HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Si les gares sont des écosystèmes digitaux, pourquoi ne retrouve-t-on pas les bagages perdus ?
No full textInternational audienceL’un des aspects anxiogènes du voyage en train est la perte de bagages et d’effets personnels. Cette expérience à la fois banale et fréquente (plus de 8 000 interventions de la police des gares par an) est en forte croissance (+30 % d’objets perdus depuis deux ans dans le réseau ferré français) (Poingt 2021). Elle crée un double problème. Du côté des voyageurs, la perte cause du stress (certains se font dérober leur thèse, d’autres un stradivarius). Du côté des gestionnaires de gare, elle perturbe le fonctionnement de la gare, car l’élément n’est plus rapporté au bureau des objets trouvés (les derniers ont fermé après les attentats de 2015). Depuis, la réponse est fournie par les industries de la communication technologique et de nombreux instruments sont déployés : caméras, applications de tracking, etc.La transformation de la manière dont se gère le problème des objets trouvés permet d’aborder la grande gare métropolitaine comme une construction socio-technique au sein de laquelle s’articulent des flux logistiques, donc physiques (les mouvements de trains, de piétons et d’innombrables sacs, valises, etc.) et des flux d’informations immatériels. Elle permet d’explorer ces infrastructures comme des écosystèmes digitaux dont on cherchera à comprendre les logiques de production et les conditions d’interaction avec les voyageurs
Capitalization of energy labels on the French housing market
No full textThe building sector in France accounts for nearly a third of national emissions and energy consumption, necessitating energy retrofits and renovations to align with the government's emission reduction targets. Energy Performance Certificates (EPCs), play crucial roles in guiding this transformation. This paper explores the market capitalization of home energy labels in mainland France. Including spatial and dynamism heterogeneity and using public data on energy efficiency and property transfers, the results show that top ranked dwellings have an average 14\% significant premium compared to middle rank ones. The bottom ranked houses are priced 6\% lower than middle rank ones. In the case of apartments however, the results for bottom ranks are mitigated and un-significant in most cases. The premium seems to be higher in magnitude in bigger and more dynamic cities. Overall, the results may suggest that more specific data at a thin granularity may be required in order to increase the quality of prediction, especially in capturing heterogeneity
Fondations et optimisation d'échantillonneurs de Langevin
No full textMolecular dynamics allows to compute many physical quantities of interest, such as the pressure of a system or the heat capacity of a material, starting from a microscopic description of systems. These calculations are based on estimating the average of observables over a given thermodynamic ensemble. In practice, one needs to sample a high-dimensional probability measure. For this purpose, the scientific community relies on Markov Chain Monte Carlo (MCMC) algorithms, where the averages of observables are estimated using trajectory averages. In this thesis, we focus on MCMC algorithms based on discretizations of Langevin dynamics, which are stochastic differential equations modeling a physical system interacting with its environment via energy exchanges at a fixed temperature.The numerical integration of these dynamics is very costly. Indeed, the system is generally composed of many particles, and its evolution requires the computation of numerous interaction forces. The time step used for discretizing the dynamics must be small enough to correctly account for all the system's frequencies, and for the integration to be stable in the first place. Moreover, the reliability of estimates of mean values for observables is ensured only after long integration times of the dynamics, thus necessitating a large number of iterations. To limit the computational cost, the first chapters of this thesis investigate the introduction of a non-constant diffusion coefficient in Langevin dynamics, in order to accelerate the convergence of algorithms.In the first work of this thesis, we analyze Hamiltonian Monte Carlo (HMC) algorithms for any Hamiltonian, and show how to generically unbias them; in particular the Riemann Manifold HMC (RMHMC) algorithm which allows the introduction of a non-constant diffusion into kinetic Langevin dynamics. In the second work, we study the optimization of the diffusion coefficient for overdamped Langevin dynamics. We propose a methodology to compute an optimal diffusion, study its behavior in the homogenized limit (with periodic boundary conditions), and obtain a simple analytical expression. In the third work, we suggest a diffusion that optimizes the exploration of the effective dynamics of overdamped Langevin dynamics, as seen through the lens of a collective variable. We show how this diffusion adapts to high-dimensional systems, can be learned on-the-fly during simulations and can be used in combination with the RMHMC algorithm.The final chapter of this manuscript is devoted to the latest advances in the community of diffusion models for estimating normalization constants. We show how these methods can be adapted to compute free energy differences, and test them on a simple molecular dynamics system. We also suggest some ideas for algorithms that could be more suitable for high-dimensional systems.La dynamique moléculaire permet de calculer de nombreuses quantités physiques d'intérêt, comme la pression d'un système ou la capacité thermique d'un matériau, à partir d'une modélisation à l'échelle microscopique. Ces calculs reposent sur l'estimation de moyennes d'observables dans un ensemble thermodynamique donné. En pratique, il s'agit d'échantillonner une mesure de probabilité en dimension grande. Pour cela, la communauté scientifique s'appuie sur les algorithmes de type Markov Chain Monte Carlo (MCMC), où les moyennes d'observables sont estimées par des moyennes trajectorielles. Dans cette thèse, nous étudions principalement les algorithmes MCMC qui reposent sur des discrétisations de dynamiques de Langevin, qui sont des équations différentielles stochastiques modélisant un système physique en interaction avec son environnement via des échanges d'énergie à une température fixée. L'intégration numérique de ces dynamiques s'avère très coûteuse. En effet, les systèmes physiques considérés sont généralement composés de beaucoup de particules et l'intégration en temps de leur évolution requiert de calculer de nombreuses forces d'interactions, le pas de temps utilisé pour la discrétisation de la dynamique devant être suffisamment petit pour correctement rendre compte de tous les mouvements du système et garantir la stabilité de l'intégration numérique. De plus, la fiabilité des estimations des moyennes d'observables est assurée seulement après des long temps d'intégration de la dynamique, demandant ainsi d'effectuer un grand nombre d'itérations. Afin de limiter le coût de calcul, nous étudions dans les premiers chapitres de cette thèse l'introduction d'un coefficient de diffusion non constant dans les dynamiques de Langevin, afin d'accélérer la convergence des algorithmes. Dans le premier travail de cette thèse, nous analysons les algorithmes Hamiltonian Monte Carlo (HMC) pour un Hamiltonien quelconque, et montrons comment les débiaiser génériquement, en particulier pour l'algorithme Riemann Manifold HMC (RMHMC) qui permet d'introduire une diffusion non constante dans la dynamique de Langevin cinétique. Dans le deuxième travail, nous étudions l'optimisation du coefficient de diffusion pour des dynamiques de Langevin suramorties. Nous proposons une méthodologie pour calculer une diffusion optimale, étudions son comportement dans la limite homogénéisée (avec des conditions aux bords périodiques), et obtenons une expression analytique simple de cette diffusion limite. Dans le troisième travail, nous proposons une diffusion permettant d'optimiser l'exploration de la dynamique effective de Langevin suramortie vue par le prisme d'une variable collective. Nous montrons comment cette diffusion s'adapte aux systèmes en grandes dimensions, peut être apprise au cours d'une simulation et peut être utilisée conjointement avec l'algorithme RMHMC.Le dernier chapitre de ce manuscrit est consacré aux dernières avancées dans la communauté des modèles de diffusion pour l'estimation de constantes de normalisation. Nous montrons comment ces méthodes peuvent être adaptées pour calculer une différence d'énergie libre, et testons ces méthodes sur un cas simple de dynamique moléculaire. Nous proposons aussi quelques pistes de réflexion vers des algorithmes potentiellement plus adaptés à des systèmes en dimension grande
Comportement Thermo-Hygro-Mécanique du bois soumis à des sollicitations accidentelles d'incendie
No full textWood is a sustainable, low-carbon construction material with substantial environmental benefits. However, as a hygroscopic material, its mechanical properties are significantly influenced by moisture content. Although Eurocode 5 considers certain climatic factors in the design of timber structures, it does not address the impact of moisture content or the development of moisture gradients within the material, both of which can alter its mechanical performance. Moreover, fire presents a serious risk to timber structures due to the combustible nature of wood.This study investigates the thermo-hygro-mechanical behavior of wood when exposed to fire, with an emphasis on the effects of temperature and moisture on its mechanical properties. The research examines how thermal, and moisture gradients affect the compressive strength and axial modulus of elasticity of wood, both at the material and structural scales.Initially, experimental tests are conducted to analyze the mechanical properties of wood under various homogeneous and heterogeneous temperature and moisture conditions. The results reveal that thermo-hygro gradients significantly influence the mechanical performance of wood, particularly its compressive strength during thermal exposure. Subsequently, structural-scale tests on timber columns validate the findings from the material scale and provide deeper insights into the effects of thermal and moisture diffusion on the behavior of timber structures.Finally, this research presents a predictive model for the behavior of wood under fire conditions, providing a framework for enhancing the safety and design of timber structures exposed to fire, supported by adapted design guidelines.Le bois, en tant que matériau de construction durable à faible empreinte carbone, présente des avantages environnementaux considérables. Cependant, en raison de son caractère hygroscopique, ses propriétés mécaniques sont fortement influencées par sa teneur en humidité. Bien que l'Eurocode 5 intègre certains aspects climatiques dans la conception des structures en bois, il néglige l'influence de l'humidité et des gradients hydriques pouvant se développer à l'intérieur du matériau, altérant ainsi ses performances mécaniques. Par ailleurs, le bois, en tant que matériau combustible, est particulièrement vulnérable aux incendies, ce qui représente une menace pour la sécurité des structures. Cette thèse porte sur l'étude du comportement thermo-hygro-mécanique du bois exposé à des conditions d’incendie, avec une attention particulière sur l'impact de la température et de l'humidité sur ses propriétés mécaniques. Les effets des gradients thermiques et hydriques sur la résistance en compression et le module d'élasticité axial du bois sont examinés, tant à l'échelle du matériau qu'à l'échelle structurale. Dans un premier temps, des essais expérimentaux sont menés afin d'analyser les propriétés mécaniques du bois sous diverses conditions de température et d'humidité, homogènes et hétérogènes. Les résultats montrent que les gradients thermo-hydriques ont une influence significative sur les performances mécaniques du bois, notamment sa résistance en compression en situation d’incendie. Ensuite, des essais à l'échelle structurale sur des poteaux en bois permettent de valider les conclusions obtenues à l'échelle du matériau et d’approfondir la compréhension des effets de la diffusion thermique et hydrique sur le comportement global des structures en bois. Enfin, cette thèse propose un modèle prédictif du comportement du bois sous conditions d’incendie, apportant ainsi des perspectives pour améliorer la conception et la sécurité des bâtiments en bois face aux risques d'incendie, en s’appuyant sur des règles de dimensionnement adaptées.' étude du comportement thermo-hygro-mécanique du bois exposé à des conditions d’incendie, avec une attention particulière portée à l'impact de la température et de l'humidité sur ses propriétés mécaniques. Les effets des gradients thermiques et hydriques sur la résistance en compression et le module d'élasticité axial du bois sont examinés, tant à l'échelle du matériau qu'à l'échelle structurale. Dans un premier temps, des essais expérimentaux sont menés afin d'analyser les propriétés mécaniques du bois sous diverses conditions de température et d'humidité, homogènes et hétérogènes. Les résultats montrent que les gradients thermo-hydriques ont une influence significative sur les performances mécaniques du bois, notamment sa résistance en compression en situation d’incendie. Ensuite, des essais à l'échelle structurale sur des poteaux en bois permettent de valider les conclusions obtenues à l'échelle du matériau et d’approfondir la compréhension des effets de la diffusion thermique et hydrique sur le comportement global des structures en bois. Enfin, cette thèse propose un modèle prédictif du comportement du bois sous conditions d’incendie, apportant ainsi des perspectives pour améliorer la conception et la sécurité des bâtiments en bois face aux risques d'incendie, en s’appuyant sur des règles de dimensionnement adaptées
Finite Volumes for the Stefan-Maxwell Cross-Diffusion System
No full textInternational audienceThe aim of this work is to propose a provably convergent finite volume scheme for the so-called Stefan-Maxwell model, which describes the evolution of the composition of a multi-component mixture and reads as a cross-diffusion system. The scheme proposed here relies on a two-point flux approximation, and preserves at the discrete level some fundamental theoretical properties of the continuous models, namely the non-negativity of the solutions, the conservation of mass and the preservation of the volume-filling constraints. In addition, the scheme satisfies a discrete entropy-entropy dissi-pation relation, very close to the relation which holds at the continuous level. In this article, we present this scheme together with its numerical analysis, and finally illustrate its behaviour with some numerical results
Pro-business arbitration with ISDS
No full textIn this paper, we investigate the Investor-State Dispute Resolution Settlement (ISDS) framework, which governs dispute resolution between foreign investors and host states in many bilateral and multilateral trade agreements. We show that ISDS delivers fair justice in a one-shot setting. In a repeated-interaction setting however, it is prone to collusion to the benefit of all parties except the host states. Three factors are determinant: First, the investors are the sole parties able to file cases; Second, arbitrators' earning prospects depend on the investors' filing cases; And finally, treaties leave substantial discretion to arbitration courts in their interpretation of treaties' provisions. We give conditions for pro-business collusion between investors and arbitrators to develop and we show how it makes it profitable for foreign investors to file high-stake claims against states in response to new environmental, social or health regulations. Further, we address regulatory chill and show how the fear of ISDS attacks can hold back welfare improving regulation in the host country. Finally, we extend the model to show how regulatory chill affect policy-making in other countries in which the investor operates with similar activities
Fredholm backstepping for critical operators and application to rapid stabilization for the linearized water waves
No full textFredholm-type backstepping transformation, introduced by Coron and Lü, has become a powerful tool for rapid stabilization with fast development over the last decade. Its strength lies in its systematic approach, allowing to deduce rapid stabilization from approximate controllability. But limitations with the current approach exist for operators of the form |Dx| α for α ∈ (1, 3/2]. We present here a new compactness/duality method which hinges on Fredholm's alternative to overcome the α = 3/2 threshold. More precisely, the compactness/duality method allows to prove the existence of a Riesz basis for the backstepping transformation for skew-adjoint operator verifying α > 1, a key step in the construction of the Fredholm backstepping transformation, where the usual methods only work for α > 3/2. The illustration of this new method is shown on the rapid stabilization of the linearized capillary-gravity water wave equation exhibiting an operator of critical order α = 3/2
Effect of temperature on the mechanical properties of fine-grained soils - A review
No full textInternational audienceAbstract Swelling of clay–sulfate rocks is a serious and devastating geo-hazard, often causing damage to geotechnical structures. Therefore, understanding underlying swelling processes is crucial for the safe design, construction, and maintenance of infrastructure. Planning appropriate countermeasures to the swelling problem requires a thorough understanding of the processes involved. We developed a coupled hydro-mechanical (HM) model to reproduce the observed heave in the historic city of Staufen in south-west Germany, which was caused by water inflow into the clay–sulfate bearing Triassic Grabfeld Formation (formerly Gipskeuper = “Gypsum Keuper”) after geothermal drilling. Richards’ equation coupled to a deformation process with linear kinematics was used to describe the hydro-mechanical behavior of clay–sulfate rocks. The mathematical model is implemented into the scientific open-source framework OpenGeoSys. We compared the model calculations with the measured long-term heave records at the study site. We then designed a sensitivity analysis to achieve a deeper insight into the swelling phenomena. The synthetic database obtained from the sensitivity analysis was used to develop a machine learning (ML) model, namely least-squares boosting ensemble (LSBoost) model coupled with a Bayesian optimization algorithm to rank the importance of parameters controlling the swelling. The HM model reproduced the heave observed at Staufen with sufficient accuracy, from a practical point of view. The ML model showed that the maximum swelling pressure is the most important parameter controlling the swelling. The other influential parameters rank as Young’s modulus, Poisson’s ratio, overburden thickness, and the initial volumetric water content of the swelling layer
Fate of nitrogen in French human excreta: current waste and agronomic opportunities for the future
No full textInternational audienceNitrogen (N) is essential for plant growth and protein synthesis but global reactive N losses, mainly from food systems, induce strong environmental impacts. N losses after human excretion are often overlooked because, in Western societies, they partly occur as inert N2, following denitrification in wastewater treatment plants (WWTP), and losses in waters are often small compared to diffuse agricultural emissions. Yet N from human excretions could be used for crop fertilization, potentially with very high recycling rates via source separation. In this study we use unique operational data from the ∼20,000 French WWTPs to produce a N mass-balance of excretions in the French sanitation system. Even though 75 % of WWTPs' sludge is spread on crops, only 10 % of the excreted N is recycled and 50 % of N is lost to the atmosphere, mainly through WWTP nitrification-denitrification. The remaining 40 % ends up in water or in diffuse losses in the ground, of which about half is lost outside of the WWTPs' discharge system, through sewers storm water and individual autonomous systems. While WWTPs removal efficiency increased in the 2000s, it has been followed by a decade of stagnation, reaching 70 % at the national level. This national average hides regional discrepancies, from 60 to 85 % in the 6 French water agencies basins. These differences closely correlate with the classification as “N sensitive areas” and is mainly due to large WWTPs which handle most of the N load. Recycling all N in excretions could supply 10 % of domestic protein consumption in the current French food system, and up to 30 % if it is prioritized towards crop production for human consumption. Redesigning the food system (decrease of nutrient losses, more plant-based diets) could further increase this contribution
The concentration of personal wealth in Italy 1995-2016
No full textInternational audienceItaly is one the countries with the highest wealth-to-income ratio in the developed world, but knowledge about the size distribution of wealth is currently limited. In this paper we estimate the distribution of personal wealth between 1995 and 2016, a period of economic turbulence and structural reforms. For this, we use a novel source on the full records of inheritance tax files, combined with surveys and national accounts. Unlike available statistics from household surveys alone, our estimates point to a sharp inversion of fortunes between the top and the bottom of the wealth distribution since the mid-1990s. Whereas the level of wealth concentration in Italy is in line with other European countries, its time trend appears more in line with the U.S., showing a large increase. Moreover, Italy stands out as one of the countries with the strongest decline in the wealth share of the bottom 50% of the population. A range of alternative series of wealth concentration, including estimates applying no adjustments and imputations, confirm our main findings. The paper also sheds new light on the determinants of wealth inequality trends. First, we show that although average wealth increases with age, dispersion within age groups remains veryhigh; hence age plays a marginal role in explaining wealth concentration. Second, we show that house prices explain little of the change in wealth across the distribution since 1995. Changes in equity prices account for a large share of wealth growth above the 99th percentile. However, all in all, changes in the volume of assets and savings appear to be the predominant force behind the increase in wealth inequality, even at the top. The probability of top earners to climb to the top of the wealth distribution has doubled since the 2000s. Third, we document the growing role of life-time wealth transfers receipts, their increasing concentration at the top, and their increasingly favourable tax treatment for the wealthy