Portail des publications scientifiques IMT Mines Alès
Not a member yet
5198 research outputs found
Sort by
Determining the kinetic of the oxidation of metallic solid particles
No full textInternational audienceThe use of non-nano iron particles is currently under investigation for several industrial applications, including carbon-free power production. Nevertheless, poor understanding from the fundamental point of view has been reached on the phenomena ruling the oxidation, with detrimental effects on the performance and safety of reactive systems. In this perspective, this work presents an experimental campaign investigating the oxidation of iron particles in a quiescent regime, exposed to external heat fluxes. Different boundary conditions were explored to this scope, monitoring the sample mass and heat release rate along with the tests. The acquired measurements were integrated with Scanning Electron Microscopy analyses. A mechanicistic model identifying the production of FeO, Fe3O4, and Fe2O3 was proposed and corroborated by the comparison of the collected data, showing excellent agreement between the enthalpy of reactions deriving from experiments and literature. The variations in oxidized structures lead to a regime ruled by intraparticle diffusion of oxygen, once large diameters are explored. The investigation of the kinetic and ignitability of iron powder allows for a technical-based approach for a proper, optimized, and safe design of reactive and storing systems involving iron and iron oxide particles, debottlenecking their use at industrial scales
Exploring the Role of Stigma and Reduced Facial Expressions in Shaping First Impressions of Individuals with Schizophrenia
No full textInternational audienceAbstract Background and Hypothesis First impressions are rapid judgments based on limited information, significantly influencing one’s willingness to interact. Social difficulties in schizophrenia may stem from negative first impressions formed by healthy individuals. We hypothesized that healthy individuals would quickly develop negative first impressions of those with schizophrenia, driven in part by reduced facial expressions and stigma. Study Design Twenty individuals with schizophrenia and 20 age- and sex-matched healthy controls were filmed responding to a question for 30 seconds. Three hundred participants from the general population rated their first impressions of the 40 video recordings. The diagnosis of individuals with schizophrenia was either retained, disclosed, or misrepresented as autistic. To isolate the effect of facial expressions, stimuli were presented without sound using a point light display. Study Results Individuals with schizophrenia received more negative first impression ratings than healthy controls, and higher negative symptoms were linked to poorer first impression ratings. OpenFace analysis confirmed that facial expressions significantly shaped impressions, particularly mouth movement. Additionally, ratings worsened when participants’ schizophrenia diagnosis was disclosed, highlighting the harmful impact of stigma. Conversely, when mislabeled as having autism, first impression ratings improved. Conclusions This study highlights the compounded negative effects of facial expressions and schizophrenia-related stigma on first impressions. Given their significant influence on social interactions, future research should explore strategies to mitigate these effects
TREB: Temporal Refinement of Egocentric Body Pose
No full textInternational audienceAccurate egocentric body-pose estimation is essential for delivering immersive experiences in Virtual, Augmented, and Mixed Reality (VR/AR/MR) applications. A major challenge in this setting arises from the limited field of view provided by head-mounted cameras, which often leads to self-occlusion and poor visibility of certain body joints—particularly in the lower body. Current state-of-the-art methods typically perform pose estimation frame-by-frame in a view-aligned manner. While this is computationally efficient and performs well for visible joints, it struggles to estimate occluded joints due to the lack of contextual information. In this work, we explore the role of temporal information in improving joint estimation. We propose a lightweight model that refines pose predictions by leveraging a large temporal window. Our results show that incorporating such module improves the accuracy of joint estimation without incurring a large computation overhead
Interbrain synchrony trajectories during teamwork
No full textInternational audienceTheoretical models posit that teamwork, the social interaction in which individuals collaborate to reach a common goal, matures and progresses during a task episode. Due to the limitations of the traditional psychological tools (i.e. questionnaires), this evolution of teamwork has never been verified. As inter-brain synchrony (IBS), the degree of similarity between individuals' brain activity, appears to be a relevant measure of teamwork, we used IBS to examine its trajectories during a task episode. We hypothesised that IBS would increase over time. Forty dyads completed a 30-minute cooperative task with asymmetric roles. fNIRS hyperscanning recorded brain activity in six regions within the prefrontal and temporoparietal areas. IBS was quantified using wavelet transform coherence. Linear mixed models, with time as a fixed effect, slope and intercept as random effects for dyads tested the effect of time on IBS. Control analyses using permuted dyads confirmed IBS appearance during teamwork. No significant effect of time on IBS was found; however, there was significant heterogeneity in IBS trajectories between teams. There is no overall increase or decrease in IBS in the sample. Further research is needed to understand why some teams show an increase in IBS over time, while others show a decrease
Organisations alternatives : faire face aux injonctions contradictoires
No full textInternational audienceSession thématique ouverte.Cette session vise à mettre en lumière et enrichir l’analyse des tensions contradictoires qui peuvent traverser les organisations dites « alternatives » (société à mission, entreprise de l’ESS, associations, etc.) : autonomie vs. contrôle, financiarisation vs. logique de long terme, réindustrialisation vs. concurrence des territoires, soutenabilité forte vs. greenwashing, etc.• Est-il possible de faire émerger une typologie de ces injonctions contradictoires ?• Quels outils conceptuels transdisciplinaires pour les éclairer ? Comment les tensions que l’on peut théoriser au niveau macro sont-elles concrètement vécues par les acteurs ?• Comment les organisations alternatives inventent de nouvelles pratiques pour y faire face
Digital Twin engineering guided by MBSSE and Digital Thread
No full textInternational audienceThe engineering of complex systems produces largevolumes of heterogeneous data from multiple stakeholders. Theyare gathered throughout the system lifecycle, is organized withina Digital Thread, ensuring traceability and consistency. Thisdata can be reused during the operational phase through aDigital Twin (DT), which serves purposes such as control,optimization, and maintenance. However, developing a DTremains costly, even though valuable models and data arealready available in the Digital Thread. This highlights the needto capitalize on existing resources to improve efficiency. Thiswork extends previous research in Digital Twin engineering byexploring the link between Digital Thread and Digital Twin. Itproposes a methodology and case study based on systemsengineering and Model-Based Systems Engineering to supportcoherent and reusable Digital Twin System design
Data Leakage and Deceptive Performance: A Critical Examination of Credit Card Fraud Detection Methodologies
No full textInternational audienceThis study critically examines the methodological rigor in credit card fraud detection research, revealing how fundamental evaluation flaws can overshadow algorithmic sophistication. Through deliberate experimentation with improper evaluation protocols, we demonstrate that even simple models can achieve deceptively impressive results when basic methodological principles are violated. Our analysis identifies four critical issues plaguing current approaches: (1) pervasive data leakage from improper preprocessing sequences, (2) intentional vagueness in methodological reporting, (3) inadequate temporal validation for transaction data, and (4) metric manipulation through recall optimization at precision’s expense. We present a case study showing how a minimal neural network architecture with data leakage outperforms many sophisticated methods reported in literature, achieving 99.9% recall despite fundamental evaluation flaws. These findings underscore that proper evaluation methodology matters more than model complexity in fraud detection research. The study serves as a cautionary example of how methodological rigor must precede architectural sophistication, with implications for improving research practices across machine learning applications. Compared to several recent studies reporting near-perfect recall (often exceeding 99%) using complex deep models, our corrected evaluation with a simple MLP baseline yields more modest but reliable metrics, exposing the overestimation common in flawed pipelines
Impact du vieillissement hygrothermique sur le comportement en fatigue des matériaux composites pultrudés à forte fraction volumique de fibres
No full textNational audienceAujourd'hui, le secteur des transports traverse de profondes transformations, principalement guidées par l'impératif de réduire l'empreinte carbone. Dans ce contexte, le projet INFRAlight a pour objectif de concevoir une infrastructure ferroviaire en composite pultrudé conduisant ainsi à la réduction de l'empreinte carbone. Pour remplir cet objectif, deux matériaux sont envisagés. Le premier est composé d'une matrice de polyamide 6 (PA6) et de fibres de verre et le second est composé d'une résine méthacrylate (Elium C195E) et de fibres de verre. Plusieurs études ont démontré que le polymère Elium est plus durable que le PA6 dont la sensibilité à l'humidité conduit à une chute de ses propriétés mécaniques. Dans cette étude, un vieillissement hygrothermique accéléré à 50 °C et 80 %hr a été réalisé afin d'étudier son effet sur le comportement en fatigue des matériaux pultrudés. Les premiers résultats de fatigue réalisés sur les composites pultrudés non vieillis ont montré que les composites à haute fraction volumique de fibres présentent une durée de vie en fatigue significativement courte
SkeltyMLOps: A Reference Architecture for Collaborative MLOps
No full textNational audienceWith the increasing adoption of machine learning (ML), the need to manage, in a tailored manner, the complete lifecycle of this new type of software arises. MLOps (Machine Learning Operations) extends DevOps to address the specific challenges of development and life-long management of ML-powered software. Despite the increasing attention to this domain, practitioners still lack a unified, accessible, modular, and open-source architecture to support MLOps projects. To address this gap, this paper proposes SkeltyMLOps, a reference architecture designed to promote collaboration between diverse actors involved in MLOps processes. The architecture is derived from a thorough literature review, through which we extracted and clustered, thanks to LLMs, a comprehensive list of MLOps actors and activities. These clusters were then used to guide the architectural decomposition and component design of SkeltyMLOps. The originality of our proposed reference architecture is (i) that it clearly aligns its components with the responsibilities of MLOps actors, covering the identified dimensions of MLOps processes, and (ii) mediates collaboration by orchestrating interactions between the different types of actors. This architecture provides a foundation for building collaborative, actor-centric ML-powered software