Luleå University of Technology Publications
Not a member yet
39356 research outputs found
Sort by
ASTrA: Adversarial Self-supervised Training with Adaptive-Attacks
No full textExisting self-supervised adversarial training (self-AT) methods rely on hand-crafted adversarial attack strategies for PGD attacks, which fail to adapt to the evolving learning dynamics of the model and do not account for instance-specific characteristics of images. This results in sub-optimal adversarial robustness and limits the alignment between clean and adversarial data distributions. To address this, we propose ASTrA (Adversarial Self-supervised Training with Adaptive-Attacks), a novel framework introducing a learnable, self-supervised attack strategy network that autonomously discovers optimal attack parameters through exploration-exploitation in a single training episode. ASTrA leverages a reward mechanism based on contrastive loss, optimized with REINFORCE, enabling adaptive attack strategies without labeled data or additional hyperparameters. We further introduce a mixed contrastive objective to align the distribution of clean and adversarial examples in representation space. ASTrA achieves state-of-the-art results on CIFAR10, CIFAR100, and STL10 while integrating seamlessly as a plug-and-play module for other self-AT methods. ASTrA shows scalability to larger datasets, demonstrates strong semi-supervised performance, and is resilient to robust overfitting, backed by explainability analysis on optimal attack strategies. Project page for source code and other details at https://prakashchhipa.github.io/projects/ASTrA.Full text license: CC BY</p
Framtagande av ett gemensamt arbetssätt för insamling av relevant data för att öka tillgängligheten vid ett produktionsavsnitt : En studie vid produktionsanläggningen på Scania Luleå
No full textDetta examensarbete har genomförts vid Scanias produktionsanläggning i Luleå och syftar till att ta fram ett nytt gemensamt arbetssätt för insamling av relevant data med målet att öka tillgängligheten i ett specifikt produktionsavsnitt. Arbetet har sin grund i ett identifierat behov av förbättrad datakvalitet och tydligare strukturer för att kunna analysera driftstopp för långtidsåtgärder och därmed effektivisera det förebyggande underhållet och minska stopptider. Genom en kombination av observationer, intervjuer, dokumentstudier och deltagande i förbättringsmöten har nuläget analyserats och de största utmaningarna identifierats. Dessa innefattar bland annat otydlig ansvarsfördelning, variation i rapporteringspraxis samt brister i förståelsen för hur tillgängligheten påverkas av datakvalitet. Studien visar också att drifttid rapporteras på ett sätt som försvårar analys, exempelvis när stopp orsakats på planerade raster ändå registreras som akuta driftstörningar. Resultatet av examensarbete är ett konkret förslag på ett nytt gemensamt arbetssätt för långtidsåtgärder för att öka tillgängligheten. Arbetssättet omfattar en tydlig struktur för ansvarsfördelning, förslag på förbättrad rapportslogik, samt en modell för uppföljning och visualisering som syftar till att skapa engagemang och långsiktighet i förbättringsarbetet. Det nya arbetssättet är framtaget med hänsyn till Scanias befintliga organisation och rutiner och är avsett att vara implementerbart och hållbart över tid. Slutresultatet är att ett gemensamt, standardiserat och tydligt förankrat arbetssätt utgör en viktig grund för att öka tillgängligheten i produktionen, vilket är syftet med detta examensarbete. Det möjliggör bättre analys, prioritering av åtgärder och bidrar samtidigt till ökat samarbete mellan olika funktioner inom anläggningen.
Arbetsgivares hinder och möjligheter associerat med anställning av internationell kompetens samt lämpliga strategier och stödmekanismer : En kvalitativ flerfallsstudie i Norrbottens region
No full textI takt med globaliseringens framväxt har behovet av internationell kompetens blivit alltmer central för svenska arbetsgivare, särskilt i regioner med begränsad lokal arbetskraftsförsörjning där demografiska utmaningar, kompetensbrist och geografiskt läge skapar särskilda förutsättningar. I denna flerfallstudie lyfts Luleå, en mindre urban kommun, som är ett särskilt relevant fall på grund av dess aktuella rekryteringsbehov och de samarbetsmöjligheter som finns mellan kommunen och lokala arbetsgivare. Denna studie undersöker hur rekryteringsprocesser för internationell arbetskraft kan utvecklas och effektiviseras med fokus på att bemöta specifika hinder och möjligheter som uppstår i sådana sammanhang. Syftet är att identifiera strategier och stödmekanismer inom rekryteringskedjan och därefter utveckla ett ramverk för internationell rekrytering som i första hand riktar sig till arbetsgivare, men som även kan tillämpas av andra relevanta aktörer inom rekrytering och etablering. Studien bygger på en kvalitativ metodansats och utgår från semistrukturerade intervjuer med arbetsgivare, specialister och aktörer inom det kommunala och regionala stödsystemet. Det empiriska materialet har analyserats med hjälp av relevanta teoretiska perspektiv, vilket möjliggör en mångdimensionell förståelse av de strukturella, individuella, organisatoriska samt teknologiska faktorer som påverkar rekryteringsprocessen. Resultatet visar att rekrytering av internationell arbetskraft präglas av komplexa hinder som berör bland annat språkliga samt kulturella skillnader, bristande administrativa och juridiska stödstrukturer samt en avsaknad av långsiktiga integrationsinsatser. I samband med detta lyfts också strategier och stödmekanismer som kan underlätta rekryteringen som flexibilitet i språkutbildning, samverkan mellan aktörer samt tillämpning av digitala verktyg. Det framtagna ramverket som presenteras, från förarbete till långsiktig etablering, betonar vikten av att se rekryteringen som en sammanhängande och samhällsknuten process. Slutsatserna bidrar till ett antal rekommendationer till Luleå kommun, som möjligheter relaterat till tillämpning av flyttbidrag, att erbjuda administrativt och juridiskt stöd till arbetsgivare samt utveckla flexibla SFI-lösningar. Studien bidrar därför till en fördjupad förståelse för hur kommuner och arbetsgivare tillsammans har möjlighet att arbeta mer strategiskt gällande att attrahera, rekrytera och integrera internationell kompetens, och inte bara i Luleå, utan även i liknande kontexter över hela Sverige. With the rise of globalization, the demand for international talent has become more important to Swedish employers. Especially in regions with limited local labor supply and where challenges related to demographic, skills shortages, and geographical location create particular conditions. In this multiple case study, Luleå, a smaller urban municipality, is highlighted as a particularly relevant case due to its current recruitment needs and the collaboration opportunities that exist between municipality and local employers. This study examines how recruitment processes for international labor can be developed and efficiently focusing on addressing specific obstacles and opportunities that arise in such contexts. The purpose is to identify strategies and support mechanisms within the recruitment chain and thereafter develop a framework for international recruitment that is primarily intended for employers, but can also be applied by other relevant actors involved in recruitment and establishment. This study is based on a qualitative methodological approach and is derived from semi-structured interviews with employers, specialists, and actors within the municipal and regional support systems. The empirical material has been analyzed using relevant theoretical perspectives, enabling a multidimensional understanding of the structural, individual, organizational, and technological factors that affect the recruitment process. The results show that recruitment of international labor is characterized by complex challenges including linguistic and cultural differences, inadequate administrative and legal support structures, and a lack of long-term integration efforts. Alongside this, strategies and support mechanisms that facilitate recruitment are also highlighted, such as flexibility in language training, cooperation between actors, and the application of digital tools. The developed framework presented, from preparatory work to long-term establishment, emphasizes the importance of viewing recruitment as a cohesive and community-linked process. The conclusion contributed to a number of recommendations for Luleå municipality, including possibilities related to the application of relocation grants, offering administrative and legal support to employers, and developing flexible SFI (Swedish for Immigrants) solutions. Therefore, the study contributes to a deepened understanding of how municipalities and employers together have the opportunity to work more strategically regarding attracting, recruiting, and integrating international competence, not only in Luleå but also in similar contexts across Sweden.
Effects of Physiological Loading from Patient-Derived Activities of Daily Living on the Wear of Metal-on-Polymer Total Hip Replacements
No full textThe current pre-clinical testing standards for total hip replacements (THRs), ISO standards, use simplified loading waveforms that do not fully replicate real-world biomechanics. These standards provide a benchmark of data that may not accurately predict in vivo wear, necessitating the evaluation of physiologically relevant loading conditions. Previous studies have incorporated activities of daily living (ADLs) such as walking, jogging and stair negotiation into wear simulations. However, these studies primarily used simplified adaptations that increased axial forces and applied accelerated sinusoidal waveforms, rather than fully replicating the complex kinematics experienced by THR patients. To address this gap, this study applied patient-derived ADL profiles—jogging and stair negotiation—using a three-station hip simulator, obtained through 3D motion analysis of total hip arthroplasty patients, processed via a musculoskeletal multibody modelling approach to derive realistic hip contact forces (HCFs). The results indicate that jogging significantly increased wear rates compared to the ISO walking gait waveform, with wear increasing from 15.24 ± 0.55 to 28.68 ± 0.87 mm3/Mc. Additionally, wear was highly sensitive to changes in lubricant protein concentration, with an increase from 17 g/L to 30 g/L reducing wear by over 60%. Contrary to predictive models, stair descent resulted in higher volumetric wear (8.62 ± 0.43 mm3/0.5 Mc) compared to stair ascent (4.15 ± 0.31 mm3/0.5 Mc), despite both profiles having similar peak torques. These findings underscore the limitations of current ISO standards in replicating physiologically relevant wear patterns. The application of patient-specific loading profiles highlights the need to integrate ADLs into pre-clinical testing protocols, ensuring a more accurate assessment of implant performance and longevity.Validerad;2025;Nivå 2;2025-07-08 (u2);Full text: CC BY license;</p
Assessing the business impacts of the COVID-19 pandemic and the Russia–Ukraine war: the role of corporate sustainability and financial performance
No full textPurpose: This study aims to examine whether corporate sustainability practices act as a shield ensuring financial performance in times of crisis. Design/methodology/approach: Data from 471 European firms during 2018–23 were categorized into three periods: (1) before the COVID-19 pandemic, (2) during the pandemic and (3) during the Russia–Ukraine War. In this regard, pooled ordinary least squares, two-step generalized method of moments, Kruskal–Wallis and Mann–Whitney non-parametric tests were performed to examine the impact of crises and whether corporate sustainability practices act as a shield to ensure financial performance. Findings: The proponents of the resource-based view (RBV) have argued that corporate sustainability practices create additional valuable resources that might work as a shield for ensuring financial performance even in times of crisis. While many studies have examined the impact of the recent pandemic, few compare the impacts of the war’s repercussions on business performance. This study found a distinction between natural (Covid) and human-induced crises (war) on corporate financial performance. The results suggest that sustainability practices might work as a shield during a natural crisis but not during a human-induced one. Practical implications: The authors observed that throughout the COVID-19 period, policymakers extended assistance to businesses, but during the subsequent European geopolitical crisis, their ability to offer such support has been absent. This absence might have a negative impact on corporate financial performance as government support increases investor confidence which artificially boosts market valuations. In addition, with the experience of COVID-19, investors might consider issues other than corporate sustainability practices when evaluating the market value of a firm. Even though we have seen an increasing trend in sustainability practices since the start of COVID-19 and continuing through the Russia–Ukraine war, these practices are not significantly reflected in or valued by investors. Originality/value: This study sheds light on the comparative impact of the COVID-19 pandemic and the Russia–Ukraine War upon business performance and raises the bar on the theoretical understanding through a RBV.Full text license: CC BY</p
Medarbetarskap i praktiken : Ansvar, delaktighet och engagemang i relation till organisationens mål
No full textDenna studie syftar till att undersöka hur anställda på en lokal reseavdelning inom en större organisation upplever sitt medarbetarskap samt identifiera vilka förutsättningar de anser främjar ett aktivt och engagerat medarbetarskap i relation till organisationens mål. Studien genomfördes i en verksamhetskontext präglad av högt tempo och tydliga leveranskrav, vilket ställde särskilda krav på fungerande samarbete och ansvarstagande. För att besvara studiens syfte och forskningsfrågor har sex semistrukterade intervjuer med anställda, däribland en avdelningschef genomförts. Studien har ett kvalitativt angreppssätt med ett målstyrt urval och har analyserats tematiskt enligt Braun och Clarke (2006) modell. Analysen har tolkats med hjälp av Hällstén och Tengblads (2006) modell medarbetarskapshjulet och Schein (2010) teori om organisationskultur. Studiens teoretiska ramverk har använts för att fördjupa förståelsen av hur medarbetarskap formas i samspelet mellan individens drivkrafter, organisatoriska strukturer och den vardagliga kultur som råder på arbetsplatsen. Fokus har legat på teman som ansvar, delaktighet, trygghet, utveckling och ledarskap. Resultatet visar att ansvarstagande ses som en självklar del i arbetet och att detta ansvar upplevs både individuellt och kollektivt. Trygghet i arbetsgruppen, ett närvarande ledarskap och möjligheter till utveckling beskrivs som centrala förutsättningar för engagemang och delaktighet. Studien visar att medarbetarskap växer fram i en vardag där både struktur och kultur samverkar. Resultatet belyser också hur medarbetarskapets uttryck kan variera beroende på individens erfarenhet, roll och arbetsrelationer samt hur medarbetarskapet formas i mötet med den konkreta arbetsvardagen. När tillit, öppenhet och inflytande ges utrymme stärks känslan av meningsfullhet och viljan att bidra till verksamhetens och organisationens mål.
Anomaly detection in time series data for Edge computing
No full textThe world is progressing towards increasingly large and complex networks of smart devices. The digital networks provides the infrastructure required to operate all of a society’s most important and basic needs such as water and power. The smart devices themselves are highly autonomous units operating at the networks’ edge and are able to generate or process all kindsof data and can operate other equipment as well. One common processing task is analysing large amounts of raw sensor data, looking for anomalous events or signals leaving theiroperating boundaries. Matrix Profile is a novel algorithm that can perform this data analysis. It is able to discover and flag both large events over time and singular points in a range of time series data. The algorithm can handle most kinds of data, without requiring extensive training or making large adjustments to parameters. This work investigates a variant of the Matrix Profile, which runs with a small, internal buffer that allows the algorithm to handle streaming data. The purpose of the investigation is to measure the effectiveness of the algorithm and see if it can analyse sensor data while runningon a smart device. Once implemented, the algorithm analysed examples with well known outputs that could confirm the produced results as being valid. A Raspberry Pi equipped with a sensor board could then run the algorithm, mimicking a real scenario, and analysed live sensor data while measuring the performance too. This investigation could conclude that the Matrix Profile, once adapted, shows good indications at being efficient enough to run directly on small devices. This allows the possibility of offloading the centralised data analysis from the core of large networks and instead distribute the analysing tasks to the sensors themselves
Production of LiOH from Li2CO3 extracted from recycled Li-ion cells
No full textThe need for lithium hydroxide (LiOH) is increasing with the production of lithium-ion cells. The conversion of lithium carbonate (Li2CO3) to LiOH is an important step to recycle Li-ion cells. The foundation for this project was an EAF (Electric Arc Furnace) dust that had been collected from pyrometallurgical recycling of lithium-ion batteries. Using precipitation with slaked lime (Ca(OH)2), Lithium hydroxide could be refined from the EAF dust. The precipitation process showed a good separation of lithium (Li) from other metals like cobalt (Co), manganese (Mn), nickel (Ni), iron (Fe), magnesium (Mg) and copper (Cu). At the cost of reduced recovery, the final lithium hydroxide concentration could be enhanced by increasing the initial addition of EAF dust. The purity of the final products however needs to be improved to reach the standard for battery production. Overall, the process is promising but needs more testing.Efterfrågan på litiumhydroxid (LiOH) har ökat med produktionen av litium-jonbatterier. Konvertering av litiumkarbonat (Li2CO3) till LiOH är ett viktigt steg för att återvinna li-jon batterier. Grunden för detta projekt var ett EAF pulver som hade producerats från pyro-metallurgisk återvinning av litium-jon batterier. Med hjälp av utfällning med släckt kalk (Ca(OH)2) kunde litiumhydroxid utvinnas från koncentratet. Processen hade en bra separation av litium (Li) från andra metaller så som kobolt (Co), mangan (Mn), nickel (Ni), järn (Fe), magnesium (Mg) och koppar (Cu). På bekostnad av utbytet kunde den slutliga koncentrationen av litiumhydroxid ökas med en större tillsats av EAF pulvret. Renheten hos den slutgiltiga produkten måste dock förbättras för att nå standarden till batteriproduktion. I helhet är processen lovande men behöver mer testning
Augmenting Large Language Models with Domain-Specific Insight: Establishing SC-KAE Framework for Improved Real-World Application of LLMs in Supply Chain
No full textRecent advancements in artificial intelligence (AI) and natural language processing (NLP) have transformed digital interactions through the emergence of large language models (LLMs). These models demonstrate impressive capabilities in generating human-like text and handling general-purpose queries. However, their application in specialized domains, such as supply chain management (SCM), remains challenging due to limitations in comprehending domain-specific terminology, handling complex data structures, and navigating unique operational contexts. Addressing these issues requires tailored solutions that augment LLMs with domain-specific knowledge. This research explores the integration of Knowledge Graphs (KGs) into Retrieval-Augmented Generation (RAG) pipelines to enhance the performance of LLMs in domain-specific tasks. Using SCM as a test domain, the study investigates how KGs can provide structured, factual context to improve the accuracy, relevance, and robustness of LLM-generated responses. The proposed framework combines the generative strengths of LLMs with the factual grounding of KGs. It incorporates semantic entity extraction, subgraph construction, and knowledge augmentation to create a unified context for reasoning. Two datasets are used to evaluate the approach: a novel SCM benchmark dataset covering eight core supply chain functions (such as procurement, inventory management, logistics etc.) and the LTU chatbot QA dataset. Performance is measured using standard metrics like ROUGE and METEOR, as well as truthfulness scores assessed by LLM-based evaluation. The study evaluated the performance of various models, including smaller open-weight models like Gemma3, Llama3.1, GPT-OSS, and Qwen3, alongside a larger, state-of-the-art model such as GPT-5 (Nano). Results demonstrated that KG integration enhanced performance compared to traditional RAG approaches, with smaller models achieving notable gains that reduced the performance gap with larger models. This underscores the potential of KGs to enable cost-effective and scalable LLM-based solutions by leveraging structured, domain-specific knowledge. While the results are promising, challenges remain. The accuracy of the system heavily depends on the completeness and quality of the KG. Future work will focus on optimizing KG construction, improving retrieval efficiency, and exploring the framework’s applicability to other domains like healthcare and finance
Emotions towards two national minority languages in Sweden : Insights from language heritage communities
No full textThis article explores the lived experiences of participants with heritage language background in Norrbotten County, Northern Sweden, with regard to two national minority languages – Finnish and Meänkieli. In the study, explanations of participants’ visualised and multicoloured language portraits were analysed to gain insight into the how the participants have experienced these languages and the emotions these experiences have evoked. Participants valued these languages, but felt excluded in their relationship to them, because only previous generations have had natural access to them. The loss of the language and attitudes about this loss were explained through both individual and collective experiences, such as Swedification and inadequate support. Participants expressed a vision of the future in which they could become part of the minority language community. To achieve such a future, education is essential. It can maintain and strengthen language skills and foster cooperation between informal environments and formal educational spheres. Our findings indicate that through collective action, more positive attitudes towards minority languages and fruitful education could be achieved. Validerad;2025;Nivå 1;2025-11-06 (u8);Full text license: CC BY</p