Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Value of Information Analysis for Improved Decisions on Infiltration and Inflow to Wastewater Systems
No full textInfiltration and inflow (I/I) of excess water into wastewater systems pose significant
technical, economic, and environmental challenges. This Master’s thesis explores the
application of value of information (VoI) analysis as a decision-support tool to improve
I/I-water management in private and separate sewer systems in Gothenburg. By com bining cost-benefit analysis with probabilistic modeling and Bayesian updating, a VoI
framework is developed to quantify the economic value of acquiring additional infor mation. This is achieved by comparing optimal decisions before and after considering
potential new data, to determine whether methods such as smoke and dye testing are
economically justified. Using data from the Department of Sustainable Waste and Wa ter at the City of Gothenburg (Kretslopp och vatten), the model assesses the viability of
three decision alternatives: intervention, no intervention, or further investigation. The
results show that interventions are recommended in 67% of analyzed subareas, while
further investigation is justified in 20%. Sensitivity analysis highlights the influence
of several parameters, such as costs for measures and film inspections, detection accu racy, and especially property owner compliance, which plays a decisive role in cost effectiveness. While the model is applied to Gothenburg, its structure is generalizable
and adaptable to other municipalities facing similar challenges. The thesis demonstrates
that VoI analysis can enhance municipal decision-making by supporting more targeted,
risk-aware, and cost-effective I/I-water management, ultimately contributing to more
resilient and sustainable wastewater infrastructure systems
Exploration and Optimization of Radiance Cascades for Real-Time Applications
No full textThis thesis introduces the fundamentals of Radiance Cascades in screen space for 2D, with more focus on the implementation of a less explored approach called Screen- Space Probes with World-Space Intervals (SPWI) to solve single shot diffuse global illumination. A conventional 3D probe grid faces significant memory constraints; however, SPWI gives a solution to capture off-screen indirect light more memory efficiently. The implementation diverges from Alexander Sannikov’s original approach in a few aspects. It employs ray-tracing instead of ray marching, which is just personal preference and it also incorporates Sannikov’s post-publication novel depth aware up-scaling method "Bilinear 3D", which he released after the release of the original Radiance Cascades paper. This work presents comparative analysis between different upscaling techniques and documents experimental approaches that, while ultimately not incorporated in the final implementation, offer valuable insights for future research. The findings contribute to the foundational understanding of this approach and establish a framework for further improvements in Radiance Cascade with SPWI for real-time rendering
Antibacterial activity of photothermal gold nanorods combined with antimicrobial peptides
No full textBiomaterials, used in medical implants for example, are important tools in modern medicine to ensure health and quality of life. Despite their benefits, biomaterials currently face two major problems, the difficulty of treating persistent biomaterialassociated infections and the rapid increase in bacterial antimicrobial resistance. It can be concluded that there is an urgent need for development of biomaterials able to prevent biomaterial-associated infections without relying on conventional antibiotics. Promising solutions are the implementation of antibacterial surface modifications of biomaterials and the use of antimicrobial peptides.
The focus of this thesis was to perform in vitro evaluation of a material modification combining the antibacterial activity of surfaces functionalised with photothermal gold nanorods (AuNRs) in the presence of antimicrobial peptides (AMPs). The
evaluation was performed by antimicrobial susceptibility testing, agar plate models and fluorescence microscopy. Significant antibacterial activity could be observed when irradiating AuNRs with near-infrared light in the presence of AMPs when
tested against Staphylococcus aureus. A slight synergistic relationship between the AuNRs and AMPs could be observed, but further testing is needed to confirm the effect. This thesis provides important insight into the antibacterial activity of surfaces functionalised with photothermal AuNRs when in the presence of AMPs
Identifying Key Factors for a Successful Trainee Program
No full textOne common way to introduce newly graduated students into workplaces is through
trainee programs. Although there is a wide variety of trainee programs, there is a
lack of research on the factors influencing their effectiveness. The aim of this thesis
has been to identify key factors for a successful trainee program from the trainees
perspective. The aim has been evaluated and in relations to the trainees’ learning
experience, motivation and how they experienced the transition from trainee pro-
gram to regular employment. The result is anchored in the perception of former
trainees, and to ensure a broad perspective, 13 former trainees from 6 different large
organizations in Sweden were interviewed. The interviews were semi-structured and
analyzed using a thematic analysis. The analysis yielded the following insights: the
former trainees valued being given space to explore and learn within the organiza-
tions, in addition to having a supportive structure provided by the trainee programs.
The former trainees believed that the trainee programs facilitated skill development
and helped them establish a network that gave them an advantage in their work
after the programs. Furthermore, they appreciated the strong cohesion with the
other trainees at their organization. Along with how the trainee programs accom-
modated for the trainees’ needs, a successful trainee program was also linked to the
trainees’ perceived ability to contribute to the organization. Thus, this thesis high-
lights essential factors for a positive trainee experience and offers useful guidance
for organizations developing their own trainee programs
Methods for Optimizing BERT Model on Edge Devices - Accelerating Biomedical NLP with Pruned and Quantized BERT Models
No full textNamed-entity recognition (NER) of clinical efficacy endpoints in oncology abstracts
supports downstream discovery pipelines at AstraZeneca. Yet, the fine-tuned transformer
models currently used are too slow and over parameterized for large-scale
CPU deployment. This thesis evaluated whether post-training model compression
techniques can accelerate inference without retraining or harming extraction quality.
In the first stage of this project, standard BERT and BioBERT were individually
pruned with a three-stage, Fisher-guided structured pruning workflow at three levels
of sparsity. Subsequently, in the second stage, dynamic 8-bit integers quantization
using ONNX Runtime was applied to standard BERT, BioBERT, and DistilBERT.
The third stage involved combining both pruning and quantization, further optimizing
the pre-trained standard BERT and BioBERT transformers. Experiments were
run on annotated MEDLINE sentences covering 25 efficacy labels, with F1 score
and inference latency per sample serving as primary metrics.
A 25% structured-sparsity level yielded no measurable drop in F1 score, and the additional
8-bit integers dynamic step cut latency further. The best configuration, 25%-
pruned+8-bit integers BioBERT, reduced mean CPU inference time from 32.52 ms
to 12.02 ms (2.6-fold speed-up) while accuracy fell only from 0.982 to 0.980 and F1
score from 0.954 to 0.948.
The Post-training structured pruning combined with 8-bit integers dynamic quantization
makes the oncology-NER pipeline about three times faster in inference time
on standard CPUs without compromising the extraction quality or needing special
hardware or libraries
Capability determination of in-process monitoring for laser powder bed fusion
No full textThis thesis investigates the capability of optical tomography (OT) as an in-situ process
monitoring tool for detecting spatter-induced defects in components produced
via Laser Powder Bed Fusion (LPBF). The study evaluates whether OT imaging,
combined with custom image processing techniques, can reliably identify internal
porosities typically verified through high-resolution X-ray Computed Tomography
(XCT).
Two Inconel 718 demonstrator parts were manufactured using an EOS M290 LPBF
machine, with process parameters deliberately adjusted to promote defect formation.
OT images captured during the build were processed using a custom MATLAB script
that applied Laplacian of Gaussian (LoG) and erosion filtering to identify thermal
anomalies. These anomalies were compiled into point cloud datasets and digitally
compared with XCT data of the segmented demonstrator.
The results revealed limited correlation between OT-based indications and actual
porosities detected through XCT, with alignment errors, image noise, and the resolution
of OT data identified as key limitations. Despite this, the study highlights
spatial patterns in OT data that may still be indicative of defect-prone regions,
suggesting potential for further development. The findings underscore the need for
enhanced filtering techniques, improved data alignment, and the integration of additional
OT data types to improve detection accuracy.
This work contributes to the ongoing development of in-situ monitoring systems
for LPBF, particularly in critical applications like aerospace, where early defect
detection is vital for part certification and safety assurance
Analysis and Generation of Wikidata Descriptions Focusing on Bangla Language
No full textWe present a Grammatical Framework (GF)-based resource grammar for Bangla designed to automatically generate structured natural language descriptions for Wikidata entities. The system covers multiple entity types including cities, universities, islands, lakes, and humans. Unlike statistical or black-box models, our approach uses a rule-based grammar that guarantees grammatical correctness and structural consistency. Evaluations on more than 76,000 entities demonstrate high coverage (over 99%) and strong alignment with source descriptions, as shown by multilingual embedding similarity. Our results show that the generated Bangla descriptions not only complement existing entries but often exceed them in semantic consistency. This work offers a practical solution for enhancing low-resource language content in multilingual knowledge bases
New Hasselblad center
No full textThis thesis has investigated how Hasselblad’s historical heritage
and distinctive design language can be translated into an
architectural expression for a new Hasselblad Center in Gothenburg.
Rooted in the city where Hasselblad cameras were first designed and
manufactured, the project responds to the Hasselblad Foundation’s
need for dedicated exhibition galleries, research facilities, public
amenities, and a venue for its prestigious awards ceremony. The
ambition has been to honour Hasselblad’s cultural significance,
clarify the Foundation’s dual commitment to photography and
the natural sciences, and enrich Gothenburg with a landmark
institution.
Iterative cycles of traditional drawings, digital and physical
models, and visualisations have been used in the study with a
research-by-design methodology. Charles Felix Lindberg’s Plats was
identified by site analysis as an underutilised urban node that
is close to both public green space and major streets. Volume
studies and early explorations abstracted key camera components—
the circular lens, the cuboid body, and the modular film back—
into elemental forms. Interviews with Foundation stakeholders and
scientists, excursions to pertinent projects, and literature studies
on embodied experience informed programmatic requirements and
spatial strategies.
The resulting design features a circular courtyard—invoking the
camera lens—as an accessible “oasis” that mediates between Avenyn’s
grandeur and the intimate interior. Inspired by the modularity of
the 500C, the underground galleries with a flexible exhibition wall
system enable reconfiguration and colour change without wasting
material. A research library, a photographic archive, offices, an
auditorium, a restaurant, and a café are unified within a coherent
composition of spaces evoking an abstracted 500C camera. Researcher
workspaces flank public circulation, visually expressing the
Foundation’s scientific mission.
This thesis has demonstrated how heritage can be abstracted into
architectural form to create a facility that is both expressive and
operationally sound. By integrating theoretical insights on sensory
experience and stakeholder needs, the new Hasselblad Center emerges
as a project in aligning institutional identity with spatial
experience—offering a study for future museum projects seeking to
embody their unique legacies
Bridging the Resolution Gap: Conditional Generative Model for High- Fidelity Time-Series in Automotive Applications
No full textThe automotive sector increasingly relies on high-resolution vehicle data for advanced analytics, yet readily available customer vehicle data is often sparse and lowresolution due to hardware, cost, and privacy constraints. Existing time-series generative models frequently struggle with long sequences (exceeding 1000 timesteps) and lack robust mechanisms for controlled generation using continuous conditional inputs. This thesis aims to bridge this data resolution gap by developing and evaluating advanced machine learning models capable of generating high-fidelity, longsequence time-series data from low-resolution conditional inputs. We evaluated current state-of-the-art time-series generative models for their efficacy
in handling long sequences, identifying Variational Autoencoders (VAEs), particularly TimeVAE, as the most promising base due to their fast training time and superior predictive performance. A novel conditional generative model was then developed by integrating TimeVAE with a Transformer encoder, adapting the PatchTST architecture to encode the conditional information. Key architectural enhancements, including Rotary Positional Embeddings, Layer Normalization, and specifically, a residual integration approach for incorporating conditional information, were explored. Further improvements like "free bits" and "conditional prior" were implemented to mitigate posterior collapse and enhance overall model performance. Our findings indicate that state-of-the-art models generally underperform on long sequences , and naive conditional integration is ineffective due to complex gradient flows. However, the proposed residual integration improved the model’s ability to leverage conditional information. The combined application of "free bits" and "conditional prior" alleviated posterior collapse, leading to a more robust model. Crucially, the purely data-driven model was able to generate physically plausible long-sequence time-series data, with three key physical metrics showing less than 16% deviation from real signals, without explicit physics-based training. This demonstrates a viable solution for controlled high-fidelity time-series generation in automotive application