. The extended synopsis argues that overcoming these limitations requires moving from simple digitization toward structured knowledge representation. A multi-stage approach begins with foundational AI technologies, including Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG), which improve information retrieval but cannot capture the full complexity of safety report interrelationships. Building Knowledge Graphs (KGs) offers the necessary next step by transforming heterogeneous, unstructured, and semi-structured documents into a connected, queryable network. KGs enable precise tracing and visualization of dependencies across datasheets, simulations, experimental results, standards, and regulatory requirements. Such structured representations would allow automatic detection of changes, propagation of effects across related documents and validation of conditions using AI-supported tools, reducing manual workload, and improving safety and consistency. Human error remains a significant factor in drafting and reviewing safety reports. A digital quality infrastructure could reduce the number of iterations and further streamline the overall process. Integrating AI into this workflow has the potential not only to optimize assessments but also to improve their robustness by increasing the interpretability of documentation and thereby enhancing overall safety. This preliminary study examines the readiness and requirements for intelligent documentation analysis systems that support regulatory compliance for transport package safety. By analysing current documentation workflows, it demonstrates how LLM-based tools can interpret complex safety reports and identify critical interdependencies, and why KG-based architectures are essential for managing these dependencies reliably.