HAL Arts et Métiers
Not a member yet
14127 research outputs found
Sort by
A Python/Fortran Implementation of the Lattice‐Boltzmann Kernel on Multiple GPU Using the OpenACC Framework
No full textInternational audienceABSTRACT The increasing availability of GPU accelerated architectures for high‐performance computing presents new opportunities for scientific software but also challenges due to the complexity of porting legacy codes to accelerator platforms. Directive‐based programming models such as OpenACC offer a minimally intrusive pathway to exploit GPU acceleration without requiring extensive rewriting of existing codes. The current work presents a comprehensive performance and portability study of a LatticeBoltzmann Method solver (PyLB) originally written in Python, Mpi4Py, and Fortran for CPU architectures, which is ported to GPUs using OpenACC directives applied to the Fortran routines. The performance of the solver is evaluated on NVIDIA V100, A100, and H100 GPUs available on the Jean Zay supercomputer from Institute for Development and Resources in Intensive Scientific Computing (IDRIS) in France. Roofline analysis and extensive strong and weak scalability tests are conducted, showing that the GPU‐enabled version of PyLB scales efficiently across multiple GPUs. The solver achieves performance on the H100 GPU equivalent to thousands of CPU cores and shows strong energy and carbon efficiency advantages over traditional CPU‐based simulations. The implementation is validated using classical benchmarks, including the decaying Taylor‐Green vortex and the flow over a 3‐D sphere. The results confirm the physical accuracy of the GPU port while highlighting its computational and environmental advantages
Data-driven surrogate modeling for accelerated multiscale FE 2 simulation of nonlinear thermomechanical behavior in composite materials
No full textInternational audienceGlass fiber reinforced composites exhibit highly nonlinear and path-dependent thermomechanical responses under coupled loading. Accurately predicting this behavior using traditional multiscale frameworks such as FE remains computationally prohibitive, particularly for complex geometries involving constituent-level interactions. To overcome these limitations, this study investigates the Thermomechanical Elasto-Plastic Artificial Neural Network (ThEP-ANN), a novel data-driven multiscale surrogate designed to accelerate fully coupled thermomechanical simulations of such materials. The ThEP-ANN framework circumvents the complexity of explicit constitutive modeling by directly predicting macroscopic quantities (e.g., stresses and energy rates), rather than relying on free-energy potentials. This strategy significantly reduces the computational cost associated with high-dimensional automatic differentiation. Crucially, the surrogate explicitly incorporates the evolution of internal variables, enabling an accurate representation of history-dependent and path-integral material responses. The ThEP-ANN is implemented non-intrusively into the commercial finite element code Abaqus via a Meta-UMAT subroutine, allowing seamless integration into macroscopic simulations. The framework is assessed through a three-stage, fully numerical validation strategy, comprising generalization to unseen loading paths, mesoscopic validation at the Representative Volume Element (RVE) level corresponding to a single macroscopic Gauss point, and macroscopic structural-level simulations on a finite element model. The results demonstrate excellent agreement with high-fidelity FE solutions while achieving orders-of-magnitude reductions in computational cost. This work establishes a practical pathway for efficient large-scale numerical analyses of nonlinear thermomechanical composite structures
Shot-peening simulations with artificial surface defect using multiple impacts and eigenstrain reconstruction method
No full textInternational audienc
Impact of thermal history on the crystalline structure and morphology of PA11 powder for PBF-LB/P process
No full textInternational audiencePolyamide 11 (PA11), a bio-based thermoplastic, is widely used in laser-based powder bed fusion (PBF-LB/P), an additive manufacturing process also known as selective laser sintering (SLS). However, during SLS, most of the powder remains unsintered and undergoes prolonged thermal exposure in oxidative environments, raising concerns about degradation and limiting powder reuse. The behavior of PA11 under similar conditions, particularly its potential crystalline changes, remains insufficiently explored. Therefore, this study investigates the effects of isothermal treatments at relevant temperatures for PBF-LB/P on both stabilized and unstabilized PA11 powders. Using complementary techniques, Small-Angle X-ray Scattering, Wide-Angle X-ray Scattering, and Differential Scanning Calorimetry, we analyzed crystalline structural and morphological changes. Our results show that oxidative degradation affects crystal surfaces, increasing interfacial energy, and reducing melting temperature and enthalpy. It also lowers the Brill transition temperature, pointing to an underexplored impact of oxidation on PA11. These findings reveal that oxidation can progressively modify the crystalline regions, challenging the traditional view that degradation occurs only in amorphous phase. This emphasizes the importance of improved thermal control during the PBF-LB/P process to enable more sustainable and efficient powder reuse in additive manufacturing
A harmonic balance normal form parametrisation for single mode reduction of nonlinear vibrating systems
No full textInternational audienceThis paper introduces a model-order reduction technique for lightly damped nonlinear vibrating systems. By combining calculation details that are specific to the harmonic balance method, the asymptotic numerical method, and the normal form style parametrisation for invariant manifolds, a complete procedure that can cope with single-mode reduction is detailed. Introducing harmonic decomposition in the process allows for a different treatment of the temporal information of the solution, which comes with advantages as compared to normal form expansions based on polynomial expansions. The computation proceeds with two nested loops on both the harmonics and the polynomial degree expansion. A decisive advantage of the procedure is its ability to compute a new expansion from a known solution, which allows the derivation of amplitudedependent piecewise reduced order models (ROMs), together with an integrated procedure that can switch from the invariant manifolds computation attached to either fixed points or limit cycles. Once the validity limit of a first expansion is met, the procedure can restart from a point where convergence is reached and produce a new ROM. This feature has the potential to overcome the well-known limitations of asymptotic expansions associated with the parametrisation method for invariant manifolds, and is derived here only for conservative systems. The whole analysis also clearly establishes the links existing between the normal form approach and computations based on the harmonic balance combined with the asymptotic numerical method. Examples of increasing complexity, starting from a Duffing equation, a two-degree-of-freedom system and a finite element beam model, are analysed, and comparisons with existing techniques are provided
EDUCAUSE 2025 French Delegation Report - EN: Visits to Vanderbilt University Arizona State University
No full textFor twelve years now, the French EDUCAUSE Delegation has offered the French Higher Educationcommunity an international perspective on digital issues, bringing together complementary profiles capable ofassessing key trends and major prospects. This approach has taken the form of continuous participation inthe EDUCAUSE Annual Conferences since 2013, which has been significantly enriched over the years. Froman initial philosophy of simple participation, the members of this Delegation have gradually become involvedat different levels of EDUCAUSE, both within the conferences and beyond (see infographic below).We now have twenty-four presentations that have been selected on a range of topical issues. Several of ourmembers are also involved each year as proposal reviewers for the EDUCAUSE Annual Conference, as wellas members of the respective Program Committees (in 2019, 2023, and 2026). Finally, some of us are part ofthe Expert Panels involved in the EDUCAUSE Horizon Report and the annual Top 10 Issues, are active invarious Community Groups (XR and Learning Spaces in particular), or are involved as writers of referencearticles and translators of tools.We should also mention the more than thirty on-site visits that have been conducted since 2013 before orafter the EDUCAUSE Annual Conferences, which, beyond the remarkable welcome we have always received,have provided us with invaluable insight into the field through high-quality exchanges
Towards adaptive sustainable scheduling within lithium-ion battery production
No full textInternational audienceThis study studies the increasing complexity of modern manufacturing scheduling, where efficiency, quality, and sustainability must be jointly optimized under flexible machine and operator constraints. Integrating real-time feedback from digital twins into optimization frameworks has emerged as a powerful approach, enabling adaptive and data-informed decision-making. By combining exact methods and metaheuristics, such frameworks can navigate the multi-objective landscape of contemporary production systems effectively. Looking forward, the adoption of surrogate models offers a promising alternative to further enhance performance. By approximating expensive simulations or high-fidelity digital twin responses, surrogate models can significantly reduce computational costs while maintaining solution accuracy
Methodological Proposal for the Deployment of PLM Software Supported by a Digital Adoption Platform and AI Models
No full textInternational audienc
Understanding the bi-modal fatigue behavior of the case-hardened M50NiL steel
No full textInternational audienceThe gears used in aircraft engines are typically made from high-strength steels reinforced by thermochemical treatments (TCT). These treatments increase surface fatigue strength through microstructural modifications, enhancing hardness and adding compressive residual stresses. In some cases, the combination of material, TCT, and applied stress can lead to a bi-modal fatigue behavior, notably in failures at the gear tooth root. This work investigates the bi-modal fatigue response of M50NiL case-hardened steel by characterizing and analyzing crack initiation mechanisms to propose a relevant fatigue modeling approach. A comprehensive experimental fatigue test campaign was carried out on notched specimens under plane bending and on gear specimens using a Single Tooth Bending Fatigue (STBF) method. The resulting Wöhler diagram shows significant scatter in fatigue life for several stress levels, suggesting a bi-modal behavior with two distinct populations. Fractographic analyses confirmed the competition between two different crack initiation mechanisms depending on stress level and number of cycles to failure. A statistical analysis using a mixture model also indicates that a bimodal distribution best represents the results. Accordingly, a probabilistic model is proposed to describe the bi-modal fatigue behavior from a global perspective, based on the maximum applied hot-spot surface stress for a fixed stress ratio. Finally, a complementary local stress analysis shows that the combined effect of stress and material property distributions significantly influences local maximum stress variation. Correcting for these factors reduces scatter in the bi-modal stress levels
