Digitala Vetenskapliga Arkivet - Academic Archive On-line
Not a member yet
734120 research outputs found
Sort by
Enhancing railway infrastructure monitoring with AI: A machine learning approach for event detection
No full textThis study presents a machine learning-based framework for detecting critical events in railway infrastructure by analyzing vibration signals from trackside accelerometers. Traditional maintenance is often reactive and labor-intensive, but this approach uses continuous sensing and data analytics to enable proactive, real-time monitoring. The research leverages a comprehensive pipeline that includes data preprocessing, segmentation of time-series data into one-second intervals labeled as "event" or "no-event", and the extraction of statistical, temporal, and spectral features like crest factor and kurtosis. Key contribution of this work is the systematic evaluation of 72 algorithm-feature selection configurations. Twelve diverse classification algorithms were compared, including tree-based, linear, and neural network models. Extensive hyperparameter optimization was performed to benchmark performance using metrics such as accuracy, precision, recall, and F1-score. The Multi-Layer Perceptron (MLPClassifier) achieved a peak cross-validation accuracy of 98.89% with the full feature set. The study also found that comparable accuracy (98.67%) could be achieved with a 47% dimensionality reduction using Recursive Feature Elimination (RFE) with only eight features, demonstrating a balance between efficiency and performance. The findings provide actionable insights for developing scalable, high-performance event detection systems.Full text license: CC BY 4.0;</p
Strain engineering of ScN thin films and its effect on optical, electrical, and thermoelectric properties
No full textScandium nitride (ScN) is a cubic NaCl-structured, degenerate, narrow-bandgap, n-type semiconductor that exhibits remarkable semiconducting, thermoelectric and plasmonic properties. However, its properties are sensitive to several types of defects, such as crystal defects, morphology, intentional or unintentional doping. For the purpose of reducing the deposition temperature of ScN, a series of films were deposited in the temperature range of 250–850 °C using a high-power impulse magnetron sputtering technique. While the stoichiometry and crystal structure remained unaffected in the sample series, the optical and electrical properties were affected when the temperature was decreased. Using in-depth XRD, optical and electrical characterizations, the effect of strain and dislocations on the semiconductor properties of ScN was evaluated. A reduction in the deposition temperature from 850 °C to 450 °C yielded a slow change in the electrical and optical properties, while a drastic change occurred for the films deposited below 450 °C. The main cause of the deterioration of the electrical transport properties (σ/10 000; n/100, and µ/100) was attributed to a high dislocation density (1011 cm−2) along with a rhombohedral distortion of the ScN cell (α: 90° → 88.6°), which was the main cause of the variation in the electrical transport. The presence of dislocations/crystal defects in the film generated defect states near the edges of the valence and conduction bands, softening the edges and impacting the electron density and mobility. The best thermoelectric properties of ScN were obtained when it was grown at 850 °C and were further modified and altered by strain engineering.Funding Agencies|Funding Agencies|Fundao para a Cincia e a Tecnologia [LA/P/0037/2020 (10.54499/LA/P/0037/2020), UID/50025/2025 (10.54499/UID/50025/2025)]; Vetenskapsrdet [2019-00191, 2021-03826]; Carl Tryggers Stiftelse fr Vetenskaplig Forskning [CTS14:310, CTS16:303, CTS20:272, CTS23:2746, CTS25:3972]; Agence Nationale de la Recherche [ANR-11-LABEX-0017-01, ANR-18-EURE-0010, ANR-21-EXES-0013]; Stiftelsen fr Strategisk Forskning [RIF14-0053]; Energimyndigheten [436061]; Knut och Alice Wallenbergs Stiftelse [KAW-2020.0196]</p
Quality Evaluation of Generative AI Systems : Processes, Metrics, Methods, and Frameworks for Industrial Software Engineering
No full textGenerative Artificial Intelligence (GenAI) is being rapidly adopted in software engineering, introducing a paradigm shift toward human-AI co-creation. However, the non-deterministic, probabilistic, and often black-box nature of GenAI models presents challenges for traditional software quality assurance. Conventional verification and validation techniques are insufficient to handle outputs that are neither predictably correct nor incorrect, but rather stochastically plausible. This discrepancy creates an urgent need for practical processes, metrics, and new governance frameworks to evaluate and manage the quality of GenAI systems in industrial environments.This thesis examines how industrial organizations adopt GenAI, identify metrics, and evaluate system qualities in alignment with ISO quality standards. Case studies were employed to explore real-world adoption processes, identify context-specific industrial metrics, and uncover practical insights within organizations. A snowballing literature review was conducted to systematically identify, categorize, and synthesize academic metrics for evaluating the output of GenAI systems. Finally, a controlled experiment was designed to quantitatively test the efficiency (e.g., E2E generation time) and effectiveness (e.g., accuracy) of GenAI agent choices. The main contributions of this thesis are a synthesized actionable model and framework grounded in both industrial practice and quality standards. The first contribution is a four-stage adoption model, denoted as the IMRM model (Innovate → considerations, Measure → metrics, Realize → values, Manage → improvements) that integrates early-stage risk assessment (e.g., legal, security, and licensing) andquality evaluation throughout the GenAI adoption and usage.The second contribution presents a detailed framework that connects risks andmetrics to concrete decision support, justifying the business value (e.g., quality gates) and technical trade-offs of GenAI solutions. The third contribution provides a structured mapping of GenAI quality to ISO/IEC 25010, 25023, and 25059 characteristics, attempting to ground practical evaluation needs within a standardized vocabulary. This thesis concludes that a structured quality evaluation process, which prioritizes risks and context, is a valuable approach intended to support building the business confidence required to leverage GenAI for efficient and effective software engineering in industry
Tense and Aspect in Multilingual Romance Language Education
No full textThis book brings together multilingualism, tense and aspect, and Romance language education. The chapters present theoretical and empirical research on the teaching and learning of French, Spanish, Italian, and Catalan in different educational contexts with data collected in Europe, the Americas, and Asia from learners with various linguistic backgrounds. With its clearly delineated sections on learning, teaching, and sociolinguistic variation, the volume makes an important contribution to the rich field of inquiry of second and third languages that is of significance to researchers, teachers, and learners alike
Exploring the impact of Italian, English and Swedish on the L3 acquisition of tense and aspect in French : A longitudinal case study
No full text‘I am a teacher of multilingual mathematics students’: A study onmultilingual mathematics teachers in multilingual mathematicsclassrooms — with a focus on professional identity
No full textThis thesis focuses on multilingual mathematics teachers in multilingualmathematics classrooms through the lens of professional identity. The aimis to increase the knowledge about the formation and development ofmultilingual mathematics teachers’ professional identities in relation tomultilingual mathematics teaching. A case study approach is adopted, inwhich Neda, a multilingual primary school mathematics teacher, and Aza,a multilingual mathematics teacher at the lower secondary level in Sweden,have been followed over two academic years. The Patterns of Participation (PoP) framework is used as a conceptual andanalytical framework. As a conceptual framework, it enables theinterpretation of teachers’ professional identities through their shiftingexperiences of being, becoming, and belonging as multilingualmathematics teachers in multilingual schools and classroom interactions.Analytically, PoP enables the construction of several practices and figuredworlds relevant to teachers’ identities, as well as the illustration andconnection of practices and figured worlds significant for teachers inmultilingual classroom interactions. The findings reveal that multiple, occasionally theoretically conflictingpractices and figured worlds are relevant for the teachers in relation tomultilingual mathematics teaching. A multilingual approach to teaching,as well as a language-focused approach to mathematical concepts andcommunication, is emphasised. However, the use of students’ mothertongue is less visible during classroom interactions. Teachers balancebetween the use of students’ mother tongue and the language ofinstruction, especially during interactions with newly arrived mathematicsstudents. Teachers’ identity trajectories over two academic years of school change,though, highlight the teachers’ shifting experiences of being a multilingualmathematics teacher more than being a multilingual mathematics teacher.The cases make visible that teachers being or becoming recognised asqualified and successful mathematics teachers shift positively when theirsense of belonging is aligned within the school and among colleagues.Experiences of being, becoming and belonging as a multilingualmathematics teacher in the school are linked to the teachers’ professionaland educational language-related and multilingual experiences in the past. In conclusion, being a multilingual mathematics teacher is entangled inteachers’ professional experiences and educational realities, whichforeground the official language of instruction for multilingualmathematics teaching. However, the identity development of multilingualmathematics teachers in schools can be enhanced by embracing theirexperiences and expertise as both multilingual and mathematics educators
Molecular dynamics simulations of a hexagonal liquid crystal phase to study drug partitioning and release mechanisms
No full textLiquid crystal nanoparticles (LCNPs), such as hexosomes based on an internal hexagonal phase (HII), enhance lipid nanoparticle-mediated drug delivery by improving drug solubility, stability and absorption. LCNPs can also be tailored for specific biological environments by incorporating non-ester-linker lipids into the HII nanostructure. In this study, we developed an HII model system with a 90:10 phytantriol:farnesol ratio based on experimental data and conducted all-atom molecular dynamics simulations. The model remained stable across various water-to-lipid ratios, and the structural effects observed were consistent with prior experimental data. We used this model to examine the localization and interactions of antibiotics vancomycin and clarithromycin. Clarithromycin, being highly lipophilic, associated mainly with the lipid phase, while vancomycin localized at the water-lipid interface due to its amphiphilic nature. An extended HII system with repeating units enclosed in Pluronic F127 polymers was also constructed. Simulations showed that hydrogen bonding between Pluronic F127 and water facilitated water influx into the HII phase, causing interfacial reorganization. To investigate drug release, we performed umbrella sampling simulations. The resulting energy profiles indicated that polymer-water-lipid interactions lowered the energy barrier for vancomycin release compared to clarithromycin. This was confirmed by in vitro release studies, where vancomycin exhibited a higher release rate. Overall, this model provides molecular-level insights into drug loading, partitioning, and release from HII systems, supporting the design of more effective drug delivery formulations
Barriers and opportunities for dynamic adaptation of coastal railways
No full textRailways are vital for climate mitigation as a low-carbon mode of transport but must also be resilient to climate change impacts. This study explores dynamic adaptation strategies for coastal railway infrastructure in Sweden, focusing on Trelleborg. Stakeholder workshops informed adaptation pathways using a Dynamic Adaptive Pathways Planning Light (DAPP)-light approach, which enhances stakeholder engagement and supports long-term decision-making under uncertainty. A Windows of Opportunity framework was applied to assess governance conditions for implementation. Findings show that railway adaptation must be integrated with broader urban planning. Key barriers include fragmented responsibilities, rigid land use policies, limited funding, and reluctance to engage in long-term planning. Opportunities lie in reframing existing policies, fostering intersectoral collaboration, and adopting incremental planning.
Learning 3D Texture-Aware Representations for Parsing Diverse Human Clothing and Body Parts
No full textExisting methods for human parsing into body parts and clothing often use fixed mask categories with broad labels that obscure fine-grained clothing types. Recent open-vocabulary segmentation approaches leverage pretrained text-to-image (T2I) diffusion model features for strong zero-shot transfer, but typically group entire humans into a single person category, failing to distinguish diverse clothing or detailed body parts. To address this, we propose Spectrum, a unified network for part-level pixel parsing (body parts and clothing) and instance-level grouping. While diffusion-based open-vocabulary models generalize well across tasks, their internal representations are not specialized for detailed human parsing. We observe that, unlike diffusion models with broad representations, image-driven 3D texture generators maintain faithful correspondence to input images, enabling stronger representations for parsing diverse clothing and body parts. Spectrum introduces a novel repurposing of an Image-to-Texture (I2Tx) diffusion model—obtained by fine-tuning a T2I model on 3D human texture maps—for improved alignment with body parts and clothing. From an input image, we extract human-part internal features via the I2Tx diffusion model and generate semantically valid masks aligned to diverse clothing categories through prompt-guided grounding. Once trained, Spectrum produces semantic segmentation maps for every visible body part and clothing category, ignoring standalone garments or irrelevant objects, for any number of humans in the scene. We conduct extensive cross-dataset experiments—separately assessing body parts, clothing parts, unseen clothing categories, and full-body masks—and demonstrate that Spectrum consistently outperforms baseline methods in prompt-based segmentation.QC 20251219</p
Decentraliserade behandlingssystem för bad-, disk- och tvättvatten: funktion, mikrobiella risker och mikroplaster
No full textGreywater originates from kitchen sinks, dishwashers, handbasins, showers, and laundry. Greywater can account for 70–90% of the domestic wastewater volume and contains organics, nutrients, pathogenic microorganisms, micropollutants, and microplastics. Effective treatment can unlock the potential of greywater for non-potable reuse purposes like urban landscaping or irrigation. The overall aim of this thesis was to investigate on-site greywater treatment systems which included on-site systems, two green walls, and a treatment wetland, and investigate the treatment in terms of organic matter, nitrogen (N), phosphorus (P), pathogenic microorganisms and microplastics (MPs), including the potential resource recovery and safe reuse of greywater. Among the eight on-site systems (1–5 persons) investigated, commercial systems included three type A, two type B, and C system. Type D was a conventional sand filter. After the pre-treatment septic tanks, the treatment unit of type A consisted of a geotextile-fitted trickling filter over a sand bed, type B contained a mineral wool filter, and type C had fine-meshed plastic filters. The two green wall studies were conducted at a testbed facility, RecoLab, which received greywater from a newly developed urban city district (ca 1000 people). The treatment of an indoor vertical flow (VF) green wall with five filter materials (pumice, biochar, hemp fiber, spent coffee grounds, and composted fiber soil (a paper industry byproduct)) was investigated with the flow rates of 4.5, 9, and 18 L/d. The outdoor horizontal flow (HF) green wall with four levels filled with biochar, pumice, and LECA as filter material was investigated for one year, using a subsurface horizontal flow of 430 L/d. A long-term evaluation of the performance of a constructed wetland for treating greywater from a residential building (ca 100 people) in Norway was conducted, using data from the period 2001–2024. The constructed wetland consisted of a biofilter with Filtralite® material and a horizontal subsurface filter with Filtralite®P, for enhanced phosphorus removal. The treatment efficiency of the systems was highly influenced by the filter material and flow rates, while seasonal temperature changes had a low impact. All the systems demonstrated effective treatment of greywater and met the local discharge guideline of 80% BOD reduction and <3mg/L of P in the effluent. However, only the VF green wall and constructed wetland could produce an effluent with <1 mg P/L, a limit for facilities located in sensitive areas. Among the studied filter materials, sand, biochar, and Filtralite® were the most efficient (log10 reduction up to 4) in the bacteria Escherichia coli, enterococci, Clostridium perfringens, Pseudomonas aeruginosa, Legionella spp., and met the European Commission’s guideline for reuse of reclaimed water in agriculture. The quantitative microbial risk assessment (QMRA) on effluent greywater from the constructed wetland, for multiple exposure scenarios (16 exposures/year) of accidental ingestion of 1 mL, indicated safe reuse in a water cascade during the summer season with regard to E. coli and C. perfringens. In addition, using TED-/Py-GC/MS, high variability of MPs was observed in greywater from the different sources of generation, while all the filter material of the respective treatment systems effectively retained the MPs, except for mineral wool and hemp. The findings of this thesis could contribute to the development of a more resource-efficient wastewater management and Water-Food-Energy nexus by demonstrating the potential of decentralized greywater treatment systems