Pith. sign in

REVIEW 4 major objections 5 minor 48 references

PIML-OFEM: A New Large-Scale Structural Analysis Method Based on Problem-Independent Machine Learning and Overlapping Finite Element Technique

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read PIML-OFEM claims that a machine-learned overlapping finite element method can keep only eight corner-node unknowns per substructure while recovering displacement and strain-energy fields close to fine-scale finite element results, at lower

desk verdict A coherent and useful reduced-order FEM/Mech learning method, but the headline accuracy gain over the linear-boundary-interpolation baseline is confounded by an unablated fine-scale PCG residual-correction step. read the letter →

arxiv 2607.22019 v1 pith:NVPRVY4B submitted 2026-07-24 math.NA cs.NAphysics.comp-ph

Yilin Guo , Chang Liu , Zongliang Du , Jin Liu , Jingyu Feng , Xinyang Zhang , Yang Li , Tianxing Yang
show 2 more authors
Changyu Shen Xu Guo
This is my paper · ORCID
classification math.NAcs.NAphysics.comp-ph MSC 65N3065N5574S0568T07
keywords problem-independentmachinelearningoverlappingfiniteelementsoversamplednumericalbasisfunctionssubstructurecondensationU-Nettopologyoptimizationheterogeneousmaterialsreduced-ordermodeling
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

PIML-OFEM tries to settle a longstanding trade-off in substructure-based reduced-order modeling: keeping few degrees of freedom while still resolving complex local deformation. It does so by computing numerical basis functions on an expanded oversampling domain, so no linear displacement assumption is imposed on the target substructure boundary, and by stitching the independent local bases together through an overlapping partition-of-unity formulation that restores global continuity. A U-Net learns the map from local Young's-modulus fields to these oversampled bases, so online local solves are replaced by batched network inference. In the paper's examples this preserves fine-scale displacement and strain-energy accuracy, cuts online analysis time compared with direct finite element analysis, and lets a density-based topology optimization run at high resolution with small filter radii.

What carries the argument

The load-bearing mechanism is the oversampled numerical basis function matrix phi_tilde = Phi (M(Phi))^{-1}, which maps a substructure's eight corner degrees of freedom to all its fine-scale nodal displacements while satisfying Kronecker-delta normalization. Because each basis vector comes from a local boundary value problem on an extended domain, it encodes heterogeneity beyond the substructure boundary. A partition-of-unity weighting w_l = P_l / sum_l P_l blends the independently built local displacement fields into a single continuous global field, and a U-Net, a convolutional encoder-decoder, predicts the local bases from the Young's-modulus field. The overlapped grid remains Cartesian,

What would settle it

Take an intermediate density field from a PIML-OFEM topology-optimization run, freeze it, and compare the U-Net-predicted oversampled basis matrices against exactly computed ones for every substructure; then solve both condensed systems and compare displacement and elemental strain-energy errors with the reported 0.0066-level accuracy. If the predicted basis error grows substantially on such structured fields, or if the optimized design changes character when exact bases are substituted, the central accuracy and stability claims are falsified.

Watch

Extended reading notes

Core claim

The core claim is that the accuracy ceiling of corner-node substructure reduction is set by the boundary interpolation, not by the reduced space itself. By solving local elasticity problems on an extended domain and restricting the solutions to the target substructure, the method obtains eight basis vectors per two-dimensional substructure that can represent bending, twisting, and nearby-void effects without adding boundary degrees of freedom. These oversampled bases are normalized to satisfy Kronecker-delta and rigid-body conditions, then coupled through partition-of-unity weights into a globally continuous displacement field. The numerical results report an average displacement error of 0.

Load-bearing premise

The U-Net was trained only on independent uniform random Young's-modulus fields in [1e-6, 1] and must generalize to the structured, often low-volume-fraction material fields produced by topology optimization; the paper reports no out-of-distribution study for that transition, and if prediction degrades there the reported accuracy and iteration stability would not follow.

Editorial extensions

If this is right

  • Global systems shrink to corner-node degrees of freedom, so analyses with millions of fine elements can be solved in seconds rather than tens of seconds.
  • Elemental strain-energy accuracy in the reported examples is sufficient to support sensitivity-based topology optimization without accumulating interface bias.
  • The same trained U-Net can be reused across loads, boundary conditions, and structural geometries within the training material range, because it predicts local operators rather than global solutions.
  • Density-based topology optimization can proceed with filter radii near the square root of 3 times the element size, preserving fine-scale features that larger filters would erase.
  • The 16-type precomputed stiffness assembly keeps online assembly work linear in element count, so the cost model remains predictable for large-scale problems.

Reading between the lines

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

  • A direct extension the paper leaves implicit is adaptive accuracy control: cheap U-Net predictions could be flagged when local heterogeneity exceeds training support, and only those substructures re-solved exactly; the reported OFEM-SUB comparison gives an upper cost bound for such an adaptive scheme.
  • Because the learning task is material-to-basis rather than problem-to-solution, the same architecture may transfer to other deterministic local coefficients, such as variable Poisson ratio or nonlinear constitutive laws, though the paper does not test this.
  • A risk not addressed by the examples is distribution shift: topology-optimization density fields are highly structured, often with long runs of solid and void, whereas training fields are independent random per element; evaluating the network on intermediate optimization designs would be a direct out-of-distribution check.
  • The dimensional bottleneck is the output size: in three dimensions the basis matrix grows with the number of corner degrees of freedom per substructure, so practical relevance depends on whether convolutional prediction remains accurate and memory-feasible at that scale.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes PIML-OFEM, a reduced-order method for large-scale 2D linear elasticity and topology optimization. Each substructure retains only its four corner nodes (8 DOFs); oversampled numerical basis functions are constructed on extended local domains, normalized to satisfy Kronecker-delta conditions, and coupled into a globally continuous field through a partition-of-unity overlapping finite element formulation. A U-Net is trained on local Young's modulus fields to predict the oversampled basis offline, so that online analysis requires only U-Net inference, assembly of the condensed system, and a small number of PCG residual-correction iterations on the original fine-scale equations. Numerical examples on a cantilever beam, a complex void topology, and high-resolution SIMP topology optimization report close agreement with fine-scale FEM, lower online cost than FEM, and better displacement accuracy than a PIML substructure method with linear boundary interpolation (error 0.0066 vs 0.0199 at comparable cost).

Significance. If fully validated, the combination of oversampled problem-independent bases with overlapping FEM and a U-Net surrogate is a meaningful contribution: it preserves a low-dimensional corner-node representation while avoiding prescribed boundary displacement assumptions, and the offline precomputation of the 16 overlapping element types gives an efficient assembly path. The paper is also commendable for providing reproducible training details (seed 42, dataset sizes, loss terms) and for including an OFEM-SUB baseline that isolates the cost of exact online oversampling. However, the central accuracy and efficiency claims are currently not fully established because the reported PIML-OFEM results include an unspecified fine-scale residual-correction step, and because the U-Net's generalization to the structured, low-volume-fraction fields encountered in topology optimization is not tested out-of-distribution. The paper's framework is plausible, but the evidence as presented is insufficient to support the abstract's attribution of the accuracy gain.

major comments (4)
  1. [§5 (residual correction) and Table 5.1] The text states that after solving the condensed system, the reconstructed displacement is used as an initial guess for 'a small number of residual corrections' with PCG on the original fine-scale equilibrium equations, adding ~1-2 s. Table 5.1 reports PIML-OFEM at 12.06 s with error 0.0066, but no iteration count, stopping tolerance, or ablation is given. If most of the error reduction comes from these fine-scale PCG iterations, the comparison against PIML-EMS-LBC (11.14 s, unstated whether it receives the same correction) does not validate the oversampled basis or U-Net. Please report: (i) PIML-OFEM error and time without residual correction; (ii) PIML-EMS-LBC and OFEM-SUB with the same correction; (iii) PCG iteration count/tolerance and contribution to final error.
  2. [§4.2–§4.4 and §5.4 (generalization)] The U-Net is trained on i.i.d. uniform random modulus fields E~U[10^-6,1] with m=10, l=6, fixed seed, yet it is applied to structured hole/void fields and, in topology optimization, to evolving density fields with large uniform low-stiffness regions and sharp interfaces. The claim in §4.4 that the model is reused across loads, boundaries, and structures without retraining is load-bearing, but no out-of-distribution error analysis is provided. Please report displacement and elemental strain-energy errors (relative to fine-scale FEM) at representative topology-optimization iterations and for the structured fields of §5.2–5.3, with and without residual correction.
  3. [§4.1 (nine models) vs §5 (one model)] Section 4.1 says 'Samples are generated respectively for nine different types of substructures to conduct training, yielding nine machine learning models,' while §5 says 'The same trained model was used in all examples.' This is ambiguous: are nine U-Nets trained, one per substructure type, and then reused? How are boundary-adjacent substructures handled when their oversampling domain extends outside the domain? Please clarify the architecture and training protocol and, if applicable, state which model is used for each of the nine substructure types.
  4. [§5.3, strain energy claim] The claim that 'the relative error of elemental strain energy of the solid area is generally kept at a low level (below 8×10^-3)' is based on a single example, and the text attributes the result to 'oversampled shape functions, overlapping compatibility mechanism, and small amount of residual correction' jointly. Since strain energy is the quantity that enters topology-optimization sensitivities, please provide quantitative strain-energy error statistics for all examples (e.g., mean and 95th percentile) and separate the residual-correction contribution, or the statement remains anecdotal.
minor comments (5)
  1. [Throughout] There are several typographical artifacts: 'vvirtual', 'vinternal', 'vgrouping', '𝐍𝑠' and similar. Please proofread the final text. Also, 'PIML-EMS-LBC' is never expanded; please define 'EMS' and 'LBC' at first use.
  2. [§2.2, Eq. (14)] The invertibility of T = M(phi) is asserted for 'regular substructures' with a note on handling ill-conditioning. A brief numerical study of the condition number of T across the training distribution would be useful, especially for near-void substructures with E_min = 10^-6.
  3. [§2.2, virtual material extrapolation] The mirror-symmetric (fixed boundary) and low-stiffness (free boundary) extrapolation rules are introduced without sensitivity analysis. Since boundary substructures are common in all examples, a short comparison of alternative extrapolations (e.g., constant low-stiffness for both) would strengthen confidence in the method's robustness.
  4. [§3.2, Eq. (23)] The condition that the denominator sum P_l(x) is nonzero is essential; the paper cites [45] for a zero-denominator case. It would be helpful to state explicitly why the proposed overlap pattern (non-uniform substructure mesh) guarantees this for all possible material distributions, or to provide a patch-test verification.
  5. [§5.1, error metrics] The displacement error uses a per-node relative error with epsilon_u = 1e-12; for nodes with near-zero reference displacement this can be large. Reporting also the normalized L2 error or the relative error of the total displacement norm would aid interpretation.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the U-Net is a supervised emulator of local BVPs, and the global accuracy claims are not fitted inputs; only minor non-load-bearing self-citations appear.

full rationale

The derivation chain is self-contained at the level that matters. The oversampled basis functions are defined by solving local elasticity BVPs on extended domains (Sec. 2.2, Eqs. (10)-(16)), and the U-Net is trained offline on those exact local solutions with labels generated by the same BVP workflow (Sec. 4.2). The global displacement is then obtained by a Galerkin projection onto the predicted bases (Eqs. (31)-(33)). No parameter of the reduced model is fitted to the displacement errors reported in Table 5.1, and the reference FEM solutions are not used in training. The claims of problem independence follow from the local nature of the BVP and are not circular. The self-citations to [27-29] and [37] are framework references (PIML paradigm, rigid-body completion); they are not used as a uniqueness theorem or as a substitute for the derivation, so they are not load-bearing. The only step worth flagging is the residual-correction PCG step introduced at the start of Sec. 5: after condensation the reconstructed field is used as an initial guess for 'a small number of residual corrections on the original fine-scale equilibrium equations.' This can confound the attribution of the 0.0066 accuracy gain versus PIML-EMS-LBC if the baseline did not receive the same correction, but it is a solver refinement on the original equations, not a fitted parameter or a definitional equivalence, and it does not make the central claim true by construction. Overall: no circular reduction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 1 invented entities

The central claim rests on standard linear-elastic finite element machinery plus two load-bearing assumptions that are not proven in the paper: (1) the partition-of-unity overlapping construction always yields a valid denominator and linear completeness; (2) the U-Net generalizes from random training fields to structured design fields. The trained network and its data are not released, and the residual-correction step adds an uncontrolled fine-scale solve to the reported accuracy and cost.

free parameters (5)
  • U-Net weights theta = not released; trained on 20,000 samples
    The network parameters are fitted to training data generated by local oversampling solves; no weights are provided.
  • U-Net training hyperparameters = lr=2e-4, weight decay=1e-5, batch size=64, lambda_data=1.0, lambda_K=0.1
    Chosen by hand in Section 4.2; they affect prediction quality and therefore the reported accuracy.
  • Substructure size m and oversampling layers l = m=10, l=6
    The model is trained only for these values (Section 4.2/5), so all numerical claims apply to this configuration.
  • Uniform-modulus reuse threshold = 10^-3
    Substructures with modulus range below this threshold bypass the U-Net and reuse precomputed uniform bases (Section 4.4).
  • Residual-correction stopping criterion = not specified
    The PCG residual-correction step in Section 5 is described only as a 'small number' of iterations with '1-2 s' cost; no tolerance or iteration count is given.
assumptions (6)
  • domain assumption Two-dimensional linear elasticity with small deformations and fixed Poisson's ratio is an adequate model for the target applications.
    The entire method is built on this setting; Section 6 lists 3D, geometric nonlinearity, material nonlinearity, and dynamics as future work.
  • domain assumption The partition-of-unity weights defined in Eq. (23) satisfy non-negativity and sum to one, and the constructed overlapping meshes avoid zero denominators.
    Used to build the globally continuous displacement field in Section 3.2; relies on the overlapping-FEM framework of Bathe et al. [40-48].
  • ad hoc to paper The matrix T = M(phi) in Eq. (14) is invertible for the substructures encountered in the numerical examples.
    The Kronecker-delta normalization requires C = T^{-1}; the paper states ill-conditioned cases must be handled by regularization or sample removal.
  • ad hoc to paper The U-Net trained on i.i.d. uniform random modulus fields generalizes to structured heterogeneous and topology-optimized fields within the training modulus range.
    Assumed in the online workflow (Section 4.4) and by the statement that no model is retrained for new structures, loads, or boundary conditions.
  • ad hoc to paper Virtual material extrapolation outside the global boundary (mirror-symmetric for fixed boundaries, low-stiffness for free boundaries) does not bias the oversampled bases.
    Invoked in Section 2.2 when oversampling domains extend beyond the physical boundary; the paper gives no quantitative evidence of negligible bias.
  • ad hoc to paper The post-processing PCG residual correction converges sufficiently in a small number of iterations that the reported accuracy and timing reflect the proposed method.
    Introduced in Section 5 before the examples; no tolerance or iteration count is reported, so the contribution of the correction to the quoted errors is unknown.
invented entities (1)
  • Virtual material extrapolation E_vir^j
    purpose: Fills the oversampling layers that extend beyond the global structural boundary when constructing local numerical basis functions.
    A numerical device only; the paper states it does not alter the true material distribution or boundary conditions. It has no falsifiable handle outside the method.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PIML-OFEM: A New Large-Scale Structural Analysis Method Based on Problem-Independent Machine Learning and Overlapping Finite Element Technique." pith.science (2026). https://pith.science/paper/NVPRVY4B

@misc{pith2026260722019,
  author       = {Pith},
  title        = {Pith review of: PIML-OFEM: A New Large-Scale Structural Analysis Method Based on Problem-Independent Machine Learning and Overlapping Finite Element Technique},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NVPRVY4B}},
  note         = {Machine review of arXiv:2607.22019}
}
read the original abstract

High-resolution analysis and design of large-scale heterogeneous structures require accurate reduced-order models and efficient online computation. Existing multiscale methods must repeatedly construct local basis functions for different material distributions, whereas substructure-based problem-independent machine learning (PIML) methods can be limited by prescribed boundary displacement interpolation. We propose PIML-OFEM, an overlapping finite element method accelerated by problem-independent machine learning. Each substructure retains only its corner-node degrees of freedom. Oversampled numerical basis functions are constructed by solving local elasticity problems on extended domains and restricting the solutions to the target substructure, eliminating prescribed displacement interpolation on its boundary. Independently constructed local bases are blended through a partition-of-unity overlapping formulation to obtain a globally continuous displacement field. A U-Net learns the mapping from local Young's modulus distributions to numerical basis functions, replacing repeated online local solves and allowing the model to be reused across load cases and global boundary conditions. Numerical examples show close agreement with fine-scale finite element results in displacement and elemental strain energy. PIML-OFEM reduces online computational cost relative to direct finite element analysis and improves accuracy over PIML substructure models based on linear boundary interpolation. In topology optimization, the method supports stable high-resolution iterations with small filter radii and preserves fine-scale features, including local patterns resembling rank-2 microstructures. The framework provides an efficient physics-data approach for large-scale heterogeneous structural analysis and high-resolution topology optimization.

Figures

Figures reproduced from arXiv: 2607.22019 by the authors.

Figure 13
Figure 13. Fig.13 [PITH_FULL_IMAGE:figures/full_fig_p041_13.png] view at source ↗
Figure 14
Figure 14. Comparison of cantilever beam deformation under different discretization and reduction strategies. (a) Result of classical FEM analysis. (b) Result of the PIML method based on the linear boundary displacement assumption for substructures [27-28]. (c) Result of PIML-OFEM analysis, in which the overall structure is discretized into 8×4 mutually overlapping substructures, each containing 10×10 fine-scale elements; the … view at source ↗
Figure 15
Figure 15. Fig.15 [PITH_FULL_IMAGE:figures/full_fig_p044_15.png] view at source ↗
Figures from the paper (4 more)
Figure 16
Figure 16. Figure 16: Fig.16 [PITH_FULL_IMAGE:figures/full_fig_p045_16.png]
Figure 17
Figure 17. Figure 17: Fig.17 [PITH_FULL_IMAGE:figures/full_fig_p047_17.png]
Figure 18
Figure 18. Figure 18: Fig.18 [PITH_FULL_IMAGE:figures/full_fig_p049_18.png]
Figure 19
Figure 19. Figure 19: (b-c) respectively give the optimization results obtained with PIML-OFEM at filter radii of 2 times and √3 times the fine element size. The average single-step analysis time of PIML-OFEM is about 33 s, significantly lower than classical FEM, while also being able to o…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 16 canonical work pages

  1. [1]

    N. Aage, E. Andreassen, B. S. Lazarov, and O. Sigmund, Giga -voxel computational morphogenesis for structural design, Nature 550, 84 (2017). https://doi.org/10.1038/nature23911

    doi:10.1038/nature23911 2017

  2. [2]

    Duraisamy, G

    K. Duraisamy, G. Iaccarino, and H. Xiao, Turbulence Modeling in the Age of Data, Annu. Rev. Fluid Mech. 51, 357 (2019). https://doi.org/10.1146/annurev-fluid-010518-040547

    doi:10.1146/annurev-fluid-010518-040547 2019

  3. [3]

    J. Hu, Z. Lu, and Y . Yang, Improving prediction of preferential concentration in particle-laden turbulence using the neural -network interpolation, Phys. Rev. Fluids 9, 034606 (2024). https://doi.org/10.1103/PhysRevFluids.9.034606

    work page doi:10.1103/physrevfluids.9.034606 2024

  4. [4]

    Y . Wang, Z. Li, Z. Y uan, W. Peng, T. Liu, and J. Wang, Prediction of turbulent channel flow using Fourier neural operator-based machine-learning strategy, Phys. Rev. Fluids 9, 084604 (2024). https://doi.org/10.1103/PhysRevFluids.9.084604

    doi:10.1103/physrevfluids.9.084604 2024

  5. [5]

    K. Guo, Z. Yang, C. H. Yu, and M. J. Buehler, Artificial intelligence and machine learning in design of mechanical materials, Mater. Horiz. 8, 1153 (2021). https://doi.org/10.1039/D0MH01451F

    work page doi:10.1039/d0mh01451f 2021

  6. [6]

    H. Jin, E. Zhang, and H. D. Espinosa, Recent Advances and Applications of Machine Learning in Experimental Solid Mechanics: A Review, Appl. Mech. Rev. 75, 061001 (2023). https://doi.org/10.1115/1.4062966

    work page doi:10.1115/1.4062966 2023

  7. [7]

    C. S. Ha, D. Yao, Z. Xu, C. Liu, H. Liu, D. Elkins, M. Kile, V . Deshpande, Z. Kong, M. Bauchy, and X. Zheng, Rapid inverse design of metamaterials based on prescribed mechanical behavior through machine learning, Nat. Commun. 14, 5765 (2023). https://doi.org/10.1038/s41467-023- 40854-1

    doi:10.1038/s41467-023- 2023

  8. [8]

    Deshpande, J

    S. Deshpande, J. Lengiewicz, and S. P . A. Bordas, Probabilistic deep learning for real-time large deformation simulations, Comput. Methods Appl. Mech. Eng. 398, 115307 (2022). https://doi.org/10.1016/j.cma.2022.115307

    arXiv 2022

Show all 48 references
  1. [10]

    P. T. Nguyen, Y . Heider, D. M. Kochmann, and F. Aldakheel, Deep learning -aided inverse design of porous metamaterials, Comput . Methods Appl. Mech. Eng. 449, 118499 (2026). https://doi.org/10.1016/j.cma.2025.118499

    2026

  2. [11]

    X. Lei, C. Liu, Z. Du, W. Zhang, and X. Guo, Machine Learning-Driven Real-Time Topology Optimization Under Moving Morphable Component-Based Framework, J. Appl. Mech. 86, 011004 (2019). https://doi.org/10.1115/1.4041319

    2019 doi

  3. [12]

    D. Geng, J. Yan, Q. Xu, Q. Zhang, M. Zhou, Z. Fan, and H. Li, Real-Time structure topology optimization using CNN driven Moving Morphable component method, Eng. Struct. 290, 116376 (2023). https://doi.org/10.1016/j.engstruct.2023.116376

    2023

  4. [13]

    S. Cai, Z. Mao, Z. Wang, M. Yin, and G. E. Karniadakis, Physics -informed neural networks (PINNs) for fluid mechanics: a review, Acta Mech. Sin. 37, 1727 (2021). https://doi.org/10.1007/s10409-021-01148-1

    2021 doi

  5. [14]

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang, Physics - informed machine learning, Nat. Rev. Phys. 3, 422 (2021). https://doi.org/10.1038/s42254-021- 00314-5

    2021 doi

  6. [15]

    J. Song, W. Cao, F. Liao, and W. Zhang, VW -PINNs: A volume weighting method for PDE residuals in physics -informed neural networks, Acta Mech. Sin. 41, 324140 (2025). https://doi.org/10.1007/s10409-024-24140-x

    2025 doi

  7. [16]

    L. Lu, P. Jin, G. Pang, Z. Zhang, and G. E. Karniadakis, Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators, Nat. Mach. Intell. 3, 218 (2021). https://doi.org/10.1038/s42256-021-00302-5

    2021 doi

  8. [17]

    L. Lu, X. Meng, Z. Mao, and G. E. Karniadakis, DeepXDE: A Deep Learning Library for Solving Differential Equations, SIAM Rev. 63, 208 (2021). https://doi.org/10.1137/19M1274067

    2021 doi

  9. [18]

    L. Lu, R. Pestourie, W. Yao, Z. Wang, F. Verdugo, and S. G. Johnson, Physics-Informed Neural Networks with Hard Constraints for Inverse Design, SIAM J. Sci. Comput . 43, B1105 (2021). https://doi.org/10.1137/21M1397908

    2021 doi

  10. [19]

    S. Saha, Z. Gan, L. Cheng, J. Gao, O. L. Kafka, X. Xie, H. Li, M. Tajdari, H. A. Kim, and W. K. Liu, Hierarchical Deep Learning Neural Network (HiDeNN): An artificial intelligence (AI) framework for computational science and engineering, Comput. Metho ds Appl. Mech. Eng. 373, ...

    2021

  11. [20]

    Zhang, Y

    L. Zhang, Y . Lu, S. Tang, and W. K. Liu, HiDeNN -TD: Reduced -order hierarchical deep learning neural networks, Comput. Methods Appl. Mech. Eng. 389, 114414 (2022). https://doi.org/10.1016/j.cma.2021.114414

    2022

  12. [21]

    H. Li, S. Knapik, Y . Li, C. Park, J. Guo, S. Mojumder, Y . Lu, W. Chen, D. W. Apley, and W. K. Liu, Convolution Hierarchical Deep-Learning Neural Network Tensor Decomposition (C-HiDeNN- TD) for high -resolution topology optimization, Comput . Mech. 72, 363 (2023). https://doi...

    2023 doi

  13. [22]

    Y . Lu, H. Li, L. Zhang, C. Park, S. Mojumder, S. Knapik, Z. Sang, S. Tang, D. W. Apley, G. J. Wagner, and W. K. Liu, Convolution Hierarchical Deep -learning Neural Networks (C -HiDeNN): finite elements, isogeometric analysis, tensor decomposition, and beyond, Comput. Mech. 72...

    2023 doi

  14. [23]

    Ramuhalli, L

    P. Ramuhalli, L. Udpa, and S. S. Udpa, Finite -Element Neural Networks for Solving Differential Equations, IEEE Trans. Neural Netw. 16, 1381 (2005). https://doi.org/10.1109/TNN.2005.857945

    2005

  15. [24]

    C. Wu, C. Liu, Y . Guo, and X. Guo, DFENN: A penalty-free variational framework coupling finite elements and neural networks via interface condensation, J. Mech. Phys. Solids 215, 106703 (2026). https://doi.org/10.1016/j.jmps.2026.106703

    2026

  16. [25]

    F. V . Senhora, H. Chi, Y . Zhang, L. Mirabella, T. L. E. Tang, and G. H. Paulino, Machine learning for topology optimization: Physics-based learning through an independent training strategy, Comput. Methods Appl. Mech. Eng. 398, 115116 (2022). https://doi.org/10.1016/j.cma.20...

    2022

  17. [26]

    H. Chi, Y . Zhang, T. L. E. Tang, L. Mirabella, L. Dalloro, L. Song, and G. H. Paulino, Universal machine learning for topology optimization, Comput. Methods Appl. Mech. Eng. 375, 112739 (2021). https://doi.org/10.1016/j.cma.2019.112739

    2021

  18. [27]

    Huang, Z

    M. Huang, Z. Du, C. Liu, Y . Zheng, T. Cui, Y . Mei, X. Li, X. Zhang, and X. Guo, Problem- independent machine learning (PIML) -based topology optimization —A universal approach, Extreme Mech. Lett. 56, 101887 (2022). https://doi.org/10.1016/j.eml.2022.101887

    2022

  19. [28]

    Huang, T

    M. Huang, T. Cui, C. Liu, Z. Du, J. Zhang, C. He, and X. Guo, A Problem-Independent Machine Learning (PIML) enhanced substructure -based approach for large -scale structural analysis and topology optimization of linear elastic structures, Extreme Mech. Lett. 63, 102041 (2023)....

    2023

  20. [29]

    Huang, C

    M. Huang, C. Liu, Y . Guo, L. Zhang, Z. Du, and X. Guo, A mechanics-based data-free Problem Independent Machine Learning (PIML) model for large -scale structural analysis and design optimization, J. Mech. Phys. Solids 193, 105893 (2024). https://doi.org/10.1016/j.jmps.2024.105893

    2024

  21. [30]

    Y . Guo, Z. Du, C. Liu, W. Zhang, S. Tang, W. Hao, Z. Zhao, C. Wu, Z. Xu, T. Bian, Z. Dai, Z. Yang, W. Huo, C. Shen, and X. Guo, AI -enhanced and data -driven mechanical analysis and structural topology optimization, Sci. China Technol. Sci. (2026). https://doi.org/10.1007/s11...

    2026 doi

  22. [31]

    R. R. Craig Jr. and M. C. C. Bampton, Coupling of substructures for dynamic analyses, AIAA J. 6, 1313 (1968). https://doi.org/10.2514/3.4741

    1968 doi

  23. [32]

    J. A. Gutierrez and A. K. Chopra, A substructure method for earthquake analysis of structures including structure‐soil interaction, Earthq. Eng. Struct. Dyn. 6, 51 (1978). https://doi.org/10.1002/eqe.4290060107

    1978 doi

  24. [33]

    X. Ma, M. Huang, Z. Du, Y . Guo, C. Liu, Y . Mei, and X. Guo, A high-performance parallel algorithm based on problem independent machine learning (PIML) for large -scale topology optimization, Acta Mech. Sin. 42, 425942 (2026). https://doi.org/10.1007/s10409-025-25942-x

    2026 doi

  25. [34]

    C. Liu, W. Xu, W. Huo, Y . Guo, and X. Guo, Surface lattice structure design via computational conformal mapping and structural optimization, Comput. Methods Appl. Mech. Eng. 451, 118680 (2026). https://doi.org/10.1016/j.cma.2025.118680

    2026

  26. [35]

    X. Cao, W. Xu, S. Zhao, Y . Guo, C. Liu, and X. Guo, Design of 3D Shell -Graded Infill Structures Based on Moving Morphable V oid and Triply Periodic Minimal Surfaces, In: X. Feng and K. Zhou (eds.), Computational and Experimental Simulations in Engineer ing: Proceedings of IC...

    2025 doi

  27. [36]

    Zhang, M

    L. Zhang, M. Huang, C. Liu, Z. Du, T. Cui, and X. Guo, Problem -independent machine learning-enhanced structural topology optimization of complex design domains based on isoparametric elements, Extreme Mech. Lett. 72, 102237 (2024). https://doi.org/10.1016/j.eml.2024.102237

    2024

  28. [37]

    Y . Guo, C. Liu, Z. Du, Y . Jia, C. Jiang, X. Guo, and C. Shen, High-Generalization AI-Enhanced mechanical analysis and topology optimization via cubic bézier interpolation of substructure boundary displacements, Comput. Methods Appl. Mech. Eng. 456, 118955 (2026). https://doi...

    2026

  29. [38]

    H. W. Zhang, Y . Liu, S. Zhang, J. Tao, J. K. Wu, and B. S. Chen, Extended multiscale finite element method: its basis and applications for mechanical analysis of heterogeneous materials, Comput. Mech. 53, 659 (2014). https://doi.org/10.1007/s00466-013-0924-x

    2014 doi

  30. [39]

    H. W. Zhang, J. K. Wu, J. Lü, and Z. D. Fu, Extended multiscale finite element method for mechanical analysis of heterogeneous materials, Acta Mech. Sin. 26, 899 (2010). https://doi.org/10.1007/s10409-010-0393-9

    2010 doi

  31. [40]

    K. J. Bathe and L. Zhang, The finite element method with overlapping elements – A new paradigm for CAD driven simulations, Comput. Struct. 182, 526 (2017). https://doi.org/10.1016/j.compstruc.2016.10.020

    2017 doi

  32. [41]

    W. L. Nicomedes and K. J. Bathe, An effective overlapping finite element for three-dimensional incompressible linear elasticity, Comput. Methods Appl. Mech. Eng. 457, 118613 (2026). https://doi.org/10.1016/j.cma.2025.118613

    2026

  33. [42]

    W. L. Nicomedes, K. J. Bathe, F. J. S. Moreira, and R. C. Mesquita, Overlapping finite elements for the Navier -Stokes equations, Comput. Struct. 299, 107343 (2024). https://doi.org/10.1016/j.compstruc.2024.107343

    2024

  34. [43]

    Lee and K

    S. Lee and K. J. Bathe, An enhancement of overlapping finite elements, Comput. Struct. 260, 106704 (2022). https://doi.org/10.1016/j.compstruc.2021.106704

    2022

  35. [44]

    Huang and K

    J. Huang and K. J. Bathe, On the convergence of overlapping elements and overlapping meshes, Comput. Struct. 244, 106429 (2021). https://doi.org/10.1016/j.compstruc.2020.106429

    2021

  36. [45]

    Huang and K

    J. Huang and K. J. Bathe, Overlapping finite element meshes in AMORE, Adv. Eng. Softw. 144, 102791 (2020). https://doi.org/10.1016/j.advengsoft.2020.102791

    2020

  37. [46]

    Huang and K

    J. Huang and K. J. Bathe, Quadrilateral overlapping elements and their use in the AMORE paradigm, Comput. Struct. 222, 25 (2019). https://doi.org/10.1016/j.compstruc.2019.05.011

    2019 doi

  38. [47]

    Zhang, K

    L. Zhang, K. T. Kim, and K. J. Bathe, The new paradigm of finite element solutions with overlapping elements in CAD – Computational efficiency of the procedure, Comput. Struct. 199, 1 (2018). https://doi.org/10.1016/j.compstruc.2018.01.003

    2018 doi

  39. [48]

    Zhang and K

    L. Zhang and K. J. Bathe, Overlapping finite elements for a new paradigm of solution, Comput. Struct. 187, 64 (2017). https://doi.org/10.1016/j.compstruc.2017.03.008

    2017 doi

  40. [49]

    J. P. Groen and O. Sigmund, Homogenization‐based topology optimization for high‐resolution manufacturable microstructures, Int. J. Numer. Methods Eng. 113, 1148 (2018). https://doi.org/10.1002/nme.5575 PIML-OFEM：一种基于问题无关机器学习与重叠有限元技术的新 型大规模结构分析方法 郭一麟，刘畅，杜宗亮，刘进，冯靖宇，张欣阳，李阳，杨添行， 申...

    2018 doi

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.