Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
Privacy Risks in Text Masking Models for Anonymization
No full textLarge Language Models (LLMs) are increasingly employed to anonymize texts containing
Personal Identifiable Information (PII), often relying on Named Entity Recognition
(NER) to identify and remove sensitive data. This thesis explores the privacy
risks associated with such text masking models by evaluating their vulnerability to
Membership Inference Attacks (MIAs) and extraction attacks. MIAs are attempting
to identify whether or not a data point was part of the training dataset, knowledge
of the membership can in certain scenarios be a breach of privacy. Two state-of-theart
MIAs have been used to conduct attacks on text masking models. This study
also proposes a framework based on multi-armed bandits for performing extraction
attacks and evaluates two different strategies within this framework. The results
from the MIAs indicate that there is some risk of revealing information regarding
the training data. The extraction attacks did not yield great results in terms of
performance but indicate that the concept could possibly be useful if developed
further
Oscillation Detection on High-Resolution Time Series Data
No full textField test vehicles at Volvo Group log between 500 and 1000 signals at resolutions up to 100 Hz, some of which originate from actuators. Oscillations in actuator signals can cause system instabilities and lead to components being worn out prematurely. This thesis investigates the suitability of unsupervised anomaly detection techniques for identifying such oscillations by framing them as a specific type of anomaly. Par ticularly, the focus is on evaluating the performance of a One-Class Support Vector Machine (OC-SVM) and a transformer-based model (TranAD). The available data is unlabeled and consists of high-frequency time series data collected from two main sources: field test vehicles and test cells. To complement this, a number of man ual data recordings were provided by domain experts at Volvo Group, containing examples of oscillations. These recordings, combined with synthetically generated oscillations, were used to create a labeled test set. OC-SVM and TranAD were trained on both field test data and test cell data, with the best OC-SVM model be ing trained on field test data and the most effective TranAD model being trained on test cell data. Although both models are able to detect oscillations, they also cap ture other types of anomalies and sometimes misclassify normal data as anomalous. Overall, TranAD demonstrates the most promising result in detecting oscillatory behaviour. Since both OC-SVM and TranAD were able to detect oscillations, but also other types of anomalies, a valuable extension to this work would therefore be some sort of clustering as a postprocessing step. Despite some limitations, the mod els successfully identified oscillatory patterns that had not previously been detected at Volvo Group
Challenges and Opportunities with Continuous Electrode Slurry Mixing in Lithium-Ion Battery Manufacturing
No full textThe growing demand for lithium-ion batteries has intensified the need for efficient and sustainable manufacturing
methods. This literature review investigates the feasibility and potential advantages of continuous
slurry mixing compared to the traditional batch mixer for production of lithium-ion batteries.
Key performance parameters such as slurry quality, operational efficiency, and production cost are evaluated
to compare the two mixing systems. By drawing from scientific literature as well as insights from
researchers and industry experts, this study finds that continuous mixing offers clear advantages over
batch mixing, particularly in terms of slurry quality and operational efficiency. Furthermore, the study
explores why the batch mixer (planetary mixer) remains the industry standard and goes on to assess if
the continuous mixer is better suited for potential changes to slurry composition. The findings suggest
that continuous mixing is indeed better equipped to handle variations in slurry composition. However,
its limited adoption at industrial scale is primarily due to lack of large-scale studies replicating laboratory
and pilot-scale experiments, resulting in insufficient data on the industry-scale performance of the
twin-screw extruder
Quality assurance in light incontinence products
No full textThe master’s thesis investigates the procedure of specifying and monitoring the key
materials Super Absorbent Polymers (SAP) and pulp in light incontinence products at
Essity. The project focuses on four phases within the value chain from product development,
design verification, process validation to running production. The aim was to evaluate
existing procedures to identify potential gaps and opportunities for improvement.
The research approach was based on the DMAIC framework originating from the Six Sigma
philosophy, integrating qualitative methods and data from semi-structured interviews,
workshops with quantitative data from laboratory testing on prototype products. Statistical
tools were also implemented to analyze variation and correlation and to perform capability
studies.
The study identified that variation in SAP and pulp amount and distribution affects the
product performance. Due to process-related influences, such as environmental conditions
and technical factors, variation will always be present in the production system. Therefore,
the importance of robust design strategies was highlighted together with the value of
proper monitoring of SAP and pulp amount and distribution across the value chain.
Control measures based on gathered information were proposed to enhance process control
and stability, emphasizing the importance of centering processes around target values
rather than only specification limits. While the limited sample size and test scope restrict
broad generalization, the results provide a solid foundation for further investigation and
practical improvements when it comes to quality assurance. Additionally, the research
highlights the value statistical tools can bring within the organization to sustain quality
improvements.
Overall, this thesis contributes to a deeper understanding of SAP and pulp management
in hygiene products and supports Essity’s efforts to optimize product consistency and
customer satisfaction through data-driven quality contro
Precision Manufacturing of Radio Frequency Microwave Filter Structures Using Additive Manufacturing Technologies
No full textThere is a growing demand for high-speed communication, to enable this high frequency
transmission links are utilized. These links employ microwave radio frequency
waveguide filters. Conventional manufacturing of the filters face limitations
in producing multi-cavity structures with complex geometries. Additive manufacturing
processes offer design freedom, reduced material wastage and rapid prototyping,
making it an alternative to produce these filters.
This study investigates the feasibility to manufacture the waveguide filters with the
frequencies E-Band (65 - 67 GHz) and D-Band (130 - 135 GHz) using two metal
Additive manufacturing processes, Binder Jetting Technology (BJT) and Powder
Bed Fusion - Laser Beam (PBF-LB). The filters were manufactured using a Markforged
PX100™ BJT printer by Bosch GmbH and an EOS M100 PBF-LB printer
at Chalmers University of Technology. The BJT filters were printed in two phases,
the first was a trial phase to determine the surface roughness measurements that
were done on selected surfaces of the parts to determine a suitable orientation for
the final print. In the second and final phase, a total of 40 filters were printed, 20
for each band. For PBF-LB, 10 parts were printed 5 for each band. All 40 BJT
and 6 PBF-LB filters underwent electrical characterization using a Vector Network
Analyzer. Five BJT filters of each frequency bands that reached near-target passband
frequency and low loss magnitude were selected for surface treatment with
electroless copper plating to improve performance.
Results showed that BJT-manufactured E-Band filters achieved insertion losses
within the target range while maintaining the required passband frequency, whereas
a subset of D-Band BJT filters met the desired specifications. PBF-LB-manufactured
filters resulted in higher losses and dimensional inaccuracies due to deviations in internal
geometry. The loss magnitudes were reduced by copper plating by nearly half
for both E-Band and D-Band filters. Overall, BJT demonstrated greater suitability
to manufacture high-frequency waveguide filter
Closing the Loop Enabling EV Battery Circularity through Data Transparency
No full textThe transition to electric vehicles (EVs) is rapidly accelerating in Europe due to the climate targets and growing demand for sustainable transformation. The rise in EV adoption is increasing the number of batteries reaching end-of-life (EoL), which causes various challenges such as inefficiencies, waste, and safety risks. These issues result from a lack of alignment across the value chain. This thesis aims to address this misalignment by investigating three research questions: (1) identifying the state-of-art of EoL processes and the its challenges associated with product and data problems, (2) exploring tools and mechanisms enabling data exchange to mitigate these problems, and (3) demonstrating how Battery-as-a-service (BaaS) business model can incentivize data sharing through the identified tools and mechanisms. The thesis applies various qualitative methods, including a systematic literature review, semi-structured interviews with industry experts, and workshops to collect the data. The findings show that the main barriers to effective EoL management stem from uncircular product designs and limited. data transparency across the value chain. Several tools and mechanisms for enhancing data transparency were identified and evaluated using the SWOT framework. The BaaS model was examined through the Circular Business Model Canvas (CBMC) and System Dynamics (SD) modelling, which helped visualize incentive structures for circular designs and collaborations. Future research should acknowledge that these results are based on the current industry, which is subject to change due to high industry uncertainties and fast-spaced developments. Engaging a more diverse group of stakeholders from diverse regions could also strengthen validity of findings. Moreover, key elements of the BaaS model require in-depth analysis to determine their viability and feasibility indifferent market environments at scale. Finally, quantifying and contextualizing the SD model can provide a strong basis for strategic decision-making
Čoarvemátta - Conceptual Model
No full textČoarvemátta comes from the Sámi words for horn and root, after the innermost and strongest part of the reindeer antlers. This symbolizes different qualities and strengths and represents elements that unite, as Čoarvemátta will be a unifying force for the institutions that share the building: the Sami National Theatre Beaivváš and the Sami High School and Reindeer Herding School.
For the concept model, we wanted to display one of the key ideas behind the design: the open, warm core connecting the three wings. This central core echoes the central fire in the traditional Lavvú, and that reference is highlighted in the warm red colors of the materials, as it is in the warm red light
Reuse of Industrial Wastewater Semiconductor Industry
No full textSemiconductor manufacturing processes generate large volumes of wastewater that of ten contain hazardous substances and fine particulate matter. The reuse of this industrial
wastewater through membrane filtration technologies has become an important focus
due to the growing demand for sustainable treatment solutions. This study presents
novel findings on industrial wastewater treatment and reuse using two different experi mental setups.
In the first phase of the study, industrial wastewater samples were taken and analyzed
from four different wastewater streams. Key parameters were tested from the permeate
and retentate to identify composition in water quality to provide a foundation for selecting
the most suitable treatment technologies. The results highlighted a complex composition
and high concentrations of pollutants, such as dissolved ions from inorganic compounds,
total solids and fluoride.
Based on wastewater characterization and literature review, two separate treatment
trains were chosen and tested in laboratory to assess industrial wastewater reclamation.
Experimental Setup 1 consisted of coagulation and flocculation, ceramic membranes
(UF), and polymeric membranes (RO). Experimental Setup 2 consisted of coagulation
and flocculation, ceramic membranes (UF), polymeric membranes (NF) and Direct Con tact Membrane Distillation (DCMD). The performance of these were analyzed based on
results from water quality, permeate flux, and water recovery ratios.
In both Setups, during the coagulation and flocculation phase of the treatment, total sol ids are aggregated into larger particles decreasing both TS and turbidity values signifi cantly. However, the removal was not as efficient during the coagulation phase in Exper imental Setup 1 as in Setup 2, except for COD. In Experimental Setup 1, the treatment
from UF to RO shows more rapid reductions in conductivity, total solids (TS), and NH3-
N. This suggests that the processes during and after UF are highly effective in removing
both suspended particles and dissolved ions. In contrast, Experimental Setup 2 shows
slower, more gradual improvements after UF. The changes in conductivity, TS, turbidity,
ammonia, COD, and TOC from UF to NF and from NF to DCMD are steadily decreasing.
Both setups showed similar overall recovery rates between 54 % and 56 % (RO 2nd pass
and DCMD respectively). However, Setup 1 had much lower water flux than Setup 2 for
UF (39.3 L/m2h and 99.2 L/m2h respectively). This difference likely results from variation
in membrane pore size, though material differences may also have contributed.
While both treatment trains were effective, Setup 1 with ultrafiltration (UF) and reverse
osmosis (RO) could offer a more accessible option for wastewater treatment and reuse,
since the technologies have been studied already with good results in semiconductor
industry. Nevertheless, Setup 2 with DCMD also showed very good results for the water
quality. It is now being developed for full-scale use, which could make it a viable option
3D Scanning of naturally corroded steel bars. An assessment of corrosion behaviour and correlations to critical cross-sections
No full textThe present study focuses on the effect of corrosion of steel reinforcement in reinforced concrete (RC) structures. Corrosion has a large impact on the structural performance of RC structures. A cross-section reduction of the steel bar occurs due to the oxidation of steel which unequivocally convey a non-uniform distribution of pits along the bar length. For a specific length, it is possible to identify a critical cross-section (CCS), which attains the highest loss of area. This thesis defines CCS and illustrates how the different characterizing parameters of the same can be described from the measured average corrosion level of a specific bar length.
To cover this objective naturally corroded steel bars from a real bridge situated in Stallbacka were employed. After structural testing of the edge beams of the bridge, the steel reinforcement was extracted, cut into suitable lengths, and subsequently cleaned. Thereafter the average corrosion level of each specific bar was obtained by using two different techniques, gravimetric and 3D scanning. By weight loss method accurate measurements of the average corrosion level were obtained; by 3D scanning discrete corrosion level in each cross-section was quantified. In addition, variables such as pit depth, pit length, pit width, and, corroded perimeter, were measured from the CCS by combined physical and numerical post-processing of the 3D scanning techniques. These measurements led to a precise description of the CCS.
Finally, relations between all the measured parameters to the average corrosion level were obtained, hence possible predictions of the CCS from the estimated average corrosion level of a specific bar length could be described. Furthermore, in further works, the tensile properties of the utilized steel bars can be correlated to the CCS description and later estimate the reduction of capacity of the steel bars because of the presence of corrosion