HAL - UPEC / UPEM
Not a member yet
57238 research outputs found
Sort by
Cramér–Rao bound‐informed training of neural networks for quantitative MRI
No full textXiaoxia Zhang, Quentin Duchemin, and Kangning Liu contributed equally to this workInternational audienceNeural networks are increasingly used to estimate parameters in quantitative MRI, in particular in magnetic resonance fingerprinting. Their advantages over the gold standard non-linear least square fitting are their superior speed and their immunity to the non-convexity of many fitting problems. We find, however, that in heterogeneous parameter spaces, i.e. in spaces in which the variance of the estimated parameters varies considerably, good performance is hard to achieve and requires arduous tweaking of the loss function, hyper parameters, and the distribution of the training data in parameter space. Here, we address these issues with a theoretically well-founded loss function: the Cram\'er-Rao bound (CRB) provides a theoretical lower bound for the variance of an unbiased estimator and we propose to normalize the squared error with respective CRB. With this normalization, we balance the contributions of hard-to-estimate and not-so-hard-to-estimate parameters and areas in parameter space, and avoid a dominance of the former in the overall training loss. Further, the CRB-based loss function equals one for a maximally-efficient unbiased estimator, which we consider the ideal estimator. Hence, the proposed CRB-based loss function provides an absolute evaluation metric. We compare a network trained with the CRB-based loss with a network trained with the commonly used means squared error loss and demonstrate the advantages of the former in numerical, phantom, and in vivo experiments
Two-dimensional metamaterials as meta-foams for optimized surface-enhanced solar steam generation
No full textInternational audienceWe report on a new class of metamaterials, which consists of meta-foams, optimized for high performance in enhanced solar steam generation. The design of these meta-foams involves precise dimensional control of a periodic two-dimensional micro-pore array. Proof-of-concept of meta-foams efficiency is demonstrated using two fabrication technologies. The first one involves microscale plasma etching onto a nanostructured black silicon substrate. As an alternative low-cost technique, we also use 3D printing of a graphene-containing polymer material. The experimental validation shows that the best evaporation rate reaches 1.34 kg/(h·m2) under 1 sun illumination, in open-air conditions under normal temperature and relative humidity conditions of 20 °C and 58%, an unmatched value so far under similar conditions, while the theoretical limit is estimated at 1.5 kg/(h·m2). This corresponding to a 89% conversion efficiency. Experimental data concur well with the predicted performance obtained by numerical simulations. The proposed model simultaneously accounts for both heat and mass transfer, including photothermal conversion and water phase change, to guide the optimization towards the maximization of the evaporation rate, an important figure of merit in many applications, including solar heat harvesting and solar water purification
Homogénéisation numérique par Transformée de Fourier Rapide des composites avec discontinuité d’interface et des effets de couche limite en milieux multiporeux
No full textThis work concerns the modeling of the effective properties of composites and porous media within the framework of the numerical homogenization method based on Fast Fourier Transform (FFT). The first part of the work deals with the consideration of interface discontinuities between the matrix and the inclusions. The approach fits within the framework of the FFT variational approaches with an enrichment of the discretization space in order to account for the interface discontinuities. The approach is developed to simulate the jump of temperature for composites conductors (Kapitza type interface) and the jump of displacement in elastic composites (adhesive type interfaces). The method is based on the use of the periodic Green’s functions and the shape coefficients to describe the geometry of the inclusions and the interfaces. The linear system is solved by means of the iterative gradient method combined with the TFR algorithm to achieve the convolution product with the Green operator. We study the convergence and the precision of our results in the case of composites with rectangular, cylindrical and spherical inclusion shapes by comparisons with the finite element and analytical solutions. In the second part of this work, we propose to determine the effective transport pro- perties of porous media with two populations of pores (micropores and macropores). A double homogenization approach is then adopted. At the micropore scale, the flow is simulated by the Stokes equations. At the intermediate scale, Brinkman’s equations are used to describe the flow in the matrix containing the micropores and the Stokes equations is considered for the flow in the macropores. The central point of the present work concerns the identification of Brinkman constitutive parameters which govern the boundary layer effects in relation with the geometry of the microstructure. These coefficients are identified by by the comparison with full calculations simulating the Stokes flow at both the micropores and macropores scales considering biporous solid made up of aligned cylinders with circular and rectangular cross-sectionsCe travail concerne la modélisation des propriétés effectives de composites et de milieux poreux dans le cadre des techniques d’homogénéisation numérique par Transformée de Fourier Rapide (TFR). Le premier volet porte sur la prise en compte des discontinuités d’interface entre la matrice et les inclusions. L’approche s’inscrit dans le cadre des approches variationnelles des méthodes TFR avec un enrichissement de l’espace de discretisation pour prendre en compte les discontinuités d’interface. L’approche est développée pour simuler des discontinuités du champ de température pour les composites conducteurs de la chaleur (interface de type Kapitza) et des discontinuités du déplacement dans les composites élastiques (interfaces adhésives). La méthode repose sur l’utilisation des fonctions de Green du milieu périodique et des coefficients de forme pour décrire la géométrie des inclusions et des interfaces. La résolution du système linéaire repose sur le schéma itératif du gradient conjugué conjointement avec l’algorithme de TFR pour réaliser le produit de convolution avec l’opérateur de Green. On étudie la convergence et la précision des résultats pour des inclusions de forme rectangulaire, cylindrique et sphérique par des comparaisons avec la méthode des éléments finis et des solutions analytiques. Dans une seconde partie, on propose des modélisations des propriétés de transporte dans les milieux à deux échelle de porosité. Une démarche par double homogénéisation est alors adoptée. A l’échelle des micropores on simule l’écoulement par les équations de Stokes. A l’échelle intermédiaire, on utilise les équations de Brinkman pour décrire l’écoulement dans la matrice contenant les micropores et les équations de Stokes pour l’écoulement dans les macropores. Le point central du travail concerne l’identification des paramètres du modèle de Brinkman qui régissent les effets de couche limite en lien avec la géométrie de la microstructure. Ceux-ci sont obtenus par identification avec des calculs simulant l’écoulement de Stokes aux deux échelles simultanées pour des réseaux de cylindres de sections circulaire et rectangulair
Using Didactic Models to Design Adaptive Pathways to Meet Students' Learning Needs in an Online Learning Environment
No full textInternational audienc
Etat de l’art sur la présence de microplastiques dans les eaux destinées à la consommation humaine
No full textInternational audienc
SERS Tags Derived from Silver Nanoparticles and Aryl Diazonium Salts for Cell Raman Imaging
No full textInternational audienceSurface functionalization of silver nanoparticles by Raman reporters derived from aryl diazonium salts offers new opportunities for the design of Surface-Enhanced Raman Spectroscopy (SERS) labels for Cell Raman Imagin
Resilience of Social-Infrastructural Systems: Functional Interdependencies Analysis
No full textInternational audienceCritical infrastructures serve human activities and play an essential role in societies. Infrastructural systems are not isolated but are interdependent with regard to social systems, including those of public health and economic and sustainable development. In recent years, both social and infrastructural systems have frequently been in dysfunction due to increasing natural or human-made disasters and due to the internal and external dependencies between system components. The interconnectedness between social-infrastructural systems (socio-economic systems and technical-infrastructural systems), implies that the damage to one single system can extend beyond its scope. For that reason, cascading dysfunction can occur and increase system vulnerability. This article aims to study the functional interdependencies between social-infrastructural systems and to propose a methodology to analyse and improve the resilience of these systems. Combining Actor Network Theory and the Functional Models approach, the social-infrastructural Interdependence Resilience (SIIR) framework was proposed. To assess the applicability of the approach, the framework was applied to study the interdependence of a social-infrastructural system in the Nantes Metropolis. The studied system was composed of the local Highway Infrastructure (an infrastructural system) and the Emergency Medical Service (a social system). The results (1) show the feasibility of SIIR for investigating the interdependencies of two urban systems, and (2) provide a guideline for decision-makers to improve the functional interdependencies of urban systems
Covid-19 pandemic: do surgical masks impact respiratory nasal functions?
No full textInternational audienceNo abstract availabl
Promenade sémiotique et mathématique
No full textInternational audienceThe science is by essence highly semiotised. Our ways of thinking and communicating science are based on permanent transformations from one system of signs to another, such as scriptural, graphic, symbolic, oral and gestural signs. The semiotic focus studied in the book allows us to grasp part of the complexity of teaching and learning phenomena by focusing the attention of researchers on the variety of possible interpretations of signs that circulate within the science classroom.This introductory chapter introduces the reader to a semiotic reflection based on Peirce's framework, using examples, particularly in mathematics (equality, algebra, numeration, plane geometry). It illustrates the crucial role of the semiotic component of mathematics in its epistemological genesis, and the interest of a semiotic perspective in didactics.Les sciences sont par essence fortement sémiotisées. Nos façons de penser et de communiquer les sciences s’appuient sur des transformations permanentes d’un système de signes à un autre comme les signes scripturaux, graphiques, symboliques, oraux, gestuels. La focale sémiotique étudiée dans cet ouvrage permet d’appréhender une part de la complexité des phénomènes d’enseignement et d’apprentissage en centrant l’attention des chercheurs et des chercheuses sur la variété des interprétations possibles des signes qui circulent au sein de la classe de science.Ce chapitre introductif initie le lecteur à une réflexion sémiotique en appui sur le cadre de Peirce à partir d’exemples, notamment mathématiques (égalité, algèbre, numération, géométrie plane). Il illustre le rôle crucial de la composante sémiotique des mathématiques, dans leur genèse épistémologique, et l’intérêt d’une perspective sémiotique en didactique