https://sam.ensam.eu:443 The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. Wed, 15 Apr 2026 22:03:47 GMT 2026-04-15T22:03:47Z http://hdl.handle.net/10985/19602 Extended Mean-Field Homogenization of Viscoelastic-Viscoplastic Polymer Composites Undergoing Hybrid Progressive Degradation Induced by Interface Debonding and Matrix Ductile Damage CHEN, Qiang; CHATZIGEORGIOU, George; MERAGHNI, Fodil In this contribution, a probabilistic micromechanics damage framework is presented to predict the macroscopic stress-strain response and progressive damage in unidirectional glass-reinforced thermoplastic polymer composites. Motivated by different failure modes observed experimentally, the damage mechanism in the vicinity of the fibers (namely, the interphase) is characterized by initiating and growing voids. The mechanisms can be formulated through a Weibull probabilistic density function. In contrast, the ductile progressive degradation of matrix initial stiffness is analyzed via the continuum damage theory. To accommodate different damage mechanisms in the matrix and the interphase, a three-phase Mori-Tanaka (MT) method and transformation field analysis approach (TFA) are established within a unified framework that allows simulation of both ductile and discrete damages in different phases. Moreover, the rate-dependent viscoelastic and viscoplastic response of the polymer matrix phase is modelled through a phenomenological model consisting of four Kelvin-Voigt branches and a viscoplastic branch under the thermodynamics framework. The reliability and efficiency of the modified mean-field damage model, based on TFA and Mori-Tanaka scheme, are assessed by comparing the simulated stress-strain response against full-field Abaqus simulations under both unidirectional and multiaxial nonproportional loading paths at different loading rates. The developed model provides an efficient alternative to the finite-element based full-field homogenization schemes or other mean-field micromechanics techniques that may be compared, as well as a framework for a potential extension of the theory for simulating damage evolution in composites with random reinforcement orientations. Wed, 01 Jan 2020 00:00:00 GMT http://hdl.handle.net/10985/19602 2020-01-01T00:00:00Z CHEN, Qiang CHATZIGEORGIOU, George MERAGHNI, Fodil In this contribution, a probabilistic micromechanics damage framework is presented to predict the macroscopic stress-strain response and progressive damage in unidirectional glass-reinforced thermoplastic polymer composites. Motivated by different failure modes observed experimentally, the damage mechanism in the vicinity of the fibers (namely, the interphase) is characterized by initiating and growing voids. The mechanisms can be formulated through a Weibull probabilistic density function. In contrast, the ductile progressive degradation of matrix initial stiffness is analyzed via the continuum damage theory. To accommodate different damage mechanisms in the matrix and the interphase, a three-phase Mori-Tanaka (MT) method and transformation field analysis approach (TFA) are established within a unified framework that allows simulation of both ductile and discrete damages in different phases. Moreover, the rate-dependent viscoelastic and viscoplastic response of the polymer matrix phase is modelled through a phenomenological model consisting of four Kelvin-Voigt branches and a viscoplastic branch under the thermodynamics framework. The reliability and efficiency of the modified mean-field damage model, based on TFA and Mori-Tanaka scheme, are assessed by comparing the simulated stress-strain response against full-field Abaqus simulations under both unidirectional and multiaxial nonproportional loading paths at different loading rates. The developed model provides an efficient alternative to the finite-element based full-field homogenization schemes or other mean-field micromechanics techniques that may be compared, as well as a framework for a potential extension of the theory for simulating damage evolution in composites with random reinforcement orientations. http://hdl.handle.net/10985/25081 Fully coupled nonlinear thermomechanical modeling of composites using mean-field Mori–Tanaka scheme combined with TFA theory CHATZIGEORGIOU, George; MERAGHNI, Fodil; CHEN, Qiang This article aims at proposing a new mean-field homogenization framework for the study of composites undergoing fully coupled thermomechanical processes. Strongly dissipative phenomena during high or moderate cyclic loading conditions in a structural component made of a composite material cause significant interplay between mechanical and thermal fields. The proposed framework attempts to address such effect by combining the Mori-Tanaka scheme and the Transformation Field Analysis (TFA) theory and by developing a multiscale framework capable of taking into account thermomechanically coupled processes. The numerical simulations performed in the examples section and validations with computations using periodic homogenization and full-structure analysis demonstrate the proposed strategy’s accuracy and robustness. The numerical simulation of a tube shows the model’s ability to simulate cyclic loading conditions with significantly less computational cost than the alternative FE 2 computation strategies. This drastic computational time reduction is due to the semi-analytical formalism of the micromechanics methodology. Sat, 01 Jun 2024 00:00:00 GMT http://hdl.handle.net/10985/25081 2024-06-01T00:00:00Z CHATZIGEORGIOU, George MERAGHNI, Fodil CHEN, Qiang This article aims at proposing a new mean-field homogenization framework for the study of composites undergoing fully coupled thermomechanical processes. Strongly dissipative phenomena during high or moderate cyclic loading conditions in a structural component made of a composite material cause significant interplay between mechanical and thermal fields. The proposed framework attempts to address such effect by combining the Mori-Tanaka scheme and the Transformation Field Analysis (TFA) theory and by developing a multiscale framework capable of taking into account thermomechanically coupled processes. The numerical simulations performed in the examples section and validations with computations using periodic homogenization and full-structure analysis demonstrate the proposed strategy’s accuracy and robustness. The numerical simulation of a tube shows the model’s ability to simulate cyclic loading conditions with significantly less computational cost than the alternative FE 2 computation strategies. This drastic computational time reduction is due to the semi-analytical formalism of the micromechanics methodology. 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/10985/19877 Electromechanical response of multilayered piezoelectric BaTiO3/PZT-7A composites with wavy architecture TU, Wenqiong; CHEN, Qiang Electromechanical laminated composites with piezoelectric phases are increasingly being explored as multifunctional materials providing energy conversion between electric and mechanical energies. The current work explores thus-far undocumented combined microstructural effects of amplitude-to-wavelength ratio, volume fraction, poling direction of piezoelectric phases on both the homogenized properties and localized stress/electric field distributions in multilayered configurations under fully coupled electro-mechanical loading. In particular, the Multiphysics Finite-Volume Direct Averaging Micromechanics (FVDAM) and its counterpart, an in-house micromechanical multiphysics finite-element model, are utilized to investigate the homogenized and localized responses of wavy multilayered piezoelectric BaTiO3/PZT-7A architectures. These two methods generate highly agreeable results. Moreover, we critically examine the convergence of the finite-volume and finite element-based approaches via the Average Stress Theorem and Average Electric Displacement Theorem. The comparison shows the finite volume-based approach possesses a better numerical convergence. This study illustrates the FVDAM’s ability toward the analysis and design of engineered multilayered piezoelectric materials with wavy architecture. Fri, 01 Jan 2021 00:00:00 GMT http://hdl.handle.net/10985/19877 2021-01-01T00:00:00Z TU, Wenqiong CHEN, Qiang Electromechanical laminated composites with piezoelectric phases are increasingly being explored as multifunctional materials providing energy conversion between electric and mechanical energies. The current work explores thus-far undocumented combined microstructural effects of amplitude-to-wavelength ratio, volume fraction, poling direction of piezoelectric phases on both the homogenized properties and localized stress/electric field distributions in multilayered configurations under fully coupled electro-mechanical loading. In particular, the Multiphysics Finite-Volume Direct Averaging Micromechanics (FVDAM) and its counterpart, an in-house micromechanical multiphysics finite-element model, are utilized to investigate the homogenized and localized responses of wavy multilayered piezoelectric BaTiO3/PZT-7A architectures. These two methods generate highly agreeable results. Moreover, we critically examine the convergence of the finite-volume and finite element-based approaches via the Average Stress Theorem and Average Electric Displacement Theorem. The comparison shows the finite volume-based approach possesses a better numerical convergence. This study illustrates the FVDAM’s ability toward the analysis and design of engineered multilayered piezoelectric materials with wavy architecture. http://hdl.handle.net/10985/19895 Deep learning in heterogeneous materials: Targeting the thermo-mechanical response of unidirectional composites CHEN, Qiang; TU, Wenqiong; MA, Meng In this communication, a multi-task deep learning-driven homogenization scheme is proposed for predicting the effective thermomechanical response of unidirectional composites consisting of a random array of inhomogeneity. Toward this end, 40 000 repeating unit cells (RUCs) comprising an arbitrary number of locally irregular inclusions are generated over a wide range of fiber volume fractions. The finite-volume direct averaging micromechanics is then employed to evaluate the homogenized thermo-mechanical moduli of each RUC. Subsequently, a two-dimensional deep convolution neural network (CNN) is constructed as a surrogate model to extract the statistical correlations between the RUC geometrical information and the corresponding homogenized response. The RUC images together with their homogenized moduli are divided into two datasets in a ratio of 9:1 with the former part used for training the CNN model and the latter part used for verification. The results presented in this contribution demonstrate that the deep CNN predictions exhibit remarkable correlations with the theoretical values generated by the finite-volume micromechanics, with a maximum relative prediction error of less than 8%, providing good support for the data-based homogenization approach. Wed, 01 Jan 2020 00:00:00 GMT http://hdl.handle.net/10985/19895 2020-01-01T00:00:00Z CHEN, Qiang TU, Wenqiong MA, Meng In this communication, a multi-task deep learning-driven homogenization scheme is proposed for predicting the effective thermomechanical response of unidirectional composites consisting of a random array of inhomogeneity. Toward this end, 40 000 repeating unit cells (RUCs) comprising an arbitrary number of locally irregular inclusions are generated over a wide range of fiber volume fractions. The finite-volume direct averaging micromechanics is then employed to evaluate the homogenized thermo-mechanical moduli of each RUC. Subsequently, a two-dimensional deep convolution neural network (CNN) is constructed as a surrogate model to extract the statistical correlations between the RUC geometrical information and the corresponding homogenized response. The RUC images together with their homogenized moduli are divided into two datasets in a ratio of 9:1 with the former part used for training the CNN model and the latter part used for verification. The results presented in this contribution demonstrate that the deep CNN predictions exhibit remarkable correlations with the theoretical values generated by the finite-volume micromechanics, with a maximum relative prediction error of less than 8%, providing good support for the data-based homogenization approach. http://hdl.handle.net/10985/19876 Fully-coupled electro-magneto-elastic behavior of unidirectional multiphased composites via finite-volume homogenization CHEN, Qiang; CHEN, Weiqiu; WANG, Guannan The effective and localized electro-magneto-elastic behavior of periodic unidirectional composites is investigated in this work. Instead of adopting the classical micromechanics models or variational principle-based finite-element (FE) techniques, the finite-volume direct averaging micromechanics (FVDAM) is extended to fulfill this task by incorporating fully-coupled electro-magneto-elastic constitutive relations. Consistent with the mathe­matical homogenization theories, the mechanical displacements and electric/magnetic potentials are partitioned using two-scale expansion involving the macroscopic and fluctuating contributions. The generalized local stiffness matrices are constructed explicitly by relating the surface-averaged tractions, electric displacements, and magnetic inductions to the surface-averaged mechanical displacements and electric/magnetic potentials, followed which the continuity and periodic boundary conditions are applied. The homogenized coefficients and localized stress, electric/magnetic field distributions are validated extensively against the exact generalized Eshelby solution and an in-house FE program, as well as the experimental measurements, where perfect agreements are observed for all cases. The efficiency and convergence of the proposed multiphysics FVDAM (MFVDAM) are tested by comparing the execution time and localized fields as a function of mesh discretization, with the multiphysics FE (MFEM) results as references. Besides, the MFVDAM is encapsulated into the particle swarm optimization algorithm to deduce the optimal fiber volume fraction at which maximum magnetoelectric coupling effect may occur in the composite system with piezomagnetic ceramics reinforced with piezoelectric unidirectional fibers. Fri, 01 Jan 2021 00:00:00 GMT http://hdl.handle.net/10985/19876 2021-01-01T00:00:00Z CHEN, Qiang CHEN, Weiqiu WANG, Guannan The effective and localized electro-magneto-elastic behavior of periodic unidirectional composites is investigated in this work. Instead of adopting the classical micromechanics models or variational principle-based finite-element (FE) techniques, the finite-volume direct averaging micromechanics (FVDAM) is extended to fulfill this task by incorporating fully-coupled electro-magneto-elastic constitutive relations. Consistent with the mathe­matical homogenization theories, the mechanical displacements and electric/magnetic potentials are partitioned using two-scale expansion involving the macroscopic and fluctuating contributions. The generalized local stiffness matrices are constructed explicitly by relating the surface-averaged tractions, electric displacements, and magnetic inductions to the surface-averaged mechanical displacements and electric/magnetic potentials, followed which the continuity and periodic boundary conditions are applied. The homogenized coefficients and localized stress, electric/magnetic field distributions are validated extensively against the exact generalized Eshelby solution and an in-house FE program, as well as the experimental measurements, where perfect agreements are observed for all cases. The efficiency and convergence of the proposed multiphysics FVDAM (MFVDAM) are tested by comparing the execution time and localized fields as a function of mesh discretization, with the multiphysics FE (MFEM) results as references. Besides, the MFVDAM is encapsulated into the particle swarm optimization algorithm to deduce the optimal fiber volume fraction at which maximum magnetoelectric coupling effect may occur in the composite system with piezomagnetic ceramics reinforced with piezoelectric unidirectional fibers. http://hdl.handle.net/10985/22397 Homogenization of size-dependent multiphysics behavior of nanostructured piezoelectric composites with energetic surfaces CHEN, Qiang; CHATZIGEORGIOU, George; MERAGHNI, Fodil; JAVILI, Ali Surface piezoelectricity considering the extended Gurtin--Murdoch coherent interface model has been incorporated into the composite cylinder assemblage (CCA), generalized self-consistent method (GSCM), as well as the multiphysics finite-element micromechanics (MFEM), for simulating the size-dependent multiphysics response of nanoporous materials wherein interface stress and electric displacement prevail. In the case of the CCA/GSCM model, the coherent interface model is implemented through the generalized Young--Laplace equations that govern the variation of the surface stress and the surface electric displacement. Three loading modes are utilized to identify the closed-form solutions for a complete set of Hill’s moduli, and piezoelectric and dielectric constants. In the case of the MFEM, surface piezoelectricity is incorporated directly through additional surface energies associated with the elements that stretch along the interface. In order to assess the accuracy of the developed computational approaches, the generalized Kirsch problem under far-field transverse electric displacement loading is developed for recovering electric displacement concentration in the vicinity of the pore boundary. Homogenized properties are generated and critically examined for a broad variety of parameters and dimensions, predicted by the CCA/GSCM and MFEM methods. It is shown that all the predicted effective properties of these two families of homogenization techniques are similar except for the transverse shear moduli where they show marked differences that are reminiscent of what has been observed in the absence of surface electricity. Sat, 01 Jan 2022 00:00:00 GMT http://hdl.handle.net/10985/22397 2022-01-01T00:00:00Z CHEN, Qiang CHATZIGEORGIOU, George MERAGHNI, Fodil JAVILI, Ali Surface piezoelectricity considering the extended Gurtin--Murdoch coherent interface model has been incorporated into the composite cylinder assemblage (CCA), generalized self-consistent method (GSCM), as well as the multiphysics finite-element micromechanics (MFEM), for simulating the size-dependent multiphysics response of nanoporous materials wherein interface stress and electric displacement prevail. In the case of the CCA/GSCM model, the coherent interface model is implemented through the generalized Young--Laplace equations that govern the variation of the surface stress and the surface electric displacement. Three loading modes are utilized to identify the closed-form solutions for a complete set of Hill’s moduli, and piezoelectric and dielectric constants. In the case of the MFEM, surface piezoelectricity is incorporated directly through additional surface energies associated with the elements that stretch along the interface. In order to assess the accuracy of the developed computational approaches, the generalized Kirsch problem under far-field transverse electric displacement loading is developed for recovering electric displacement concentration in the vicinity of the pore boundary. Homogenized properties are generated and critically examined for a broad variety of parameters and dimensions, predicted by the CCA/GSCM and MFEM methods. It is shown that all the predicted effective properties of these two families of homogenization techniques are similar except for the transverse shear moduli where they show marked differences that are reminiscent of what has been observed in the absence of surface electricity. http://hdl.handle.net/10985/25184 Adaptive deep homogenization theory for periodic heterogeneous materials WU, Jiajun; CHEN, Qiang; JIANG, Jindong; CHATZIGEORGIOU, George; MERAGHNI, Fodil We present an adaptive physics-informed deep homogenization neural network (DHN) approach to formulate a full-field micromechanics model for elastic and thermoelastic periodic arrays with different microstructures. The unit cell solution is approximated by fully connected multilayers via minimizing a loss function formulated in terms of the sum of residuals from the stress equilibrium and heat conduction partial differential equations (PDEs), together with interfacial traction-free or adiabatic boundary conditions. In comparison, periodicity boundary conditions are directly satisfied by introducing a network layer with sinusoidal functions. Fully trainable weights are applied on all collocation points, which are simultaneously trained alongside the network weights. Hence, the network automatically assigns higher weights to the collocation points in the vicinity of the interface (particularly challenging regions of the unit cell solution) in the loss function. This compels the neural networks to enhance their performance at these specific points. The accuracy of adaptive DHN is verified against the finite element and the elasticity solution respectively for elliptical and circular cylindrical pores/fibers. The advantage of the adaptive DHN over the original DHN technique is justified by considering locally irregular porous architecture where pore–pore interaction makes training the network particularly slow and hard to optimize. Mon, 01 Jul 2024 00:00:00 GMT http://hdl.handle.net/10985/25184 2024-07-01T00:00:00Z WU, Jiajun CHEN, Qiang JIANG, Jindong CHATZIGEORGIOU, George MERAGHNI, Fodil We present an adaptive physics-informed deep homogenization neural network (DHN) approach to formulate a full-field micromechanics model for elastic and thermoelastic periodic arrays with different microstructures. The unit cell solution is approximated by fully connected multilayers via minimizing a loss function formulated in terms of the sum of residuals from the stress equilibrium and heat conduction partial differential equations (PDEs), together with interfacial traction-free or adiabatic boundary conditions. In comparison, periodicity boundary conditions are directly satisfied by introducing a network layer with sinusoidal functions. Fully trainable weights are applied on all collocation points, which are simultaneously trained alongside the network weights. Hence, the network automatically assigns higher weights to the collocation points in the vicinity of the interface (particularly challenging regions of the unit cell solution) in the loss function. This compels the neural networks to enhance their performance at these specific points. The accuracy of adaptive DHN is verified against the finite element and the elasticity solution respectively for elliptical and circular cylindrical pores/fibers. The advantage of the adaptive DHN over the original DHN technique is justified by considering locally irregular porous architecture where pore–pore interaction makes training the network particularly slow and hard to optimize. http://hdl.handle.net/10985/23925 Cycle jump technique combined with mean-field micromechanics towards predicting the cyclic response of PA66/GF composites under viscoelastic- viscoplastic regime and damage mechanisms CHEN, Qiang; CHATZIGEORGIOU, George; ROBERT, Gilles; MERAGHNI, Fodil This work proposes a probabilistic micromechanics damage framework to predict the uniaxial and cyclic stress-strain response and progressive damage in random glass-reinforced polyamide composites. Motivated by different microscopic degradation modes observed experimentally, the damage mechanism in the vicinity of the fibers is characterized by the onset and the coalescence of voids, whose evolution can be formulated through a Weibull probabilistic density function. In contrast, the ductile progressive degradation of matrix initial stiffness is analyzed via the continuum damage theory. Towards this end, a 2N+1-phase Mori-Tanaka (MT) method combined with the transformation field analysis approach (TFA) is established within a unified framework. Moreover, the rate-dependent viscoelastic and viscoplastic response of the polymer matrix phase is formulated through a phenomenological model consisting of four Kelvin-Voigt branches and a viscoplastic branch under the thermodynamics framework. Comparison of numerical predictions with experimental data demonstrates the model’s capabilities. In a second step of this work, the micromechanics scheme is combined with the cycle-jump technique in order to simulate moderate and high cycle fatigue tests. This modeling strategy is validated through comparison with experimental results. Sat, 01 Jul 2023 00:00:00 GMT http://hdl.handle.net/10985/23925 2023-07-01T00:00:00Z CHEN, Qiang CHATZIGEORGIOU, George ROBERT, Gilles MERAGHNI, Fodil This work proposes a probabilistic micromechanics damage framework to predict the uniaxial and cyclic stress-strain response and progressive damage in random glass-reinforced polyamide composites. Motivated by different microscopic degradation modes observed experimentally, the damage mechanism in the vicinity of the fibers is characterized by the onset and the coalescence of voids, whose evolution can be formulated through a Weibull probabilistic density function. In contrast, the ductile progressive degradation of matrix initial stiffness is analyzed via the continuum damage theory. Towards this end, a 2N+1-phase Mori-Tanaka (MT) method combined with the transformation field analysis approach (TFA) is established within a unified framework. Moreover, the rate-dependent viscoelastic and viscoplastic response of the polymer matrix phase is formulated through a phenomenological model consisting of four Kelvin-Voigt branches and a viscoplastic branch under the thermodynamics framework. Comparison of numerical predictions with experimental data demonstrates the model’s capabilities. In a second step of this work, the micromechanics scheme is combined with the cycle-jump technique in order to simulate moderate and high cycle fatigue tests. This modeling strategy is validated through comparison with experimental results. http://hdl.handle.net/10985/23502 Physically informed deep homogenization neural network for unidirectional multiphase/multi-inclusion thermoconductive composites JIANG, Jindong; WU, Jiajun; CHEN, Qiang; CHATZIGEORGIOU, George; MERAGHNI, Fodil Elements of the periodic homogenization framework and deep neural network were seamlessly connected for the first time to construct a new micromechanics theory for thermoconductive composites called physically informed Deep Homogenization Network (DHN). This method utilizes a two-scale expansion of the temperature field of spatially uniform composites in terms of macroscopic and fluctuating contributions. The latter is estimated using deep neural network layers. The DHN is trained on a set of collocation points to obtain the fluctuating temperature field over the unit cell domain by minimizing a cost function given in terms of residuals of strong form steady-state heat conduction governing differential equations. Novel use of a periodic layer with several independent periodic functions with adjustable training parameters ensures that periodic boundary conditions of temperature and temperature gradients at the unit cell edges are exactly satisfied. Automatic differentiation is utilized to correctly compute the fluctuating temperature gradients. Homogenized properties and local temperature and gradient distributions of unit cells reinforced by unidirectional fiber or weakened by a hole are compared with finite-element reference results, demonstrating remarkable correlation but without discontinuities associated with temperature gradient distributions in the finite-element simulations. We also illustrate that the DHN enhanced with transfer learning provides a substantially more efficient and accurate simulation of multiple random fiber distributions relative to training the network from scratch. Mon, 01 May 2023 00:00:00 GMT http://hdl.handle.net/10985/23502 2023-05-01T00:00:00Z JIANG, Jindong WU, Jiajun CHEN, Qiang CHATZIGEORGIOU, George MERAGHNI, Fodil Elements of the periodic homogenization framework and deep neural network were seamlessly connected for the first time to construct a new micromechanics theory for thermoconductive composites called physically informed Deep Homogenization Network (DHN). This method utilizes a two-scale expansion of the temperature field of spatially uniform composites in terms of macroscopic and fluctuating contributions. The latter is estimated using deep neural network layers. The DHN is trained on a set of collocation points to obtain the fluctuating temperature field over the unit cell domain by minimizing a cost function given in terms of residuals of strong form steady-state heat conduction governing differential equations. Novel use of a periodic layer with several independent periodic functions with adjustable training parameters ensures that periodic boundary conditions of temperature and temperature gradients at the unit cell edges are exactly satisfied. Automatic differentiation is utilized to correctly compute the fluctuating temperature gradients. Homogenized properties and local temperature and gradient distributions of unit cells reinforced by unidirectional fiber or weakened by a hole are compared with finite-element reference results, demonstrating remarkable correlation but without discontinuities associated with temperature gradient distributions in the finite-element simulations. We also illustrate that the DHN enhanced with transfer learning provides a substantially more efficient and accurate simulation of multiple random fiber distributions relative to training the network from scratch.