Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

The paper argues that the thermodynamic framework embedded in a neural constitutive model—dissipation potential, generalized standard materials, or metriplectic dynamics—is itself an inductive bias that changes how well the model generalize

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-02 19:17 UTC pith:C2CPPA4X

load-bearing objection Useful controlled comparison of thermodynamic learning frameworks, but the attribution claim is weaker than the abstract suggests; still worth refereeing. the 4 major comments →

arxiv 2603.02645 v2 pith:C2CPPA4X submitted 2026-03-03 cond-mat.mtrl-sci

A hierarchy of thermodynamics learning frameworks for inelastic constitutive modeling

Reese E. Jones , Jan N. Fuhg This is my paper
classification cond-mat.mtrl-sci
keywords inelasticityinternal state variablesdissipation potentialgeneralized standard materialsmetriplectic dynamicsneural ordinary differential equationsthermodynamically consistent neural networksinductive bias
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper sets out to determine whether the thermodynamic scaffolding inside a neural constitutive model has a measurable effect on learning and generalization, independent of the network architecture. The authors implement three frameworks—dissipation potential (DP), generalized standard materials (GSM), and metriplectic (MP)—within a shared neural-potential architecture using invariant inputs and neural ordinary differential equations, with matched parameter counts. They train and test on three high-fidelity simulated datasets: an elastoplastic alloy, a viscoelastic composite, and a rate-dependent crystal plasticity polycrystal. All three frameworks predict held-out stress accurately, but their relative performance shifts by dataset: the least-constrained DP model does best on the hardest microstructural data, while the most-constrained GSM model shows a marginal advantage on homogeneous materials with closed-form behavior. The paper's central claim is that this dataset-dependent pattern shows thermodynamic model selection acts as a genuine inductive bias in data-driven constitutive modeling.

Core claim

The paper claims that when DP, GSM, and MP formulations are each realized in the same neural-potential architecture with matched parameters, the differences in held-out prediction error reflect thermodynamic structure rather than architecture. The hierarchy runs from least restrictive (DP: convexity in the conjugate force with flexible state dependence) to most restrictive (GSM: additional convex duality and normality), with MP distinguished by an operator-based splitting into energy-conserving and entropy-producing dynamics. On three representative microstructural datasets all three models give accurate stress predictions; DP is marginally best on the elastoplastic alloy, while GSM underper

What carries the argument

The common machinery is a pair of neural potentials—a free energy (or internal energy for MP) and a dissipation potential—built from invariant inputs and differentiated to produce stress, conjugate force, and internal-state flow, with evolution integrated as a neural ordinary differential equation. DP uses flow as the gradient of the dissipation potential in the conjugate force; GSM adds convex duality and normality so that flow and force are dual variables; MP evolves states through the sum of a skew-symmetric energy-conserving operator L and a symmetric entropy-producing operator M, with degeneracy enforced by projection. In the comparison, L and M are fixed and non-learned so that the ope

Load-bearing premise

The headline comparison stands or falls on the assumption that nearly identical neural potentials, matched parameter counts, and fixed non-learned L and M operators are enough to remove architectural and optimization confounding, so that measured error differences reflect thermodynamic structure alone—an assumption the paper states as its design goal rather than proving directly, especially given its own note that the observable stress is insensitive to some internal-state dy

What would settle it

Retrain the metriplectic model with learnable metric and symplectic operators (within the same parameter budget) on the same three datasets; if its validation error drops materially or the DP/GSM/MP ranking changes, the fixed-operator parity assumption is the likely cause of the observed ordering. A complementary check is to generate datasets with known non-associative flow and test whether GSM's error worsens exactly as its normality restriction predicts.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • All three thermodynamic frameworks can be trained on simulation data to predict held-out inelastic stress traces accurately, so thermodynamic consistency by construction does not prevent learning.
  • The least-constrained DP model is a strong default for complex microstructural data, while the most-constrained GSM model pays off when data conforms to associative, normal-flow structure.
  • Thermodynamic model selection becomes a meaningful design choice: the same neural code base generalizes differently depending on which thermodynamic structure is embedded.
  • The metriplectic framework can match the other two on these benchmarks while cleanly separating reversible and irreversible dynamics, giving it a distinct set of internal trajectories that could be exploited for operator-based extensions.
  • On data that is not cleanly associative or that hides some internal-state dynamics, stricter thermodynamic restrictions cost some accuracy, indicating a trade-off between transparency and expressive flexibility.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial: The reversal between microstructural and homogeneous datasets suggests the value of thermodynamic constraints tracks how well the data generator itself conforms to the theory; a direct test would vary non-associativity or rate-dependence in generated data and watch whether GSM's relative error tracks it.
  • Editorial: Fixing L and M to non-learned forms was done for parameter parity, so the MP results likely understate the formulation's ceiling; allowing learned operators with the same degeneracy projections is the natural follow-up that could change the ranking.
  • Editorial: Because the observable stress is only partially sensitive to the hidden states, the discovered internal-state trajectories should not be over-interpreted; comparing latent spaces through invariants rather than raw trajectories would be a more robust way to ask whether the frameworks find shared representations.
  • Editorial: The implicit-standard-material bipotential structure, identified in the paper as a superclass containing DP, GSM, and metric MP, suggests a concrete falsification target: if a neural bipotential implementation matches or beats all three frameworks on the same benchmarks, the differences among the three may be artifacts of the separable ansatz rather than irreducible thermodynamic constr

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a unified comparison of three thermodynamically consistent inelastic constitutive modeling frameworks—dissipation potential (DP), generalized standard materials (GSM), and metriplectic (MP)—implemented within a common neural potential architecture using invariant representations and neural ODEs. The authors aim to attribute differences in predictive performance to thermodynamic structure (duality, normality, convexity, operator-based evolution) rather than to architectural variation. They train and evaluate the three models on three RVE-generated datasets (elastoplastic alloy, viscoelastic composite, viscoplastic polycrystal) and on four closed-form benchmark datasets, reporting held-out RMSE for 10 independent training runs. The central claim is that thermodynamic model selection acts as an inductive bias that aids generalization.

Significance. If the attribution claim holds, the paper would provide a valuable controlled comparison of thermodynamic frameworks for data-driven constitutive modeling, with practical implications for model selection in inelasticity. Strengths include a thoughtful theoretical review (Sec. 3), a unified implementation (Sec. 4), matched potential parameterizations, near-equal parameter counts, 10 independent trainings, held-out validation, and an additional closed-form benchmark set. However, the central claim is currently conditional: the comparison design does not fully eliminate implementation-level confounds, and the absence of code/data release prevents independent verification. The paper nonetheless makes a useful contribution by sharpening the question of how thermodynamic structure influences learnability and generalization.

major comments (4)
  1. [Sec. 4, Eq. (61)] The claim that observed RMSE rankings (Fig. 10) reflect thermodynamic structure is undermined by the arbitrary choice of fixed MP operators L and M in Eq. (61). The metriplectic formalism only requires skew-symmetry of L and symmetric positive-definiteness of M (with degeneracy conditions, Eq. (34)); it does not prescribe the specific values chosen here. Different valid operators—e.g., diagonal M with unequal entries, or L=0—define different function classes and can change the expressiveness of the MP model on a given dataset. Since the stated design goal is to attribute performance differences to thermodynamic structure rather than architecture, the paper needs either (i) a sensitivity analysis over L and M choices, or (ii) a principled argument that the chosen fixed operators are neutral. Without this, the MP ranking is confounded with this implementation choice.
  2. [Sec. 4, Eq. (46) vs. Sec. 3, Eqs. (25)-(26)] The GSM dissipation potential is restricted to φ_GSM(k, I_dotC), omitting dependence on κ and E that is permitted by the general GSM theory (Eqs. (25)-(26); see also Table 2, 'dependence on state: restricted'). This is an implementation choice, not a requirement of GSM. Because GSM is the worst performer on the RVE datasets (Fig. 10a), the ranking may reflect this omitted dependency rather than the GSM thermodynamic structure itself. To support the paper's central claim, the authors should either implement GSM with the full allowed state dependence, or provide evidence (e.g., a DP model with the same input restriction) that the omitted dependencies are not responsible for the observed performance gap.
  3. [Sec. 6 / Table 7] The 'parameter parity' design does not remove architectural confounding. Table 7 shows input dimensions of 9/18/9/12 for the potentials across frameworks, with hidden widths adjusted to match total parameter counts. Input dimension is part of the architecture; changing it alters the inductive bias even at equal parameter count. The ISV dimension (6) is acknowledged as 'the most arbitrary' (Sec. 6) and chosen from prior studies. Without a sensitivity study varying ISV dimension and hidden widths, the equal-parameter-count condition is insufficient to attribute performance differences to thermodynamic structure. Please report learning curves and rankings for at least one alternative ISV dimension and network width.
  4. [Secs. 5-7] No code or data availability is stated. All conclusions rest on 10 independent trainings on three RVE datasets drawn from the authors' prior work (e.g., Refs. [51], [57], [80]) plus an additional closed-form benchmark set; without release of the data-generation scripts, training code, and random seeds, the central comparison cannot be independently verified. The body text itself acknowledges weak identifiability of the hidden states (Sec. 7), so the qualitative internal-state and dissipation analysis in Secs. 6.1-6.3 cannot serve as a substitute. Please include a data/code availability statement or explain any restrictions.
minor comments (5)
  1. [Abstract / Sec. 8] The standalone abstract says the unified setting 'ensures that performance differences can be attributed to thermodynamic structure,' while the in-paper abstract says it 'reduces architectural confounding and allows us to assess, to the extent possible'; Sec. 8 similarly states the results 'demonstrate' an inductive-bias effect. Please align these statements to the more cautious one.
  2. [Table 3 / Sec. 6.1] The abbreviation 'GS' appears in text and Fig. 10 instead of 'GSM'; also, the 'T=1' assumption for MP in Table 3 should be explained (is T dimensionless? normalized?).
  3. [Sec. 5, Eq. (65)] The data-generation statement says 256 trajectories with 200 steps and an 80/10/10 split, but the exact numbers of training, in-training test, and validation trajectories used are not given; please state the exact counts.
  4. [Sec. 4, Eqs. (53)-(56)] The normalization of the GSM dissipation potential in Eqs. (55)-(56) contains a nonzero intercept p_φ(k) that is affine in I4; the notation p_φ(k) in Eq. (56) is a slight abuse since the derivative is evaluated at k=0, I_dotC=0. Please clarify.
  5. [Fig. 8] The free-energy visualization in Fig. 8 is described as a 'diagnostic' but the color scale for the third column (I3) is unclear; consider adding a colorbar or explicit potential values.

Circularity Check

0 steps flagged

No significant circularity: the central comparison is an empirical benchmark against independent RVE simulations; only minor non-load-bearing self-citations and acknowledged implementation confounds.

full rationale

The paper's central claim—that thermodynamic structure acts as an inductive bias—is supported by held-out predictions on data generated from classical J2 plasticity, the UPM viscoelastic model, and crystal plasticity, not by the learned potentials themselves. Training uses an 80/10/10 split, and reported errors are on post-training validation trajectories (Sec. 6), so no fitted quantity is renamed as a prediction. The three frameworks share the same ISNN/neural-ODE scaffolding (Sec. 4, Table 3), and the differences in inputs and flow rules are exactly the 'thermodynamic structure' being tested; the outcome in Fig. 10 is empirically determined and dataset-dependent, not forced by construction. The paper cites its own prior work for the ISNN architecture [96], the datasets [51,106], and ISV-size heuristics [51,57,80], but none of these citations supplies the comparative result itself; the claim stands or falls on the new experiments. Acknowledged limitations—the arbitrary ISV dimension, fixed L/M operators, restricted GSM dissipation inputs, and hidden-state non-identifiability (Secs. 4, 6, 7)—are potential confounds for the attribution to 'thermodynamic structure,' and the full-text abstract appropriately says 'reduces architectural confounding' rather than the summary abstract's stronger 'ensures.' Confounding is a validity risk, not a circular derivation: no equation in the paper reduces an output to an input by definition, and no prediction is identical to a fitted parameter by construction.

Axiom & Free-Parameter Ledger

3 free parameters · 7 axioms · 0 invented entities

The central claim relies on standard continuum thermodynamics assumptions (local equilibrium, simple materials, potential-based dissipation) and on a fair-comparison design choice (identical architecture, fixed operators). These are all reasonable within the stated scope but limit the breadth of the conclusion.

free parameters (3)
  • Internal state variable dimension = 6
    Chosen 'based on previous studies' (Sec. 6); determines expressiveness of all three models and could bias the comparison if too small or too large.
  • Fixed MP operators L and M = L skew with off-diagonal ±1; M=2 diagonal, 1 off-diagonal (Eq. 61)
    Hand-selected rather than learned; this restriction may handicap the MP model and is a free design choice in the comparison.
  • Training hyperparameters = Adam lr=0.001, 80,000 epochs, patience 8,000, 2 layers, Softplus
    Manual selections 'guided by preliminary studies'; the fair-tuning assumption underpins the attribution of performance differences to thermodynamic structure.
axioms (7)
  • domain assumption Local equilibrium thermodynamics: S=∂Eψ and η=−∂Tψ (Coleman–Noll procedure)
    Invoked for ISV/DP frameworks in Sec. 3.1 (Eqs. 6–8) to identify stresses with free-energy derivatives; if violated in real materials, the compared frameworks may not apply.
  • domain assumption Simple material behavior with fading memory: response is a functional of first-gradient deformation history at a point
    Sec. 3.2; excludes non-local/gradient effects and restricts scope to homogeneous materials.
  • domain assumption Existence of Helmholtz free energy ψ(E,κ,T) and convex dissipation potentials; LF duality for GSM
    Sec. 3.3; all models are built on these potentials; GSM requires dual convex potentials satisfying the Biot relation Eq. (23).
  • domain assumption Metriplectic degeneracies: L skew-symmetric, M symmetric positive definite, L∂s=0, M∂e=0
    Sec. 3.3 Eqs. (30)–(37); the MP dissipation inequality Γ=Th·Mh≥0 depends on these operator properties.
  • domain assumption Isotropy and invariant parameterization: potentials depend only on scalar invariants of C and Cdot; ISVs are scalar invariants
    Sec. 4 Eq. (41); restricts models to isotropic response under uniaxial loading, a modeling choice not tested on anisotropic data.
  • domain assumption RVE datasets are representative high-fidelity ground truth
    Sec. 5; data generated from J2 plasticity, UPM viscoelasticity, and crystal-plasticity FE simulations; if these generators are themselves inconsistent with the fitted frameworks, conclusions about real materials weaken.
  • domain assumption Internal-state dynamics are adequately identifiable from a single stress component for fair comparison
    The models observe only the 11 stress component but fit a 6-dimensional hidden state; Sec. 7 acknowledges ambiguities in the learned dynamics.

pith-pipeline@v1.3.0-alltime-deepseek · 26066 in / 14865 out tokens · 139544 ms · 2026-08-02T19:17:27.122370+00:00 · methodology

0 comments
read the original abstract

Recent advances in physics-augmented neural networks have enabled thermodynamically consistent data-driven constitutive modeling of complex inelastic materials. Most existing approaches, however, implicitly adopt a specific thermodynamic framework and embed structural assumptions such as normality, dual dissipation potentials, or other structure from manually constructed models directly into the learning architecture. Consequently, differences in predictive performance may arise not only from data or network design, but also from the underlying theoretical assumptions. In this work, we present a unified comparison of several thermodynamically consistent inelastic modeling frameworks from a machine learning perspective. We consider internal-variable formulations with dissipation potential, generalized standard materials, and metriplectic structures, and we analyze their structural assumptions, admissible dependencies, convexity requirements, and implications for dissipation and evolution. Each framework is implemented within a common neural potential architecture based on invariant representations and neural ordinary differential equations. This unified setting ensures that performance differences can be attributed to thermodynamic structure rather than architectural variation. The models are trained and evaluated on three representative inelastic datasets generated from high-fidelity representative volume element simulations: an elastoplastic alloy, a viscoelastic composite, and a rate-dependent crystal plasticity polycrystal. By isolating the role of thermodynamic structure, we assess how restrictions such as duality, normality, operator-based evolution, and convexity influence learnability, expressiveness, stability, and generalization.

Figures

Figures reproduced from arXiv: 2603.02645 by Jan N. Fuhg, Reese E. Jones.

Figure 1
Figure 1. Figure 1: Schematic of the DP framework component DP GSM MP energy ψ(IC, κ) ψ(IC, κ) ϵ(IC, κ) minimum at IC = 1 IC = 1 IC = 1 monotone & convex I1, I2 I1, I2 I1, I2 convex I3 I3 I3 free κ κ κ dissipation ϕ(k, IC,C˙ , κ) ϕ(k, IC˙ ) ϕ(IC,C˙ , κ) minimum at k = 0 k = 0, IC˙ = 0 – monotone & convex – IC˙ – convex k k – free IC,C˙ , κ – IC,C˙ , κ stress S = 2∂Cψ S = 2∂Cψ − 2∂C˙ ϕ S = 2∂Cψ conjugate force k = −∂κψ k = −∂κ… view at source ↗
Figure 2
Figure 2. Figure 2: AlSi10Mg alloy. (a) representative volume element (red: Al, gray:Si), (b) representative hysteresis curves of a random ensemble where realizations are distinguished by color. properties of the constituents [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Silicone-glass bead composite. (a) representative volume element showing spherical glass inclusions, (b) [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FCC Fe polycrystal. (a) representative volume element showing polycrystalline structure colored by grain [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: EP prediction comparison. A representative randomly-selected subset of the validation data is shown for [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: VE prediction comparison. A representative randomly-selected subset of the validation data is shown for [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: VP prediction comparison. A representative randomly-selected subset of the validation data is shown for [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Representative qualitative diagnostic of the learned free energy [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Representative qualitative diagnostic of the learned dissipation potential [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Summary of validation errors for 10 independent trainings. [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Interval and fuzzy physics-augmented neural networks (iPANN and fPANN) for uncertainty quantification and propagation in constitutive modeling

    cs.LG 2026-07 conditional novelty 5.0

    A neural-network constitutive model whose lower/mean/upper free-energy branches are trained to enclose noisy stress observations, plus a fuzzy alpha-cut interpolation for adjustable conservatism.

Reference graph

Works this paper leans on

115 extracted references · 7 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Heinz Linhart and Walter Zucchini.Model selection.John Wiley & Sons, 1986

    1986

  2. [2]

    Model selection techniques: An overview.IEEE Signal Processing Magazine, 35(6):16–34, 2018

    Jie Ding, Vahid Tarokh, and Yuhong Yang. Model selection techniques: An overview.IEEE Signal Processing Magazine, 35(6):16–34, 2018

    2018

  3. [3]

    A review on data-driven constitutive laws for solids.Archives of Computational Methods in Engineering, 32(3):1841–1883, 2025

    Jan N Fuhg, Govinda Anantha Padmanabha, Nikolaos Bouklas, Bahador Bahmani, WaiChing Sun, Nikolaos N Vlassis, Moritz Flaschel, Pietro Carrara, and Laura De Lorenzis. A review on data-driven constitutive laws for solids.Archives of Computational Methods in Engineering, 32(3):1841–1883, 2025

    2025

  4. [4]

    Finite electro-elasticity with physics-augmented neural networks.Computer Methods in Applied Mechanics and Engineering, 400:115501, 2022

    Dominik K Klein, Rogelio Ortigosa, Jesús Martínez-Frutos, and Oliver Weeger. Finite electro-elasticity with physics-augmented neural networks.Computer Methods in Applied Mechanics and Engineering, 400:115501, 2022

    2022

  5. [5]

    Viscoelasticty with physics-augmented neural networks: model formulation and training methods without prescribed internal vari- ables.Computational Mechanics, 74(6):1279–1301, 2024

    Max Rosenkranz, Karl A Kalina, Jörg Brummund, WaiChing Sun, and Markus Kästner. Viscoelasticty with physics-augmented neural networks: model formulation and training methods without prescribed internal vari- ables.Computational Mechanics, 74(6):1279–1301, 2024

    2024

  6. [6]

    Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics.Computer Methods in Applied Mechanics and Engineering, 426:116973, 2024

    Jan Niklas Fuhg, Reese Edward Jones, and Nikolaos Bouklas. Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics.Computer Methods in Applied Mechanics and Engineering, 426:116973, 2024. 23 Learning frameworks for modeling inelasticityA PREPRINT

    2024

  7. [7]

    Antoine Benady, Emmanuel Baranger, and Ludovic Chamoin. Nn-mcre: A modified constitutive relation error framework for unsupervised learning of nonlinear state laws with physics-augmented neural networks.Inter- national Journal for Numerical Methods in Engineering, 125(8):e7439, 2024

    2024

  8. [8]

    Experimental learning of a hypere- lastic behavior with a physics-augmented neural network.Experimental Mechanics, 64(9):1465–1481, 2024

    Clément Jailin, Antoine Benady, Remi Legroux, and Emmanuel Baranger. Experimental learning of a hypere- lastic behavior with a physics-augmented neural network.Experimental Mechanics, 64(9):1465–1481, 2024

    2024

  9. [9]

    Inverse design of anisotropic microstructures using physics-augmented neural networks.Journal of the Mechanics and Physics of Solids, page 106161, 2025

    Asghar A Jadoon, Karl A Kalina, Manuel K Rausch, Reese Jones, and Jan Niklas Fuhg. Inverse design of anisotropic microstructures using physics-augmented neural networks.Journal of the Mechanics and Physics of Solids, page 106161, 2025

    2025

  10. [10]

    Polyconvex physics- augmented neural network constitutive models in principal stretches.International Journal of Solids and Struc- tures, page 113469, 2025

    Adrian Buganza Tepole, Asghar Arshad Jadoon, Manuel Rausch, and Jan Niklas Fuhg. Polyconvex physics- augmented neural network constitutive models in principal stretches.International Journal of Solids and Struc- tures, page 113469, 2025

    2025

  11. [11]

    A physics-augmented neural network framework for finite strain incompressible viscoelasticity.arXiv preprint arXiv:2511.02959, 2025

    Karl A Kalina, Jörg Brummund, and Markus Kästner. A physics-augmented neural network framework for finite strain incompressible viscoelasticity.arXiv preprint arXiv:2511.02959, 2025

    Pith/arXiv arXiv 2025

  12. [12]

    Recovering mullins damage hyperelastic behaviour with physics augmented neural networks.Journal of the Mechanics and Physics of Solids, 193:105839, 2024

    Martin Zlati ´c and Marko ˇCana ¯dija. Recovering mullins damage hyperelastic behaviour with physics augmented neural networks.Journal of the Mechanics and Physics of Solids, 193:105839, 2024

    2024

  13. [13]

    Kevin Linka, Markus Hillgärtner, Kian P Abdolazizi, Roland C Aydin, Mikhail Itskov, and Christian J Cy- ron. Constitutive artificial neural networks: A fast and general approach to predictive data-driven constitutive modeling by deep learning.Journal of Computational Physics, 429:110010, 2021

    2021

  14. [14]

    A new family of constitutive artificial neural networks towards automated model discovery.Computer Methods in Applied Mechanics and Engineering, 403:115731, 2023

    Kevin Linka and Ellen Kuhl. A new family of constitutive artificial neural networks towards automated model discovery.Computer Methods in Applied Mechanics and Engineering, 403:115731, 2023

    2023

  15. [15]

    Theory and implementation of inelastic constitutive artificial neural networks.Computer Methods in Applied Mechanics and Engineering, 428:117063, 2024

    Hagen Holthusen, Lukas Lamm, Tim Brepols, Stefanie Reese, and Ellen Kuhl. Theory and implementation of inelastic constitutive artificial neural networks.Computer Methods in Applied Mechanics and Engineering, 428:117063, 2024

    2024

  16. [16]

    Accounting for plasticity: An extension of inelastic constitutive artificial neural networks.arXiv preprint arXiv:2407.19326, 2024

    Birte Boes, Jaan-Willem Simon, and Hagen Holthusen. Accounting for plasticity: An extension of inelastic constitutive artificial neural networks.arXiv preprint arXiv:2407.19326, 2024

    Pith/arXiv arXiv 2024

  17. [17]

    Polyconvex inelastic constitutive artificial neural networks.PAMM, 24(3):e202400032, 2024

    Hagen Holthusen, Lukas Lamm, Tim Brepols, Stefanie Reese, and Ellen Kuhl. Polyconvex inelastic constitutive artificial neural networks.PAMM, 24(3):e202400032, 2024

    2024

  18. [18]

    Kian P Abdolazizi, Kevin Linka, and Christian J Cyron. Viscoelastic constitutive artificial neural networks (vcanns)–a framework for data-driven anisotropic nonlinear finite viscoelasticity.Journal of computational physics, 499:112704, 2024

    2024

  19. [19]

    Prakash Thakolkaran, Yaqi Guo, Shivam Saini, Mathias Peirlinck, Benjamin Alheit, and Siddhant Kumar. Can kan cans? input-convex kolmogorov-arnold networks (kans) as hyperelastic constitutive artificial neural net- works (cans).Computer Methods in Applied Mechanics and Engineering, 443:118089, 2025

    2025

  20. [20]

    Hamiltonian neural networks.Advances in neural information processing systems, 32, 2019

    Samuel Greydanus, Misko Dzamba, and Jason Yosinski. Hamiltonian neural networks.Advances in neural information processing systems, 32, 2019

    2019

  21. [21]

    Lagrangian neural networks.arXiv preprint arXiv:2003.04630, 2020

    Miles Cranmer, Sam Greydanus, Stephan Hoyer, Peter Battaglia, David Spergel, and Shirley Ho. Lagrangian neural networks.arXiv preprint arXiv:2003.04630, 2020

    Pith/arXiv arXiv 2003

  22. [22]

    Symplectic recurrent neural networks.arXiv preprint arXiv:1909.13334, 2019

    Zhengdao Chen, Jianyu Zhang, Martin Arjovsky, and Léon Bottou. Symplectic recurrent neural networks.arXiv preprint arXiv:1909.13334, 2019

    Pith/arXiv arXiv 1909

  23. [23]

    Thermodynamically consistent data-driven computa- tional mechanics.Continuum Mechanics and Thermodynamics, 31(1):239–253, 2019

    David González, Francisco Chinesta, and Elías Cueto. Thermodynamically consistent data-driven computa- tional mechanics.Continuum Mechanics and Thermodynamics, 31(1):239–253, 2019

    2019

  24. [24]

    Efficiently parameterized neural metriplectic systems

    Anthony Gruber, Kookjin Lee, Haksoo Lim, Noseong Park, and Nathaniel Trask. Efficiently parameterized neural metriplectic systems. InThe Thirteenth International Conference on Learning Representations, 2025

    2025

  25. [25]

    Thermodynamics with internal state variables.The journal of chemical physics, 47(2):597–613, 1967

    Bernard D Coleman and Morton E Gurtin. Thermodynamics with internal state variables.The journal of chemical physics, 47(2):597–613, 1967

    1967

  26. [26]

    Cambridge university press, 1994

    Jean Lemaitre and Jean-Louis Chaboche.Mechanics of solid materials. Cambridge university press, 1994

    1994

  27. [27]

    Sur les matériaux standard généralisés.Journal de mécanique, 14(1):39–63, 1975

    Bernard Halphen and Quoc Son Nguyen. Sur les matériaux standard généralisés.Journal de mécanique, 14(1):39–63, 1975

    1975

  28. [28]

    Bracket formulation for irreversible classical fields.Physics Letters A, 100(8):423–427, 1984

    Philip J Morrison. Bracket formulation for irreversible classical fields.Physics Letters A, 100(8):423–427, 1984

    1984

  29. [29]

    Dynamics and thermodynamics of complex fluids

    Hans Christian Öttinger and Miroslav Grmela. Dynamics and thermodynamics of complex fluids. II. illustra- tions of a general formalism.Physical Review E, 56(6):6633, 1997. 24 Learning frameworks for modeling inelasticityA PREPRINT

    1997

  30. [30]

    A review on data-driven constitutive laws for solids.Archives of Computational Methods in Engineering, pages 1–43, 2024

    Jan N Fuhg, Govinda Anantha Padmanabha, Nikolaos Bouklas, Bahador Bahmani, WaiChing Sun, Nikolaos N Vlassis, Moritz Flaschel, Pietro Carrara, and Laura De Lorenzis. A review on data-driven constitutive laws for solids.Archives of Computational Methods in Engineering, pages 1–43, 2024

    2024

  31. [31]

    Reflections on the motive power of fire, and on machines fitted to develop that power.Paris: Bachelier, 108(1824):1824, 1824

    Sadi Carnot. Reflections on the motive power of fire, and on machines fitted to develop that power.Paris: Bachelier, 108(1824):1824, 1824

  32. [32]

    James Prescott Joule. Xxxii. on the calorific effects of magneto-electricity, and on the mechanical value of heat. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 23(152):263–276, 1843

  33. [33]

    Ueber die bewegende kraft der wärme und die gesetze, welche sich daraus für die wärmelehre selbst ableiten lassen.Annalen der Physik, 155(3):368–397, 1850

    Rudolf Clausius. Ueber die bewegende kraft der wärme und die gesetze, welche sich daraus für die wärmelehre selbst ableiten lassen.Annalen der Physik, 155(3):368–397, 1850

  34. [34]

    Hermann Ludwig Ferdinand von Heimholtz.Ueber die Erhaltung der Kraft: eine physikalische Abhandlung: vorgetragen in der Sitzung der physikalischen Gesellschaft zu Berlin am 23sten Juli 1847. G. Reimer, 1847

  35. [35]

    Xv.—on the dynamical theory of heat, with numerical results deduced from mr joule’s equivalent of a thermal unit, and m

    William Thomson. Xv.—on the dynamical theory of heat, with numerical results deduced from mr joule’s equivalent of a thermal unit, and m. regnault’s observations on steam.Transactions of the Royal Society of Edinburgh, 20(2):261–288, 1853

  36. [36]

    Xxii.—on the dynamical evidence of the molecular constitution of bodies.Journal of the Chemical Society, 28:493–508, 1875

    J Clerk-Maxwell. Xxii.—on the dynamical evidence of the molecular constitution of bodies.Journal of the Chemical Society, 28:493–508, 1875

  37. [37]

    Graphical methods in the thermodynamics of fluids.Transactions of the Connecticut Academy of Arts and Sciences, 1:309–342, 1873

    Josiah Willard Gibbs. Graphical methods in the thermodynamics of fluids.Transactions of the Connecticut Academy of Arts and Sciences, 1:309–342, 1873

  38. [38]

    Reciprocal relations in irreversible processes

    Lars Onsager. Reciprocal relations in irreversible processes. I.Physical review, 37(4):405, 1931

    1931

  39. [39]

    Reciprocal relations in irreversible processes

    Lars Onsager. Reciprocal relations in irreversible processes. II.Physical review, 38(12):2265, 1931

    1931

  40. [40]

    Introduction to thermodynamics of irreversible processes.Journal of The Electrochemical Society, 110(4):97C, 1963

    Ilya Prigogine and Pierre Van Rysselberghe. Introduction to thermodynamics of irreversible processes.Journal of The Electrochemical Society, 110(4):97C, 1963

    1963

  41. [41]

    Courier Corporation, 2013

    Sybren Ruurds De Groot and Peter Mazur.Non-equilibrium thermodynamics. Courier Corporation, 2013

    2013

  42. [42]

    Springer, 2008

    Georgy Lebon, David Jou, and José Casas-Vázquez.Understanding non-equilibrium thermodynamics: foun- dations, applications, frontiers. Springer, 2008

    2008

  43. [43]

    Quelques réflexions au sujet de la physique expérimentale

    Pierre Duhem.II. Quelques réflexions au sujet de la physique expérimentale. Hermann, 2015

    2015

  44. [44]

    The thermal conductivity of liquids under pressure

    Percy Williams Bridgman. The thermal conductivity of liquids under pressure. InProceedings of the American Academy of Arts and Sciences, volume 59, pages 141–169. JSTOR, 1923

    1923

  45. [45]

    Historical review of internal state variable theory for inelasticity

    Mark F Horstemeyer and Douglas J Bammann. Historical review of internal state variable theory for inelasticity. International Journal of Plasticity, 26(9):1310–1334, 2010

    2010

  46. [46]

    The thermodynamics of irreversible processes

    Carl Eckart. The thermodynamics of irreversible processes. i. the simple fluid.Physical Review, 58(3):267, 1940

    1940

  47. [47]

    How the internal state of a plastically deformed body is to be described in a continuum theory

    E Kröner. How the internal state of a plastically deformed body is to be described in a continuum theory. 1965

    1965

  48. [48]

    Thermodynamics with internal variables

    Gerard A Maugin and Wolfgang Muschik. Thermodynamics with internal variables. part i. general concepts. 1994

    1994

  49. [49]

    A theory of finite viscoelasticity and numerical aspects.International journal of solids and structures, 35(26-27):3455–3482, 1998

    Stefanie Reese and Sanjay Govindjee. A theory of finite viscoelasticity and numerical aspects.International journal of solids and structures, 35(26-27):3455–3482, 1998

    1998

  50. [50]

    Continuum thermodynamics.Journal of applied mechan- ics, 50:1010–1020, 1983

    Paul Germain, Quoc Son Nguyen, and Pierre Suquet. Continuum thermodynamics.Journal of applied mechan- ics, 50:1010–1020, 1983

    1983

  51. [51]

    Reese E Jones, Ari L Frankel, and KL Johnson. A neural ordinary differential equation framework for modeling inelastic stress response via internal state variables.Journal of Machine Learning for Modeling and Computing, 3(3), 2022

    2022

  52. [52]

    Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations.Advances in neural information processing systems, 31, 2018

    2018

  53. [53]

    Augmented neural odes.Advances in neural information processing systems, 32, 2019

    Emilien Dupont, Arnaud Doucet, and Yee Whye Teh. Augmented neural odes.Advances in neural information processing systems, 32, 2019

    2019

  54. [54]

    Automated discovery of generalized standard mate- rial models with euclid.Computer Methods in Applied Mechanics and Engineering, 405:115867, 2023

    Moritz Flaschel, Siddhant Kumar, and Laura De Lorenzis. Automated discovery of generalized standard mate- rial models with euclid.Computer Methods in Applied Mechanics and Engineering, 405:115867, 2023

    2023

  55. [55]

    Convex neural networks learn generalized standard material models.Journal of the Mechanics and Physics of Solids, 200:106103, 2025

    Moritz Flaschel, Paul Steinmann, Laura De Lorenzis, and Ellen Kuhl. Convex neural networks learn generalized standard material models.Journal of the Mechanics and Physics of Solids, 200:106103, 2025. 25 Learning frameworks for modeling inelasticityA PREPRINT

    2025

  56. [56]

    Data-driven anisotropic finite viscoelasticity using neural ordinary differential equations.Computer methods in applied mechanics and engineering, 411:116046, 2023

    Vahidullah Taç, Manuel K Rausch, Francisco Sahli Costabal, and Adrian Buganza Tepole. Data-driven anisotropic finite viscoelasticity using neural ordinary differential equations.Computer methods in applied mechanics and engineering, 411:116046, 2023

    2023

  57. [57]

    A physics-augmented neural network frame- work for modeling and detecting thermo-visco-plastic behavior.arXiv preprint arXiv:2512.09284, 2025

    Reese E Jones, Asghar Jadoon, D Thomas Seidl, and Jan N Fuhg. A physics-augmented neural network frame- work for modeling and detecting thermo-visco-plastic behavior.arXiv preprint arXiv:2512.09284, 2025

    arXiv 2025

  58. [58]

    A complement to neural networks for anisotropic inelasticity at finite strains

    Hagen Holthusen and Ellen Kuhl. A complement to neural networks for anisotropic inelasticity at finite strains. Computer Methods in Applied Mechanics and Engineering, 450:118612, 2026

    2026

  59. [59]

    Ham- mering at the entropy: A generic-guided approach to learning polymeric rheological constitutive equations using pinns.arXiv preprint arXiv:2409.07545, 2024

    David Nieto Simavilla, Andrea Bonfanti, Imanol GarcÃa de Beristain, Pep Español, and Marco Ellero. Ham- mering at the entropy: A generic-guided approach to learning polymeric rheological constitutive equations using pinns.arXiv preprint arXiv:2409.07545, 2024

    Pith/arXiv arXiv 2024

  60. [60]

    Port-metriplectic neural net- works: thermodynamics-informed machine learning of complex physical systems.Computational Mechanics, 72(3):553–561, 2023

    Quercus Hernández, Alberto Badías, Francisco Chinesta, and Elías Cueto. Port-metriplectic neural net- works: thermodynamics-informed machine learning of complex physical systems.Computational Mechanics, 72(3):553–561, 2023

    2023

  61. [61]

    Formulation of thermoelastic dissipative material behavior using generic.Continuum Me- chanics and Thermodynamics, 23(3):233–256, 2011

    Alexander Mielke. Formulation of thermoelastic dissipative material behavior using generic.Continuum Me- chanics and Thermodynamics, 23(3):233–256, 2011

    2011

  62. [62]

    Data-driven generic modeling of poroviscoelastic materials.Entropy, 21(12):1165, 2019

    Chady Ghnatios, Iciar Alfaro, David González, Francisco Chinesta, and Elias Cueto. Data-driven generic modeling of poroviscoelastic materials.Entropy, 21(12):1165, 2019

    2019

  63. [63]

    Mark Schiebl and Peter Betsch. Structure-preserving space-time discretization of large-strain thermo- viscoelasticity in the framework of generic.International Journal for Numerical Methods in Engineering, 122(14):3448–3488, 2021

    2021

  64. [64]

    Markus Hütter and Bob Svendsen. On the formulation of continuum thermodynamic models for solids as general equations for non-equilibrium reversible-irreversible coupling.Journal of Elasticity, 104(1):357–368, 2011

    2011

  65. [65]

    Continuum damage mechanics: combining thermodynamics with a thoughtful characterization of the microstructure.Acta mechanica, 201(1):297–312, 2008

    Markus Hütter and Theo A Tervoort. Continuum damage mechanics: combining thermodynamics with a thoughtful characterization of the microstructure.Acta mechanica, 201(1):297–312, 2008

    2008

  66. [66]

    Finite anisotropic elasticity and material frame indifference from a nonequilibrium thermodynamics perspective.Journal of Non-Newtonian Fluid Mechanics, 152(1-3):45–52, 2008

    Markus Hütter and Theo A Tervoort. Finite anisotropic elasticity and material frame indifference from a nonequilibrium thermodynamics perspective.Journal of Non-Newtonian Fluid Mechanics, 152(1-3):45–52, 2008

    2008

  67. [67]

    Statistical-mechanics based modeling of anisotropic viscoplastic defor- mation.Mechanics of Materials, 80:37–51, 2015

    Markus Hütter and Theo A Tervoort. Statistical-mechanics based modeling of anisotropic viscoplastic defor- mation.Mechanics of Materials, 80:37–51, 2015

    2015

  68. [68]

    Why generic?Journal of non-newtonian fluid mechanics, 165(17-18):980–986, 2010

    Miroslav Grmela. Why generic?Journal of non-newtonian fluid mechanics, 165(17-18):980–986, 2010

    2010

  69. [69]

    The principle of fading memory.Archive for Rational Mechanics and Analysis, 18(5):343–366, 1965

    C-C Wang. The principle of fading memory.Archive for Rational Mechanics and Analysis, 18(5):343–366, 1965

    1965

  70. [70]

    The non-linear field theories of mechanics

    Clifford Truesdell and Walter Noll. The non-linear field theories of mechanics. InThe non-linear field theories of mechanics, pages 1–579. Springer, 2004

    2004

  71. [71]

    The thermodynamics of elastic materials with heat conduction and viscosity

    Bernard D Coleman and Walter Noll. The thermodynamics of elastic materials with heat conduction and viscosity. InThe foundations of mechanics and thermodynamics: Selected papers, pages 145–156. Springer, 1974

    1974

  72. [72]

    Polyconvex neural network models of thermoelasticity.Journal of the Mechanics and Physics of Solids, 192:105837, 2024

    Jan N Fuhg, Asghar Jadoon, Oliver Weeger, D Thomas Seidl, and Reese E Jones. Polyconvex neural network models of thermoelasticity.Journal of the Mechanics and Physics of Solids, 192:105837, 2024

    2024

  73. [73]

    A new mathematical theory of simple materials.Archive for Rational Mechanics and Analysis, 48(1):1–50, 1972

    Walter Noll. A new mathematical theory of simple materials.Archive for Rational Mechanics and Analysis, 48(1):1–50, 1972

    1972

  74. [74]

    Springer Science & Business Media, 1993

    Donald R Smith.An introduction to continuum mechanics-after Truesdell and Noll, volume 22. Springer Science & Business Media, 1993

    1993

  75. [75]

    Vito Antonio Cimmelli, Antonio Sellitto, and Vita Triani. A generalized coleman–noll procedure for the ex- ploitation of the entropy principle.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 466(2115):911–925, 2010

    2010

  76. [76]

    Extended irreversible thermodynamics.Reports on Progress in Physics, 51(8):1105, 1988

    David Jou, José Casas-Vázquez, and Georgy Lebon. Extended irreversible thermodynamics.Reports on Progress in Physics, 51(8):1105, 1988

    1988

  77. [77]

    Extended irreversible thermodynamics revisited (1988-98)

    David Jou, Jose Casas-Vazquez, and Georgy Lebon. Extended irreversible thermodynamics revisited (1988-98). Reports on Progress in Physics, 62(7):1035, 1999. 26 Learning frameworks for modeling inelasticityA PREPRINT

    1988

  78. [78]

    Thermodynamics of elastic-plastic materials as a theory with internal state variables.Journal of Applied Physics, 40(8):3207–3218, 1969

    J Kratochvil and OW Dillon Jr. Thermodynamics of elastic-plastic materials as a theory with internal state variables.Journal of Applied Physics, 40(8):3207–3218, 1969

    1969

  79. [79]

    Internal state variable theory

    DL McDowell. Internal state variable theory. InHandbook of Materials Modeling: Methods, pages 1151–1169. Springer, 2005

    2005

  80. [80]

    An attention-based neural ordinary differential equation framework for modeling inelastic processes.arXiv preprint arXiv:2502.10633, 2025

    Reese E Jones and Jan N Fuhg. An attention-based neural ordinary differential equation framework for modeling inelastic processes.arXiv preprint arXiv:2502.10633, 2025

    Pith/arXiv arXiv 2025

Showing first 80 references.