Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Conservative Compiler Optimisations
No full textModern functional programming is built upon heavy abstraction that compilers are not always able to optimise away. A two-stage compiler has been proposed to solve this, but lacking any optimisation this remains clunky to work in. We specify a
number of optimisations in a way that they can be guaranteed to always be applied so that code generated by the first stage can be made simpler without causing any unnecessary overhead
The Mechanical Behavior of the Glued-in Plate Connection: An experimental study on bonded timber connections
No full textGlued-in plates is a novel connection for timber construction and limited research
has been done on the subject. The connection consists of a timber element with
a slot, in which a steel plate is mounted and set with adhesive. Previous research
shows that this connection has many advantages compared to traditional timber
fastening methods. Similarities can be found with the glued-in rod connection,
where both connections are made with the same three materials and have the same
basic mechanical behavior.
In this project, an experimental study was conducted where different parameters
were investigated to gain further insights in the load carrying capacity of the glued in plates connection.
Geometrical parameters were changed depending on the test series, and one of the main focuses was to investigate the properties of the glue dowel
formed in the perforation of the steel plate.
Results from the experimental testing shows that glue dowels increase the load
capacity of the connection and contributes a less brittle failure compared to unper forated plates.
The shear area of the glue dowel has a linear increasing correlation to
the load capacity. Both plate thickness and embedment length have an influence on
load capacity up to a certain limit and then plateaus. Distributing the perforation
of the plate, gives a higher load capacity due to a better stress distribution in the
adhesive. Although, the perforation should not be placed closer than 2.5 times the
dowel diameter to the edge of the plate to ensure that there is no failure in thetimber.
The connection is proven to exhibit high stiffness and strength. However, there
are many areas that need studying before the connection can be used in real life applications.
One of the areas is how the application of adhesive should be con ducted for reliable results and practical use, while also studying the occurrence of
air bubbles. Furthermore, a design standard with demands, requirements and calcculation method should be developed as to consider all aspects and variables of the
connection design
Assessing annoyance in automotive seat adjustments; perception and prediction modeling of subjective annoyance responses using advanced approaches; a comparison of regression methods and neural networks
No full textIn the automotive industry, acoustic comfort is a key aspect of perceived quality,
especially in high-end vehicles where even subtle noise such as squeaks, buzzes, or
rattles can negatively impact user satisfaction. Sound Quality (SQ) assessments help
ensure a premium experience but typically rely on subjective jury testing, which is
time-consuming and depends on expert judgment that may not reflect everyday user
expectations.
This thesis aims to develop a predictive model for annoyance ratings based on objective
acoustic parameters. This is a challenging task, as annoyance is inherently
subjective and influenced by various perceptual factors. While traditional methods
such as linear regression can estimate simple perceptual attributes like loudness,
they fall short when modeling more complex, non-linear characteristics. To overcome
these limitations, machine learning approaches, including neural networks and
random forests, are investigated and compared to linear and polynomial regression
models.
The study focuses on seat adjustment mechanisms as a use case. These sounds
are relatively easy to isolate and analyze, show noticeable variation across vehicle
brands, and are less affected by external noise sources. This makes them a suitable
candidate for controlled testing and model development. Using those sounds,
a large scale listening test is made to assess annoyance and be able to train the
models. Results demonstrate that machine learning models can successfully predict
perceived annoyance based on objective metrics, offering a promising alternative to
traditional jury testing. Such models could significantly improve the efficiency and
scalability of SQ evaluations in the automotive industry
Enhancing Proof Development in Liquid- Haskell: Implementation and Evaluation of Typed Holes
No full textRefinement type systems provide mechanisms for specifying and verifying programs beyond what mainstream type systems can express. LiquidHaskell extends Haskell with refinement types, and beyond that, has evolved to be used as a theorem prover, akin to Agda or Idris. Unfortunately, it lacks a fundamental feature present in most proof assistants: the ability to inspect goals during proof development, for example, with typed holes. In this thesis, I implement typed hole support for LiquidHaskell addressing this limitation and taking an initial step towards more interactivity in LiquidHaskell
Multivariate anomaly detection with LSTM layered Variational Autoencoder
No full textThe aim of this thesis was to develop and evaluate the effectiveness of a recurrent neural
network layered autoencoder model for detecting anomalies in multivariate time-series
data, with a focus on improving the accuracy and reliability of diagnostic data for Volvo
Penta’s boats. The primary goal was to leverage the relationships and correlations between
signals to identify deviations that traditional models may fail to detect. The
model’s performance was assessed in terms of its ability to learn the structure of normal
data, detect synthetic anomalies, and provide meaningful insights without relying on labeled
datasets.
The study highlights the limitations of traditional evaluation metrics, which are often unsuitable
for unsupervised learning approaches like the model used. Instead, the model’s
effectiveness was demonstrated through reconstruction error analysis and its ability to
handle the complexities of multivariate time-series data. Challenges such as data dimensionality,
sequence length optimization, and noise handling were addressed to enhance
the model’s robustness. The findings suggest that while the model excels at identifying
synthetic anomalies and capturing temporal relationships, further work is needed
to generalize its capabilities to real-world scenarios. This research lays the groundwork
for improving diagnostic processes and supports the development of more adaptive and
reliable anomaly detection systems
Is This Data Point In your Training Set? Similarity-based Inference Attacks: Performance Evaluation of Range Membership Attacks to Audit Privacy Risks in Machine Learning Models
No full textMembership Inference Attacks (MIAs) pose a serious threat to the privacy of machine learning (ML) models by determining whether a specific data point was used during model training. A recent and powerful variant, the Range Membership Inference Attack (RaMIA), assesses privacy risks over a range of semantically similar data points. However, its practical application is limited by a high query overhead, as it requires querying the target model for every sample in the range. This thesis proposes and evaluates a novel approach designed to overcome this limitation by combining range queries with group testing principles to reduce the number of queries sent to the target model without losing the attack performance and making the attack more stealthy. Instead of testing every sample, this method first groups similar data points based on their extracted features and then queries only a small number of strategically chosen representatives.
All the experiments are conducted on the CIFAR-10 dataset, comparing its performance against the standard RaMIA baseline. The results demonstrate that RaMIA with group testing successfully reduces the number of queries by 84% in a setting of 50 augmentations. This work reveals that even minor enhancements in query design and decoding strategy can lead to substantial gains in auditing. Moreover, we provide practical recommendations for tuning key hyperparameters and integrate our attack into the LeakPro framework for reproducibility and broader adoption in privacy auditing of ML models
Measurements of different Silicon Car bide Transistors and their Varying Con duction Losses when used in Three-Phase Inverters
No full textAbstract
In this bachelor thesis, the spread of RDS(on) is measured and quantified for five SiC MOSFET samples from each of three different manufacturers. The resulting spread in on-state resistance was then applied theoretically to a three-phase inverter model in order to calculate the corresponding conduction loss variation. The conduction loss was determined using an algebraic equation based on the MOSFET parameters, specifically the on-state resistance and drain-source current. Finally, the measured maximum and minimum conduction losses were compared to typical values from each manufacturer’s datasheet, with the results expressed as relative percentage deviations, scaled to represent the total conduction loss across all six MOSFETs in a three-phase inverter. The results show that the spread of RDS(on) was significant within each manufacturer, with standard deviations ranging from 3.1 mΩ for Onsemi to 6.4 mΩ for Infineon. This spread was calculated for two gate voltages. When the measured RDS(on) values were used to calculate conduction losses, the variation resulted in substantial
differences in total inverter loss. For Infineon, the conduction loss ranged from –34.8% to +42.4% relative to the typical datasheet-based value. STM exhibited the largest deviation, reaching up to +52.2%. In contrast, Onsemi showed a more modest variation, with losses between –7.7% and +26.5%. These findings highlight the importance of accounting for RDS(on) spread when estimating conduction losses in practical three-phase inverter designs
Distance to Singularity for Skew-Symmetric Matrix Pencils
No full textThe singularity of a matrix or matrix pencil is easily affected by small errors in
numerical calculations. Therefore, it is motivated to consider the distance to singularity,
rather than whether or not the matrix (pencil) is singular. In the case
of unstructured matrices, the distance to singularity is well known. For structured
matrices and matrix pencils, however, the question is more complex. In this thesis,
a numerical method for determining the distance to the nearest skew-symmetric
matrix pencil of given maximal rank is presented. The task is formulated as a
minimization problem using the skew-symmetric Kronecker Canonical form and a
rank-1 decomposition of skew-symmetric matrix pencils. Four different algorithms
for solving the minimization problem are proposed, using the so-called vec-trick, QRdecomposition,
singular value decomposition, and the GUPTRI form [5, 6]. These
algorithms perform well compared to state of the art methods
DEVELOPMENT OF HIKING FOOTWEAR FOR WOMEN
No full textThis thesis was conducted in collaboration with the Italian brand Dolomite and focuses on the development of a hiking shoe specifically tailored for young Scandinavian women. The project adopts a user-centered design approach, with both functional and aesthetic needs at its core. The objective was to create user requirements and a concept for a hiking boot designed to appeal to the female target group. The project began with extensive research, including interviews, surveys and observational studies with women from the intended user group. Insights gained from this research were used to develop a representative persona, which served as the foundation for idea generation and concept development. A series of concept proposals were formulated based on the collected data and systematically evaluated using a Pugh matrix. User interviews and group discussions provided additional qualitative insights into preferences regarding fit, lightness and visual style.
The final concept is a hiking shoe that combines functionality, comfort and Scandinavian minimalism. Its design is carefully adapted to the anatomical needs of the target group and their aesthetic preferences, strengthening Dolomite’s product portfolio towards a growing market segment
Web Design Styles and Design Recommendations: Balancing Aesthetics and Usability
No full textThis project explores different characteristics of classic and modern web design, and how it affects user experience. The web design recommendations presented in this project are intended to support designers and businesses in merging classic and modern design styles. Usability testing, expert interviews, and thematic analysis was performed, and three websites were developed: one with a classic design, one with a modern style, and a third hybrid version that integrates elements from both approaches. The project was limited to front-end development for desktop use, targeting Swedish adults aged 19–30. Comprehensive accessibility measures were excluded, and SEO testing focused on differences between styles, rather than optimization. The results show that the classic design was valued for its familiarity, clarity, and efficiency, while the modern design stood out for its visual appeal, engagement, and immersive qualities. The hybrid solution, guided by the developed web design recommendations, effectively integrated the strengths of both styles, resulting in a balanced and user-friendly interface. These findings suggest that blending classic and modern design principles can enhance user experience without compromising aesthetics, offering practical guidance for those seeking to create accessible and engaging web interfaces