Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Optimization of Deep Neural Networks for Efficient Resource Utilization
No full textDeep neural networks (DNNs) are widely used in computer vision tasks such as image classification and semantic segmentation, but their high computational and memory demands limit deployment on resource-constrained edge devices. This thesis explores quantization as a model compression technique to improve inference efficiency while minimizing accuracy loss. Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT) were applied to MobileNetV2 and ResNet50 for classification on the Mini-ImageNet dataset, and to FCN-ResNet18 for segmentation on the Cityscapes dataset. Additionally, mixed-precision QAT was investigated using first-order gradient-based sensitivity analysis to assign per-layer bit-widths. Maintaining activation precision at or above 6 bits during mixed-precision QAT enabled substantial compression—up to 7.8×—while keeping accuracy degradation under 1%. ResNet50 and MobileNetV2 attained compression ratios of 6.3× and 5.2×, respectively. FCN-ResNet18 preserved 57.3% mIoU with 7.8× compression and under 1% accuracy drop compared to the FP32 baseline. Conversely, reducing activation precision to 4 bits led to notable performance degradation, especially in lightweight models and segmentation tasks. Experiments were conducted on NVIDIA Tesla T4 GPU. The results demonstrate strong potential for deploying quantized DNNs on integer-based hardware such as mobile devices, embedded systems, and FPGAs
Outsourcing Electrode Manufacturing: A Technical and Economic Evaluation
No full textThe demand for lithium-ion batteries (LIBs) has been continuing to grow, and that is the reason why battery manufacturers like NOVO Energy face a strategic decision: whether to continue producing electrodes in-house or to outsource their production. This thesis investigated the decision by exploring both the technical and economic implications of outsourcing electrode manufacturing in the context of LIB production. The study used academic literature and expert interviews to assess four strategic scenarios: fully In-house production, fully Outsourcing, Gradual Vertical Integration, and Parallel Make-and-Buy.
Each scenario was analyzed through the lens of key industry factors like cost efficiency, scalability, quality control, and intellectual property protection. Outsourcing can offer short-term benefits like lower initial investment and possibly faster ramp-up, but it also has drawbacks related to long-term control over processes and maintaining the product quality consistency. On the other hand, keeping production in-house brings tighter integration and long-term competitiveness but needs more upfront resources and causes higher early-stage risk.
The thesis suggested a staged approach that combines Gradual Vertical Integration and Parallel Make-and-Buy based on the findings. By starting with external suppliers for the electrodes and then transitioning to in-house production over time, NOVO Energy can manage risks and costs while aiming towards a controlled and
scalable production mode
Flip-chip Assembly and Reliability Testing
No full textThis thesis investigates the assembly and reliability processes of two different flip chip technologies, C4 solder bumps and gold stud bumps. The thesis mainly focuses on how underfill affects the mechanical and electrical reliability of the flip chip assemblies when going through temperature cycling. The reliability of the assemblies were tested using various methods such as measuring resistance, performing shear tests and cross sections.
The result of the project clearly shows that underfill significantly improves both mechanical and electrical reliability. This was especially clear on the C4 chips were the ones with underfill seemed almost unaffected by the temperature cycling while the chips without underfill deteriorated significantly over time. For the Au stud bumps assemblies, problems with underfill application were observed, mainly because of low standoff height. Even though there were challenges the underfill still improved the mechanical reliability of the Au stud bump assemblies. The thesis concludes that the underfill process is crucial to achieve reliable flip chip assemblies and recommends further improvements of the assembly processes to reduce defects
Utformning och definition av flexibla VVS-system: Möjligheter, utmaningar och branschens perspektiv
No full textI byggbranschen pratas det mycket om att VVS-systemen ska vara flexibla så att dess tekniska
livslängd kan bli längre och att det ska vara anpassningsbart för olika verksamheter eller
hyresgäster med olika behov. Just nu saknas det en definition av vad flexibla VVS-system är.
Syftet med vårt arbete är att försöka ta reda på hur branschen, alltså konsulter och
fastighetsägare, anser vilka faktorer och aspekter som spelar in för att skapa ett flexibelt
system. Målet innefattar också att försöka skapa en gemensam definition av själva begreppet.
För att undersöka detta har studien huvudsakligen baserats på intervjuer med konsulter och
fastighetsägare, kompletterat med en kortare litteraturstudie.
Resultat visar att VVS-branschen delar en gemensam syn på vad begreppet flexibelt system
innebär; att ett VVS-system ”bör kunna anpassas till framtida förändringar med minimala
ingrepp”. Det finns dock olika tolkningar av hur detta bäst ska uppnås och tillämpas i
praktiken. Några av de lösningar som beskrivs som flexibla, såsom placering av don i
fönsterzon, modultänk och centraliserade system, är redan etablerade standarder i branschen.
Samtidigt finns det olika tankar och skolor när det gäller andra lösningar. Exempelvis vid val
av styrsystem, kylsystem, ringmatning, dimensionerad överkapacitet för lokal- och
hyresgästförändringar samt klimatanpassning till högre sommartemperaturer.
Flexibilitet uppfattas ofta som en önskvärd funktion, men nedprioriteras i projekteringsfasen
på grund av faktorer som kostnad, energieffektivitet och beställarens prioriteringar. Studien
identifierar tre huvudsakliga perspektiv på flexibilitet: specifika tekniska lösningar (till
exempel val av system och komponenter), planering (till exempel moduluppbyggnad och
placering av installationer) samt strategiskt förhållningssätt (till exempel att matcha lokalens
utformning med framtida verksamheter). Dessa nivåer används ofta parallellt, men utan att det
tydliggörs vilken typ av flexibilitet som avses, vilket bidrar till begreppsförvirring.
Slutsatsen är att flexibilitet ofta är beroende av projektets specifika förutsättningar samt
beställarens vision och budget, vilket begränsar konsulternas inflytande. För att minska
behovet av ombyggnationer och främja långsiktiga lösningar krävs en ökad medvetenhet om
hur flexibilitet praktiskt ska tillämpas i projekt samt ett långsiktigare tänkande kring framtida
klimatförutsättningar och systemens hållbara funktion
On-chip cache coherency evaluation for space applications: Snoop with AHB bus vs. directory with CHI NoC
No full textWith the end of frequency scaling as a solution for more efficient processors, bigger
multi-core systems are getting increasingly popular, including in the space domain.
To achieve bigger high-performing multi-core systems, more efficient on-chip data
transfer systems are required which in turn are dependent on systems scaling. Increasing
the size of a multi-core system introduces more complexity in upholding coherency.
Different protocols and hardware mechanisms have been used for supporting
coherent systems, such as snooping and directory-based, which can be supported
by different interconnects. Although NoC-based interconnects and directory-based
mechanisms have been shown to successfully scale multi-core systems, past research
lacks a comparison of the coherence overhead introduced in NoC versus bus interconnects.
This thesis project has performed an evaluation and comparison of two protocols
for upholding coherency, AHB and CHI, used by Frontgrade Gaisler in their current
and future state-of-the-art radiation hardened microprocessors. Two evaluation
frameworks have been developed, one for each protocol, with configurable simulation
stimuli targeted for cache coherency. The limited experiments we managed to
perform in this project identified scenarios and topologies where the more expensive
CHI system will not deliver a higher and more scalable performance compared to
the more simple AHB system. This provides a valuable foundation for the next
step, that is performing quantitative analysis of the topologies where CHI delivers
superior performance as the system scales to a larger number of cores. However, due
to technical challenges we leave that part for future work. Improvements such as
increasing parallelism by using non-blocking components and sliced and distributed
last-level-cache, as well as using more advanced NoC routers with reduced latency
should be part of the future work that demonstrates the advantages of CHI over
AHB system. Performance gains in the evaluated CHI system could still be seen
when using a coherency protocol that allows for a write-back policy instead of writethrough
mAuth: Secure Authorization and Authentication Protocol for Native Apps
No full textOAuth and OIDC are well-established for handling user authentication and authorization, which are industry standards today. However, the user experience of native mobile apps remains a challenge due to the use of browser redirection. It is hard for users to determine if they can trust the web page that pops up during the login process. To improve on this, we have designed a protocol called mAuth that performs user authentication and authorization on mobile phones without the use of
browser redirection. This protocol follows the best current practice (BCP) of OAuth and the FAPI standard. The analysis of the protocol showed that mAuth follows the BCPs for OAuth and FAPI through the use of attestation, demonstrating proof of possession (DPoP) with the client instance key and following the basis of the authorization code flow. From the analysis of the user experience of the theoretical protocol, which is based on our point of view, we found that it achieves the goal of better user experience. It also provides flexibility for the developer as they can choose between three different flows depending on their security and user experience demands
Exploring Variations in Frequent Passing Grade Tests in Mathematics
No full textFrequent Passing Grades tests (FPG) have been proven to be beneficial for enhancing students
learning. In this study students at an upper secondary school were exposed to high-stakes and
low-stakes FPG-tests as well as variations in follow-up work and timing for feedback with the
purpose to find out which variations are more beneficial for increasing students’ perceived
learning. Students were then given two surveys to fill out with questions about their
experiences. The responses from the surveys were compiled and analysed using thematic
analysis. Additionally, data was collected about how many students passed the FPG-tests.
The result from the surveys indicates that students perceive they learn more from having high-
stakes FPG-tests. Students express they retain the material for longer when having low-stakes
FPG-tests as well as feeling less stressed. Students with lower grades feel they learn more from
teacher-led assessment reviews whereas students with higher grades were more inclined to
choose self-assessment as the follow-up work, they learn the most from. The relationship
between type of follow-up work students feel they learn from, and their academic performance
was found to be statistically significant. However, regarding the timing of feedback a
statistically significant relationship between when students prefer feedback, and their grades
could not be found. Although, students with higher grades seemed to prefer feedback
immediately while students with lower grades preferred feedback to be delayed, this difference
is likely due to random variation
Positron Emission Tomography Image Reconstruction with Time-of-Flight Data Using a SPLAT System Model
No full textAbstract
This thesis presents the development and evaluation of a physics-based system model tailored for image reconstruction in a novel time-of-flight (TOF) positron emission tomography (PET) brain scanner. The scanner design features a continuous, multilayered monolithic thin detector, offering improved spatial resolution and sensitivity compared to conventional PET systems. To fully leverage these hardware advances, a custom system model called the SPLAT model was implemented. The SPLAT system model integrates key resolution loss mechanisms, including non-collinearity, detector blur, and coincidence timing resolution (CTR), using a Gaussian framework. In addition, positron range and photon attenuation, both implemented in a form viable for 3D, were incorporated to enable more realistic reconstruction. A simulation tool for PET list-mode (LM) data was developed based on this model,
enabling controlled evaluation of reconstruction performance. Two simulation strategies were implemented to discretize the continuous detector domain. The forward and backprojection operators were implemented with performance optimizations in C and integrated into an iterative maximum-likelihood expectation-maximization (ML-EM) reconstruction pipeline. Image quality was quantitatively assessed using normalized RMSE metrics, with analyses explicitly linking reconstruction accuracy to characteristics of the modeled point spread function
Solo but not alone; a co housing solution for single parents
No full textThis master’s thesis has explored how co-housing can be designed to support single
parents by both reducing overall housing costs and contributing to more
community, cooperation, and access to social safety nets. The background is that
single parents often have a weaker economic position, limited social networks, and
greater everyday responsibilities. This affects not only the parents themselves but
can also influence their children's well-being and future opportunities. Traditional
forms of housing are rarely adapted to their life situation, and there is a limited
range of housing options that promote sharing and community without
compromising privacy.
The work has been conducted through a research-based design process where
literature studies, statistics, surveys, reference projects, and study visits formed the
basis of the analysis. The results have been compiled into design strategies focusing
on reducing private living space in favor of shared functions, as well as creating
social areas that encourage spontaneous meetings and cooperation in everyday life.
The final design proposal presents a co-housing model where private apartments
are complemented by shared spaces for cooking, play, and social interaction. By
sharing resources and spaces, both construction costs and environmental impact
can be reduced, while providing residents with access to social support that can
ease everyday burdens. However, the intention was not to create an “isolated”
housing project solely for single parents, as a diverse residential environment is
considered valuable. By including both private and shared spaces for all residents,
the project aimed to strengthen the sense of community within the building while
also reducing housing costs.
The work demonstrates that co-housing can function as a daily social safety net,
provided that the architectural design balances community with the need for
privacy. An important conclusion is that co-housing should not be seen merely as a
niche form of housing but as a relevant alternative in a housing market where many
life situations currently lack suitable solutions. The study also emphasizes the
importance of creating housing with mixed target groups, where single parents,
older people, and other households can support each other in a shared
environment
Sveriges befolkning: På randen till utdöende eller evig tillväxt?
No full textNyligen kom ett pressmeddelande från Statiska Centralbyrån (SCB) att Sverige har den lägs ta siffran på antal barn per kvinna i modern tid, 1.43. Därför är det relevant att spekulera
kring hur Sveriges framtida befolkningsmängd kan se ut, både på kort och lång sikt. I detta
arbete har begrepp undersökts som populationsstorlek, fruktsamhetstal, försörjningskvot och
befolkningspyramid, och hur dessa ter sig över tid.
Fyra scenarier analyserades: oförändrad fruktsamhet, fortsatt nedgång, övergång till ökning
samt en cyklisk utveckling med både minskningar och ökningar. Modeller har byggts upp på
grunder av matematiska koncept som Galton–Watson–processer och Leslie–matriser, som be skrivs med begrepp som sannolikhetsgenererande funktioner och LF–fördelningen (linear frac tional offspring distribution). Därifrån har populationer simulerats i programmeringsverktyget
Python för att illustrera resultaten och till slut jämföra två huvudmodeller, Galton–Watson–
modellen och Leslie–modellen.
Resultatet är att i båda modellerna överlever befolkningen fallet med en strikt ökande frukt samhet. För båda modellerna gäller att i resterande scenarion dör befolkningen efter lång tid
ut. Framtiden är dock mer slumpmässig än vad någon av modellerna speglar, de olika fallen
bör mer ses som scenarion värda att tänka på