Search

International audience; This paper addresses structured covariance matrix estimation under t-distribution. Covariance matrices frequently reveal a particular structure due to the considered application and taking into account this structure usually improves estimation accuracy. In the framework of robust estimation, the t-distribution is particularly suited to describe heavy-tailed observation. In this context, we propose an efficient estimation procedure for covariance matrices with convex structure under t-distribution. Numerical examples for Hermitian Toeplitz structure corroborate the theoretical analysis.

Dans le cadre de la simulation du comportement mécanique des grandes structures en béton armé par éléments finis, la régularisation des paramètres de traction du modèle de comportement par la méthode de Hillerborg est aujourd’hui largement acceptée. Des limites peuvent néanmoins être constatées sur des structures dont le fonctionnement est principalement gouverné par l’écrasement du béton en compression. La définition d’une énergie de fissuration en compression est encore largement discutée dans la communauté scientifique, mais l’absence de régularisation en compression peut conduire à une forte dépendance de la réponse au maillage. Cette contribution présente une nouvelle loi de comportement sur la base d’un critère de Mazars, intégrant à la fois une régularisation en traction et en compression, en modifiant la loi d’évolution de l’endommagement en compression. Son application sur des cas-tests représentatifs montre l’intérêt d’introduire une régularisation en compression pour réduire la dépendance de la solution numérique au maillage

International audience; In this paper we present an error estimator for unilateral contact problems solved by a Neumann–Neumann Domain Decomposition algorithm. This error estimator takes into account both the spatial error due to the finite element discretization and the algebraic error due to the domain decomposition algorithm. To differentiate specifically the contribution of these two error sources to the global error, two quantities are introduced: a discretization error indicator and an algebraic error indicator. The effectivity indices and the convergence of both the global error estimator and the error indicators are shown on several examples.

International audience; The reduction of pollutant emissions from internal combustion engines, and in particular greenhouse gases, must be achieved by optimizing the combustion process and choosing fuels with a low carbon impact. Heterogeneous mixture combustions such as Diesel induce a difficulty in managing of the pollutants formation even if in its principle the output efficiency remains favorable. Therefore, it is necessary to distinguish this last point, which refers to the compression ratio, in order to move towards homogeneous combustion with a high compression ratio. We then speak generically of HCCI combustion (Homogeneous Charge Compression Ignition), whose process takes place in the context of auto ignition.An important challenge of this combustion is the control of the ignition timing. This paper presents a concept to control the initiation of combustion of a reactive gas mixture with low susceptibility to auto-ignition, which meets the expectation of high efficiency, via the introduction of biomass particles. From a schematic point of view, these particles will play the role of spark plugs distributed throughout the combustion chamber, and should have an auto ignition time much lower than that of the fuel.For this purpose, a particle injector was conceived to directly inject a controlled quantity of biomass powder into the reactive mixture directly. This part is a challenging point. The reproducibility and viability of the particle injector were qualitatively validated by image analysis acquired at high frequency and high resolution. Thus, the system was integrated into an RCM (Rapid Compression Machine) chamber, allowing diphasic gas/solid biomass combustion study. This preliminary test was investigated by high-speed video on a lean diluted biomass enriched with hydrogen mixture (CH4/CO2/H2) at low fuel-air equivalence ratio (Φ = 0.6), with different auto-ignition delays in the condition of experimentation. It was shown that the direct injection of a small amount of biomass particles was able to promote its reactivity and initiate the combustion process. This study highlights the promising effects of the injection of powder to control the beginning of homogeneous combustion, by bringing locally a spot of energy, distributed in the whole chamber. The control given by a global contribution of weak energy is distinguished from the process of combustion by auto-ignition of the fuel mixture whose properties of equivalence air-fuel ratio or dilution will complete the optimization of the process.

Dimension reduction (DR) methods provide systematic approaches for analyzing high-dimensional data. A key requirement for DR is to incorporate global dependencies among original and embedded samples while preserving clusters in the embedding space. To achieve this, we combine the principles of optimal transport (OT) and principal component analysis (PCA). Our method seeks the best linear subspace that minimizes reconstruction error using entropic OT, which naturally encodes the neighborhood information of the samples. From an algorithmic standpoint, we propose an efficient block-majorization-minimization solver over the Stiefel manifold. Our experimental results demonstrate that our approach can effectively preserve high-dimensional clusters, leading to more interpretable and effective embeddings. Python code of the algorithms and experiments is available online.

International audience

International audience; Image synthesis in the context of radio interferometric data can be expressed as a signal reconstruction from incomplete Fourier measurements. Most imaging techniques for radio interferometry lie in minimizing the least square error between the reconstructed image and the observed data assuming an additive white gaussian noise. In this paper, we derive an expectation-maximization based imaging algorithm that handles the presence of outliers in the observed data. Subsequently, we propose a new generic image synthesis algorithm based on the expectation-maximization algorithm, leading to a computationally efficient method.

International audience; The extent to which the introduction of subwavelength spatial modulation of electromagnetic properties improves absorption performances is studied. The proposed absorber represents an evolution from the Salisbury screen, whereby the uniform resistive layer is replaced by a metagrating. A periodic supercell that supports only the specular reflection is first designed, and load impedances are then engineered to suppress this diffraction mode. To experimentally demonstrate the concept, four prototypes are fabricated and tested in the microwave domain around 10 GHz. Furthermore, the performances assessed by a merit factor derived from Rozanov’s bound show that the use of metagratings opens up good perspectives for improving the state of the art. Our findings can pave the way toward the development of high performance absorbers for applications across a broad frequency spectrum.

National audience; L'apprentissage de distance propose d'optimiser la distance de Mahalanobis pour des problèmes de classification. Plusieurs algorithmes associés à ce problème s'interprètent comme des estimateurs d'une matrice de covariance. Partant de ce constat, nous proposons un nouveau problème d'estimation appelé Robust Geometric Metric Learning (RGML) qui vise à estimer une matrice de covariance par classe et leur barycentre riemannien. Deux algorithmes associés à ce problème d'estimation sont développés : RGML gaussien et RGML Tyler. Ces algorithmes optimisent deux fonctions de coût sur la variété riemannienne des matrices symétriques définies positives et sa sous-variété à déterminant unitaire. Enfin, les algorithmes proposés sont appliqués sur des jeux de données réels. De bonnes performances ainsi qu'une robustesse aux données mal étiquetées sont obtenues.

National audience; -L'imagerie radar à travers murs (TWRI) est un domaine de recherche en cours qui vise à étudier l'intérieur d'un bâtiment depuis son extérieur. Dans le cadre le plus courant, il cherche à détecter ou à surveiller des cibles fixes. S'éloignant des techniques classiques de radar à synthèse d'ouverture (SAR), des problèmes de regression parcimonieuse ont été proposés pour résoudre ce problème. Ces méthodes reposent sur une étape de prétraitement où une séparation appropriée des sous-espaces mur et cibles est effectuée pour supprimer la réponse du mur obstruant la vue du radar sur la pièce. Dans ce travail, nous explorons des méthodes en une étape utilisant des méthodes de décomposition jointe en composantes de faible rang et parcimonieuse via Robust PCA (RPCA). La nouveauté est une reconstruction en une étape par un problème d'inversion structuré en produit de Kronecker pour lequel nous adaptons l'algorithme Alternating Direction Method of Multipliers (ADMM). Nous validons et comparons notre méthode sur des simulations.

International audience; Les radio-interféromètres sont des réseaux phasés produisant des images à haute résolution à partir de la matrice de covariance des mesures. L'étalonnage de tels instruments est nécessaire et constitue une tâche critique. Toutefois, les radio-interféromètres modernes représentent un nouveau défi pour la radioastronomie car ils sont plus sensibles aux interférences de radiofréquence (RFI). En effet, la présence de RFI dans le processus de calibration induit des biais dans l'estimation des erreurs instrumentales. L'objectif de cet article est de proposer une alternative pour atténuer les effets des RFI. Pour cela, nous proposons d'abord un modèle permettant de prendre en compte la présence de RFI dans les données sur plusieurs canaux de fréquences grâce à un modèle de rang faible. Nous réalisons ensuite une estimation par maximum de vraisemblance des paramètres de calibration avec un algorithme d'Espérance Maximisation Généralisée sur Espaces Alternés (SAGE). Pour cet algorithme, nous dérivons deux ensembles originaux de données complètes, pour lesquels nous obtenons des solutions explicites. Des simulations numériques montrent un gain en performance significatif sur des données corrompues par des RFI comparé aux méthodes de l'état de l'art.

National audience; Cet article a pour but de présenter un nouvel algorithme pour améliorer l'estimation des phases SAR interférométriques (InSAR) dans le contexte de séries temporelles. Basée sur l'estimateur du maximum de vraisemblance d'un modèle gaussien multivarié, l'estimation des phases InSAR est résolue en utilisant une approche de Descente en Bloc de Coordonnées (DBC). Par rapport aux approches classiques, l'amélioration principale réside dans l'estimation conjointe de la cohérence interférométrique et des phases InSAR au lieu de l'utilisation d'un estimateur annexe de cohérence, obtenue

International audience; The prediction of the load of a structure with a defect remains to this day a great challenge. In this study, we propose to highlight the harmfulness of a sharp V-notch through a comparative study of the stress to failure obtained by the TLS model and by the coupled criterion. Indeed, the harmfulness of the notch is more important when its size is higher than the characteristic length of the material. Moreover,we showed that the prediction of the stress to failure by the coupled criterion is overestimated when the size of the notch becomes smaller than the characteristic length of the material.; La prédiction de la charge critique d’une structure présentant un défaut reste à ce jour un grand challenge. Dans cette étude, nous proposons de mettre en évidence la nocivité d’une entaille aigüe en V à travers une étude comparative de la contrainte à la rupture obtenue par le modèle TLS et par le critère couplé. En effet, la nocivité de l’entaille est plus importante quand sa taille est supérieure à la longueur caractéristique du matériau. Nous avons montré que les prédictions de la contrainte à la rupture par le critère couplé sont surestimées quand la taille de l’entaille devient plus petite par rapport à la longueur caractéristique du matériau.

This paper proposes new algorithms for the metric learning problem. We start by noticing that several classical metric learning formulations from the literature can be viewed as modified covariance matrix estimation problems. Leveraging this point of view, a general approach, called Robust Geometric Metric Learning (RGML), is then studied. This method aims at simultaneously estimating the covariance matrix of each class while shrinking them towards their (unknown) barycenter. We focus on two specific costs functions: one associated with the Gaussian likelihood (RGML Gaussian), and one with Tyler's M -estimator (RGML Tyler). In both, the barycenter is defined with the Riemannian distance, which enjoys nice properties of geodesic convexity and affine invariance. The optimization is performed using the Riemannian geometry of symmetric positive definite matrices and its submanifold of unit determinant. Finally, the performance of RGML is asserted on real datasets. Strong performance is exhibited while being robust to mislabeled data., Published in EUSIPCO 2022. Best student paper award

International audience; This study focuses on the experimental characterization of the physical and mechanical properties of FGM/FGPM. These materials present a controlled variation of the properties at the thickness level. Ni/BaTiO$_3$ samples have been developed by the SPS sintering technique. The variation of thermal conductivity across the thickness was determined using the transient planar source method. However, the small size of the samples does not allow the use of standard techniques to measure the elastic constants. An alternative protocol based on the use of ultrasonic waves has been proposed to overcome this difficulty; Cette étude porte sur la caractérisation expérimentale des propriétés physiques et mécaniques de matériaux FGM /FGPM. Ces matériaux présentent une variation contrôlée des propriétés au niveau de l'épaisseur. Des échantillons Ni/BaTiO3 ont ainsi été élaborés par la technique originale de frittage SPS. La variation de la conductivité thermique dans l'épaisseur a été déterminée à l'aide de la méthode de la source plane transitoire. Cependant, la petite taille des échantillons ne permet pas d'utiliser les techniques standard pour mesurer les constantes élastiques. Un protocole alternatif basé sur l'utilisation des ondes ultrasons est donc proposé pour palier à cette difficulté.

International audience; To provide for an ever-increasing need for camouflage, experts in stealth have been envisioning for a few years to use metasurfaces and their unconventional electromagnetic properties to serve as efficient cloaking devices. To prepare for the use of surface-mounted electronic components, useful for frequency-selecting purposes, we take interest in an equivalent circuit analysis of metasurfaces. Indeed, metasurfaces using electronic components invite one to combine the passive behavior of the host structure with that of the lumped elements. The equivalent circuit analysis of the hosting surface deprived of electronic components has so far been dedicated to the case of metasurfaces constituted by rectangular or square unit cells. This work, therefore, aims at extending the former study on Cartesian unit cells to the coaxial geometry of interest. The validity of such a method is then assessed by comparison with the numerical simulation results obtained using a finite-element method.

International audience; In this paper, we consider a domain decomposition algorithm associated to a finite element method to approximate a unilateral frictionless contact problem between two elastic bodies. We present a global error estimator that takes into account of the error introduced by finite element analysis as well as the error introduced by the iterative resolution of the domain decomposition algorithm. The control of these error sources is a key point in order to introduce adaptive techniques and we propose error indicators that estimate the contribution of each source of error.

International audience; no abstract

International audience; no abstract

International audience; This paper presents a new classification framework for both first and second order statistics, i.e. mean/location and covariance matrix. In the last decade, several covariance matrix classification algorithms have been proposed. They often leverage the Riemannian geometry of symmetric positive definite matrices (SPD) with its affine invariant metric and have shown strong performance in many applications. However, this geometry assumes a zero mean. In practice, it is often estimated and then removed in a preprocessing step. This is of course damaging for applications where the mean is a discriminative feature. Unfortunately, the distance associated to the affine invariant metric for both mean and covariance matrix remains unknown. Leveraging previous works on geodesic triangles, we propose two affine invariant divergences that use both statistics. Then, we derive an algorithm to compute the associated Riemannian centers. Finally, a divergence based Nearest centroid, applied on the crop classification dataset Breizhcrops, shows the interest of the proposed framework.

International audience; This study presents a general simulation methodology to assess both full and partial composite action of steel-concrete-steel (SCS) structures. This refined methodology, using 3D finite elements, is applied to two three-point bending beams in which a different composite action is provided through variation of the number of studs. The comparison to experimental results validates the methodology and the global and local behavior can be reproduced. However, the important calculation cost reduces the use of this numerical strategy to more complex structures. A simplified methodology is proposed with 1D finite elements to represent the connectors. This modelling choice allows to greatly reduce the computational cost. It also imply a reduction of the reproduced phenomena. Particularly, local damage of the concrete core around the dowels is slightly different. Nevertheless, this strategy allows to accurately reproduce the global behavior as well as the failure modes of SCS beams with full and partial composite action.

International audience; In order to evaluate the cracking process in large reinforced and pre-stressed concrete structures, a predictive model of concrete damage with refined mesh and a nonlinear law can be required. Because of the computational load, such modelling is not applicable directly on large-scale structures whose characteristic dimensions are over several meters. To overcome this difficulty, a method based on the static condensation, called “ASC” has been proposed. It concentrates the computational effort on the damaged area, which can evolvedue to crack propagation or initiation. In order to reach a higher level of representativeness, mesh refinement is integrated into the ASC method in the proposed contribution. The method first uses a coarse initial mesh and refines it only in the domain of interest (DI) to reach a density of mesh adapted to a non-local model (order of the centimeter). This principle of refinement, integrated in the ASC method, is described. It is then applied to a notched bending beam. The use of mesh refinement combined to ASC method allow to keep results similar to a calculation with a fine meshon the entire structure. However, it also allows a saving in computational time and in memory occupation that increases with the increase in the mesh density of the refined zones.

International audience

International audience; In order to evaluate the cracking process in large reinforced and pre-stressed concrete structures, a predictive simulation of concrete damage with a refined mesh and a nonlinear law is required. Because of the computational load, such modelling is not applicable directly on large-scale structures whose characteristic dimensions are over several tens of meters. To deal with this type of structure, an adaptive static condensation method (ASC), which concentrates the computational effort on the damaged area (domain of interest) only, is proposed. The ASC method is first presented. A particular attention is paid to the needed developments to extend its domain of application up to prestressed concrete structures. The method is then applied to a three point bending prestressed beam and validated by comparison to a classical finite element damage computation. The numerical efficiency of the proposed approach is evaluated with, on this application, a time saving factor of about 15.An automatic mesh partitioning method is then implemented to make the method even more efficient. Based on physical considerations, it takes into account the expected shape and the evolution of the damaged regions during the loading. It is applied to three test cases with different geometries and loading to discuss the calibration process through a sensitivity analysis. A unique set of parameters is finally proposed and the efficiency of the method is demonstrated. As a conclusion; a complete automatic method is made available for the simulation of large scale reinforced and prestressed concrete structures

International audience

International audience; To provide for an ever-increasing need for camouflage, experts in stealth have been envisioning for a few years to use metasurfaces and their unconventional electromagnetic properties [S. B. Glybovski et al., "Metasurfaces: From microwaves to visible," 2016] to serve as efficient cloaking devices. To prepare for the use of surface-mounted electronic components, useful for frequency-selecting purposes, we took interest in an equivalent circuit analysis of metasurfaces. Indeed, the works of [R. Dubrovka and R. Donnan, "Equivalent circuit of FSS loaded with lumped elements using modal decomposition equivalent circuit method," 2011] on metasurfaces using electronic components invite one to combine the passive behavior of the host structure with that of the lumped elements. The equivalent circuit analysis of the hosting surface deprived of electronic components can be done using the works of [F. Mesa et al., "Unlocking complexity using the ECA: The equivalent circuit model as an efficient and physically insightful tool for microwave engineering," 2018] on passive metasurfaces. This project therefore aims at translating the results of [F. Mesa et al., "Unlocking complexity using the ECA: The equivalent circuit model as an efficient and physically insightful tool for microwave engineering," 2018] onto the coaxial geometry of interest here. The validity of such a method is then assessed by comparison with the simulation results obtained using a finite-element method.

International audience; Radio interferometers are phased arrays producing high-resolution images from the covariance matrix of measurements. Calibration of such instruments is necessary and is a critical task. This is how the estimation of instrumental errors is usually done thanks to the knowledge of referenced celestial sources. However, the use of high sensitive antennas in modern radio interferometers (LOFAR, SKA) brings a new challenge in radio astronomy because they are more sensitive to Radio Frequency Interferences (RFI). The presence of RFI during the calibration process generally induces biases in state-of-the-art solutions. The purpose of this paper is to propose an alternative to alleviate the effects of RFI. For that, we first propose a model to take into account the presence of RFI in the data across multiple frequency channels thanks to a low-rank structured noise. We then achieve maximum likelihood estimation of the calibration parameters with a Space Alternating Generalized Expectation-Maximization (SAGE) algorithm for which we derive originally two sets of complete data allowing closed-form expressions for the updates. Numerical simulations show a significant gain in performance for RFI corrupted data in comparison with some more classical methods.

Angular and spectral separations of thermal radiation have become a key challenge in solar concentration or thermal management of sources radiating at extremely high or low temperatures. Reflections obtained from electromagnetic theory in an open cavity geometry increase the emission and absorption compared to a flat surface due to the cavity effect. In this paper, in order to obtain the directional emission of geometric surfaces (V-Grooves) using ray tracing and studying the propagation of light, a new algorithm is developed. The numerical simulations take into account the materials properties of both facets of the V-shape, thus simulating an original asymmetric and a multilayer V-shape and providing a very interesting directive thermal emission behavior. We evaluated the emission behavior from the reflection and emission coefficients of different rays at different angles for different parameters (materials properties, wavelength, and geometry). The simulations of a V-groove showed that due to the different reflections occurring inside an aluminum V-cavity with an aperture angle of 28°, the emissivity was well enhanced by 86% in the normal direction compared to a flat surface made of the same material. Moreover, using the original asymmetric opposite-sided materials (Al and Si) in a V- groove, it was possible to separate and control the emission by focusing the radiation or directing different spectral bands in different directions.

International audience; This paper presents a new algorithm for improving the estimation of interferometric SAR (InSAR) phases in the context of time series and phase linking approach. Based on maximum likelihood estimator of a multivariate Gaussian model, the estimation of the InSAR phases is solved using a Block Coordinate Descent algorithm. Compared to the state-of-the-art approaches, the main improvement lies on the joint estimation of the covariance matrix and the InSAR phases instead of using a plug-in coherence estimate obtained from the sample covariance of the data or the modeling of the temporal decorrelation of the target under observation. Results of synthetic simulations confirm the improvement brought by the proposed estimator.

International audience; Afin d'évaluer le phénomène de fissuration dans les grandes structures en béton armé et précontraint, un modèle prédictif de l'endommagement du béton avec un maillage raffiné et une loi de comportement non linéaire peut être requis. En raison du cout de calcul, une telle modélisation est difficilement applicable à l’échelle de la structure complète. Pour résoudre cette difficulté, une méthode basée sur la condensation statique, qui concentre l'effort de calcul sur la zone endommagée, a été proposée. Dans ce travail, la méthode a été développée en y intégrant le principe de raffinement automatique du domaine d'intérêt (DI).

International audience; Change detection (CD) for remotely sensed images of the Earth has been a popular subject of study in the past decades. With the increase in the number of spatial missions with embedded synthetic aperture radar (SAR) sensors, the amount of readily available observations has now reached the "big data" era. This chapter introduces several families of elliptical distributions that can be used to model multivariate SAR images. It describes several dissimilarity functions based on covariance matrices, which are then compared for CD on the considered datasets. The chapter presents an extension of a statistical detection methodology that allows us to account for low-rank structures in the covariance matrix, whose interest is also illustrated on the real dataset. The Kullback–Leibler divergence is a popular measure between probability density functions, encountered notably in SAR change detection problems. The local covariance matrix of pixel patches appears to be relevant feature to analyze multivariate image time series.

International audience; The present work exploits a simulation-based design procedure to reduce the scattering of a metallic cylinder in desired directions using conformal sparse metasurfaces.

Self-heating phenomenonrepresents thermal evolution in materials resulting from mechanical loading. It is presentlyinvestigated during fatigue tests at 20 kHz. The aim of this article is to compare the variation of temperature, in the range 20 to 80 ∘C, of a material body due to self-heating computed from an innovative point of view: requiring a covariant formulation of the thermomechanical behavior, we propose a spacetime approach for heat conduction and heat equation based on spacetime thermodynamics. In this article, we prove the feasibility to use such a spacetime framework to solve thermal problems. Thus computational methods performed with FEniCS platform, using finite element method, lead to a numerical approach for investigating self-heating. The integral form of the spacetime problem with the adapted boundary conditions directly usable are provided for heat conduction problem. The identification process of the parameters obtained from the experimental data of fatigue tests is also particularly studied to obtain the best accuracy and reliability of the numerical simulations, presently applied to metallic material (C65 steel) and 1D beam (60 mm length). Moreover, by use of optimization techniques, some numerical values for these input parameters are obtained, especially the characteristic time (in the range 40 to 130 s) and the heat source related to thermomechanical phenomena (intrinsic dissipation in the range 0.3 to 12 ∘C ⋅ s− 1). They have been successfully used to obtain spacetime temperature distributions in comparison to experimental measurements.

The aim of the present in vitro study was to evaluate specific mechanical and physicochemical properties of two calcium silicate based sealers, (AH Plus Bioceramic “AHPB”; Well-Root ST “WRST”), and a conventional resin sealer (AH Plus “AHP”). These aims were accomplished by assessing the porosity, pH, compression strength, roughness, wettability and cell attachment of the tested materials. The results were compared statistically using the one-way ANOVA test. Higher pH values were obtained in both AHPB and WRST compared to AHP at 3, 24 and 72 h (p < 0.05). A greater level of porosity and wettability was detected for both AHPB and WRST compared to the resin sealer AHP (p < 0.05). Evident cell growth characterized by elongated morphology was observed on the surface of AHPB and WRST, while only a thin layer of cells was seen on the surface of AHP. A significant lower compression strength and modulus were obtained in the specimens created using AHPB compared to those made with AHP and WRST (p < 0.05). The removal of calcium silicates may be quite tricky during endodontic retreatment. In conclusion, considering the limitations of the present in vitro study, both calcium silicate sealers demonstrated good physicochemical properties. However, the lower compression strength and modulus of AHPB may facilitate its removal and make the retreatment procedures considerably easier. Keywords: calcium silicate based sealer; resin sealer; mechanical properties; porosity

International audience; This paper studies a statistical model for heteroscedastic (i.e., power fluctuating) signals embedded in white Gaussian noise. Using the Riemannian geometry theory, we propose an unified approach to tackle several problems related to this model. The first axis of contribution concerns parameters (signal subspace and power factors) estimation, for which we derive intrinsic Cramér-Rao bounds and propose a flexible Riemannian optimization algorithmic framework in order to compute the maximum likelihood estimator (as well as other cost functions involving the parameters). Interestingly, the obtained bounds are in closed forms and interpretable in terms of problem's dimensions and SNR. The second axis of contribution concerns the problem of clustering data assuming a mixture of heteroscedastic signals model, for which we generalize the Euclidean K-means++ to the considered Riemannian parameter space. We propose an application of the resulting clustering algorithm on the Indian Pines segmentation problem benchmark.

International audience; In this paper, robust mean and covariance matrix estimation are considered in the context of mixed-effects models. Such models are widely used to analyze repeated measures data which arise in several signal processing applications that need to incorporate possible individual variations within a common behavior of individuals. In this context, most algorithms are based on the assumption that the observations follow a Gaussian distribution. Nevertheless, in certain situations in which the data set contains outliers, such assumption is not valid and leads to a dramatic performance loss. To overcome this drawback, we design an expectation-conditional maximization either algorithm in which the heterogeneous component is considered as a part of the complete data. Then, the proposed algorithm is cast into a parallel scheme w.r.t. the individuals in order to mitigate the computational cost and a possible central processor overload. Finally, the proposed algorithm is extended to deal with missing data which refers to the situation where part of the individual responses are unobserved. Numerical simulations are conducted to assess the performance of the proposed algorithm regarding robust regression estimators, probabilistic principal component analysis and its recent robust version.

Steel-Concrete-Steel (SCS) structures are sandwich composite structures combining steel plates and a concrete core through a connection system. Their high strength, stiffness, durability, sustainability and modularity make them an increasingly competitive choice of structure for the construction of high-rise buildings, submerged tunnels or even in the nuclear field. To enable their use and to better understand, characterize and predict their behavior, numerical studies are conducted. This work presents a refine methodology to modelling SCS structures. The developed strategy allows, among other, to finely reproduce the response of the concrete around the connectors, an area in which this material is exposed to significant localized compression and shear stresses, as well as the shear behavior of the connectors, one of the phenomena at the origin of the failure of the connection system. Applied first to a push-out test, then to SCS beams in bending, these modelling allow to observe the variations of behavior that can be seen between full and partial composite action SCS structures. Based on this refine modelling, a simplified modelling strategy is also developed and applied to the study of SCS structural connection. This study allows to analyze the response and to identify the weak points of this type of complex structure.; Les structures « Steel-Concrete-Steel » (SCS) sont des structures composites combinant des plaques d’acier et un cœur de béton à l’aide d’un système de connexion. Leur fortes résistance, rigidité, durabilité, étanchéité et leur modularité en font un choix de structure de plus en plus concurrentiel pour la construction d'immeubles de grande hauteur, des tunnels submergés ou encore dans le domaine du nucléaire. Pour permettre leur utilisation et afin de mieux comprendre, caractériser et prédire leurs comportements, des études numériques sont réalisées. Ces travaux présentent une méthodologie fine de modélisation des structures SCS. La stratégie développée permet entre autres, de reproduire finement la réponse du béton autour des connecteurs, zone dans laquelle ce matériaux subit d’importantes sollicitations de compression et cisaillement localisées, ainsi que le cisaillement des connecteurs, l’un des phénomènes à l’origine de la ruine de la zone de connexion. Appliquée dans un premier temps à des essais push-out, puis à des poutres SCS en flexion, ces modélisations permettent d’observer les variations de comportement observables entre des structures SCS à action composite complète et partielle. À partir de cette modélisation fine, une stratégie de modélisation simplifiée est également développée et appliquée à l’étude de jonction d’ossature SCS. Cette étude permet d’analyser la réponse et d’identifier les points faibles de ce type de structure complexe.

The densification and grain growth behavior of a ZnS nanopowder during spark plasma sintering (SPS) were investigated under the pressure of 125 MPa and constant heating rate of 25 °C min−1. At a sintering temperature of 550 °C, almost fully dense ZnS ceramics (98.2%) could be obtained with grain size as low as 75 nm. At higher temperature, ZnS is subject of a very important grain growth that is detrimental to the mechanical properties. A maximum Vickers microhardness of 4.2 GPa was measured, which is 70% higher than the current state of the art. Retaining the grain size below 100 nm appears to be the key to developing competitive ZnS functional ceramics.

International audience; In this paper, we propose a new estimator of the covariance matrix parameters when observations follow a mixture of a deterministic Compound-Gaussian (CG) and a white Gaussian noise. In particular, the covariance matrix of the CG contribution is assumed to be expressed as the Kronecker product of two low-rank matrices, which is a structure often involved in MIMO array processing. The proposed estimator is then obtained by maximizing the likelihood of the data with the use of a specifically tailored block Majorization-Minimization (MM) algorithm. Finally, the method is evaluated in terms of adaptive filtering on a MIMO-STAP radar setting, showing important improvements over standard processing.

International audience; In order to evaluate the cracking process in large reinforced and prestressed concrete structures, a predictive model of concrete damage with refined mesh and a nonlinear law can be required. Because of the computational load, such modelling is not applicable directly on large-scale structures whose characteristic dimensions are over several meters. To solve this problem, a method based on the static condensation, which concentrates the computational effort on the damaged area has been proposed. The so-called “zones of interest” are evolutionary to be adapted to the cracking process. The method has been developed, improved and implemented in the finite element code Cast3m. It follows different steps including an initial cutting of the mesh into sub-zones, a condensation of these zones on their borders and the definition of the first “zone of interest” by a preliminary linear computation. The borders of « non-risky » zones are then condensed at the interface of this initial zone of interest (“double condensation”). The nonlinear computation is finally applied only on the zone(s) of interest, consequently reducing the computational load. At the end of each calculation time, a potential evolution of the zones of interest is considered through either criteria of “propagation” or “apparition” (evolving method). In this contribution, the “adaptive condensation method” is presented and the influence of the input parameters is discussed. Among them, the initial decomposition of the mesh into sub-zones plays a key role in the efficiency of the method. In order to improve the initial method, an automatic division procedure is proposed to enhance the initial mesh cutting. It is especially shown how this procedure, based on a limited number of parameters, can significantly decrease both the user effort and the computational time. The method is finally applied on test cases of different complexity: a 2D three-point bending beam, a simplified reinforced concrete containment and a prestressed beam. Both global and local results are compared to “reference” simulation using classical finite element methods. Similar results are obtained, even for very local quantities (damage), with a variable calculation time saving factor ranging from 4 to 15 for the case of the prestressed beam.

The purpose of the present study was to evaluate the physicochemical properties and antibacterial activity of three calcium silicate cements. Mineral trioxide aggregate (MTA Biorep "BR"), Biodentine (BD) and Well-Root PT (WR) materials were investigated using scanning electron microscopy (SEM) at 24, 72 and 168 h of immersion in phosphate buffered saline (PBS). The antibacterial activity against Enterococcus faecalis (E. faecalis), the solubility, roughness, pH changes and water contact angle were also analyzed. All results were statistically analyzed using a one-way analysis of variance test. Statistically significant lower pH was detected for BD than WR and BR (p < 0.05). No statistical difference was found among the three materials for the efficacy of kill against E. faecalis (p > 0.05). Good antibacterial activity was observed (kill 50% of bacteria) after 24 h of contact. The wettability and the roughness of BR were higher than for the other cements (p < 0.05). BD was more soluble than WR and BR (p < 0.05). In conclusion, the use of bioceramic cements as retrograde materials may play an important role in controlling bacterial growth and in the development of calcium phosphate surface layer to support healing. Moreover, the premixed cement was easier to use than powder-liquid cement. Keywords: calcium silicate cement; powder–liquid cement; premixed cement; retrograde materials.

International audience; The information geometry of the zero-mean tdistribution with Kronecker-product structured covariance matrix is derived. In particular, we obtain the Fisher information metric which shows that this geometry is identifiable to a product manifold of S ++ p (positive definite symmetric matrices) and sS ++ p (positive definite symmetric matrices with unit determinant). From this result, we obtain the geodesics and the Riemannian gradient. Finally, this geometry makes it possible to propose an on-line covariance matrix estimation algorithm well adapted to large datasets. Numerical experiments show that optimal results are obtained for a reasonable number of data.

Sandwich structures are widely used in many industrial applications and especially in light aviation. The local buckling phenomenon named “wrinkling” is one of the primary causes of compressive failure of such structures. Its calculation is a difficult practical problem since this phenomenon cannot be captured by the GFEM (Global Finite Element Model) classically used for aircraft structure sizing. Therefore, pre-sizing involves the use of a wrinkling model, which can be found in the literature. In practice, such models are used with high safety factors by the industry. This paper proposes an evaluation of analytical wrinkling formulas in an industrial setting. Realistic applications involve a framework (3D stress state, orthotropy, skin asymmetry) far from the assumptions on which most of the analytical formulations are based. The case study is a sandwich composite beam subjected to uni-axial compressive load. Limitations and assumptions of the analytical wrinkling models studied are quantified and a discussion on the relevance of using simple formulas for the design of sandwich structured composites is developed.

We consider the problem of jointly estimating the location and scatter matrix of a Compound Gaussian distribution with unknown deterministic texture parameters. When the location is known, the Maximum Likelihood Estimator (MLE) of the scatter matrix corresponds to Tyler’s M-estimator, which can be computed using fixed point iterations. However, when the location is unknown, the joint estimation problem remains challenging since the associated standard fixed-point procedure to evaluate the solution may often diverge. In this paper, we propose a stable algorithm based on Riemannian optimization for this problem. Finally, numerical simulations show the good performance and usefulness of the proposed algorithm.

International audience; In this article, the active vibration control of a smart structure made out of a single functionally graded piezoelectric material layer, equipped with a network of discrete electrodes, is studied. The material properties vary continuously across the direction of thickness, so that top and bottom surfaces consist of pure PZT4 and the mid surface is composed of pure aluminium. The percolation phenomenon is taken into account. A functionally graded piezoelectric material plate finite element based on the first-order shear deformation theory hypothesis and layer-wise approximation for electric potential is implemented. An optimization procedure is considered to define the relevant electrodes for actuators and sensors, based on controllable and observable criteria. An adaptative fuzzy controller system is used, activating with relevance the actuators according to the most excited eigenmodes. Simulations show the effectiveness of this kind of concept.

International audience; This paper proposes a new strategy based on the proper generalized decomposition method to solve the forced vibration problem in two-dimensional (2D) layered beams. The approach redefines the problem by approximating the displacement field as a sum of separated functions of x, z, and ω and proposes an iterative algorithm that solves three 1D problems at each iteration. Several numerical tests with different slenderness ratios and boundary conditions are considered to validate the method. The results in terms of modal parameters and frequency response functions are analyzed and compared with exact elasticity solution and finite element simulations with very good agreements.

International audience; A beam model is developed for the static analysis of composite steel-concrete beams and twin-girder decks which takes into account the effects of shear lag phenomenon as well as the interfacial slip between the concrete slab and steel joists. A sinus distribution is considered for the axial displacements of the concrete slab and steel joists. This kinematic represents the transverse shear strain in the concrete slab and steel joists using a cosine function. The considered kinematic satisfies the boundary conditions of transverse shear stress on the top and bottom surfaces and does not need any shear correction factor. The shear lag effects are introduced by prescribing the warping shape of the slab cross-section and taking its intensity as axial unknowns. A conforming three-nodded 25-dof beam element is derived for numerical simulations. The presented model has been validated by comparing the obtained results with elasticity solutions, 3D finite element results and other analytical and numerical models available in the open literature.Keywords: Interfacial slip; Shear lag; Sinus model; Steel-concrete beams; Twin-girder decks

Hypothesis: The setting time and mechanical properties of cements are a major technical concern for a long time in civil engineering. More recently those practical problems became a major concern for biomedical applications -in bone surgery and in dentistry- in particular concerning the setting time which should be minimized. The possibility to add organic additives to interact with the different constituting ions in cements constitutes a way to modify the setting kinetics. We made the assumption that a hydrolysable polyphenol like tannic acid could modify the setting time and the physical properties of Mineral Trioxide Aggregate (MTA). Experiments: Tannic acid is added in variable proportions to the water used to set MTA. The formation of the hybrid organic-mineral cements is investigated using a combination of structural, chemical and mechanical methods. X-ray tomography was also used to investigate the changes in porosity and pore size distribution upon incorporation of tannic acid in MTA based cements. The hydrophilicity of the cements was evaluated by measuring the permeation kinetics of small water droplets. Findings: We found that tannic acid allowed to reduce markedly the setting time of MTA based cements. The obtained cements have an increased hydrophilicity and display excellent resistance to compression. The number of pores but not the average pore size is also affected. The possible roles of tannic acid in modifying the cement properties are discussed. Keywords: Mechanical properties; Mineral trioxide aggregate; Natural additives; Setting time; Tannic acid; X-ray tomography.

International audience; Many anticounterfeiting tags rely on images intricate in different color and/or different polarization states to create optical effects invisible in conventional observation conditions. We build here an imaging micropolarimeter based on the dual rotating compensator system and using the color capabilities of cameras to yield the full polarimetric information for any color coordinates in one acquisition step. After correction for the camera response, the full color image reconstructed from the polarimetric response is validated by the comparison with images obtained in an optical microscope fitted with polarizers.

In this paper, robust mean and covariance matrix estimation are considered in the context of mixed-effects models. Such models are widely used to analyze repeated measures data which arise in several signal processing applications that need to incorporate a same global individuals behavior with a possible individual variations. In this context, most algorithms are based on Gaussian assumption of the observations. Nevertheless, in certain situations in which there exist outliers within the data set, such assumption is not valid and leads to a dramatic performance loss. To overcome this drawback, we design an expectation-conditional maximization either algorithm in which the heterogeneous component is considered as part of the complete data. Then, the proposed algorithm is cast into a parallel scheme, w.r.t. the individuals, in order to alleviate the computational cost and a possible central processor overload. Finally, the proposed algorithm is extended to deal with missing data which refers to the situation where part of the individual responses are unobserved. Numerical simulations are performed in order to assess the performance of the proposed algorithm in regard to robust regression estimators, probabilistic principal component analysis and its recent robust version.