Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Learning to mitigate reliance on features with missing values in interpretable prediction models
No full textIn the healthcare area, it is common for datasets to contain observations that are missing for the corresponding features. Predicting outcomes with such datasets in supervised learning tasks often results in outcomes that are heavily influenced by these missing values. This thesis modifies two original machine learning algorithms and introduces two novel models: the Least Absolute Shrinkage and Selection Operator Mitigating Reliance (LASSOMR) and the Decision Tree Mitigating Reliance (DTMR). Both models are designed to reduce dependency on features with missing values during predictions. This reduction is achieved by penalizing features that have missing values, thereby decreasing the model’s reliance on these features. The synthetic dataset and real-world dataset are used to explore that DTMR and LASSOMR models give a larger penalty to the features that have larger missing ratios. As a result, the coefficient value of the features becomes less leading to the goal of relying less on features having missing values. Additionally, real-world datasets with missing values evaluate the performance of these models against baseline methods, confirming that the models perform comparably while effectively mitigating reliance on missing value features
Frihamnskyrkan - Sermon of the Mount
No full textThe model shows an abstract version of the Sermon on the mount, which served as an inspiration to the architect when designing the church hall. Similar to its reference in the bible the goal was to gather many different people up on the ”hill” and listen to the sermon of the church
Sensorbaserad verifiering av renoverade betongkonstruktioner
No full textDetta examensarbete undersöker möjligheten till att installera och använda
fiberoptiska sensorer, kallade Distributed Optical Fibre Sensors (DOFS), för att
verifiera och övervaka den strukturella hälsan av renoverade betongkonstruktioner
med fokus på marina miljöer. Arbetet har genomförts i samarbete med Göteborgs
Hamn och Chalmers forskningsbaserade projekt Sens-IT, där syftet är att möjliggöra
proaktivt underhåll av infrastruktur genom kontinuerlig datainsamling.
Projektet omfattar litteraturstudier, teoretiska beräkningar samt ett praktiskt
genomförande av fiberoptikinstallation och renoveringsgjutning av pålar från
Göteborgs Hamn. Den teoretiska delen innefattar litteraturstudier kring betong i
utsatta miljöer, renovering av betongkonstruktioner, armeringskorrosion,
exponeringsklasser, samt krypning och krympning.
Den praktiska delen innefattar ett flertal moment med syftet att renovera befintliga
betongpålar och integrera fiberoptiska sensorer för framtida strukturell övervakning.
Fyra påldelar, ursprungligen från två marina betongpålar förbereddes genom kapning
till hanterbara längder och vattenbilning för fiberoptikinstallation, formbyggnation,
gjutning. Provning av betongens hållfasthet gjordes även genom kubtest.
Resultatet är ett tekniskt genomförbart metodförslag för fiberoptikinstallation vid
renovering av betongkonstruktioner. Tekniken har potential att bidra till mer hållbart
och kostnadseffektivt underhåll av hamninfrastruktur genom att möjliggöra tidig
upptäckt av skador och tillståndsförändringar.
Ett betydande bidrag till Sens-IT projektet är att detta examensarbete möjliggör
långtidsövervakning av renoverade betongelement, något som inte tidigare varit
möjligt. Hittills har projektet enbart omfattat nygjutna betongelement vilket begränsat
möjligheterna till att utvärdera sensorernas prestanda i återställda eller skadade
element. Då Göteborgs Hamn genomför mycket renoveringar av konstruktioner kan
detta bidra med ny information som tas i beaktning vid underhållsarbetet
Impact of Road Work Zones on Traffic Flow and Safety - A VISSIM-Based Analysis of Driving Behavior and Risk Factors
No full textTo achieve Vision Zero and eliminate all fatalities and severe injuries in road traffic, it is
necessary to improve road safety for both road users and road workers. Accidents and
incidents that occur in work zones could be prevented by following national regulations
and implementing measures such as putting up signs, barriers, and speed limits. Further,
the work zone safety is closely related to driving behavior.
The aim of this study is to examine how work zones affect traffic flow and road safety
with a focus on the Swedish driving behavior and national regulations. This study fills
a research gap addressing the lack of simulation studies on work zones in Sweden. A
literature study and interviews were conducted to present the regulations and understand
the current situation regarding road safety. A case study area was observed and recorded
in connection with a work zone. Machine learning was used to extract parameters
from the Swedish traffic flow, which was used to calibrate a simulation scenario that
correlated with a general Swedish work zone traffic flow. The model was improved
by changing parameters that mimic a lower speed limit in the work zone and driving
behavior with earlier merging.
It is found that Swedish drivers generally exhibit non-aggressive driving behaviors, in cluding gap acceptance, adherence to speed limits, and early merging. There is a vari ation of risks of work zone safety, where several situations are believed to occur due
to stressed drivers or a lack of information. Safety issues due to driving behavior were
tested in the traffic simulation tool VISSIM, where an improved design simulation sce nario illustrated a work zone where the speed limit was reduced and drivers merged
earlier compared to the calibrated and adjusted scenario. The improvements impacted
travel time by 2.2%, an insignificant increase compared to the enhanced safety to which
the lower speed contributes
Characterisation of a 4680 Cylindrical Cell: Insights Into Cell Design and Performance
No full textLi-ion batteries are expected to tackle the demands of powering the automotive industry
shift toward electrification. Selecting the right components by considering
attributes like performance, safety, and longevity is crucial. These batteries come
in various forms: cylindrical, pouch, and prismatic, each with its own set of pros
and cons in terms of manufacturing, packaging, energy and power. Recently, there
has been a proposal for cylindrical cells of the 46xx series, which allows increased
capacity, potentially making them more viable for battery packs. This project aims
to provide insights into the design and electrical testing of a 4680 cylindrical Li-ion
cell with a lithium iron phosphate (LFP) cathode. This cell chemistry has recently
gained more interest within the automotive sector due to its cost and safety.
Insights into the cell design have been obtained through a thorough cell-teardown
process. Unlike prismatic cells, which typically use aluminum cans, the cylindrical
cell features a steel can, presenting unique challenges during the disassembly process.
To address this, two methodologies for opening cylindrical cells were developed and
reviewed. Additionally, the design for the cell considered within this study differs
from existing tabless architectures for 4680 cylindrical cells found in literature.
The materials harvested from the cell teardown were characterized to obtain information
about the cell components, such as electrodes, electrolyte, separators,
tabs, etc. Furthermore, electrical testing methods such as galvanostatic charge and
discharge, hybrid pulse power characterization, and electrochemical impedance spectroscopy
(EIS) provided additional valuable insights into cell performance.
This study contributes to the benchmarking of 4680 cylindrical cell series, which are
otherwise scarcely documented in literature
Informationsdriven organisationskultur för effektiv vakanshantering i kommersiella fastigheter; processer, verktyg och nyckelfaktorer; en fallstudie om hur datadrivet arbete minskar vakanser och stärker affärsnyttan
No full textThis study explores how commercial real-estate companies can lower office-vacancy rates
through improved information-management processes and more coordinated work routines.
The aim is to deepen understanding of the mechanisms that create vacancies in the Gothenburg
region and to pinpoint critical leverage points where targeted actions can reduce vacancy levels
and enhance the day to day working strategies.
The project is designed as a qualitative single case study of one property company, combining
semi-structured interviews and document analysis with methodological triangulation to
strengthen validity. Data were gathered from key roles including area and leasing managers,
technical and commercial property managers, and business development staff to capture
perspectives across the leasing value chain.
Based on our conducted case study, three overarching patterns emerge that constitute key
components of the study’s findings. These patterns highlight organizational practices,
information management, and the role of organizational culture in the handling of vacant
premises:
1. The study shows that working methods and information management related
to vacant premises vary between different functions and individuals, reflecting
a decentralized organizational practice.
2. The results indicate variation in the presence of formalized structures for the
documentation, updating, and systematic follow-up of vacancy-related
information.
3. The organizational culture is characterized by a high degree of flexibility and
individual adaptation, creating room for independent action but also posing
challenges for effective coordination and information flow.
IV
Together, these observations form the foundation for our continued analysis and the
development of proposals for improvement measures.
The study concludes that technical solutions such as an integrated Business Intelligence system
must be paired with a strong information culture in which data sharing, standardisation, and
follow-up are embedded in reward and management systems. This dual strategy is expected to
reduce person dependence, shorten the time from vacancy to lease signing, and strengthen
competitiveness in a market with growing office supply
The Genetic Compatibility between Antibiotic Resistance Genes and Bacterial Hosts: Evaluated by measuring differences in k-mer distributions and comparing gene codon usage with tRNA availability
No full textAntibiotic resistance is a growing global health concern, driven by bacteria exchanging
antibiotic resistance genes (ARGs) through horizontal gene transfer. The factors
influencing the spread of ARGs across bacteria are not entirely understood, though
genetic compatibility has been proposed as a contributing factor. This project aimed
to explore genetic compatibility between ARGs and bacterial genomes by creating
two metrics. The first metric, the 5mer score, was created by looking at nucleotide
composition, specifically comparing 5-mer distributions using Euclidean distance.
The second metric, the tRNA score, was created by comparing codon usage in the
genes with the tRNA availability in the bacterial hosts. The results showed that
both scores capture certain aspects of genetic compatibility and that higher compatibility
correlates with increased likelihood of horizontal gene transfer. Although
some transfers have occurred with poor scores, this suggests that transfers can still
take place despite lower genetic compatibility, for example under evolutionary pressure.
Gene length was identified as an important factor to take into account when
working with 5-mers. Further studies include implementing the 5mer score in machine
learning to determine the spread of ARGs, and refining the tRNA score due
to its limitations, including how the scores were determined
Optimizing Night Driving Simulations: A Comparative Study of Light Simulation Software
No full textNight driving simulations are essential tools for the development and validation of advanced automotive lighting systems. These simulations require not only photometrically accurate representations of light-material interactions but also real-time performance suitable for iterative design and driver-in-the-loop evaluations. This thesis presents a comparative study of three rendering platforms: Ansys AVxcelerate Headlamp, Synopsys LucidDrive, and Unity, to assess their capabilities in simulating nighttime driving environments with high visual fidelity.
The study benchmarks these tools across several dimensions, including rendering performance, photometric accuracy, and perceptual similarity, using standardized test scenes and real-world photometric profiles. A particular focus is placed on evaluating Unitys real-time ray tracing capabilities, enhanced by ReSTIR, against the more static and proprietary pipelines of AVxcelerate and LucidDrive. Experiments utilize standardized test scenes and automotive-grade models under controlled nighttime conditions, profiling performance via structural similarity of rendering results, GPU utilization, frame rate, and memory consumption.
The results underscore the limitations of commercial tools in handling dynamic lighting scenarios and complex BRDFs, while highlighting the flexibility and performance of open rendering frameworks. This work provides a reproducible benchmarking methodology and lays the foundation for future research on hybrid rendering strategies, perceptual validation models, and real-time simulation of intelligent headlight systems