pith. sign in

arxiv: 2605.22845 · v1 · pith:DXH5ATKZnew · submitted 2026-05-14 · 💻 cs.CE · cs.LG

Cross-attention-based bipartite graph neural network for coupled nodal and elemental field prediction in large-deformation sheet material forming

Yingxue Zhao , Haoran Li , Haosu Zhou , Tobias Pfaff , Nan Li This is my paper

Pith reviewed 2026-05-25 00:34 UTC · model grok-4.3

classification 💻 cs.CE cs.LG
keywords bipartite graph neural networkcross-attentionfinite element surrogatesheet metal forminglarge deformationcoupled field predictionmesh nodes and elementsrollout stabilization
0
0 comments X

The pith

A bipartite graph network with cross-attention predicts nodal displacements and elemental thinning directly on their native mesh domains.

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

The paper proposes a graph neural network that represents mesh nodes and elements as two distinct sets of entities connected by directed edges. Cross-attention conditioned on edge geometry allows messages to pass bidirectionally between nodal kinematics and elemental deformation states, so both fields are predicted without post-processing interpolation. This structure is meant to preserve the node-element coupling that drives finite-element updates in large-deformation sheet forming. A hierarchical version adds downsampling and upsampling to handle larger meshes. If the approach holds, surrogates could generate consistent multi-field rollouts faster than full simulators while respecting the native discretisation of each quantity.

Core claim

The central claim is that a cross-attention-based bipartite graph neural network (CAtt-BiGNN) models nodes and elements as separate vertex sets linked by directed node-element edges; an edge-aware cross-attention processor then computes adaptive coupling weights from geometric features, enabling simultaneous prediction of displacement increments on nodes and thinning on elements. The hierarchical extension CAtt-BiUGNN adds graph coarsening and refinement to improve long-range information flow on bigger meshes. Experiments on two forming cases show improved balance between the two field types relative to node-centred baselines and bipartite ablations, with the hierarchical model strongest on

What carries the argument

CAtt-BiGNN, a bipartite graph with nodes and elements as distinct vertex partitions connected by directed edges whose features drive an edge-aware cross-attention processor for bidirectional state exchange between kinematic and deformation quantities.

If this is right

  • Displacement and thinning predictions remain more balanced than those from node-centred graph models because each field is updated on its native discretisation.
  • The hierarchical CAtt-BiUGNN variant reduces error accumulation on larger meshes by propagating information through successive coarsening and refinement stages.
  • Adaptive Gaussian noise injection during rollout improves stability of long prediction sequences without altering the core message-passing architecture.
  • The same bipartite cross-attention structure can serve as a drop-in surrogate component for any finite-element workflow that alternates between nodal and elemental updates.

Where Pith is reading between the lines

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

  • The same node-element bipartite construction could be applied to other coupled problems such as fluid-structure interaction where kinematic and stress fields live on different mesh entities.
  • Because the attention weights are learned from geometry alone, the model may transfer across material models provided the training data cover the relevant deformation range.
  • Extending the hierarchy to multiple coarsening levels might further improve scaling on industrial-size meshes without increasing the number of message-passing layers.

Load-bearing premise

The node-element coupling inside the finite-element update can be captured by directed edges between separate node and element entities plus cross-attention, without needing the simulator's full constitutive or contact rules.

What would settle it

Train the model on one forming geometry and material, then evaluate rollout error on a second geometry with different boundary conditions or thickness; if the displacement and thinning errors both stay within the range reported for the original test cases, the claim holds.

Figures

Figures reproduced from arXiv: 2605.22845 by Haoran Li, Haosu Zhou, Nan Li, Tobias Pfaff, Yingxue Zhao.

Figure 1
Figure 1. Figure 1: Overview of the proposed workflow. 4. Dataset generation The FE simulations were recorded at 11 discrete timesteps, following the standard output settings commonly used in commercial sheet-forming solvers. The first timestep corresponds to the undeformed initial configura￾tion, while the remaining ten timesteps describe the progressive deformation of the blank throughout the forming process. Such temporal … view at source ↗
Figure 2
Figure 2. Figure 2: Dome case study (a) simulation setup (b) geometry parameter variables (c) [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Corner case study (a) simulation setup (b) geometry parameter variables (c) [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Bipartite graph construction. (a) FE mesh with mesh nodes, FE mesh edges, [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: CAtt-BiGNN architecture and autoregressive rollout. At timestep n, the encoded [PITH_FULL_IMAGE:figures/full_fig_p022_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Cross-attention-based CAtt-BiGNN processor block. Each processor layer up [PITH_FULL_IMAGE:figures/full_fig_p024_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Hierarchical CAtt-BiUGNN architecture. The original fine bipartite graph is [PITH_FULL_IMAGE:figures/full_fig_p029_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Graph coarsening used for the downsampling and upsampling operations in [PITH_FULL_IMAGE:figures/full_fig_p031_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Dome case: evolution of test-set mean errors over the 10 autoregressive rollout [PITH_FULL_IMAGE:figures/full_fig_p040_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Final-timestep visual comparison for a representative dome test case. Columns [PITH_FULL_IMAGE:figures/full_fig_p041_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Corner case: evolution of test-set mean errors over the 10 autoregressive rollout [PITH_FULL_IMAGE:figures/full_fig_p045_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Final-timestep visual comparison for a representative corner test case. Columns [PITH_FULL_IMAGE:figures/full_fig_p046_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Attention-deviation maps at the last processor layer and final rollout timestep [PITH_FULL_IMAGE:figures/full_fig_p047_13.png] view at source ↗
read the original abstract

Finite element simulations of large-deformation sheet material forming involve node-element coupling between nodal kinematics and element-level deformation measures. Machine-learning surrogates can accelerate such simulations, but most graph-based models use node-centred representations. This representation is indirect for element-level quantities, which are often recovered from nodal predictions by interpolation or post-processing. It may also obscure the node-element coupling structure that underlies the finite element update. This work proposes a cross-attention-based bipartite graph neural network (CAtt-BiGNN) for coupled prediction of nodal displacement increments and elemental thinning. The graph represents mesh nodes and elements as distinct but connected entities, linked by directed node-element edges, so that nodal and elemental fields are predicted on their native discretisation domains. An edge-aware cross-attention processor conditions adaptive node-element coupling weights on geometric edge features, enabling bidirectional message passing between nodal kinematic states and elemental deformation states. A hierarchical extension, CAtt-BiUGNN, combines the CAtt-BiGNN with graph downsampling-upsampling to improve information propagation on larger meshes. Adaptive Gaussian noise is further evaluated as an optional rollout-stabilisation strategy. The models are tested on two representative forming cases with different graph sizes. CAtt-BiGNN improves the balance between displacement and thinning prediction relative to node-centred baselines and bipartite ablation variants, while CAtt-BiUGNN gives the strongest overall performance in the larger-graph setting. The results indicate that the proposed model provides an effective surrogate framework for large-deformation sheet material forming.

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 / 1 minor

Summary. The paper proposes a cross-attention-based bipartite graph neural network (CAtt-BiGNN) that represents mesh nodes and elements as distinct entities connected by directed edges, using edge-aware cross-attention to predict coupled nodal displacement increments and elemental thinning in large-deformation sheet forming FE simulations. A hierarchical extension (CAtt-BiUGNN) with graph downsampling-upsampling is introduced for larger meshes, along with optional adaptive Gaussian noise for rollout stabilization. On two forming cases, CAtt-BiGNN is claimed to improve the balance between displacement and thinning predictions relative to node-centred baselines and bipartite ablations, with CAtt-BiUGNN strongest on larger graphs.

Significance. If the performance claims hold under detailed scrutiny, the bipartite formulation offers a direct way to model node-element coupling on native discretizations, which could yield more consistent surrogates for forming processes than post-processed node-only models. The hierarchical variant and stabilization strategy address practical scalability issues in graph-based surrogates.

major comments (2)
  1. [Abstract] Abstract: the central empirical claim (improved balance between displacement and thinning) is presented without any quantitative metrics, loss definitions, training details, or ablation tables. This prevents verification of whether the reported gains are supported by the data or are statistically meaningful.
  2. [Method] Method (cross-attention processor description): the model relies on directed node-element edges and edge-aware cross-attention to capture coupling, yet the finite-element update depends on explicit constitutive integration, contact detection, and tangent operators that are not injected into the graph. It is unclear whether implicit learning from trajectories suffices for the claimed surrogate fidelity, especially under changes in material parameters or friction (directly relevant to the skeptic concern on constitutive rules).
minor comments (1)
  1. [Abstract] Notation for the bipartite entities and edge features could be introduced with a small diagram or table to clarify the directed node-to-element and element-to-node connections.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below, providing clarifications and indicating planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claim (improved balance between displacement and thinning) is presented without any quantitative metrics, loss definitions, training details, or ablation tables. This prevents verification of whether the reported gains are supported by the data or are statistically meaningful.

    Authors: We agree that the abstract would be strengthened by including key quantitative metrics. The full loss definitions (Section 3.3), training details (Section 4.2), ablation tables (Table 2), and performance metrics comparing displacement and thinning errors (Tables 1 and 3) are already present in the manuscript. In the revised version we will add specific numerical improvements (e.g., relative error reductions) to the abstract to support the central claim. revision: yes

  2. Referee: [Method] Method (cross-attention processor description): the model relies on directed node-element edges and edge-aware cross-attention to capture coupling, yet the finite-element update depends on explicit constitutive integration, contact detection, and tangent operators that are not injected into the graph. It is unclear whether implicit learning from trajectories suffices for the claimed surrogate fidelity, especially under changes in material parameters or friction (directly relevant to the skeptic concern on constitutive rules).

    Authors: The proposed model is a data-driven surrogate trained end-to-end on FE simulation trajectories; the bipartite graph and edge-aware cross-attention learn the observed node-element coupling implicitly from the data rather than through explicit injection of constitutive integration, contact, or tangent operators. This design choice enables direct prediction on native mesh domains and yields the reported improvements for the fixed material and friction parameters used in the two test cases. We acknowledge that generalization to unseen material parameters or friction coefficients is not evaluated and would require retraining or domain adaptation. We will add an explicit limitations paragraph in the revised discussion section addressing this point. revision: partial

Circularity Check

0 steps flagged

Empirical model comparison with no derivation chain reducing to inputs

full rationale

The paper proposes a bipartite GNN architecture (CAtt-BiGNN) and its hierarchical extension, then reports empirical performance on two forming cases. No equations, uniqueness theorems, or first-principles derivations are presented that could reduce to self-definitions, fitted parameters renamed as predictions, or self-citation chains. All claims rest on direct comparison of rollout errors against node-centred baselines and ablations; the central result (improved balance between displacement and thinning) is an observed metric difference, not a quantity forced by construction inside the model. This is the standard non-circular case for an architecture paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The claim rests on the effectiveness of the invented bipartite cross-attention processor and the assumption that geometric edge features suffice to learn the coupling; no external benchmarks or formal derivations are supplied in the abstract.

invented entities (1)
  • CAtt-BiGNN cross-attention processor no independent evidence
    purpose: To learn adaptive bidirectional node-element coupling weights from geometric edge features
    The processor is introduced by the paper as the core novel component; no independent evidence outside the reported experiments is given.

pith-pipeline@v0.9.0 · 5822 in / 1361 out tokens · 28104 ms · 2026-05-25T00:34:12.839722+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · 2 internal anchors

  1. [1]

    Politis, A

    D. Politis, A. Foster, N. Li, L. Wang, J. Lin, A. Foster, D. Szegda, Pre- diction of thinning behavior for complex-shaped, lightweight alloy pan- els formed through a hot stamping process, in: Advanced High Strength Steel and Press Hardening: Proceedings of the 2nd International Con- ference (ICHSU2015), World Scientific, 2016, pp. 395–401

    work page 2016

  2. [2]

    T. J. R. Hughes, The Finite Element Method: Linear Static and Dy- namic Finite Element Analysis, Prentice-Hall, Englewood Cliffs, NJ, 1987

    work page 1987

  3. [3]

    Bathe, Finite Element Procedures, Prentice Hall, 2006

    K. Bathe, Finite Element Procedures, Prentice Hall, 2006. URLhttps://books.google.co.uk/books?id=rWvefGICfO8C

    work page 2006

  4. [4]

    H. R. Attar, A. Foster, N. Li, Development of a deep learning platform for sheet stamping geometry optimisation under manufacturing con- 53 straints, Engineering Applications of Artificial Intelligence 123 (2023) 106295

    work page 2023

  5. [5]

    H. Zhou, H. Li, Y. Zhao, P. R. Childs, N. Li, Image-based artificial intelligence-driven modelling for blank shape optimisation in sheet metal forming, Materials & Design (2025) 114302

    work page 2025

  6. [6]

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, L. Yang, Physics-informed machine learning, Nature Reviews Physics 3 (2021) 422–440. doi:10.1038/s42254-021-00314-5

    work page doi:10.1038/s42254-021-00314-5 2021

  7. [7]

    Y. Zhao, H. Li, H. Zhou, H. R. Attar, T. Pfaff, N. Li, A review of graph neural network applications in mechanics-related domains, Artificial In- telligence Review 57 (11) (2024) 315

    work page 2024

  8. [8]

    arXiv preprint arXiv:2003.049191(1), 1–34 (2020)

    J. Willard, X. Jia, S. Xu, M. Steinbach, V. Kumar, Integrating physics-based modeling with machine learning: A survey (03 2020). doi:10.48550/arXiv.2003.04919

    work page doi:10.48550/arxiv.2003.04919 2020

  9. [9]

    V. K. Dubey, C. E. Haese, O. Gültekin, D. Dalton, M. K. Rausch, J. Fuhg, Graph neural network surrogates for contacting deformable bodies with necessary and sufficient contact detection, Computer Methods in Applied Mechanics and Engineering 448 (2026) 118413. doi:https://doi.org/10.1016/j.cma.2025.118413

    work page doi:10.1016/j.cma.2025.118413 2026

  10. [10]

    L. Xing, P. Gardoni, G. Song, Y. Zhou, Deep learning-based surrogate capacity models and multi-objective fragility es- timates for reinforced concrete frames, Computer Methods in Applied Mechanics and Engineering 440 (2025) 117928. doi:https://doi.org/10.1016/j.cma.2025.117928

    work page doi:10.1016/j.cma.2025.117928 2025

  11. [11]

    Sanchez-Gonzalez, J

    A. Sanchez-Gonzalez, J. Godwin, T. Pfaff, R. Ying, J. Leskovec, P. Battaglia, Learning to simulate complex physics with graph net- works, in: International conference on machine learning, PMLR, 2020, pp. 8459–8468

    work page 2020

  12. [12]

    Pfaff, A

    T. Pfaff, A. Sanchez-Gonzalez, P. W. Battaglia, et al., Learning mesh- based simulation with graph neural networks, International Conference on Learning Representations (2021). 54

    work page 2021

  13. [13]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, P. S. Yu, A comprehensive survey on graph neural networks, IEEE transactions on neural networks and learning systems 32 (1) (2020) 4–24

    work page 2020

  14. [14]

    M. M. Bronstein, J. Bruna, T. Cohen, P. Veličković, Geometric deep learning: Grids, groups, graphs, geodesics, and gauges, arXiv preprint arXiv:2104.13478 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  15. [15]

    Veličković, Everything is connected: Graph neural networks, Current Opinion in Structural Biology 79 (2023) 102538

    P. Veličković, Everything is connected: Graph neural networks, Current Opinion in Structural Biology 79 (2023) 102538

    work page 2023

  16. [16]

    J. Zhou, G. Cui, S. Hu, Z. Zhang, C. Yang, Z. Liu, L. Wang, C. Li, M. Sun, Graph neural networks: A review of methods and applications, AI open 1 (2020) 57–81

    work page 2020

  17. [17]

    Gilmer, S

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, G. E. Dahl, Neural message passing for quantum chemistry, in: International conference on machine learning, Pmlr, 2017, pp. 1263–1272

    work page 2017

  18. [18]

    Dalton, D

    D. Dalton, D. Husmeier, H. Gao, Physics-informed graph neural net- work emulation of soft-tissue mechanics, Computer Methods in Applied Mechanics and Engineering 417 (2023) 116351

    work page 2023

  19. [19]

    Y. Fei, S. Qin, W. Liao, H. Guan, X. Lu, Graph neural network-assisted evolutionary algorithm for rapid optimization design of shear-wall struc- tures, Advanced Engineering Informatics 65 (2025) 103129

    work page 2025

  20. [20]

    Patrignani, S

    L. Patrignani, S. T. Pinho, Graph neural networks with hybrid local- global attention for effective prediction of mechanical response in struc- tures, Computer Methods in Applied Mechanics and Engineering 452 (2026) 118753. doi:https://doi.org/10.1016/j.cma.2026.118753

    work page doi:10.1016/j.cma.2026.118753 2026

  21. [21]

    Y. Zhao, Q. Chen, H. Li, H. Zhou, H. R. Attar, T. Pfaff, T. Wu, N. Li, Recurrent u-net-based graph neural network (rugnn) for accurate de- formation predictions in sheet material forming, Advanced Engineering Informatics 69 (2026) 104021

    work page 2026

  22. [22]

    Y. Zhao, H. Li, H. Zhou, H. R. Attar, T. Pfaff, N. Li, Rapid prediction of material deformation in hot stamping of battery box geometries using graph neural network, in: Journal of Physics: Conference Series, Vol. 3104, IOP Publishing, 2025, p. 012057. 55

    work page 2025

  23. [23]

    Q. Chen, J. Cao, W. Lin, S. Zhu, S. Wang, Predicting dynamic re- sponses of continuous deformable bodies: A graph-based learning ap- proach, Computer Methods in Applied Mechanics and Engineering 420 (2024) 116669

    work page 2024

  24. [24]

    Black, A

    N. Black, A. R. Najafi, Learning finite element convergence with the multi-fidelity graph neural network, Computer Methods in Applied Me- chanics and Engineering 397 (2022) 115120

    work page 2022

  25. [25]

    Battaglia, R

    P. Battaglia, R. Pascanu, M. Lai, D. Jimenez Rezende, et al., Interaction networks for learning about objects, relations and physics, Advances in neural information processing systems 29 (2016)

    work page 2016

  26. [26]

    Sanchez-Gonzalez, N

    A. Sanchez-Gonzalez, N. Heess, J. T. Springenberg, J. Merel, M. Ried- miller, R. Hadsell, P. Battaglia, Graph networks as learnable physics engines for inference and control, in: International conference on ma- chine learning, PMLR, 2018, pp. 4470–4479

    work page 2018

  27. [27]

    X. Fu, F. Zhou, D. Peddireddy, Z. Kang, M. B.-G. Jun, V. Aggarwal, An finite element analysis surrogate model with boundary oriented graph embedding approach for rapid design, Journal of Computational Design and Engineering 10 (3) (2023) 1026–1046

    work page 2023

  28. [28]

    H. Zhou, Y. Zhao, H. Li, T. Pfaff, N. Li, A multi-level graph-based surrogate model for real-time high-fidelity sheet forming simulations, Advanced Engineering Informatics 66 (2025) 103458

    work page 2025

  29. [29]

    Deshpande, S

    S. Deshpande, S. Bordas, J. Lengiewicz, Magnet: A graph u-net ar- chitecture for mesh-based simulations, arXiv preprint arXiv:2211.00713 (2022)

    work page arXiv 2022

  30. [30]

    T. Li, S. Zou, X. Chang, L. Zhang, X. Deng, Finite volume graph net- work (fvgn): Predicting unsteady incompressible fluid dynamics with finite volume informed neural network, arXiv preprint arXiv:2309.10050 (2023)

    work page arXiv 2023

  31. [31]

    R. Gao, I. K. Deo, R. K. Jaiman, A finite element-inspired hypergraph neural network: Application to fluid dynamics sim- ulations, Journal of Computational Physics 504 (2024) 112866. doi:https://doi.org/10.1016/j.jcp.2024.112866. 56

    work page doi:10.1016/j.jcp.2024.112866 2024

  32. [32]

    Belytschko, W

    T. Belytschko, W. K. Liu, B. Moran, K. Elkhodary, Nonlinear finite elements for continua and structures, John wiley & sons, 2014

    work page 2014

  33. [33]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

    work page 2017

  34. [34]

    Graph Attention Networks

    P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y. Bengio, Graph attention networks, arXiv preprint arXiv:1710.10903 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  35. [35]

    Zheng, D

    K. Zheng, D. J. Politis, L. Wang, J. Lin, A review on forming techniques for manufacturing lightweight complex—shaped aluminium panel com- ponents, International Journal of Lightweight Materials and Manufac- ture 1 (2) (2018) 55–80

    work page 2018

  36. [36]

    Zheng, J

    K. Zheng, J. Lee, D. J. Politis, N. Li, J. Lin, T. A. Dean, An analyti- cal investigation on the wrinkling of aluminium alloys during stamping using macro-scale structural tooling surfaces, The International Journal of Advanced Manufacturing Technology 92 (1) (2017) 481–495

    work page 2017

  37. [37]

    N. Li, C. Sun, N. Guo, M. Momhamed, J. Lin, M. Takeki, Damage inves- tigation of boron steel at hot stamping conditions, Procedia Engineering 81 (2014) 1744–1749

    work page 2014

  38. [38]

    Dassault Systèmes, United States of America, Getting Started with Abaqus Interactive Edition (2008)

    work page 2008

  39. [39]

    J. L. D. Z. Marciniak, S. J. Hu, Mechanics of Sheet Metal Forming, Butterworth-Heinemann, Great Britain, 2002

    work page 2002

  40. [40]

    Hosford, R

    W. Hosford, R. Caddell, Metal Forming Mechanics and Metallurgy,

    work page

  41. [41]

    doi:10.1017/CBO9780511811111

    work page doi:10.1017/cbo9780511811111

  42. [42]

    Liang, S

    B. Liang, S. Gao, W. Zhang, An integrated preforming-performance model for high-fidelity performance analysis of cured woven composite part with non-orthogonal yarn angles, Chinese Journal of Aeronautics 35 (6) (2022) 367–378

    work page 2022

  43. [43]

    H. R. Attar, H. Zhou, A. Foster, N. Li, Rapid feasibility assessment of components to be formed through hot stamping: A deep learning approach, Journal of Manufacturing Processes 68 (2021) 1650–1671. doi:https://doi.org/10.1016/j.jmapro.2021.06.011. 57

    work page doi:10.1016/j.jmapro.2021.06.011 2021

  44. [44]

    Mohamed, N

    M. Mohamed, N. Li, L. Wang, O. E. Fakir, J. Lin, T. Dean, J. Dear, An investigation of a new 2d cdm model in predicting failure in hfqing of an automotive panel, in: 4th International Conference on New Forming Technology (ICNFT 2015), EDP Sciences, 2015, p. 05011. 58

    work page 2015