https://sam.ensam.eu:443 The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. Sun, 17 May 2026 03:33:39 GMT 2026-05-17T03:33:39Z http://hdl.handle.net/10985/26061 Advanced Meta-Modeling framework combining Machine Learning and Model Order Reduction towards real-time virtual testing of woven composite laminates in nonlinear regime EL FALLAKI IDRISSI, Mohammed; PASQUALE, Angelo; MERAGHNI, Fodil; PRAUD, Francis; CHINESTA SORIA, Francisco This paper presents an advanced meta-modeling framework that efficiently combines Machine Learning and Model Order Reduction (MOR) techniques for real-time virtual testing of woven composite materials. The framework is specifically de signed to develop a multiparametric solution capable of accurately predicting the macroscopic nonlinear stress–strain curves of woven composite laminates submitted to loading–unloading paths. It takes into account five key microstructural parameters: yarn weft width, yarn warp width, yarn spacing, fabric thickness as well as the reinforcement orientation. The methodology employs the Proper Orthogonal Decomposition (POD) technique to decompose the stress–strain curves, extracting principal features that effectively characterize the overall composite’s response. Subsequently, a Random Forest machine learning model is applied to interpolate these features across the microstructural parameter space, allowing for rapid retrieval of corresponding features for any new laminate configuration in the nonlinear regime. A key advantages of this approach is its capacity to dynamically generate extensive virtual test databases, in real-time, across a wide range of composite laminate configurations. This capability provides a comprehensive and efficient tool for analyzing and optimizing composite performance while substantially reducing both experimental and computational costs. Furthermore, to enhance usability for engineers and researchers, this multiparametric solution has been integrated into a user-friendly Graphical User Interface (GUI) application. This GUI empowers users to easily explore various laminate configurations, visualize results, and conduct virtual testing, establishing the framework as a powerful tool for real-time virtual testing and in-depth analysis of microstructural effects on composite materials. Sat, 01 Mar 2025 00:00:00 GMT http://hdl.handle.net/10985/26061 2025-03-01T00:00:00Z EL FALLAKI IDRISSI, Mohammed PASQUALE, Angelo MERAGHNI, Fodil PRAUD, Francis CHINESTA SORIA, Francisco This paper presents an advanced meta-modeling framework that efficiently combines Machine Learning and Model Order Reduction (MOR) techniques for real-time virtual testing of woven composite materials. The framework is specifically de signed to develop a multiparametric solution capable of accurately predicting the macroscopic nonlinear stress–strain curves of woven composite laminates submitted to loading–unloading paths. It takes into account five key microstructural parameters: yarn weft width, yarn warp width, yarn spacing, fabric thickness as well as the reinforcement orientation. The methodology employs the Proper Orthogonal Decomposition (POD) technique to decompose the stress–strain curves, extracting principal features that effectively characterize the overall composite’s response. Subsequently, a Random Forest machine learning model is applied to interpolate these features across the microstructural parameter space, allowing for rapid retrieval of corresponding features for any new laminate configuration in the nonlinear regime. A key advantages of this approach is its capacity to dynamically generate extensive virtual test databases, in real-time, across a wide range of composite laminate configurations. This capability provides a comprehensive and efficient tool for analyzing and optimizing composite performance while substantially reducing both experimental and computational costs. Furthermore, to enhance usability for engineers and researchers, this multiparametric solution has been integrated into a user-friendly Graphical User Interface (GUI) application. This GUI empowers users to easily explore various laminate configurations, visualize results, and conduct virtual testing, establishing the framework as a powerful tool for real-time virtual testing and in-depth analysis of microstructural effects on composite materials. http://hdl.handle.net/10985/22637 Multiparametric modelling of composite materials based on non-intrusive PGD informed by multiscale analyses: Application for real-time stiffness prediction of woven composites EL FALLAKI IDRISSI, Mohammed; PRAUD, Francis; CHAMPANEY, Victor; CHINESTA SORIA, Francisco; MERAGHNI, Fodil In this paper, a multiparametric solution of the stiffness properties of woven composites involving several microstructure parameters is performed. For this purpose, non-intrusive PGD-based methods are employed. From offline pre-computed solutions generated through a full-field multiscale modeling, the proposed method approximates the multidimensional solution as a sum of products of one-dimensional functions each depending on a single variable. The present work aims at providing an accurate approximation of this multiparametric solution with lower computational cost for dataset generation. Thus, a comparative analysis of three non-intrusive PGD formulations (SSL, s-PGD and ANOVA-PGD) is carried out. The obtained results reveal and demonstrate that the ANOVA-PGD model works well for approximating the stiffness properties over the entire parameter space, i.e., along its boundary as well as inside it, by using only few pre-computed high-fidelity solutions. Finally, a GUI application is developed to exploit this multiparametric solution by incorporating other composite weave architectures. This application could be easily used by engineers and composite designers, to deduce, in real-time, the macroscopic properties of woven composite for a given set of microstructural parameters by simply varying the cursors and without any microstructure generation and meshing nor FE computations using periodic homogenization. Thu, 01 Sep 2022 00:00:00 GMT http://hdl.handle.net/10985/22637 2022-09-01T00:00:00Z EL FALLAKI IDRISSI, Mohammed PRAUD, Francis CHAMPANEY, Victor CHINESTA SORIA, Francisco MERAGHNI, Fodil In this paper, a multiparametric solution of the stiffness properties of woven composites involving several microstructure parameters is performed. For this purpose, non-intrusive PGD-based methods are employed. From offline pre-computed solutions generated through a full-field multiscale modeling, the proposed method approximates the multidimensional solution as a sum of products of one-dimensional functions each depending on a single variable. The present work aims at providing an accurate approximation of this multiparametric solution with lower computational cost for dataset generation. Thus, a comparative analysis of three non-intrusive PGD formulations (SSL, s-PGD and ANOVA-PGD) is carried out. The obtained results reveal and demonstrate that the ANOVA-PGD model works well for approximating the stiffness properties over the entire parameter space, i.e., along its boundary as well as inside it, by using only few pre-computed high-fidelity solutions. Finally, a GUI application is developed to exploit this multiparametric solution by incorporating other composite weave architectures. This application could be easily used by engineers and composite designers, to deduce, in real-time, the macroscopic properties of woven composite for a given set of microstructural parameters by simply varying the cursors and without any microstructure generation and meshing nor FE computations using periodic homogenization. http://hdl.handle.net/10985/26243 Multiscale Thermodynamics-Informed Neural Networks (MuTINN) for nonlinear structural computations of recycled thermoplastic composites SEKKAL, S.E.; EL FALLAKI IDRISSI, Mohammed; MERAGHNI, Fodil; CHATZIGEORGIOU, George; CHINESTA SORIA, Francisco Fiber-reinforced thermoplastic composites are increasingly valued for their light-weight properties, mechanical performance, and recyclability, yet the recycling process introduces microstructural heterogeneities that degrade their mechanical behavior. To address the challenges from a modeling point of view, this study proposes a Multiscale Thermodynamics-Informed Neural Network (MuTINN) approach to predict the nonlinear, anisotropic response of recycled glass fiber-reinforced polyamide 6 composites, with the primary aim of enabling structural simulations in significantly reduced time compared to traditional FE² approaches. The MuTINN framework integrates thermodynamic principles with artificial neural networks (ANNs) to capture the evolution of internal state variables and Helmholtz free energy, eliminating the need for memory-based networks. Finite element simulations of representative volume elements (RVEs) under diverse loading conditions are utilized to provide off-line data for the MuTINN. The latter accurately predicts stress, strain, and energy quantities, accounting for the anisotropic and heterogeneous nature of recycled materials. While trained using numerical simulations at 0◦ and 90◦ orientation specimens, the proposed framework succesfully predicts the response for specimens with 45◦ orientation with error in the maximum stress level up to 1.6%. The model is implemented into commercial finite element analysis (FEA) software via a Meta-UMAT framework, allowing efficient macroscale simulations. Validation against experimental data and finite element-based periodic homogenization confirms the framework’s accuracy for structural computations. Tue, 01 Apr 2025 00:00:00 GMT http://hdl.handle.net/10985/26243 2025-04-01T00:00:00Z SEKKAL, S.E. EL FALLAKI IDRISSI, Mohammed MERAGHNI, Fodil CHATZIGEORGIOU, George CHINESTA SORIA, Francisco Fiber-reinforced thermoplastic composites are increasingly valued for their light-weight properties, mechanical performance, and recyclability, yet the recycling process introduces microstructural heterogeneities that degrade their mechanical behavior. To address the challenges from a modeling point of view, this study proposes a Multiscale Thermodynamics-Informed Neural Network (MuTINN) approach to predict the nonlinear, anisotropic response of recycled glass fiber-reinforced polyamide 6 composites, with the primary aim of enabling structural simulations in significantly reduced time compared to traditional FE² approaches. The MuTINN framework integrates thermodynamic principles with artificial neural networks (ANNs) to capture the evolution of internal state variables and Helmholtz free energy, eliminating the need for memory-based networks. Finite element simulations of representative volume elements (RVEs) under diverse loading conditions are utilized to provide off-line data for the MuTINN. The latter accurately predicts stress, strain, and energy quantities, accounting for the anisotropic and heterogeneous nature of recycled materials. While trained using numerical simulations at 0◦ and 90◦ orientation specimens, the proposed framework succesfully predicts the response for specimens with 45◦ orientation with error in the maximum stress level up to 1.6%. The model is implemented into commercial finite element analysis (FEA) software via a Meta-UMAT framework, allowing efficient macroscale simulations. Validation against experimental data and finite element-based periodic homogenization confirms the framework’s accuracy for structural computations. http://hdl.handle.net/10985/24581 PGD non-intrusive pour la simulation multiparamétrique en temps réel du comportement non-linéaire des composites à renforts tissés intégrant les paramètres microstructuraux EL FALLAKI IDRISSI, Mohammed; PRAUD, Francis; CHINESTA SORIA, Francisco; MERAGHNI, Foudil La modélisation multi-échelle non-linéaire par éléments finis des composites reste aujourd’hui un défi dans des applications industrielles. En effet, son utilisation nécessite une puissance de calcul élevée et donc souvent couteuse en temps, ce qui constitue un verrou majeur rendant impossible la prise de décision en temps réel. C’est pourquoi, la construction de modèles paramétriques est un outil primordial pour pouvoir analyser et optimiser les structures composites. À cette fin, une approche basée sur la PGD non-intrusive est proposée afin d’offrir une solution multi-paramétrique capable d’évaluer quasi-instantanément la réponse macroscopique non-linéaire des composites tissés en fonction de certains paramètres de microstructure. La solution paramétrique est développée pour un matériau composite à renfort tissé (sergé 2,2) et à matrice polyamide 66 pour des séquences d’empilement symétriques et équilibrées de [± 0 ]s et [± 45 ]s . Cette solution permet d’analyser le comportement macroscopique du composite non linéaire lors de divers types de sollicitations, notamment monotones, charge-décharge et en fluage-recouvrance. Elle prend en considération à la fois le caractère viscoélastique-viscoplastique de la matrice et les dommages subis par les torons. Sat, 01 Jul 2023 00:00:00 GMT http://hdl.handle.net/10985/24581 2023-07-01T00:00:00Z EL FALLAKI IDRISSI, Mohammed PRAUD, Francis CHINESTA SORIA, Francisco MERAGHNI, Foudil La modélisation multi-échelle non-linéaire par éléments finis des composites reste aujourd’hui un défi dans des applications industrielles. En effet, son utilisation nécessite une puissance de calcul élevée et donc souvent couteuse en temps, ce qui constitue un verrou majeur rendant impossible la prise de décision en temps réel. C’est pourquoi, la construction de modèles paramétriques est un outil primordial pour pouvoir analyser et optimiser les structures composites. À cette fin, une approche basée sur la PGD non-intrusive est proposée afin d’offrir une solution multi-paramétrique capable d’évaluer quasi-instantanément la réponse macroscopique non-linéaire des composites tissés en fonction de certains paramètres de microstructure. La solution paramétrique est développée pour un matériau composite à renfort tissé (sergé 2,2) et à matrice polyamide 66 pour des séquences d’empilement symétriques et équilibrées de [± 0 ]s et [± 45 ]s . Cette solution permet d’analyser le comportement macroscopique du composite non linéaire lors de divers types de sollicitations, notamment monotones, charge-décharge et en fluage-recouvrance. Elle prend en considération à la fois le caractère viscoélastique-viscoplastique de la matrice et les dommages subis par les torons. http://hdl.handle.net/10985/27122 Micromechanics-Informed Neural Networks for Periodic Homogenization of Thermocondcutive Behavior in Unidirectional Composites with Cylindrically Orthotropic Graphite Fibers XIAO, Ce; CHEN, Qiang; EL FALLAKI IDRISSI, Mohammed; YANG, Zhibo; CHEN, Xuefeng; CHATZIGEORGIOU, George; MERAGHNI, Fodil A micromechanics-informed neural network framework is developed for homogenization of periodic unidirectional thermoconductive composites with cylindrically orthotropic fibers. The framework hard-imposes the steady-state governing heat conduction equations within the network architecture, enabling accurate capture of singular heat flux fields at the fiber center that are challenging for conventional approaches. In contrast, continuity and periodicity conditions are enforced via boundary collocation points in the loss function. Validation against finite element simulations across a wide range of fiber volume fractions shows that accurate and converged temperature distributions can be achieved after 9000 training epochs using 8-16 harmonic terms. Additional higher-order harmonics are difficult to train reliably and may degrade predictions. While strong agreement is observed in the matrix heat flux distributions, noticeable discrepancies persist in the fiber phase due to varying ability to capture the singular heat flux fields. Furthermore, uniform collocation points converge faster than random points during solution refinement. Finally, transfer learning is employed to accelerate training for new configurations, allowing the network to achieve comparable accuracy after only 2000 training epochs, which is substantially fewer than the 9,000 epochs required when training from scratch. Sat, 01 Nov 2025 00:00:00 GMT http://hdl.handle.net/10985/27122 2025-11-01T00:00:00Z XIAO, Ce CHEN, Qiang EL FALLAKI IDRISSI, Mohammed YANG, Zhibo CHEN, Xuefeng CHATZIGEORGIOU, George MERAGHNI, Fodil A micromechanics-informed neural network framework is developed for homogenization of periodic unidirectional thermoconductive composites with cylindrically orthotropic fibers. The framework hard-imposes the steady-state governing heat conduction equations within the network architecture, enabling accurate capture of singular heat flux fields at the fiber center that are challenging for conventional approaches. In contrast, continuity and periodicity conditions are enforced via boundary collocation points in the loss function. Validation against finite element simulations across a wide range of fiber volume fractions shows that accurate and converged temperature distributions can be achieved after 9000 training epochs using 8-16 harmonic terms. Additional higher-order harmonics are difficult to train reliably and may degrade predictions. While strong agreement is observed in the matrix heat flux distributions, noticeable discrepancies persist in the fiber phase due to varying ability to capture the singular heat flux fields. Furthermore, uniform collocation points converge faster than random points during solution refinement. Finally, transfer learning is employed to accelerate training for new configurations, allowing the network to achieve comparable accuracy after only 2000 training epochs, which is substantially fewer than the 9,000 epochs required when training from scratch.