Université de technologie de Troyes open archive
Not a member yet
10722 research outputs found
Sort by
Energy Mix based on the National Master Plan for Energy and Changes of Life Cycle GHG Emission Factors
No full textInternational audienceThe energy mix serves as the basis for calculating the greenhouse gas (GHG) emission factor of electricity, and therefore has a significant impact not only on the evaluation of GHG-related policies but also on the carbon footprint of industries that use electricity in life cycle assessments (LCA). As such, there is a need to annually publish GHG emission factors that reflect changes in the energy mix. The purpose of this study is to present IPCC- and LCA-based GHG emission factors and GHG emissions for the energy sector based on changes in the energy mix, and to assess whether the nationally determined contribution (NDC) can be achieved under these conditions. To achieve this, the study applies Intergovernmental Panel on Climate Change (IPCC) and LCA-based GHG emission factors for each power source to the energy mix outlined in the 10th Basic Plan for Electricity Supply and Demand, calculating annual (2018~2036) national electricity GHG emission factors and emissions, and analyzing the feasibility of meeting the NDC target. The analysis revealed that GHG emission factors fluctuate significantly with changes in the energy mix, underscoring the need for annual calculations. Under the planned energy mix, GHG emissions from the energy transition sector are projected to reach 159.9 million tons CO2eq, which exceeds the NDC target of 149.9 million tons CO2eq. However, reducing coal-fired power generation by 10% and replacing it with offshore wind and solar power could make achieving the target feasible. Additionally, the LCA-based GHG emission factors indicate that expanding offshore wind and solar power instead of relying on hydrogen and ammonia in the energy mix could achieve a 2.5% reduction. Therefore, adopting methodologies such as those used in this study to calculate annual GHG emission factors would allow efforts to transition the energy mix to be immediately reflected. Furthermore, when planning the energy mix, an LCA-based approach rather than an IPCC-based approach would provide a more effective means of responding to practical environmental regulation
Deep Learning for Forecasting Patient Visits: A Comparative Study of CNNs, xLSTM, and Transformers
No full textInternational audiencePatient forecasting is crucial for healthcare management, enabling optimal resource allocation, improved patient outcomes, and better decision-making in emergency settings. While traditional statistical models like ARIMA continue to be widely used due to their interpretability, their limitations in processing complex and dynamic healthcare data have led to the growing adoption of deep learning (DL) techniques. Long Short-Term Memory (LSTM)-based models have shown promising results in the literature, particularly in capturing temporal dependencies. However, their ability to adapt to sudden fluctuations in patient arrivals remains a major challenge, which is critical for effective hospital planning. This study compares several modern deep learning architectures and evaluates their effectiveness in healthcare time series forecasting, aiming to improve predictive accuracy while addressing the challenge of capturing extreme variations in emergency department patient visits. The models assessed include xLSTM variants (xLSTM-Time, xLSTM-Mixer), PatchTST, Temporal Fusion Transformer (TFT), RWKV-TS, Temporal Convolutional Networks (TCN), and TimesNet. Our results indicate that PatchTST achieves the lowest forecasting error (RMSE = 13.18, MAPE = 7.17%), demonstrating superior adaptability to complex temporal patterns. However, all models struggle with extreme daily fluctuations, emphasizing the need for more advanced hybrid approaches and the integration of external data sources to enhance robustness and reliability in real-world healthcare applications
Forensics Analysis of Residual Noise Texture in digital Images for Detection of Deepfake
No full textInternational audienceThis paper proposes an original approach for the automatic detection of AI-generated images, using features derived from noise residuals artefacts. Contrary to most current research that leverages sophisticated deep learning models to further improve performance, this study highlights the distinct noise residual characteristics in deepfakes, facilitating the identification of AI-generative images. Our findings highlight some limitations of image models, which can be used for forensic analysis and for future AI-based text-to-image generative models. Broad numerical results on a large and diverse dataset show the interest of the identified features as well as the relevance of the present method
MultiAgentNetSim: Empowering Next-Generation Network Modeling with Multi-Agent Simulation
No full textInternational audienceThe increasing complexity of next-generation networks necessitates advanced simulation techniques, with multi-agent simulation (MAS) emerging as a highly effective solution. MAS enables a thorough analysis of micro-level interactions and intricate interdependencies inherent in modern networks, along with their implications – complexities that traditional simulation methods are increasingly unable to address effectively. Motivated by this potential, MultiAgentNetSim is proposed, founded on the principles of MAS. A key feature of MultiAgentNetSim is its capacity to facilitate realistic simulations of complex network scenarios, with network slicing serving as a prominent example explored within this article. Beyond serving as a simulation environment, MultiAgentNetSim functions as a decision-support tool, providing dynamic inputs for algorithm training, and a robust framework for evaluating algorithmic performance. A notable example is its integration with a dynamic pricing algorithm within the network slicing scenario. The simulation inputs deliver expressive and realistic data, enabling a more informed pricing strategy. This strategy, explored through the metric of operator profit, can be assessed across multiple metrics and continuously optimized using feedback from the platform, making it an invaluable asset in modern network management
Advances in Fiber Optic Surface-Enhanced Raman Spectroscopy Sensors
No full textInternational audienc
Target-Specific Domain Adaptation via Geometry-Correlation Prediction for Point Cloud
No full textInternational audiencePoint cloud datasets suffer from a large domain discrepancy due to variations in data acquisition procedures, sensor perspectives, and realistic noise. To address this issue, domain adaptation technologies have emerged to improve scalability and generalization for point cloud models. We propose a novel method named GeoCo-TSDA (Geometry-Correlation Prediction and Target-Specific Domain Adaptation) which boosts the performance of domain adaptation with a geometry-correlation prediction task and a target-specific self-training strategy. We design a self-supervised task that predicts the geometry correlation, which is obtained by the covariance of the local point clusters and is related to a variety of geometric properties, enabling the model to learn more complete and robust features. Moreover, existing domain adaptation methods commonly focus on aligning feature space among domains. However, owing to the internal distribution gap among domains, merely aligning features at the domain level falls short of achieving optimal performance for the target domain. We tackle this problem by adding a target domain specific training procedure that focuses on further adapting the model to fit the internal distribution of the target domain. For the rationality of our method, we provide theoretical and empirical analysis. For the effectiveness of our method, we conduct experiments on commonly used benchmark PointDA-10 and GraspNetPC-10, and on both datasets our model achieves SOTA performance among point cloud domain adaptation methods and considerable elevation compared to the baseline model
Apprentissage Profond pour la Prévision des Visites de Patients : Une Étude Comparative entre xLSTM et Transformers
No full textInternational audienceApprentissage Profond pour la Prévision des Visites de Patients : Une Étude Comparative entre xLSTM et Transformer
Hyperglosae : L’hypertexte nelsonien à la rencontre des pratiques des traducteurs et des ethnographes
No full textInternational audienceThis paper deals with the feedback from two long-term research and technology projects: one on the instrumentation of translators' practices, the other on that of ethnographers. Despite the differences in their initial objectives and target disciplines, these two projects have gradually converged, model after model, use after use, to a theory of the intellectual work instrumentation or, at the very least, that of the supports and gestures required for interpretive work. This theoretical convergence is now materializing in the ongoing convergence between the two platforms around a common hypertextual infrastructure, built to reproduce the fundamental properties of Xanadu and reappropriate the visual forms stemming from the age-old tradition of text interpretation.Cet article présente le retour d'expérience croisé de deux projets scientifiques et techniques au long cours : l'un portant sur l'instrumentation des pratiques des traducteurs, l'autre sur celle des ethnographes. Malgré la différence de leurs objectifs initiaux et des disciplines visées, ces deux projets ont peu à peu convergé, comme si se dessinait, modèle après modèle, usage après usage, l'esquisse d'une théorie de l'instrumentation du travail intellectuel ou, tout au moins, celle des supports et des gestes nécessaires à un travail d'interprétation. Cette convergence théorique se matérialise aujourd'hui dans une convergence en cours entre les deux plateformes autour d'une infrastructure hypertextuelle commune, construite pour reproduire les propriétés fondamentales de Xanadu et se réapproprier les formes visuelles issues de la tradition millénaire de l'interprétation des textes
Electrical Characterization of HV (10 Kv) Power 4H-SiC Bipolar Junction Transistor
No full textInternational audienceIn this paper, the static and dynamic characterization of a High Voltage (10kV) 4H-SiC Bipolar Junction Transistor (BJT) is presented. Using a high-voltage source in vacuum conditions, a breakdown voltage of 11 kV was measured. Results showed that both large and small BJTs exhibit similar on-state resistance per unit area and collector current density of 55 A.cm-2 . The current gain increases with a decrease in temperature, indicating reduced charge carrier recombination at lower thermal energies. Also, BJT have been characterized in switching mode at 1 kV. The study concludes that 4H-SiC BJT demonstrates promising electrical performance for high-efficiency applications in harsh environments
"Cadres et Outils pour la Conception en Soutenabilité Forte" - Séminaire D-TechnoSS
No full textCette planche synthétise sous forme d’illustrations graphiques les réflexions menées lors de la table ronde “Cadres et Outils pour la Conception en Soutenabilité Forte” dans le cadre du séminaire D-TechnoSS - 8 juillet 2025Depuis les années 1970, le rapport Meadows a mis en évidence la crise socio-écologique actuelle. La dernière décennie a vu une prise de conscience collective et la nécessité de lutter contre le changement climatique, ce qui entraîne une mobilisation pour développer des connaissances, des réglementations et des technologies visant à intensifier la décarbonation de la société. En réalité, le changement climatique n’est qu’une des six limites planétaires déjà dépassées aujourd’hui (Richardson et al., 2023 ; Rockström et al., 2009 ; Steffen et al., 2015). La plupart des actions mises en place pour répondre aux défis socio-écologiques restent axées sur des approches techno-centrées et/ou ne remettent pas en question le modèle économique actuel. Ces actions, fondées sur l’économie circulaire, les technologies vertes, l’éco-conception, les modèles économiques traditionnels et les stratégies de décarbonation, demeurent insuffisantes pour relever les défis socio-écologiques mis en évidence par l’économie du donut (Brozovic, 2020 ; de Oliveira Neto et al., 2018 ; Raworth, 2017 ; Vilochani et al., 2024). Il est devenu important de développer de nouvelles démarches de co-construction des solutions soutenables entre les concepteurs et les consommateurs et des modes de consommation plus durables pourraient sans doute favoriser une transition vers une soutenabilité forte. Ce séminaire a pour objectif d’approfondir le concept de soutenabilité forte ainsi que sa relation avec les comportements du consommateur et les produits et services intégrant les enjeux socio-écologiques en conception. Il s'inscrit dans le cadre du projet D-TechnoSS financé par l’ANR et mené par l’Université de Technologie de Troyes, l’École des Mines de Saint-Étienne et l’Université Paris Nanterre. Il vise à développer une meilleure compréhension des mécanismes de changements des comportements afin d'adopter des démarches de co-construction de soutenabilité forte