https://sam.ensam.eu:443 The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. Sun, 17 May 2026 02:24:31 GMT 2026-05-17T02:24:31Z http://hdl.handle.net/10985/26841 Hybrid homogenization neural networks for periodic composites CHEN, Qiang; ZHAO, Wenhui; XIAO, Ce; YANG, Zhibo; CHATZIGEORGIOU, George; MERAGHNI, Fodil; CHEN, Xuefeng 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. Sat, 01 Nov 2025 00:00:00 GMT http://hdl.handle.net/10985/26841 2025-11-01T00:00:00Z CHEN, Qiang ZHAO, Wenhui XIAO, Ce YANG, Zhibo CHATZIGEORGIOU, George MERAGHNI, Fodil CHEN, Xuefeng 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. 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.