Embargo Until 01/May/2026 - Author postprint (1.83 MB)
NOTICE: this is the author’s version of a work that was accepted for publication in Computer Methods in Applied Mechanics and Engineering. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Computer Methods in Applied Mechanics and Engineering 449 Part A (2026), DOI: 10.1016/j.cma.2025.118554
[en] Data-driven approaches make the development of surrogates of complex heterogeneous material responses possible. After being trained using a previously generated data-set during an offline stage, the surrogates can be used as a material law to conduct structural simulations during the online stage. Nevertheless, in view of accounting for the uncertainty and variability of the heterogeneous materials, the surrogates should be able to account for the micro-structure variability, which remains a challenge. Among the possible surrogate candidates, (Interaction-Based-)Deep-Material Networks ((IB-)DMN) offer the advantage that they can extrapolate the response for new material model parameters of the heterogeneous material phases and for arbitrary loading histories outside of their training range. This advantage results from their thermodynamics consistency and from the fact that the IB-DMN learnable parameters represent solely the micro-structure organization and not the phases material response. However, a trained IB-DMN remains an image of a given micro-structure spatial organization realization in terms of clustering etc. A new microstructure realization thus requires a new training process, limiting the interest of the IB-DMN for stochastic multi-scale analyses. In order to address this limitation, we define the learnable or topological arameters of the IB-DMN from a combination of convolution encoder and neural network, with the micro-structure image serving as input data. After training a CNN encoder-decoder,
the encoder part allows extracting the feature vectors of the heterogeneous material directly from micro-structure images. These feature vectors then serve as input of a trained feedforward neural network (FNN) that predicts the topological parameters of the IB-DMN, yielding a “Image to IB-DMN” framework. The methodology is first illustrated in the context of Unidirectional (UD) composites, for which Stochastic Volumes Elements (SVEs) serve as images of the micro-structure realizations. In a second step we show that the machine learning tools can be trained by considering simultaneously composite families of different inclusion shapes such as circular, elliptical and squared. Despite training considering only elastic data, the predictions for a complex pressure-sensitive elasto-plastic model remain accurate. These results demonstrate the complementary roles of the two networks: the CNN encoder–decoder efficiently extracts reduced feature vectors from micro-structure images with diverse inclusion geometries, and the FNN accurately maps these features to the topological parameters of the IB-DMN, establishing a robust, end-to-end image-to-model framework capable of generalizing across different micro-structural configurations.
Research Center/Unit :
A&M - Aérospatiale et Mécanique - ULiège
Disciplines :
Mechanical engineering
Author, co-author :
Wu, Ling ; Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational & Multiscale Mechanics of Materials (CM3)
Noels, Ludovic ; Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational & Multiscale Mechanics of Materials (CM3)
Language :
English
Title :
Convolutional neural network-based mapping of material micro-structures to deep material networks for non-linear mechanical response prediction
Publication date :
01 February 2026
Journal title :
Computer Methods in Applied Mechanics and Engineering
H2020 - 958174 - M-ERA.NET3 - ERA-NET for research and innovation on materials and battery technologies, supporting the European Green Deal.
Name of the research project :
This research has been funded by the Walloon Region under the agreement no. 2010092-CARBOBRAKE in the context of the M-ERA.Net Join Call 2020 funded by the European Union under the Grant Agreement no. 958174.
Funders :
SPW - Public Service of Wallonia EC - European Commission European Union
Funding number :
2010092; 958174
Funding text :
This research has been funded by the Walloon Region under the agreement no. 2010092-CARBOBRAKE in the context of the M-ERA.Net Join Call 2020 funded by the European Union under the Grant Agreement no. 958174.
Graham-Brady, L.L., Arwade, S.R., Corr, D.J., Gutiérrez, M.A., Breysse, D., Grigoriu, M., Zabaras, N., Probability and materials: from nano- to macro-Scale: a summary. Probab. Eng. Mech. 21:3 (2006), 193–199, 10.1016/j.probengmech.2005.10.005 Probability and Materials: from Nano- to Macro-Scale.
Ostoja-Starzewski, M., Kale, S., Karimi, P., Malyarenko, A., Raghavan, B., Ranganathan, S.I., Zhang, J., Chapter two - Scaling to RVE in random media. Advances in Applied Mechanics, 49, 2016, Elsevier, 111–211, 10.1016/bs.aams.2016.07.001.
Noels, L., Chapter one - Toward stochastic multiscale methods in continuum solid mechanics. Advances in Applied Mechanics, 55, 2022, Elsevier, 1–254, 10.1016/bs.aams.2022.03.001.
Mariani, S., Martini, R., Ghisi, A., Corigliano, A., Beghi, M., Overall elastic properties of polysilicon films: a statistical investigation of the effects of polycrystal morphology. Int. J. Multiscale Comput. Eng. 9:3 (2011), 327–346.
Geißendörfer, M., Liebscher, A., Proppe, C., Redenbach, C., Schwarzer, D., Stochastic multiscale modeling of metal foams. Probab. Eng. Mech. 37 (2014), 132–137, 10.1016/j.probengmech.2014.06.006.
Savvas, D., Stefanou, G., Papadrakakis, M., Determination of RVE size for random composites with local volume fraction variation. Comput. Methods Appl. Mech. Eng. 305 (2016), 340–358, 10.1016/j.cma.2016.03.002.
Stefanou, G., Savvas, D., Papadrakakis, M., Stochastic finite element analysis of composite structures based on mesoscale random fields of material properties. Comput. Methods Appl. Mech. Eng. 326 (2017), 319–337, 10.1016/j.cma.2017.08.002.
Wu, L., Chung, C.N., Major, Z., Adam, L., Noels, L., From SEM images to elastic responses: a stochastic multiscale analysis of UD fiber reinforced composites. Compos. Struct. 189 (2018), 206–227, 10.1016/j.compstruct.2018.01.051.
Wu, L., Lucas, V., Nguyen, V.-D., Golinval, J.-C., Paquay, S., Noels, L., A stochastic multi-scale approach for the modeling of thermo-elastic damping in micro-resonators. Comput. Methods Appl. Mech. Eng. 310 (2016), 802–839, 10.1016/j.cma.2016.07.042.
Lucas, V., Golinval, J.-C., Voicu, R.C., Danila, M., Gavrila, R., Müller, R., Dinescu, A., Noels, L., Wu, L., Propagation of material and surface profile uncertainties on MEMS micro-resonators using a stochastic second-order computational multi-scale approach. Int. J. Numer. Methods Eng. 111:1 (2017), 26–68, 10.1002/nme.5452.
Hun, D.-A., Guilleminot, J., Yvonnet, J., Bornert, M., Stochastic multiscale modeling of crack propagation in random heterogeneous media. Int. J. Numer. Methods Eng. 119:13 (2019), 1325–1344, 10.1002/nme.6093.
Doghri, I., Ouaar, A., Homogenization of two-phase elasto-plastic composite materials and structures: study of tangent operators, cyclic plasticity and numerical algorithms. Int. J. Solids Struct. 40:7 (2003), 1681–1712.
Molinari, A., El Houdaigui, F., Toth, L.S., Validation of the tangent formulation for the solution of the non-linear eshelby inclusion problem. Int. J. Plast. 20:2 (2004), 291–307.
Wu, L., Noels, L., Adam, L., Doghri, I., A combined incremental-secant mean-field homogenization scheme with per-phase residual strains for elasto-plastic composites. Int. J. Plast. 51 (2013), 80–102, 10.1016/j.ijplas.2013.06.006.
Wu, L., Nguyen, V.-D., Adam, L., Noels, L., An inverse micro-mechanical analysis toward the stochastic homogenization of nonlinear random composites. Comput. Methods Appl. Mech. Eng. 348 (2019), 97–138, 10.1016/j.cma.2019.01.016.
Calleja Vázquez, J.M., Wu, L., Nguyen, V.-D., Noels, L., A micromechanical mean-field homogenization surrogate for the stochastic multiscale analysis of composite materials failure. Int. J. Numer. Methods Eng. 124:23 (2023), 5200–5262, 10.1002/nme.7344.
Geers, M.G.D., Kouznetsova, V.G., Brekelmans, W.A.M., Multi-scale computational homogenization: trends and challenges. J. Comput. Appl. Math. 234:7 (2010), 2175–2182, 10.1016/j.cam.2009.08.077 Fourth International Conference on Advanced COmputational Methods in ENgineering (ACOMEN 2008).
Lefik, M., Schrefler, B.A., Artificial neural network as an incremental non-linear constitutive model for a finite element code. Comput. Methods Appl. Mech. Eng. 192:28 (2003), 3265–3283, 10.1016/S0045-7825(03)00350-5 Multiscale Computational Mechanics for Materials and Structures.
Stoffel, M., Bamer, F., Markert, B., Neural network based constitutive modeling of nonlinear viscoplastic structural response. Mech. Res. Commun. 95 (2019), 85–88, 10.1016/j.mechrescom.2019.01.004.
Ghavamian, F., Simone, A., Accelerating multiscale finite element simulations of history-dependent materials using a recurrent neural network. Comput. Methods Appl. Mech. Eng., 357, 2019, 112594, 10.1016/j.cma.2019.112594.
Wu, L., Nguyen, V.D., Kilingar, N.G., Noels, L., A recurrent neural network-accelerated multi-scale model for elasto-plastic heterogeneous materials subjected to random cyclic and non-proportional loading paths. Comput. Methods Appl. Mech. Eng., 369, 2020, 113234, 10.1016/j.cma.2020.113234.
Bonatti, C., Mohr, D., On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids. J. Mech. Phys. Solids, 158, 2022, 104697, 10.1016/j.jmps.2021.104697.
Wu, L., Noels, L., Self-consistency reinforced minimal gated recurrent unit for surrogate modeling of history-dependent non-linear problems: application to history-dependent homogenized response of heterogeneous materials. Comput. Methods Appl. Mech. Eng., 424, 2024, 116881, 10.1016/j.cma.2024.116881.
Heidenreich, J.N., Mohr, D., Extended minimal state cells (EMSC): self-consistent recurrent neural networks for rate- and temperature dependent plasticity. Int. J. Plast., 188, 2025, 104305, 10.1016/j.ijplas.2025.104305.
Zhongbo, Y., Hien, P.L., Pre-trained transformer model as a surrogate in multiscale computational homogenization framework for elastoplastic composite materials subjected to generic loading paths. Comput. Methods Appl. Mech. Eng., 421, 2024, 116745, 10.1016/j.cma.2024.116745.
Kibrete, F., Trzepieciński, T., Gebremedhen, H.S., Woldemichael, D.E., Artificial intelligence in predicting mechanical properties of composite materials. J. Compos. Sci., 7(9), 2023 https://www.mdpi.com/2504-477X/7/9/364.
Dornheim, J., Morand, L., Nallani, H.J., Helm, D., Neural networks for constitutive modeling: from universal function approximators to advanced models and the integration of physics. Arch. Comput. Methods Eng. 31:2 (2024), 1097–1127.
Masi, F., Stefanou, I., Multiscale modeling of inelastic materials with thermodynamics-based artificial neural networks (TANN). Comput. Methods Appl. Mech. Eng., 398, 2022, 115190, 10.1016/j.cma.2022.115190.
Maia, M.A., Rocha, I.B.C.M., Kerfriden, P., van der Meer, F.P., Physically recurrent neural networks for path-dependent heterogeneous materials: embedding constitutive models in a data-driven surrogate. Comput. Methods Appl. Mech. Eng., 407, 2023, 115934, 10.1016/j.cma.2023.115934.
Pyrialakos, S., Kalogeris, I., Papadopoulos, V., Multiscale analysis of nonlinear systems using a hierarchy of deep neural networks. Int. J. Solids Struct., 271–272, 2023, 112261, 10.1016/j.ijsolstr.2023.112261.
Kalogeris, I., Pyrialakos, S., Kokkinos, O., Papadopoulos, V., Stochastic optimization of carbon nanotube reinforced concrete for enhanced structural performance. Eng. Comput. 39:4 (2023), 2927–2943, 10.1007/s00366-022-01693-8.
Lu, X., Yvonnet, J., Papadopoulos, L., Kalogeris, I., Papadopoulos, V., A stochastic FE2 data-Driven method for nonlinear multiscale modeling. Materials, 14(11), 2021, 10.3390/ma14112875.
Rao, C., Liu, Y., Three-dimensional convolutional neural network (3D-CNN) for heterogeneous material homogenization. Comput. Mater. Sci., 184, 2020, 109850, 10.1016/j.commatsci.2020.109850.
Aldakheel, F., Elsayed, E.S., Zohdi, T.I., Wriggers, P., Efficient multiscale modeling of heterogeneous materials using deep neural networks. Comput. Mech. 72:1 (2023), 155–171, 10.1007/s00466-023-02324-9.
Eidel, B., Deep CNNs as universal predictors of elasticity tensors in homogenization. Comput. Methods Appl. Mech. Eng., 403, 2023, 115741, 10.1016/j.cma.2022.115741.
Gavallas, P., Stefanou, G., Savvas, D., Mattrand, C., Bourinet, J.-M., CNN-Based prediction of microstructure-derived random property fields of composite materials. Comput. Methods Appl. Mech. Eng., 430, 2024, 117207, 10.1016/j.cma.2024.117207.
Storm, J., Rocha, I.B.C.M., van der Meer, F.P., A microstructure-based graph neural network for accelerating multiscale simulations. Comput. Methods Appl. Mech. Eng., 427, 2024, 117001, 10.1016/j.cma.2024.117001.
Liu, Z., Wu, C.T., Exploring the 3D architectures of deep material network in data-driven multiscale mechanics. J. Mech. Phys. Solids 127 (2019), 20–46, 10.1016/j.jmps.2019.03.004.
Gajek, S., Schneider, M., Böhlke, T., On the micromechanics of deep material networks. J. Mech. Phys. Solids, 142, 2020, 103984, 10.1016/j.jmps.2020.103984.
Wu, L., Adam, L., Noels, L., Micro-mechanics and data-driven based reduced order models for multi-scale analyses of woven composites. Compos. Struct., 270, 2021, 114058, 10.1016/j.compstruct.2021.114058.
Nguyen, V.D., Noels, L., Interaction-based material network: a general framework for (porous) microstructured materials. Comput. Methods Appl. Mech. Eng., 389, 2022, 114300, 10.1016/j.cma.2021.114300.
Gajek, S., Schneider, M., Böhlke, T., An FE–DMN method for the multiscale analysis of short fiber reinforced plastic components. Comput. Methods Appl. Mech. Eng., 384, 2021, 113952, 10.1016/j.cma.2021.113952.
Nguyen, V.D., Noels, L., Micromechanics-based material networks revisited from the interaction viewpoint; robust and efficient implementation for multi-phase composites. Eur. J. Mech. A. Solids, 91, 2022, 104384, 10.1016/j.euromechsol.2021.104384.
Wan, W.-N., Wei, T.-J., Su, T.-H., Chen, C.-S., Decoding material networks: exploring performance of deep material network and interaction-based material networks. J. Mech. 40 (2024), 796–807, 10.1093/jom/ufae053.
Huang, T., Liu, Z., Wu, C.T., Chen, W., Microstructure-guided deep material network for rapid nonlinear material modeling and uncertainty quantification. Comput. Methods Appl. Mech. Eng., 398, 2022, 115197, 10.1016/j.cma.2022.115197.
Meyer, N., Gajek, S., Görthofer, J., Hrymak, A., Kärger, L., Henning, F., Schneider, M., Böhlke, T., A probabilistic virtual process chain to quantify process-induced uncertainties in sheet molding compounds. Compos. Part B: Engineering, 249, 2023, 110380, 10.1016/j.compositesb.2022.110380.
Li, T., Micromechanics-informed parametric deep material network for physics behavior prediction of heterogeneous materials with a varying morphology. Comput. Methods Appl. Mech. Eng., 419, 2024, 116687, 10.1016/j.cma.2023.116687.
Jean, J.G., Su, T.-H., Huang, S.-J., Wu, C.-T., Chen, C.-S., Graph-enhanced deep material network: multiscale materials modeling with microstructural informatics. Comput. Mech., 2024, 10.1007/s00466-024-02493-1.
Wu, L., Noels, L., Stochastic deep material networks as efficient surrogates for stochastic homogenisation of non-linear heterogeneous materials. Comput. Methods Appl. Mech. Eng., 441, 2025, 117994, 10.1016/j.cma.2025.117994.
Nguyen, V.-D., Wu, L., Noels, L., Unified treatment of microscopic boundary conditions and efficient algorithms for estimating tangent operators of the homogenized behavior in the computational homogenization method. Comput. Mech. 59:3 (2017), 483–505, 10.1007/s00466-016-1358-z.
Nguyen, V.D., Lani, F., Pardoen, T., Morelle, X.P., Noels, L., A large strain hyperelastic viscoelastic viscoplastic damage constitutive model based on a multimechanism nonlocal damage continuum for amorphous glassy polymers. Int. J. Solids Struct. 96 (2016), 192–216, 10.1016/j.ijsolstr.2016.06.008.
Bonet, J., Burton, A.J., A simple orthotropic, transversely isotropic hyperelastic constitutive equation for large strain computations. Comput. Methods Appl. Mech. Eng. 162:1–4 (1998), 151–164, 10.1016/S0045-7825(97)00339-3.
Wu, L., Tjahjanto, D., Becker, G., Makradi, A., Jérusalem, A., Noels, L., A micro–meso-model of intra-laminar fracture in fiber-reinforced composites based on a discontinuous galerkin/cohesive zone method. Eng. Fract. Mech. 104 (2013), 162–183, 10.1016/j.engfracmech.2013.03.018.
L. Wu, L. Noels, Data of “Convolutional Neural Network-Based Mapping of Material Micro-Structures to Deep Material Networks for Non-Linear Mechanical Response Prediction”, 2025, 10.5281/zenodo.17182395.
