HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
Multiphysics numerical modeling of semi-crystalline polymer and microstructure informed metal alloys
No full textInternational audienc
Prediction of fretting fatigue damage under variable loading blocks: Effect of plasticity
No full textInternational audienceThis research paper investigates the lifetime span and the crack propagation mechanism for a steel cylinder/plane contact under severe variable plastic fretting fatigue conditions that are representative of high pressure dynamic flexible pipe risers. An FEA model was used to predict the total life of the contact by separating the crack nucleation and the crack propagation mechanisms. Good correlation between numerical predictions and experimental results was obtained. Since fretting is well known for generating important stress gradients, a non-local critical distance method is applied to estimate the number of cycles required for crack nucleation. The crack propagation was addressed using Kujawski’s fatigue crack driving parameter with Paris law. The variable loading conditions were investigated considering decreasing tangential force and fatigue stress loading blocks and comparing various elastic and elastic-plastic material responses. Since significant plastic deformations are generated during the initial loading blocks (the highly charged block), better predictions of crack extension or fretting fatigue endurance are achieved using elastoplastic hypotheses (monotonic or cyclic hardening). By contrast, elastic assumption which is less accurate provides systematic conservative predictions
Quelle mobilisation des données du passé pour penser le futur des forêts ?: Etat des lieux et expérimentation inter- et transdisciplinaire au sein du FORESTT-HUB (PEPR FORESTT)
No full textInternational audienceLes forêts sont des socio-écosystèmes façonnés par une longue coévolution avec les sociétés humaines. En les replaçant dans ce temps long, les sciences du passé (histoire environnementale, écologie historique, paléoécologie, archéologie, génétique) sont pertinentes pour éclairer leur fonctionnement actuel et accompagner les choix de gestion. Leurs apports sont multiples : soutien à la biologie de la conservation et à l’écologie de la restauration, participation aux changements de paradigmes, et mise en perspective des enjeux écologiques via l’histoire longue de la réflexivité environnementale et des controverses. De même, les modélisations climatiques et environnementales et les prospectives se nourrissent des données historiques pour construire une gamme de scénarii futurs et aider à la prise de décision. Pourtant, la pertinence des données passées et les modalités selon lesquelles elles sont mobilisées pour penser les forêts de demain ont peu été explorées.Cette communication questionne la place des données du passé dans la construction de récits du futur (modélisations, prospectives…) en identifiant pourquoi, comment, pour qui et par qui celles-ci sont mobilisées. Elle s’intéresse aux différences épistémologiques entre les récits de trajectoires et controverses passées étayés sur une analyse critique de données, et les récits rétrospectifs servant de base à la construction de prospectives. Elle s’interroge sur le rôle du chercheur en sciences du passé dans la mobilisation de ces données (simple passeur de savoir ou implication active à la prospective) et sur son positionnement vis-à vis de l’action. Nous nous appuierons sur les recherches inter- et transdisciplinaires du FOREST-HUB pour explorer l’hypothèse selon laquelle les récits sur les trajectoires socio-écologiques et les controverses passées peuvent favoriser le dialogue entre chercheurs et acteurs forestiers (propriétaires, gestionnaires, usagers), contribuer à dépasser certains blocages, modifier des perceptions, fournir de nouveaux leviers et innover pour penser d’autres futurs pour les forêts
Open Review of "A gradient-enhanced approach for stable finite element approximations of reaction-convection-diffusion problems"
No full textThis is the Open Review of article https://doi.org/10.46298/jtcam.15788 published by JTCA
Energy-Consistent Multi-Input Multi-Output Multi-Horizon Extreme Learning Machines with Embedded Reconciliation for Multi-Source Power Forecasting
No full textInternational audienceForecasting energy production from multiple sources is essential for balancing supply and demand, especially in island electricity grids with limited interconnections and no nuclear baseload capacity. This work introduces a multi-input, multi-output, multi-horizon (MIMO-MH) framework that combines Extreme Learning Machines (ELM) with forecast reconciliation. The ELM core couples the robustness of linear closed-form training with the expressiveness of random feature mappings, enabling deep-learning-like accuracy at negligible computational cost. The proposed MIMO-MH jointly predicts thermal, hydropower, solar PV, wind, and import generation over 1-24 h horizons, ensuring built-in coherence across both sources and lead times. Validation on seven years of Corsican grid data (2016-2022, hourly resolution) shows: (i) consistent forecasts across all horizons, (ii) 30-45% error reduction versus persistence, and outperforming NeuralProphet and TimeGPT-1, and (iii) subminute training times compatible with operational deployment. Because MIMO-MH forecasts are inherently reconciled, they naturally preserve physical energy balances without any post-processing. This effectively enforces the principle of energy conservation across sources and horizons, making forecast models directly actionable tools for dispatch, reserve sizing, and market bidding. By linking statistical coherence to operational reliability, this framework bridges predictive modeling and grid management, addressing a long-standing limitation of renewable integration in non-interconnected systems.</div
Identification of microprotein-coding intronic polyadenylation isoforms and function in genotoxic anticancer drug response
No full textInternational audienceBackground: Many transcript isoforms generated by intronic polyadenylation (IPA) encode isoforms of canonical proteins. Microproteins are an emerging class of small proteins translated from small open reading frames (sORFs) in noncoding RNAs and mRNAs, but their production by IPA isoforms is unknown.Results: Here, by crossing 3′-seq, Ribo-Seq, and mass-spectrometry data, we identify 297 genes with a microprotein-coding IPA isoform terminating in a 5′UTR intron (coined miP-5′UTR-IPA isoform). By 3′-seq and long-read RNA-seq analyses in lung cancer cells treated with cisplatin, a DNA-cross-linking anticancer drug, we find that cisplatin globally favors the expression of (miP-5′UTR-)IPA isoforms relative to fulllength mRNAs, mainly by decreasing the latter through an inhibition of transcription processivity in a FANCD2 and senataxin-dependent manner. The cisplatin-regulated miP-5′UTR-IPA isoform in the PRKAR1B gene is translated, as it is associated with light polysome fractions and contains Ribo-Seq-supported sORFs in its alternative last exon, and the microprotein (PRKAR1B-IPA-miP2) encoded by its sORF#2 is detected by Western blot and immunofluorescence. CRISPR editing of either the IPA site or the sORF#2 initiation site leads to decreased cell growth inhibition by cisplatin and camptothecin, another genotoxic drug. Mechanistically, PRKAR1B-IPA-miP2 promotes p53 protein induction by cisplatin. Finally, 70 miP-5′UTR-IPA isoforms are detected in normal cells, and 143 are upregulated by cisplatin.Conclusions: Here, we show that IPA isoforms are a novel source of microproteins, and we reveal the novel paradigm of miP-5′UTR-IPA genes that produce both a canonical full-length mRNA and a microprotein-coding IPA isoform
Two-Dimensional Stochastic Structural Geomodeling with Deep Generative Adversarial Networks
No full textInternational audienceAbstract Structural geological modeling is aimed at finding a representation of geological units. This is a complex ill-posed problem, and the data may be sparse and of varying quality, leading to multiple geological models consistent with them. Despite continuous advances for decades in geological modeling, recent studies still show some unresolved industrial challenges. Consequently, this work aims to improve geological modeling by proposing a novel approach with a focus on improving uncertainty management. A stochastic approach is developed based on deep generative methods, namely generative adversarial networks (GANs). Thanks to a synthetic dataset, a GAN is trained to generate two-dimensional unconditional geomodels that are plausible. Subsequently, a Bayesian inversion is performed with a Metropolis-adjusted Langevin algorithm (MALA) to produce geomodels consistent with field data. The proposed approach is validated on different conditioning data. For each case, the approach is able to successfully produce a variety of geomodels that can be linked to different geological settings. The uncertainties on geological units are measured by Shannon entropy on the generated models
Group Morphology Fixed Points on Homogenous Spaces for Deep Learning Equivariant Networks
No full textInternational audienceThis paper presents a theoretical framework that integrates mathematical morphology with deep learning, focusing on the construction of neural network layers that inherently converge to fixed points through iterative application. Drawing from the principles of idempotence and convergence in complete lattices, we propose a class of nonlinear operators that can be embedded into deep architectures to enhance stability and reduce parameter complexity. The framework is extended to group-equivariant settings on homogeneous spaces, enabling the design of layers that respect symmetries in the data. We formalize the construction of equivariant fixed-point layers using group convolutions and max-plus algebra operators, and we characterize their convergence properties. This work lays the mathematical foundation for future implementations of fixed-point layers in deep learning, particularly in contexts where equivariance and stability are desirable
Thermal shock resistance of a tungsten diamond composite under extreme heat loads
No full textInternational audienceThermal shock resistance of a tungsten diamond composite under extreme heat loads References Limitations of bulk tungsten PFCs The very high thermal conductivity of Chemical Vapor Deposition (CVD) diamond (~2000 W.m -1 .K -1 at 20°C) allows outstanding performances under thermal shocks with T retention [2]. Unfortunately, its erosion properties are comparable to graphite [3].2 .s 0.5 ) to simulate ITER-relevant fast transients like ELMs, disruptions and runaways, • up to 20 cycles of 1 s at 100 MW.m -2 to simulate reattachment in the ITER divertor.</p
