HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
Predictive posterior sampling from non-stationnary Gaussian process priors via Diffusion models with application to climate data
No full textBayesian models based on Gaussian processes (GPs) offer a flexible framework to predict spatially distributed variables with uncertainty. But the use of nonstationary priors, often necessary for capturing complex spatial patterns, makes sampling from the predictive posterior distribution (PPD) computationally intractable. In this paper, we propose a two-step approach based on diffusion generative models (DGMs) to mimic PPDs associated with non-stationary GP priors: we replace the GP prior by a DGM surrogate, and leverage recent advances on training-free guidance algorithms for DGMs to sample from the desired posterior distribution. We apply our approach to a rich non-stationary GP prior from which exact posterior sampling is untractable and validate that the issuing distributions are close to their GP counterpart using several statistical metrics. We also demonstrate how one can fine-tune the trained DGMs to target specific parts of the GP prior. Finally we apply the proposed approach to solve inverse problems arising in environmental sciences, thus yielding state-of-the-art predictions.</div
Converse Lyapunov Theorem for Input-to-State Stability of Linear Integral Difference Equations
No full textInternational audienceThis paper investigates necessary and sufficient Lyapunov conditions for Input-to-State Stability (ISS) of Linear Integral Difference Equations in the presence of an additional exogenous signal. Building on recent research in the literature pertaining to necessary conditions for the exponential stability of difference equations, we introduce a quadratic Lyapunov functional that incorporates the derivative of the so-called Lyapunov matrix. We demonstrate that the ISS of the considered class of systems is contingent upon the existence of an ISS Lyapunov functional. The Lyapunov analysis hinges on the properties of the fundamental matrix and the delay Lyapunov matrix.</div
Localization of Embrittlement in CMAS-Infiltrated Thermal Barrier Coatings via Laser Shock Experiments
No full textInternational audienc
Beyond Meta-Appearances: Philosophical and Bioinspired Foundations for Non-Dualistic Robotics
No full textSlides of my flash talk at the IEEE RoboSoft workshop on Bioinspired Autonomy: Philosophical aspects meet technological challengesInternational audienceCurrent robotic paradigms are shaped by meta-appearances, where articulated structures—such as robotic hands—are designed to replicate human biomechanics without achieving true integration between material structure and intelligence. This dualistic separation between control (software) and embodiment (hardware) mirrors historical philosophical models that treated cognition as independent of materiality. In contrast, biological intelligence emerges through an indivisible synthesis of form and function.Drawing on philosophical anthropology, this work reframes bioinspired robotics through a non-dualistic, non-reductionist ontology, inspired by historical perspectives on embodiment, tool-making, and the evolution of intelligence. The human hand, as the archetype of dexterous control, has served as the foundation for articulated robotics, yet its robotic counterparts remain constrained by predefined movement patterns and external computation.We explore how active materials may enable robotics to overcome the limitations of meta-appearances, fostering sensorimotor synchrony between structure and intelligence. Unlike articulated robots that function as externally controlled mechanical systems, bioinspired material intelligence could allow for real-time adaptation, where cognition is not merely simulated but emerges dynamically from the material itself.By integrating philosophical inquiry with bioinspired robotics, this work contributes to the discussion on how autonomy in artificial systems can evolve beyond functional imitation, toward a paradigm where intelligence and embodiment co-develop as a unified whole
Bottom and Top Internodes Subjected to Interactions with Genotype in Miscanthus: Impact of Biochemical Composition and Anatomy on Stem-Based Composites Mechanical Properties
No full textInternational audienceMiscanthus (Miscanthus Andersson) is a perennial grass for which biomaterials market has taken growing interest. Our objective was to evaluate the effect of stem internode position in Miscanthus × giganteus and Miscanthus sinensis and the impact of its anatomy and biochemical composition on internode-based composites' mechanical properties. Stems' bottom and top internodes were sampled for two genotypes of each species in two different years and separately added to a polypropylene matrix, and the mechanical properties of the internode-reinforced composites were measured. Before composite production, the internodes were extensively phenotyped for biochemical composition and anatomy. Stems' bottom and top internode-based composites yielded different modulus (3203 and 2988 MPa, respectively), while tensile strength was similar (36.4 and 36.5 MPa, respectively). Significant genotype × internode interactions occurred for most variables, mainly due to differences among species, since both Miscanthus sinensis clones proved to be more stable than both Miscanthus × giganteus clones for modulus (4% and 10.2%, respectively). Regarding tensile strength, the species showed small but opposite differences between internodes. Tensile strength and modulus were rather close only in the top internodes, where good mechanical properties were associated with the lowest values of vascular bundles number and section area and highest parenchyma tissue, while opposite results were obtained in the bottom ones, only for tensile strength. Miscanthus sinensis species proved to be interesting for the stability improvement of composite mechanical properties. It appears essential for experimental purposes to stratify the sampling by internode in order to be representative of the whole stem
On the ε-Euler-Maruyama scheme for time-inhomogeneous jump-driven SDEs
No full textInternational audienceWe consider a class of general SDEs with a jump integral term driven by a time-inhomogeneous Poisson random measure. We propose a two-parameters Euler-type scheme for this SDE class and prove an optimal rate for the strong convergence with respect to the -norm and for the weak convergence, considering integration over uniform time-steps.One of the primary issues to address in this context is the approximation of the noise structure when it can no longer be expressed as the increment of random variables. We extend the Asmussen--Rosiński approach to the case of a fully dependent jump coefficient and time-dependent Poisson compensation, handling contribution of jumps smaller than with an appropriate Gaussian substitute and exact simulation for the large jumps contribution. For any , under hypotheses required to control the -moments of the process, we obtain a strong convergence rate of order . Under standard regularity hypotheses on the coefficients, we obtain a weak convergence rate of , where is the Blumenthal--Getoor index of the underlying Lévy measure. We compare this scheme with the Rubenthaler's approach where the jumps smaller than are neglected, providing strong and weak rates of convergence in that case too. The theoretical rates are confirmed by numerical experiments afterwards. We apply this model class for some anomalous diffusion model related to the dynamics of rigid fibres in turbulence
3D mosaicity of a single-crystal nickel-based superalloy by lab-based diffraction contrast tomography
No full textInternational audienceThe 3D microstructure of a second-generation single-crystal nickel-based superalloy, René N5, has been analyzed using laboratory-based X-ray diffraction contrast tomography (lab-based DCT). This experiment has demonstrated the precise capabilities of lab-based DCT in resolving subgrain boundaries with a misorientation angle of less than 1°, achieving an angular accuracy as fine as 0.1°. The performance of the lab-based DCT has been compared with standard and widely used electron backscatter diffraction (EBSD) analysis. Obtaining the 3D microstructure non-destructively enabled the segmentation of the network of nickel-based single-crystal dendrites, opening up new opportunities for studying crystal mosaicity
In situ and ex situ characterization of microstructure evolution of a γ-γ’ coating on CMSX-4 Plus superalloy during thermal cycling: Insights into Pt diffusion and phase transformations
No full textInternational audienceThe microstructure evolution of a Pt-rich γ-γ’ coating deposited on a third generation CMSX-4 Plus nickel-based superalloy was investigated during relatively short thermal cycles (from 300 °C to 1100 °C with a dwell time of 5 min at 1100 °C) using both ex situ and in situ characterization techniques. Ex situ analyses, including Scanning Electron Microscopy (SEM) and Energy-Dispersive X-Ray Spectroscopy (EDS) and postmortem X-Ray Diffraction (XRD), demonstrated that Pt and Al interdiffusion induced by thermal cycling leads to significant microstructure changes in the coating. Lattice parameters of γ and γ’ phases were revealed to correlate with Pt content. In a novel approach, in situ XRD using laboratory equipment was employed to monitor microstructural evolution during 250 thermal cycles. This unique in situ analysis highlighted the microstructural differences at low and high temperatures within each cycle. For the first time, it was observed that partial dissolution of γ’ precipitates occurs at high temperatures, altering the local chemical composition of both γ and γ’ phases
New normal, new norms: Towards sustainable and resilient global logistics and supply chain management
No full textInternational audienceGlobal logistics and supply chains (GLSC)—a vital element in the strategic decisions of multinational corporations and governments—have received unprecedented attention in recent years. On one hand, the climate crisis has intensified the imperative to reduce greenhouse gas emissions across all sectors, particularly in freight transportation and logistics. A rapidly growing number of firms in the sector have committed to achieving the target of Net Zero Emissions by 2050 and are seeking effective, efficient strategies and solutions to reach this goal. On the other hand, recent disruptive events, most notably the COVID-19 pandemic, rising tariffs and trade wars, resurging protectionism, and geopolitical tensions, have clearly exposed the modern supply chain’s vulnerabilities and profoundly challenged the prevailing management practices that have shaped global operations for decades. As a result, GLSC management is undergoing a fundamental transformation, and a return to the pre-disruption status quo appears increasingly unlikely in the near term. A central question now facing both scholars and practitioners is how GLSC can adapt effectively to this new normal. This Special Issue brings together original and impactful research that rethinks the future of GLSC, investigates its ongoing transformation, examines the evolving landscape of management research in this domain, and identifies emerging paradigms, approaches, and models that are driving the transition. These insights are expected to shape the future of GLSC toward a model of sustainable, human-centric, and resilient management—a vision aligned with the vision of Industry 5.0
