Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Synthesis and Characterization of Polyglycerol-based Surfactants
No full textPolyglycerol-based surfactants have gained recognition as a renewable and safe alternative to polyethylene glycol-based surfactants, which are derived from fossil resources. Polyglycerol esters have a complex and polydisperse structure upon
synthesis, which makes it difficult to pinpoint which characteristic of the compound is responsible for a certain property. It is therefore desirable to obtain monoesters as this would facilitate a more accurate comparison of the performance
of different surfactants. To promote monoester formation, optimization of the synthesis was performed by fine-tuning the reaction conditions. The synthesized polyglycerol esters were characterized using various techniques. The results indicated that a higher molar ratio of polyglycerol to fatty acid promoted monoester formation. However, it is yet to be confirmed whether one-phase synthesis (in solvent) results in a higher monoester content than two-phase synthesis (solvent-free). The surfactants were also tested for their cleaning efficacy. A particular formulation of PG6C10 ester performed comparably to that of Berol OX 91-6 (a polyethylene glycol-based surfactant), indicating that polyglycerol esters are a promising green alternative to fossil-based surfactants
Designing for Digital Inclusion - Supporting Individuals with Cognitive Disability Through Digital Planning Tools
No full textThis thesis explores how a digital interface can improve support for individuals with cognitive disabilities in their participation at Daily Activity Centres (DACs), and how such support can promote equality, quality
of life, and societal participation. DACs are part of Sweden’s LSS services and provide meaningful occupation for individuals unable to engage in paid employment. Despite these support structures, people
with cognitive disabilities often face digital exclusion and limited participation in society.
The project was carried out in collaboration with Välfärdsteknik Sverige, the developer of Boet, a digital planning tool originally designed for group homes and now being explored for use in DACs. To understand
the specific context, challenges, and needs within DACs, observations and interviews were conducted with both staff and participants. These insights were synthesised into design goals and guidelines for creating
a digital interface tailored to the DAC environment. Building on these findings, an iterative design process was carried out, resulting in a conceptual design for a web-based planning interface for staff. The design
was evaluated in collaboration with DAC personnel to validate its relevance and usability.
The results demonstrate how a digital planning tool, when grounded in user needs and used responsibly, can improve structured planning, individualised support, and digital inclusion at DACs. Ultimately, such
tools have the potential to enhance autonomy and well-being as well as contribute to a more equal and inclusive society for individuals with cognitive disabilities
NextGen AI agent for perceived quality of cars
No full textThis thesis details an AI-powered tool intended to assist designers in evaluating Percived
Quality in the early stages of the design of the development of vechicles. The
tool integrates machine learning models, image analysis, and a Retrieval-Augmented
Generation system to provide real-time, context-aware insights based on customer
data and design attributes. Enabling proactive decision-making aims to reduce
costly design iterations and align products more closely with user expectations
Designing for Perceived Safety in In-Vehicle Touch Interfaces - A Mixed-Methods Study
No full textAs digital in-vehicle interfaces become increasingly central to the driving experience, ensuring that they are perceived as safe is essential. This thesis explores how
perceived safety can be influenced through the design of human-machine interfaces (HMIs), with a particular focus on touch-based interaction in the centre stack display
(CSD) of modern cars. A mixed-methods approach was applied, starting with a literature review and expert interviews to define perceived safety, identify appropriate evaluation tools, and
pinpoint key interface design factors. These insights informed the development of four interface concepts, varying in navigation structure (flat and nested) and button size.
Nineteen participants took part in a user test, evaluating the concepts in a real vehicle on a closed test track. Both objective data (glance behaviour via eye-tracking) and subjective input
(questionnaires on perceived safety, usability, and workload, as well as interviews) were collected. The results formed a set of design and evaluation guidelines grounded in theory and
empirical findings. Use flat layouts for quick tasks, particularly when stationary. Consider nested structures when the goal is to reduce the duration of individual glances. Increase button sizes
to around 35×35 mm to support confidence and perceived safety. From an evaluation perspective, use a combination of methods to build a more complete understanding of perceived safety.
Interpret quantitative and subjective data alongside qualitative user narratives. While standardised tools such as UMUX-Lite and NASA-TLX offer helpful benchmarks, complement them with task-specific
prompts and interviews to capture richer feedback. Furthermore, when comparing similar concepts, use comparative ranking over scalar ratings to reveal clearer preferences. Lastly, test under realistic
conditions to ensure findings are grounded in real-world use
Signal Processing Techniques for Step Counting, Activity Classification, and Distance Measurement Using a Single IMU
No full textThe aim of this thesis was to develop and validate an offline method for monitoring
human walking and running using a single, low-cost Inertial Measurement Unit
(IMU). We designed signal-processing algorithms to count steps, estimate distance
and speed, and classify activity level. All from the 3-axis accelerometer and gyroscope
data. Raw signals were filtered with low-pass Finite Impulse Response (FIR)
filter, zero-velocity updates were applied at each detected gait event and dominant
stride frequencies were extracted via Fast Fourier Transform (FFT) over sliding windows.
In test with five subject on a 75 m straight path, step-count accuracy averaged
98% for walking and 94% for running. The distance estimates reached 96% accuracy
in walking and 91% in running. Activity classification achieved 100% accuracy in
controlled trials and 89% in mixed scenarios
Coupled Design Optimization of Compact Heat Exchangers in Aviation. A design study using a generalized heat exchanger model, computational fluid dynamics and Bayesian optimization
No full textThe transition to hydrogen-fuelled aviation presents significant challenges in thermal management, particularly in the integration of high-performance, compact heat exchangers. As the European Union aims to achieve net-zero emissions by 2050, hydrogen-powered gas turbine engines are a promising solution. Future civil and military engines will face increased thermal loads, necessitating the integration of megawatt-class heat exchangers while maintaining aerodynamic efficiency.
This master’s thesis addresses this problem through the optimization of duct geometries with integrated finned heat exchangers. The primary focus of the thesis is to investigate how performance varies with heat exchanger inlet area and total duct length. To enable an objective comparison of losses across different designs, the internal heat exchanger geometry was updated between each CFD iteration to converge the solution to a specified performance target.
The results of this study provide new insights into the underlying trade-offs. Heat exchangers with a larger area result in lower losses over the heat exchanger matrix but incur increased losses in the ducts. Additionally, shorter ducts lead to higher losses over the heat exchanger due to the reduced diffusive capacity, while duct losses remain largely unchanged. Another notable trend in the optimized designs is the presence of significant recirculation regions in all duct geometries, highlighting the strong diffusive capacity of the heat exchanger.
The study also includes a limited investigation into the effects of the transversal fins of the heat exchanger by selecting a fixed geometry from the previous optimization trials and removing the sink term associated with the fins. This modification increases the normalized losses from 1.35 to 1.37, indicating that the pressure drop across the heat exchanger matrix is the primary driver of diffusion, rather than the finned structure itself
Following the Tail: A Comparative Analysis of Job Crafting in Academic Setting
No full textThis thesis explores how job crafting practices differ between teaching faculty and PhD
students at Chalmers University of Technology, with a focus on perceived autonomy, flexibility,
and the relational and structural factors that influence proactive role shaping. While both groups
engage in job crafting, they do so under different conditions. Our findings show that faculty
members, who typically enjoy higher autonomy are more likely to engage in proactive job
crafting by aligning their tasks with personal goals and institutional expectations. In contrast,
PhD students experience more limited opportunities as their roles are in many cases shaped
by supervisory relationships and organizational structure.
We introduced the concept of ‘’follow-the tail job crafting’’ to describe how PhD students often
adjust within predefined boundaries, following the lead of their supervisor rather than fully
reshaping their roles. Their ability to craft is often dependent on how much freedom and support
their supervisors provide. Supportive supervision was found to be a key enabler of job crafting,
while rigid leadership tends to reinforce disengagement and role passivity.
By applying Wrzesniewski and Dutton’s (2001) job crafting framework and integrating theories
such as self-determination theory, this thesis provides both theoretical and practical
contributions. One of the most important insights is the central role the supervisor and PhD
students' relationship has in enabling or limiting proactive job crafting. We argue that academic
institutions should raise awareness among supervisors about the potential of job crafting and
create structures that encourage autonomy, reflection, and co-creation. This could support
greater engagement, motivation, and well-being among PhD students.
Although this study was limited to one department at Chalmers University, the findings offer
valuable implications for other academic environments seeking to create a more flexible and
supportive environment for PhD students, helping them grow into confident, proactive and
independent researchers. However, we also see the potential to use the findings in other
contexts with apprenticeship-like relationships
Sub-networks and Spectral Anisotropy in Deep Neural Networks
No full textDeep neural networks (DNNs) have achieved remarkable success across diverse domains,
yet the fundamental reasons behind their efficacy and ability to generalize
remain elusive. This thesis examines how over-parameterized DNNs learn and generalize
by investigating two interconnected phenomena: the emergence of sparse,
critical sub-networks (aligned with the Lottery Ticket Hypothesis) and the structural
symmetry-breaking. Additionally, we explore the geometric structure of the
parameter space, with a particular focus on the anisotropy of the Fisher Information
Matrix (FIM) spectrum.
We demonstrate that different layers in a deep network exhibit varying degrees of
symmetry breaking, which we link to the presence of sub-networks that encapsulate
the model’s core representational capacity. Using two distinct criteria—magnitudebased
and change-based—we identify critical sub-networks and show that, despite
the over-parameterization of DNNs, these sparse sub-networks play a central role in
achieving high performance.
By analyzing the spectrum of the FIM, we reveal that DNNs evolve along a limited
number of dominant eigendirections, spanning a subspace where training dynamics
converge. This finding highlights an intrinsic anisotropy in the parameter manifold.
Furthermore, we investigate how this anisotropy correlates with the emergence of
sub-networks and the internal structure of the subspace.
Overall, this thesis provides a novel perspective on the roles of implicit regularization,
loss landscape geometry, and sparse substructures in modern deep neural
networks, offering insights into the geometric nature of DNNs