HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Deep Learning Detection and Classification of Red Blood Cells: Towards a Universal Dataset
No full textWe evaluate emerging machine learning models for pattern recognition, focusing on the YOLOv11 architecture for detecting and classifying red blood cell shapes. Our analysis targets two characteristic morphologies observed under flow: slipper and parachute . A key challenge in this task is the development of a robust and diverse dataset. To address this, we employ synthetic image generation using a cut-and-paste approach, introducing variations in cell overlap and arrangements of microfluidic channels disposition to alleviate data scarcity and reduce cross-dataset bias. We generate these datasets with U-Net and Cellpose segmentation models, and rigorously assess YOLOv11 performance on two benchmarks: (i) a controlled dataset for evaluating classification accuracy, and (ii) a challenging, visually heterogeneous dataset for assessing generalization. Results show that the model achieves high precision for distinct cell types in controlled settings, but exhibits reduced performance on the unseen dataset, highlighting a trade-off between specialized accuracy and broad applicability in complex microscopy scenarios
Optimising Supercritical Carbon Dioxide Extraction of Rosmarinic Acid from Rosmarinus officinalis L. and Enhancing Yield Through Soxhlet Coupling
No full textInternational audienceRosmarinic acid (RA) is a bioactive phenolic compound prevalent in various medicinal plants, renowned for its significant pharmacological properties. This study aims to optimise the extraction conditions of this compound from Rosmarinus officinalis L. using the response surface methodology (RSM) with a three-variable, three-level Box–Behnken design. Optimising the parameters for supercritical CO2 (scCO2) extraction focused on pressure (150 to 350 bar), temperature (40 to 80 °C), and co-solvent weight percentage (5 to 15% ethanol), evaluating their impact on overall yield and RA content. The optimal conditions determined were a pressure of 150 bar, a temperature of 80 °C, and 15% ethanol, yielding a total extract of 21.86 ± 1.55%, with an RA content of 3.43 ± 0.13 mg/g dry matter (DM). Scanning electron microscopy revealed that the scCO2 treatment induced microcracks on the surface of the rosemary powder, enhancing the fluid’s ability to penetrate the plant matrix. By employing the combined scCO2-Soxhlet method, the RA content increased to 5.78 mg/g DM. Furthermore, the final extract obtained using the Soxhlet post-scCO2 treatment contained only trace amounts of carnosic acid (0.38 ± 0.10 mg/g DM) and carnosol (0.38 ± 0.20 mg/g DM), compared to the crude extract obtained solely with Soxhlet, which exhibited significantly higher concentrations of 8.45 ± 2.98 mg/g DM of carnosol and 16.67 ± 0.94 mg/g DM of carnosic acid. This work highlighted an innovative extraction strategy based on the coupling of scCO2 and Soxhlet, which significantly increased RA content while reducing concentrations of other compounds such as CA and CAR. This approach makes it possible to produce RA-enriched extracts, offering considerable potential for future large-scale applications and commercialisation
An original approach to generate periodic Representative Volume Elements with anisotropic heterogeneous microstructure: application to skeletal muscle
No full textInternational audienceSkeletal muscle is an organ whose hierarchical, multiscale structure greatly influences the overall mechanical response. Complementary to mechanical experiments, finite element modeling is increasingly used to study the influence of its constituents across different scales. To develop such a multiscale model, particular attention must be paid not only to the scale transition, but also to the definition of the structure and its mechanical behavior at different scales (macroscopic, microscopic, submicron). One of the most effective approaches is to define a Representative Volume Element (RVE) including smaller scale components and their respective mechanical behavior laws, likely to be altered through pathologies.In this study, an original approach for periodic RVE generation dedicated to multiscale modeling of the skeletal muscle is proposed. From optical microscopy cross-section images of mouse skeletal muscle and single fiber experiments, the RVE integrates parameters related to fiber type distribution, geometric and mechanical characteristics. The key features of this geometry are spatial periodicity, rounded edges and inclusion of experimentally measured probabilistic distributions of the extracellular matrix (ECM), slow and fast muscle fibers. Smooth variation of the mechanical properties between the muscle fibers and the ECM are implemented to avoid unrealistic and purely numerical stress accumulation at these interfaces through the definition of transition layers between the different microcomponents. By the inclusion of custom geometrical and material features, this original model allows the multiscale and multicomponent analysis of different muscle phenotypes and can also be used for other heterogeneous anisotropic materials such as fiber reinforced composites
Machine Learning-Based Netmob25 Mobility Model for Network Simulators
No full textInternational audienceUser mobility pattern has a great impact on mobile network performances. In network simulation, researchers need tools to generate realistic user trajectories. In this paper, we propose to use Variational Autoencoders to capture the characteristics of the Netmob25 mobility dataset. This machine learning model is then used to build a new mobility model in the NS3 network simulator to generate a number of trajectories in different types as needed
Effect of Friction Material on Vehicle Brake Particle Emissions
No full textInternational audienceThis study focuses on the influence of different brake pad formulations on the emission of particulate matter coming from car braking systems. The brake particles were characterised using a pin-on-disc bench and some particle measuring devices such as CPC, APS, SMPS and a PM10 sampling unit. Seven samples of brake pad materials of different compositions (1 NAO and 6 Low Steel) were tested against grey cast iron discs. The results presented in this work show differences in particle number concentration and PM10 emission factor between the different friction materials tested. Three friction materials, LS04, LS06 and NAO01, reduce particle number emissions by up to 71% and PM10 emissions by up to 57%. On the other hand, this reduction in particulate emissions goes along with a reduction of 20% to 27% in the coefficient of friction. The microscopic analyses carried out on the test parts (pins and discs) show differences between the most emissive and the least emissive friction pairs, which may explain the differences observed in particle emissions. Correlations between the emission of particles and the concentration of iron of the PM10, as well as the steel fibre content in the formulas, were found
Detection of Low-Velocity Impact Damage in Woven-Fabric Reinforced Thermoplastic Composite Laminates by Deep-Learning Classification Trained on Terahertz-Imaging Data
No full textNational audienceTerahertz (THz) imaging is gaining attention as a nondestructive testing technique for assessing damage due to its high axial resolution and nonionizing nature, presenting a promising alternative to conventional methods such as ultrasound and X-ray imaging. Its practical implementation, however, remains limited by the reliance on expert interpretation and the frequent need for validation using supplementary techniques such as X-ray microcomputed tomography (µCT), particularly for complex damage modes. This study focuses on woven-fabric-reinforced thermoplastic composites subjected to low-velocity impact, which typically causes barely visible impact damage (BVID). The damage is subtle yet critical, potentially leading to failure under subsequent loading. The multilayered and spatially distributed characteristics of BVID make it especially challenging to identify. To overcome these challenges, this work integrates deep learning with pulsed THz time-of-flight tomography (TOFT) imaging to enable automated damage detection in composite laminates. In contrast to existing research that mainly targets delamination using A- or C-scan data, this study emphasizes the detection of low-velocity impact damage by leveraging THz B-scans, which offer nondestructive depth-resolved cross-sectional imaging. The training dataset is labeled by correlating THz TOFT scans with X-ray CT images used as ground truth. A transfer learning approach, based on convolutional neural network (CNN) architectures, is employed for binary classification to distinguish damaged from undamaged regions. The resulting classifier achieves over 95 % accuracy, demonstrating the viability of this method for industrial applications such as quality assurance and in-service inspection of composite structures
Brassinosteroid Synthesis and Perception Differently Regulate Phytohormone Networks in Arabidopsis thaliana
No full textInternational audienceBrassinosteroids (BRs) are essential regulators of plant development and stress responses, but the distinct contributions of BR biosynthesis and signaling to hormonal crosstalk remain poorly defined. Here, we investigated the effects of the BR biosynthesis inhibitor brassinazole (BRZ) and the BR-insensitive mutant bri1-6 on endogenous phytohormone profiles in Arabidopsis thaliana. Using multivariate analysis and targeted hormone quantification, we show that BRZ treatment and BRI1 disruption alter hormone balance through partially overlapping but mechanistically distinct pathways. Principal component analysis (PCA) and hierarchical clustering revealed that BRZ and the bri1-6 mutation do not phenocopy each other and that BRZ still alters hormone profiles even in the bri1-6 mutant, suggesting potential BRI1-independent effects. Both BRZ treatment and the bri1-6 mutation tend to influence cytokinins and auxin conjugates divergently. On the contrary, their effects on stress-related hormones converge: BRZ decreases salicylic acid (SA), jasmonic acid (JA), and abscisic acid (ABA) in the WT leaves; similarly, bri1-6 mutants show reduced SA, JA, and ABA. These results indicate that BR biosynthesis and BRI1-mediated perception may contribute independently to hormonal reprogramming, with BRZ eliciting additional effects, possibly via metabolic feedback, compensatory signaling, or off-target action. Hormone correlation analyses revealed conserved co-regulation clusters that reflect underlying regulatory modules. Altogether, our findings provide evidence for a partial uncoupling of BR levels and BR signaling and illustrate how BR pathways intersect with broader hormone networks to coordinate growth and stress responses
Analyse multi-échelle de la microstructure de l’alliage Fe-49Co-2V élaboré par fusion laser sur lit de poudre
No full textInternational audienceLes rotors et stators des moteurs électriques sont usuellement fabriqués par empilement de tôles ferromagnétiques laminées et isolées électriquement. Cependant, cette approche limite la liberté de conception des composants. La fabrication additive, via le procédé de fusion laser sur lit de poudre (L-PBF), peut répondre à cette limitation. Néanmoins, en l’absence de couches isolantes, le composant présente des pertes magnétiques élevées, qui augmentent avec la fréquence [1]. L’alliage Fe-49Co-2V (m%) est particulièrement prometteur pour augmenter la densité de puissance des moteurs électriques. Pour obtenir des propriétés magnétiques optimales (aimantation et perméabilité élevées, faible coercitivité), il est essentiel de favoriser la formation de la phase B2 et de maximiser la taille des grains [2]. Ces objectifs peuvent être atteints par l’optimisation des paramètres de lasage et les traitements thermiques [3]. Nos travaux de thèse visent à réduire les pertes par courants de Foucault dans l’alliage Fe-49Co-2V en intégrant des couches isolantes, telles que des fentes d’air, via le procédé L-PBF. Par ailleurs, nous cherchons à comprendre le lien entre le procédé, la microstructure et les propriétés magnéto-électriques. L’un des premiers objectifs est de maîtriser et d’optimiser la microstructure du matériau massif. La poudre utilisée est caractérisée en détail et les paramètres de lasage sont optimisés pour maximiser la densité des pièces. Une analyse microstructurale multi-échelles permet de caractériser les bains de fusion, les tailles et textures des grains (MEB, EBSD) (Fig.1a) et la présence de nano-précipités dans la phase B2 (MET-EDX et diffraction) (Fig.1b). L’effet de différents traitements thermiques sur les microstructures sera également présenté, ainsi que les premières propriétés magnéto-électriques
Intégration de l'impact des technologies de l'industrie 4.0 dans les politiques de maintenance pour une maintenance soutenable : feuille de route et cas d'étude
No full textInternational audienceMaintenance decision-making has traditionally focused on economic criteria, yet the growing demand for carbon neutrality highlights the need to address all three dimensions of sustainability (economic, environmental, and social) within manufacturing industries. Although Industry 4.0 (I4.0) enabling technologies are widely recognized for their potential benefits, their full sustainability impacts remain poorly understood. Existing studies often emphasize their positive contributions but lack precise quantification of both their positive and negative effects. Moreover, these analyses tend to focus exclusively on the use phase, neglecting impacts during manufacturing and end-of-life stages. This article proposes a structured roadmap for evaluating the lifecycle impact of I4.0 technologies on maintenance policies. By considering multiple scenarios, this approach quantifies their effects across all dimensions of sustain- ability, ensuring that the benefits realized during use outweigh the negative impacts from manufacturing and disposal. To illustrate its applicability, a preliminary use case is presented using a vibration test bench equipped with IoT sensors. Looking ahead, these sensors are set to generate fault data under varying conditions, which will be used to test maintenance scenarios. Additionally, as outlined in the roadmap, a life cycle assessment (LCA) is planned for the sensor to provide a comprehensive assessment of its sustainability impact. This case study serves to demonstrate the roadmap’s relevance and its potential to support sustainable maintenance decision-making, laying the foundation for integrating I4.0 enabling technologies into maintenance strategies while avoiding undesirable rebound effects that could compromise sustainability goals.La prise de décision dans la maintenance s'est traditionnellement concentrée sur des critères économiques. Cependant, dans la perspective d'atteindre la neutralité carbone d'ici 2050, cette prise de décision doit considérer l'ensemble des dimensions de la soutenabilité (économique, environnementale et sociale). Pour atteindre cet objectif, l'industrie 4.0 apparaît comme un levier potentiel majeur. Or, si les technologies de l'Industrie 4.0 (IA, IoT, réalité augmentée, infonuagique, etc.) sont reconnues pour leurs bénéfices potentiels dans la maintenance tels que l'amélioration de la performance et la réduction des temps d'arrêt, leurs impacts globaux sur la soutenabilité restent encore peu quantifiés. Les études existantes se concentrent souvent sur les avantages liés à la phase d'utilisation, en négligeant les effets négatifs potentiels sur tout le cycle de vie, de la fabrication à la fin de vie, ainsi que le risque d'effets rebond.Cet article propose une feuille de route structurée pour évaluer, de façon précise et globale, l'impact de ces technologies dans la maintenance dans une perspective de maintenance soutenable. En tenant compte de multiples scénarios de maintenance, cette approche quantifie leurs effets dans toutes les dimensions de la soutenabilité, en veillant à ce que les bénéfices constatés lors de l'utilisation l'emportent sur les effets négatifs associés à la fabrication, à l'utilisation et à la gestion en fin de vie de ces technologies. Pour illustrer cette feuille de route, un premier cas d'étude est présenté, basé sur un simulateur de défauts mécaniques équipé de capteurs IoT. Ces capteurs génèrent des données de défaillances dans des conditions variées, permettant une évaluation complète de l'ensemble du cycle de vie de ces technologies. Plusieurs scénarios de maintenance seront construits et testés, en tenant compte de l'utilisation ou non de ces capteurs, dans le but de démontrer la pertinence et l'applicabilité concrète de la feuille de route proposée
Les Techniques ont-elles une patrie ?: Retour sur le nationalisme et l'ethnocentrisme dans l'histoire des techniques
No full textInternational audienceL'histoire des techniques n'a pas échappé aux réflexes ethnocentriques, voire nationalistes, mais elle semble s'y être attachée de manière particulièrement durable. Elle a même longtemps constitué l'un des domaines par excellence dans lequel s'est exprimé l'indexation des territoires selon leur degré d'avancement ou de retard sur la trajectoire euro-centrée du progrès. La multiplication des revendications de cultures techniques nationales a longtemps cadré l'historiographie. Peu à peu, cependant, les recherches historiennes ont mis au jour les impasses de ces lectures restrictives et politiques des pratiques techniques. Cet ouvrage propose un premier bilan réflexif sur les effets des approches nationales et ethnocentrées et sur les conditions de leur dépassement. L'enjeu est d'importance : il s'agit non seulement de sortir du récit piégé d'une sociodicée nationale par les techniques, mais également d'engager une analyse plurivoque des ancrages pratiques, des complexions artefactuelles, des circulations de savoir-faire. Pour ce faire, l'ouvrage éclaire les logiques historiques et politiques profondes qui enracinent le nationalisme technicien et sa performativité dominatrice, prenant part à une dynamique internationale très active en histoire globale des techniques. Les contributions abordent des thèmes comme l'alimentation, l'agriculture, l'habillement, les convertisseurs énergétiques ou encore la danse, mettant en lumière la diversité caractéristique de l'histoire des techniques à travers le monde. Variant les angles, les approches et les géographies, l'ouvrage adopte un large empan chronologique allant du Moyen Âge à nos jours. Ce livre s'adresse à toute personne curieuse de comprendre comment les techniques ont pu servir d'instrument à la revendication identitaire et nationale