256869 research outputs found
Sort by
Microbial Community Profiling of Concrete
No full textConcrete is the most widely used construction material worldwide, yet its production and disposal pose significant environmental challenges due to high carbon emissions and limited recyclability. While microbial colonization of concrete is often associated with structural deterioration, recent research has highlighted the potential of microorganisms to contribute positively to concrete recycling and self-healing. In this study, we investigated the bacterial and fungal communities inhabiting urban concrete samples using amplicon-based taxonomic profiling targeting the 16S rRNA gene and internal transcribed spacer (ITS) region. Our analyses revealed a diverse assemblage of microbial taxa capable of surviving the extreme physicochemical conditions of concrete. Several taxa were associated with known metabolic functions relevant to concrete degradation, such as acid and sulphate production, as well as biomineralization processes that may support crack repair and surface sealing. These findings suggest that concrete-associated microbiomes may serve as a reservoir of biological functions with potential applications in sustainable construction, including targeted biodegradation for recycling and biogenic mineral formation for structural healing. This work provides a foundation for developing microbial solutions to reduce the environmental footprint of concrete infrastructure
Linkage Isomerism in a Homoleptic Fe(II) Complex with BODIPY-1H-Tetrazole Ligands
Get PDFWe examined the coordination behavior of the sterically demanding 4,4-difluoro-1,3,5,7-tetramethyl-8-[(1H-tetrazol-1-yl)methyl]-4-bora-3a,4a-diaza-s-indacene ligand L in octahedral Fe(II) coordination compounds. Using four different solvents – ClCH₂CN, BrCH₂CN, CH₃OH, and (CH₃)₂CO – five unique crystal structures were obtained and characterized by single-crystal X-ray diffraction. In ClCH₂CN and BrCH₂CN, L coordinated equatorially through N4 and apically through its less basic N3 atom, yielding a pseudo-homoleptic [Fe(Lᴺ⁴)₄(Lᴺ³)₂]²⁺ architecture. This represents the first documented example of monodentate 1H-tetrazole N3/N4 linkage isomerism in an Fe(II) complex. Conversely, O-donor solvents (CH₃OH and (CH₃)₂CO) resulted in the formation of a heteroleptic [Fe(L)4(solvent)2]2+ motif, in which the solvent molecules coordinated through their O-atoms and occupied two coordination sites. The CH₃OH adduct underwent a topotactic single-crystal-to-single-crystal transformation upon air exposure, exchanging CH₃OH for H₂O, and was stabilized by extensive hydrogen bonding. Unlike classical [Fe(1-alkyl-1H-tetrazole)₆]²⁺ systems, none of the investigated coordination compounds exhibited SCO properties, instead remained locked in a single spin state. These findings demonstrate that steric bulk can override the ligand's N4 coordination preference to Fe(II), either by enforcing N3 coordination or by permitting coordination of co-ligands, and thus provide guidance for the design of 1H-tetrazole-based Fe(II) materials
Constraint Learning for Non-confluent Proof Search
No full textProof search in non-confluent tableau calculi, such as the connection tableau calculus, suffers from excess backtracking, but simple restrictions on backtracking are incomplete. We adopt constraint learning to reduce backtracking in the classical first-order connection calculus, while retaining completeness. An initial constraint learning language for connection-driven search is iteratively refined to greatly reduce backtracking in practice. The approach may be useful for proof search in other non-confluent tableau calculi
Advanced salt hydrate thermochemical heat pipe: Experimental validation and numerical analysis
No full textThis study investigates the use of salt hydrate in a heat pipe as a thermochemical substance for the enhancement of thermal energy transfer, both numerically and experimentally. The main objective is to assess the impact of chemical (hydration/dehydration) reactions on heat conduction and convection within the heat pipe. For this, relying on its high thermal stability, solubility, energy density, and expected positive boiling characteristics, CaCl₂.6H₂O solution was selected as the working fluid. A lab-scale setup of the concept was fabricated and evaluated in the experimental phase. 2D thermal-fluid simulation of two-phase boiling and condensation phenomena was conducted to investigate thermophysical properties, including thermal conductivity, specific heat capacity, evaporation rate, and vapor pressure. Numerical analysis was used to evaluate the effect of salt addition on heat transfer, evaporation rate, and boiling regime stability. Results show that the CaCl₂ solution decreases thermal resistance by 35.4% at 80 °C and by 48.7% at 100 °C compared to pure water. In temperature range of 80–100 °C, the heat transfer mechanisms conduction and convection are dominant, not the chemical reaction. The increased vaporization enthalpy and boiling stability compensate for the solution's lower vapor pressure and thermal conductivity. The presence of salt is proven to prevent unstable boiling modes such as geyser boiling, thereby enhancing the transient behavior of the heat pipe. An analysis of different CaCl₂.6H2O concentrations shows that increasing salt concentration decreases bubble size and dry-out regions below saturation levels where no precipitation occurs
Environmental laboratory aging of bitumen: Contrasting a parametric approach to field aging
Get PDFBitumen aging is a frequently discussed topic among the scientific community. The consensus is that standardized laboratory aging procedures like the Pressure Ageing Vessel (PAV) test do not reflect environmental field conditions, resulting in discrepancies between real-world and predicted aging behavior. Therefore, the Viennese Binder Aging (VBA) method incorporates aging-inducing factors (light, humidity, and reactive oxygen species (ROS)) present in the troposphere into a laboratory setup. This paper uses a parametric approach focused on overall trends to compare the combined and individual effects of these factors on two unmodified binders under different exposure modes. Visual documentation, Fourier Transform Infrared (FTIR) spectroscopy, and Dynamic Shear Rheometer (DSR) results were used to compare VBA aged samples to PAV aged samples and field aged data with varying exposure times. ROS induced strong aging with pronounced visible cracking. Direct light exposure led to the formation of a passivating layer, which reflected field aging well, while indirect light exposure had only limited impact. Water alone did not accelerate aging, but in combination with ROS promoted binder redistribution. While binder A showed no increased aging, binder B showed the overall highest aging degree after exposure to water and ROS. Binder A was most affected by direct light and ROS exposure. These differences highlight the VBA’s ability to distinguish binder-specific sensitivities and to identify the dominant aging drivers across binders. Thus, this study supports the understanding of realistic environmental aging drivers by providing a method that helps bridge the gap between laboratory aging and field aging
Die Reibungswirkung von interkalierten MXenen auf Titanbasis für biomedizinische Anwendungen
No full textThis thesis investigates the tribological performance of organically intercalated Ti3C2Tx MXene coatings applied to a novel Ti64-5HAP-5CMC composite substrate developed for implant applications. The study aims to evaluate the influence of MXene intercalation chemistry and coating methodology on friction behaviour, wear mechanisms, and tribolayer stability under ambient sliding conditions.Intercalated MXene powders were characterised using PXRD, XPS, and TEM, confirming successful intercalation and increased interlayer spacing, with ODA-intercalated MXene exhibiting a larger interlayer distance compared to HDA-intercalated MXene. Surface roughness and morphology of the coated substrates were analysed using optical profilometry.Tribological testing demonstrated a significant reduction and stabilisation of the coefficient of friction for MXene-coated samples compared to the uncoated reference material. Among the investigated systems, spraycoated ODA-intercalated MXene exhibited the lowest and most stable friction values in short-term tests and was therefore selected for extended long-term investigations. Stable low-friction behaviour was maintained over 24h, 72h. While the wear track width on the sample surface remained comparable to the reference material, a clear reduction in counterbody wear was observed for ODA-intercalated MXene coatings.Comprehensive wear track analyses using LSM, Raman spectroscopy, SEM, and TEM revealed the formation of a mechanically mixed tribolayer. Raman spectroscopy confirmed the presence of the organic intercalant within the wear track after sliding. SEM and TEM investigations showed that MXene flakes undergo pronounced bending, folding, and partial amorphisation while largely retaining their lamellar structure. TEM-EDX analysis demonstrated elemental intermixing between the MXene coating and the porous substrate, as well as reduced material transfer from the counterbody for ODA-intercalated systems. Overall, the results demonstrate that organic intercalation, particularly with ODA, significantly enhances the tribological performance of MXene coatings on the investigated implant substrate
Examination of LEIS sputtering depth profiles by Monte Carlo simulations on silicon oxide and zinc oxide
No full textMonokulare 3D-Schätzung menschlicher Körperhaltungen zur Beobachtung von Fahrzeuginnenräumen unter Verwendung synthetischer Bilder
No full textEiner der entscheidendsten Aspekte der Fahrzeugherstellung ist die Gewährleistung der Sicherheit der Fahrzeuginsassen. Da fahrerbedingte Faktoren wie Müdigkeit und Ablenkung zu einem Großteil der Unfälle beitragen, ist die Überwachung der Fahrer*innen wesentlich, um die Verkehrssicherheit zu verbessern. Fortschritte im Bereich des maschinellen Sehens haben den Einsatz kostengünstiger Bildsensoren zur Implementierung von Fahrerüberwachungssystemen ermöglicht. In dieser Arbeit interessierten wir uns für die Schätzung der 3D-Pose von Fahrer*innen mit dem Ziel, menschliche Skelettdarstellungen aus Eingabebildern mithilfe von Deep-Learning-Methoden zu rekonstruieren. Da Deep Learning jedoch große Datenmengen erfordert, ist die Erfassung realer Datensätze kostspielig und herausfordernd. Synthetische Daten bieten eine attraktive Alternative, die die Menge an benötigten realen Daten verringern kann, ohne die Genauigkeit zu beeinträchtigen. Unser Ansatz folgt einem dreistufigen Framework zur 3D-Pose-Schätzung. Die Pose-Schätzungspipeline besteht aus vorgefertigten Modellen für die Personenerkennung und die 2D-Pose-Schätzung. Anschließend verwendeten wir synthetische Daten, um verschiedene 2D-zu-3D-Human-Pose-Lifting-Modelle basierend auf unterschiedlichen neuronalen Netzwerkarchitekturen für die letzte Stufe vorzutrainieren. Schließlich wurden diese Modelle mit zunehmenden Mengen realer Daten feinabgestimmt. Ein Experiment mit Drive&Act als Benchmark-Datensatz zeigte Genauigkeitsgewinne für vortrainierte Modelle bei jeder Menge realer Daten, obwohl diese Gewinne mit zunehmender Menge realer Daten abnahmen. Hybride Modelle wie GraphMLP und GraFormer erzielten die besten Ergebnisse, wenn sie mit geringen bis mittleren Mengen realer Daten trainiert wurden, während JointFormer, ein Transformer-Modell, die anderen übertraf, wenn das vollständige reale Datenset verwendet wurde. Darüber hinaus stellten wir fest, dass das nur mit dem synthetischen Datensatz vortrainierte Lifting-Modell selbst dann eine angemessene Pose-Schätzungsleistung erreichte, wenn keine 3D-Pose-Annotationen für die Ziel-Realweltdaten verfügbar waren, beispielsweise wenn deren Erfassung zu kostspielig ist. Insgesamt deuten die Ergebnisse klar auf den Vorteil der Verwendung synthetischer Daten zur Verbesserung der Genauigkeit der 3D-Fahrer*innen-Pose-Schätzung hin, insbesondere wenn 3D-Pose-Annotationen für reale Datensätze nur eingeschränkt verfügbar sind.One of the most crucial aspects of vehicle manufacturing is ensuring passenger safety. As driver-related factors such as fatigue and distraction contribute to a majority of accidents, monitoring drivers becomes essential to improve road safety. Advances in computer vision have enabled the use of affordable image sensors to implement driver monitoring systems. In this work, we were interested in estimating 3D driver pose with the goal of reconstructing human skeletal representations from input images using deep learning methods. However, deep learning requires large amounts of data, and real-world dataset collection is expensive and challenging. Synthetic data offers an appealing alternative that might reduce the amount of real-world data needed while maintaining accuracy. Our approach adopts a three-stage 3D pose estimation framework. The pose estimation pipeline consists of off-the-shelf models for both human detection and 2D pose estimation. Then, we used synthetic data to pre-train various 2D-to-3D human pose lifting models based on different neural network architectures for the last stage. Finally, we fine-tuned these models with increasing amounts of real-world data. An experiment with Drive&Act as a benchmark dataset revealed accuracy gains for pre-trained models with any amount of real-world data, though these gains diminished as more real data became available. Hybrid models like GraphMLP and GraFormer performed best when trained on low to moderate amounts of real-world data, while JointFormer, a transformer model, outperformed others when trained with the full real-world dataset. In addition, we found that the lifter pre-trained only with the synthetic dataset still achieved reasonable pose estimation performance even when 3D pose annotations for the target real-world data were not available, such as when they are too costly to obtain. Overall, the findings clearly suggest the advantage of using synthetic data for improving the accuracy of 3D driver pose estimation, especially when 3D pose annotations for real-world datasets are limited
