Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Experimental validation of the strut-and-tie design for reinforced concrete massive structures by deformation flow analysis
No full textInternational audienceThe design of reinforced concrete structures follows prescriptions of codes and standards (e.g., ACI and Eurocode). The strut-and-tie method is the only approach detailed in standards for the design of reinforcement of massive structures. Stut-and-tie methods are currently showing a strong interest and several techniques are based on automatic strategies and topological optimization. However, this method is based on a strong hypothesis associated with linear elastic analyses. In particular, the occurrence of pre-existing cracks is not considered when drawing the strut-and-tie model. This work aims to understand the role of an initially cracked state and, thus, the loading history in the design and choice of strut and tie directions. An experimental test is analyzed for a real size corbel loaded in two different inplane directions with forces not applied simultaneously. Hence, the specimen was cracked when the second force was applied. The corbel has minimal reinforcement to avoid brittle failure but should not affect the crack directions. The instrumentation used, combining digital image correlation and optical fiber sensing, allows for monitoring during the test and provides insight into the cracking history and the deformations in visible surfaces and in the bulk. Digital image correlation was used over the entire top surface of the corbel to measure displacement fields and thus strain fields, as well as information on cracking during the test. Optical fibers were installed within the specimen to measure strains in bulk. This measurement allows for the detection of cracks in the sample volume and thus complements digital image correlation measurements for fibers close to the surface. The results provide key data on the flow history of mechanical fields within the structure under different loads. This information helped clarify the domain of validity of strut-and-tie fundamental assumptions and may improve this design approach
Socially Supervised Representation Learning: the Role of Subjectivity in Learning Efficient Representations
No full textInternational audienceDespite its rise as a prominent solution to the data inefficiency of today's machine learning models, self-supervised learning has yet to be studied from a purely multi-agent perspective. In this work, we propose that aligning internal subjective representations, which naturally arise in a multi-agent setup where agents receive partial observations of the same underlying environmental state, can lead to more data-efficient representations. We propose that multi-agent environments, where agents do not have access to the observations of others but can communicate within a limited range, guarantees a common context that can be leveraged in individual representation learning. The reason is that subjective observations necessarily refer to the same subset of the underlying environmental states and that communication about these states can freely offer a supervised signal. To highlight the importance of communication, we refer to our setting as \textit{socially supervised representation learning}. We present a minimal architecture comprised of a population of autoencoders, where we define loss functions, capturing different aspects of effective communication, and examine their effect on the learned representations. We show that our proposed architecture allows the emergence of aligned representations. The subjectivity introduced by presenting agents with distinct perspectives of the environment state contributes to learning abstract representations that outperform those learned by a single autoencoder and a population of autoencoders, presented with identical perspectives of the environment state. Altogether, our results demonstrate how communication from subjective perspectives can lead to the acquisition of more abstract representations in multi-agent systems, opening promising perspectives for future research at the intersection of representation learning and emergent communication
Model order reduction of nonlinear MEMS structures: a high order invariant manifold approach
No full textInternational audienceA high order direct parametrisation of invariant manifolds is exploited to operate dimensionality reduction for vibrating structures subjected to geometric nonlinearities. The method defines a nonlinear coordinate change between the nodal degrees-of-freedom and the normal coordinates, hence expressing the dynamics in an invariant-based span of the phase space. The method is applied to study micro-electro-mechanical systems (MEMS) subjected to geometric nonlinearities and internal resonance. Remarks on the computational performance are reported to highlight the efficiency of the technique
Loading Classification for Fatigue Design Applied to Automotive Time-Series
No full textInternational audienc
Kleene Algebra to Compute Invariant Sets of Dynamical Systems
No full textInternational audienceIn this paper, we show that a basic fixed point method used to enclose the greatest fixed point in a Kleene algebra will allow us to compute inner and outer approximations of invariant-based sets for continuous-time nonlinear dynamical systems. Our contribution is to provide the definitions and theorems that will allow us to make the link between the theory of invariant sets and the Kleene algebra. This link has never be done before and will allow us to compute rigorously sets that can be defined as a combination of positive invariant sets. Some illustrating examples show the nice properties of the approach
Estimates of Methane Release From Gas Seeps at the Southern Hikurangi Margin, New Zealand
No full textInternational audienceThe highest concentration of cold seep sites worldwide has been observed along convergent margins, where fluid migration through sedimentary sequences is enhanced by tectonic deformation and dewatering of marine sediments. In these regions, gas seeps support thriving chemosynthetic ecosystems increasing productivity and biodiversity along the margin. In this paper, we combine seismic reflection, multibeam and split-beam hydroacoustic data to identify, map and characterize five known sites of active gas seepage. The study area, on the southern Hikurangi Margin off the North Island of Aotearoa/New Zealand, is a well-established gas hydrate province and has widespread evidence for methane seepage. The combination of seismic and hydroacoustic data enable us to investigate the geological structures underlying the seep sites, the origin of the gas in the subsurface and the associated distribution of gas flares emanating from the seabed. Using multi-frequency split-beam echosounder (EK60) data we constrain the volume of gas released at the targeted seep sites that lie between 1,110 and 2,060 m deep. We estimate the total deep-water seeps in the study area emission between 8.66 and 27.21 × 10 6 kg of methane gas per year. Moreover, we extrpolate methane fluxes for the whole Hikurangi Margin based on an existing gas seep database, that range between 2.77 × 10 8 and 9.32 × 10 8 kg of methane released each year. These estimates can result in a potential decrease of regional pH of 0.015–0.166 relative to the background value of 7.962. This study provides the most quantitative assessment to date of total methane release on the Hikurangi Margin. The results have implications for understanding what drives variation in seafloor biological communities and ocean biogeochemistry in subduction margin cold seep sites
COVID-19: Current challenges regarding medical healthcare supplies and their implications on the global additive manufacturing industry
No full textInternational audienceThe covid-19 outbreak has caused a shortage of masks and other healthcare products for the general public around the globe. In addition, it has also affected the supply of personal protective equipment (PPE) used by healthcare services because of a sudden increase in their demand. This significant disruption in the global supply chain of these products resulted in, leaving many staff and patients without protection. The additive manufacturing (AM) industry is going through extraordinary times and can provide emergency responses to help deal with the global crisis caused by the COVID-19 pandemic. The objective of the present work is therefore to perform an up-to-date review to determine the capacity of AM to provide exclusive benefits for the medical healthcare supplies sector to fight this current situation. In this review, it is found that AM technology has proved that it can be used as a volume manufacturing technology for the ongoing crisis. However, the standardization and certification are appeared to represent the main challenges for adopting the AM in healthcare against COVID-19. Furthermore, additively manufactured materials for medical applications must be developed for medical environments. Most printed medical products for COVID-19 require biocompatibility evaluation and shall prove their ability to sterilize. Finally, this review concluded that AM technology can fulfill the requirements of face masks and ventilator parts for healthcare systems for proper controlling and treating of COVID-19 patients when the safety and efficacy of these devices are ensured
Efficient jib-mainsail fluid-structure interaction modelling – Validations with semi-rigid sails experiments
No full textInternational audiencehe present work takes place in the framework of the “Jib Sea” project. The main purpose of the project is to develop a new sail design made of articulated composite panels, for large merchant ships. The French laboratory “ENSTA Bretagne, IRDL” is involved in this industrial project for its expertise on fluid-structure interactions modelling. A fast and robust approach to model fluid-structure interactions for yacht sails is presented. Specifically, interaction effects between the jib and the mainsail are considered in the flow model presented. This is achieved using the lifting-line theory combined with a discrete vortex method, involving distributions of lumped-vortex elements along sail sections. The flow model is coupled with a finite element analysis of the structure, using shell elements for the modelling of sail membranes, beam stringers for battens modelling and a quasi-static resolution based on a dynamic backward Euler scheme. The flow model based on the Lifting-Line Theory is presently validated by a commercial software tools using the Unsteady Reynolds-Average Navier-Stokes equations. Numerical comparisons with experiments are conducted on a 50 m2 composite mainsail prototype and a conventional jib, built and hoisted on an onshore balestron rig. Measurements, such as strain gauges or cable tensions, are synchronized with a wind sensor. These data collected together enable both global and local numerical-experimental comparisons for forces and moments, providing a validation of the proposed fluid-structure interactions modelling of ship sails. A good match between experimental and numerical modelling is observed on local comparisons as relative differences are all less than 25% for TWA∈[−20;20] and TWS<10 kn. Global comparison results exhibit validations with experiments for |TWA|<10 deg and TWS<10 kn, where numerical-experimental relative differences are less than 10%
Finite element formulation of a homogenized beam for reticulated structure dynamics
No full textInternational audienc
Learning Sensorimotor Agency in Cellular Automata
No full texthttps://developmentalsystems.org/sensorimotor-lenia/In this blogpost ( https://developmentalsystems.org/sensorimotor-lenia/ ), we explore the concepts of embodiment, individuality, self-maintenance and sensorimotor agency within a cellular automaton (CA) environment. Whereas those concepts are central in theoretical biology and cognitive science, it remains unclear how such behaviors can emerge in a CA-like environment made only of low-level particles and physical rules. We present a novel set of tools (based on curriculum learning, diversity search and gradient descent over a differentiable CA) to automatically learn the rules leading to the emergence of such behaviors. Our method is able to discover robust self-organizing agents with strong coherence and generalization to out-of-distribution changes, reminiscent of the robustness of living systems to maintain specific functions despite environmental and body perturbations.https://developmentalsystems.org/sensorimotor-lenia