Publications

Reliable detection of barely visible impact damage is critical to ensure the structural integrity of composite components in service, particularly in safety-critical applications such as pressure vessels and transportation systems. This study presents a solution for detecting such damage in woven glass fiber-reinforced thermoplastic composites using terahertz (THz) time-of-flight tomography and convolutional neural networks. THz provides non-contact, non-ionizing, high-axial-resolution imaging of subsurface and back-surface damage, addressing key limitations of surface-based inspection methods. While THz imaging alone may not always permit conclusive damage identification, we bridge this gap by training neural network classifiers on depth-resolved THz B-scan images using ground truth from co-located X-ray micro-computed tomography. Among several pretrained architectures tested via transfer learning, DenseNet-121 exhibits the highest accuracy. The model remains robust even when trained on truncated B-scans excluding surface indentation features, confirming its ability to detect structural anomalies located internally or on the back surface. This is particularly relevant for applications where back-side access is not feasible. Experimental validation is performed on impacted glass-fiber-reinforced thermoplastic coupons prepared in accordance with ASTM D7136, with damage severity quantified through force–displacement data and micro-tomographic analysis. Labeling for supervised learning conforms to acceptance criteria from industrial standards for composite pressure vessels (ASME BPVC Section X, CGA C-6.2), ensuring regulatory alignment and enabling deployment in quality control workflows. The proposed method minimizes the need for expert interpretation or secondary validation and offers direct applicability to in-service inspection and manufacturing quality control.

Cet article propose une exploration approfondie de la réalité virtuelle (RV) en tant que technologie émergente dans l’enseignement du génie mécanique, en s’appuyant sur sa mise en œuvre sur le campus des Arts et Métiers de Metz. Il met en lumière une série de projets dirigés par des étudiants, qui exploitent la RV pour concevoir des environnements d’apprentissage immersifs. Des études expérimentales menées auprès d’étudiants de premier cycle indiquent que la RV améliore la compréhension des systèmes mécaniques complexes et favorise l’apprentissage collaboratif. L’article présente les principaux avantages de cette technologie, tels qu’un engagement accru des étudiants et une meilleure visualisation, tout en abordant les défis, notamment le mal des transports, les coûts élevés des équipements et la nécessité de former les enseignants. Les perspectives d’avenir sont également examinées, notamment l’extension des applications de la RV à d’autres disciplines académiques, la création d’expériences d’apprentissage interdisciplinaires, et l’intégration de la RV dans les dispositifs pédagogiques existants. L’étude souligne le potentiel transformateur de la RV dans l’enseignement supérieur et sa capacité à préparer les étudiants aux environnements numériques de demain.

Wire Arc Additive Manufacturing (WAAM) is an efficient technology for producing metal parts. It offers high deposition rates, low material waste and reduced machining costs. However, the process's complex multi-physics nature presents several challenges, particularly with regard to residual stresses, deposition accuracy, and internal defects. Internal defects that occur during the WAAM process are difficult to detect in real time or immediately after fabrication without the use of non destructive testing methods.This study focuses on enhancing an experimental protocol that employs a monitoring method analysing the geometry of each deposited layer using a laser sensor. This approach, combined with segmentation techniques, aims to identify and locate internal defects. Specifically, the proposed method relies on segmenting individual layers and comparing normalised values to detect and localise defects. This paper presents the developed methodology and an initial validation of its effectiveness.

Dans un contexte industriel en mutation, les entreprises manufacturières doivent allier flexibilité et performance. Depuis 2017, une recherche appliquée en partenariat avec un équipementier automobile a permis de co-concevoir des produits et des processus d’assemblage pour développer des systèmes de production modulaires et reconfigurables. Une nouvelle approche de modélisation stabilise les architectures produits, facilitant l’identification de familles de produits et l’optimisation des systèmes d’assemblage multiproduits. L’importance du positionnement et du guidage dans l’assemblage a été mise en avant, ainsi qu’un outil d’optimisation pour la sélection des ressources et l’affectation des tâches. Des solutions concrètes, telles que des lignes de production modulaires et des outils adaptables, ont été testées avec succès, réduisant les temps de reconfiguration et améliorant l’efficacité. Le transfert de connaissances a renforcé l’expertise régionale, favorisant l’innovation et la compétitivité dans l’usine du futur.

In this paper, we compare two parametric model order reduction methods, the multi-moment matching method and the interpolation of projection subspaces method for the magneto-quasistatic (MQS) and electro-quasistatic (EQS) problems derived from Maxwell’s equations and discretized with the Finite Element (FE) method. The two problems considered are both governed by the differential–algebraic equations. The material characteristic parameters as well as the geometry parameters have been considered. The applications are two realistic test cases: an EQS model of a transformer bushing under insulation defect uncertainty and a MQS model of a planar inductor with geometric and material variations. The result shows that both methods approximate well global quantities, such as the current or the voltage, as well as the local quantities like field distributions. The multi-moment matching method remains always faster in the online stage, since the reduced basis is not parameter dependent, requiring no reduced basis calculation. The multi-moment matching method requires an affine decomposition of the FE model, which is not easy to obtain when considering geometry parameters. A hybrid method is proposed and tested leading to more accurate results than the interpolation of projection subspaces method but much easier to implement than the multi-moment matching method.

Boundary layers of adiabatic and isothermal curved walls are investigated for a supersonic turbine cascade, including the effects of shock-boundary layer interactions (SBLIs). Wall-resolved large eddy simulations (LES) are performed for a linear cascade of blades with an inlet Mach number of $M_\infty = 2.0$ and Reynolds number based on the axial chord $Re_\infty = 200\,000$. The wall to inlet temperature ratio of the isothermal case is $T_w/T_{\infty}=0.75$, representing a cooled wall. An assessment of the effects of pressure gradient, thermal boundary conditions and SBLIs is presented in terms of the downstream variation of mean flow quantities such as density, temperature, and momentum profiles. The different thermal boundary conditions affect the density and temperature profiles along the boundary layer, where cooling increases the density of the gas near the wall, and reduces its temperature and viscosity. Both of these effects make the momentum profiles fuller and, hence, the boundary layer of the isothermal case is less prone to separate than that of the adiabatic wall. The mean density profiles are also affected by pressure gradients induced by the convex and concave curvatures of the blade, which lead to expansion and compression of the flow, respectively. The analysis of separate terms from the momentum balance equation explains the behavior of various physical mechanisms in the inner and outer regions of the supersonic boundary layers. The importance of mean flow advection, compressibility, and Reynolds stresses is presented in terms of flow acceleration and deceleration. The impact of the SBLIs in the momentum balance mechanisms is also investigated, showing that a combination of compressions and expansions impact the boundary layers by redirecting the flow toward the wall due to the shock formations.

The study performs large-eddy simulations of supersonic free jet flows using the Discontinuous Galerkin Spectral Element Method (DGSEM). The main objective of the present work is to assess the resolution requirements for adequate simulation of such flows with the DGSEM approach. The study looked at the influence of the mesh and the spatial discretization accuracy on the simulation results. The present analysis involves four simulations, incorporating three different numerical meshes and two different orders of spatial discretization accuracy. The numerical meshes are generated with distinct mesh topologies and refinement levels. Detailed descriptions of the grid generation and refinement procedures are presented. The study compares flow property profiles and power spectral densities of velocity components with experimental data. The results show a consistent improvement in the computed data as the simulation resolution increases. This investigation revealed a trade-off between mesh and polynomial refinement, striking a balance between computational cost and the accuracy of large-eddy simulation results for turbulent flow analyses.

This review provides a comprehensive examination of artificial intelligence methods applied to the design, optimization, and performance prediction of composite-based hydrogen storage vessels, with a focus on composite overwrapped pressure vessels. Targeted at researchers, engineers, and industrial stakeholders in materials science, mechanical engineering, and renewable energy sectors, the paper aims to bridge traditional mechanical modeling with evolving AI tools, while emphasizing alignment with standardization and certification re­quirements to enhance safety, efficiency, and lifecycle integration in hydrogen infrastructure. The review begins by introducing HSV types, their material compositions, and key design challenges, including high-pressure durability, weight reduction, hydrogen embrittlement, leakage prevention, and environmental sustainability. It then analyzes conventional approaches, such as finite element analysis, multiscale modeling, and experimental testing, which effectively address aspects like failure modes, fracture strength, liner damage, dome thickness, winding angle effects, crash behavior, crack propagation, charging/discharging dynamics, burst pressure, durability, reliability, and fatigue life. On the other hand, it has been shown that to optimize and predict the characteristics of hydrogen storage vessels, it is necessary to combine the conventional methods with artificial intelligence methods, as conventional methods often fall short in multi-objective optimization and rapid predictive analytics due to computational intensity and limitations in handling uncertainty or complex datasets. To overcome these gaps, the paper evaluates hybrid frameworks that integrate traditional techniques with AI, including machine learning, deep learning, artificial neural networks, evolutionary algorithms, and fuzzy logic. Recent studies demonstrate AI’s efficacy in failure prediction, design optimization to mitigate structural risks, structural health monitoring, material property evaluation, burst pressure forecasting, crack detection, com­posite lay-up arrangement, weight minimization, material distribution enhancement, metal foam ratio optimi­zation, and optimal material selection. By synthesizing these advancements, this work underscores AI’s potential to accelerate development, reduce costs, and improve HSV performance, while advocating for physics-informed models, robust datasets, and regulatory alignment to facilitate industrial adoption.

Faced with the rising number of electric vehicles, the recycling of permanent magnet (PM) rotors of electrical machines is a pivotal concern since PM, for this, application are generally made with Critical Raw Materials, i.e. rare earth materials. Therefore, the development of effective End of Life strategies for PM is essential to mitigate the environmental impact associated with their production and meet the rising demand sustainably. This paper presents a method to reconstruct the magnetization state of the PM on site within the rotor based on external field measurements. This information will be really useful to evaluate the PM state of health in order to evaluate the possibility of reuse of the rotor or to recycle the PM. The process of reconstruction is based on an inverse method and it has been fully simulated using A Finite Element (FE) model for the rotor. It is shown on different rotor topologies (surface PM mounted rotor, PM buried rotor…) that it is possible to determine the magnetization state of the PM.

A new physics-informed deep homogenization neural network (DHN) framework is proposed to identify the homogenized and local behaviors in periodic heterogeneous microstructures. To achieve this, the displacement field is decomposed into averaged and fluctuating contributions, with the local unit cell solution obtained via neural networks subject to periodic boundary conditions. The periodic microstructures are divided into sub­domains representing the fiber and matrix phases, respectively. A key contribution of the proposed method is the marriage of elasticity solution and physics-informed neural network to each phase of the composite, namely, the fiber phase as a mesh-free component whose fluctuating displacements are expanded using a discrete Fourier transform, and the matrix phase using material points with fluctuating displacements handled through fully connected neural network layers. The interfacial continuity conditions are enforced by minimizing the traction and displacement differences at separate material points along the interface. Transfer learning is exploited further to facilitate training new microstructures from pre-trained geometry. This hybrid formulation inherently satisfies stress equilibrium equations within the fiber, while efficiently handling the periodic boundary conditions of hexagonal and square unit cells via a series of trainable sinusoidal functions. The innovative use of distinct neural network architectures enables accurate and efficient predictions of displacement and stress when discontinuities are present in the solution fields across the interface. We validate the proposed DHN with the finite-element predictions for unidirectional composites comprised of elastic fiber significantly stiffer than the matrix, under various volume fractions and loading conditions.