Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Integrated Finite Element Neural Network (IFENN) for Phase-Field Fracture with Minimal Input and Generalized Geometry-Load Handling

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that one CNN trained on two load increments from a single-notch tension test can replace the phase-field solver for unseen geometries, multiple cracks, and new loading paths.

desk verdict A cheap spatial-only CNN surrogate for phase-field fracture that genuinely generalizes across geometries, but the two-snapshot training doesn't yet support the 'arbitrary loading paths' claim. read the letter →

arxiv 2505.19566 v1 pith:NWEHAJJZ submitted 2025-05-26 cs.CE

Panos Pantidis , Lampros Svolos , Diab Abueidda , Mostafa E. Mobasher This is my paper · ORCID
classification cs.CE MSC 74R1068T0765M60
keywords IFENNphase-fieldfracturephysics-informedconvolutionalneuralnetworkPICNNhybridFEM-MLsolverunsupervisedtrainingcracknucleationgeneralization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a single physics-informed convolutional network, trained for about five minutes on strain-energy-density maps from just two load increments of one single-notch tension simulation, can replace the phase-field PDE solver inside a staggered finite-element loop for a wide class of problems it never saw: new rectangular domains, more than one crack, different mesh resolutions, and different loading paths. The network is trained unsupervised, with the residual of the phase-field equation as its loss, and it uses no temporal features at all. If the claim holds, the expensive phase-field solve of a coupled fracture simulation is replaced by a cheap network prediction, while the equilibrium equation stays with FEM to keep the solution robust. A reader should care because this is a concrete recipe for making path-dependent fracture simulations substantially faster and more generalizable than existing hybrid FEM-ML approaches.

What carries the argument

The load-bearing mechanism is a physics-informed convolutional neural network whose kernels are made symmetric by construction: each $5\times5$ filter has only six independent learnable entries, forcing translational and rotational invariance in the learned $H\to\varphi$ map. The input is a pixel representation of the strain energy density $H$ on the Gauss-point grid, and the output is a pixel map of the phase-field $\varphi$, constrained to $[0,1]$ by a final sigmoid. The network loss is the $\ell^2$ norm of the residual of the phase-field PDE, with the Laplacian of $\varphi$ computed by a fixed nine-point finite-difference convolution filter. At inference, the network is embedded in a staggered IFENN loop in which FEM solves equilibrium, $H$ is capped, irreversibility is imposed on both $H$ and $\varphi$, and the predicted $\varphi$ is passed through a Gaussian smoothing filter before being projected back to the finite elements.

What would settle it

Keep the element size fixed but double the phase-field length scale $\ell_c$ in the single-notch tension problem and compare IFENN with FEM: because the paper reports the network preserves the $\ell_c/\ell_{\mathrm{elem}}$ ratio rather than $\ell_c$, the IFENN crack width would stay roughly unchanged while the FEM crack width doubles, directly showing the network did not learn the physical length scale.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the coupling between strain energy density $H$ and phase-field $\varphi$ only needs to be learned locally and spatially, near the fracture process zone, and that this single local map is enough to advance a crack through entire unseen simulations when embedded in the IFENN loop. The same PICNN, trained on increments 300 and 310 of a single-notch tension case, reproduces the FEM force-displacement response and crack paths for single- and double-notch specimens, symmetric and asymmetric crack configurations, rectangular domains, three different mesh densities, and loading histories with 350, 700, and 1500 increments. The authors also report that the network predicts the formation of a second crack it never saw, that it tends to preserve the characteristic-to-element length ratio rather than the absolute physical length scale, and that a Gaussian smoothing filter leaves a small residual stiffness at the end of the curves.

Load-bearing premise

The load-bearing premise is that the $H\to\varphi$ mapping learned from exactly two snapshots of one coarse single-notch simulation is stationary and sufficient for every later load increment and every unseen configuration, so that iterating the network hundreds of times inside IFENN does not drift.

Editorial extensions

If this is right

  • If the claim holds, a single five-minute training run is enough to build a phase-field fracture surrogate that can be reused across many unseen simulations, so the offline cost of hybrid modeling essentially disappears.
  • Because the network is fully convolutional and has no flattening layer, the same trained model can be evaluated on finer meshes and on arbitrary rectangular domains without retraining.
  • The removal of temporal features makes the surrogate indifferent to load incrementation, so the same network works with coarse or adaptive time stepping and even with changed load paths.
  • The observed tendency of the network to preserve the $\ell_c/\ell_{\mathrm{elem}}$ ratio rather than $\ell_c$ implies that crack widths predicted on finer meshes are slightly thinner; this is a limitation the paper identifies and leaves for future work.
  • The double-notch tests show the same network can nucleate and drive a second, unseen crack from the local $H$ field, indicating the local coupling carries enough information to trigger new damage sites.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An extension the paper does not test: instead of using the PICNN prediction as the final phase-field value, use it as an initial guess for one or two FEM corrections of the phase-field equation; the residual-loss formulation already produces residuals, so this could remove the smoothing-filter drift and residual stiffness at almost no extra cost.
  • The ratio-preserving behavior suggests a principled fix: train the PICNN on $H$-$\varphi$ maps from several $\ell_c/\ell_{\mathrm{elem}}$ values so the pixel width is decoupled from element size, which would give the network true physical length-scale awareness.
  • Because the method only needs a rectangular structured grid, it could probably be ported to any Cartesian-grid PDE system with a scalar history variable, such as gradient damage or plasticity localization, but that is a conjecture going beyond the paper.
  • The paper's own future-work list (damage nucleation in training, irregular geometries, unstructured meshes) defines the immediate boundary of the claimed generality; the two-snapshot training would need to be re-examined before those extensions.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces an IFENN (Integrated Finite Element Neural Network) variant for phase-field fracture in which the phase-field variable is computed by a physics-informed convolutional neural network (PICNN), while the equilibrium equation is still solved with FEM in a staggered loop. The PICNN is trained in an unsupervised manner by minimizing the residual of the phase-field PDE (Eqs. 15-16) on exactly two strain-energy-density profiles extracted from a coarse single-notch tension simulation (SNT1, increments 300 and 310). The authors then deploy the same trained PICNN inside IFENN across a range of test cases: three loading step sizes for the single-notch specimen, two finer meshes, a symmetric double-notch tension (SDNT) problem, and an asymmetric double-notch tension (ADNT) problem with rotated notches and a different rectangular aspect ratio. The central claim is that the local spatial H-to-phi mapping learned from these two snapshots is sufficient, without any temporal features, to model crack propagation across these unseen scenarios with excellent agreement to FEM.

Significance. If the central claims were fully supported, the contribution would be significant: training a convolutional surrogate for the phase-field subproblem in about five minutes and then reusing it across geometries, meshes, and loading-step schemes would address a real bottleneck in hybrid FEM-ML fracture modeling. The paper has several genuine strengths: the training is unsupervised with respect to FEM phase-field labels, using only the PDE residual; the network is fully convolutional, which permits variable input sizes; the double-symmetric kernel design is a principled way to embed translational/rotational invariance; and the force-displacement comparisons in Figs. 10, 12, 13, 14, 17, and 19 show qualitatively good agreement, including the challenging ADNT case with crack coalescence. The reported computational-time comparison (Fig. 15) also indicates increasing savings at finer meshes. However, the breadth of the generalization claims as stated in the abstract and introduction exceeds the evidence presented, and the absence of quantitative error metrics makes it difficult to assess the fidelity of the hybrid solver.

major comments (5)
  1. [Abstract and Section 5.1] The claim of 'arbitrary loading paths' is not supported by the evidence. The three SNT1 cases (I, II, III) differ only in the size and number of monotonic displacement increments; there is no unloading, cyclic, non-proportional, or otherwise non-monotonic loading. Please either add a genuinely non-monotonic loading test or restrict the claim to 'arbitrary monotonic loading step sizes' throughout the abstract, highlights, and Section 5.1.
  2. [Section 4 and Section 5] Training on exactly two H-profiles (increments 300 and 310) of one SNT1 simulation does not by itself constrain the H-to-phi mapping on the space of H fields encountered online, which includes multi-crack configurations, rotated cracks, and different mesh resolutions. The low residual at those two snapshots is not evidence that the learned operator is stationary or transferable; the paper asserts this stationarity as an inductive bias rather than demonstrating it. Please provide held-out validation, e.g., residual norms or phase-field errors on untrained increments of the same simulation and on the SDNT/ADNT configurations, and report quantitative error metrics (e.g., L2 errors in force-displacement curves and phase-field fields) in addition to the visual comparisons.
  3. [Section 3.3] The online stability of the method relies on several hand-tuned post-processing steps: capping H at 5e4 or 1e5, applying a Gaussian smoothing filter with k=5 and sigma=2, and enforcing irreversibility on both H and phi. These modifications change the effective learned mapping, and their parameters appear to be free choices. A sensitivity analysis with respect to these parameters is needed to establish that the reported agreement is not contingent on a narrow tuning of the post-processing; without it, the claim that the raw PICNN mapping is sufficient is weakened.
  4. [Section 5.1, mesh-refinement discussion] The paper itself notes in the SNT2/SNT3 study that the PICNN preserves the ratio lc/lelem rather than the physical length scale lc, leading to thinner cracks on finer meshes (Fig. 14b). This indicates that the network did not learn the actual phase-field length-scale behavior, which is a load-bearing aspect of the physics being modeled. The deviation should be quantified (e.g., crack-width or phi-profile errors at different mesh resolutions) and discussed more prominently as a limitation, since it directly affects the method's credibility for mesh-independent predictions.
  5. [Section 5.2 and Section 6] The highlights claim that the PICNN 'predicts the creation and nucleation of many cracks,' but the formulation explicitly excludes damage initiation and nucleation (Section 6 states this as future work), and all test cases start from pre-existing notches. The ADNT case also shows phase-field values stagnating at 0.7-0.75 after coalescence, preventing the reaction force from dropping to zero. Please temper the claims about nucleation and full crack formation, and discuss whether the stagnation after coalescence indicates a systematic limitation of the pixel-based local mapping in interactions between multiple crack fronts.
minor comments (5)
  1. [Throughout] Several figure captions are duplicated or mislabeled, e.g., 'Fig. 1: Comparison between FEM and IFENN...' appears over Figures 11, 12, 13, 17, and 19; these should be corrected to the actual figure numbers.
  2. [Section 5.2] The phrase 'predicts the formulation of both cracks' should be 'predicts the formation of both cracks'.
  3. [Section 3.2.4, Eq. (13)] The indices in Eq. (13) appear inconsistent: the y-direction finite difference terms should involve y_{i+1} and y_{i-1} or the notation should be clarified to avoid confusion with the x-index.
  4. [Abstract and Section 1.3] The wording 'arbitrary rectangular domains' is too broad given that all domains are axis-aligned rectangles with structured square meshes; please replace 'arbitrary' with 'varying rectangular' or similar.
  5. [Section 5.1] The discussion of the artificial stiffening caused by the Gaussian smoothing filter is placed after the results are presented; it would be clearer to state this known artifact earlier, when the smoothing filter is introduced in Section 3.3.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the PICNN is trained against the phase-field PDE residual and validated on FEM benchmarks not used as training labels.

full rationale

The derivation chain is not circular. The PICNN is trained by minimizing the PDE residual of Eq. 16, computed from Eq. 15, rather than by fitting to FEM phase-field labels from the benchmark cases. The two H-profiles used for training (increments 300 and 310 of the SNT1-I analysis) serve as inputs from one reference simulation, while the validation cases (SNT1-II/III, SNT2/3, SDNT, and ADNT) are solved with the same network and compared against FEM solutions that were not used in training. The self-citations (refs. 57–61) describe the prior IFENN framework and are contextual, not load-bearing; no central claim is justified by an author-supplied uniqueness theorem or by citing an unverified ansatz. The paper itself acknowledges that the two-snapshot sufficiency is an empirical finding and that post-hoc corrections (H capping, irreversibility enforcement on phi, and Gaussian smoothing) are needed for stability, which is a robustness limitation rather than a circularity. The broad generalization claim may be under-supported, but it does not reduce to its own inputs by construction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the phase-field AT2 PDE and on a strong empirical premise that two snapshots determine the full spatial H-to-phi map. No new physical entities are introduced. Several post-hoc corrections, H capping, phi irreversibility, Gaussian smoothing, and activation threshold, act as hand-tuned stabilizers and are counted as free parameters because the final accuracy depends on them.

free parameters (5)
  • H input cap value = 5e4 or 1e5
    Strain energy density is capped to a maximum before entering the PICNN; chosen by hand to keep inference within the training range, no principled rule given.
  • Gaussian smoothing kernel size and sigma = k=5, sigma=2
    Post-processing filter applied after every PICNN prediction to widen predicted crack profiles; tuned empirically against FEM behavior.
  • IFENN activation threshold = phi_nodal,max = 0.99
    The switch from FEM to IFENN occurs when the maximum nodal phase-field reaches 0.99; this threshold is selected ad hoc.
  • Network architecture hyperparameters = 4 layers, 24 channels, lr=1e-4, 10000 epochs
    Layer count, channel count, learning rate and epoch count are chosen without a systematic hyperparameter study.
  • Training snapshot indices = inc = [300, 310]
    Two load increments selected from the propagation stage of SNT1-I; no sensitivity analysis to this choice is reported.
assumptions (5)
  • domain assumption Phase-field AT2 model with hybrid isotropic/anisotropic split (Eq. 7) is the governing physics.
    The entire method is built on this PDE; no alternative fracture model is considered.
  • ad hoc to paper The spatial H-to-phi mapping can be learned from two static snapshots and used time-invariantly.
    Section 3.2.1 states this sufficiency as a proposal; it is the key modeling premise of the paper.
  • domain assumption Gauss-point values can be mapped one-to-one to pixel centers with negligible error.
    Section 3.2.2 acknowledges the Gauss-point and pixel-center mismatch is 'assumed to be negligible'.
  • ad hoc to paper Double-symmetric 5x5 convolutional kernels are expressive enough to capture arbitrary crack front orientations.
    Kernel symmetry is built in by construction in Section 3.2.3 to enforce invariance; ADNT tests with vertical and diagonal cracks show degradation, with phase-field stagnation around 0.7 to 0.75.
  • domain assumption Structured uniform square mesh and rectangular domains suffice for target applications.
    The pixel representation requires grid-aligned elements; non-rectangular and unstructured geometry is deferred to future work in Section 6.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Integrated Finite Element Neural Network (IFENN) for Phase-Field Fracture with Minimal Input and Generalized Geometry-Load Handling." pith.science (2026). https://pith.science/paper/NWEHAJJZ

@misc{pith2026250519566,
  author       = {Pith},
  title        = {Pith review of: Integrated Finite Element Neural Network (IFENN) for Phase-Field Fracture with Minimal Input and Generalized Geometry-Load Handling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NWEHAJJZ}},
  note         = {Machine review of arXiv:2505.19566}
}
read the original abstract

We present a novel formulation for modeling phase-field fracture propagation based on the Integrated Finite Element Neural Network (IFENN) framework. IFENN is a hybrid solver scheme that utilizes neural networks as PDE solvers within FEM, preserving accuracy via residual minimization while achieving speed-up via swift network predictions and reduction of the size of system of equations in coupled problems. In this work, we introduce a radically new formulation of IFENN in which the phase-field variable is calculated using physics-informed convolutional networks (PICNNs), while the equilibrium equation is still solved using FEM to maintain the solver robustness. Unlike conventional approaches, which rely on sequence or time-dependent models, we eliminate the need to include temporal features in the training setup and inference stage. Instead, we show that it is sufficient to learn only the spatial coupling between the strain energy density and the phase-field variable in the vicinity of the fracture process zone, and utilize this information along the advancing crack simulation. We train a single CNN in a purely physics-based, unsupervised manner on just two load increments from a single-notch tension problem, with a total training time of only 5 minutes. Following this exceptionally minimal and fast training, we show that the same PICNN can (when embedded within IFENN) model crack propagation in a very wide range of unseen scenarios, including arbitrarily rectangular domains, single and multiple interacting cracks, varying mesh densities, and arbitrary loading paths. The proposed formulation delivers breakthroughs that address many of the limitations in the existing literature of hybrid modeling, introducing a new paradigm for the development of generalizable, physics-consistent hybrid models that are applicable to fracture and other coupled problems.

Figures

Figures reproduced from arXiv: 2505.19566 by the authors.

Figure 1
Figure 1. Schematic of a generic domain Ω with a sample crack that is represented by the phase-field variable [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. IFENN workflow. a. Stage A: a swift FEM analysis on a simplified version of the target problem is performed, to generate the training data b. Stage B: a problem-specific neural network is selected and trained c. Stage C: the computationally expensive target problem is solved with the hybrid solver. of this stage is to generate the training data that are necessary in the subsequent network training. The nature and ty… view at source ↗
Figure 3
Figure 3. Schematic transformation of the strain energy density profile evaluated at the Gauss points to an equivalent pixel-based [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (15 more)
Figure 4
Figure 4. Figure 4: PICNN architecture and training process 3.2.4. PICNN training In order to construct the PICNN loss function, we first need to compute the Laplacian of ϕ at every pixel. We approach this task using the finite difference method, and below we elaborate on our approach whi…
Figure 5
Figure 5. Figure 5: a. Double-symmetric arrangement of 5 × 5 kernels used at each layer, resulting in 6 independent trainable entries per kernel. b. Schematic grid of pixels and the finite difference approximation using their centers as reference points. c. Laplacian convolution filters f…
Figure 6
Figure 6. Figure 6: Schematic overview of PICNN-IFENN for phase-field fracture. The PICNN is trained on an ultra-low amount of [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Geometric and loading/boundary condition details for the single-notch tension (SNT) problem [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Strain energy density profiles of the two load increments used in training. The graphs on the left depict Gauss point [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: a. Evolution of PICNN training loss function b. True, predicted and absolute error ϕ-maps for the two load increments used in training. training. The network is trained for 10000 epochs using the Adam optimizer with a learning rate lr = 10−4 . We highlight that the tot…
Figure 11
Figure 11. Figure 11: Comparison between FEM and IFENN for the SNT1 – I (350 increments): close-up 3D views of [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 1
Figure 1. Figure 1: Comparison between FEM and IFENN for the SNT – II (700 increments) problem. a. Vertical reaction force-displacement curves. b. Phase-field contours from FEM (top) and IFENN (bottom) at various load levels. IFENN activated a. b [PITH_FULL_IMAGE:figures/full_fig_p019_1.png]
Figure 1
Figure 1. Figure 1: Comparison between FEM and IFENN for the SNT – III (1500 increments) problem. a. Vertical reaction force-displacement curves. b. Phase-field contours from FEM (top) and IFENN (bottom) at various load levels. IFENN activated a. b [PITH_FULL_IMAGE:figures/full_fig_p020_1.png]
Figure 14
Figure 14. Figure 14: Comparison between FEM and IFENN for varying characteristic-to-element length ratios for the SNT problem. [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: Computational time for the phase-field variable as the ratio of characteristic length to element size increases. The [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 16
Figure 16. Figure 16: Geometric and loading/boundary condition details for the symmetric double-notch tension (SDNT) problem [PITH_FULL_IMAGE:figures/full_fig_p022_16.png]
Figure 1
Figure 1. Figure 1: Comparison between FEM and IFENN for the SDNT problem. a. Vertical reaction force-displacement curves. b. Phase-field contours from FEM (top) and IFENN (bottom) at various load levels. a. b. IFENN on [PITH_FULL_IMAGE:figures/full_fig_p023_1.png]
Figure 18
Figure 18. Figure 18: Geometric and loading/boundary condition details for the asymmetric double-notch tension (ADNT) problem [PITH_FULL_IMAGE:figures/full_fig_p023_18.png]
Figure 19
Figure 19. Figure 19: Comparison between FEM and IFENN for the ADNT problem. [PITH_FULL_IMAGE:figures/full_fig_p024_19.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. I-FENN with DeepONets: accelerating simulations in coupled multiphysics problems

    cs.CE 2025-08 conditional novelty 5.0 of 10

    A hybrid solver that replaces the coupled temperature/pressure equation with a trained operator network inside an FEM framework reduces compute by 35-43% while keeping field errors under 5% on unseen loads in thermoel...

Reference graph

Works this paper leans on

78 extracted references · 64 canonical work pages · cited by 1 Pith paper

  1. [1]

    Zhang, Z

    N. Zhang, Z. Luo, Z. Chen, F. Liu, P. Liu, W. Chen, L. Wu, L. Zhao, Thermal–hydraulic–mechanical– chemical coupled processes and their numerical simulation: a comprehensive review, Acta Geotechnica 18 (12) (2023) 6253–6274

    work page 2023

  2. [2]

    Li, Z.-Y

    K.-Q. Li, Z.-Y. Yin, State of the art of coupled thermo–hydro-mechanical–chemical modelling for frozen soils, Archives of Computational Methods in Engineering (2024) 1–58

    work page 2024

  3. [3]

    Sedmak, Computational fracture mechanics: An overview from early efforts to recent achievements, Fatigue & Fracture of Engineering Materials & Structures 41 (12) (2018) 2438–2474

    A. Sedmak, Computational fracture mechanics: An overview from early efforts to recent achievements, Fatigue & Fracture of Engineering Materials & Structures 41 (12) (2018) 2438–2474

    work page 2018

  4. [4]

    T. Rabczuk, Computational methods for fracture in brittle and quasi-brittle solids: state-of-the-art review and future perspectives, International Scholarly Research Notices 2013 (1) (2013) 849231

    work page 2013

  5. [5]

    Ambati, T

    M. Ambati, T. Gerasimov, L. De Lorenzis, A review on phase-field models of brittle fracture and a new fast hybrid formulation, Computational Mechanics 55 (2015) 383–405

    work page 2015

  6. [6]

    Diehl, R

    P. Diehl, R. Lipton, T. Wick, M. Tyagi, A comparative review of peridynamics and phase-field models for engineering fracture mechanics, Computational Mechanics 69 (6) (2022) 1259–1293

    work page 2022

  7. [7]

    Marigo, Revisiting brittle fracture as an energy minimization problem, Journal of the Mechanics and Physics of Solids 46 (8) (1998) 1319–1342

    J.-J. Marigo, Revisiting brittle fracture as an energy minimization problem, Journal of the Mechanics and Physics of Solids 46 (8) (1998) 1319–1342

    work page 1998

  8. [8]

    Bourdin, G

    B. Bourdin, G. A. Francfort, J.-J. Marigo, Numerical experiments in revisited brittle fracture, Journal of the Mechanics and Physics of Solids 48 (4) (2000) 797–826

    2000

Show all 78 references
  1. [9]

    A. A. Griffith, Vi. the phenomena of rupture and flow in solids, Philosophical transactions of the royal society of london. Series A, containing papers of a mathematical or physical character 221 (582-593) (1921) 163–198

    1921

  2. [10]

    H. Ren, X. Zhuang, C. Anitescu, T. Rabczuk, An explicit phase field method for brittle dynamic fracture, Computers & Structures 217 (2019) 45–56

    2019

  3. [11]

    D. Chu, X. Li, Z. Liu, Study the dynamic crack path in brittle material under thermal shock loading by phase field modeling, International Journal of Fracture 208 (2017) 115–130

    2017

  4. [12]

    Svolos, C

    L. Svolos, C. A. Bronkhorst, H. Waisman, Thermal-conductivity degradation across cracks in coupled thermo-mechanical systems modeled by the phase-field fracture method, Journal of the Mechanics and Physics of Solids 137 (2020) 103861

    2020

  5. [13]

    Zhou, C.-C

    S.-W. Zhou, C.-C. Xia, Propagation and coalescence of quasi-static cracks in brazilian disks: an insight from a phase field model, Acta Geotechnica 14 (4) (2019) 1195–1214

    2019

  6. [14]

    Spatschek, E

    R. Spatschek, E. Brener, A. Karma, Phase field modeling of crack propagation, Philosophical Magazine 91 (1) (2011) 75–95. 25

    2011

  7. [15]

    P. K. Kristensen, C. F. Niordson, E. Mart ´ ınez-Pa˜ neda, Applications of phase field fracture in modelling hydrogen assisted failures, Theoretical and Applied Fracture Mechanics 110 (2020) 102837

    2020

  8. [16]

    Bleyer, C

    J. Bleyer, C. Roux-Langlois, J.-F. Molinari, Dynamic crack propagation with a variational phase-field model: limiting speed, crack branching and velocity-toughening mechanisms, International Journal of Fracture 204 (1) (2017) 79–100

    2017

  9. [17]

    P. Li, W. Li, B. Li, S. Yang, Y. Shen, Q. Wang, K. Zhou, A review on phase field models for fracture and fatigue, Engineering Fracture Mechanics 289 (2023) 109419

    2023

  10. [18]

    Svolos, H

    L. Svolos, H. M. Mourad, G. Manzini, K. Garikipati, A fourth-order phase-field fracture model: Formulation and numerical solution using a continuous/discontinuous galerkin method, Journal of the Mechanics and Physics of Solids 165 (2022) 104910

    2022

  11. [19]

    Schapira, L

    Y. Schapira, L. Radtke, S. Kollmannsberger, A. D¨ uster, Performance of acceleration techniques for staggered phase-field solutions, Computer methods in applied mechanics and engineering 410 (2023) 116029

    2023

  12. [20]

    Zhuang, S

    X. Zhuang, S. Zhou, G. Huynh, P. Areias, T. Rabczuk, Phase field modeling and computer implementation: A review, Engineering Fracture Mechanics 262 (2022) 108234

    2022

  13. [21]

    T. Wick, Modified newton methods for solving fully monolithic phase-field quasi-static brittle fracture propagation, Computer Methods in Applied Mechanics and Engineering 325 (2017) 577–611

    2017

  14. [22]

    P. K. Kristensen, E. Mart ´ ınez-Pa˜ neda, Phase field fracture modelling using quasi-newton methods and a new adaptive step scheme, Theoretical and Applied Fracture Mechanics 107 (2020) 102446

    2020

  15. [23]

    Hofacker, C

    M. Hofacker, C. Miehe, Continuum phase field modeling of dynamic fracture: variational principles and staggered fe implementation, International journal of fracture 178 (2012) 113–129

    2012

  16. [24]

    T. Jin, Z. Li, K. Chen, A novel phase-field monolithic scheme for brittle crack propagation based on the limited-memory bfgs method with adaptive mesh refinement, International Journal for Numerical Methods in Engineering 125 (22) (2024) e7572

    2024

  17. [25]

    Bharali, S

    R. Bharali, S. Goswami, C. Anitescu, T. Rabczuk, A robust monolithic solver for phase-field fracture integrated with fracture energy based arc-length method and under-relaxation, Computer Methods in Applied Mechanics and Engineering 394 (2022) 114927

    2022

  18. [26]

    K. Li, Y. Mehmani, A multiscale preconditioner for crack evolution in porous microstructures: Accelerating phase-field methods, International Journal for Numerical Methods in Engineering 125 (11) (2024) e7463

    2024

  19. [27]

    Badri, G

    M. Badri, G. Rastiello, E. Foerster, Preconditioning strategies for vectorial finite element linear systems arising from phase-field models for fracture mechanics, Computer Methods in Applied Mechanics and Engineering 373 (2021) 113472. 26

    2021

  20. [28]

    Svolos, L

    L. Svolos, L. Berger-Vergiat, H. Waisman, Updating strategy of a domain decomposition preconditioner for parallel solution of dynamic fracture problems, Journal of Computational Physics 422 (2020) 109746

    2020

  21. [29]

    Rannou, C

    J. Rannou, C. Bovet, Domain decomposition methods and acceleration techniques for the phase field fracture staggered solver, International Journal for Numerical Methods in Engineering 125 (17) (2024) e7544

    2024

  22. [30]

    Gupta, U

    A. Gupta, U. M. Krishnan, T. K. Mandal, R. Chowdhury, V. P. Nguyen, An adaptive mesh refinement algorithm for phase-field fracture models: Application to brittle, cohesive, and dynamic fracture, Computer Methods in Applied Mechanics and Engineering 399 (2022) 115347

    2022

  23. [31]

    X. Yang, N. Guo, Z. Yang, A finite-volume implementation of the phase-field model for brittle fracture with adaptive mesh refinement, Computers and Geotechnics 165 (2024) 105921

    2024

  24. [32]

    Saberi, H

    H. Saberi, H. Saberi, T. Q. Bui, Y. Heider, M. N. Nguyen, A multi-level adaptive mesh refinement method for phase-field fracture problems, Theoretical and Applied Fracture Mechanics 125 (2023) 103920

    2023

  25. [33]

    Agrawal, B

    V. Agrawal, B. Runnels, Block structured adaptive mesh refinement and strong form elasticity approach to phase field fracture with applications to delamination, crack branching and crack deflection, Computer Methods in Applied Mechanics and Engineering 385 (2021) 114011

    2021

  26. [34]

    Jaccon, B

    A. Jaccon, B. Prabel, G. Moln´ ar, J. Bluth´ e, A. Gravouil, Adaptive mesh refinement and cycle jumps for phase-field fatigue fracture modeling, Finite Elements in Analysis and Design 224 (2023) 104004

    2023

  27. [35]

    Herrmann, S

    L. Herrmann, S. Kollmannsberger, Deep learning in computational mechanics: a review, Computational Mechanics 74 (2) (2024) 281–331

    2024

  28. [36]

    H. Jin, E. Zhang, H. D. Espinosa, Recent advances and applications of machine learning in experimental solid mechanics: A review, Applied Mechanics Reviews 75 (6) (2023) 061001

    2023

  29. [37]

    N. N. Vlassis, R. Ma, W. Sun, Geometric deep learning for computational mechanics part i: Anisotropic hyperelasticity, Computer Methods in Applied Mechanics and Engineering 371 (2020) 113299

    2020

  30. [38]

    D. W. Abueidda, S. Koric, R. A. Al-Rub, C. M. Parrott, K. A. James, N. A. Sobh, A deep learning energy method for hyperelasticity and viscoelasticity, European Journal of Mechanics-A/Solids 95 (2022) 104639

    2022

  31. [39]

    Mendizabal, P

    A. Mendizabal, P. M´ arquez-Neila, S. Cotin, Simulation of hyperelastic materials in real-time using deep learning, Medical image analysis 59 (2020) 101569

    2020

  32. [40]

    Huang, J

    D. Huang, J. N. Fuhg, C. Weißenfels, P. Wriggers, A machine learning based plasticity model using proper orthogonal decomposition, Computer Methods in Applied Mechanics and Engineering 365 (2020) 113008

    2020

  33. [41]

    D. P. Jang, P. Fazily, J. W. Yoon, Machine learning-based constitutive model for j2-plasticity, International Journal of Plasticity 138 (2021) 102919. 27

    2021

  34. [42]

    J. N. Fuhg, L. van Wees, M. Obstalecki, P. Shade, N. Bouklas, M. Kasemer, Machine-learning convex and texture-dependent macroscopic yield from crystal plasticity simulations, Materialia 23 (2022) 101446

    2022

  35. [43]

    Goswami, C

    S. Goswami, C. Anitescu, S. Chakraborty, T. Rabczuk, Transfer learning enhanced physics informed neural network for phase-field modeling of fracture, Theoretical and Applied Fracture Mechanics 106 (2020) 102447

    2020

  36. [44]

    Zheng, T

    B. Zheng, T. Li, H. Qi, L. Gao, X. Liu, L. Yuan, Physics-informed machine learning model for computational fracture of quasi-brittle materials without labelled data, International Journal of Mechanical Sciences 223 (2022) 107282

    2022

  37. [45]

    Manav, R

    M. Manav, R. Molinaro, S. Mishra, L. De Lorenzis, Phase-field modeling of fracture with physics-informed deep learning, Computer Methods in Applied Mechanics and Engineering 429 (2024) 117104

    2024

  38. [46]

    L. Lu, P. Jin, G. Pang, Z. Zhang, G. E. Karniadakis, Learning nonlinear operators via deeponet based on the universal approximation theorem of operators, Nature machine intelligence 3 (3) (2021) 218–229

    2021

  39. [47]

    D. W. Abueidda, P. Pantidis, M. E. Mobasher, Deepokan: Deep operator network based on kolmogorov arnold networks for mechanics problems, Computer Methods in Applied Mechanics and Engineering 436 (2025) 117699

    2025

  40. [48]

    Goswami, M

    S. Goswami, M. Yin, Y. Yu, G. E. Karniadakis, A physics-informed variational deeponet for predicting crack path in quasi-brittle materials, Computer Methods in Applied Mechanics and Engineering 391 (2022) 114587

    2022

  41. [49]

    Kiyani, M

    E. Kiyani, M. Manav, N. Kadivar, L. De Lorenzis, G. E. Karniadakis, Predicting crack nucleation and propagation in brittle materials using deep operator networks with diverse trunk architectures, Computer Methods in Applied Mechanics and Engineering 441 (2025) 117984

    2025

  42. [50]

    Aldakheel, E

    F. Aldakheel, E. S. Elsayed, Y. Heider, O. Weeger, Physics-based machine learning for computational fracture mechanics, Machine Learning for Computational Science and Engineering 1 (1) (2025) 18

    2025

  43. [51]

    D. K. Klein, R. Ortigosa, J. Mart ´ ınez-Frutos, O. Weeger, Finite electro-elasticity with physics-augmented neural networks, Computer Methods in Applied Mechanics and Engineering 400 (2022) 115501

    2022

  44. [52]

    Dammaß, K

    F. Dammaß, K. A. Kalina, M. K¨ astner, Neural networks meet phase-field: A hybrid fracture model, Computer Methods in Applied Mechanics and Engineering 440 (2025) 117937

    2025

  45. [53]

    S. Thel, L. Greve, B. van de Weg, P. van der Smagt, Introducing finite element method integrated networks (femin), Computer Methods in Applied Mechanics and Engineering 427 (2024) 117073

    2024

  46. [54]

    S. K. Mitusch, S. W. Funke, M. Kuchta, Hybrid fem-nn models: Combining artificial neural networks with the finite element method, Journal of Computational Physics 446 (2021) 110651. 28

    2021

  47. [55]

    R. E. Meethal, A. Kodakkal, M. Khalil, A. Ghantasala, B. Obst, K.-U. Bletzinger, R. W¨ uchner, Finite element method-enhanced neural network for forward and inverse problems, Advanced modeling and simulation in engineering sciences 10 (1) (2023) 6

    2023

  48. [56]

    Zhang, D

    W. Zhang, D. S. Li, T. Bui-Thanh, M. S. Sacks, Simulation of the 3d hyperelastic behavior of ventricular myocardium using a finite-element based neural-network approach, Computer methods in applied mechanics and engineering 394 (2022) 114871

    2022

  49. [57]

    Pantidis, M

    P. Pantidis, M. E. Mobasher, Integrated finite element neural network (I-FENN) for non-local continuum damage mechanics, Computer Methods in Applied Mechanics and Engineering 404 (2023) 115766

    2023

  50. [58]

    Pantidis, H

    P. Pantidis, H. Eldababy, C. M. Tagle, M. E. Mobasher, Error convergence and engineering-guided hyperparameter search of pinns: Towards optimized I-FENN performance, Computer Methods in Applied Mechanics and Engineering 414 (2023) 116160

    2023

  51. [59]

    Pantidis, H

    P. Pantidis, H. Eldababy, D. Abueidda, M. E. Mobasher, I-FENN with temporal convolutional networks: Expediting the load-history analysis of non-local gradient damage propagation, Computer Methods in Applied Mechanics and Engineering 425 (2024) 116940

    2024

  52. [60]

    D. W. Abueidda, M. E. Mobasher, Variational temporal convolutional networks for I-FENN thermoelasticity, Computer Methods in Applied Mechanics and Engineering 429 (2024) 117122

    2024

  53. [61]

    D. W. Abueidda, M. E. Mobasher, I-FENN for thermoelasticity based on physics-informed temporal convolutional network (pi-tcn), Computational Mechanics 74 (6) (2024) 1229–1259

    2024

  54. [62]

    K. Pham, H. Amor, J.-J. Marigo, C. Maurini, Gradient damage models and their use to approximate brittle fracture, International Journal of Damage Mechanics 20 (4) (2011) 618–652

    2011

  55. [63]

    Miehe, F

    C. Miehe, F. Welschinger, M. Hofacker, Thermodynamically consistent phase-field models of fracture: Variational principles and multi-field fe implementations, International journal for numerical methods in engineering 83 (10) (2010) 1273–1311

    2010

  56. [64]

    N. P. van Dijk, J. J. Espadas-Escalante, P. Isaksson, Strain energy density decompositions in phase-field fracture theories for orthotropy and anisotropy, International Journal of Solids and Structures 196 (2020) 140–153

    2020

  57. [65]

    Vicentini, C

    F. Vicentini, C. Zolesi, P. Carrara, C. Maurini, L. De Lorenzis, On the energy decomposition in variational phase-field models for brittle fracture under multi-axial stress states, International Journal of Fracture 247 (3) (2024) 291–317

    2024

  58. [66]

    C. Kuhn, A. Schl¨ uter, R. M¨ uller, On degradation functions in phase field fracture models, Computational Materials Science 108 (2015) 374–384. 29

    2015

  59. [67]

    J. M. Sargado, E. Keilegavlen, I. Berre, J. M. Nordbotten, High-accuracy phase-field models for brittle fracture based on a new family of degradation functions, Journal of the Mechanics and Physics of Solids 111 (2018) 458–489

    2018

  60. [68]

    Svolos, J

    L. Svolos, J. N. Plohr, G. Manzini, H. M. Mourad, On the convexity of phase-field fracture formulations: Analytical study and comparison of various degradation functions, International Journal of Non-Linear Mechanics 150 (2023) 104359

    2023

  61. [69]

    Miehe, M

    C. Miehe, M. Hofacker, F. Welschinger, A phase field model for rate-independent crack propagation: Robust algorithmic implementation based on operator splits, Computer Methods in Applied Mechanics and Engineering 199 (45-48) (2010) 2765–2778

    2010

  62. [70]

    I. A. Baratta, J. P. Dean, J. S. Dokken, M. Habera, J. HALE, C. N. Richardson, M. E. Rognes, M. W. Scroggs, N. Sime, G. N. Wells, Dolfinx: the next generation fenics problem solving environment (2023)

    2023

  63. [71]

    M. S. Alnæs, A. Logg, K. B. Ølgaard, M. E. Rognes, G. N. Wells, Unified form language: A domain-specific language for weak formulations of partial differential equations, ACM Transactions on Mathematical Software (TOMS) 40 (2) (2014) 1–37

    2014

  64. [72]

    Archana, P

    R. Archana, P. E. Jeevaraj, Deep learning models for digital image processing: a review, Artificial Intelligence Review 57 (1) (2024) 11

    2024

  65. [73]

    Nanni, G

    L. Nanni, G. Maguolo, S. Brahnam, M. Paci, An ensemble of convolutional neural networks for audio classification, Applied Sciences 11 (13) (2021) 5796

    2021

  66. [74]

    Alsobhani, H

    A. Alsobhani, H. M. ALabboodi, H. Mahdi, Speech recognition using convolution deep neural networks, in: Journal of Physics: Conference Series, Vol. 1973, IOP Publishing, 2021, p. 012166

    1973

  67. [75]

    N. S.-N. Lam, Spatial interpolation methods: a review, The American Cartographer 10 (2) (1983) 129–150

    1983

  68. [76]

    M. L. Minion, A projection method for locally refined grids, Journal of Computational Physics 127 (1) (1996) 158–178

    1996

  69. [77]

    R. S. Teegavarapu, T. Meskele, C. S. Pathak, Geo-spatial grid-based transformations of precipitation estimates using spatial interpolation methods, Computers & Geosciences 40 (2012) 28–39

    2012

  70. [78]

    X. Zhao, Z. Gong, Y. Zhang, W. Yao, X. Chen, Physics-informed convolutional neural networks for temperature field prediction of heat source layout without labeled data, Engineering Applications of Artificial Intelligence 117 (2023) 105516. 30

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.