Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Urban forest: Buildings that capture and utilise carbon dioxide
No full textThe built environment is a major contributor to the global CO₂ emissions (United Nations Environment Programme, 2023), making it essential for us to explore new ways of designing our built environment in a way that not only reduces its emissions but also actively participates in the reduction of greenhouse gases in our atmosphere. This thesis investigates the potential of buildings to function as carbon sinks by integrating Carbon Capture into architectural design. By treating facades as active components in carbon sequestration, the study envisions buildings as part of an “urban forest” that removes CO₂ from the atmosphere, much like trees in a natural ecosystem.
The thesis builds upon existing carbon capture technologies, developed by Dr Klaus Lackner (2009) at Columbia University, and explores their potential architectural integration through a design-driven case study where filters serve a dual purpose of offering shade to reduce solar heat gain while simultaneously capturing CO₂. In an effort to address one of the major challenges of carbon capture, what to do with the captured CO₂, this thesis also explores, beyond sequestration, how the captured CO₂ can be repurposed within the building itself, creating a closed-loop system. The thesis uses data from non building devices to calculate the magnitude and possibility of integrating carbon capture on a buildings facade.
Drawing on data from Lackner (2009), this thesis develops a design proposal located on a site in Gothenburg showing that the building’s closed-loop system can sequester and reuse more CO₂ than is emitted during its construction. With this thesis design and in the given context, the proposed design could capture and utilize approximately 200 tonnes of CO₂ annually
Machine Learning for PROTAC Decomposition and Enhanced Degradation Prediction
No full textPROTACs (Proteolysis Targeting Chimeras) are bifunctional molecules composed of three components that mediate the degradation of target proteins, and are widely used in drug discovery. This project explores the application of machine learning in two key aspects: splitting PROTAC molecules into their three components (E3 ligase, linker, and POI), and predicting the degradation potential of PROTACs on target proteins. We evaluated an existing splitting model using internal data from AstraZeneca. Given recent updates to the public PROTAC dataset, we retrained the degradation prediction model on the expanded data. Additionally, we are transitioning the model from a binary classification task to a regression approach to directly predict degradation-related values such as DC50 and Dmax. We also investigated whether the solvent-accessible surface area (SASA) of lysine residues on the target protein influences degradation outcomes, though no clear relationship was observed
Evaluation of the Clay Cutting Test - Assessment of a new method for measuring undrained shear strength in soils through comparative testing
No full textThe Clay Cutting Test (CC-Test) is a newly developed laboratory method to assess
the undrained shear strength of soil samples. Unlike other tests that provide a single
value of strength, the CC-Test presents a continuous strength profile by measuring
the resistance as a thin wire cuts through the soil sample. This thesis evaluates the
CC-Test through a combination of archive data comparison and in-depth study of
strength profiles from clay samples of varying depths.
In the archive compilation, representative values of shear strength from the CC-Test
were compared to those obtained with direct simple shear (DSS) and fall cone test.
For strengths below 25 kPa a good agreement between the methods was observed.
At higher strengths, the fall cone becomes less reliable and the DSS data are limited.
A parameter study indicated that density, water content and thread diameter have
an impact on the measured resistance. However, in the range of typical soils, the
deviation of the parameters is within ± 10% of the measured strength.
The in-depth analysis focused on 12 stiffer clay samples from 7-35 m depth. For each
sample, uniaxial compression test (UCT), CC-Test and fall cone test were conducted
on the same specimen, to minimize deviation due to natural variation of the soil. The
CC-Test yielded higher values than the fall cone and generally lower than the UCT.
For samples below 20 m, even the upper quartile of the CC-Test strength profile
could be considered conservative compared to UCT. The correlation between the
CC-Test measurements and the UCT was strongest near the locations corresponding
to the ends of the shear plane. In some tests, oscillating patterns appeared in the
measured strength curves, implying an effect of minor material changes within the
soil. These patterns were also present in reference tests, indicating they are not
caused by prior loading or sample disturbance.
The CC-Test shows potential as a complement to established methods for deter mining undrained shear strength due to its simple setup, fast execution, and ability
to present strength variation within the soil specimen. This makes it particularly
useful in quality control of soil samples. For it to be accepted in the industry, clear
guidelines on how to interpret the data is crucial. Further development, such as
measuring the remoulded shear strength with the CC-Test would further strengthen
its applicability
Lågrangmatriskomplettering: En jämförelse av två algoritmer
No full textLågrangmatriskomplettering innefattar algoritmer som fyller ut saknade värden i en matris
under antagandet att den kompletta matrisen är av låg rang. Rapporten har undersökt två
olika algoritmer för långrangmatriskomplettering, singular value thresholding (SVT) och nor malized iterative hard thresholding (NIHT), på slumpmässigt genererad data och ett urval av
databasen Netflix prize data. Rapportens syfte är att bestämma vilken av dessa två algoritmer
som lämpar sig bättre för komplettering av Netflix-datan och slumpmässigt genererad data.
För att mäta detta undersöktes hur nära algoritmerna konvergerar till de kompletta matriser na i termer av bland annat RMSE samt hur lång tid det tar för de olika algoritmerna att köra
givet olika parameterval. Eftersom både NIHT och SVT använder sig av singulärvärdesdekom position som steg i algoritmen undersöktes även hur olika numeriska metoder för att beräkna
dessa påverkar precisionen och tiden det tar att köra algoritmerna.
Rapporten visade att SVT var snabbare och gav högre precision än NIHT när det kommer
till att komplettera Netflix-databasen. Däremot visar NIHT god precision att komplettera
slumpmässigt genererad data och kan även göra det snabbare än SVT om en tillräckligt god
uppskattning av rangen anges i förväg. Testerna visade även att NIHT kan ge bättre resultat
om vissa parametrar i algoritmen justeras, vilket kan vara av intresse för vidare forskning.
Nyckelord - Lågrangmatriskomplettering, normalized iterative hard thresholding, singular va lue thresholding, singulärvärdesdekomposition
Advanced driver assistance systems (ADAS) usage and resulting safety effects
No full textThe development of active safety systems such as Adaptive Cruise Control (ACC),
Lane Keeping Aid (LKA), and Volvo’s Pilot Assist (PA) has the potential to improve
road safety. These systems assist drivers in performing driving tasks through the
use of cameras, radar, and Light Detection and Ranging (LIDAR). However, to
what extent are these systems actually used in real-world driving? When and where
are they activated, and how do their usage differ among different drivers? What
measurable safety effects can be achieved by using these systems in various driving
scenarios?
This thesis investigates usage patterns of Advanced Driver Assistance Systems (ADAS),
with a particular focus on time gap settings of PA and ACC, types of system deactivations,
and lane deviation in comparison to manual driving. In addition, the
thesis explores general ADAS usage and the contexts in which these systems are engaged.
Typical driving scenarios, such as car following, car approaching, and cut-in
scenarios are defined and analyzed to understand how ADAS operates within each
scenario.
The result showed that the usage of ADAS, both ACC and PA, are used to a
greater extent on high speed roads such as highways and expressways compared to
rural roads. Additionally, drivers who engage ACC and PA during car following
and car approaching scenarios tend to maintain greater safety margins in terms of
time gap and Time to Collision (TTC) compared to those driving manually. The
analysis further revealed that drivers who set larger time gaps when using ACC or
PA also tend to maintain greater time gaps during manual driving. Drivers with PA
activated also showed to maintain the lowest mean lane deviations, while manual
driving resulted in the highest mean lane deviations on highways, expressways, and
rural roads
Leveraging Generative AI for Predictive Maintenance - Building a Knowledge Base for Fault Diagnosis
No full textFault diagnosis is a complex challenge for industrial production. This thesis develops and evaluates a predictive maintenance assistant integrating large language models (LLM) with retrieved-generation techniques (RAG). By constructing a unified knowledge base comprising sensor data, event logs, and equipment manuals, the system enhances fault diagnosis in industrial settings. The system analyzes sensor data and links it to event logs, matching sensor data with faults. Meanwhile, it connects faults with equipment manuals via RAG, forming a unified knowledge framework. It generates readable and accurate fault diagnostics via LLMs and searching relevant technical documents. It is adaptive, transferable, and capable of integrating specific knowledge. Experiments conducted using a simulated drone assembly production line demonstrate significant improvements in diagnostic accuracy, interpretability, and reliability, effectively addressing common issues such as hallucinations and unsupported claims found in traditional LLM applications. The findings highlight the practical feasibility of deploying advanced AI-driven predictive maintenance solutions, emphasizing the importance of semantic richness and structured knowledge integration
Unified Calendar and Task Management: A Web Application Integrating Google Services
No full textThis thesis project follows the design, development and implementation of a React based web application inspired by features of physical organizational systems. The primary objective was a unified interface capable of gathering information from
various online services, specifically Google Calendar and Google Task. The project used the React framework for front-end development, integrating with respective Google API’s through Google OAuth 2.0 for authentication and data retrieval. The report follows an adapted agile, sprint-based methodology for solo development, focusing on iterative feature implementation from core calendar functionality to API integration and user interface enhancements