HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
Impact of oxic and thermal transient phases on corrosion of carbon steel in different cementitious media: insights from new in situ experiments
No full textInternational audienceThe impact of oxic and thermal transient phases on corrosion of carbon steel in a cementitious environment was studied through three in situ experiments (Tournemire underground research laboratory, France). For 2 years, heated metallic samples (80°C) were placed in direct or indirect contact with two different cementitious materials: a low-pH bentonitic cement grout (BCG) and a Portland cement paste material (CEM I). Mineralogical and microstructural analyses were carried out in an attempt to identify the combined effects of pH, chemistry and microstructure properties associated with such specific cementitious media on steel corrosion mechanisms. Additionally, in situ electrical resistance corrosion sensors allowed continuous monitoring of the corrosion rates corresponding to each of the three field experiments. Post-mortem characterization indicated that metallic samples embedded in low-pH BCG were heavily damaged and exhibited high corrosion rates. Conversely, steel samples in contact with a highly alkaline CEM I environment appeared to be much less impacted by corrosion processes and revealed extremely low corrosion rate values. A comparison between these field experiments observations and results previously obtained through complementary laboratory mock-up tests finally enabled the evaluation of the impact that variations in geometrical/design aspect existing between in situ and laboratory tests can induce on material degradation
Thermo-mechanical simulation of L-PBF process at part-scale by coupling grain structure calculation and crystal viscoplasticity
No full textInternational audienceAn integrated numerical framework is proposed to perform a thermomechanical analysis of Laser Powder Bed Fusion (L-PBF) at the part scale, covering the entire construction process. The novelty lies in the fact that the thermomechanical analysis is driven by the evolving grain structure, also predicted at the part scale. The grain structure is first generated using a hybrid - Cellular Automaton (CA) method, which enables the generation of the grain structure on a CA grid at part-scale while incorporating the detailed scanning trajectories. For the thermo-mechanical simulation, a layer-by-layer thermal analysis includes the non-exposed powder, whereas the mechanical analysis only considers the substrate and the part under construction. A crystal plasticity model is employed for the nickel-based superalloy Inconel 718 (IN718), utilizing the CA grain structure referenced by Euler angles. A reduced grain approach is proposed, wherein Euler angles from the CA grid are projected onto the mechanical mesh. The mechanical mesh size is chosen appropriately to balance computational efficiency and accuracy. The calibration of the parameters of the crystal plasticity laws is conducted from room temperature up to 1100 C using tests performed on homogenized IN718 material. Tensile tests on the representative volume elements (RVEs) of grain structures with different laser scanning trajectories are performed to study the mechanical response for each specified texture. Finally, thermo-mechanical simulations are applied to the L-PBF construction of a propeller with single-crystal and polycrystalline grain structures. At each time increment, the temperature field is transferred to the mechanical mesh, and grains are activated according to the part under construction. The influence of local grain texture on anisotropic stress distribution and distortion is analyzed and discussed
Paradoxes as a compass in the unknown: the case of knowledge transfer in the context of demographic transition
No full textInternational audienc
Evaluation of AI-based Forecasting Models for Electricity Demand at Household Level: Focus on Generative AI Models
No full textInternational audienceShort-term forecasting of electricity consumption, from a few minutes to day ahead, is critical for distribution system operators (DSOs) to manage the electricity grid, balance supply and demand, and integrate renewable energy sources. For a long time, forecasting approaches used to rely on statistical methods (ARIMA, exponential smoothing) or classical "lightweight" machine learning models (Gradient Boosting, Random Forest). These models require extensive feature engineering and show limited accuracy and robustness. The recent emergence of Time Series Foundation Models (TSFMs), mostly based on transformer architectures and pre-trained on millions of time series are capable of "zero-shot" forecasting, which means without specific training, offers a promising alternative for forecasting on unseen data . However, their practical applicability to specific tasks such as residential consumption forecasting is still poorly evaluated, particularly regarding forecasting accuracy and computational efficiency. This work presents a comprehensive benchmark comparing 12 models, from naive baselines to state-of-the-art TSFMs, on residential electricity consumption data at 15-minute resolution. Results show that fine-tuned TSFMs achieve the lowest NRMSE (0.58 ± 0.14), outperforming classical ML approaches (XGBoost: NRMSE 0.65), while zero-shot variants remain competitive without any task-specific training. The high MAPE values (≈45-75 %) observed across the models is mainly due to the nature of the forecasting at household scale
Laboratoire numérique : workflow digital en pratique et intégration de l’IA en pathologie de routine, à travers l’exemple de la prostate
No full textNational audienceDigital pathology is a major technological revolution for pathology. It modernizes routine practices and paves the way for the integration of artificial intelligence (AI) solutions for diagnostic and research purposes. At Rennes University Hospital, digital pathology has been routinely deployed since 2020, and an AI solution for the detection of prostate adenocarcinoma (Galen® Prostate, Ibex) has been integrated since July 2023. In this article, we review our experience in Rennes and assess both the impact of digitization on the various professions within the department and the prospective use of AI for routine diagnosis. The concordance between AI and pathologists was 93.2% for the detection of high-probability cancer and 99% for low-probability slides. Among slides with intermediate probability (43% of the total), cancer was confirmed in 4.7% of cases. For Gleason grading, the concordance rate was 76.6%. To date, the integration of AI has not changed the use of immunohistochemistry. A 10% failure rate related to pre-analytical artifacts was observed and is an area for improvement in our practices. Thus, the effectiveness of digital pathology and the use of AI models are closely dependent on pre-analytical quality and its organizational integration.La pathologie numérique est une révolution technologique majeure pour l’anatomie et cytologie pathologiques. Elle modernise les pratiques de routine et ouvre la voie à l’intégration de solutions d’intelligence artificielle (IA) à visée diagnostique et de recherche. Au CHU de Rennes, la pathologie numérique est déployée en routine depuis 2020, et une solution d’IA pour la détection de l’adénocarcinome prostatique (Galen® Prostate, Ibex) est intégrée depuis juillet 2023. Nous proposons à travers cet article un retour sur notre expérience rennaise et une évaluation tant de l’impact de la numérisation sur les différents corps de métier du service que de l’utilisation prospective d’une IA à visée diagnostique en routine. La concordance entre l’IA et les pathologistes était de 93,2 % pour la détection de cancer à forte probabilité, et de 99 % pour les lames à faible probabilité. Parmi les lames à probabilité intermédiaire (43 % du total), le cancer a été confirmé dans 4,7 % des cas. Pour le grading de Gleason, la concordance atteignait 76,6 %. L’intégration de l’IA n’a pas à ce jour modifié le recours à l’immunohistochimie. Un taux d’échec de 10 % lié à des artefacts pré-analytiques a été observé et est un axe d’amélioration de nos pratiques. Ainsi, l’efficacité de la pathologie numérique et de l’utilisation de modèles d’IA dépend étroitement de la qualité pré-analytique et de son intégration organisationnelle
Resilience analysis for energy foresight scenarios: addressing the deep uncertainties of hydrogen deployment in industrial hubs
No full textInternational audienceHydrogen is emerging as a promising solution for reducing greenhouse gas emissions in key industrial sectors. However, its deployment faces significant uncertainties related to technological progress, economic viability, political decisions, and geopolitical dynamics. This study adopts a scenario-based approach to assess the resilience of a hydrogen ecosystem in an industrial hub. A two-step methodology is developed to explore how an optimised multi-energy system would respond to unanticipated disruptions such as fuel price shocks, policy changes, or slower-than-expected technological advancements. First, an integrated capacity expansion and operation model determines optimal system capacities assuming perfect foresight. Second, the resulting energy system is evaluated under an alternative context, where only operational decisions are optimised. Special attention is given to electricity price modelling, due to its crucial impact on electrolytic hydrogen production. Nine scenarios are investigated, reflecting major categories of deep uncertainty relevant to long-term energy planning, including policy directions, technological innovation, and resource availability. Results show that the most pessimistic scenarios in terms of hydrogen cost are also the most robust. Scenarios with more electrolysis capacities are generally more expensive but less sensitive to disruptions. In contrast, scenarios relying on early development of a European hydrogen transportation infrastructure are cheaper but less resilient. For policymakers, these findings highlight the need for coherent strategies integrating electricity and hydrogen systems to balance cost-efficiency, resilience, and environmental goals. Clear strategic direction, combined with support for local hydrogen industries, will be vital to securing Europe’s energy sovereignty and long-term sustainability in the face of evolving uncertainties
Revisiting the meanings of the Critical Zone through the OZCAR research infrastructure example, definitions and evolutions
No full textOZCARInternational audienceSince its first definition by the National Research Council in 2001, the concept of Critical Zone has known undeniable success over the last quarter of a century. A success that is often reflected by the evolution and diversification of its meanings. Recently, Lee et al. (2023) proposed a review that literally focuses on “the meanings of the Critical Zone”. Through an extensive review of the literature across the disciplines and journals, they have identified three loosely overlapping meanings. An ontological meaning, where the Critical Zone is mostly seen as the Earth’s spatial interface where geochemical and biological activity sustains life. An epistemic meaning, where the Critical Zone is considered a product of collaborative efforts between scientific communities to build a whole-system knowledge data-base and library. And finally, an anthropocenic meaning, where the Critical Zone is the vulnerable home of the human species. In this contribution, we aim at revisiting these three meanings through the creation and development of the French network OZCAR (Critical Zone Observatories: Research and Application).Created in 2015 to enhance the collaborations between Critical Zone observatories (Gaillardet et al., 2018), OZCAR is a French Research Infrastructure that gathers 23 national observation services and +120 study sites in metropolitan France and on 5 continents. If most observation services existed prior to the creation of OZCAR, we have seen major evolutions over the last decade as the OZCAR community developed and bloomed. Originally conceived as a spatial definition (ontological meaning), the “Critical Zone” words in OZCAR became a vast collaborative effort to develop the whole system approach and data base (epistemic meaning). It is now also fostering transformative research aimed at preserving our planet’s habitability, i.e., the giant spaceship in which we all live together (anthropocenic meaning).References:Lee, R. M., Shoshitaishvili, B., Wood, R. L., Bekker, J., & Abbott, B. W. (2023). The meanings of the Critical Zone. Anthropocene, 42, 100377.,doi:10.1016/j.ancene.2023.100377. Gaillardet, J., Braud, I., Hankard, F., Anquetin, S., Bour, O., Dorfliger, N., et al. (2018). OZCAR: The French network of critical zone observatories. Vadose Zone Journal, 17(1), 1-24, doi:10.2136/vzj2018.04.0067
Les blocs verticaux des chaos gréseux du Sud de l'Ilde-de-France: éléments de reconnaissance d'une origine anthropique
No full textNational audienceThe problem of recognizing an anthropogenic origin of upright megaclasts in Paris Basin rocky lag deposits is a delicate task. We advocate a method grounded inthe sedimentological mechanisms that could potentially produce upright rock megaclastsduring the erosion of cliffs or sandy slopes. It is a prospective approach thataims to locate, within a rocky chaos, isolated megaclasts or arrangements of rock claststhat seem to be incompatible or out of place with the sedimentological mechanismsof erosion. The analytical procedure is detailed. Several examples of upright rock clastsare analysed and exemples from the archaeological literature have been reconsideredin this context.Le problème de la reconnaissance d'une origine anthropique des blocs verticaux au sein de chaos rocheux du bassin de Paris est une question délicate. Nous proposons dans cet article une méthode qui s'appuie sur les mécanismes sédimentologiques capables de redresser les blocs rocheux lors de l'érosion des falaises ou des pentes sableuses. Elle est prospective et vise à localiser dans des chaos des blocs de grès isolés ou des arrangements de blocs qui semblent incompatibles ou incohérents avec les mécanismes sédimentologiques d'érosion. La méthode d'analyse est détaillée. Plusieurs exemples de roches verticales sont analysés et des exemples de la littérature archéologique sont reconsidérés
Endommagement microstructural d’un composite hybride thermoplastique en fatigue : corrélation qualitative avec un critère de fatigue basé sur la vitesse de déformation de fluage.
No full textInternational audienceThis study investigates the fatigue behaviour of hybrid PA66 composites reinforced with short and continuous fibres through fatigue testing and in-situ microtomography. Damage initiates in the PA66GF50 short-fibre composite, propagates as a macro-crack towards the UD reinforcement, and leads to fibre fracture, fibre-matrix debonding, and rapid crack growth. A strong correlation exists between cyclic creep stages, at macroscopic level, and damage: primary stage with negligible damage, secondary stage, which defines the cyclic creep strain-rate criterion, is characterised by the initiation and limited progression of micro-damage, and tertiary stage with macro-crack propagation. Weaknesses lie in the short-fibre composite and fibre-matrix interface, rather than between the two polyamides