LUTPub (LUT University)
Not a member yet
24632 research outputs found
Sort by
Intrusion Detection System for Internet of Things Using Image Classification
Get PDFThe Internet of Things (IoT) is a fast-moving technology that is gradually being integrated into our daily lives. As communication protocols and network technologies evolve, the vulnerability of IoT devices to cyberattacks also increases, fueling the need to address this pressing problem. In this work, we propose an intrusion detection system based on a residual neural network with inductive transfer learning. This learning approach is designed to detect cyberattacks on IoT devices by visually encoding the CIC-IoT-2023 dataset from multivariate numerical data to visual formats (images). Extensive numerical experiments are carried out using the well-known dataset CIC-IoT-2023, which consists of 34 classes. Furthermore, the ensuing results demonstrate the effectiveness of our proposed solution, which achieves an accuracy of 99.35% with a latency of 70.9 ms, a detection time of 99.6 s for the entire dataset, and executes 316.82 predictions per second, outperforming existing solutions in terms of the ability to distinguish between the 34 classes of IoT cyberattacks while reducing overfitting.Post-print / Final draf
Lääkeaineiden väärinkäytön tunnistaminen biologisista matriiseista korkean erotuskyvyn nestekromatografia : tandemmassaspektrometrialla
No full textThis thesis provides a literature review on the detection of pharmaceutical drug abuse from biological matrices using High - Performance Liquid Chromatography coupled with Tandem Mass Spectrometry (HPLC-MS/MS) and the challenges and strengths of this analytical method. Detection of pharmaceutical drug abuse using HPLC-MS/MS is widely applied in forensic toxicology, including in autopsies, drug-related crimes, custody cases and workplace and doping drug screening.
This thesis presents the principles of HPLC-MS/MS, commonly used biological matrices and their sample pretreatments, HPLC-MS/MS data and its interpretations as well as the challenges and strengths of this analytical method. HPLC-MS/MS is a highly selective and sensitive analytical tool providing both qualitative and quantitative analysis. However, challenges such as matrix effects, complex sample pretreatments and sample tampering, short detection windows, false results and misinterpretation of the data still remain. Despite these challenges, HPLC-MS/MS remains a useful and powerful tool in forensic toxicology when identifying pharmaceutical drug abuse.Tämä kandidaatintyö on kirjallisuuskatsaus, jossa tarkastellaan lääkeaineiden väärinkäytön tunnistamista biologista matriiseista käyttäen korkean erotuskyvyn nestekromatografia - tandemmassaspektrometriaa (HPLC-MS/MS). Tässä työssä käsitellään myös tämän analyysimenetelmän heikkouksia ja vahvuuksia. Lääkeaineiden väärinkäytön tunnistamista HPLC-MS/MS - menetelmää käyttäen hyödynnetään laajasti rikosteknisessä toksikologiassa, kuten ruumiinavauksissa, huumausaineisiin liittyvissä rikoksissa, huoltajuustapauksissa sekä työpaikka- ja dopingtestauksissa.
Työssä esitellään ensin HPLC-MS/MS:n toimintaperiaatteet, usein käytetyt biologiset matriisit ja niiden esikäsittelyt, HPLC-MS/MS dataa ja sen tulkintaa sekä analyysimenetelmän haasteet ja vahvuudet. HPLC-MS/MS on erittäin selektiivinen ja herkkä analyysimenetelmä, joka voi tuottaa sekä kvalitatiivista että kvantitatiivista dataa. Menetelmään liittyy kuitenkin haasteita, kuten matriisihäiriöitä, vaikeita näytteiden esikäsittelyjä ja näytteiden manipulointia, lyhyitä lääkkeiden havaitsemisikkunoita, vääriä tuloksia sekä datan tulkintavirheitä. Näistä haasteista huolimatta HPLC-MS/MS on hyödyllinen ja tehokas työkalu, kun halutaan tutkia lääkeaineiden väärinkäyttöä rikosteknisessä toksikologiassa
A Qualitative Study of Sensemaking Through Data Exploration
Get PDFSystems for non-expert audiences to interact with and use data, such as open data portals and other data exploration tools, are increasing in number; however, there has not been much research on how non-expert audiences make sense of data via these exploration tools. This paper presents a qualitative study of non-experts sensemaking of local data via a prototype of an interactive data exploration tool that was created based upon a series of design principles. As an outcome of this study we present our insights on sensemaking via interactive data exploration tools and we provide recommendations for improvements to the series of design principles that the prototype was built upon.Post-print / Final draf
Experimental characterization of open loop passive heat removal
Get PDFThis study characterized the functioning of an open loop gravity-driven containment passive heat removal system of a large pressurized water reactor by means of experiments performed at the PASI test facility of LUT University, Finland. PASI is height-scaled 1:2 compared to the reference system and comprises a containment vessel, a heat exchanger with 15 tubes, a water pool, and interconnecting riser and downcomer pipelines.
The objective of the study was to determine whether or not an open natural circulation loop is sensitive to disturbances that could make it an unreliable safety system. Particularly, the aim was to address the concerns about the small driving forces and unstable operation that might threaten the structural integrity of such a system. Specifically, the behavior of the natural circulation flow in the loop, and the system’s heat removal capability, were investigated.
The quasi-steady behavior of the system was characterized in the single-phase and twophase modes. Periodic testability of the system in a real power plant was demonstrated. Dynamic loads to the piping caused by flashing-based flow oscillations were assessed and found to be small. The conditions for the stable operation of the system were identified; they depend on the riser flooding and reaching of the CCFL criterion in the riser. In the quasi-steady conditions, the amplitude of flow oscillations is somewhat constant, but the frequency is roughly proportional to the heating power. The impact of the gravity head was examined; the pool water level affects the oscillation frequency and amplitude slightly. Finally, the system response to an added loop flow resistance was studied.
Based on the study, the open loop heat removal system was found to be very robust and not sensitive to distractions such as added loop flow resistance. Heat transfer from the containment was efficient in all tested conditions. Furthermore, the system was proven to be strongly self-regulating: the system immediately finds a new operation point after a change in, e.g., heating power or flow resistance. Most probably, due to the low heating power originating from reactor decay heat, in an accident situation, this type of safety system would operate in a quasi-steady mode accompanied by continuous, undamped two-phase flow oscillations.ei tietoa saavutettavuudest
Viitekehys generatiivisen tekoälyn käyttöönottoon yrityksen liiketoiminnassa
No full textGenerative artificial intelligence (GenAI) is rapidly evolving and is considered as one of the most significant technological breakthroughs of this decade. Despite the rapid development, the value-adding integration of GenAI into business operations remains in its early stages. To address this implementation challenge, this thesis presents a practical framework for implementing GenAI into business processes. In addition, this thesis answers questions about which business departments benefit the most from GenAI and how these benefits can be achieved.
The literature review discusses GenAI’s capabilities, ethics and its potential from a business perspective. In addition, the review examines general process optimization methods, investigates their implementation into business processes and discusses how GenAI can become part of this value chain. In the empirical part of the study, the potential of GenAI is evaluated within a focal company operating in the aviation sector, utilizing the insights from the literature review, interviews, work study and qualitative analysis.
This thesis presents a practical three-level framework enabling the focal company to integrate GenAI into its business processes and support its workforce. A key finding is that GenAI can be used in three main ways: independent general use, task-specific use, and use within automation solutions. The study shows different departments gaining varying benefits depending on how GenAI is applied.Generatiivinen tekoäly (GenAI) kehittyy huomattavalla vauhdilla ja sitä pidetään yhtenä tämän vuosikymmenen merkittävimpänä teknologisena läpimurtona. Vaikka GenAI teknologiana kehittyy vauhdilla, on sen implementointi yrityksiin tuottavasti vielä alkumatkassa. Tämän implementoinnin haasteen ratkaisemiseksi tutkielmassa kehitetään viitekehys GenAI:n käyttöönottoa varten. Lisäksi vastataan kysymyksiin siitä, mitkä yritysten liiketoimintaosastot hyötyvät GenAI:sta eniten ja miten nämä hyödyt ovat saavutettavissa.
Työn kirjallisuuskatsauksessa käsitellään GenAI:n ominaisuuksia, etiikkaa ja potentiaalia liiketoiminnan näkökulmasta. Lisäksi katsauksessa käsitellään yleisiä prosessien kehittämismalleja, arvioidaan näiden kytkeytymistä liiketoimintaan ja tarkastellaan tapoja, joilla GenAI voidaan liittää osaksi tätä arvoketjua. Työn empiirisessä osassa käsitellään ilmailualan kohdeyrityksen mahdollisuuksia hyödyntää GenAI:ta, perustuen kirjallisuuskatsauksen tuloksiin, haastatteluihin, työntutkimukseen ja laadulliseen analyysiin.
Työn tuloksena luodaan kolmiportainen käytännönläheinen viitekehys, jonka avulla kohdeyritys voi hyödyntää GenAI:n potentiaalia liiketoimintaprosessien kehittämiseen ja henkilöstön tukemiseen. Keskeisenä havaintona todetaan, että GenAI:ta voidaan käyttää kolmella keskeisellä tavalla: omatoimisesti useisiin yleisiin työtehtäviin, tehtäväkohtaisesti valittuihin prosesseihin sekä osana automaatioratkaisuja. Tutkimuksessa havaitaan, että liiketoiminnan eri osastot hyötyvät käyttötavoista vaihtelevissa määrin
Proceedings of the 8th International Seminar on ORC Power Systems
No full textThis volume presents the proceedings of the 8 th International Seminar on Organic Rankine Cycle Power Systems (ORC 2025), taking place in Lappeenranta, Finland, from 9th to 11th September 2025. The event is hosted by LUT University on behalf of the Knowledge Centre on Organic Rankine Cycle Technology (KCORC). The ORC Seminar is a recurring international conference dedicated to Organic Rankine Cycle technology, bringing together researchers, industry professionals, and policymakers to share recent developments and practical insights. The contributions in this book reflect the current state of ORC research and applications, covering topics such as system design, components, working fluids, and integration into energy systems. The proceedings also highlight the role of ORC technology in the transition toward more sustainable energy solutions.Publishers versio
Vertaileva analyysi koneoppimisen malleista taloudellisen ahdingon luokittelussa
No full textFinancial distress prediction is a critical area of study within financial risk management, influencing decision-making for various stakeholders, including financial institutions, regulators, investors, and corporate managers. Traditional models, such as logistic regression, have been widely used to assess financial distress, but recent advancements in machine learning offer promising improvements in predictive accuracy and model efficiency.
The research conducted evaluates multiple models, including logistic regression, K-nearest neighbors, decision trees, support vector machines, and neural networks. The results indicate that these models exhibit strong predictive capabilities, with all models achieving an accuracy exceeding 67%. Among the evaluated models, traditional logistic regression performed the weakest, highlighting the limitations of conventional statistical methods in handling complex financial data. KNN, while effective, exhibited marginally lower performance than the decision tree model. Neural networks emerged as the most effective model, achieving the highest accuracy and AUC. This underscores their superior predictive performance in financial distress forecasting. These findings reinforce the practical utility of machine learning in financial risk assessment. While neural networks demonstrate the highest predictive accuracy, their limited explainability raises concerns regarding the transparency of the model's decision-making process. In contrast, decision trees provide clearer reasoning behind predictions, making them valuable for stakeholders requiring transparent decision-making frameworks.Taloudellisen ahdingon ennustaminen on keskeinen tutkimusalue rahoitusriskien hallinnassa, ja se vaikuttaa monien sidosryhmien päätöksentekoon, kuten rahoituslaitosten, sääntelyviranomaisten, sijoittajien ja yritysjohtajien. Perinteisiä malleja, kuten logistista regressiota, on käytetty laajasti taloudellisen ahdingon arvioinnissa, mutta viimeaikaiset koneoppimisen edistysaskeleet tarjoavat lupaavia parannuksia ennustetarkkuudessa ja mallien tehokkuudessa.
Tässä tutkimuksessa arvioitiin useita malleja, mukaan lukien logistinen regressio, KNN, päätöspuut, SVM ja NN. Tulokset osoittavat, että näillä malleilla on vahva ennustekyky, ja kaikki mallit saavuttivat yli 67 %:n tarkkuuden. Arvioiduista malleista perinteinen logistinen regressio suoriutui heikoimmin, mikä korostaa perinteisten tilastollisten menetelmien rajoituksia monimutkaisen taloudellisen datan käsittelyssä. KNN malli oli tehokas, mutta sen suorituskyky jäi hieman päätöspuiden alapuolelle. NN osoittautui tehokkaimmiksi, saavuttaen parhaan tarkkuuden ja AUC:n, mikä korostaa niiden ylivoimaista ennustesuorituskykyä taloudellisen ahdingon ennustamisessa. Nämä havainnot vahvistavat koneoppimisen käytännön hyödyn rahoitusriskien arvioinnissa. Vaikka NN saavuttaa korkeimman ennustetarkkuuden, niiden rajallinen selitettävyys herättää huolta mallin päätöksenteon läpinäkyvyydestä. Sen sijaan päätöspuut tarjoavat selkeät perustelut ennusteilleen, mikä tekee niistä arvokkaita sidosryhmille, jotka tarvitsevat läpinäkyviä päätöksenteon välineitä
Computationally efficient modeling of internal combustion engine for system-level studies of vehicle power trains
Get PDFThe power train development of vehicles, non-road mobile machines (NRMM), and heavyduty machines focuses on hybrid electric power trains and combustion engine technology for new, environmentally friendly fuels. Concurrently, development needs are increasing, and research and development time is shortening. Numerical simulation methods offer significant benefits for machine design, and their use is growing in mobile machine design tasks. The share of battery electric power trains and hydrogen and fuel cell-powered machines is increasing. However, it seems that combustion engines will be needed for decades in different applications where charging possibilities are limited or the operation time of electric or hydrogen-based power trains is limited.
This study aims to improve the simulation models of combustion engines for initial-phase power train studies. Three computationally efficient methods for simulating combustion engines have been developed and presented. The first method presents a static fuel consumption map estimation method from a specific fuel consumption curve, and the second method improved it by adding the dynamic torque response of the turbocharged engine, utilizing different transfer functions and time series-based models. The presented methods were validated by comparing the simulated results to the actual engine experiments. The third method utilized a reduced engine model, which was calibrated to produce reasonably accurate results of a real engine. Two of these research studies used a statistical approach to predicting the exhaust emission and fuel injection profiles.
Key findings indicate that computationally efficient and easily calibrated models can predict modern NRMM combustion engine fuel consumption and dynamic behavior. This study also expands the static fuel consumption estimation to different electric-hybrid NRMM applications and duty-cycle simulation for good accuracy. Estimating emissions and fuel injection profiles using machine learning methods shows that engine output can be predicted well; however, the accuracy of the emission formation and fuel injection profiles includes uncertainties, and further research is needed.ei tietoa saavutettavuudest
The cost-benefit analysis model for evaluating potential investment in automating water bottling line segment
Get PDFThis thesis has evaluated the economic viability of automating the PET bottle blowing segment within a water bottling production line. The study was done on a real-world case example from a bottling company in Bosnia and Herzegovina. It has been investigated whether replacing the two current operating semi-automated machines with one fully automated PET blowing machine would improve operational efficiency and justify the required capital investment. The analysis applies a structured cost-benefit model that incorporates actual production data into the financial indicators, Net Present Value (NPV), Benefit-Cost Ratio (BCR), Payback Period (PBP), and Internal Rate of Return (IRR) over a 10-year period. Results of the analysis show that automation would reduce labour needs from four operators to just one operator, it would lower the cost of the bottle from 0.15 BAM to 0.13 BAM, and would cut energy consumption per bottle by over 10%. The investment has been calculated to yield an NPV of 165,356 BAM and a BCR of 1.29. Furthermore, sensitivity analysis shows that even under more favourable and unfavourable conditions, the project still remains economically viable. The study concludes that investment in the automation of the PET blowing process presents an economically justified decision for the company
Generatiivisen tekoälyn energiankulutus datakeskusten kautta : energiahuollon haasteet ja EU:n poliittiset toimintamallit
No full textThis master’s thesis examines, through a literature review, the energy demand of GPT-based chatbots enabled by Generative AI (GenAI) in the context of rising energy consumption, energy security, and European Union (EU) legislation. The study addresses chatbot usage throughout their lifecycle, the role of data centers, the concept of energy security, and existing EU policy guidelines relevant to data centers and GenAI.
The results indicate that GenAI language models is rapidly increasing. This trend is closely linked to data centers energy consumption. Which is projected to reach approximately 5 % of the EU’s total electricity use by 2030. Despite collective technological improvements, the growing energy intensity of AI workloads cannot be fully offset by technology alone. Peak- load management and grid flexibility emerge as critical factors for energy security, particularly as the share of renewable energy production increases. EU policy measures, such as the Energy Efficiency Directive and the AI Act, emphasize transparency, documentation, and efficiency standards, but point-out that energy-specific obligations for AI developers remain fragmented.
The research concludes that GenAI development is a both a driver of energy challenges and as a potential mitigation tool. While its energy consumption raises concerns, leveraging AI in smart grids and load management also offers opportunities. Strengthening EU regulation will be essential to achieve a balance between AI development and sustainable energy use.Tässä diplomityössä selvitettiin kirjallisuuskatsauksen avulla generatiivisen tekoälyn mahdollistamien GPT-pohjaisten chatbottien energiantarvetta kasvavan energiankulutuksen, energiaturvallisuuden ja Euroopan unionin lainsäädännön näkökulmasta. Tutkimukseen sisältyivät chatbotin toiminta koulutus- ja päättelyvaiheessa, datakeskusten rooli, energiaturvallisuuden määrittely sekä Euroopan unionin olemassa oleva lainsäädäntö koskien datakeskuksia ja generatiivista tekoälyä.
Tulokset osoittavat, että generatiivisen tekoälyn kielimallien käyttö on yleistymässä nopeasti. Tämä kehitys on sidoksissa merkittävästi datakeskusten energiankulutukseen, jonka arvioidaan saavuttavan noin 5 % EU:n kokonaissähkönkulutuksesta vuoteen 2030 mennessä. Huolimatta kollektiivisesta teknologian kehityksestä, tekoälytyökuormien energiaintensiteetin kasvu ei ole täysin kompensoitavissa vain teknologialla. Huippukuormien hallinta ja sähköverkkojen joustavuus nousevat keskeisiksi tekijöiksi energiaturvallisuuden näkökulmasta, erityisesti uusiutuvan energian osuuden tuotannon kasvaessa. EU:n politiikkatoimet, kuten energiatehokkuusdirektiivi ja tekoälyasetus, korostavat läpinäkyvyyttä, dokumentointia ja tehokkuusstandardeja, mutta paljastavat, että energiaan liittyvät velvoitteet tekoälykehittäjille ovat edelleen hajanaisia.
Johtopäätöksenä voidaan todeta, että generatiivisen tekoälyn kehitys toimii sekä energiahaasteiden ajurina että mahdollisena rankaisuna. Vaikka sen energiankulutus herättää huolta, tekoälyn hyödyntäminen älykkäissä sähköverkoissa ja kuormanhallinnassa tarjoavat myös mahdollisuuksia. EU:n sääntelyn vahvistaminen on olennaista, jotta tekoälyn ja kestävän energiankäytön välillä voidaan saavuttaa tasapaino