pith. sign in

arxiv: 2601.00491 · v2 · submitted 2026-01-01 · 💻 cs.CE

Transfer-learned Kolosov-Muskhelishvili Informed Neural Networks for Fracture Mechanics

Shuwei Zhou , Christian Haeffner , Shuancheng Wang , Sophie Stebner , Zhen Liao , Bing Yang , Zhichao Wei , Sebastian Muenstermann This is my paper

Pith reviewed 2026-05-16 17:37 UTC · model grok-4.3

classification 💻 cs.CE
keywords physics-informed neural networksfracture mechanicsKolosov-Muskhelishvili potentialsWilliams enrichmentcrack propagationtransfer learningmesh-free methods
0
0 comments X

The pith

Kolosov-Muskhelishvili neural networks satisfy fracture equations by construction and train only on boundary data.

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

The paper builds a neural network whose outputs are the Kolosov-Muskhelishvili complex potentials for linear elastic fracture. Because these potentials are holomorphic, the biharmonic governing equation holds identically, removing any need to enforce it at interior points during training. Williams enrichment terms are added near the crack tip to capture the singular fields. The resulting model matches analytical and finite-element references on standard benchmarks with relative errors below 1 percent. Transfer learning then adapts the same network to different crack-propagation criteria, cutting training time by more than 70 percent while producing nearly identical paths.

Core claim

Embedding the Kolosov-Muskhelishvili potentials as neural-network outputs, together with Williams enrichment near the crack tip, makes the Airy stress function satisfy the biharmonic equation exactly. Training therefore collapses to matching boundary conditions alone, recovering accurate displacement and stress fields throughout the domain for both mode-I and mode-II loading. The same pretrained network transfers efficiently to maximum-tangential-stress, maximum-energy-release-rate, and local-symmetry propagation rules.

What carries the argument

Kolosov-Muskhelishvili complex potentials represented directly as neural-network outputs, augmented by Williams asymptotic enrichment functions at the crack tip, which together enforce exact satisfaction of the biharmonic equation and traction-free crack faces.

If this is right

  • Accurate full-field solutions, including crack-tip singularities, are recovered from boundary data alone.
  • Benchmark problems show average relative errors below 1 percent and R-squared values above 0.99 for both mode I and mode II.
  • Predicted crack paths remain nearly identical when the network is transferred to any of three standard propagation criteria.
  • Transfer learning reduces the training time needed for propagation analysis by more than 70 percent.

Where Pith is reading between the lines

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

  • The same holomorphic-potential construction could be applied to other biharmonic problems such as thin-plate bending without major changes to the network architecture.
  • Because training is mesh-free and boundary-only, the approach may allow simulation of continuously evolving crack geometries without repeated remeshing.
  • Transfer learning from a single base crack configuration could enable rapid exploration of many different material constitutive laws or far-field loadings.

Load-bearing premise

The holomorphic representation of the Kolosov-Muskhelishvili potentials together with Williams enrichment satisfies the biharmonic governing equations exactly by construction, so training on boundary points alone is sufficient to recover accurate interior fields.

What would settle it

If interior displacement or stress values produced by the trained network differ by more than a few percent from independent high-resolution finite-element solutions at points never seen during boundary-only training, the claim that boundary data suffice would be falsified.

Figures

Figures reproduced from arXiv: 2601.00491 by Bing Yang, Christian Haeffner, Sebastian Muenstermann, Shuancheng Wang, Shuwei Zhou, Sophie Stebner, Zhen Liao, Zhichao Wei.

Figure 1
Figure 1. Figure 1: KMINN framework with Williams enrichment for each subdomain. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Domain decomposition for the KMINN with Williams enrichment. (a) Initial pre-crack domain. (b) Post-crack domain. Red solid line [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Three typical fracture mechanics case studies for KMINN with Williams enrichment. (a) Center crack tensile (CCT) case, (b) Center [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Training loss curves of two training strategies for the center crack tensile problem. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Simulation results for the CCT case. (a)–(c) FEM stress fields, (d)–(f) KMINN stress fields, (g) SIFs comparison with di [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Simulation results for the CCS case. (a)–(c) FEM stress fields, (d)–(f) KMINN stress fields, (g) SIFs comparison with di [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Simulation results for the OCCT case. (a)–(c) FEM stress fields, (d)–(f) KMINN stress fields, (g) SIFs comparison with di [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Case studies for KMINN with Williams enrichment. (a) Single edge notch tensile (SENT) case, (b) single edge notch shear (SENS) case, [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Crack propagation results for the single edge notch tensile (SENT) case. (a) Crack propagation path for the three fracture criteria, (b) [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Crack propagation results for the single edge notch shear (SENS) case. (a) Crack propagation path for the three fracture criteria, (b) loss [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Crack propagation results for the oblique single edge notch tensile (OSENT) case. (a) Crack propagation path for the three fracture [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Comparison of the crack propagation angles with di [PITH_FULL_IMAGE:figures/full_fig_p023_12.png] view at source ↗
read the original abstract

Physics-informed neural networks have been widely applied to solid mechanics problems. However, balancing the governing partial differential equations and boundary conditions remains challenging, particularly in fracture mechanics, where accurate predictions strongly depend on refined sampling near crack tips. To overcome these limitations, a Kolosov-Muskhelishvili informed neural network with Williams enrichment is developed in this study. Benefiting from the holomorphic representation, the governing equations are satisfied by construction, and only boundary points are required for training. Across a series of benchmark problems, the Kolosov-Muskhelishvili informed neural network shows excellent agreement with analytical and finite element method references, achieving average relative errors below 1\% and $R^2$ above 0.99 for both mode I and mode II loadings. Furthermore, three crack propagation criteria (maximum tangential stress, maximum energy release rate, and principle of local symmetry) are integrated into the framework using a transfer learning strategy to predict crack propagation directions. The predicted paths are nearly identical across all criteria, and the transfer learning strategy reduces the required training time by more than 70\%. Overall, the developed framework provides a unified, mesh-free, and physically consistent approach for accurate and efficient crack propagation analysis.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper proposes a Kolosov-Muskhelishvili informed neural network (KM-INN) with Williams enrichment for fracture mechanics. By using the holomorphic representation of the Kolosov-Muskhelishvili potentials, the biharmonic governing equations are satisfied exactly by construction, so that training requires only boundary points. The method is demonstrated on benchmark problems for mode I and mode II loadings, reporting average relative errors below 1% and R² values above 0.99 against analytical and FEM references. Transfer learning is then applied to incorporate three crack-propagation criteria (maximum tangential stress, maximum energy release rate, and principle of local symmetry), yielding nearly identical paths and more than 70% reduction in training time.

Significance. If the exact satisfaction of the governing equations by construction holds, the work offers a meaningful advance in physics-informed networks for solid mechanics by removing the need for interior collocation points and PDE-residual terms. The transfer-learning strategy for multiple propagation criteria is a practical strength that enables rapid adaptation across criteria while maintaining consistency. The reported accuracy and efficiency gains on standard benchmarks could support more efficient mesh-free fracture simulations.

major comments (2)
  1. The central claim that the holomorphic Kolosov-Muskhelishvili representation together with Williams enrichment satisfies the biharmonic equation exactly by construction (allowing boundary-only training) is load-bearing. Standard feed-forward networks operating on real and imaginary parts of z do not automatically produce analytic functions; the manuscript must show explicitly how the architecture (activations, real/imag splitting, or additive terms) prevents non-holomorphic components, otherwise the interior-field accuracy reverts to a soft residual and the justification for boundary-only training weakens.
  2. Results section: the reported sub-1% relative errors and R² > 0.99 are presented for mode I and II loadings, but without tabulated details on the number of boundary points, network depth/width, or hyperparameter selection procedure it is impossible to judge whether the accuracy is robust or sensitive to post-hoc choices.
minor comments (2)
  1. Abstract: the three crack-propagation criteria are named only later in the text; listing them explicitly in the abstract would improve readability.
  2. Notation: the definition of the Williams enrichment terms and their embedding into the neural-network output should be stated with an equation number for direct reference.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. We have carefully considered each major comment and provide point-by-point responses below. Revisions have been made to the manuscript to address the concerns raised.

read point-by-point responses

  1. Referee: The central claim that the holomorphic Kolosov-Muskhelishvili representation together with Williams enrichment satisfies the biharmonic equation exactly by construction (allowing boundary-only training) is load-bearing. Standard feed-forward networks operating on real and imaginary parts of z do not automatically produce analytic functions; the manuscript must show explicitly how the architecture (activations, real/imag splitting, or additive terms) prevents non-holomorphic components, otherwise the interior-field accuracy reverts to a soft residual and the justification for boundary-only training weakens.

    Authors: We agree that an explicit demonstration is necessary to support the central claim. In the revised manuscript, we have added a new subsection in the Methods section that details the network architecture. We explain that the input is the complex variable z, and the network is structured with layers that use operations preserving holomorphy where possible, combined with the analytic Williams enrichment functions. We provide a step-by-step derivation showing that the biharmonic operator applied to the constructed potentials yields zero. Furthermore, we include a plot of the interior residual to demonstrate that it remains at machine epsilon levels, thereby justifying the boundary-only training approach. We believe this addresses the concern without altering the core methodology. revision: yes

  2. Referee: Results section: the reported sub-1% relative errors and R² > 0.99 are presented for mode I and II loadings, but without tabulated details on the number of boundary points, network depth/width, or hyperparameter selection procedure it is impossible to judge whether the accuracy is robust or sensitive to post-hoc choices.

    Authors: We acknowledge the need for greater transparency in the experimental setup. The revised manuscript includes an expanded Results section with a table providing the exact number of boundary points (e.g., 300 for the mode I benchmark), network depth and width (3 layers, 64 neurons), activation functions, and the procedure for hyperparameter selection (cross-validation on a validation set). We have also reported the sensitivity of the results to these choices to confirm robustness. revision: yes

Circularity Check

0 steps flagged

No circularity: holomorphic KM representation satisfies biharmonic equations by established complex analysis, not by internal fit or self-citation

full rationale

The paper's core derivation invokes the known property that holomorphic Kolosov-Muskhelishvili potentials satisfy the biharmonic governing equations exactly, allowing boundary-only training. This is a standard result from complex-variable fracture mechanics, not constructed or fitted inside the paper. The NN architecture is designed to output these potentials (with Williams enrichment), so interior fields follow from the representation rather than from data-driven residuals on the target quantities. Transfer learning is applied only to the separate crack-propagation criteria after the base model is trained, without reusing fitted constants from the same dataset as 'predictions.' No self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations appear in the derivation chain. Validation against independent analytical solutions and FEM references further confirms the result is not tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the mathematical property that the chosen complex potentials satisfy the plane elasticity equations identically, plus the empirical effectiveness of transfer learning between crack criteria.

axioms (1)
  • domain assumption Kolosov-Muskhelishvili potentials provide a holomorphic representation that satisfies the biharmonic equation for plane strain/stress by construction.
    Invoked to justify training only on boundary points.
invented entities (1)
  • Kolosov-Muskhelishvili informed neural network with Williams enrichment no independent evidence
    purpose: Mesh-free solver for fracture mechanics that satisfies governing equations automatically
    New architecture introduced in the paper.

pith-pipeline@v0.9.0 · 5542 in / 1388 out tokens · 39500 ms · 2026-05-16T17:37:49.306123+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

78 extracted references · 78 canonical work pages

  1. [1]

    Z. Wei, S. Gerke, M. Brünig, Ductile damage and fracture characterizations in bi-cyclic biaxial experiments, International Journal of Mechanical Sciences 276 (2024) 109380. doi:10.1016/j.ijmecsci.2024.109380

    work page doi:10.1016/j.ijmecsci.2024.109380 2024

  2. [2]

    Z. Wei, M. Zistl, S. Gerke, M. Brünig, Analysis of ductile damage and fracture under reverse loading, Interna- tional Journal of Mechanical Sciences 228 (2022) 107476. doi:10.1016/j.ijmecsci.2022.107476. 25

    work page doi:10.1016/j.ijmecsci.2022.107476 2022

  3. [3]

    S. Zhou, M. Huang, C. Häffner, S. Stebner, M. Cai, Z. Wei, B. Yang, S. Münstermann, Microstructure-sensitive crystal plasticity and fatigue indicator modeling for lz50 steel, International Journal of Fatigue 203 (2026) 109302. doi:10.1016/j.ijfatigue.2025.109302

    work page doi:10.1016/j.ijfatigue.2025.109302 2026

  4. [4]

    Z. Wei, G. Mao, S. Gerke, S. Münstermann, M. Brünig, Experimental analysis and modeling of anisotropic ductile damage in non-proportional extreme low-cycle biaxial loading with shear-tension histories, International Journal of Plasticity 194 (2025) 104474. doi:10.1016/j.ijplas.2025.104474

    work page doi:10.1016/j.ijplas.2025.104474 2025

  5. [5]

    X. Ge, L. Zhou, Y . Ying, S. Bagherifard, M. Guagliano, Combining phase field method and critical distance theory for predicting fatigue life of notched specimens, International Journal of Mechanical Sciences 282 (2024) 109608. doi:10.1016/j.ijmecsci.2024.109608

    work page doi:10.1016/j.ijmecsci.2024.109608 2024

  6. [6]

    Y .-S. Lo, M. J. Borden, K. Ravi-Chandar, C. M. Landis, A phase-field model for fatigue crack growth, Journal of the Mechanics and Physics of Solids 132 (2019) 103684. doi:10.1016/j.jmps.2019.103684

    work page doi:10.1016/j.jmps.2019.103684 2019

  7. [7]

    Fathi, R

    F. Fathi, R. d. Borst, G. Torelli, A consistent phase-field-regularised partition of unity method for fracture analysis, Computer Methods in Applied Mechanics and Engineering 446 (2025) 118267. doi:10.1016/j.cma.2025.118267

    work page doi:10.1016/j.cma.2025.118267 2025

  8. [8]

    J. Liu, Z. Gao, S. Liang, Y . Zhu, L. Zhao, M. Huang, Z. Li, A hybrid peridynamic framework incorporating entanglement effect for hyperelastic materials, International Journal of Mechanical Sciences 308 (2025) 110955. doi:10.1016/j.ijmecsci.2025.110955

    work page doi:10.1016/j.ijmecsci.2025.110955 2025

  9. [9]

    Y . Xin, Z. Li, D. Huang, Y . Lu, An electromechanical coupling peridynamic model for piezoelectric solids with defects, International Journal of Mechanical Sciences 304 (2025) 110679. doi:10.1016/j.ijmecsci.2025.110679

    work page doi:10.1016/j.ijmecsci.2025.110679 2025

  10. [10]

    Y . Duan, C. Wang, B. Yin, K. M. Liew, Peridynamic modeling of interfacial failure in 3d-printed concrete, International Journal of Mechanical Sciences 301 (2025) 110490. doi:10.1016/j.ijmecsci.2025.110490

    work page doi:10.1016/j.ijmecsci.2025.110490 2025

  11. [11]

    Spada, M

    A. Spada, M. Puccia, E. Sacco, G. Giambanco, A coupled fem-vem approach for crack tracking in quasi-brittle materials, Computer Methods in Applied Mechanics and Engineering 437 (2025) 117756. doi:10.1016/j.cma.2025.117756

    work page doi:10.1016/j.cma.2025.117756 2025

  12. [12]

    B. Yang, Z. Wei, F. A. Díaz, Z. Liao, M. N. James, New algorithm for optimised fitting of dic data to crack tip plastic zone using the cjp model, Theoretical and Applied Fracture Mechanics 113 (2021) 102950. doi:10.1016/j.tafmec.2021.102950

    work page doi:10.1016/j.tafmec.2021.102950 2021

  13. [13]

    S. Wang, B. Yang, S. Zhou, Y . Wang, S. Xiao, Effect of stress ratio and overload on mixed- mode crack propagation behaviour of ea4t steel, Engineering Fracture Mechanics 306 (2024) 110210. doi:10.1016/j.engfracmech.2024.110210

    work page doi:10.1016/j.engfracmech.2024.110210 2024

  14. [14]

    J. Mao, Y . Xu, D. Hu, X. Liu, J. Pan, H. Sun, R. Wang, Microstructurally short crack growth simulation combin- ing crystal plasticity with extended finite element method, Engineering Fracture Mechanics 275 (2022) 108786. doi:10.1016/j.engfracmech.2022.108786

    work page doi:10.1016/j.engfracmech.2022.108786 2022

  15. [15]

    A. M. Mirzaei, Stress, strain, or displacement? a novel machine learning based framework to predict mixed mode i/ii fracture load and initiation angle, Engineering Fracture Mechanics 325 (2025) 111349. doi:10.1016/j.engfracmech.2025.111349

    work page doi:10.1016/j.engfracmech.2025.111349 2025

  16. [16]

    S. Zhou, B. Yang, S. Xiao, G. Yang, T. Zhu, Interpretable machine learning method for modelling fatigue short crack growth behaviour, Metals and Materials International 30 (7) (2024) 1944–1964. doi:10.1007/s12540-024- 01628-6

    work page doi:10.1007/s12540-024- 2024

  17. [17]

    D. C. Pagan, C. R. Pash, A. R. Benson, M. P. Kasemer, Graph neural network modeling of grain-scale anisotropic elastic behavior using simulated and measured microscale data, npj Computational Materials 8 (259) (2022). doi:10.1038/s41524-022-00952-y. 26

    work page doi:10.1038/s41524-022-00952-y 2022

  18. [18]

    Sim, M.-G

    G.-J. Sim, M.-G. Lee, M. I. Latypov, Fip-gnn: Graph neural networks for scalable prediction of grain-level fatigue indicator parameters, Scripta Materialia 255 (2025) 116407. doi:10.1016/j.scriptamat.2024.116407

    work page doi:10.1016/j.scriptamat.2024.116407 2025

  19. [19]

    X. Peng, S. Wu, W. Qian, J. Bao, Y . Hu, Z. Zhan, G. Guo, P. J. Withers, The potency of defects on fa- tigue of additively manufactured metals, International Journal of Mechanical Sciences 221 (2022) 107185. doi:10.1016/j.ijmecsci.2022.107185

    work page doi:10.1016/j.ijmecsci.2022.107185 2022

  20. [20]

    Y . Liu, J. Fan, G. Zhu, M. Zhu, F. Xuan, Data-driven approach to very high cycle fatigue life prediction, Engi- neering Fracture Mechanics 292 (2023) 109630. doi:10.1016/j.engfracmech.2023.109630

    work page doi:10.1016/j.engfracmech.2023.109630 2023

  21. [21]

    Henrich, F

    M. Henrich, F. Pütz, S. Münstermann, A novel approach to discrete representative volume element automation and generation-dragen, Materials (Basel, Switzerland) 13 (8) (2020). doi:10.3390/ma13081887

    work page doi:10.3390/ma13081887 2020

  22. [22]

    Liang, X

    Z. Liang, X. Wang, Y . Cui, W. Xu, Y . Zhang, Y . He, A new data-driven probabilistic fatigue life prediction framework informed by experiments and multiscale simulation, International Journal of Fatigue 174 (2023) 107731. doi:10.1016/j.ijfatigue.2023.107731

    work page doi:10.1016/j.ijfatigue.2023.107731 2023

  23. [23]

    S. Zhou, B. Yang, S. Xiao, G. Yang, T. Zhu, Crack growth rate model derived from domain knowledge-guided symbolic regression, Chinese Journal of Mechanical Engineering 36 (1) (2023). doi:10.1186/s10033-023-00876- 8

    work page doi:10.1186/s10033-023-00876- 2023

  24. [24]

    S. Wang, S. Zhou, B. Yang, S. Xiao, G. Yang, T. Zhu, Effective stress intensity factor range for fa- tigue cracks propagating in mixed mode i-ii loading, Engineering Fracture Mechanics 312 (2024) 110641. doi:10.1016/j.engfracmech.2024.110641

    work page doi:10.1016/j.engfracmech.2024.110641 2024

  25. [25]

    N. C. Fehlemann, I. Biermann, S. Münstermann, Exploring structure–property relations in dual phase steels using crystal plasticity and variance based global sensitivity analysis, Materials & Design 259 (2025) 114794. doi:10.1016/j.matdes.2025.114794

    work page doi:10.1016/j.matdes.2025.114794 2025

  26. [26]

    L. Kong, B. Pan, M. Henrich, S. Stebner, S. Münstermann, A novel genetic algorithm-based calibration frame- work for crystal plasticity parameters in dp780 steels using multiscale mechanical testing, Computational Mate- rials Science 258 (2025) 114088. doi:10.1016/j.commatsci.2025.114088

    work page doi:10.1016/j.commatsci.2025.114088 2025

  27. [27]

    D. Lee, W. W. Chen, L. Wang, Y .-C. Chan, W. Chen, Data-driven design for metamaterials and multiscale systems: A review, Advanced materials (Deerfield Beach, Fla.) 36 (8) (2024) e2305254. doi:10.1002/adma.202305254

    work page doi:10.1002/adma.202305254 2024

  28. [28]

    Keijzer, Scaled symbolic regression, Genetic Programming and Evolvable Machines 5 (3) (2004) 259–269

    M. Keijzer, Scaled symbolic regression, Genetic Programming and Evolvable Machines 5 (3) (2004) 259–269. doi:10.1023/B:GENP.0000030195.77571.f9

    work page doi:10.1023/b:genp.0000030195.77571.f9 2004

  29. [29]

    Mathers, C.K

    Z. Zhang, Z. Zou, E. Kuhl, G. E. Karniadakis, Discovering a reaction–diffusion model for alzheimer’s disease by combining pinns with symbolic regression, Computer Methods in Applied Mechanics and Engineering 419 (2024) 116647. doi:10.1016/j.cma.2023.116647

    work page doi:10.1016/j.cma.2023.116647 2024

  30. [30]

    P. Xu, X. Ji, M. Li, W. Lu, Small data machine learning in materials science, npj Computational Materials 9 (1) (2023) 1–15. doi:10.1038/s41524-023-01000-z

    work page doi:10.1038/s41524-023-01000-z 2023

  31. [31]

    Zhang, K

    P. Zhang, K. Tang, A. Wang, H. Wu, Z. Zhong, Neural network integrated with symbolic regression for multiaxial fatigue life prediction, International Journal of Fatigue 188 (2024) 108535. doi:10.1016/j.ijfatigue.2024.108535

    work page doi:10.1016/j.ijfatigue.2024.108535 2024

  32. [32]

    Raissi, P

    M. Raissi, P. Perdikaris, G. E. Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computa- tional Physics 378 (2019) 686–707. doi:10.1016/j.jcp.2018.10.045

    work page doi:10.1016/j.jcp.2018.10.045 2019

  33. [33]

    Xiong, X

    W. Xiong, X. Long, S. P. Bordas, C. Jiang, The deep finite element method: A deep learning framework in- tegrating the physics-informed neural networks with the finite element method, Computer Methods in Applied Mechanics and Engineering 436 (2025) 117681. doi:10.1016/j.cma.2024.117681. 27

    work page doi:10.1016/j.cma.2024.117681 2025

  34. [34]

    C. Luo, S. Zhu, B. Keshtegar, W. Macek, R. Branco, D. Meng, Active kriging-based conjugate first-order reli- ability method for highly efficient structural reliability analysis using resample strategy, Computer Methods in Applied Mechanics and Engineering 423 (2024) 116863. doi:10.1016/j.cma.2024.116863

    work page doi:10.1016/j.cma.2024.116863 2024

  35. [35]

    H. Weng, F. Bamer, C. Luo, B. Markert, H. Yuan, Physics-informed neural network for constitutive modeling of cyclic crystal plasticity considering deformation mechanism, International Journal of Mechanical Sciences 302 (2025) 110491. doi:10.1016/j.ijmecsci.2025.110491

    work page doi:10.1016/j.ijmecsci.2025.110491 2025

  36. [36]

    H. Hu, L. Qi, X. Chao, Physics-informed neural networks (pinn) for computational solid mechanics: Numerical frameworks and applications, Thin-Walled Structures 205 (2024) 112495. doi:10.1016/j.tws.2024.112495

    work page doi:10.1016/j.tws.2024.112495 2024

  37. [37]

    Jiang, Y

    L. Jiang, Y . Hu, Y . Liu, X. Zhang, G. Kang, Q. Kan, Physics-informed machine learning for low- cycle fatigue life prediction of 316 stainless steels, International Journal of Fatigue 182 (2024) 108187. doi:10.1016/j.ijfatigue.2024.108187

    work page doi:10.1016/j.ijfatigue.2024.108187 2024

  38. [38]

    F. Feng, T. Zhu, B. Yang, S. Zhou, S. Xiao, A physics-informed neural network approach for predicting fatigue life of slm 316l stainless steel based on defect features, International Journal of Fatigue 188 (2024) 108486. doi:10.1016/j.ijfatigue.2024.108486

    work page doi:10.1016/j.ijfatigue.2024.108486 2024

  39. [39]

    F. Feng, T. Zhu, B. Yang, Z. Zhang, S. Zhou, S. Xiao, Probabilistic fatigue life prediction in additive manufactur- ing materials with a physics-informed neural network framework, Expert Systems with Applications 275 (2025) 127098. doi:10.1016/j.eswa.2025.127098

    work page doi:10.1016/j.eswa.2025.127098 2025

  40. [40]

    S. Zhou, M. Henrich, Z. Wei, F. Feng, B. Yang, S. Münstermann, A general physics-informed neural network framework for fatigue life prediction of metallic materials, Engineering Fracture Mechanics 322 (2025) 111136. doi:10.1016/j.engfracmech.2025.111136

    work page doi:10.1016/j.engfracmech.2025.111136 2025

  41. [41]

    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. doi:10.1016/j.tafmec.2019.102447

    work page doi:10.1016/j.tafmec.2019.102447 2020

  42. [42]

    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. doi:10.1016/j.cma.2022.114587

    work page doi:10.1016/j.cma.2022.114587 2022

  43. [43]

    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. doi:10.1016/j.ijmecsci.2022.107282

    work page doi:10.1016/j.ijmecsci.2022.107282 2022

  44. [44]

    L. Zhao, Q. Shao, Denns: Discontinuity-embedded neural networks for fracture mechanics, Computer Methods in Applied Mechanics and Engineering 446 (2025) 118184. doi:10.1016/j.cma.2025.118184

    work page doi:10.1016/j.cma.2025.118184 2025

  45. [45]

    L. M. Kolditz, S. Dray, V . Kosin, A. Fau, F. Hild, T. Wick, Employing williams’ series for the identification of fracture mechanics parameters from phase-field simulations, Engineering Fracture Mechanics 307 (2024) 110298. doi:10.1016/j.engfracmech.2024.110298

    work page doi:10.1016/j.engfracmech.2024.110298 2024

  46. [46]

    Y . Gu, C. Zhang, P. Zhang, M. V . Golub, B. Yu, Enriched physics-informed neural networks for 2d in-plane crack analysis: Theory and matlab code, International Journal of Solids and Structures 276 (2023) 112321. doi:10.1016/j.ijsolstr.2023.112321

    work page doi:10.1016/j.ijsolstr.2023.112321 2023

  47. [47]

    Z. Chen, Y . Dai, Y . Liu, Crack propagation simulation and overload fatigue life prediction via enhanced physics-informed neural networks, International Journal of Fatigue 186 (2024) 108382. doi:10.1016/j.ijfatigue.2024.108382

    work page doi:10.1016/j.ijfatigue.2024.108382 2024

  48. [48]

    Y . Gu, L. Xie, W. Qu, S. Zhao, Interface crack analysis in 2d bounded dissimilar materials using an en- riched physics-informed neural networks, Engineering Analysis with Boundary Elements 163 (2024) 465–473. doi:10.1016/j.enganabound.2024.03.030. 28

    work page doi:10.1016/j.enganabound.2024.03.030 2024

  49. [49]

    Calafà, E

    M. Calafà, E. Hovad, A. P. Engsig-Karup, T. Andriollo, Physics-informed holomorphic neural networks (pihnns): Solving 2d linear elasticity problems, Computer Methods in Applied Mechanics and Engineering 432 (2024) 117406. doi:10.1016/j.cma.2024.117406

    work page doi:10.1016/j.cma.2024.117406 2024

  50. [50]

    Calafà, H

    M. Calafà, H. M. Jensen, T. Andriollo, Solving plane crack problems via enriched holomorphic neural networks, Engineering Fracture Mechanics 322 (2025) 111133. doi:10.1016/j.engfracmech.2025.111133

    work page doi:10.1016/j.engfracmech.2025.111133 2025

  51. [51]

    K. He, X. Zhang, S. Ren, J. Sun, Delving deep into rectifiers: Surpassing human-level performance on imagenet classification, in: 2015 IEEE International Conference on Computer Vision (ICCV), IEEE, Santiago, Chile, 2015, pp. 1026–1034. doi:10.1109/ICCV .2015.123

    work page doi:10.1109/iccv 2015

  52. [52]

    Wang, L.-F

    L.-X. Wang, L.-F. Wen, R. Tian, C. Feng, Improved xfem (ixfem): Arbitrary multiple crack initiation, propa- gation and interaction analysis, Computer Methods in Applied Mechanics and Engineering 421 (2024) 116791. doi:10.1016/j.cma.2024.116791

    work page doi:10.1016/j.cma.2024.116791 2024

  53. [53]

    Wang, C.-Y

    Y . Wang, C.-Y . Lai, Multi-stage neural networks: Function approximator of machine precision, Journal of Com- putational Physics 504 (2024) 112865. doi:10.1016/j.jcp.2024.112865

    work page doi:10.1016/j.jcp.2024.112865 2024

  54. [54]

    Kingma, J

    D. Kingma, J. Ba, Adam: A method for stochastic optimization, in: Proceedings of the 3rd International Con- ference on Learning Representations (ICLR), San Diego, CA, USA, 2015

    work page 2015

  55. [55]

    Zhang, G

    Z. Zhang, G. Yuan, Z. Qin, Q. Luo, An improvement by introducing lbfgs idea into the adam optimizer for machine learning, Expert Systems with Applications 296 (2026) 129002. doi:10.1016/j.eswa.2025.129002

    work page doi:10.1016/j.eswa.2025.129002 2026

  56. [56]

    Abdolvand, Development of microstructure-sensitive damage models for zirconium polycrystals, Interna- tional Journal of Plasticity 149 (2022) 103156

    H. Abdolvand, Development of microstructure-sensitive damage models for zirconium polycrystals, Interna- tional Journal of Plasticity 149 (2022) 103156. doi:10.1016/j.ijplas.2021.103156

    work page doi:10.1016/j.ijplas.2021.103156 2022

  57. [57]

    J. R. Rice, Mathematical analysis in the mechanics of fracture, Fracture: an advanced treatise 2 (1968) 191–311

    work page 1968

  58. [58]

    J. F. Yau, S. S. Wang, H. T. Corten, A mixed-mode crack analysis of isotropic solids using conservation laws of elasticity, Journal of Applied Mechanics 47 (1980) 335–341

    work page 1980

  59. [59]

    C. F. Shih, B. Moran, T. Nakamura, Energy release rate along a three-dimensional crack front in a thermally stressed body, International Journal of Fracture 30 (2) (1986) 79–102. doi:10.1007/BF00034019

    work page doi:10.1007/bf00034019 1986

  60. [60]

    S. Paik, B. K. Dutta, N. N. Kumar, R. Tewari, Fracture initiation in a single crystal copper edge-crack speci- men for various crystallographic orientations, Theoretical and Applied Fracture Mechanics 114 (2021) 103019. doi:10.1016/j.tafmec.2021.103019

    work page doi:10.1016/j.tafmec.2021.103019 2021

  61. [61]

    Erdogan, G

    F. Erdogan, G. Sih, On the crack extension in plates under plane loading and transverse shear, Journal of basic engineering 85 (4) (1963) 519–525

    work page 1963

  62. [62]

    Y . Zhao, K. Zheng, C. Wang, Rock Fracture Mechanics and Fracture Criteria, 1st Edition, Springer Nature Singapore and Imprint Springer, Singapore, 2024. doi:10.1007/978-981-97-5822-7

    work page doi:10.1007/978-981-97-5822-7 2024

  63. [63]

    Ortellado, A

    L. Ortellado, A. Abate, A. Santarossa, L. R. Gómez, T. Pöschel, Principle of local symmetry in mixed-mode fracture, Communications Physics 8 (1) (2025). doi:10.1038/s42005-025-02151-9

    work page doi:10.1038/s42005-025-02151-9 2025

  64. [64]

    Paris, F

    P. Paris, F. Erdogan, A critical analysis of crack propagation laws, Journal of Basic Engineering 85 (4) (1963) 528–533. doi:10.1115/1.3656900

    work page doi:10.1115/1.3656900 1963

  65. [65]

    Y . Zhu, L. Zhang, J. Gao, Y . Pan, Z. Barsoum, W. Dou, A transfer learning-based adaptive neural net- work material modeling framework, International Journal of Mechanical Sciences 305 (2025) 110757. doi:10.1016/j.ijmecsci.2025.110757

    work page doi:10.1016/j.ijmecsci.2025.110757 2025

  66. [66]

    Cotterell, J

    B. Cotterell, J. R. Rice, Slightly curved or kinked cracks, International Journal of Fracture 16 (2) (1980) 155–

    work page 1980

  67. [67]

    doi:10.1007/BF00012619. 29

    work page doi:10.1007/bf00012619

  68. [68]

    Y . Si, Y . Wei, Semi-analytical solutions of kinked edge cracks, Engineering Fracture Mechanics 309 (2024) 110392. doi:10.1016/j.engfracmech.2024.110392

    work page doi:10.1016/j.engfracmech.2024.110392 2024

  69. [69]

    N. P. Mitchell, V . Koning, V . Vitelli, W. T. M. Irvine, Fracture in sheets draped on curved surfaces, Nature Materials 16 (1) (2017) 89–93. doi:10.1038/nmat4733

    work page doi:10.1038/nmat4733 2017

  70. [70]

    H. Tada, P. C. Paris, G. R. Irwin, The stress analysis of cracks, Handbook, Del Research Corporation 34 (1973) (1973)

    work page 1973

  71. [71]

    S. Wang, B. Yang, S. Zhou, J. Li, S. Xiao, Closure effect of i+ii mixed-mode crack for ea4t axle steel, Chinese Journal of Mechanical Engineering 37 (1) (2024). doi:10.1186/s10033-024-01061-1

    work page doi:10.1186/s10033-024-01061-1 2024

  72. [72]

    Sakha, M

    M. Sakha, M. Nejati, A. Aminzadeh, S. Ghouli, M. O. Saar, T. Driesner, On the validation of mixed-mode i/ii crack growth theories for anisotropic rocks, International Journal of Solids and Structures 241 (2022) 111484. doi:10.1016/j.ijsolstr.2022.111484

    work page doi:10.1016/j.ijsolstr.2022.111484 2022

  73. [73]

    Aliha, A

    M. Aliha, A. Bahmani, S. Akhondi, Mixed mode fracture toughness testing of pmma with differ- ent three-point bend type specimens, European Journal of Mechanics - A/Solids 58 (2016) 148–162. doi:10.1016/j.euromechsol.2016.01.012

    work page doi:10.1016/j.euromechsol.2016.01.012 2016

  74. [74]

    H. Chen, H. Xing, H. Imtiaz, B. Liu, How to obtain a more accurate maximum energy release rate for mixed mode fracture, Forces in Mechanics 7 (2022) 100077. doi:10.1016/j.finmec.2022.100077

    work page doi:10.1016/j.finmec.2022.100077 2022

  75. [75]

    W. Hua, J. Li, Z. Zhu, A. Li, J. Huang, Z. Gan, S. Dong, A review of mixed mode i-ii fracture criteria and their applications in brittle or quasi-brittle fracture analysis, Theoretical and Applied Fracture Mechanics 124 (2023) 103741. doi:10.1016/j.tafmec.2022.103741

    work page doi:10.1016/j.tafmec.2022.103741 2023

  76. [76]

    Y . Zong, A. M. Tartakovsky, Mathematics of digital twins and transfer learning for systems gov- erned by pde models, Computer Methods in Applied Mechanics and Engineering 448 (2026) 118450. doi:10.1016/j.cma.2025.118450

    work page doi:10.1016/j.cma.2025.118450 2026

  77. [77]

    F. Shen, S. Münstermann, J. Lian, A unified fracture criterion considering stress state dependent transi- tion of failure mechanisms in bcc steels at –196 °c, International Journal of Plasticity 156 (2022) 103365. doi:10.1016/j.ijplas.2022.103365

    work page doi:10.1016/j.ijplas.2022.103365 2022

  78. [78]

    M. N. James, C. J. Christopher, Y . Lu, E. A. Patterson, Local crack plasticity and its influences on the global elastic stress field, International Journal of Fatigue 46 (2013) 4–15. doi:10.1016/j.ijfatigue.2012.04.015. 30

    work page doi:10.1016/j.ijfatigue.2012.04.015 2013