Pith. sign in

REVIEW 3 major objections 5 minor 45 references

Inference of Substructured Reduced-Order Models for Dynamic Contact from Contact-free Simulations

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A reduced-order contact model can be learned without ever simulating contact.

desk verdict A promising non-intrusive route to contact ROMs that needs a careful repair pass before it is reproducible. read the letter →

arxiv 2505.18050 v1 pith:5C2FIGKN submitted 2025-05-23 math.NA cs.NA

classification math.NAcs.NA MSC 74M1565M6065K05
keywords modelorderreductionoperatorinferencedynamiccontactmechanicsCraig-BamptonsubstructuringlinearcomplementarityproblemLagrangemultiplierspressurenon-intrusive
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that fast reduced-order models for dynamic contact problems—structures hitting an obstacle—can be built from simulation data in which no contact occurs at all. The authors combine operator inference, which fits small system matrices to snapshot data, with Craig-Bampton-style substructuring, which keeps the potential contact nodes at full fidelity and compresses only the interior. They then switch to the dual contact-pressure formulation, a linear complementarity problem solved with Lemke's algorithm. The payoff, if the claim holds, is that engineers can obtain accurate displacements and contact pressures in near-real time without access to the finite-element solver's internal matrices and without paying for expensive contact simulations during the training phase.

What carries the argument

The central object is the Craig-Bampton-like reduction ansatz $q \approx V\,(q_B^T,\hat q_I^T)^T$ with $q_I \approx \Phi_{IB} q_B + V_I \hat q_I$, where $q_B$ are the boundary (potential contact) degrees of freedom kept in full dimension and $V_I$ is a POD basis for the interior. The coupling matrix $\Phi_{IB}$ carries the argument: it maps boundary displacements to interior displacements and, when identified exactly through static unit-displacement simulations, it is what makes the dual-system Lagrange multipliers (contact pressures) accurate. The dual system is a linear complementarity problem (LCP), written as $B + A\lambda \ge 0$, $\lambda \ge 0$, $\lambda^T(B+A\lambda)=0$, with matrix $\hat A = h^2 C_B(\hat M + h^2\hat K)^{-1} C_B^T$. Because the inferred $\hat M$, $\hat K$ are symmetric positive definite, so is $\hat A$, and the LCP has a unique solution for every right-hand side.

What would settle it

Take a test structure with strong inertial boundary–interior coupling (for example a stiff, heavy interior attached to a light boundary), train the reduced model on contact-free snapshots using all three coupling strategies, then solve a contact scenario and compare the Lagrange multipliers against the full-order contact solution. If the static-simulation coupling matrix also produces contact-force errors far above 1%, or if the least-squares variants cannot be improved by enriching the training data, the claim that contact-free data are sufficient for accurate contact pressures would be refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that a substructured reduced-order model inferred from contact-free snapshots can reproduce the contact response of the full-order model, provided the boundary–interior coupling matrix is recovered well. On the beam example, the variant that computes the coupling matrix from successive static unit-displacement simulations reaches less than 1% relative error in the contact forces, while the two least-squares estimates of the coupling matrix produce much larger contact-force errors even though displacements remain accurate. The authors interpret this as evidence that the coupling matrix, not the interior reduction basis, is the critical ingredient for contact pressure accuracy. The inferred reduced matrices are forced to be symmetric positive definite, which makes the reduced linear complementarity problem well posed and uniquely solvable by Lemke's method.

Load-bearing premise

The load-bearing assumption is that the interior response is the sum of a static coupling response to boundary motion plus the response of the fixed-boundary interior system to the external load, with the inertial coupling between boundary and interior neglected; the least-squares coupling estimates rest directly on this split, and they are the ones that give poor contact-force accuracy.

Editorial extensions

If this is right

  • Contact simulations become unnecessary for ROM training: only contact-free response data and static unit-displacement solves are needed.
  • Contact pressures can be recovered in near-real time because the LCP size equals the number of contact nodes, which is small compared with the full displacement dimension.
  • The substructured ROM preserves the contact interface exactly, so the reduced model can be used for contact detection and search at low cost even when the coupling estimate is only approximate.
  • The SPD constraints on the inferred matrices guarantee a well-posed reduced contact problem, not just a stable reduced dynamics.
  • The method transfers to 3D finite-element models and to loading frequencies different from the training frequency, as the two test cases show.

Reading between the lines

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

  • The large gap in contact-force accuracy between the static-simulation coupling and the least-squares couplings suggests that the inertial term neglected in the approximation of the coupling matrix is precisely what controls the dual variable; a testable extension would add an inertial correction term to the least-squares fit and see whether contact-force errors drop.
  • For structures with large possible contact regions, the exact static-mode strategy becomes expensive; a hybrid that uses least-squares coupling for a first pass and then refines only the active contact nodes with static modes could keep the cost down while preserving accuracy.
  • The same substructured inference idea could be applied to node-to-segment contact or frictional contact, but the LCP would have to be replaced by a more general complementarity or nonsmooth system, so the uniqueness argument based on an SPD LCP matrix would no longer apply directly.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a non-intrusive model-order-reduction method for dynamic node-to-node contact in linear elasticity. The method infers a Craig-Bampton-like substructured ROM from contact-free simulation snapshots, using operator inference with symmetry and positive-definiteness constraints, and then enforces contact by solving a dual linear complementarity problem with Lemke's algorithm. Three ways of identifying the boundary--interior coupling matrix are proposed: a full-dimensional least-squares fit, a reduced-dimensional least-squares fit, and a static unit-displacement simulation approach. The method is demonstrated on two three-dimensional finite-element examples, a console beam and a piston rod, with the static-simulation coupling giving the most accurate contact forces.

Significance. If the central claim holds, the paper offers a practical route to building contact-capable reduced models from data that contain no active contact, which is valuable for digital-twin applications in structural mechanics. The combination of Craig-Bampton substructuring with operator inference is a sensible and useful extension, and the static-coupling variant achieves a genuine out-of-sample prediction, with below 1% relative contact-force error in the beam example. The paper also provides code and data availability, which is a strength for reproducibility. The main limitations are the unquantified superposition assumption behind Eq. (41) and the fact that, in the reported experiments, only the static-simulation coupling variant attains acceptable contact-force accuracy, while the two least-squares variants do not.

major comments (3)
  1. [§4.2.3, Eq. (45) and Algorithm 1] The reduced force snapshot matrix is defined inconsistently with the stated transformation. Section 4.1 states that bf = V^T f, and with V in Eq. (30) this gives bF_B = F_B + Φ_IB^T F_I. However, Eq. (45) and Algorithm 1, line 4, write bF_B = F_B + Φ_IB F_I^(1). Besides the missing transpose, the printed expression is dimensionally incompatible because Φ_IB is n_I × n_B while F_I is n_I × k. This is load-bearing for reproducibility, since the training data for the operator-inference problem depend on this definition; please correct the formula and ensure the algorithm listing matches the code used to produce the numerical results.
  2. [§4.2.2, Eq. (41)] The load-bearing modeling assumption Q_I ≈ Φ_IB Q_B + Q_I^(1) is introduced without a quantitative validity check. In an exact Craig-Bampton decomposition, the dynamic interior coordinates satisfy an equation forced by -(M_II Φ_IB + M_IB) q¨_B, so the fixed-boundary response Q_I^(1) does not equal the dynamic part of the full interior response when boundary accelerations are significant. Since Eq. (41) is used both to fit Φ_IB in Eqs. (42)-(43) and to define the interior training data bQ_I^(1) in Eq. (45) and Algorithm 1, the inferred reduced model can be systematically biased even when Φ_IB itself is exact. The paper should report a residual or error measure for Eq. (41) on the training data, such as ||Q_I - Φ_IB Q_B - Q_I^(1)||_F / ||Q_I||_F, and state a condition under which the approximation is valid.
  3. [§5.1, Fig. 7] The numerical evidence narrows the general claim made in the abstract and conclusions: the two least-squares coupling estimates produce contact-force errors that the text itself calls 'not acceptable' (Fig. 7b), while only the static-simulation coupling reaches below 1% relative error. Because accurate contact forces are the advertised advantage of the adjoint/LCP formulation, the paper should present the static-simulation variant as the main validated method and the least-squares variants as auxiliary tools, or provide a criterion for when the latter can be trusted.
minor comments (5)
  1. [§4.1, Eq. (34)] Immediately after Eq. (34), the text says q_B, f_B ∈ R^{n_I}; this should be R^{n_B}.
  2. [§4.2.2, Eq. (42)] The least-squares objective in Eq. (42) should be ||Q_I - Q_I^(1) - Φ_IB Q_B||_F^2; as printed, the plus sign before Q_I^(1) changes the problem. The same sign issue appears in Eq. (43).
  3. [§4.2.2, Eq. (43)] The snapshot matrices Q_I^(2) and Q_B^(2) in Eq. (43) are not defined; presumably they are the interior and boundary blocks of a data set used for the reduced-space least-squares fit, but this should be stated explicitly.
  4. [§5.1] The load frequency is written as '0 , 32 Hz' with a comma as decimal separator, which is inconsistent with the rest of the text.
  5. [Fig. 4 caption] The phrase 'interior subsystem snapshot set of the full-order system simulation with fixed boundary nodes' is redundant and confusing; it should simply say the snapshot set from the fixed-boundary simulation (38).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: contact forces are out-of-sample outputs of an LCP solved with reduced operators fitted to contact-free snapshots; cited prior work is re-derived rather than load-bearing.

full rationale

The paper's inference chain is: contact-free snapshots are used to build a Craig-Bampton-style basis and coupling matrix, reduced operators are fitted by constrained least squares, and contact pressures are then obtained by solving a linear complementarity problem in the dual variables. The displacements and Lagrange multipliers reported in Section 5 are not among the fitted quantities: the LCP matrix and right-hand side are assembled from the fitted reduced operators and the current external load, and Lemke's method computes lambda from complementarity conditions that were never used as training data. Thus the central contact-force prediction is a genuine out-of-sample output, not a renamed fit. The use of the authors' prior work [34] for primal-dual decoupling is not load-bearing because Section 2.2 re-derives the LCP from the KKT conditions, and the uniqueness claim is supported by the standard reference [12] together with the enforced SPD constraints. Equation (41), QI ≈ ΦIB QB + QI^(1), is a physical approximation that neglects inertial coupling between boundary and interior degrees of freedom; the paper's own Fig. 7 shows that contact-force accuracy is sensitive to how ΦIB is obtained, so this is a correctness and robustness limitation rather than a circular definition. The 'exact' coupling matrix from static simulations is the standard constraint-mode definition from Craig-Bampton substructuring, not a fitted parameter that is later reported as a prediction. Overall, no step of the derivation reduces by construction to its own inputs, and the self-citations are contextual rather than load-bearing.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The method introduces no new physical entities. Its free parameters are the reduced dimension, the training load, the boundary node set, and the time step and snapshot count, all chosen by the user and not determined by the theory. The main load-bearing assumptions are the linear-elastic node-to-node contact setting, the a priori knowledge of the contact node set, the superposition approximation in Eq. (41), and standard LCP theory.

free parameters (5)
  • Interior reduced dimension r = 2 (both examples)
    Chosen from the SVD decay of the fixed-boundary interior snapshots; controls ROM accuracy and is not determined by the theory.
  • Training load (harmonic force parameters) = beam: 0.16 Hz, 3 kN; piston rod: not fully specified
    The contact-free training data are generated under a user-selected load; the method's extrapolation to other loads is validated but not guaranteed.
  • Boundary/contact node set = 6 nodes for beam; not stated for piston rod
    The maximum possible contact node set must be known a priori and is preserved exactly; the method's efficiency depends on it being small.
  • Implicit Euler time step h = not reported in text
    The LCP matrix and the discrete ROM depend on h; the numerical sections do not state the value.
  • Snapshot count k = not reported in text
    The number of snapshots affects the least-squares inference; not stated in the numerical sections.
assumptions (6)
  • domain assumption Linear elasticity, small deformations, frictionless adhesion-free node-to-node contact.
    Stated in Section 2.1.1; the entire method is scoped to this setting.
  • domain assumption The semi-discretized system is M q_ddot + K q = f with M and K symmetric positive definite.
    Used for the LCP matrix A = h^2 C(M+h^2K)^{-1}C^T to be SPD and for unique solvability; Section 2.2 and Section 4.3.
  • domain assumption The maximum possible set of contact nodes is known a priori.
    Listed as a basic assumption in Section 4.1.
  • ad hoc to paper The interior response decomposes as QI ≈ Φ_IB QB + Q_I^(1), where Q_I^(1) is a fixed-boundary simulation with the same external force.
    Introduced in Eq. (41); neglects boundary-interior inertia coupling and underpins the LSQ coupling estimates.
  • ad hoc to paper The reduced operators can be inferred by the constrained least squares in Eq. (46), with cM, cK SPD and interior blocks fixed to those of the fixed-boundary subsystem.
    The paper does not prove that the true Craig-Bampton projection satisfies the equality constraints exactly with noisy data.
  • standard math Standard LCP theory: an LCP with an SPD matrix has a unique solution for every right-hand side.
    Used in Section 4.3 to assert convergence of Lemke's method; see [12].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Inference of Substructured Reduced-Order Models for Dynamic Contact from Contact-free Simulations." pith.science (2026). https://pith.science/paper/5C2FIGKN

@misc{pith2026250518050,
  author       = {Pith},
  title        = {Pith review of: Inference of Substructured Reduced-Order Models for Dynamic Contact from Contact-free Simulations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5C2FIGKN}},
  note         = {Machine review of arXiv:2505.18050}
}
read the original abstract

In this paper, we propose an operator-inference-based reduction approach for contact problems, leveraging snapshots from simulations without active contact. Contact problems are solved using adjoint methods, by switching to the dual system, where the corresponding Lagrange multipliers represent the contact pressure. The Craig-Bampton-like substructuring method is incorporated into the inference process to provide the reduced system matrices and the coupling of the contact and interior nodes. The maximum possible set of contact nodes must be known a priori. Characteristic properties of the inferred matrices, such as symmetry and positive definiteness, are enforced by appending additional constraints to the underlying least-squares problem. The resulting dual system, which forms a linear complementarity problem, is well-defined and can be effectively solved using methods such as Lemke's algorithm. The performance of the proposed method is validated on three-dimensional finite element models.

Figures

Figures reproduced from arXiv: 2505.18050 by the authors.

Figure 1
Figure 1. Static mode. In this case, the identification of the coupling term is exact. If the assumption about the small size of the contact region holds, the calculation of the coupling term through consecutive static simulations is not only highly accurate but also very efficient. However, as the number of contact nodes increases, the effectiveness of this method is decreasing drastically. 4.2.3 Optimization problem for sys… view at source ↗
Figure 2
Figure 2. FEM model of a console beam with a rigid plane obstacle. [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Training snapshots for the console beam example. [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Console beam example: SVD decay of the snapshot set that corresponds to the [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Console beam model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Console beam model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Console beam model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: A piston rod with a wall-obstacle. In conclusion, the proposed strategies can approximate the contact behavior of the console beam model, even though no information about the contact pressures was contained in the snapshots used for the non-intrusive ROM construction. …
Figure 9
Figure 9. Figure 9: Piston rod model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Piston rod model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 11
Figure 11. Figure 11: Piston rod model: Comparison of the contact simulation results [PITH_FULL_IMAGE:figures/full_fig_p025_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [1]

    M. S. Allen, D. Rixen, M. van der Seijs, P. Tiso, T. Abrahamsson, and R. L. Mayes. Substructuring in Engineering Dynamics . Springer Cham, 2020

  2. [2]

    Balajewiec, D

    M. Balajewiec, D. Amsallem, and C. Farhat. Projection-based model reduction for contact problems. Int. J. Numer. Meth. Engng , 106:pp. 644–663, 2015

  3. [3]

    U. Baur, P. Benner, and L. Feng. Model order reduction for linear and nonlinear systems: a system-theoretic perspective. Arch. Comput. Methods Eng., 21(4):pp. 331– 358, 2014

  4. [4]

    Belytschko and T.-J

    T. Belytschko and T.-J. R. Hughes. Computational methods for transient analysis. Vol. 1 in Computational Methods in Mechanics. Amsterdam-New York, North-Holland Publ. Co., 1983

  5. [5]

    Benner and L

    P. Benner and L. Feng. Recycling Krylov subspaces for solving linear systems with successively changing right-hand sides arising in model reduction. In Model Reduction for Circuit Simulation . Springer, 2011

  6. [6]

    Benner, P

    P. Benner, P. K¨ urschner, and J. Saak. Frequency-limited balanced truncation with low-rank approximations. SIAM J. Sci. Comput. , 38(1):A471–A499, Feb. 2016

  7. [7]

    Benner, V

    P. Benner, V. Mehrmann, and D. C. Sorensen, editors. Dimension Reduction of Large- Scale Systems, Lecture Notes in Computational Science and Engineering . Springer- Verlag,Berlin, 2005. 27

  8. [8]

    Benner, M

    P. Benner, M. Ohlberger, A. Cohen, and K. Willcox, editors. Model Reduction and Approximation: Theory and Algorithms . SIAM, Philadelphia, PA, 2017

Show all 45 references
  1. [9]

    Besselink, N

    B. Besselink, N. van de Wouw, J. M. Scherpen, and H. Nijmeijer. Model reduction for nonlinear systems by incremental balanced truncation. IEEE Trans. Autom. Control, 59(10):pp. 2739–2753, 2014

  2. [10]

    S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan. Linear Matrix Inequalities in System and Control Theory . SIAM, 1994

  3. [11]

    Breiten and T

    T. Breiten and T. Damm. Krylov subspace methods for model order reduction of bilinear control systems. Syst. Control Lett. , 59(8):pp. 443–450, 2010

  4. [12]

    R. W. Cottle, J.-S. Pang, and R. E. Stone. The Linear Complementarity Problem . SIAM, 2009

  5. [13]

    Craig and M

    R. Craig and M. Bampton. Coupling of substructures for dynamic analyses. AIAA J., pp. 1313–1319, 1968

  6. [14]

    Delhez, F

    E. Delhez, F. Nyssen, J.-C. Golinval, and A. Batailly. Reduced order modeling of blades with geometric nonlinearities and contact interactions. J. Sound Vib. , 500:116037, 2021

  7. [15]

    Diamond and S

    S. Diamond and S. Boyd. CVXPY: A Python-embedded modeling language for convex optimization. J. Mach. Learn. Res. , 17(83):pp. 1–5, 2016

  8. [16]

    Farhat, P

    C. Farhat, P. Avery, T. Chapman, and J. Cortial. Dimensional reduction of nonlinear finite element dynamic models with finite rotations and energy-based mesh sampling and weighting for computational efficiency. Int. J. Numer. Methods Eng. , 98(9):pp. 625–662, 2014

  9. [17]

    Fauque, I

    J. Fauque, I. Rami` ere, and D. Ryckelynck. Hybrid hyper-reduced modeling for contact mechanics problems. Int. J. Numer. Methods Eng. , 115(1):pp. 117–139, 2018

  10. [18]

    Filanova, I

    Y. Filanova, I. Pontes Duff, P. Goyal, and P. Benner. An operator inference oriented approach for linear mechanical systems. Mech. Syst. Signal Process., 200:110620, 2023

  11. [19]

    Francavilla and O

    A. Francavilla and O. Zienkiewicz. A note on numerical computation of elastic contact problems. Int. J. Numer. Methods Eng. , 9:pp. 913–924, 1975

  12. [20]

    Geradin and D

    M. Geradin and D. J. Rixen. Mechanical Vibrations: Theory and Application to Structural Dynamics. Wiley, 2015

  13. [21]

    P. Gill, W. Murry, and M. Wright. Practical Optimization . Academic Press, 4th edition, 1981. 28

  14. [22]

    Hartmann

    D. Hartmann. Real-time digital twins. e-print 2311.14691, arXiv, 2023. cs.CY

  15. [23]

    Hartmann, M

    D. Hartmann, M. Herz, M. Paffrath, J. Rommes, T. Tamarozzi, H. Van der Auweraer, and U. Wever. Model order reduction and digital twins , pp. 379–430. De Gruyter, Berlin, Boston, 2021

  16. [24]

    Heinrich, M

    C. Heinrich, M. Khalil, K. Martynov, and U. Wever. Online remaining lifetime esti- mation for structures. Mech. Syst. Signal Process. , 119:pp. 312–327, 2019

  17. [25]

    W. C. Hurty. Vibrations of structural systems by component mode synthesis. J. Eng. Mech. Div., 86(4):51–69, 1960

  18. [26]

    Ilyas Ahmad, P

    M. Ilyas Ahmad, P. Benner, and I. Jaimoukha. Krylov subspace methods for model reduction of quadratic-bilinear systems. IET Control Theory Appl. , 10(16):pp. 2010– 2018, 2016

  19. [27]

    Kikuchi and J

    N. Kikuchi and J. T. Oden. Contact Problems in Elasticity . SIAM, 1988

  20. [28]

    Kramer, B

    B. Kramer, B. Peherstorfer, and K. E. Willcox. Learning nonlinear reduced models from data with operator inference. Annual Review of Fluid Mechanics, 56:pp. 521–548, 2024

  21. [29]

    K¨ urschner

    P. K¨ urschner. Balanced truncation model order reduction in limited time intervals for large systems. Adv. Comput. Math. , 44(6):pp. 1821–1844, 2018

  22. [30]

    Lamperski

    A. Lamperski. Python: Lemkelcp 0.1. https://pypi.org/project/lemkelcp/

  23. [31]

    Le Berre, I

    S. Le Berre, I. Rami` ere, J. Fauque, and D. Ryckelynck. Condition number and clustering-based efficiency improvement of reduced-order solvers for contact problems using lagrange multipliers. Mathematics, 10(9):1495, 2022

  24. [32]

    Li and S.-L

    C.-X. Li and S.-L. Wu. A note on the unique solution of linear complementarity problem. Cogent Math., 3(1):1271268, 2016

  25. [33]

    Lohmann and B

    B. Lohmann and B. Salimbahrami. Ordnungsreduktion mittels Krylov- Unterraummethoden (Order Reduction using Krylov Subspace Methods). at- Automatisierungstechnik, 52(1):30–38, 2004

  26. [34]

    Manvelyan, B

    D. Manvelyan, B. Simeon, and U. Wever. An efficient model order reduction scheme for dynamic contact in linear elasticity. Comput. Mech., 10.1007/s00466-021-02068-4, 2021

  27. [35]

    Manvelyan, B

    D. Manvelyan, B. Simeon, and U. Wever. A physics-based model reduction approach for node-to-segment contact problems in linear elasticity. Int. J. Numer. Methods Eng., 2022. 29

  28. [36]

    Martynov and U

    K. Martynov and U. Wever. On polynomial hyperreduction for nonlinear structural mechanics. Int. J. Numer. Methods Eng. , 118(12):pp. 701–717, 2019

  29. [38]

    Peherstorfer and K

    B. Peherstorfer and K. Willcox. Data-driven operator inference for nonintrusive projection-based model reduction. Comput. Methods Appl. Mech. Eng. , 306:pp. 196– 215, 2016

  30. [39]

    Radermacher and S

    A. Radermacher and S. Reese. Pod-based model reduction with empirical interpolation applied to nonlinear elasticity. Int. J. Numer. Methods Eng., 107(6):pp. 477–495, 2016

  31. [40]

    D. J. Rixen. A dual Craig–Bampton method for dynamic substructuring. J. Comput. Appl. Math. , 168(1-2):pp. 383–391, 2004

  32. [41]

    J. Saak, D. Siebelts, and S. W. R. Werner. A comparison of second-order model order reduction methods for an artificial fishtail. at-Automatisierungstechnik, 67(8):pp. 648– 667, 2019

  33. [42]

    Salimbahrami, R

    B. Salimbahrami, R. Eid, and B. Lohmann. Model reduction by second order Krylov subspaces: Extensions, stability and proportional damping. In Proc. IEEE Conf. Comput. Aided Control Syst. Design, Munich, Germany , pp. 2997–3002. IEEE, 2006

  34. [43]

    Sharma, D

    H. Sharma, D. A. Najera-Flores, M. D. Todd, and B. Kramer. Lagrangian operator inference enhanced with structure-preserving machine learning for nonintrusive model reduction of mechanical systems. Comput. Methods Appl. Mech. Eng. , 423:116865, 2024

  35. [44]

    Swischuk, L

    R. Swischuk, L. Mainini, B. Peherstorfer, and K. Willcox. Projection-based model reduction: Formulations for physics-based machine learning. Comput. Fluids, 179:pp. 704–717, 2019

  36. [45]

    Van der Auweraer and D

    H. Van der Auweraer and D. Hartmann. The executable digital twin: merging the digital and the physics worlds. e-print 2210.17402v2, arXiv, 2023. cs.OH

  37. [46]

    Wriggers

    P. Wriggers. Computational Contact Mechanics . Springer Berlin, Heidelberg, 2006. 30

Pith tools

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