Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
The performance of [Emim]Br/H2O as a working pair in the absorption refrigeration system
No full textInternational audienc
Development of a Creep Mechanical Frame Setting Based on Arcan Test
No full textInternational audienceStructural adhesive bonding has several advantages compared to other assembly techniques such as welding and riveting. Yet, creep is an important long-term phenomenon that needs to be considered for the design of such joints. To investigate this aspect, experimental investigations are needed at the scale of the adhesive layer. At such a scale, Arcan setting presents several advantages compared to other existing tests (analysis of different loads, limited edge effects) and was thus chosen for the development of an experimental approach aiming at providing creep characterization of adhesive layers. A mechanically operated testing machine was thus developed and is presented herein. To measure local strains of the adhesive layer, two linear variable differential transformer sensors are fixed on a support placed on the beaks of the Arcan sample according to two directions: normal and tangential to the bonded surface. The developed system (mechanical system and strain measurement) proved to be successful in investigating the creep behavior of adhesive layers while varying several parameters (load levels, type of load, temperature, adhesive thickness)
Active Design of Diffuse Acoustic Fields in Enclosures
No full textInternational audienceThis paper presents a numerical framework for designing diffuse fields in rooms of any shape and size, driven at arbitrary frequencies. That is, we aim at overcoming the Schroeder frequency limit for generating diffuse fields in an enclosed space. We formulate the problem as a Tikhonov regularized inverse problem and propose a lowrank approximation of the spatial correlation that results in significant computational gains. Our approximation is applicable to arbitrary sets of target points and allows us to produce an optimal design at a computational cost that grows only linearly with the (potentially large) number of target points. We demonstrate the feasibility of our approach through numerical examples where we approximate diffuse fields at frequencies well below the Schroeder limit
How does the partition of unity influence SORAS preconditioner?
No full textInternational audienceWe investigate the influence of the choice of the partition of unity on the convergence of the Symmetrized Optimized Restricted Additive Schwarz (SORAS) preconditioner for the reaction-convection-diffusion equation. We focus on two kinds of partitions of unity, and study the dependence on the overlap and on the number of subdomains. In particular, the second kind of partition of unity, which is non-zero in the interior of the whole overlapping region, gives more favorable convergence properties, especially when increasing the overlap width, in comparison with the first kind of partition of unity, whose gradient is zero on the subdomain interfaces and which would be the natural choice for ORAS solver instead
Exploring low-rank structure for an inverse scattering problem with far-field data
No full textInternational audienceThe inverse scattering problem exhibits an inherent low-rank structure due to its ill-posed nature; however developing low-rank structures for the inverse scattering problem remains challenging. In this work, we introduce a novel low-rank structure tailored for solving the inverse scattering problem. The particular low-rank structure is given by the generalized prolate spheroidal wave functions, computed stably and accurately via a Sturm-Liouville problem. We first process the far-field data to obtain a post-processed data set within a disk domain. Subsequently, the post-processed data are projected onto a low-rank space given by the low-rank structure. The unknown is approximately solved in this low-rank space, by dropping higher-order terms. The low-rank structure leads to a H\"{o}lder-logarithmic type stability estimate for arbitrary unknown functions, and a Lipschitz stability estimate for unknowns belonging to a finite dimensional low-rank space. Various numerical experiments are conducted to validate its performance, encompassing assessments of resolution capability, robustness against randomly added noise and modeling errors, and demonstration of increasing stability
Hydraulic Volumetric Soft Everting Vine Robot Steering Mechanism for Underwater Exploration
No full textInternational audienc
Autotelic LLM-based exploration for goal-conditioned RL
No full textInternational audienceDesigning autotelic agents capable of autonomously generating and pursuing their own goals represents a promising endeavor for open-ended learning and skill acquisition in reinforcement learning. This challenge is especially difficult in open worlds that require inventing new previously unobserved goals. In this work, we propose an architecture where a single generalist autotelic agent is trained on an automatic curriculum of goals. We leverage large language models (LLMs) to generate goals as code for reward functions based on learnability and difficulty estimates. The goal-conditioned RL agent is trained on those goals sampled based on learning progress. We compare our method to an adaptation of OMNI-EPIC to goal-conditioned RL. Our preliminary experiments imply that our method generates a higher proportion of learnable goals, suggesting better adaptation to the goalconditioned learner
Reinforcement Learning for Aligning Large Language Models Agents with Interactive Environments: Quantifying and Mitigating Prompt Overfitting
No full textReinforcement learning (RL) is a promising approach for aligning large language models (LLMs) knowledge with sequential decision-making tasks. However, few studies have thoroughly investigated the impact on LLM agents capabilities of fine-tuning them with RL in a specific environment. In this paper, we propose a novel framework to analyze the sensitivity of LLMs to prompt formulations following RL training in a textual environment. Our findings reveal that the performance of LLMs degrades when faced with prompt formulations different from those used during the RL training phase. Besides, we analyze the source of this sensitivity by examining the model's internal representations and salient tokens. Finally, we propose to use a contrastive loss to mitigate this sensitivity and improve the robustness and generalization capabilities of LLMs
Analysis of the durability damage scenarios of air spring sleeves with axial reinforcements based on computer tomography and digital image processing
No full textInternational audienceAir springs are flexible components used in various applications such as vehicles, machinery, and industrial equipment. They are made of a multilayer composite material based on rubber and reinforcing textile cords. The service life and reliability of air springs depend heavily on their fatigue behaviour, which is influenced by various factors such as temperature, pressure, load and environmental conditions. To better understand and model the fatigue behaviour of air spring sleeves, it is necessary to identify and quantify the damage mechanisms and processes inside the individual layers. In this work, a new tool is presented that is able to distinguish the components of an air spring sleeve using volumetric CT-data. The tool is based on advanced greyscale detection technology to recognize and segment the different layers and materials in the data of a scanned sleeve. It allows to determine the geometry, thickness, and volume fractions of each layer and to detect potential defects such as cracks, delaminations, or holes. The tool is applied to various damage scenarios for the fatigue behaviour of air springs. The focus is put on fatigued axial sleeves subjected to cyclic loading. Various CT scan data sets of axial sleeves are analysed and compared to their initial state after being exposed to different load cases. It is shown how the tool helps to track the damage evolution in each layer and how to identify the critical areas. This paper provides the first results of the damage scenarios for fatigued axial sleeves. The presented tool will allow an in-depth analysis of the fatigue behaviour of axial air spring composite materials in future and can also be applied to other types of air spring sleeves or similar structures. It can provide an important contribution to improving the service life prediction and design of air springs
