University of Kaiserslautern
KLUEDO Publication Server of University of Kaiserslautern-Landau (RPTU)Not a member yet
6017 research outputs found
Sort by
Synthese von isopropylsubstituierten Cyclopentadienyl-Komplexen der 3d-Übergangsmetalle
Get PDFDie vorliegende Arbeit bietet Einblicke in die Darstellung neuer isopropylsubstituierter Cyclopentadienyl-Liganden ausgehend von den korrespondierenden Pentafulvenen sowie in der Synthese mehrkerniger Cyclopentadienyl-Komplexe mit 3d-Übergangsmetalle.
Ausgehend von 1,2,3,4-Tetraisopropyl-6-(N,N-dimethylamino)pentafulven konnten durch Addition eines Lithiumorganyls mit nachfolgender Eliminierung der Dimethylamino-Gruppe vier neue 1,2,3,4-Tetraisopropyl¬pentafulvene synthetisiert und ihre Struktureigenschaften vergleichend diskutiert werden. Die beiden Pentafulvene mit den sperrigsten Substituenten in 6-Position wurden in einem weiteren Syntheseschritt in Cyclopentadienyl-Liganden überführt, wobei das Lithiumbenzhydryltetraisopropylcyclopentadienid zusätzlich in Komplexierungen mit FeBr2(DME) und NiBr2(DME) verwendet wurde.
Für die homologe Reihe der dimeren Halbsandwichkomplexe [5CpFe(μ-X)]2 (X = Cl, Br, I) wurden die literaturbekannten Syntheserouten evaluiert und ausstehende Charakterisierungen ergänzt. Zudem wurde unter Verwendung von Cäsiumfluorid das fluoroverbrückte Dimer [5CpFe(μ-F)]2 synthetisiert und vollständig charakterisiert. Zusätzlich wurden im Zuge dieser Synthesen fünf Salz-Addukte über Röntgenstrukturanalyse nachgewiesen.
Für den in situ hergestellten Komplex [4CpFe(σ-Mes)] konnte durch Umsetzungen mit fünf verschiedenen bromoverbrückten Halbsandwichkomplexen mit dem Strukturmotiv [(CpR)M(μ-X)]2 die haptotrope Umlagerung unter Erhalt der korrespondierenden, zweikernigen Komplexe [4CpFe(µ,ƞ6:ƞ1-Mes)MBr(CpR)] erfolgreich durchgeführt werden.
Durch die Addition von bidentaten Stickstoff-Donorliganden an [4CpCo(µ-Br)]2 konnten erstmalig Komplexe mit dem Strukturmotiv [4CpCo(N-N)]Br (mit N-N = bipy, dtbbpy, phen) analysenrein hergestellt und charakterisiert werden. Über Röntgenstruktur¬analyse und DFT-Rechnungen von [4CpCo(dtbbpy)]Br und [4CpCo(phen)]Br konnte die elektronische Struktur der Komplexe mit den non-innocent-Liganden nachvollzogen werden. Die Reduktion dieser Komplexe mittels Natriumamalgam resultierte in der Darstellung der diamagnetischen Komplexe [4CpCo(dtbbpy)] und [4CpCo(phen)], in denen, wie im Fall von [4CpCoOMeBIAN], das Co(II)zentrum antiferromagnetisch mit dem Radikalanion des bidentaten Liganden gekoppelt ist
Controllable Deep Image Synthesis
Get PDFDeep learning breakthroughs have significantly advanced computer vision in the past decade, particularly in image synthesis, which involves generating and manipulating images. Image synthesis has numerous practical applications, including art generation, editing, virtual reality, video games, and computer-aided design. While learning the unconditional distribution of natural images is interesting, gaining control over the image generation process by learning a conditional distribution is essential for practical applications. This thesis presents new methods of controllable generative models for high-quality deep image synthesis, building upon and extending the progress made in generative deep learning over the past decade. The central research question is how to advance controllable deep image synthesis, which is explored through five main dimensions: understanding the current state-of-the-art, integrating existing approaches, enabling fine-grained user inputs, improving models by focusing on important image areas, and developing an efficient generation algorithm. Natural language, the primary medium through which we communicate thoughts, ideas, and feelings, is arguably the most flexible and intuitive interface for controllable image synthesis. Thus, the first part of this research reviews text-to-image synthesis models and highlights open challenges such as generating complex scenes. The second part develops hybrid models that enhance image quality and alignment by integrating text-to-image synthesis with visual question answering and proposes a framework of robust generative networks. The third part focuses on precise control over image regions, covering attribute-controlled and dense text-to-image synthesis from free-form region descriptions. The fourth part introduces methods that prioritize key image areas, such as dynamic attention-guided diffusion and a curriculum learning approach that progressively blurs object regions to stabilize training and improve quality. Finally, the last part proposes an efficient algorithm leveraging pre-trained models for high-resolution text-based image generation. In summary, this thesis contributes to the field of controllable deep image synthesis, providing new methods and insights for developing advanced generative models
Amtliche Bekanntmachung der RPTU Kaiserslautern-Landau 2025.09
Get PDFAmtliche Bekanntmachung der RPTU
Nr.9/18.12.202
Deep Learning and Sensor-Driven Learning Analytics and Augmentation
Get PDFEnhancing e-learning outcomes requires understanding students’ learning behaviors, particularly their affective and cognitive states. By recognizing and responding to these states, personalized interventions can be implemented to improve learning efficacy. This research aims to develop adaptive, learner-centric systems that integrate real-time mental state predictions with customized support, promoting both academic performance and student well-being in digital learning environments.
To address the challenges in detecting learners’ mental states, this research focuses on developing innovative deep-learning solutions that integrate multimodal sensor data, enabling more accurate predictions and personalized interventions to enhance e-learning outcomes. We have developed and implemented innovative
deep learning based models incorporating various validation methods and techniques to improve generalized prediction and personalized prediction models with significant improvements in overall accuracies. Notably, fine-tuning and user-specific calibration resulted in a substantial performance increase, with three-level stress detection using an LSTM model achieving a 56% improvement to reach 91% accuracy (F1=0.911) and a person-specific CNN-LSTM model showing a 50% improvement in detecting interest levels per participant. User-dependent approaches further enhanced accuracy, yielding gains of 13% for engagement, 19% for arousal, and 15% for valence.
These refined models provide the foundation for personalized interventions designed to enhance the learning experience. These interventions include an adapted cognitive control training for managing distractions during online learning where the findings indicated that participants who received training exhibited significantly improved comprehension (p-value = 0.0059) and reduced distraction levels (p-value = 0.021) compared to the untrained group. Additionally, an application-based feedback system provides visual cues based on the learner’s current engagement and
emotional state, enabling students to self-regulate and adapt their learning strategies in real time.
Building on these interventions, a gaze-based adaptive learning system dynamically adjusts content presentation based on real-time engagement analysis. This system utilizes AI-generated summaries via ChatGPT to provide concise and relevant information when low engagement is detected, helping to re-engage students and maintain focus. User studies demonstrated significant improvements in comprehension (p = 0.0018), engagement (p = 0.0021), and overall learning outcomes, further highlighting the effectiveness of these personalized approaches
Enhancing platform chemical production in bioelectrochemical systems
Get PDFCathodic electro-fermentation is an emerging technology with promising potential for industrial application, particularly for enhancing the production of various biotechnologically relevant metabolites. It is based on the combination of traditional industrial fermentation processes with electrochemical approaches. This dissertation investigates two specific application areas of this technology. Firstly, succinate biosynthesis with Actinobacillus succinogenes and secondly, the enhancement of butanol production during acetone-butanol-ethanol (ABE) fermentation with Clostridium acetobutylicum. Key influencing factors are systematically analysed, including the selection of different mediator molecules, the effect of various electrical potentials, the impact on intracellular redox balance, and initial concepts for process scale-up. Cathodic electro-fermentation led to an increased product yield in both processes. Regarding A. succinogenes, increased NADH availability was also observed in the stationary phase. The results of the study help select suitable mediators for electron transfer, with neutral red currently delivering the most favorable results. For electrified ABE fermentation, successful scale-up was achieved in a newly developed reactor design that allows the use of a rotating electrode serving simultaneously as a stirrer and working electrode. Despite these advances, significant challenges remain. The molecular and biochemical mechanisms underlying the product-enhancing effects of electric potentials are still not fully understood. Furthermore, process scalability continues to represent a key barrier. Nevertheless, the results clearly demonstrate the strong potential of cathodic electro fermentation. Continued research should therefore pave the way for future industrial implementation of this technology.Die kathodische Elektrofermentation ist eine neuartige Technologie mit vielversprechendem Potenzial für die industrielle Anwendung, insbesondere zur Steigerung verschiedenster biotechnologisch relevanter Stoffwechselprodukte. Sie basiert auf der Kombination traditioneller industrieller Fermentationsprozesse mit elektrochemischen Ansätzen. Die vorliegende Dissertation untersucht zwei spezifische Anwendungsbereiche dieser Technologie. Zum einen die Succinatbiosynthese mit Actinobacillus succinogenes, zum anderen die Steigerung der Produktion von Butanol in der Aceton-Butanol-Ethanol (ABE)-Fermentation mit Clostridium acetobutylicum. Es werden systematisch zentrale Einflussfaktoren analysiert, darunter die Auswahl verschiedener Mediatormoleküle, die Wirkung unterschiedlicher elektrischer Potentiale, der Einfluss auf den intrazellulären Redoxzustand sowie erste Konzepte zur Skalierung. Durch die kathodische Elektrofermentation konnte in beiden Prozessen eine Erhöhung der Produktausbeute erzielt werden. Bei A. succinogenes wurde in der stationären Phase zudem eine gesteigerte NADH-Verfügbarkeit beobachtet. Die Untersuchungen ermöglichen eine Eingrenzung geeigneter Mediatoren für den Elektronentransfer, wobei Neutralrot derzeit die besten Ergebnisse liefert. Für die elektrifizierte ABE-Fermentation gelang eine erfolgreiche Hochskalierung in ein neu entwickeltes Reaktordesign, das den Einsatz einer rotierenden Elektrode als kombinierte Rühr- und Arbeitselektrode erlaubt. Trotz dieser Fortschritte bestehen weiterhin erhebliche Herausforderungen. Die zugrunde liegenden molekularen und biochemischen Mechanismen der produktionserhöhenden Wirkung elektrischer Potentiale sind bislang nur unzureichend verstanden. Auch die Skalierbarkeit der Prozesse bleibt weiterhin eine zentrale Hürde. Dennoch verdeutlichen die Ergebnisse das große Potenzial der kathodischen Elektrofermentation. Somit sollte weitere Forschung den Weg für eine zukünftige industrielle Anwendung dieser Technologie ebnen
Artificial Neural Network-Based Methodology for Tribological Evaluation and Optimization of PEEK Composites
Get PDFPolymer-based composites are increasingly employed in tribological applications due to their lightweight nature, chemical inertness, and ease of modification compared with conventional metallic materials. However, optimizing their performance through empirical and analytical approaches remains time-consuming and resource-intensive. In this study, three artificial neural networks (ANNs) were systematically integrated into a unified tribological workflow for polyetheretherketone (PEEK)-based composites, encompassing automated data processing, quantitative analysis of worn surface morphology, and prediction of friction and wear behavior across multi-dimensional compositional spaces. The Long Short-Term Memory (LSTM) network reduced data-processing time by more than 95% compared with manual analysis, while the U-Net segmentation model achieved a third-body patch area estimation error of only 7.2% and accelerated image analysis by a factor of approximately 50000. The Feedforward Backpropagation Neural Network (FBNN) captured the complex nonlinear relationships between filler composition, test conditions, and tribological performance, enabling the prediction of the tribological behavior of more than 23000 potential composite formulations. Based on these predictions, a cost–performance evaluation framework was established to enable the efficient design of new tribomaterials. These results demonstrate the potential of ANN-based methods to accelerate the research and design of high-performance tribological composites and to support their targeted optimization for industrial applications.Polymerverbundwerkstoffe werden aufgrund ihres geringen Gewichts, ihrer chemischen Beständigkeit und der einfachen Modifizierbarkeit im Vergleich zu herkömmlichen metallischen Werkstoffen zunehmend in tribologischen Anwendungen eingesetzt. Die Optimierung ihrer Leistung auf der Grundlage empirischer und analytischer Ansätze ist jedoch nach wie vor zeit- und ressourcenintensiv. In der vorliegenden Arbeit wurden drei verschiedene künstliche neuronale Netze (KNN) systematisch in einen einheitlichen tribologischen Arbeitsablauf für Verbundwerkstoffe auf Polyetheretherketon-(PEEK)-Basis integriert. Dieser Ansatz umfasste die automatisierte Datenverarbeitung, die quantitative Analyse der Verschleißoberflächenmorphologie sowie die Vorhersage des Reibungs- und Verschleißverhaltens über mehrdimensionale Zusammensetzungsräume hinweg. Das Long Short-Term Memory (LSTM)-Netzwerk reduzierte die Datenverarbeitungszeit im Vergleich zur manuellen Auswertung um mehr als 95 %. Das U-Net-Segmentierungsmodell erreichte einen Flächenfehler der Drittkörperpatches von nur 7,2 % und beschleunigte die Bildanalyse um etwa den Faktor 50 000. Das Feedforward Backpropagation Neural Network (FBNN) erfasste die komplexen nichtlinearen Zusammenhänge zwischen Zusammensetzung, Testbedingungen und tribologischem Verhalten und ermöglichte die Vorhersage des tribologische Verhalten von mehr als 23 000 potenziellen Formulierungen. Auf der Grundlage dieser Vorhersagen wurde ein Kosten-Leistungs-Bewertungssystem entwickelt, das eine effiziente Auslegung neuer Tribomaterialien ermöglicht. Diese Ergebnisse zeigen das Potenzial KNN-basierter Methoden, die Forschung und Entwicklung von Hochleistungs-Tribocompositen zu beschleunigen und ihre gezielte Optimierung für industrielle Anwendungen zu unterstützen
The Effects of an Outdoor Learning Program, ‘GewässerCampus’, in the Context of Environmental Education
Get PDFWith education playing a role as a catalyst for change towards a more sustainable world, there is a need to develop educational concepts that enable young people to responsibly take up the challenges of future-proof development. The GewässerCampus project is related to environmental education in the context of the ecological dimension of Education for Sustainable Development. This article focuses on evaluating the GewässerCampus project by assessing current motivation, ecological knowledge, and environmental values during participation in an outdoor learning program. In total, 231 German pupils of lower and upper secondary level participated in the project. In a quasi-experimental study design, current motivation, pro-environmental and anthropogenic values (Preservation and Utilization), and knowledge were assessed before and immediately after participation in the learning program. The learning activities during the project day led to significant knowledge acquisition. Furthermore, high individual values of the test items for Preservation and
low values of the test items for Utilization were obtained. Our results show how important it is to consider the individual teaching and learning equirements of the learner group depending on the grade level, as well as the type of school, when preparing modules for environmental education in the context of sustainable development
Toward automated plantar pressure analysis: machine learning-based segmentation and key point detection across multicenter data
Get PDFPlantar pressure analysis is a pivotal tool for assessing foot function, diagnosing deformities, and characterizing gait patterns. Traditional proportion-based segmentation methods are often limited, particularly for atypical foot structures and low-quality data. Although recent advances in machine learning (ML) offer opportunities for automated and robust segmentation across diverse datasets, existing models primarily rely on data from single laboratories, limiting their applicability to multicenter datasets. Furthermore, the prediction of relevant landmarks on the plantar pressure profile has not been explored. This study addresses these gaps by exploring ML-based approaches for anatomical zone segmentation and landmark detection in plantar pressure analysis, including 758 plantar pressure samples from 460 individuals (197 females, 263 males) collected from multiple centers during static and dynamic conditions using two distinct systems. The datasets were further standardized and augmented. The plantar surface was segmented into four regions (hallux, metatarsal area 1, metatarsal areas 2–5, and the heel) using a U-Net model, and deep learning regression models predicted the key points, such as interdigital space coordinates and the center of metatarsal area 1. The results underscore the U-Net’s capacity to attain an accuracy comparable to that of experts (Median Dice Scores ≥ 0.88), particularly in regions with well-defined plantar pressure boundaries. Metatarsal area 1 exhibited unique characteristics because of its ambiguous boundaries, with expert reviews playing a valuable role in enhancing accuracy in critical cases. Using a regression model (Median Euclidean distance = 7.72) or an ensemble model (Median Euclidean distance = 5.26) did not improve calculating the center of metatarsal area 1 directly from the segmentation model (Median Euclidean distance = 4.47). Furthermore, regression-based approaches generated higher errors in key point detection of the interdigital space 2–3 (Median Euclidean distance = 10.06) than in metatarsal area 1 center (Median Euclidean distance = 7.72). These findings emphasize the robustness of the proposed segmentation and key point prediction models across diverse datasets and hardware setups. Overall, the proposed methods facilitate the efficient processing of large, multicenter datasets across diverse hardware setups, significantly reducing the reliance on extensive human labeling, lowering costs, and minimizing subjective bias through ML-driven standardization. Leveraging these strengths, this work introduces a novel framework that integrates multicenter plantar pressure data for both segmentation and landmark detection, offering practical value in clinical and research settings by enabling standardized, automated analyses across varying hardware configurations
Weighted Hartree-Fock-Bogoliubov method for interacting fermions: An application to ultracold Fermi superfluids
Get PDFFor several decades it has been known that divergences arise in the ground-state energy and chemical potential of unitary superfluids, where the scattering length diverges, due to particle-hole scattering. Leading textbooks and research articles recognize that there are serious issues but ignore them due to the lack of an approach that can regularize these divergences. We find a solution to this difficulty by proposing a general method, called the weighted Hartree-Fock Bogoliubov theory, to handle multiple decomposition channels originating from the same interaction. We distribute the interaction in weighted channels determined by minimization of the action, and we apply this idea to unpolarized Fermi superfluids. Using our method, we solve a long-standing difficulty in the partitioning of the interaction into Hartree, Fock, and Bogoliubov channels for Fermi superfluids, and we obtain a phase diagram at the saddle-point level, which contains multichannel nonperturbative corrections. In particular, we find a previously overlooked superfluid phase for weak interactions, which is dominated by particle-hole processes, in addition to the usual superfluid phase only containing particle-particle physics
Isogeometric analysis of the lithosphere under topographic loading: Igalith v1.0.0
Get PDFThis paper presents methods from isogeometric finite-element analysis for numerically solving problems in geoscience involving partial differential equations. In particular, we consider the numerical simulation of shells and plates in the context of isostasy. Earth's lithosphere is modeled as a thin elastic shell or plate floating on the asthenosphere and subject to topographic loads. We demonstrate the computational methods on the isostatic boundary value problem posed on selected geographic locations. For Europe, the computed lithospheric depression is compared with available Mohorovičić depth data. We also perform parameter identification for the effective elastic thickness of the lithosphere, the rock density, and the topographic load that are most plausible to explain the measured depths. An example of simulating the entire lithosphere of the Earth as a spherical shell using multi-patch isogeometric analysis is presented, providing an alternative to spherical harmonics for solving partial differential equations on a spherical domain. The numerical results serve to showcase the features and capabilities of isogeometric methods rather than to provide insightful predictions since a fairly simple model is used for the loading of the lithosphere