HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
De la pensée écologique à l’ingénieur contemporain : Entretien avec Dominique Bourg
No full textMerci à Frédéric Huet pour sa relecture à la fois critique et constructive.International audienceCet entretien avec Dominique Bourg constitue le premier texte de cette nouvelle rubrique des Cahiers Costech : « Les défis de la pensée écologique ». Il introduit à sa manière aux problèmes que cette rubrique entend affronter : la singularité et les limites de la pensée écologique ; sa capacité – ou son incapacité – à déjouer la logique du capitalisme contemporain ; les conséquences à en tirer s’agissant de l’ingénierie soutenable qu’il convient aujourd’hui de promouvoir. Dominique Bourg y souligne particulièrement l’importance de l’agriculture, aussi bien du point de vue des rapports de domination et des pratiques techniques dont elle est porteuse que s’agissant de la pensée et de l’anthropocentrisme qu’elle rend possible
A modified TimeGAN-based data augmentation approach for the state of health prediction of Lithium-Ion Batteries
No full textInternational audienceLithium-ion batteries are critical components of zero-emission electro-hydrogen generators (GEH2), where accurate performance prediction is essential for ensuring optimal operation and enabling effective predictive maintenance. Data-driven models have become increasingly prominent for predicting the State of Health (SOH) of lithium-ion batteries due to their high accuracy and reduced development time. However, in hybrid systems like GEH2, where the battery frequently remains inactive while the fuel cell supplies most of the power, the available battery data is limited. This data scarcity presents a significant challenge for achieving accurate SOH prediction. To address this challenge, we propose a novel data augmentation approach that integrates Time-series Generative Adversarial Network with a Transformer and a Gated Recurrent Unit to enhance data availability and improve prediction accuracy. This new approach enhances the model’s ability to capture long-term temporal dependencies within multivariate battery parameters while effectively addressing irregular time intervals, a common challenge in real-world batteries datasets. We evaluated the proposed approach using real-world industrial datasets from four distinct GEH2 batteries and two additional batteries from the publicly available NASA dataset. The performance of SOH prediction was assessed using a Long Short-Term Memory (LSTM) model trained on augmented data generated by various data augmentation techniques. The results consistently demonstrate that our approach outperforms all competing methods, highlighting its superior ability to enhance data for lithium-ion batteries. These findings highlight the effectiveness of our approach in enhancing predictive accuracy and robustness, making it highly suitable for real-world battery applications
Robust health indicator extraction and RUL prediction for PEMFCs under highly dynamic industrial conditions
No full textInternational audienceProton Exchange Membrane Fuel Cells (PEMFCs) are increasingly deployed in clean energy systems, such as GEH2 hydrogen generators, where they operate under highly dynamic and unpredictable load conditions. Accurate prediction of their Remaining Useful Life (RUL) is essential for ensuring reliable, cost-effective, and proactive maintenance strategies. However, conventional voltage-based Health Indicators (HIs) are highly sensitive to power fluctuations and fail to provide consistent degradation trends in real-world industrial scenarios, particularly when system usage varies significantly across different clients, as in the GEH2 case. In this paper, we propose a scalable two-stage framework for RUL prediction of PEMFCs operating under such conditions. First, we introduce a machine learning-based method to extract a degradation-specific Health Indicator directly from voltage measurements, effectively filtering out transient operational effects. Second, we develop a hybrid deep learning architecture that combines Transformer networks and Gated Recurrent Units (GRUs) to model temporal dependencies and provide accurate RUL predictions under dynamic conditions. The proposed approach is validated on a real-world industrial dataset collected from three PEMFC stacks deployed in GEH2 systems operating under highly variable conditions. Comparative results show that our method consistently outperforms baseline machine learning and deep learning models, achieving superior accuracy, robustness, and generalization across diverse mission profiles
Vector Map Quality Metrics for Contextual Autonomous Driving Systems
No full textInternational audienceEnsuring safety in autonomous driving requires continuous map maintenance supported by reliable quality indicators. In this context, it is crucial to identify when and where map updates should be triggered, for instance through crowdsourced data, and under which conditions a new map compilation should be deployed. This paper focuses on effective metrics for assessing the quality of vector maps and guiding such decisions. We present a new metric called GOSPAM designed to measure map discrepancies in terms of location errors, existence, and completeness. Through detailed simulations on both point and polyline feature maps, we analyze its sensitivity to common map degradation such as bias, false positives, false negatives, and coordinate errors. The results demonstrate that GOSPAM offers a unified and interpretable measure that effectively captures various forms of map deviation, making it a strong candidate for map quality assessment in automotive applications
Classification hiérarchique et détection de données hors distribution pour la vision
No full textThis thesis addresses the reliability challenges of neural networks, which have become ubiquitous but remain "black boxes" that are difficult to interpret, particularly problematic in critical applications. It explores two complementary approaches: first, hierarchical classification that integrates structural information during training to reduce the severity of classification errors through novel training loss methods and label smoothing techniques; second, Out-of-Distribution (001)) detection that develops post-hoc methods to identify aberrant samples without modifying the model architecture, including an in-depth study of the impact of underlying classifier performance. An additional contribution proposes an exploratory method in anomaly detection using "sister classes" to differentiate multiple types of anomalies, although it showed limitations on real image data. The overall objective is to improve confidence and predictability of neural network classifiers in contexts where reliability is crucial.Cette thèse aborde la problématique de fiabilité des réseaux de neurones, qui sont devenus omniprésents mais restent des "boîtes noires" difficiles à interpréter, particulièrement problématiques dans les applications critiques. Elle explore deux approches complémentaires : premièrement, la classification hiérarchique qui intègre des informations structurelles pendant l'entraînement pour réduire la gravité des erreurs de classification via de nouvelles méthodes de perte d'entraînement et de lissage d'étiquettes ; deuxièmement, la détection Hors-Distribution (OOD) qui développe des méthodes post-hoc pour identifier les échantillons aberrants sans modifier l'architecture du modèle, incluant une étude approfondie de l'impact de la performance du classificateur sous-jacent. Une contribution additionnelle propose une méthode exploratoire en détection d'anomalies utilisant des "classes sœurs" pour différencier plusieurs types d'anomalies, bien qu'elle ait montré des limitations sur les données d'images réelles. L'objectif global est d'améliorer la confiance et la prévisibilité des classificateurs à réseaux de neurones dans des contextes où la fiabilité est cruciale
Uncertainty in quantitative bipolar argumentation frameworks
No full textInternational audienceOnline deliberation platforms allow people to exchange their opinions around a specified issue and to vote on these opinions in order to reach a collective decision. Argumentation allows to structure and analyse user input for these platforms. A debate can be represented by a quantitative bipolar argumentation framework where votes on each argument of the debate are aggregated into an initial weight. One of the main challenges these platforms face is sparse voting i.e. participants vote on a few number of arguments, leading to an imbalance of the number of votes between the arguments. In this paper, we propose a methodology that handles sparse voting in online debates, by introducing imprecise quantitative bipolar argumentation frameworks that incorporate uncertainty on the initial weights. Specifically, we leverage votes on arguments to initialize weight intervals that represent the uncertainty on the initial weights, using the imprecise Dirichlet model. We use four state-of-the-art bipolar gradual semantics to generate a final acceptability interval on each argument and we introduce several properties to study the effect of these semantics on the uncertainty on each argument's final evaluation. Our methodology allows for a more robust representation of argument strength in the presence of limited data
Role of activation progress on textural properties of biochar and their impact on tar cracking
No full textSource Agritrop Cirad (https://agritrop.cirad.fr/613528/)International audienceThe gasification process converts biomass into syngas and produces biochar, a porous solid by-product, potentially with a high surface area, ideal for tar cracking applications. In this study, biochars with varying textural properties were produced by activating them under CO2 and H2O atmospheres to different extents of conversion. The relationship between these properties—particularly surface area and pore size distribution—and the conversion of a model tar was then examined. Results show that while the activation agent (CO2 or H2O), which impacts pore size distribution, had minimal influence on tar conversion, the activation progress, controlling specific surface area, was crucial. At 850 °C, toluene conversion ranged from 35 % for biochar with 300 m2/g surface area to nearly 100 % for biochar with 800 m2/g. Additionally, activation improved gas quality, with higher concentrations of H2, CO, and CH4 correlating with increased biochar surface area
Atypical Homogeneous Rheology of a High-Entropy Metallic Glass Challenges Standard Free Volume Models
No full textInternational audienceMetallic glasses (MGs) exhibit exceptional mechanical properties but their application is often limited by brittleness. At elevated temperatures near the glass transition (T), they undergo homogeneous viscoplastic deformation, a regime commonly described using free volume (FV) theory. Despite its prevalence, the quantitative accuracy and applicability of FV models, particularly for transient behaviors, remain subjects of investigation. This study examines the homogeneous rheology of a LaCeYNiAl High-Entropy MG (HEMG) between 475 K and 490 K, and critically assesses the relevance of two prominent FV model formulations. Experimental characterization included dynamic mechanical analysis and uniaxial tensile tests across various strain rates. The tensile data were subsequently analyzed using two elasto-viscoplastic constitutive frameworks incorporating distinct FV evolution kinetics: Spaepen’s original formulation (Model 1) and the bimolecular annihilation kinetics proposed by Van den Beukel/Sietsma (Model 2). Our analysis reveals that Model 1, when applied to steady-state flow, yields physically inconsistent negative parameters, questioning its validity for homogeneous deformation. Model 2 demonstrates better qualitative agreement with experimental stress-strain curves but fails to accurately reproduce stress overshoot features. Furthermore, fitting Model 2 necessitates unphysically low Young’s modulus values and results in unusual negative apparent activation energies for key kinetic parameters, suggesting limitations in the model structure (e.g., neglecting explicit viscoelasticity) or potentially unique behavior in HEMGs. These findings highlight significant shortcomings of standard FV models in quantitatively capturing the homogeneous deformation of this HEMG, particularly transient effects, and underscore the need for more refined constitutive descriptions
Shaping the Future of Mobility Culture: An Expert-Driven Policy Roadmap
No full textInternational audienceDaily mobility patterns vary across regions but share key similarities, shaped by spatial, social, economic, and cultural factors. Mobility can be seen as a “culture” in itself, and in Europe, this culture is currently undergoing significant evolution. The Horizon 2020 project REBALANCE, involving six participants from six European countries until 2022, sought to explore the values and culture of future mobility. The project’s goal was to encourage a paradigm shift in mobility policy, urging greater attention to social values in decision-making and alignment with the Sustainable Development Goals (SDGs).REBALANCE initiated discussions about the foundations of current mobility policies. By reflecting on existing frameworks, it brought together a broad network of experts and stakeholders to explore new mobility cultures and policies. The objective was to converge on a sustainable, shared vision and create a roadmap for a new mobility paradigm. What set REBALANCE apart was its focus on engaging high-level interdisciplinary thinkers who have traditionally not been involved in transport research. Philosophers, sociologists, psychologists, geographers, and legal experts played a crucial role in challenging the assumptions underlying conventional mobility paradigms. In addition to this interdisciplinary engagement, the project involved a group of visionaries and 14 transport experts. A key outcome was the creation of a Manifesto for a New Mobility Culture.The project aimed to significantly influence European public policies on mobility and transport. It examined the evolution of mobility culture, identifying emerging trends through expert insights. These trends were translated into future scenarios, distinguishing between desirable and probable trajectories. Through surveys across European countries and extensive input from experts and thinkers, REBALANCE articulated a vision for the future of mobility. This vision was formalized in a manifesto addressed to European policymakers, calling for integration into future public policies. Supported by a roadmap containing concrete actions, timelines, and policy guidance, the manifesto offered a plan to break free from outdated paradigms. Additionally, it underscored the importance of communication in shaping mobility culture, examined the public interest in mobility issues, and proposed new methods for evaluating mobility projects to better address the necessary changes.This presentation will focus on the roadmap and guide for future European mobility policies. It will detail the methodology employed to translate the REBALANCE vision into the manifesto and, ultimately, into the roadmap and policy guide. An innovative workshop, inspired by the Delphi method, was conducted to bring together experts, including scientists, associations, and industry representatives, to achieve a high level of consensus at every stage of the process
Outils d'accompagnement à la mise en conformité des dispositifs médicaux de diagnostic in vitro legacy devices au Règlement Européen 2017/746
No full textSince its entry into force on May 5, 2017, Regulation (EU) 2017/746 on in vitro diagnostic medical devices has imposed considerable challenges on manufacturers and notified bodies, indirectly impacting the availability of devices on the market.On the one hand, manufacturers are now required to meet increased demands, necessitating precise interpretation and rigorous integration into their practices. On the other hand, the number of notified bodies, already limited, has further decreased due to the new accreditation conditions stipulated by the regulation (UE) 2017/746. These bodies, overwhelmed by a massive influx of files to process, struggle to meet the deadlines set by the European Council and Parliament for conformity assessments. To facilitate the transition of manufacturers to this new regulation, guides developed by the Medical Device Coordination Group (MDCG) provide valuable recommendations. In this context, our project aims to support manufacturers by providing two tools : a chronological template-style mapping that quickly explains the conditions for benefiting from transitional periods and what compliance with Regulation (EU) 2017/746 entails. These tools will enable manufacturers to optimally plan their compliance before the end of the transitional period.these tools are based on an in-depth analysis of regulatory requirements and take into account the needs expressed by industry professionals, gathered through a field survey.Depuis son entrée en vigueur le 5 mai 2017, le Règlement (UE) 2017/746 sur les dispositifs médicaux de diagnostic in vitro impose des défis considérables aux fabricants et aux organismes notifiés, impactant indirectement la disponibilité des dispositifs sur le marché.D’une part, les fabricants doivent désormais répondre à des exigences accrues, nécessitant une interprétation précise et une intégration rigoureuse dans leurs pratiques. D’autre part, le nombre d’organismes notifiés, déjà limité, a encore diminué en raison des nouvelles conditions d’accréditation prévues par le règlement (UE) 2017/746. Ces organismes, confrontés à un afflux massif de dossiers à traiter, peinent à respecter les délais fixés par le Conseil et le Parlement européens pour les évaluations de conformité. Pour faciliter la transition des fabricants vers cette nouvelle règlementation, des guides élaborés par le Medical Device Coordination Group (MDCG) apportent des recommandations précieuses. Dans ce contexte, notre projet vise à soutenir les fabricants en leur fournissant deux outils : d'une part, une cartographie qui permet d’expliquer rapidement au fabricant les conditions pour bénéficier des périodes transitoires et en quoi consiste la mise en conformité au règlement (UE) 2017/746 ; d'autre part, un outild’autodiagnostic permettant de faire ressortir les écarts entre les pratiques des fabricants de dispositifs médicaux de diagnostic in vitro (DMDIV) legacy devices sous la directive et les exigences du règlement. Ces outils reposent sur une analyse approfondie des exigences règlementaires et la prise en compte des besoins exprimés par les professionnels du secteur et collectés lors d’une enquête terrain