Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
System-level monitoring and diagnosis of starvation faults in solid oxide electrolyzers
No full textInternational audienceThis paper presents a model-based methodology for real-time fault detection in Solid Oxide Cells (SOCs), focusing on fuel-side starvation in electrolysis mode. Steam starvation occurs when consumption exceeds supply, while hydrogen starvation arises from an insufficient inlet fraction; both risking electrode degradation. With limited measurements, a lumped model with a square-root Unscented Kalman Filter is adopted to estimate gas partial pressures, stack temperature, and deviations from the nominal inlet fuel flow rate, using stack voltage, outlet air temperature, and outlet hydrogen flow rate measurements. Faults are flagged when the estimated states exceed adaptive thresholds derived from nominal predictions via moving-window filter. For quantification, the observer is augmented with two states capturing deviations in inlet steam and hydrogen flow rates. Furthermore, the total resistance is calculated: it increases under steam starvation and remains nearly unchanged under hydrogen starvation. Experimental validation confirms accurate, real-time detection and identification, helping achieve safer SOEC operatio
Towards a Smarter Homophone Correction Tool: A Case Study in Khmer Writing
No full textInternational audienceHomophone errors are a common challenge in written communication, affecting both high-resource languages, such as English, and low-resource languages, such as Khmer. These errors are often difficult to detect because they require contextual understanding rather than simple spelling correction. While existing spelling correction tools enhance text accuracy, they do little to improve users’ long-term writing skills, often leading to an over-reliance on automated corrections. This study aims to bridge this gap by investigating the challenges of homophone usage, specifically among Khmer users, and proposing a foundational theoretical blueprint for future solution development. Through a questionnaire-based survey, we analyzed the prevalence of homophone errors and their impact on Khmer speakers. Additionally, we conducted an experimental study using Typing Tracker, where participants transcribed audio-recorded articles to determine their ability to correctly use homophones in context. Based on these insights, we introduce Sor-Ser, an innovative conceptual approach that integrates Natural Language Processing (NLP) with Learning Analytics (LA) techniques. This preliminary framework provides a foundation for addressing homophone errors while enhancing writing proficiency. By addressing both error correction and skill development, Sor-Ser provides a potential pathway for improving Khmer writing accuracy while fostering long-term proficiency and confidence
Kink-bands as drivers of hygroscopic response in flax fibres: A multi-scale investigation
No full textInternational audienceFlax fibres are increasingly explored as sustainable reinforcements in high-performance composites due to their remarkable stiffness and strength. However, their hygroscopic nature and complex hierarchical organisation pose challenges for dimensional stability and long-term durability. Among the intrinsic microstructural heterogeneities, kink-bands, localised deformations associated with cellulose misalignment and increased porosity, may affect water uptake and swelling. Despite their potential impact, the effect of kink-bands on water sorption and moisture-induced structural changes has not been directly investigated. This study investigates the influence of kink-band density on the hygroscopic behaviour of flax fibres using an integrated multi-scale approach combining Dynamic Vapour Sorption (DVS), Environmental Scanning Electron Microscopy (ESEM), X-ray microtomography, and solid-state Nuclear Magnetic Resonance (ss-NMR). Three flax fibre batches with distinct kink-band frequencies were prepared through controlled mechanical processing. DVS results revealed no significant difference in overall water sorption capacity between batches, indicating that kink-bands do not markedly affect global hygroscopicity. However, ESEM and X-ray tomography analyses highlighted heterogeneous swelling behaviour at the bundle and fibre scales, with kink-band regions exhibiting lower expansion than intact zones, suggesting a compensating effect of internal pores on local deformation. Solid-state NMR analyses revealed that, although fibril dimensions and overall crystallinity were unchanged, fibres with higher kink-band content exhibited increased water accessibility and subtle nanoscale polymer reorganisation. This suggests that kink-bands promote local water penetration and interactions with surrounding non-cellulosic components, without altering overall bulk hygroscopicity. Together, these results indicate that kink-bands modulate local moisture dynamics, providing new insights into the structural–functional relationships governing flax fibre behaviour in humid environments. They are also relevant for natural fibre composites, as understanding how kink-bands influence local moisture dynamics can inform predictions of dimensional stability, fibre–matrix interactions, and long-term durability
Édition numérique du Dictionnaire universel françois et latin, vulgairement appellé Dictionnaire de Trévoux
No full textCette plateforme propose une interface de consultation numérique du Dictionnaire universel françois et latin, vulgairement appellé Dictionnaire de Trévoux
Breaking the 3D Dataset Bottleneck: Fast Scalable Generation of Aligned 3D Assets from Scratch for Category 6D Pose Estimation and Robotic Grasping
No full textInternational audienceWhile 2D vision has been revolutionized by large-scale datasets like ImageNet, 3D vision remains constrained by the scarcity of high-quality, canonically aligned data. We introduce the first scalable, automated framework that generates complete category-level 6D pose datasets directly from text prompts, bypassing the need for existing 3D assets. Our method overcomes key challenges by: (1) ensuring reliable, scalable asset generation via a controlled text-to-image-to-3D pipeline; (2) enforcing built-in canonical alignment through depth-conditioned generation, achieving a 96\% pose consistency rate; and (3) enabling large-scale 6D annotation via mixed reality rendering. The pipeline produces high-quality, aligned 3D meshes in under 3 minutes per object—a 5–20 speedup over traditional scanning. We generate over 1,000 instances for each of the 153 categories in the Omni6Dpose benchmark, culminating in 153,000 aligned meshes—a >40 increase in instances per category over previous aligned real-world datasets. Extensive evaluation demonstrates competitive zero-shot sim2real transfer on the NOCS 6D pose benchmark and superior robotic grasping performance in both simulation and real-world zero-shot transfer, where aligned meshes prove essential for success. We release the largest publicly available aligned 3D mesh dataset, largest category-level 6D pose dataset, grasping simulation environments, and open-source pipeline, providing a critical step toward foundation models for 3D understanding and enabling efficient, unlimited generation of task-specific 3D data from scratch
DEFORMATION-AWARE SIMULATOR FOR HANDHELD ULTRASOUND IMAGING
No full textInternational audienceRealistic ultrasound simulation is essential for training and algorithm development. However, most existing methods generate single-frame images from thin slices of the probe's field of view, neglecting tissue deformation and spatial continuity across frames. We propose a physics-based simulation pipeline, fully compatible with any scatterer-based ultrasound simulator, which employs finite element modeling to update local scatterer maps under large-scale stability constraints, given the probe trajectory and tissue geometry. The framework produces continuous and anatomically accurate ultrasound sequences with physically consistent speckle evolution, providing a practical tool for training, visualization, and the development of freehand ultrasound reconstruction algorithms.</div
Habiter Babel: Un manifeste pour une Ingénierie du Sens responsable
No full textInternational audienceThis book deconstructs the illusions of semantic automation, AI "understanding," and universal ontologies. Core thesis: meaning is neither computable, nor storable, nor transferable.For data architects, AI practitioners, enterprise architects, and systems engineers confronting the failure of "well-designed" solutions.Topics: Semantic interoperability, knowledge graphs, ontology engineering, AI language models, meaning governance, federated systems, enterprise architecture, complex systems.Format: PDF ebook, 65 pages, EnglishAuthor: Dr. Nicolas Figay - Expert in interoperability and enterprise architectures (Airbus Defence and Space) - LIRIS UMR 5205 CNRS / INSA Lyon / Université Claude Bernard Lyon 1Not a manual. A manifesto. No turnkey solutions—only conceptual lucidity for those who refuse to let technology decide meaning in their place.Ce livre n’est ni un guide pratique, ni un manuel sur l’IA, les ontologies ou les knowledge graphs.Habiter Babel est un essai-manifeste sur les confusions conceptuelles qui traversent aujourd’hui les discours sur :la sémantique,l’interopérabilité,l’intelligence artificielle,et les systèmes dits « intelligents ».La thèse est simple, mais exigeante :Le sens n’est ni calculable, ni stockable, ni transférable.À partir de cette affirmation, le livre déconstruit :les promesses d’unification sémantique,les fantasmes d’automatisation du sens,la naturalisation abusive des modèles et des ontologies.Il propose une autre posture :non pas résoudre Babel, mais l’habiter —en assumant la pluralité des mondes, la nécessité de la traduction, et la responsabilité humaine dans les systèmes socio-techniques.À qui s’adresse ce livreCe texte s’adresse à des lecteurs déjà confrontés à la complexité :architectes data / SI,responsables IA,chercheurs,décideurs techniques,ingénieurs confrontés à l’échec de solutions pourtant « bien conçues ».Il n’est pas destiné :aux débutants,aux amateurs de recettes,ni aux discours de solutionnisme technologique.FormatPDF – lecture individuelleAucune promesse d’outillage ou de méthode clé en main.AuteurDr Nicolas FigayArchitecte du sens et médiateur conceptuel des systèmes complexes
Unsupervised local learning based on voltage-dependent synaptic plasticity for resistive and ferroelectric synapses
No full textInternational audienceThe deployment of AI on edge computing devices faces significant challenges related to energy consumption and functionality. These devices could greatly benefit from brain-inspired learning mechanisms, allowing for real-time adaptation while using low-power. In-memory computing with nanoscale resistive memories may play a crucial role in enabling the execution of AI workloads on these edge devices. In this study, we introduce voltage-dependent synaptic plasticity (VDSP) as an efficient approach for unsupervised and local learning in memristive synapses based on Hebbian principles. This method enables online learning without requiring complex pulse-shaping circuits typically necessary for spike-timing-dependent plasticity (STDP). We show how VDSP can be advantageously adapted to three types of memristive devices (TiO2, HfO2-based metal-oxide filamentary synapses, and HfZrO4-based ferroelectric tunnel junctions (FTJ)) with disctinctive switching characteristics. System-level simulations of spiking neural networks incorporating these devices were conducted to validate unsupervised learning on MNIST-based pattern recognition tasks, achieving state-of-the-art performance. The results demonstrated over 83% accuracy across all devices using 200 neurons. Additionally, we assessed the impact of device variability, such as switching thresholds and ratios between high and low resistance state levels, and proposed mitigation strategies to enhance robustness
An Interpretable Model for Multi-Target Predictions with Ordinal Outputs
No full textInternational audienceMulti-Target Prediction (MTP) aims to predict a class for multiple targets from a single input instance. In this paper, we focus on the general ordinal outputs setting, where targets may have different numbers of ordered classes. This scenario is under-explored yet critical for human-centered applications such as educational assessment or psychological profiling. In these domains, targets span multiple aspects of an evaluation, and capturing latent relationships between seemingly independent targets is essential for deriving meaningful user profiles. Beyond prediction accuracy, it is crucial that the resulting profiles are interpretable, as these applications directly impact humans. We introduce IMPACT, a novel method that extends existing binary MTP frameworks (such as CD-BPR) to the multi-class domain through a newly designed loss function. Rooted in a Bayesian modeling framework, IMPACT jointly embeds user profiles and targets within a shared vector space, providing theoretical rigor while explicitly optimizing for both predictive accuracy and interpretability. Furthermore, IMPACT offers a geometric interpretation of the embedding learning dynamics, giving insight into how the model captures relationships between users and targets and providing an intuitive understanding of profile formation in the latent space. Experimental results show that IMPACT outperforms state-of-the-art approaches in terms of profile interpretability while maintaining competitive prediction accuracy. An ablation study highlights the contribution of each component, demonstrating the benefits of extending the framework to multiple ordered classes, the Bayesian formulation, and the geometric interpretability in enhancing both performance and transparency
On a mode-matching technique for acoustic scattering by a cascade of cambered vanes
No full textInternational audienceA two-dimensional mode-matching technique is developed to compute the scattering of an acoustic wave by a cascade of staggered and cambered vanes in subsonic regime, such as those encountered in axial-flow fan stages. Apart from the need to reproduce a more realistic geometry in analytical modeling, introducing vane camber is a relevant way of retrieving the global evolution of the mean flow away from the vanes, simply by mass-flow conservation through the expanding inter-vane channels. This prevents mean-flow discontinuity and introduces a realistic variation of the equivalent dipole sources along the vane chord. The expansion of the cross-section along the inter-vane channels, induced by curvature, is taken into account by multiple-scale analysis, assuming slow variations of the geometry. The validity of the model is assessed by extensive comparisons with high-fidelity numerical results, with and without flow. The assumptions used in the analytical model are found to be suited to modern geometries of outlet guide vanes