HAL Portal UTC Université de Technologie de Compiègne
Not a member yet
11652 research outputs found
Sort by
Panneaux non-tissés en fibres naturelles pour l'absorption acoustique : caractérisation et prédiction en fonction de la masse volumique
No full textMatériaux poreux et métamatériaux acoustiques; GABE - Acoustique du Bâtiment et de l'Environnement: GAPSUS - Acoustique Physique, Sous-Marine et Ultra-Sonore: GVB - Vibro acoustique et Contrôle du BruitNational audienceCe travail compare les propriétés acoustiques de panneaux non tissés en fibres naturelles de posidonie, d'Alfa et de chanvre. La performance acoustique est étudiée en fonction de la densité et de l'épaisseur des panneaux par des approches expérimentales et de modélisation. Tout d'abord, la porosité et la résistivité au passage de l’air sont déterminées expérimentalement pour différentes densités et comparées avec les valeurs extrapolées à partir des propriétés à plus faible densité. Des mesures du tube d'impédance sont ensuite effectuées pour déterminer le coefficient d'absorption acoustique en incidence normale et pour valider la pertinence d’un modèle de fluide équivalent pour ces matériaux. Pour permettre une comparaison à iso-densité et iso-épaisseur, le modèle de fluide équivalent est alimenté par une porosité et une résistivité extrapolées par le modèle analytique présenté dans la partie précédente. Les propriétés d'absorption acoustique de panneaux de 40 mm et 80 mm d'épaisseur pour des densités de 40 kg.m-3 et 80 kg.m-3 sont ainsi comparées pour les trois matériaux. Nous montrons que les fibres naturelles de posidonie, d'Alfa et de chanvre ont des performances acoustiques très proches et peuvent contribuer au développement de matériaux d'absorption acoustique plus durables
Strain energy harvesting performances of piezoelectric composite transducers in rolling tires: a numerical study
No full textTire-pavement contact is responsible for mechanical strain energy which can be used to power autonomous smart devices in the context of automotive industry. Power harvesting capabilities of piezoelectric composites included in the host structure are numerically evaluated from finite element (FE) simulations. The transducer consists of a network of rectangular piezoelectric parallelepipeds embedded in a soft polymer matrix. At the local scale of the composite, optimal transducer geometries are identified with a brute force research. The procedure highlights a large sensitivity to the mechanical load, that justifies an analysis of the intrusiveness of the sensor integration in the host structure. Computations are performed at the tire scale to discuss the influence of the sensor on the strain energy distribution. The results show that a trade-off has to be made between the harvesting efficiency and the changes in mechanical behaviour of the host structure. These variations are considered with care in view of preserving the functional properties of the tire.</div
Hierarchical Copula-based Conformal Prediction and Exact Validity via Nested Prediction Regions
No full textInternational audienceEmpirical, Archimedean and vine copulas have been repeatedly investigated and leveraged to infer conformal prediction regions for multivariate predictions, but they do not provide finite-size guarantees when the estimated copula is biased or misspecified. To address this limitation, we start with copula-based conformal prediction regions that are always nested and we leverage this property to counteract this copula-estimation bias, via an additional conformal re-calibration step. Furthermore, we introduce a simpler class of semi-parametric copulas (i.e., hierarchical Archimedean copulas) as an alternative to the more complex vine copulas for which incorporating prior knowledge is difficult. Using synthetic data sets, we compare biased and debiased copula-based conformal prediction methods, and we report the impact of the data size and the impact of the number of output dimensions. Using real data, we leverage prior knowledge via this simpler class of copulas. In these experiments, we observe that this additional re-calibration step effectively eliminates the estimation bias of empirical and semi-parametric copulas when its computations are precise (enough) and the data size is large enough. The debiased hierarchical Archimedean copulas yield performances that are comparable to the results of debiased vine copulas
Multivariate spatial conditional U -quantiles: a Bahadur–Kiefer representation
No full textInternational audienceQuantiles constitute a core concept in probability theory and theoretical statistics, providing an indispensable instrument in a wide array of applications. Although the univariate notion of quantiles is intuitively clear and mathematically well established, extending this concept to a multivariate framework poses significant theoretical and practical challenges.A well-established approach to extending univariate quantiles to the multivariate setting is the \textit{spatial} (or \textit{geometric}) framework, whose empirical counterparts exhibit notable robustness and admit an elegant Bahadur-Kiefer representation. Independently, another generalization of univariate quantiles leads to \textit{-quantiles}, which naturally encompass classical estimators such as the Hodges--Lehmann estimator for a central tendency. In this study, we bridge these perspectives by introducing \textit{multivariate conditional spatial -quantiles} and deriving their corresponding Bahadur-Kiefer representation. This representation enables us to establish fundamental theoretical properties, including weak convergence and a law of the iterated logarithm. These results are proved under some standard structural conditions on the Vapnik-Chervonenkis classes of functions and some mild conditions on the model. The uniform limit theorems discussed in this paper are key tools for further developments in data analysis involving empirical process techniques
Multi-scale mechanical properties of Al–Mg–Zr–Sc alloys fabricated by direct metal laser sintering: Towards single-material composites
No full textInternational audienceThis study proposes a methodology for experimentally deriving parameter-dependent mechanical properties of DMLS-printed aluminum. Eighty-five specimens were produced using seventeen distinct parameter sets, designed through a Design of Experiments approach, and subjected to non-destructive and destructive testing. The results reveal a strong correlation between printing parameters, porosity, and mechanical performance, with average pore areas ranging from 0.037% to 21.161% and varied pore types (spherical pores, keyhole pores, and lack of fusion). Correspondingly, mechanical properties showed a broad range: tensile strength (105.3 MPa–459.0 MPa), Young's modulus (23.38 GPa–69.77 GPa), and yield strength (105.2 MPa–444.8 MPa). Single-material composites were fabricated by integrating dense and porous structures within a single geometry. Tensile testing of these composites showed that geometry and the cross-sectional ratio of ductile material influence mechanical properties significantly (tensile strengths ranging from 89.8 MPa to 112.2 MPa), with stress-strain responses displaying atypical behavior due to the brittle, porous matrix and embedded ductile truss-lattice structures
Limit theorems for conditional U-statistics analysis on hyperspheres for missing at random data in the presence of measurement error
No full textInternational audienceMissing data and measurement errors are prevalent challenges in modern statistical analyses,mainly when observations lie on complex structures like unit hyperspheres. To address theseissues, we introduce a comprehensive framework for conditional U -statistics of general order,tailored explicitly for data missing at random and contaminated by measurement errors insuch settings. We propose a novel deconvolution method for these conditional U -statistics and,for the first time, investigate its convergence rate and asymptotic distribution. Our unifiedapproach establishes general asymptotic properties under broad model conditions, enabling usto derive asymptotic confidence intervals based on the estimator’s distribution. To demonstratethe practical significance of our framework, we provide new insights into the Kendall rankcorrelation coefficient and address discrimination problems
Gender recognition with aging using HD-sEMG signals
No full textInternational audienceAs gender recognition is key in advancing personalized medicine, this study explores the use of high-density surface electromyography (HD-sEMG) signals for gender recognition during the Sit-to-Stand (STS) exercise, utilizing a combination of time, frequency, and time-frequency domain features with machine learning classifiers. A comprehensive methodology is presented, including signal preprocessing, feature extraction, and classification through conventional classifiers (K-NN, SVM, LR, DT, RF) and a hybrid CNN-KNN model, leveraging the Stockwell Transform for time-frequency image representation of the signal. Data from 64 participants across five age groups were analyzed using a 5-fold cross-validation process to ensure robustness. The CNN-KNN model achieved the highest accuracy of 99.08% ± 1.12, significantly outperforming traditional models. Additionally, the study highlights the impact of aging on gender recognition, underscoring the importance of age-aware models for accurate predictions. This work demonstrates the potential of HD-sEMG signals for both clinical and biometric applications involving gender and age-specific analysis
Development of a Smart Renewable Energy-Based System for the Automation of Microclimate Management in Poultry Farms in West Africa: Application of the Tak-Avipack1 Prototype to the Beninese Context
No full textInternational audienceAvailability and optimization of energy consumption are essential for the success of poultry activities. This is a strategic problem for food security in Benin and more broadly in West Africa. This article presents Tak-Avipack 1, an intelligent system designed to ensure availability at lower cost of energy while guaranteeing the main functionalities necessary for the adequate development of poultry: thermal regulation, lighting, hygiene and biosecurity etc. Based on an integrated IoT architecture, Tak-Avipack 1 incorporates environmental sensors (temperature, humidity, NH3, CO2, CO, PM2.5), a high-efficiency catalytic gas heater and dynamically controlled LED lighting. Its food is provided by three energy sources: photovoltaics, the conventional network (if available) and gas (which can be butane or biogas). These systems are optimally sized, and their intelligent hybridization guarantees continuous operation in rural areas. A local decision-making algorithm adjusts thermal parameters, air and lighting flows in real time, minimizing energy consumption. With the GSM / GPRS resilient connectivity and an offline mode with local storage, the system remains functional in the absence of a network. An economic assessment carried out on a model farm with 1,000 weighted hens shows a return on investment of less than six months, with an expected increase of 15% of egg production and a 20% reduction in mortality. Tak-Avipack 1 thus represents an appropriate, accessible and scalable solution to support the transition to tropicalized poultry cultivation
The Scientific and Technological Contribution of African FabLabs: Innovations, Impacts and Perspectives
No full textInternational audienceThis study examines the scientific and technological contribution of African within the context of local development. Through a mixed approach combining literature review, questionnaire surveys and semi-structured interviews, it analyzes the activities of FabLabs located in sub-Saharan Africa (Madagascar, DRC, Togo, Burkina Faso) between May 2018 – July 2024. The results reveal a diversity of equipment (3D printers, CNC, electronic tools), software (Fusion 360, KiCad, Arduino) and innovative projects in the fields of energy, health, agriculture and education. Initiatives such as SolarFocus, Plastikôo or W.Afate illustrate a frugal and sustainable approach, based on recycling and adaptation to local resources. In parallel, several FabLabs have produced scientific publications on educational commons, digital inclusion and low -tech models. Despite structural constraints (limited access to electricity, precarious funding, weak institutional recognition), these FabLabs promote technical skills acquisition, professional integration, and social innovation. Their pedagogical model, based on practice and interdisciplinarity, offers an alternative to traditional education systems. To strengthen their impact, the study recommends the creation of a pan-African network, integration into public policies, and the development of South-South partnerships positioning FabLabs as catalysts of technological sovereignty and community resilience
Revisiting the Attacker’s Knowledge in Inference Attacks Against Searchable Symmetric Encryption
No full textInternational audienceEncrypted search schemes have been proposed to address growing privacy concerns. However, several leakage-abuse attacks have highlighted some security vulnerabilities. Recent attacks assumed an attacker’s knowledge containing data “similar” to the indexed data. However, this vague assumption is barely discussed in literature: how likely is it for an attacker to obtain a “similar enough” data?Our paper provides novel statistical tools usable on any attack in this setting to analyze its sensitivity to data similarity. First, we introduce a mathematical model based on statistical estimators to analytically understand the attackers’ knowledge and the notion of similarity. Second, we conceive statistical tools to model the influence of the similarity on the attack accuracy. We apply our tools on three existing attacks to answer questions such as: is similarity the only factor influencing accuracy of a given attack? Third, we show that the enforcement of a maximum index size can make the “similar-data” assumption harder to satisfy. In particular, we propose a statistical method to estimate an appropriate maximum size for a given attack and dataset. For the best known attack on the Enron dataset, a maximum index size of 200 guarantees (with high probability) the attack accuracy to be below 5%