HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Test Setup for Measuring the Impact of Gate Stress on a GaN-HEMT's Output Characteristic
No full textInternational audiencePrecisely estimating conduction losses requires knowledge of the device's output characteristic in the ohmic region. For GaN-HEMTs, it dynamically depends on time, temperature, and applied biases. To study the impact of gate bias over time, a transient measurement approach is applied, that can acquire the start of the output characteristic with a single pulse. This ensures neglectable impact of temperature changes and trapping state on the single measurement. In addition, the short acquisition time allows studying the impact of the bias in the milliseconds range. Therefore, the measurement pulses are altered with stress intervals. While reference measurements for a traditional Si-MOSFET under constant gate bias show almost no change in the output characteristic over time, test results for a commercial GaN-HEMT reveal changes for low gate voltages. This demonstrates that the user should at least be aware of such trapping problems effecting measurements, which may lead to differences in characterization and application
Synthetic Peptide Antibodies via a Rational Approach Based on Disulfide‐Stabilized α‐Helical Peptides, for the Recognition of the Intrinsically Disordered Protein NUPR1
No full textInternational audienceAbstract Nuclear Protein 1 (NUPR1) is a ubiquitous protein playing an important role in cancer and acute kidney injury. Its specific targeting by natural or synthetic antibodies like molecularly imprinted polymers (MIPs), is therefore of interest. NUPR1 is an intrinsically disordered protein (IDP), such that it displays a high degree of flexibility and an unstable secondary/tertiary structure, resulting in a continuous fluctuation of its conformation in the free state. These characteristics are not in favor of the creation of homogeneous binding sites during molecular imprinting, so that imprinting using peptide epitopes is investigated. Based on an in silico rational approach, two α‐helices from the model structure of NUPR1, as predicted by AlPhaFold, are selected. Two cysteine residues are added at both ends of the epitopes to form a disulfide bond, which provides high stability to the α‐helix. The template peptides possess the same 3D structure as the epitope fragments in NUPR1. Consequently, they are effective in producing MIP nanogels that cross‐react with high affinity (IC 50 1 n m ) only with NUPR1. The work indicates that α‐helices, besides the preferred flexible loops, can be considered as viable template epitopes for MIPs, opening new opportunities for the obtention of selective antipeptide MIP nanogels for IDPs
Management and valorization of storage in electrical networks
No full textInternational audienceThis presentation addresses the management and valorization of energy storage in electrical grids, highlighting its key role in integrating renewable energy sources. In the face of intermittency in solar and wind power, storage smooths production fluctuations, ensures network stability, and enhances electric system flexibility. Case studies are presented before concluding with environmental issues, particularly the impact of material extraction and carbon footprint, while offering perspectives on sustainable solutions to support the energy transition
Influence of the flammable cloud geometry on the gas explosion effects
No full textInternational audienceIn the context of industrial large cloud explosions such as the Buncefield accident, it is commonly accepted that the flammable cloud spreads over a large area on the ground but has a limited height. This can, therefore, be considered as the limiting dimension of the cloud. In this work at a small scale, Leyer highlighted the influence of the limited height of the flammable cloud in the case of cylindrical cloud explosions. Without prejudging the combustion mechanisms, this paper aims to present the influence of this limited dimension of the flammable cloud on the flame dynamics to assess more precisely the overpressure distances of a UVCE by better determining the energy involved in the explosion. The analysis is based on comparing the flame position over time from the fast video films and the overpressure signals recorded in the flammable clouds. The explosions examined are methane and hydrogen-free jets, methane jets interacting with the ground and rows of obstacles, and large propane clouds obstructed by rows of tree trunks
Cell Membrane Vesicle Camouflaged Artificial Cells
No full textInternational audienceThe integration of artificial and mammalian cells into semi-synthetic aggregates remains a challenge in bottom-up synthetic biology. Here, the fabrication of cell membrane vesicles (CMV) from HepG2 cells and their use as a coating for alginate microgels to produce camouflaged artificial cells (ACs) is demonstrated. These ACs are used for the assembly of either synthetic aggregates or semi-synthetic aggregates. In the first case, a predator-defendant and a liver-like synthetic aggregates are investigated, showing promising initial steps toward complex synthetic aggregates. In the other case, the camouflaged ACs show enhanced integration with HepG2 cells. The encapsulation of a reactive oxygen species (ROS) scavenger artificial enzyme in the ACs shows protection against tert-butyl hydroperoxide in terms of HepG2 cell viability, proliferation, and mitochondrial health in semi-synthetic aggregates. Taken together, this effort is a substantial step forward in combining mammalian cells and ACs in the same aggregate where the latter act as support unit
Etudier les imaginaires pluriels du numérique et de l’hydrogène vert pour penser la transition écologique en monde d’ingénieurs.: Comprendre les tensions entre innovation technologique et sobriété
No full textInternational audienceLa transition écologique est paradoxale par bien des aspects. Si elle s’est affirmée depuis plusieurs années comme un impératif nécessaire au maintien de l’habitabilité de la terre, les technologies promues dans ce cadre sont également contestées pour leurs conséquences sur l’environnement, qu’il s’agisse d’extractivisme (Pitron 2019), d’impact sur la biodiversité (Oiry 2019) ou de reproduction de rapports coloniaux (Blanc 2020). Ce paradoxe est à l’origine de nombreux débats d’experts ayant des visions différentes de la manière de mener les transitions, et notamment de la place que ces techniques prennent dans les stratégies de lutte contre les désordres environnementaux. Ces visions différentes s’ancrent dans des imaginaires (Sebbah et al. 2023) et des rationalités multiples (Hecht 1998; Jasanoff and Kim 2009) dont la pluralité a parfois été sous-estimée par les sciences sociales au profit de lectures globalisantes des mondes experts (comme le grand « lobby du nucléaire » par exemple). Cette communication vise à montrer la pluralité des imaginaires à l’œuvre dans deux mondes experts que sont ceux des innovateurs du numérique et des industriels de l’hydrogène « vert ». Dans ces deux domaines, le modèle dominant promu par de grands industriels souvent transnationaux entre en concurrence avec des initiatives portées par d’autres acteurs (citoyens, locaux, associatifs etc.). Or, étudier ces différents imaginaires prend tout son sens en école d’ingénieurs, au sein de la Fabrique de la Pensée Critique, pour outiller les mondes ingénieurs à envisager une pluralité de futurs possibles et penser ensemble les organisations sociales que ces imaginaires contribuent à façonner
Preliminary results of YOLO-Based Recognition of Surface Markers in Aquatic Environments for Equine Motion Analysis Applications
No full textInternational audienc
Imagerie cérébrale dans le diagnostic de la maladie d'Alzheimer
No full textAlzheimer's disease (AD) is a progressive neurodegenerative pathology that is the most common form of dementia. Early diagnosis is a crucial factor in optimising care for sufferers and slowing the progression of symptoms. This dissertation looks at the different brain imaging modalities used to diagnose AD. Structural magnetic resonance imaging (MRI) and positron emission tomography (PET) are the two main imaging methods used to identify specific biomarkers of the disease, such as brain atrophy, the deposition of amyloid plaques and the accumulation of tau protein. The integration of imaging data with biological and clinical approaches improves the accuracy of diagnosis, particularly at the early stages of the disease. However, challenges remain, particularly concerning the cost and accessibility of these technologies. Finally, this thesis highlights the prospects offered by technological advances, such as electroencephalogram (EEG) and retinal imaging, which could enable early diagnosis to improve patient’s quality of life.La maladie d’Alzheimer (MA) est une pathologie neurodégénérative progressive qui constitue la forme la plus courante de démence. Son diagnostic précoce représente un enjeu crucial pour optimiser la prise en charge des personnes atteintes et ralentir la progression des symptômes. Ce mémoire traite des différentes modalités d’imagerie cérébrale utilisées dans le diagnostic de la MA. Parmi celles-ci, L’imagerie par résonance magnétique (IRM) structurelle et la tomographie par émission de positons (TEP) sont les deux principales méthodes d’imagerie utilisées pour identifier les biomarqueurs spécifiques de la maladie, tels que l’atrophie du cerveau, le dépôt de plaques amyloïdes et l’accumulation de protéine tau. L’intégration des données d’imagerie avec des approches biologiques et cliniques renforce la précision du diagnostic, en particulier au stade précoce de la maladie. Cependant, des défis persistent, notamment concernant le coût et l’accessibilité de ces technologies. Enfin, ce mémoire met en évidence les perspectives offertes par les avancées technologiques, comme l’électroencéphalogramme (EEG) ou l’imagerie rétinienne, qui pourraient permettre un diagnostic précoce afin d’améliorer la qualité de vie des patients
Extraction de relations multi-étiquettes en utilisant des modèles pré-entraînés et des couches de Transformer
No full textNational audienceWe present in this paper the BTransformer18 model, a deep learning architecture designed for multi-label relation extraction in French texts. Our approach combines the contextual representation capabilities of pre-trained language models from the BERT family, such as BERT, RoBERTa, and their French counterparts CamemBERT and FlauBERT, with the power of Transformer encoders to capture long-range dependencies between tokens. Experiments conducted on the TextMine'25 challenge dataset show that our model achieves superior performance, particularly when using CamemBERT-Large, reaching an F1-macro score of 0.654, outperforming the results obtained with FlauBERT-Large. These findings demonstrate the effectiveness of our approach for automatic extraction of complex relations in intelligence reports.Nous présentons dans cet article le modèle BTransformer18, une architecture d'apprentissage profond conçue pour l'extraction de relations multi-étiquettes dans des textes en français. Notre approche combine les capacités de représentation contextuelle de modèles de langage pré-entraînés de la famille BERT, tels que BERT, RoBERTa, ainsi que leurs versions francophones Camem-BERT et FlauBERT, avec la puissance des encodeurs Transformer pour capturer les dépendances à long terme entre les tokens. Les expérimentations menées sur le jeu de données du défi TextMine'25 montrent que notre modèle atteint des performances supérieures, en particulier en utilisant CamemBERT-Large, avec un score F1-macro de 0,654, surpassant les résultats obtenus avec FlauBERT-Large. Ces résultats démontrent l'efficacité de notre approche pour l'extraction automatique de relations complexes dans des rapports de renseignement