Pith. sign in

REVIEW 3 major objections 5 minor 72 references

A Foundation Model for Material Fracture Prediction

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A single transformer predicts material fracture across simulators, materials, and tasks.

desk verdict A credible foundation-model idea with strong in-distribution results, but the paper's own Table S3 contradicts the headline claim of generalizing to crack-density variations. read the letter →

arxiv 2507.23077 v1 pith:AM2I7CDG submitted 2025-07-30 cs.LG cond-mat.mtrl-sciphysics.geo-ph

Agnese Marcato , Aleksandra Pachalieva , Ryley G. Hill , Kai Gao , Xiaoyu Wang , Esteban Rougier , Zhou Lei , Vinamra Agrawal
show 9 more authors
Janel Chua Qinjun Kang Jeffrey D. Hyman Abigail Hunter Nathan DeBardeleben Earl Lawrence Hari Viswanathan Daniel O'Malley Javier E. Santos
This is my paper · ORCID
classification cs.LGcond-mat.mtrl-sciphysics.geo-ph
keywords fracturepredictionfoundationmodeltransformermultimodalarchitectureLLMembeddingsphase-fieldsimulationfinite-discreteelementmethodtransferlearning
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

The paper argues that fracture prediction, long fragmented into simulator-specific pipelines, can be unified in a single foundation model. It presents a transformer-based architecture that takes a spatial fracture field, on structured or unstructured meshes, together with language-model embeddings of the simulation text input deck, and predicts failure patterns. Pretraining first on an inexpensive rule-based surrogate before fine-tuning on phase-field simulations cuts validation error by more than 70 percent in low-data regimes, and the same pretrained model fine-tunes to new materials with as few as one sample, to time-to-failure prediction, to temporal fracture evolution, and to unstructured-mesh finite-discrete-element simulations. If the approach holds, a practitioner could replace simulator-specific screening sweeps with a single adaptable model that runs on one GPU.

What carries the argument

The load-bearing mechanism is an encoder-decoder transformer with two fused input channels. Spatial data, such as Cartesian grid nodes or unstructured mesh edge centers, are flattened into token sequences with sine-cosine positional encodings and cross-attended to a fixed set of learnable latent tokens, producing a size-invariant representation; a pretrained large language model converts the textual input deck, containing material, boundary conditions, solver, and target, into context vectors that are fused with the latent tokens. The second piece of machinery is the multi-fidelity curriculum: pretraining on an on-the-fly rule-based surrogate that grows orthogonal cracks until a percolating path forms, then fine-tuning on phase-field damage fields, teaches generic topological operations like crack growth and coalescence before the model sees expensive high-fidelity data. The decoder then branches into task-specific heads for spatial failure fields, temporal trajectories, or a scalar time-to-failure, all without architectural changes.

What would settle it

A decisive check would be to take the pretrained model and evaluate it on phase-field simulations with curved, branched, or non-percolating crack topologies, which the surrogate never generates; if validation loss jumps far above the reported 0.003 while in-distribution performance stays flat, the surrogate-to-physics transfer is doing less work than claimed. The reported low-fracture-density mean absolute errors of 0.59 to 1.33 in Table S3 are already such a case, since sparse single cracks fall outside the training range of 6 to 30 initial cracks.

Watch

Extended reading notes

Core claim

The central claim is that a single multimodal transformer, trained on a mixture of cheap rule-based crack-growth data and high-fidelity phase-field simulations, learns a transferable representation of fracture that is not tied to one simulator, mesh type, material, or task. On the full two-million-simulation phase-field pretraining the model reaches a validation loss of 0.003; fine-tuned, it predicts time-to-failure with an R-squared of 0.9522, an improvement of about 30 percent over the prior baseline, and adapts to unstructured-mesh combined finite-discrete element simulations with a mean absolute error of 0.0218. The model also generalizes to unseen materials such as titanium with a single fine-tuning sample, and to unseen initial crack configurations, which the authors attribute to the transformer's global receptive field.

Load-bearing premise

The data-efficiency and cross-simulator story depends on the assumption that the cheap rule-based surrogate's fracture patterns, straight orthogonal cracks growing until they connect, are representative enough of real phase-field and finite-discrete-element fracture fields that representations learned on them transfer; the paper's own out-of-distribution low-density errors, roughly ten times larger than in-distribution ones, show one regime where that assumption already strains.

Editorial extensions

If this is right

  • Pretraining on cheap rule-based data lets the model match the performance of models trained on roughly ten times more high-fidelity data, cutting data requirements for fracture machine learning by an order of magnitude.
  • A new material can be added by naming it in the input deck and fine-tuning only the decoder for about three minutes, with a single sample sufficing for titanium.
  • The same backbone predicts time-to-failure with an R-squared of 0.9522 and can interpolate unseen intermediate states along the fracture trajectory.
  • Fracture prediction transfers from Cartesian grids to unstructured meshes, so expensive finite-discrete-element screening sweeps can be replaced by single-GPU inference.
  • Because the architecture is mesh-agnostic and context is injected through text, the same pretrained model can be extended to new simulators, geometries, and quantities of interest without redesign.

Reading between the lines

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

  • If the surrogate-pretraining strategy transfers broadly, the same recipe of cheap proxy pretraining followed by physics fine-tuning could be applied to other damage and PDE-governed problems where high-fidelity data is scarce, such as fatigue or corrosion.
  • The input-deck language channel implies that model behavior can be steered at inference time by editing the text description, which offers a cheap way to probe which material and boundary-condition attributes the model actually relies on.
  • The paper's own limitation, that training covers only extensional loading, suggests a concrete next test: fine-tune on shear- or compression-dominated finite-discrete-element data and measure whether the learned tensile-fracture representation helps or hinders.
  • One-sample fine-tuning worked for titanium partly because its failure behavior overlaps with pretraining materials; concrete required about 20 samples, so the few-shot claim is material-dependent and should be characterized per material before relying on it.
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

3 major / 5 minor

Summary. This manuscript proposes a multimodal transformer-based 'foundation model' for predicting fracture propagation in two-dimensional simulations. The encoder maps structured or unstructured spatial meshes via cross-attention to latent tokens; LLaMA-3.1 embeddings of templated input decks provide material, loading, and solver context; a task-specific decoder outputs final fracture fields, temporal trajectories, or time-to-failure scalars. Training uses a two-stage curriculum: on-the-fly rule-based surrogate data followed by phase-field simulations across five materials and axial/biaxial loading. The authors report strong in-distribution accuracy (validation loss 0.003, mean MAE 0.031), time-to-failure R-squared of 0.9522, successful HOSS unstructured-mesh fine-tuning (MAE 0.0218), and one-sample fine-tuning to titanium. They also claim out-of-distribution generalization to unseen initial fracture patterns, new materials, and new mesh types.

Significance. If the out-of-distribution and data-efficiency claims held as stated, this would be a useful step toward general-purpose simulation surrogates for fracture mechanics. The strongest features are the concrete engineering results: the multi-fidelity curriculum with on-the-fly surrogate pretraining, the cross-attention encoder that natively accepts unstructured meshes, and the use of frozen LLM embeddings to condition on textual metadata. These are genuine contributions, and the public dataset release is a valuable asset. However, the paper's central 'generalization across initial fracture patterns' claim is contradicted by its own Table S3, so the significance currently rests on in-distribution and narrow-transfer evidence rather than on the broad out-of-distribution capabilities advertised in Section 2 and the abstract.

major comments (3)
  1. [Section 2, Table S3] The claim that the model generalizes to 'variations in crack density' is contradicted by Table S3. For low fracture density, the MAEs are 0.209 to 0.590 (biaxial) and 0.232 to 1.332 (axial) across materials, versus a mean of 0.031 for training-like configurations, i.e., roughly 7 to 40 times larger. More seriously, the steel/axial entry of 1.332 exceeds the maximum possible L1 error for fields confined to [0,1] as stated in Section S6.2, which indicates either unbounded raw outputs or an inconsistent error computation. The authors should clip outputs to the physical range, recompute Table S3, and either support the out-of-distribution claim with corrected numbers or remove the crack-density bullet from Section 2.
  2. [Section 2 and Section S6.2] There is a dataset-size discrepancy that affects the scaling claims. Section 2 states that pretraining used a 'full dataset of 2 million phase-field simulations' and Section S3 repeats a value of 2,000,000, while Section S6.2 reports 'over 400,000 simulations' for the phase-field dataset. The seven-million-step, 64-H100 pretraining cost is justified by dataset scale, so this discrepancy must be reconciled. If the 2M figure includes on-the-fly surrogate samples, the text should say so; if the phase-field count is 400k, the scaling and data-efficiency claims should be re-expressed accordingly.
  3. [Section 3.1, Figure 5] The abstract claims that the model 'dramatically reduc[es] data needs compared to standard ML,' but the one-sample titanium and concrete fine-tuning experiments are not compared against any standard ML baseline or a from-scratch transformer trained on the same samples. Stabilization of MAE after one titanium sample, which the authors partly attribute to overlap with pretraining materials, does not by itself quantify a data-efficiency advantage relative to conventional training. Add such a comparison or soften the claim.
minor comments (5)
  1. [Figure 6] The figure caption labels interpolated predictions as 'red,' while the main text in Section 3.2 describes them as 'orange'; please harmonize the color naming.
  2. [Abstract and Section 3.3] The abstract's 'less than 10% mean relative error' for time-to-failure is not stated in Section 3.3; report the relative-error value or remove the claim from the abstract.
  3. [Table S3] Please define 'low fracture density' and clarify whether it corresponds to the 'single initial fracture' case described in Section 2, since the training distribution is 6 to 30 initial fractures.
  4. [Equation (1)] At first use of the L1 metric, specify the domain of y_i and y_hat_i (e.g., both confined to [0,1] for normalized fields), because these bounds are needed to interpret the reported L1 values.
  5. [Section 3.3] The statement of a 'performance increase of 30% [38]' should specify the metric (R2, MAE, or relative error) and the comparison protocol, since the cited work may use a different error definition.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: all central metrics are held-out evaluations, and no load-bearing claim reduces by construction to a fitted input or self-citation.

full rationale

This is an empirical machine-learning study; the central quantities (validation loss 0.003, time-to-failure R2 of 0.9522, and HOSS unstructured-mesh MAE 0.0218) are measured on held-out data from the same simulators used for training, and the small-sample adaptation results are genuine fine-tuning evaluations rather than parameters renamed as predictions. The rule-based surrogate pretraining is compared directly against phase-field-only training in Figure 2 and is an independent data-generation method, not a fitted input. No load-bearing claim rests on a self-citation: reference [38] is used only as a baseline to compare against, and the HOSS simulator citation [56] is external tooling rather than an assumed conclusion. The possible concern from Table S3 (low crack-density MAEs of 0.209 to 1.332, including values exceeding the [0,1] field range) is a generalization and correctness issue, not a circular derivation, so it does not raise the circularity score.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced. The rule-based surrogate is a data generator, not a physical claim about the world, and the LLM embeddings are a representational tool, so neither counts as an invented physical entity. No free parameters enter a derived law; the model weights are learned and the architecture size is chosen from a scaling study.

assumptions (4)
  • domain assumption Ground-truth fracture behavior is represented by the phase-field equations (S1)-(S3) with the material parameters in Table S5.
    The high-fidelity training and test labels are generated by this solver; the model inherits its limitations, including small-strain kinematics and extensional loading.
  • ad hoc to paper The rule-based surrogate (S6.1) produces fracture topologies representative of the phase-field and FDEM solutions.
    Section 2 relies on this for pretraining transfer; Table S3 low-density errors suggest the overlap is imperfect.
  • ad hoc to paper LLaMA-3.1 text embeddings of the templated input decks provide sufficient material and loading information for generalization to new materials.
    This is used in the textual input deck encoder (Methods); the one-shot titanium transfer depends on it.
  • domain assumption Final fracture patterns are deterministic functions of the initial crack configuration, material, and boundary conditions.
    The model is trained to map initial state to final state; stochasticity or path dependence in the simulators is ignored.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Foundation Model for Material Fracture Prediction." pith.science (2026). https://pith.science/paper/AM2I7CDG

@misc{pith2026250723077,
  author       = {Pith},
  title        = {Pith review of: A Foundation Model for Material Fracture Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AM2I7CDG}},
  note         = {Machine review of arXiv:2507.23077}
}
read the original abstract

Accurately predicting when and how materials fail is critical to designing safe, reliable structures, mechanical systems, and engineered components that operate under stress. Yet, fracture behavior remains difficult to model across the diversity of materials, geometries, and loading conditions in real-world applications. While machine learning (ML) methods show promise, most models are trained on narrow datasets, lack robustness, and struggle to generalize. Meanwhile, physics-based simulators offer high-fidelity predictions but are fragmented across specialized methods and require substantial high-performance computing resources to explore the input space. To address these limitations, we present a data-driven foundation model for fracture prediction, a transformer-based architecture that operates across simulators, a wide range of materials (including plastic-bonded explosives, steel, aluminum, shale, and tungsten), and diverse loading conditions. The model supports both structured and unstructured meshes, combining them with large language model embeddings of textual input decks specifying material properties, boundary conditions, and solver settings. This multimodal input design enables flexible adaptation across simulation scenarios without changes to the model architecture. The trained model can be fine-tuned with minimal data on diverse downstream tasks, including time-to-failure estimation, modeling fracture evolution, and adapting to combined finite-discrete element method simulations. It also generalizes to unseen materials such as titanium and concrete, requiring as few as a single sample, dramatically reducing data needs compared to standard ML. Our results show that fracture prediction can be unified under a single model architecture, offering a scalable, extensible alternative to simulator-specific workflows.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 65 canonical work pages

  1. [1]

    Case Studies in Construction Materials, 03848 (2024)

    Qu, Z., Zhang, Y., Liu, Z., Si, R., Wu, J.: A review on early-age cracking of concrete: Causes and control. Case Studies in Construction Materials, 03848 (2024)

    work page 2024

  2. [2]

    Engineering Failure Analysis 120, 105013 (2021)

    Niazi, H., Eadie, R., Chen, W., Zhang, H.: High ph stress corrosion cracking initiation and crack evolution in buried steel pipelines: A review. Engineering Failure Analysis 120, 105013 (2021)

    work page 2021

  3. [3]

    : Understanding hydraulic fracturing: a multi-scale problem

    Hyman, J., Jim´ enez-Mart ´ ınez, J., Viswanathan, H.S., Carey, J.W., Porter, M.L., Rougier, E., Karra, S., Kang, Q., Frash, L., Chen, L., et al. : Understanding hydraulic fracturing: a multi-scale problem. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 374(2078), 20150426 (2016) 28

    work page 2016

  4. [4]

    Advanced Materials 21(20), 2103–2110 (2009)

    Launey, M.E., Ritchie, R.O.: On the fracture toughness of advanced materials. Advanced Materials 21(20), 2103–2110 (2009)

    work page 2009

  5. [5]

    Engineering fracture mechanics 23(1), 289–298 (1986)

    Kalthoff, J.: Fracture behavior under high rates of loading. Engineering fracture mechanics 23(1), 289–298 (1986)

    work page 1986

  6. [6]

    Journal of the Mechanics and Physics of Solids 48(1), 175–209 (2000)

    Silling, S.A.: Reformulation of elasticity theory for discontinuities and long-range forces. Journal of the Mechanics and Physics of Solids 48(1), 175–209 (2000)

    work page 2000

  7. [7]

    Computer Physics Communications 87(1-2), 236–252 (1995)

    Sulsky, D., Zhou, S.J., Schreyer, H.L.: Application of a particle-in-cell method to solid mechanics. Computer Physics Communications 87(1-2), 236–252 (1995)

    work page 1995

  8. [8]

    International Journal for Numerical Methods in Engineering 45(5), 601–620 (1999)

    Belytschko, T., Black, T.: Elastic crack growth in finite elements with minimal remeshing. International Journal for Numerical Methods in Engineering 45(5), 601–620 (1999)

    work page 1999

Show all 72 references
  1. [9]

    Journal of the Mechanics and Physics of Solids 48(4), 797–826 (2000)

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

    2000

  2. [10]

    International Journal for Numerical Methods in Engineering 44(9), 1267–1282 (1999)

    Ortiz, M., Pandolfi, A.: Finite-deformation irreversible cohesive elements for three-dimensional crack analysis. International Journal for Numerical Methods in Engineering 44(9), 1267–1282 (1999)

    1999

  3. [11]

    In: AIAA SCITECH 2023 Forum (2023)

    Scoggins, J.B., Wignall, T.J., Nakamura-Zimmerer, T., Bibb, K.: Multihierarchy gaussian process models for probabilistic aerodynamic databases using uncertain nominal and off-nominal configuration data. In: AIAA SCITECH 2023 Forum (2023)

    2023

  4. [12]

    Physics of Fluids 36(5), 056116 (2024)

    Tao, G., Fan, C., Wang, W., Guo, W., Cui, J.: Multi-fidelity deep learning for aerodynamic shape optimization using convolutional neural networks. Physics of Fluids 36(5), 056116 (2024)

    2024

  5. [13]

    arXiv preprint arXiv:2108.07258 (2021)

    Bommasani, R., Hudson, D.A., Adeli, E., Altman, R., Arora, S., Arx, S., Bern- stein, M.S., Bohg, J., Bosselut, A., Brunskill, E., et al.: On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258 (2021)

    2021 arXiv

  6. [14]

    : Language models are few-shot learners

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. : Language models are few-shot learners. Advances in neural information processing systems 33, 1877–1901 (2020)

    2020

  7. [15]

    Nature 641(7969), 1180–1186 (2025) 29

    Bodnar, C., Bruinsma, W.P., Lucic, A., Stanley, M., Allen, A., Brandstetter, J., Garvan, P., Riechert, M., Weyn, J.A., Dong, H., Gupta, J.K., Thambiratnam, K., Archibald, A.T., Wu, C., Heider, E., Welling, M., Turner, R.E., Perdikaris, P.: A foundation model for the earth syst...

    2025

  8. [16]

    Bulletin of the American Meteorological Society 89(4), 459–470 (2008)

    Palmer, T., Doblas-Reyes, F., Weisheimer, A., Rodwell, M.: Toward seamless prediction: Calibration of climate change projections using seasonal forecasts. Bulletin of the American Meteorological Society 89(4), 459–470 (2008)

    2008

  9. [17]

    Nature 596(7873), 583–589 (2021)

    Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., ˇZ ´ ıdek, A., Potapenko, A.,et al.: Highly accurate protein structure prediction with alphafold. Nature 596(7873), 583–589 (2021)

    2021

  10. [18]

    : A deep learning approach to antibiotic discovery

    Stokes, J.M., Yang, K., Swanson, K., Jin, W., Cubillos-Ruiz, A., Donghia, N.M., MacNair, C.R., French, S., Carfrae, L.A., Bloom-Ackermann, Z., et al. : A deep learning approach to antibiotic discovery. Cell 180(4), 688–702 (2020)

    2020

  11. [19]

    Physical review letters 108(5), 058301 (2012)

    Rupp, M., Tkatchenko, A., M¨ uller, K.-R., Von Lilienfeld, O.A.: Fast and accu- rate modeling of molecular atomization energies with machine learning. Physical review letters 108(5), 058301 (2012)

    2012

  12. [20]

    arXiv preprint arXiv:2209.01712 (2022)

    Ahmad, W., Simon, E., Chithrananda, S., Grand, G., Ramsundar, B.: Chemberta- 2: Towards chemical foundation models. arXiv preprint arXiv:2209.01712 (2022)

    2022 arXiv

  13. [21]

    arXiv preprint arXiv:2401.00096 (2023)

    Batatia, I., Benner, P., Chiang, Y., Elena, A.M., Kov´ acs, D.P., Riebesell, J., Advincula, X.R., Asta, M., Avaylon, M., Baldwin, W.J., et al.: A foundation model for atomistic materials chemistry. arXiv preprint arXiv:2401.00096 (2023)

    2023 arXiv

  14. [22]

    npj Computational Materials 11(1), 61 (2025)

    Pyzer-Knapp, E.O., Manica, M., Staar, P., Morin, L., Ruch, P., Laino, T., Smith, J.R., Curioni, A.: Foundation models for materials discovery–current state and future directions. npj Computational Materials 11(1), 61 (2025)

    2025

  15. [23]

    npj Computa- tional Materials 8(1), 84 (2022)

    Pyzer-Knapp, E.O., Pitera, J.W., Staar, P.W., Takeda, S., Laino, T., Sanders, D.P., Sexton, J., Smith, J.R., Curioni, A.: Accelerating materials discovery using artificial intelligence, high performance computing and robotics. npj Computa- tional Materials 8(1), 84 (2022)

    2022

  16. [24]

    npj Computational Materials 8(1), 102 (2022)

    Gupta, T., Zaki, M., Krishnan, N.A., Mausam: Matscibert: A materials domain language model for text mining and information extraction. npj Computational Materials 8(1), 102 (2022)

    2022

  17. [25]

    ai: an open platform for automated optical chemical structure identification, segmen- tation and recognition in scientific publications

    Rajan, K., Brinkhaus, H.O., Agea, M.I., Zielesny, A., Steinbeck, C.: Decimer. ai: an open platform for automated optical chemical structure identification, segmen- tation and recognition in scientific publications. Nature communications 14(1), 5045 (2023)

    2023

  18. [26]

    In: Williams, B., Chen, Y., Neville, J

    Takeda, S., Kishimoto, A., Hamada, L., Nakano, D., Smith, J.R.: Foundation model for material science. In: Williams, B., Chen, Y., Neville, J. (eds.) Proceed- ings of the AAAI Conference on Artificial Intelligence, vol. 37, pp. 15376–15383 (2023) 30

    2023

  19. [27]

    arXiv preprint arXiv:2010.09885 (2020)

    Chithrananda, S., Grand, G., Ramsundar, B.: Chemberta: large-scale self- supervised pretraining for molecular property prediction. arXiv preprint arXiv:2010.09885 (2020)

    2020 arXiv

  20. [28]

    Nature Machine Intelligence 4(12), 1256–1264 (2022)

    Ross, J., Belgodere, B., Chenthamarakshan, V., Padhi, I., Mroueh, Y., Das, P.: Large-scale chemical language representations capture molecular structure and properties. Nature Machine Intelligence 4(12), 1256–1264 (2022)

    2022

  21. [29]

    Journal of chemical information and modeling 62(9), 2064–2076 (2021)

    Bagal, V., Aggarwal, R., Vinod, P., Priyakumar, U.D.: Molgpt: molecular gen- eration using a transformer-decoder model. Journal of chemical information and modeling 62(9), 2064–2076 (2021)

    2021

  22. [30]

    Nature Communications 15(1), 2323 (2024)

    Chang, J., Ye, J.C.: Bidirectional generation of structure and properties through a single molecular foundation model. Nature Communications 15(1), 2323 (2024)

    2024

  23. [31]

    : Inorganic materials synthesis planning with literature-trained neural networks

    Kim, E., Jensen, Z., Grootel, A., Huang, K., Staib, M., Mysore, S., Chang, H.- S., Strubell, E., McCallum, A., Jegelka, S., et al. : Inorganic materials synthesis planning with literature-trained neural networks. Journal of chemical information and modeling 60(3), 1194–1201 (2020)

    2020

  24. [32]

    npj 2D Materials and Applications 5(1), 48 (2021)

    Lew, A.J., Yu, C.-H., Hsu, Y.-C., Buehler, M.J.: Deep learning model to predict fracture mechanisms of graphene. npj 2D Materials and Applications 5(1), 48 (2021)

    2021

  25. [33]

    Computer Methods in Applied Mechanics and Engineering 395, 115021 (2022)

    Perera, R., Guzzetti, D., Agrawal, V.: Graph neural networks for simulating crack coalescence and propagation in brittle materials. Computer Methods in Applied Mechanics and Engineering 395, 115021 (2022)

    2022

  26. [34]

    Matter 3, 197–211 (2020)

    Hsu, Y., Yu, C., Buehler, M.: Using deep learning to predict fracture patterns in crystalline solids. Matter 3, 197–211 (2020)

    2020

  27. [35]

    Mechanics of Materials 171, 104309 (2022)

    Zhang, L., Wei, X.: Prediction of fatigue crack growth under variable amplitude loading by artificial neural network-based lagrange interpolation. Mechanics of Materials 171, 104309 (2022)

    2022

  28. [36]

    Mechanics of Materials 186, 104789 (2023)

    Perera, R., Agrawal, V.: Dynamic and adaptive mesh-based graph neural network framework for simulating displacement and crack fields in phase field models. Mechanics of Materials 186, 104789 (2023)

    2023

  29. [37]

    Computational Materials Science 236, 112846 (2024)

    Chen, S., Feng, S., Huang, Y., Lei, Z., Jia, X., Lin, Y., Rougier, E.: Hossnet: An efficient physics-guided neural network for simulating micro-crack propagation. Computational Materials Science 236, 112846 (2024)

    2024

  30. [38]

    Computational Materials Science 148, 46–53 (2018) 31

    Moore, B.A., Rougier, E., O’Malley, D., Srinivasan, G., Hunter, A., Viswanathan, H.: Predictive modeling of dynamic fracture growth in brittle materials with machine learning. Computational Materials Science 148, 46–53 (2018) 31

    2018

  31. [39]

    : Reduced-order modeling through machine learning and graph-theoretic approaches for brittle fracture applications

    Hunter, A., Moore, B.A., Mudunuru, M., Chau, V., Tchoua, R., Nyshadham, C., Karra, S., O’Malley, D., Rougier, E., Viswanathan, H., et al. : Reduced-order modeling through machine learning and graph-theoretic approaches for brittle fracture applications. Computational Materials...

    2019

  32. [40]

    npj Materials Degradation 5(1), 6 (2021)

    Wang, Y., Oyen, D., Guo, W., Mehta, A., Scott, C.B., Panda, N., Fern´ andez- Godino, M.G., Srinivasan, G., Yue, X.: Stressnet-deep learning to predict stress with fracture propagation in brittle materials. npj Materials Degradation 5(1), 6 (2021)

    2021

  33. [41]

    Acta Materialia 190, 105–112 (2020)

    Liu, X., Athanasiou, C.E., Padture, N.P., Sheldon, B.W., Gao, H.: A machine learning approach to fracture mechanics problems. Acta Materialia 190, 105–112 (2020)

    2020

  34. [42]

    Journal of Reinforced Plastics and Composites 39(15-16), 587–598 (2020)

    Daghigh, V., Lacy Jr, T.E., Daghigh, H., Gu, G., Baghaei, K.T., Horstemeyer, M.F., Pittman Jr, C.U.: Machine learning predictions on fracture toughness of multiscale bio-nano-composites. Journal of Reinforced Plastics and Composites 39(15-16), 587–598 (2020)

    2020

  35. [43]

    Journal of Materials Science 55, 15334–15349 (2020)

    Yan, F., Song, K., Liu, Y., Chen, S., Chen, J.: Predictions and mechanism analyses of the fatigue strength of steel based on machine learning. Journal of Materials Science 55, 15334–15349 (2020)

    2020

  36. [44]

    Fatigue & Fracture of Engineering Materials & Structures 47(8), 2752–2773 (2024)

    Frie, C., Riza Durmaz, A., Eberl, C.: Exploration of materials fatigue influence factors using interpretable machine learning. Fatigue & Fracture of Engineering Materials & Structures 47(8), 2752–2773 (2024)

    2024

  37. [45]

    Journal of Materials Science & Technology (2024)

    Wang, H., Gao, S., Wang, B., Ma, Y., Guo, Z., Zhang, K., Yang, Y., Yue, X., Hou, J., Huang, H., et al.: Recent advances in machine learning-assisted fatigue life prediction of additive manufactured metallic materials: A review. Journal of Materials Science & Technology (2024)

    2024

  38. [46]

    Scientific Reports 10(1), 16910 (2020)

    Biswas, S., Fernandez Castellanos, D., Zaiser, M.: Prediction of creep failure time using machine learning. Scientific Reports 10(1), 16910 (2020)

    2020

  39. [47]

    Journal of Petroleum Science and Engineering 210, 109963 (2022)

    Liu, R., Misra, S.: A generalized machine learning workflow to visualize mechan- ical discontinuity. Journal of Petroleum Science and Engineering 210, 109963 (2022)

    2022

  40. [48]

    Automation in Construction 133, 104022 (2022)

    Perry, B.J., Guo, Y., Mahmoud, H.N.: Automated site-specific assessment of steel structures through integrating machine learning and fracture mechanics. Automation in Construction 133, 104022 (2022)

    2022

  41. [49]

    Archives of Civil and Mechanical Engineering 25(4), 1–25 (2025) 32

    Thi, T.A.V., Dinh, H.T., Dinh, G.V., Le Cong, H., Dinh, D.N., Dinh, D.N.: Concrete crack simulation and its machine learning application in propagation prediction. Archives of Civil and Mechanical Engineering 25(4), 1–25 (2025) 32

    2025

  42. [50]

    Yosinski, J., Clune, J., Bengio, Y., Lipson, H.: How transferable are features in deep neural networks? Advances in neural information processing systems 27 (2014)

    2014

  43. [51]

    Computer Methods in Applied Mechanics and Engineering 430, 117226 (2024)

    Zhou, X.-P., Yu, X.-L.: Transfer learning enhanced nonlocal energy-informed neu- ral network for quasi-static fracture in rock-like materials. Computer Methods in Applied Mechanics and Engineering 430, 117226 (2024)

    2024

  44. [52]

    Mechanics of Materials 181, 104639 (2023)

    Perera, R., Agrawal, V.: A generalized machine learning framework for brittle crack problems using transfer learning and graph neural networks. Mechanics of Materials 181, 104639 (2023)

    2023

  45. [53]

    Expert Systems with Applications 242, 122807 (2024)

    Zhao, Z., Alzubaidi, L., Zhang, J., Duan, Y., Gu, Y.: A comparison review of transfer learning and self-supervised learning: Definitions, applications, advan- tages and limitations. Expert Systems with Applications 242, 122807 (2024)

    2024

  46. [54]

    Nature materials 20(6), 762–773 (2021)

    Dijkstra, M., Luijten, E.: From predictive modelling to machine learning and reverse engineering of colloidal self-assembly. Nature materials 20(6), 762–773 (2021)

    2021

  47. [55]

    Nature materials 18(5), 435–441 (2019)

    Ekins, S., Puhl, A.C., Zorn, K.M., Lane, T.R., Russo, D.P., Klein, J.J., Hickey, A.J., Clark, A.M.: Exploiting machine learning for end-to-end drug discovery and development. Nature materials 18(5), 435–441 (2019)

    2019

  48. [56]

    Computational Particle Mechanics 7(5), 765–787 (2020)

    Knight, E.E., Rougier, E., Lei, Z., Euser, B., Chau, V., Boyce, S.H., Gao, K., Okubo, K., Froment, M.: Hoss: an implementation of the combined finite-discrete element method. Computational Particle Mechanics 7(5), 765–787 (2020)

    2020

  49. [57]

    Wiley, Hoboken, NJ (2011)

    Munjiza, A., Knight, E.E., Rougier, E.: The Combined Finite-discrete Element Method. Wiley, Hoboken, NJ (2011)

    2011

  50. [58]

    Wiley, Chichester, West Sussex, UK (2015)

    Munjiza, A., Rougier, E., Knight, E.E.: Large Strain Finite Element Method: a Practical Course. Wiley, Chichester, West Sussex, UK (2015)

    2015

  51. [59]

    Wiley, Chichester, West Sussex, PO19 8SQ, United Kingdom (2004)

    Munjiza, A.: Computational Mechanics of Discontinua. Wiley, Chichester, West Sussex, PO19 8SQ, United Kingdom (2004)

    2004

  52. [60]

    arXiv preprint arXiv:2407.21783 (2024)

    Grattafiori, A., Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Letman, A., Mathur, A., Schelten, A., Vaughan, A., et al.: The llama 3 herd of models. arXiv preprint arXiv:2407.21783 (2024)

    2024 arXiv

  53. [61]

    Composites Part B: Engineering 169, 239–248 (2019)

    Hirshikesh, Natarajan, S., Annabattula, R.K., Mart ´ ınez-Pa˜ neda, E.: Phase field modelling of crack propagation in functionally graded materials. Composites Part B: Engineering 169, 239–248 (2019)

    2019

  54. [62]

    Acta Geotechnica 15(9), 2599–2618 33 (2020)

    Zhou, S., Zhuang, X.: Phase field modeling of hydraulic fracture propagation in transversely isotropic poroelastic media. Acta Geotechnica 15(9), 2599–2618 33 (2020)

    2020

  55. [63]

    arXiv preprint arXiv:2506.09985 (2025)

    Assran, M., Bardes, A., Fan, D., Garrido, Q., Howes, R., Komeili, M., Muckley, M., Rizvi, A., Roberts, C., Sinha, K., Zholus, A., Arnaud, S., Gejji, A., Martin, A., Hogan, F.R., Dugas, D., Bojanowski, P., Khalidov, V., Labatut, P., Massa, F., Szafraniec, M., Krishnakumar, K., ...

    2025 arXiv

  56. [64]

    Technical report (2024)

    OpenAI: GPT-4 Technical Report. Technical report (2024)

    2024

  57. [65]

    arXiv preprint arXiv:2311.11627 (2023)

    Touvron, H., al.: Llama: Open and efficient foundation language models. arXiv preprint arXiv:2311.11627 (2023)

    2023 arXiv

  58. [66]

    International Journal of Fatigue 181, 108142 (2024)

    Liu, M., Zou, T., Wang, Q., Jiang, Y., Wu, H., Pei, Y., Zhang, H., Liu, Y., Wang, Q.: Microstructure evolution, failure mechanism and life prediction of addi- tively manufactured inconel 625 superalloy with comparable low cycle fatigue performance. International Journal of Fat...

    2024

  59. [67]

    The European Physical Journal B-Condensed Matter and Complex Systems 26(2), 141–151 (2002)

    Guarino, A., Ciliberto, S., Garcimartın, A., Zei, M., Scorretti, R.: Failure time and critical behaviour of fracture precursors in heterogeneous materials. The European Physical Journal B-Condensed Matter and Complex Systems 26(2), 141–151 (2002)

    2002

  60. [68]

    Mechanics of Materials 79, 15–24 (2014)

    Hassine, M.B., Na ¨ ıt-Abdelaziz, M., Za ¨ ıri, F., Colin, X., Tourcher, C., Marque, G.: Time to failure prediction in rubber components subjected to thermal ageing: A combined approach based upon the intrinsic defect concept and the fracture mechanics. Mechanics of Materials ...

    2014

  61. [69]

    arXiv preprint arXiv:1711.05101 (2017)

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017)

    2017 arXiv

  62. [70]

    Computational Particle Mechanics 3, 215–228 (2016)

    Lei, Z., Rougier, E., Knight, E.E., Munjiza, A., Viswanathan, H.: A generalized anisotropic deformation formulation for geomaterials. Computational Particle Mechanics 3, 215–228 (2016)

    2016

  63. [71]

    Engineering Computations 33(7), 1929–1956 (2016)

    Lei, Z., Rougier, E., Knight, E.E., Frash, L., Carey, J.W., Viswanathan, H.: A non- locking composite tetrahedron element for the combined finite discrete element method. Engineering Computations 33(7), 1929–1956 (2016)

    2016

  64. [72]

    Computational Particle Mechanics 1, 307–319 (2014) 34

    Lei, Z., Rougier, E., Knight, E.E., Munjiza, A.: A framework for grand scale par- allelization of the combined finite discrete element method in 2d. Computational Particle Mechanics 1, 307–319 (2014) 34

    2014

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.