Technical University of Munich

The current Eurocode EN 1993-1-8 provides no explicit design guidance for single-sided fillet (SSF) welds subjected to bending, although such configurations occur in practical applications. This study addresses this gap through an experimental investigation of the load-bearing behaviour of SSF welds under pure bending and combined shear–bending loading. A total of 60 tests were carried out, covering different weld throat thicknesses, base materials and filler metals. The resulting dataset forms the basis for assessing the applicability of the resistance models for bending and combined shear–bending loading provided in EN 1993-1-8. The experimental results were compared with the resistance model of the directional method, and a reliability analysis of the individual resistance models was performed in accordance with EN 1990, Annex D. The results show that SSF welds can be designed in accordance with the Eurocode provisions. For pure bending, the resistance predictions remain conservative, even when a plastic bending resistance is assumed. For combined shear–bending loading, the proposed application leads to less conservative and more economical resistance predictions. The resistance formulations of EN 1993-1-8 can be applied to SSF welds under bending without considering the normal-stress criterion. • Experimental investigation of the out-of-plane bending behaviour of single-sided fillet (SSF) welds. • Test series on SSF Welds under pure bending and combined shear–bending loading. • Statistical evaluation of resistance models according to EN 1990, Annex D. • Design-oriented guidance for applying EN 1993-1-8 to SSF welds subjected to bending.

With the rapid integration of generative AI (GenAI) into higher education, concerns over cognitive offloading, overreliance, and diminished critical thinking underscore an urgent need to prioritize learning-to-learn (L2L) competencies. However, the lack of a clear and detailed conceptualization of L2L, a key 21st-century skill for lifelong learning, hinders interdisciplinary research and limits the development of informed, pedagogically sound AI applications. This paper presents a scoping review of L2L definitions within pedagogical and psychological literature, based on sources retrieved from ERIC, Scopus, and Web of Science. The review follows the PRISMA-ScR framework and identifies 21 relevant publications. We propose a novel three-layered framework organized by conceptual broadness: Dimensions (dispositions like cognitive and metacognitive skills), Processes (actionable activities such as self-regulation), and Tools (concrete strategies like retrieval practice), providing entry points for researchers from varying disciplines. We further highlight existing GenAI research efforts addressing these layers, outline current limitations, and propose directions for future exploration, with a focus on higher education. By unifying L2L theory with GenAI practice, this framework provides an actionable foundation for educational technologies in an AI-driven era. • Three-layer learning-to-learn (L2L) framework (Dimensions, Processes, Tools) to guide AI‑enhanced learning system design. • Synthesizes L2L definitions from 21 studies identified by a PRISMA-ScR review and offers actionable paradigms for teaching L2L, for example, using AI-supported systems. • Maps L2L components to GenAI application use cases in higher education. • Positions L2L as key to reducing GenAI overreliance and fostering learner agency.

Mapping air pollution across space remains challenging, even in countries with dense monitoring networks. In Germany, pollutant levels can change over short distances because of traffic, land use, and meteorological conditions, while national assessments often rely on unevenly distributed monitoring stations. This study examines how openly available satellite observations and reanalysis data can support annual modelling of NO2 and PM2.5 across Germany from 2019 to 2024. Sentinel-5P (NO2 and CO), MODIS Normalized Difference Vegetation Index (NDVI) and Multi-Angle Implementation of Atmospheric Correction (MAIAC) aerosol optical depth (AOD), and ERA5-Land meteorological variables were combined with EuroAirnet observations, and seven machine-learning algorithms were evaluated. Model performance was assessed using random cross-validation, an independent test set, and spatial cross-validation, while SHAP (Shapley Additive Explanations) values were used to interpret predictor contributions. For NO2, Random Forest achieved the highest accuracy (R2=0.68; RMSE=5.87μgm-3), with SHAP analysis identifying tropospheric NO2 and vegetation structure (NDVI) as the most influential predictors. PM2.5 proved more difficult to model at the annual scale: Gradient Boosting performed best (R2=0.50; RMSE=11.53μgm-3), with surface pressure, NDVI, and co-emitted gases emerging as key variables, while MAIAC AOD contributed little independent information when aggregated annually. A sensitivity analysis showed that including a static road-density layer improved NO2 estimates near monitoring sites but provided limited gains under spatial validation. The resulting concentration maps reproduce the main national patterns observed in the monitoring network, showing a decline in NO2 and more regionally variable behaviour for PM2.5. Although annual predictors cannot capture short-term variability or highly localised emission sources, the study provides a transparent and reproducible framework for national-scale air-quality assessment based entirely on open global datasets and highlights the potential to integrate additional Earth observation and climate reanalysis products in future research.

Sub-visible particle analysis using flow imaging microscopy combined with deep learning has proven effective in identifying particle types, enabling the distinction of harmless components such as silicone oil from protein particles. However, the scarcity of available data and severe imbalance between particle types within datasets remain substantial hurdles when applying multi-class classifiers to such problems, often forcing researchers to rely on less effective methods. The aforementioned issue is particularly challenging for particle types that appear unintentionally and in lower numbers, such as silicone oil and air bubbles, as opposed to protein particles, where obtaining large numbers of images through controlled settings is comparatively straightforward. In this work, we develop a state-of-the-art diffusion model to address data imbalance by generating high-fidelity images that can augment training datasets, enabling the effective training of multi-class deep neural networks. We validate this approach by demonstrating that the generated samples closely resemble real particle images in terms of visual quality and structure. To assess the effectiveness of using diffusion-generated images in training datasets, we conduct large-scale experiments on a validation dataset comprising 500,000 protein particle images and demonstrate that this approach improves classification performance with no observable downside. Finally, to promote open research and reproducibility, we publicly release both our diffusion models and the trained multi-class deep neural network classifiers, along with a straightforward interface for easy integration into future studies, at https://github.com/utkuozbulak/svp-generative-ai.

Post-disturbance recovery is a central element of forest resilience against intensifying disturbance regimes. Although recovery signals are strong across Central European forests, the relative roles of different factors contributing to recovery remain incompletely understood. As climate change increasingly challenges recovery, elucidating these processes is essential to adapt forest management to changing climate and disturbance regimes. We extended and applied a biologically grounded model of forest growth to remote sensing data to quantify how management shapes two key drivers of canopy recovery—disturbance legacies and post-disturbance height growth—across Bavaria, Germany. We combined 23,036 ha of quality-filtered photogrammetric canopy height model data with a Landsat-based disturbance map, a forest ownership map and environmental covariates in a Bayesian modelling framework. Post-disturbance growth rates were governed primarily by forest type and site conditions, whereas management strongly influenced disturbance legacies, i.e. the remaining post-disturbance vegetation height structure on site. Legacies varied widely across management types: Federal and set-aside forests retained the highest level of disturbance legacies, while private forests had the lowest legacy levels. Despite marginally lower growth rates, set-aside areas had recovery trajectories that were comparable to managed forests. The median recovery time to 5 m mean canopy height was 14.3 years over all forest and management types. Set-aside areas exhibited the greatest variation in recovery trajectories. We here show that (i) management affects disturbance legacies more strongly than post-disturbance tree growth, (ii) set-aside areas do not differ in recovery speed from managed areas, and (iii) legacies are diversifying forest recovery trajectories, with potential implications for future forest resilience. Our results underline that the post-disturbance reorganization window is a crucial period for management to influence long-term forest development. The framework presented here provides a scalable approach to monitor structural recovery and guide adaptive forest policy and management under increasing disturbance. • Forest management in Central Europe affects post-disturbance recovery more via legacies than tree growth rates. • Set-aside forests recover their canopy height equally fast as managed forests in Central Europe. • Homogenizing and removing disturbance legacies can reduce forest canopy variation across forest stand development. • We combined a biological growth model with remote sensing data to assess forest canopy recovery.