HAL Portal UPHF (Université Polytechnique Hauts-de-France)
Not a member yet
28473 research outputs found
Sort by
Closed-form solutions for wave propagation in hexagonal diatomic non-local lattices
No full textInternational audiencePeriodic mass–spring lattices are commonly used to investigate the propagation of waves in elastic systems, including wave localisation and topological protection in phononic crystals and metamaterials. Recent studies have shown that introducing non-neighbouring (i.e., beyond nearest neighbour) connections in these chains leads to multiple topologically localised modes, while generating roton-like dispersion relations. This paper focuses on the theoretical analysis of elastic wave propagation in hexagonal diatom mass–spring systems in which both neighbouring and non-neighbouring interactions occur through linear elastic springs. Closed-form expression for the dispersion equations are derived, up to an arbitrary order of beyond-the-nearest connections for both in-plane and out-of-plane mass displacements. This allows to explicitly determine the influence of the order of non-neighbouring interactions on the band gaps, the local minima and the slope inversions in the first Brillouin zone for the considered unit cell. All analytical solutions are numerically verified. Finally, examples are provided on how non-neighbouring connections can be exploited to enhance the localisation of topologically-protected edge modes in waveguides constructed using mirror symmetric diatomic lattices constituted by two regions with different unit cell orientations. The study provides further insight on how to design phononic crystals generating roton-like behaviour and to exploit them for topologically protected waveguiding
Machine learning and deep learning applications in the automotive manufacturing industry: A systematic literature review and industry insights
No full textInternational audienceIn the context of the automotive manufacturing industry, complexity and the extensive data generated during production pose significant challenges. With ongoing technological advancements, effectively harnessing and analyzing this data has become increasingly critical. Machine learning (ML) and deep learning (DL) have emerged as powerful tools to manage complexity and leverage data for enhanced decision-making and process optimization. This systematic literature review examines the application of ML and DL in automotive manufacturing, focusing on application domains, ML/DL model mapping, current trends, and effective implementation practices. Out of 2786 articles, 257 were analyzed, revealing key research areas: equipment optimization (31%), quality enhancement (26%), supply chain optimization (21%), and production efficiency (17%). Energy management was notably underrepresented (4%), indicating a significant opportunity for advancing energy efficiency and decarbonization efforts. Additionally, the review highlighted significant challenges in data management, including data quality, integration, and interoperability issues, which critically affect the successful deployment of ML and DL technologies. Insights from the review were shared with senior management at Toyota Motor Manufacturing France, aligning closely with their strategic vision for digital transformation. Successful implementation of ML and DL hinges on three essential pillars: standardization of manufacturing processes and data, robust IoT and big data infrastructure, and comprehensive human resource development. Embracing these pillars is crucial to navigating complexity, realizing AI’s full potential, and advancing efficiency, sustainability, and innovation in automotive manufacturing
Cancer classification through the selection of genes extracted from microarray data
No full textInternational audienceIn this work we propose to compare tree methods for feature selection in the binary classification context. We focus on the case where thenumber of variables is very large and much more important than the sample size, as is the case in most microarray data. Four classificationalgorithms were selected: Decision Tree (DT), k-Nearest Neighbors (K-NN), Neural Networks (NN), and Support Vector Machines (SVM), along withthree filter-based feature selection criteria using mutual information: MIM (Mutual Information Maximization), JMI (Joint Mutual Information), andMRMR (Max-Relevance Min-Redundancy).First, we applied these classification algorithms to the microarray datasets without any preprocessing orfeature selection, allowing us to establish a baseline for assessing the impact of preprocessing and feature selection on improving classificationperformance. The second method involved classification after data preprocessing but without feature selection, which enabled us to evaluate theimpact of preprocessing on classification performance. Finally, the last method applied classification after both preprocessing and feature selection,allowing us to measure the combined impact of preprocessing and feature selection on improving classification performance
FRET-based mesoporous organosilica nanoplatforms for in vitro and in vivo anticancer two-photon photodynamic therapy
No full textInternational audienceDesign of multifunctional photoactive periodic mesoporous organosilica nanoparticles integrating chromphores for FRET-based two photon excitation fluorescence and photodynamic therapy
Partie III. Les modèles socioéconomiques des tiers-lieux associatifs
No full textInternational audienc
Lagrange’s Method and Lagrangian’s Mechanism in Maxwell’s A Treatise on Electricity and Magnetism (1873)
No full textInternational audienceJames Clerk Maxwell advanced Electromagnetic theory, while also mathematically enlarging Michael Faraday’s works. He declined Newtonian mechanics for the formulation of new conceptual frameworks because more adequate for describing a field of new phenomena such as electromagnetism; so he also went beyond the Newtonian paradigm introducing a novelty in the relationship physics–mathematics based on new magnitudes, e.g., electric charge and energy instead of mass and force. However, through exactly which terms were these advances made? In particular, in A Treatise on Electricity and Magnetism, II, Part IV, Chapter IV–VII (1873) Maxwell aimed to formulate a dynamical justification for field equations; he focused on the fact that magnetic field appeared a complete kinetic energy system. In this paper, we analyse and describe both: (a) what he called, Lagrange’s method of reducing the ordinary dynamical equations in Mécanique analytique (1788)––including––three methods used by Maxwell for expressing the kinetic energy––detecting the existence of the terms of the form Tme, and (b) the application of the Lagrangian formulating through the equations of motion of a connected system
« L'internationale urbaine en actions : professionnels de la ville et mobilisations urbaines dans l'Espagne franquiste (années 1960-1970) »
No full textInternational audienc
L'avenir de l'emprise en droit. La poursuite de la co-construction d'une notion aux impacts potentiels multiples
No full textInternational audienc
Monolayer-Defined Flat Colloidal PbSe Quantum Dots in Extreme Confinement
No full textInternational audienceColloidal 2D PbX (X = S, Se, Te) nanocrystals are innovative materials pushing the boundaries of quantum confinement by combining crystal thicknesses down to a monolayer with additional confinement in the lateral dimension. These flat PbSe quantum dots (fQDs) exhibit telecommunication band photoluminescence (1.43–0.83 eV), which is highly interesting for fiber optic information processing. With scanning tunneling microscopy/spectroscopy (STM/STS), we probe single-layer-defined fQD populations down to one monolayer, showing an in-gap state free QD-like density of states in excellent agreement with theoretical tight-binding (TB) calculations. Cryogenic ensemble spectra match STS/STM and TB calculations and exhibit the contribution of mono-, bi-, and trilayers to the photoluminescence. Comparing the electronic band gaps with the optical ones, we derive exciton binding energies as high as 600 meV for PbSe monolayers. Our results allow for a target-oriented synthesis of a new class of QDs with record binding energies and precisely tailored optical properties at technologically relevant wavelengths
Un « petit » commun : le chemin d'exploitation
No full textInternational audience(Civ. 3e, 9 janv. 2025, no 23-20.665, D. 2025. 55 ; RDI 2025. 254, obs. J.-L. Bergel ; Dr. rur. 2025, no 3, p. 18, obs. D. Lochouarn