pith. sign in

arxiv: 2512.06779 · v2 · submitted 2025-12-07 · 💻 cs.CE

A Texture-Generalizable Deep Material Network via Orientation-Aware Interaction Learning for Polycrystal Modeling and Texture Evolution

Ting-Ju Wei , Tung-Huan Su , Chuin-Shan Chen This is my paper

Pith reviewed 2026-05-17 01:18 UTC · model grok-4.3

classification 💻 cs.CE
keywords deep material networkgraph neural networkpolycrystal modelingtexture evolutionhomogenizationsurrogate modelmultiscale simulationcrystal plasticity
0
0 comments X

The pith

A graph neural network allows deep material networks to work on any new crystal texture without retraining.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops a way to make the Orientation-aware Interaction-based Deep Material Network applicable to crystal textures it was not trained on. They introduce a Texture-Adaptive Clustering and Sampling method to represent any texture and train a Graph Neural Network to predict the needed interaction parameters for stress equilibrium. If this works, simulations of polycrystal deformation and texture evolution can proceed without retraining for each new microstructure. The results show close agreement with direct numerical simulations for nonlinear responses across various texture distributions. This opens the door to more flexible use of such models in multiscale materials engineering.

Core claim

The TACS--GNN--ODMN framework reformulates ODMN generalization as a microstructure-to-parameter inference problem. Combining a Texture-Adaptive Clustering and Sampling (TACS) scheme for texture representation with a Graph Neural Network (GNN) for inferring micromechanical equilibrium parameters enables fully parameterized ODMNs for previously unseen microstructures without retraining. Numerical results show accurate predictions of nonlinear mechanical responses and texture evolution that agree with direct numerical simulations.

What carries the argument

The combination of the Texture-Adaptive Clustering and Sampling (TACS) scheme and a Graph Neural Network that infers the micromechanical equilibrium parameters of the ODMN from a given texture representation.

Load-bearing premise

The GNN trained on a finite set of textures can accurately infer equilibrium parameters for any arbitrary unseen texture distribution while preserving the physical consistency of the original ODMN.

What would settle it

Perform a direct numerical simulation and the framework prediction for a specific texture distribution held out from the GNN training data, and check if the stress-strain curves and texture evolution match closely; disagreement would disprove the generalization claim.

Figures

Figures reproduced from arXiv: 2512.06779 by Chuin-Shan Chen, Ting-Ju Wei, Tung-Huan Su.

Figure 1
Figure 1. Figure 1: Workflow of the TACS–GNN–ODMN framework. In the offline training stage, polycrystalline RVEs are [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic architecture of the ODMN. A binary-tree material network defines the hierarchical stress [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Architecture of the GNN module for predicting ODMN micromechanical equilibrium parameters. A grain [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Training and validation error curves of the TACS–GNN–ODMN framework with [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Unseen RVEs used for testing: (a) S1, (b) S2, (c) W1, (d) W2. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Pole figures of reconstructed ODFs compared with DNS results for the four testing RVEs: (a) S1, (b) S2, (c) [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Stress–strain curves under cyclic loading for (a) S1, (b) S2, (c) W1, and (d) W2 at a strain rate of [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Pole figures after unaxial cycylic loading with an applied deformation of [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Stress–strain curves under simple shear loading for (a) S1, (b) S2, (c) W1, and (d) W2 at a strain rate of [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Pole figures after simple shear loading with an applied deformation of [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Training and validation errors of the TACS-GNN-ODMN framework for varying values of [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Reconstructed analogous unit cells for the unseen RVEs: (a) S1, (b) S2, (c) W1, and (d) W2. [PITH_FULL_IMAGE:figures/full_fig_p017_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Stress–strain curves under cyclic loading for (a) S1, (b) S2, (c) W1, and (d) W2 at a strain rate of [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Comparison of local stress distributions predicted by the inferred ODMN surrogate, the original RVE DNS, [PITH_FULL_IMAGE:figures/full_fig_p018_14.png] view at source ↗
read the original abstract

Machine learning surrogate models have emerged as a promising approach for accelerating multiscale materials simulations while preserving predictive fidelity. Among them, the Orientation-aware Interaction-based Deep Material Network (ODMN) provides a hierarchical homogenization framework in which material nodes encode crystallographic texture and interaction nodes enforce stress equilibrium under the Hill--Mandel condition. Trained solely on linear-elastic stiffness data, ODMN captures intrinsic microstructure--mechanics relationships, enabling accurate prediction of nonlinear mechanical responses and texture evolution. However, its applicability remains fundamentally limited by the absence of a parametric mapping from arbitrary microstructures to the ODMN parameter space. This limitation necessitates retraining for each new microstructure. To address this challenge, we reformulate ODMN generalization as a microstructure-to-parameter inference problem and propose the TACS--GNN--ODMN framework. The proposed framework combines a Texture-Adaptive Clustering and Sampling (TACS) scheme for texture representation with a Graph Neural Network (GNN) for inferring micromechanical equilibrium parameters. This strategy enables the construction of fully parameterized ODMNs for previously unseen microstructures without retraining. Numerical results demonstrate that the proposed framework accurately predicts nonlinear mechanical responses and texture evolution across diverse texture distributions. The predicted responses show close agreement with direct numerical simulations (DNS), highlighting the framework as a generalizable and physically interpretable surrogate model for microstructure-informed multiscale materials simulations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes the TACS-GNN-ODMN framework to overcome the retraining requirement of the Orientation-aware Interaction-based Deep Material Network (ODMN). It combines Texture-Adaptive Clustering and Sampling (TACS) for microstructure representation with a Graph Neural Network (GNN) that infers micromechanical equilibrium parameters, enabling fully parameterized ODMNs for unseen textures. The central claim is that this yields accurate predictions of nonlinear mechanical responses and texture evolution that agree closely with direct numerical simulations (DNS), while preserving physical interpretability from the original ODMN trained only on linear-elastic data.

Significance. If the numerical evidence holds under rigorous validation, the work would represent a meaningful step toward generalizable, microstructure-aware surrogate models in polycrystal mechanics. It directly addresses the per-microstructure retraining bottleneck in deep material networks and could accelerate multiscale simulations while retaining the Hill-Mandel equilibrium enforcement that makes ODMN physically interpretable.

major comments (2)
  1. [Abstract / Numerical Results] Abstract and Numerical Results section: the repeated claim of 'close agreement with DNS' and 'accurate prediction' across diverse textures is unsupported by any quantitative error metrics (e.g., relative L2 errors on stress-strain curves, pole-figure intensity differences, or texture evolution RMSE). No training/validation split ratios, cross-validation procedure, or propagation analysis of GNN inference errors into the nonlinear regime are provided, leaving the central generalization claim without load-bearing numerical substantiation.
  2. [Methodology] Methodology (TACS-GNN-ODMN pipeline): the GNN is trained solely on equilibrium parameters extracted by TACS from a finite texture collection. No architectural constraint, auxiliary loss, or post-inference projection is described that would enforce satisfaction of the stress-equilibrium and Hill-Mandel conditions at interaction nodes for textures lying outside the training distribution. This leaves open the possibility that inferred parameters produce non-zero residuals, violating the physical consistency that the original ODMN relies upon.
minor comments (2)
  1. [Introduction] Clarify in the introduction or methods whether the GNN training data include only linear-elastic stiffness tensors or also sample nonlinear constitutive responses; the current description leaves this ambiguous.
  2. [Figures / Results] Figure captions and results tables should report explicit quantitative comparison metrics (error norms, R² values) rather than relying solely on visual overlay of curves.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough and constructive review. The comments highlight important aspects of quantitative validation and physical consistency that we will address to strengthen the manuscript. Below we respond point-by-point to the major comments.

read point-by-point responses

  1. Referee: [Abstract / Numerical Results] Abstract and Numerical Results section: the repeated claim of 'close agreement with DNS' and 'accurate prediction' across diverse textures is unsupported by any quantitative error metrics (e.g., relative L2 errors on stress-strain curves, pole-figure intensity differences, or texture evolution RMSE). No training/validation split ratios, cross-validation procedure, or propagation analysis of GNN inference errors into the nonlinear regime are provided, leaving the central generalization claim without load-bearing numerical substantiation.

    Authors: We agree that the current presentation relies primarily on visual comparisons in the figures without accompanying quantitative error metrics. In the revised manuscript we will add explicit quantitative measures, including mean relative L2 errors on stress-strain responses, average pole-figure intensity differences, and texture evolution RMSE across the test textures. We will also report the training/validation split (80/20 with 5-fold cross-validation) and include a short analysis of how GNN parameter inference errors propagate into the nonlinear regime. These additions will be placed in a new subsection of the Numerical Results section. revision: yes

  2. Referee: [Methodology] Methodology (TACS-GNN-ODMN pipeline): the GNN is trained solely on equilibrium parameters extracted by TACS from a finite texture collection. No architectural constraint, auxiliary loss, or post-inference projection is described that would enforce satisfaction of the stress-equilibrium and Hill-Mandel conditions at interaction nodes for textures lying outside the training distribution. This leaves open the possibility that inferred parameters produce non-zero residuals, violating the physical consistency that the original ODMN relies upon.

    Authors: The GNN learns a mapping from TACS-derived texture descriptors to the equilibrium parameters that were originally obtained by enforcing Hill-Mandel conditions on the training textures. For unseen textures the inference is therefore approximate. We acknowledge that the manuscript does not currently describe an explicit enforcement mechanism (auxiliary loss or projection) for out-of-distribution cases. In revision we will add a verification subsection that computes the residual of the stress-equilibrium and Hill-Mandel conditions on the inferred parameters for the held-out textures and report the magnitude of any violations. If residuals are non-negligible we will introduce a lightweight post-inference projection step onto the admissible parameter manifold; otherwise we will clarify that the physical consistency is inherited from the original ODMN training and preserved to within the observed residual tolerance. revision: partial

Circularity Check

0 steps flagged

Minor self-citation on ODMN base; GNN generalization validated against independent DNS benchmarks

full rationale

The paper builds ODMN on prior interaction-based homogenization (likely self-cited) but introduces new TACS-GNN mapping from texture to equilibrium parameters. The central claim of accurate nonlinear prediction and texture evolution for unseen microstructures is supported by direct numerical simulation (DNS) comparisons on diverse distributions, which serve as external falsifiable benchmarks. No load-bearing step reduces by construction to fitted inputs or self-citation chains; the GNN acts as a learned surrogate tested for out-of-distribution performance. This qualifies as standard surrogate modeling with independent validation rather than circular derivation.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that a learned mapping from texture descriptors to ODMN interaction parameters preserves the Hill-Mandel condition and linear-elastic training fidelity for nonlinear regimes; no new physical axioms are introduced beyond those already in the base ODMN.

free parameters (2)
  • GNN architecture hyperparameters
    Number of layers, hidden dimensions, and aggregation functions chosen during training of the parameter-inference network.
  • TACS clustering parameters
    Number of clusters and sampling density used to represent arbitrary textures as input to the GNN.
axioms (1)
  • domain assumption Hill-Mandel condition holds for the homogenized response
    Invoked to justify that interaction nodes enforce stress equilibrium; inherited from the base ODMN model.

pith-pipeline@v0.9.0 · 5554 in / 1373 out tokens · 42602 ms · 2026-05-17T01:18:15.864664+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages · 1 internal anchor

  1. [2]

    Jain, Peidong Wu, Yichuan Shao, Dayong Li, and Yinghong Peng

    Guowei Zhou, Mukesh K. Jain, Peidong Wu, Yichuan Shao, Dayong Li, and Yinghong Peng. Experiment and crystal plasticity analysis on plastic deformation of az31b mg alloy sheet under intermediate temperatures: How deformation mechanisms evolve.International Journal of Plasticity, 79:19–47, 2016

    work page 2016

  2. [3]

    Bate and Y .G

    P.S. Bate and Y .G. An. Plastic anisotropy in aa5005 al–1mg: predictions using crystal plasticity finite element analysis.Scripta Materialia, 51(10):973–977, 2004

    work page 2004

  3. [4]

    Delannay, M.A

    L. Delannay, M.A. Melchior, J.W. Signorelli, J.-F. Remacle, and T. Kuwabara. Influence of grain shape on the planar anisotropy of rolled steel sheets – evaluation of three models.Computational Materials Science, 45(3):739–743, 2009. Proceedings of the 17th International Workshop on Computational Mechanics of Materials

    work page 2009

  4. [5]

    Lebensohn and C.N

    R.A. Lebensohn and C.N. Tomé. A self-consistent anisotropic approach for the simulation of plastic deformation and texture development of polycrystals: Application to zirconium alloys.Acta Metallurgica et Materialia, 41(9):2611–2624, 1993

    work page 1993

  5. [6]

    Turner and C.N

    P.A. Turner and C.N. Tomé. A study of residual stresses in zircaloy-2 with rod texture.Acta Metallurgica et Materialia, 42(12):4143–4153, 1994

    work page 1994

  6. [7]

    Wang, P.D

    H. Wang, P.D. Wu, C.N. Tomé, and Y . Huang. A finite strain elastic–viscoplastic self-consistent model for polycrystalline materials.Journal of the Mechanics and Physics of Solids, 58(4):594–612, 2010

    work page 2010

  7. [8]

    A novel grain cluster-based homogenization scheme.Modelling and Simulation in Materials Science and Engineering, 18(1):015006, dec 2009

    D D Tjahjanto, P Eisenlohr, and F Roters. A novel grain cluster-based homogenization scheme.Modelling and Simulation in Materials Science and Engineering, 18(1):015006, dec 2009

    work page 2009

  8. [9]

    Multiscale deep drawing analysis of dual-phase steels using grain cluster-based rgc scheme.Modelling and Simulation in Materials Science and Engineering, 23(4):045005, apr 2015

    D D Tjahjanto, P Eisenlohr, and F Roters. Multiscale deep drawing analysis of dual-phase steels using grain cluster-based rgc scheme.Modelling and Simulation in Materials Science and Engineering, 23(4):045005, apr 2015

    work page 2015

  9. [10]

    Roters, M

    F. Roters, M. Diehl, P. Shanthraj, P. Eisenlohr, C. Reuber, S.L. Wong, T. Maiti, A. Ebrahimi, T. Hochrainer, H.-O. Fabritius, S. Nikolov, M. Friák, N. Fujita, N. Grilli, K.G.F. Janssens, N. Jia, P.J.J. Kok, D. Ma, F. Meier, E. Werner, M. Stricker, D. Weygand, and D. Raabe. Damask – the düsseldorf advanced material simulation kit for modeling multi-physics...

    work page 2019

  10. [11]

    Temizer and P

    I. Temizer and P. Wriggers. An adaptive multiscale resolution strategy for the finite deformation analysis of microheterogeneous structures.Computer Methods in Applied Mechanics and Engineering, 200(37):2639–2661,

    work page

  11. [12]

    Special Issue on Modeling Error Estimation and Adaptive Modeling

    work page

  12. [13]

    Eisenlohr, M

    P. Eisenlohr, M. Diehl, R.A. Lebensohn, and F. Roters. A spectral method solution to crystal elasto-viscoplasticity at finite strains.International Journal of Plasticity, 46:37–53, 2013. Microstructure-based Models of Plastic Deformation

    work page 2013

  13. [14]

    Shanthraj, P

    P. Shanthraj, P. Eisenlohr, M. Diehl, and F. Roters. Numerically robust spectral methods for crystal plasticity simulations of heterogeneous materials.International Journal of Plasticity, 66(SI):31–45, MAR 2015

    work page 2015

  14. [15]

    Vidyasagar, Abbas D

    A. Vidyasagar, Abbas D. Tutcuoglu, and Dennis M. Kochmann. Deformation patterning in finite-strain crystal plasticity by spectral homogenization with application to magnesium.Computer Methods in Applied Mechanics and Engineering, 335:584–609, JUN 15 2018

    work page 2018

  15. [16]

    Lebensohn and Anthony D

    Ricardo A. Lebensohn and Anthony D. Rollett. Spectral methods for full-field micromechanical modelling of polycrystalline materials.Computational materials science, 173, FEB 15 2020. 20 APREPRINT- DECEMBER9, 2025

    work page 2020

  16. [17]

    Lallit Anand. Single-crystal elasto-viscoplasticity: application to texture evolution in polycrystalline metals at large strains.Computer Methods in Applied Mechanics and Engineering, 193(48):5359–5383, 2004. Advances in Computational Plasticity

    work page 2004

  17. [18]

    Beyerlein, and Marko Knezevic

    Milan Ardeljan, Irene J. Beyerlein, and Marko Knezevic. A dislocation density based crystal plasticity finite element model: Application to a two-phase polycrystalline hcp/bcc composites.Journal of the Mechanics and Physics of Solids, 66:16–31, 2014

    work page 2014

  18. [19]

    Beyerlein, and Marko Knezevic

    Milan Ardeljan, Irene J. Beyerlein, and Marko Knezevic. Effect of dislocation density-twin interactions on twin growth in az31 as revealed by explicit crystal plasticity finite element modeling.International Journal of Plasticity, 99:81–101, 2017

    work page 2017

  19. [20]

    Kalidindi, C.A

    S.R. Kalidindi, C.A. Bronkhorst, and L. Anand. Crystallographic texture evolution in bulk deformation processing of fcc metals.Journal of the Mechanics and Physics of Solids, 40(3):537–569, 1992

    work page 1992

  20. [21]

    Mishra, Roger D

    Marko Knezevic, Amanda Levinson, Ryan Harris, Raja K. Mishra, Roger D. Doherty, and Surya R. Kalidindi. Deformation twinning in az31: Influence on strain hardening and texture evolution.Acta Materialia, 58(19):6230– 6242, 2010

    work page 2010

  21. [22]

    Daymond, and Irene J

    Marko Knezevic, Mark R. Daymond, and Irene J. Beyerlein. Modeling discrete twin lamellae in a microstructural framework.Scripta Materialia, 121:84–88, 2016

    work page 2016

  22. [23]

    John Wiley & Sons, Ltd, 2010

    Franz Roters, Philip Eisenlohr, Thomas R Bieler, and Dierk Raabe.Crystal Plasticity Finite Element Methods: In Materials Science and Engineering. John Wiley & Sons, Ltd, 2010

    work page 2010

  23. [24]

    Cross-scale prediction from rve to component.International Journal of Plasticity, 140:102973, 2021

    Xinxin Sun, Hongwei Li, Mei Zhan, Junyuan Zhou, Jian Zhang, and Jia Gao. Cross-scale prediction from rve to component.International Journal of Plasticity, 140:102973, 2021

    work page 2021

  24. [25]

    A new ann based crystal plasticity model for fcc materials and its application to non-monotonic strain paths.International Journal of Plasticity, 144:103059, 2021

    Olga Ibragimova, Abhijit Brahme, Waqas Muhammad, Julie Lévesque, and Kaan Inal. A new ann based crystal plasticity model for fcc materials and its application to non-monotonic strain paths.International Journal of Plasticity, 144:103059, 2021

    work page 2021

  25. [26]

    A physics-constrained neural network for crystal plasticity modelling of fcc materials.Scripta Materialia, 241:115861, 2024

    Guowei Zhou, Yuanzhe Hu, Zizheng Cao, Myoung Gyu Lee, and Dayong Li. A physics-constrained neural network for crystal plasticity modelling of fcc materials.Scripta Materialia, 241:115861, 2024

    work page 2024

  26. [27]

    On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids.Journal of the Mechanics and Physics of Solids, 158:104697, 2022

    Colin Bonatti and Dirk Mohr. On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids.Journal of the Mechanics and Physics of Solids, 158:104697, 2022

    work page 2022

  27. [28]

    From cp-fft to cp-rnn: Recurrent neural network surrogate model of crystal plasticity.International Journal of Plasticity, 158:103430, 2022

    Colin Bonatti, Bekim Berisha, and Dirk Mohr. From cp-fft to cp-rnn: Recurrent neural network surrogate model of crystal plasticity.International Journal of Plasticity, 158:103430, 2022

    work page 2022

  28. [29]

    A temporal graph neural network for cross-scale modelling of polycrystals considering microstructure interaction.International Journal of Plasticity, 179:104017, 2024

    Yuanzhe Hu, Guowei Zhou, Myoung-Gyu Lee, Peidong Wu, and Dayong Li. A temporal graph neural network for cross-scale modelling of polycrystals considering microstructure interaction.International Journal of Plasticity, 179:104017, 2024

    work page 2024

  29. [30]

    Zeliang Liu and C.T. Wu. Exploring the 3d architectures of deep material network in data-driven multiscale mechanics.Journal of the Mechanics and Physics of Solids, 127:20–46, 2019

    work page 2019

  30. [31]

    On the micromechanics of deep material networks

    Sebastian Gajek, Matti Schneider, and Thomas Böhlke. On the micromechanics of deep material networks. Journal of the Mechanics and Physics of Solids, 142:103984, 2020

    work page 2020

  31. [32]

    Van Dung Nguyen and Ludovic Noels. Micromechanics-based material networks revisited from the interaction viewpoint; robust and efficient implementation for multi-phase composites.European Journal of Mechanics - A/Solids, 91:104384, 2022

    work page 2022

  32. [33]

    Interaction-based material network: A general framework for (porous) microstructured materials.Computer Methods in Applied Mechanics and Engineering, 389:114300, 2022

    Ludovic Noels et al. Interaction-based material network: A general framework for (porous) microstructured materials.Computer Methods in Applied Mechanics and Engineering, 389:114300, 2022

    work page 2022

  33. [34]

    Deep material network for thermal conductivity problems: Application to woven composites.Computer Methods in Applied Mechanics and Engineering, 431:117279, 2024

    Dongil Shin, Peter Jefferson Creveling, Scott Alan Roberts, and Rémi Dingreville. Deep material network for thermal conductivity problems: Application to woven composites.Computer Methods in Applied Mechanics and Engineering, 431:117279, 2024

    work page 2024

  34. [35]

    Deep material networks for fiber suspensions with infinite material contrast.International Journal for Numerical Methods in Engineering, 126(7):e70014, 2025

    Benedikt Sterr, Sebastian Gajek, Andrew Hrymak, Matti Schneider, and Thomas Böhlke. Deep material networks for fiber suspensions with infinite material contrast.International Journal for Numerical Methods in Engineering, 126(7):e70014, 2025

    work page 2025

  35. [36]

    Graph-enhanced deep material network: multiscale materials modeling with microstructural informatics.Computational Mechanics, 75:113–136, 2025

    Jimmy Gaspard Jean, Tung Huan Su, Szu Jui Huang, Cheng-Tang Wu, and Chuin Shan Chen. Graph-enhanced deep material network: multiscale materials modeling with microstructural informatics.Computational Mechanics, 75:113–136, 2025

    work page 2025

  36. [37]

    Decoding material networks: exploring performance of deep material network and interaction-based material networks.Journal of Mechanics, 40:796–807, 2024

    Wen-Ning Wan, Ting-Ju Wei, Tung-Huan Su, and Chuin-Shan Chen. Decoding material networks: exploring performance of deep material network and interaction-based material networks.Journal of Mechanics, 40:796–807, 2024. 21 APREPRINT- DECEMBER9, 2025

    work page 2024

  37. [38]

    Foundation model for composite microstructures: Reconstruction, stiffness, and nonlinear behavior prediction.Materials & Design, 257:114397, 2025

    Ting-Ju Wei and Chuin-Shan Chen. Foundation model for composite microstructures: Reconstruction, stiffness, and nonlinear behavior prediction.Materials & Design, 257:114397, 2025

    work page 2025

  38. [39]

    Orientation-aware interaction-based deep material network in polycrystalline materials modeling.Computer Methods in Applied Mechanics and Engineering, 441:117977, 2025

    Ting-Ju Wei, Tung-Huan Su, and Chuin-Shan Chen. Orientation-aware interaction-based deep material network in polycrystalline materials modeling.Computer Methods in Applied Mechanics and Engineering, 441:117977, 2025

    work page 2025

  39. [40]

    Machine learning enhanced analysis of ebsd data for texture representa- tion.npj Computational Materials, 10(1):133, 2024

    J Wanni, CA Bronkhorst, and DJ Thoma. Machine learning enhanced analysis of ebsd data for texture representa- tion.npj Computational Materials, 10(1):133, 2024

    work page 2024

  40. [41]

    How Attentive are Graph Attention Networks?

    Shaked Brody, Uri Alon, and Eran Yahav. How attentive are graph attention networks?arXiv preprint arXiv:2105.14491, 2021

    work page internal anchor Pith review arXiv 2021

  41. [42]

    Michael A Groeber and Michael A Jackson. Dream. 3d: a digital representation environment for the analysis of microstructure in 3d.Integrating materials and manufacturing innovation, 3(1):56–72, 2014

    work page 2014

  42. [43]

    Studying the mi- cromechanical behaviors of a polycrystalline metal by artificial neural networks.Acta Materialia, 214:117006, 2021

    Wei Dai, Huamiao Wang, Qiang Guan, Dayong Li, Yinghong Peng, and Carlos N Tomé. Studying the mi- cromechanical behaviors of a polycrystalline metal by artificial neural networks.Acta Materialia, 214:117006, 2021

    work page 2021

  43. [44]

    Beyerlein, David J

    Saiyi Li, Irene J. Beyerlein, David J. Alexander, and Sven C. V ogel. Texture evolution during multi-pass equal channel angular extrusion of copper: Neutron diffraction characterization and polycrystal modeling.Acta Materialia, 53(7):2111–2125, 2005

    work page 2005

  44. [45]

    Barrett and Marko Knezevic

    Timothy J. Barrett and Marko Knezevic. Deep drawing simulations using the finite element method embedding a multi-level crystal plasticity constitutive law: Experimental verification and sensitivity analysis.Computer Methods in Applied Mechanics and Engineering, 354:245–270, 2019

    work page 2019

  45. [46]

    Damask documentation: Phenopowerlaw aa6022-t4, 2024

    Max-Planck-Institut für Eisenforschung GmbH. Damask documentation: Phenopowerlaw aa6022-t4, 2024. Accessed: 21st June 2024

    work page 2024

  46. [47]

    Tianyu Huang, Zeliang Liu, CT Wu, and Wei Chen. Microstructure-guided deep material network for rapid nonlinear material modeling and uncertainty quantification.Computer Methods in Applied Mechanics and Engineering, 398:115197, 2022

    work page 2022

  47. [48]

    Deep material network via a quilting strategy: visualization for explainability and recursive training for improved accuracy.npj Computational Materials, 9(1):128, 2023

    Dongil Shin, Ryan Alberdi, Ricardo A Lebensohn, and Rémi Dingreville. Deep material network via a quilting strategy: visualization for explainability and recursive training for improved accuracy.npj Computational Materials, 9(1):128, 2023. Appendix A Phenomenological Crystal Plasticity Model In this work, the phenomenological crystal plasticity framework ...

    work page 2023