HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Écologie sociale et projets numériques : deux cas d’études
No full textInternational audienceCe poster s’insère dans l’axe “low tech et FAIR” pour présenter deux projets numériques qui ont été tous les deux conçus en suivant les principes de l’écologie sociale. Ces deux projets, très différents du point de vue des données traités, des objectifs et du public, offrent une idée de la variété des contextes dans lesquels les principes de l’écologie sociale peuvent s’appliquer et être pertinents.Le premier projet concerne l’édition et la valorisation du corpus d’Elisa Chimenti, écrivaine italo-marocaine francophone oubliée. L’objectif du projet est, dans ce cas, d’utiliser le numérique pour développer le traitement et l’étude d’un corpus très varié – tant du point de vue des genres que des langues, des thèmes et des conditions de conservation. Le but final étant celui de diffuser et de faire connaître l’œuvre de l’autrice, le public visé est le plus vaste possible. Le deuxième projet concerne la création d’une base de données pour un répertoire de théâtre d’Ancien Régime, à savoir la Comédie Italienne de Paris au XVIII siècle. L'objectif est de faire des analyses quantitatives et des visualisations de données concernant la programmation du théâtre à l'époque étudiée. On vise donc prioritairement un public académique de chercheur.euse.s en études théâtrales, qui connaissent déjà le contexte : la base de données leur permettra de faire des analyses croisées.Malgré leurs différences, les deux projets ont été également guidés par une volonté de répondre aux enjeux écologiques, et cela a été fait dans les deux cas en suivant la piste de l’écologie sociale. Formulée pendant les années ‘60 par Mourray Bookchin (Our synthetic environment, Bookchin 1962, The ecology of freedom, Bookchin 1982), il s’agit de la première philosophie libertaire à faire de la cause écologiste une cause sociale : une technologie est ainsi conçue comme le résultat d’une certaine organisation sociale, tendant à maintenir et reproduire la société qui l’a générée. Les principes de cette philosophie sont définis clairement : travail non hiérarchique et décentralisé, production à petite échelle et par conséquence appropriable, valorisation de l’interdépendance ainsi que de la contribution de chaque élément (complémentarité) à la composition d’une société (unité).Concrètement, dans le cadre d’un projet numérique, ces pistes peuvent se traduire dans des solutions technologiques simples, telles que les pratiques du minimal computing et des low-tech. Cela nous permet non seulement d’atteindre les défis FAIR (Findable, Accessible, Interoperable, Reusable) mais aussi d'aller au-delà, en répondant à des enjeux sociaux et en tenant en compte les besoins des communautés
Self-explaining Neural Network for Multi-criteria Sentiment Analysis
No full textInternational audienceDespite a rich literature on explainable classification, to our knowledge, there is a lack of classification methods that come with reliable user-orientated explanations supporting the predictions. To complement the existing literature on explainable classification, we propose a Self-Explaining Neural Network for Multi-Criteria Sentiment Analysis (SENN4MCSA) which consists of three key components: topic modeling, which extracts relevant topics from the training data, topics-criteria alignment, which partitions the relevant topics into the evaluation criteria given by the end-user, and self-explanation sentiment analysis, which consists of training a self-explanation classification and explanation step based on domain knowledge extracted from the topics-criteria alignment phase. More precisely, the output of the topics-criteria alignment is taken into account in the explanation step to provide user-orientated explanations supporting the prediction of the self-explanation classification. We implement SENN4MCSA by employing BERTopic, manual topicscriteria alignment based on domain experience, and SELFEXPLAIN as concrete methods for doing topic modeling, topics-criteria alignment, and self-explanation classification respectively. We assess the potential advantages of the proposed SENN4MCSA on a TripAdvisor data set.The user-orientated explanations are assessed by 4 sentiment criteria, which are often used to assess the reviews on this specific data set: room quality, value, location, and service. The empirical evidence suggests that SENN4MCSA can provide promising levels of predictive performance, and reliable user-orientated explanations.</div
iEMG-Based Diagnosis of ALS and Myopathy using 1D-CNN
No full textInternational audienceAmyotrophic Lateral Sclerosis (ALS) and myopathy are debilitating neuromuscular disorders that require accurate and timely diagnosis for effective management. Traditional electromyography (EMG)-based diagnostic methods rely on manual interpretation, which is time-consuming and prone to variability. This study proposes an approach that directly classifies EMG signals using a one-dimensional convolutional neural network (1D-CNN) without feature extraction, addressing the limitations of existing methods that depend on handcrafted features and focus primarily on binary classification. The proposed model is evaluated on a publicly available EMG dataset, achieving an overall accuracy of 99.27%, with macro and weighted precision, recall, and F1-scores exceeding 99% across ALS, myopathy, and healthy subjects. Unlike previous approaches that require extensive preprocessing, our method maintains high classification performance while reducing computational complexity, offering a clinically relevant multiclass classification framework. Although our method achieves high classification performance, it also maintains a strong balance between sensitivity and specificity, ensuring reliable and accurate neuromuscular disorder diagnosis, making it a practical tool for clinical applications. Future research will focus on improving model generalizability, expanding dataset diversity, and integrating real-time deployment for enhanced diagnostic utility
Quantification of Lung Stiffness Using Magnetic Resonance Elastography (MRE): Clinical Validation for Smokers
No full textInternational audienceTobacco-related pathologies are the most preventable diseases. The purpose is to provide personalized cartography of smoker lung stiffness using non-irradiating imaging modalities, MRI and MRE (magnetic resonance imaging/elastography). Methods: Thirty-four smokers were divided into five groups distributed with a range of pack-years (PY) of 10. All patients underwent three imaging tests (CT: computed tomography, MRI, MRE) to make possible measurements of lung density, with two modalities (CT, MR), and stiffness. CT lung density was measured using the Hounsfield number. MR density was obtained from a fast gradient echo sequence and validated with an in vitro 3D abdominal phantom. The MRE test was performed with a motion-encoding gradient (Z direction), a spin-echo echo-planar sequence and four offsets. A pneumatic driver (frequency: 50 Hz) was placed on the right lung and four axial phase images were recorded. Post-processing was then performed to record a personalized stiffness cartography. Results: CT density significantly increased in relation to PY, showing denser tissue for the heavy smokers. As MR density acquisition is less accurate than CT, a slight increase in lung density was obtained. MRE tests revealed a significant increase in stiffness with pack-year. Patient-specific lung stiffness showed inhomogeneous distribution of values. Conclusion: MRE could provide a personalized cartography of stiffness for regular uptake of the lung's mechanical behavior in smokers. The stiffness could become a biomarker for preventing future lung disease. Significance: MRE test could be an alternative to CT test for the follow-up of smokers
Mechanosynthesis and Polymerization of Styrene Derivatives Based on Building Blocks of Lignin
No full textInternational audienceGiven the constraints dictated by the environment and the current policies, it is urgent to conceive and developnovel molecular building blocks and materials from bio-sourced platforms in order to compete and replace thoseobtained from petroleum sources (styrene, bisphenol A, etc.). Bio-sourced monomers can be obtained from thetransformation of molecules extracted from five main sources, namely terpenes, carbohydrates, lignin, proteins andlipids from animal, plant, or sea origin. Thanks to the presence of alcohol, acid or amine functional groups, thesebio-sourced molecules can often be polymerized by condensation reaction and yield bio-based polyesters orpolyamides with mechanical properties that can compete with their petroleum-based counterparts.Step-growth polymerization remains therefore the most explored pathway to produce bioplastics nowadays.Recently, the number of publications reporting free-radical chain-reaction polymerization (FRP) of bio-monomershas strongly increased. However, it remains still limited, probably due to the little number of readily availableradically polymerizable bio-sourced structures. In this context, lignin, the second most abundant natural polymericconstituent of wood, accounting for around 20% of the lignocellulosic biomass11 and currently discarded as wasteby the paper industry, appears to be a suitable candidate. Indeed, the controlled degradation of lignin allows toproduce biofuels, precursors for organic synthesis or oligomers that can be applied to prepare functional materials.Among those examples, the compounds obtained from lignin degradation represent a promising alternative todesign bio-sourced polymers displaying interesting thermo-mechanical properties thanks to the presence ofaromatic cycles.In this communication, we report the synthesis of polystyrenes prepared in three steps from vanillin, 4-hydrobenzaldehyde, and syringaldehyde, compounds that can be obtained through lignin depolymerization underoxidative conditions. The synthesis involves the conversion of these biosourced platforms into polymerizablestyrene derivatives through a methylation of the hydroxyl group followed by an olefination of the aldehyde function.The monomers were first synthesized under conventional conditions using solvents. Then, the synthesis wasimproved from the sustainability point of view by using a ball mill under solventless conditions, generating muchless waste in the process. Mechanochemistry has successfully been transposed from material sciences to organicchemistry in the past decades. Compared to conventional procedures and solvent-based reactions, it usesmechanical energy to induce a chemical reaction in solvent-free conditions and can afford shorter reaction times,higher yields and more sustainable process.The three monomers were then converted into biosourced homopolymers through free radical polymerization inbulk, providing functional polystyrene derivatives with thermal properties comparable to those of common petrosourced polystyrene
An Upgraded FOS/TAC Titration Model Integrating Phosphate Effects for Accurate Assessments of Volatile Fatty Acids and Alkalinity in Anaerobic Media
No full textInternational audienceThe accurate determination of volatile fatty acids (VFAs) and total alkalinity (TAC, mostly carried by bicarbonate ions) is critical for operating anaerobic digesters. The FOS/TAC titration method developed by Nordmann is widely used due to its simplicity and affordability. This method has known limitations in dosing VFAs and TAC, since the presence of one interferes with the determination of the other, especially at higher VFA or bicarbonate concentrations. This study builds upon our prior research in 2021 by integrating the influence of phosphate (H2PO4−/HPO42−) into numerical models correcting FOS/TAC titration results. A Scilab-based program was used to assess the impact of phosphate on titration results, revealing significant biases at lower concentrations. A revised multivariate regression formula was developed, incorporating phosphate effects, and demonstrating superior accuracy. The mean absolute percentage errors (MAPE) for TAC and VFA estimation were reduced to less than 0.3%. The model maintains compatibility with standard Nordmann’s titration protocols and equipment while significantly improving reliability. These findings highlight the necessity of considering phosphate interference in FOS/TAC titration, particularly in AD systems with variable buffering conditions. The proposed correction model enhances process monitoring and control, providing a more robust tool for both research and industrial practice in anaerobic digestion
Conducted Emission Mitigation for Solid-State Transformer using Modulation and Multiple Configuration Techniques
No full textInternational audienceSolid-state transformers have several advantages over conventional 50 Hz transformers. However, the high switching speed of power electronic devices generates electromagnetic interference (EMI) that can cause malfunction. Conducted emission mitigation techniques for solid-state transformers using appropriate modulations and configurations in the QSPICE simulator are analysed in time and frequency domains. Common-mode (CM) emission and reactive power can be improved using proper modulation and configuration techniques for single and multiple solid-state transformers. The single-phase shift modulation has the lowest CM emission and the highest reactive power. In contrast, the triple-phase shift modulation demonstrates the highest CM emission and the lowest reactive power among the various modulation strategies. The CM emission of single and multiple solidstate transformers with dV/dt cancellation capability is below the IEC 62041-2020 standard limits
Engineering and Value Change
No full textThis volume consists of chapters derived from the best papers submitted to the 2023 Forum on Philosophy, Engineering, and Technology (fPET) meeting that took place in April 2023 at Delft University of Technology. Topics and fields covered within the volume include applied ethics, meta-ethics, value theory, education, responsible innovation, political philosophy and value-sensitive design. The contributors present research that addresses the challenges of engineering in a changing world. This text is of interest to students and researchers working in the fields of engineering and philosophy.Ce volume rassemblent des chapitres issus des meilleurs communications présentée à l'occasion du Forum sur la philosophie, l'ingénierie et la technologie (fPET) qui s'est tenue en avril 2023 à l'Université de technologie de Delft. Les sujets et les domaines couverts dans cet ouvrage comprennent l'éthique appliquée, la méta-éthique, la théorie des valeurs, l'éducation, l'innovation responsable, la philosophie politique et le design sensible aux valeurs. Les contributeurs et contributrices présentent des recherches qui abordent les défis de l'ingénierie dans un monde en mutation. Ce texte s'adresse aux étudiant·es et aux chercheur·es travaillant dans les domaines de l'ingénierie et de la philosophie
Identifier les témoins du recyclage des alliages ferreux par l’expérimentation : le cas de l’agglomération à chaud
No full textAn Ontology-Based Bridge between Algorithmic Thinking and Programming
No full textInternational audienceThis work uses an ontology approach to address the fundamental challenge of knowledge transfer between algorithmic thinking and the implementation of programming languages in computer science education. Traditional approaches create substantial learning gaps when students transition from structured pedagogical languages such as Pascal to contemporary languages such as Python. Through an ontology-driven approach, conceptual relationships are captured in structured semantic models, making implicit language relationships explicit and navigable. The research develops three interconnected ontologies covering general algorithms and data structure concepts, Python syntax, and Pascal syntax, enabling systematic comparison and knowledge transfer between programming paradigms