Publikationer från KTH
Not a member yet
54324 research outputs found
Sort by
Influence of liquid lead and lead-bismuth eutectic on three alumina forming austenitic (AFA) steels through slow strain rate testing
Liquid metal embrittlement (LME) in three newly developed alumina-forming austenitic (AFA) alloys, two 50 kg batches and one 5-ton heat, was studied in the temperature range 350–600 °C in liquid Pb and 140–600 °C in LBE using slow strain rate testing (SSRT) in a low-oxygen environment. No significant decrease in the engineering strain was observed in either environment. However, the presence of secondary cracks along the length of the specimen and brittle intergranular areas on the fracture surfaces indicates that the AFA alloys do show a minor degree of embrittlement above 570 °C. This appears to be related to grain boundary wetting by Pb/LBE. At temperatures below 570 °C, this wetting effect does not seem to be strong enough to induce LME in the alloys, and their ability to form a sufficiently protective oxide means that they remain unaffected by LME. The results indicate that the AFA alloy group can perform sufficiently well in liquid Pb/LBE environments, and long-term testing should be carried out to determine their viability as candidate materials for use in Pb- and LBE-based cooling systems.QC 20241023</p
Capturing the Shape and Pose of Horses in 3D
Animals play a significant role in the Earth's ecology and have lived alongside humans throughout history. Studying and understanding their movements and behaviors is important for advancing scientific knowledge and benefiting practical applications. In this thesis, we focus specifically on horses, which are key subjects in both computer vision and biological research due to their strength in speed and unique locomotion systems. Traditional systems for capturing horse motion often rely on attaching sensors or markers to the horse's body. These systems, however, are often limited to constrained environments and difficult to use in natural, unconstrained settings. In contrast, capturing horses using standard video cameras, where horses are observed in their natural environments, presents a more practical solution. However, capturing horses in 3D, specifically the 3D shape and pose, from 2D images is a highly challenging problem due to the ambiguity with only 2D data. To address these challenges, we propose model-based methods to capture the 3D shape and pose of horses from monocular images or videos. We start by presenting hSMAL, a horse-specific 3D parameterized model, capable of expressing diverse horse shapes, which is learned from 3D scan data. We also demonstrate the practical utility of this model in lameness detection, a critical veterinary task for assessing the well-being of horses. Additionally, we present a comprehensive horse motion dataset, collecting data from horses of varying shapes and performing diverse movements, using dense motion capture markers. This motion capture data allows us to animate hSMAL with real horse movements, providing details about how horses move and also tackling the common issue of limited data in animal research. Building on the proposed model and the dataset, we develop data-driven regression methods, to capture horses in 3D from monocular images and videos in an end-to-end manner. First, we integrate multimodal data, combining video clips and audio. Our findings show that incorporating audio enhances the robustness of the method, especially in situations of visual ambiguity and occlusion. Second, we integrate vision foundation models and disentanglement learning with an on-the-fly synthetic data generation pipeline. The pipeline allows the creation of paired data during network training, facilitating the learning of disentangled feature spaces. Together, these approaches enhance the generalization and adaptability of the method, improving performance on images from various domains and other four-legged animals. Through experiments on both our own collected datasets and public datasets, we demonstrate the effectiveness of the proposed methods in advancing horse-specific capture from monocular images and videos. This thesis contributes methodologies for capturing horses from standard video cameras, specifically focusing on the 3D shape and pose, opening new possibilities for animal motion capture and analysis.Djur spelar en betydande roll i jordens ekologi och har levt tillsammans med människor genom historien. Att studera och förstå deras rörelser och beteenden är viktigt för att främja vetenskaplig kunskap och gynna praktiska tillämpningar. I detta examensarbete fokuserar vi specifikt på hästar, som är nyckelämnen inom både datorseende och biologisk forskning på grund av deras styrka i hastighet och unika rörelsesystem. Traditionella system för att fånga hästens rörelser förlitar sig ofta på att fästa sensorer eller markörer på hästens kropp. Dessa system är dock ofta begränsade till begränsade miljöer och svåra att använda i naturliga, oinskränkta miljöer. Att fånga hästar med vanliga videokameror, där hästar observeras i sina naturliga miljöer, är däremot en mer praktisk lösning. Men att fånga hästar i 3D, särskilt 3D-formen och posituren, från 2D-bilder är ett mycket utmanande problem på grund av tvetydigheten med endast 2D-data. För att möta dessa utmaningar föreslår vi modellbaserade metoder för att fånga hästars 3D-form och ställning från monokulära bilder eller videor. Vi börjar med att presentera hSMAL, en hästspecifik 3D-parameteriserad modell, kapabel att uttrycka olika hästformer, som lärs från 3D-skanningsdata. Vi visar också den praktiska användbarheten av denna modell för att upptäcka hälta, en viktig veterinäruppgift för att bedöma hästars välbefinnande. Dessutom presenterar vi en omfattande datauppsättning för häströrelser, som samlar in data från hästar med olika former och utför olika rörelser med hjälp av markörer för täta rörelsefångst. Denna motion capture-data tillåter oss att animera hSMAL med riktiga häströrelser, tillhandahålla detaljer om hur hästar rör sig och även ta itu med det vanliga problemet med begränsad data inom djurforskning. Med utgångspunkt i den föreslagna modellen och datamängden utvecklar vi datadrivna regressionsmetoder för att fånga hästar i 3D från monokulära bilder och videor på ett heltäckande sätt. Först integrerar vi multimodal data, kombinerar videoklipp och ljud. Våra resultat visar att inkorporering av ljud ökar robustheten i metoden, särskilt i situationer med visuell tvetydighet och ocklusion. För det andra integrerar vi grundmodeller för vision och inlärning av disentanglement med en pipeline för generering av syntetisk data i farten. Pipelinen gör det möjligt att skapa parad data under nätverksträning, vilket underlättar inlärningen av disentangled funktionsutrymmen. Tillsammans förbättrar dessa tillvägagångssätt generaliseringen och anpassningsförmågan hos metoden, vilket förbättrar prestandan på bilder från olika domäner och andra fyrbenta djur. Genom experiment på både våra egna insamlade datamängder och offentliga dataset visar vi effektiviteten hos de föreslagna metoderna för att främja hästspecifik fångst från monokulära bilder och videor. Denna avhandling bidrar med metoder för att fånga hästar från vanliga videokameror, speciellt med fokus på 3D-formen och posituren, vilket öppnar nya möjligheter för rörelsefångning och analys av djur.QC 20241129</p
Design procedure for estimating the vertical response of end-bearing piles from free field vibrations produced by a nearby surface load
This paper investigates the vertical response of end-bearing piles in a homogeneous isotropic linear elastic soil on a rigid bedrock subjected to a vertical harmonic point load at the soil’s surface. A numerical model is used to compute vibration responses. A novel system of dimensionless parameters is established to bring insight into the influence of the relationships between soil and pile properties on the dynamic pile–soil interaction and to allow for general conclusions to be drawn. The results show the conditions under which the relationship between the axial stiffness of the pile and the stiffness of the soil has a significant influence on the end-bearing pile response. Different pile group configurations are considered where the vertical response is found to be bounded by the single pile response, justifying its use as a conservative estimate for the group response. Finally, an expression including only two dimensionless parameters, the pile slenderness ratio and the pile–soil stiffness ratio, is proposed for calculating an estimation of the vertical response of an end-bearing pile from the free field vibrations.QC 20250203</p
Aspect-ratio effect on the wake of a wall-mounted square cylinder immersed in a turbulent boundary layer
The wake topology behind a wall-mounted square cylinder immersed in a turbulent boundary layer is investigated using high-resolution large-eddy simulations (LES). The boundary-layer thickness at the obstacle location is fixed, with a Reynolds number based on cylinder height ℎ and free-stream velocity ∞ of 10,000 while the aspect ratio (AR), defined as obstacle height divided by its width, ranges from 1 to 4. The mesh resolution is comparable to DNS standards used for similar wall-mounted obstacles, though with relatively lower Reynolds numbers. The effects of AR on wake structures, turbulence production, and transport are analyzed via Reynolds stresses, anisotropy-invariant maps (AIM), and the turbulent kinetic energy (TKE)budget. In particular, the transition from ‘‘dipole’’ to a ‘‘quadrupole’’ wake is extensively examined as AR increases. With increasing AR, the wake shrinks in both the streamwise and spanwise directions, attributed to the occurrence of the base vortices (AR = 3 and 4). This change in the flow structure also affects the size of the positive-production region that extends from the roof and the flank of the obstacle to the wake core. The AIMs confirm three-dimensional wake features, showing TKE redistribution in all directions (Simonsen and Krogstad, 2005). Stronger turbulence production in AR = 3 and 4 cases highlights the role of tip and base vortices behind the cylinder. The overall aim is to refine the dipole-to-quadrupole transition as a function of AR and accounting for the incoming TBL properties. The novelty relies on proposing the momentum-thickness-based Reynolds number Re as a discriminant for assessing TBL effects on turbulent wake structures.QC 20250122</p
Embodied carbon saving of reusing concrete elements in new buildings : A Swedish pilot study
Reusing the building elements is the highest possible level of circularity for buildings that must be demolished, potentially slowing down climate change. This study explores the embodied carbon reduction of construction of a pilot building with structural elements of reused concrete. The assessment focuses on applying different methodological approaches and discussing the upscaling opportunities of reusing concrete elements from a global warming potential perspective. The assessment shows large embodied carbon savings compared to conventional building practices like recycling the concrete and building with new low-carbon and prefabricated elements. Embodied carbon saving is also obvious when applying alternative system modelling, future market projection and different allocation approaches of the production emissions of the elements. Finally, the study emphasises the need for further research in evaluating the benefits of reusing structural concrete elements broadly, like including the deconstruction impact related to elements for reuse, to be able to draw general conclusions.QC 20241011</p
Fastighetsvetenskaplig forskning vid KTH : Ett supplement, åren 2021-2024
Under 2021 sammanställdes och publicerades en forskningsöversikt (Ekbäck & Mattsson, 2021), där målsättningen var att inventera, dokumentera och systematisera den fastighetstekniska och fastighetsrättsliga forskning som bedrivits och publicerats vid avdelningen för fastighetsvetenskap (och motsvarande äldre organisatoriska enheter), institutionen för fastigheter och byggande på Kungl. Tekniska Högskolan. Den tidsmässiga omfattningen för översikten spänner från 1932 – då den första professuren inom det fastighetstekniska ämnesområdet inrättades vid KTH – fram till hösten 2021. Syftet med detta supplement är att bygga vidare på den tidigare dokumentationen och de analyser som då genomfördes. Supplementet är emellertid inte avsett att läsas fristående, utan för en helhetsbild förutsätts den tidigare rapporten ha studerats. Som underlag för detta supplement har, liksom i den tidigare publicerade forskningsöversikten, en litteratursökning genomförts. Sökmetoden ansluter i stort till metodiken i den tidigare rapporten, givetvis under betydligt enklare och mer begränsade former med hänsyn till den kortare tidsperioden och tillgängliga digitala publikationsregister. Den fullständiga publikationslistan finns som bilaga sist i rapporten. Av den tematiska analysen framgår att inriktningen mot Nya rättighets- och upplåtelseformer i dagsläget kan sägas dominera de fastighetsvetenskapliga forskningsinriktningarna. I de senare årens forskning utkristalliseras även ett nytt tema: Digitalisering av samhällsbyggnadsprocessen. Antalet forskningspublikationer har fortsatt ökat under den studerade perioden, och de engelskspråkiga och internationella studierna dominerar stort.QC 20250115</p
Case study on SAF emissions from air travel considering emissions modeling impact
The environmental impact of air travel, largely driven by fossil-fuel consumption, remains a critical subject of debate. Addressing this challenge requires immediately adopting sustainable practices to mitigate its environmental footprint. While hydrogen and hybrid-electric propulsion technologies show promise for the future, current efforts focus on Sustainable Aviation Fuels (SAF) as a viable near-term solution to reduce aviation emissions while ensuring compatibility with existing aviation infrastructure. This paper examines the environmental impact of air travel, focusing on the emissions associated with conventional fuel and SAF. Using two methodologies, namely the subsonic fuel flow method (SF2) and an improved version of it, the emissions corrected subsonic fuel flow method (EC-SF2), non-CO2 emissions trends are analyzed along a flight trajectory from Stockholm to Bordeaux. The comparison between the two methods underscores the importance of accurate emission modeling, particularly for SAF correction on emission index. The SF2 method reveals that SAF fuels with higher calorific value than conventional fuel increased total HC and CO emissions while decreasing NOx emissions. Conversely, the EC-SF2 method resulted in a more homogeneous emissions reduction trend. Our proposed methodology, which corrects both fuel flow and emission index based on SAF-specific data, could, therefore, offer a more reliable estimation of emissions behavior for SAF. These findings highlight the sensitivity of emissions modeling on environmental assessment.QC 20250218REFMA
A latent class dynamic discrete choice model for travel behaviour and scheduling
In travel behaviour modelling, latent class models are used to represent underlying discrete groupings of behavioural preferences. The paper presents a latent class extension of a dynamic discrete choice model (DDCM) and applies the model to the problem of activity demand generation and scheduling. The DDCM is a recursive multinomial logit model where agents make sequential decisions in time, maximizing the expected future utility of their decisions in a random utility maximization framework. It generates activities and their associated travel within a full day schedule, endogenously respecting agents' inherent time-space constraints. The latent class DDCM builds on the base model by representing heterogeneous lifestyle preferences. A specification of the model is estimated on a Stockholm travel survey and uses age, income level, gender, car ownership and presence of children in the household as classifying variables. The models result in classes which primarily represent modality styles, finding car-, transit- and bike-primary behavioural groups as well as a multimodal group, each linked with different socio-demographic characteristics. The models improve over non-latent class reference models and provide insight into the structure of heterogeneity in travel behaviour preferences in Stockholm.Part of ISBN 978-1853397233QC 20250304</p
Fluctuations on Global and Intermediate Scales for Orthogonal Polynomial Ensembles
Orthogonal polynomial ensembles (OPEs) arise naturally in many models of statistical mechanics, probability theory, combinatorics, and random matrix theory. An important and well-known source of examples of OPEs is the eigenvalues of random Hermitian matrices. Random matrix theory began in 1928 with John Wishart's work, aimed at analysing large datasets. Over the past several decades, the mathematical theory has seen significant advancements, and it continues to be a vibrant area of research today. Another motivation for studying OPEs comes from random tilings. Random tiling models exhibit phase transitions, where different regions of a tiling exhibit distinct behaviour (e.g., frozen vs. liquid regions). This resembles physical systems with phase boundaries, making random tiling a simple yet powerful model to study phenomena such as crystallisation. This thesis is a synthesis of two original studies of the asymptotic behaviour of OPEs. The first part of the thesis consists of an introduction, an overview of the papers, and an outlook of the topics studied. The second part of the thesis includes the original papers. Paper A studies the lozenge tilings of a hexagon with the -Racah weights, which serve as a generalisation of the uniform and -volume weights. We show that the height function for this model concentrates near a deterministic limit shape and that the global fluctuations are described by the Gaussian Free Field (GFF). Paper B is about OPEs at the mesoscopic scales, which is an intermediate scale between global and local ones. We show that a large class of OPEs have the same limiting fluctuations at the edges in the mesoscopic regime. We extend the method of Breuer and Duits (2016) to varying weights. This approach does not make smoothness assumptions on weights. Hence, our results apply to both continuous and discrete OPEs. Ortogonala polynomensembler (OPE:er) förekommer naturligt i många modeller inom statistisk mekanik, sannolikhetsteori, kombinatorik och slumpmatristeori. Hermitiska matrisers egenvärden är en viktig och välkänd källa till OPE:er. Slumpmatristeori inleddes 1928 i och med John Wisharts arbete, som syftade till att analysera stora datamängder. Under de senaste decennierna har den matematiska teorin gjort betydande framsteg och fortsätter även idag att vara ett livligt forskningsområde. En annan motivation för att studera OPE:er är slumpmässig plattläggning. Modeller för slumpmässing plattläggning uppvisar fasövergångar, där olika regioner av en plattläggning uppvisar skilda beteenden (t.ex. frusna eller flytande regioner). Detta speglar fysikaliska system med fasövergångar, vilket gör slumpmässig plattläggning till en enkel men samtidigt kraftfull modell för att studera fenomen som kristallisation. Denna avhandling är en sammanläggning av två originalartiklar om det asymptotiska beteendet hos OPE:er. Den första delen av avhandlingen består av en introduktion, en översikt över artiklarna och en utblick över de studerade ämnena. Den andra delen av avhandlingen består av de originalartiklarna. Artikel A studerar plattläggning av en hexagon med -Racah-vikter, vilka utgör en generalisering av såväl de likformiga som -volymvikterna. Vi visar att höjdfunktionen för denna modell koncentreras nära en deterministisk gränsform och att de globala fluktuationerna beskrivs av det Gaussiska fria fältet (GFF). Artikel B fokuserar på mesoskopiska skalor av OPE:er, vilket är en regim mellan de globala och lokala regimerna. Vi visar att en stor klass av OPE:er uppvisar samma fluktuationer vid kanterna i den mesoskopiska regimen. Vi vidareutvecklar metoden som introducerades av Breuer och Duits (2016) till att även omfatta varierande vikter. Metoden kräver inga antaganden om vikternas släthet, vilket gör att våra resultat gäller för både kontinuerliga och diskreta OPE:er.QC 2025-02-06</p
Toward systematic finite element reconstructions of accidents involving vulnerable road users
To combat the global fatality rates among vulnerable road users (VRUs), prioritizing research on head injury mechanisms and human tolerance levels in vehicle-to-VRU traffic collisions is imperative. A foundational step for VRU injury prevention is often to create virtual reconstructions of real-world collisions. Thus, efficient and trustworthy reconstruction tools are needed to make use of recent advances in accident data collection routines and Finite Element (FE) human body modeling. In this study, a comprehensive and streamlined reconstruction methodology, starting from a video-recorded accident, has been developed. The workflow, that includes state-of-the-art tools for personalization of human body models (HBMs) and vehicles, was evaluated and demonstrated through 20 real-world VRU collision cases. The FE models successfully replicated the vehicle damage that was observed in on-scene photographs of the post-impact vehicle, as well as impact kinematics captured in dash cam or surveillance recordings. The findings highlight how video evidence can considerably narrow down the number of plausible impact scenarios and raise the credibility of virtual reconstructions of real-world VRU collision events. More importantly, this study demonstrates how, with an efficient and systematic methodology, FE might be feasible also for large-scale VRU accident datasets