pith. sign in

arxiv: 2604.10059 · v1 · submitted 2026-04-11 · 💻 cs.CE

Data-adaptive spline surfaces for non-separable hyperelastic energy functions

Simon Wiesheier , Miguel Angel Moreno-Mateos , Paul Steinmann This is my paper

Pith reviewed 2026-05-10 16:10 UTC · model grok-4.3

classification 💻 cs.CE
keywords hyperelasticityB-spline surfacesnon-separable invariantsdata-driven constitutive modelinglinear least-squares calibrationincompressible isotropichomogeneous deformation
0
0 comments X

The pith

A bivariate B-spline surface on the invariant domain models non-separable hyperelastic energies with fast linear calibration from homogeneous deformations.

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

The paper proposes replacing the separable form of hyperelastic strain energy with a bivariate B-spline surface defined directly on the physically admissible domain of the first and second invariants. This surface captures arbitrary coupling between the invariants while keeping the representation compact. Because the energy depends linearly on the spline coefficients, fitting to stress data from simple homogeneous deformation tests reduces to a constrained linear least-squares problem that solves quickly without sensitivity to starting values. A sympathetic reader would care because the combination yields higher accuracy than separable models yet remains efficient enough for tasks that require many repeated calibrations.

Core claim

By aligning the approximation space of a bivariate B-spline surface with physically realizable states in the invariant domain, all model parameters contribute meaningfully to the constitutive response. Calibration proceeds directly from analytical stress relations obtained under homogeneous deformation modes, turning the parameter identification problem into a constrained linear least-squares task that is fast, robust, and independent of initialization. The resulting model improves accuracy over separable representations while requiring only mild regularization in sparsely sampled regions.

What carries the argument

The bivariate B-spline surface defined directly on the physically admissible invariant domain, which encodes the full non-separable dependence of the strain-energy function on the two invariants.

If this is right

  • The model achieves higher accuracy than separable representations for complex material responses.
  • Parameter identification reduces to a linear problem that solves practically instantaneously.
  • Only mild regularization is required in regions with sparse sampling.
  • The efficiency makes the approach suitable for applications requiring repeated calibration such as uncertainty quantification and interactive material characterization.

Where Pith is reading between the lines

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

  • The linear calibration could be embedded inside experimental control loops to update material parameters in real time during testing.
  • The same spline construction offers a practical middle path between restrictive analytical forms and heavy black-box models when constitutive data arrive incrementally.
  • Extensions to anisotropic hyperelasticity would require only a change in the underlying invariant set while preserving the linear fitting structure.

Load-bearing premise

Homogeneous deformation modes alone supply sufficient and representative data across the entire physically admissible invariant domain, and the spline surface can be aligned with realizable states so that all coefficients contribute meaningfully without introducing non-physical artifacts.

What would settle it

A withheld set of mixed-mode deformation tests on which the fitted spline produces stress errors larger than those of a calibrated separable model, or violates basic physical requirements such as positive definiteness of the tangent stiffness, would disprove the claimed accuracy and reliability gains.

Figures

Figures reproduced from arXiv: 2604.10059 by Miguel Angel Moreno-Mateos, Paul Steinmann, Simon Wiesheier.

Figure 1
Figure 1. Figure 1: The admissible invariant domain (a) in the ( [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Results for the separable spline model. (a) Calibrated univariate splines [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Results for the spline surface Wsurf(I¯1, I¯2) in the (I¯1, I¯2) space. (a) Calibration without curvature regularization. (b) Calibration with a mild curvature penalty. The heatmap shows the logarithmic relative parameter activation as defined in Equation (69). (c) Comparison of model predictions based on (b) with experimental Treloar data for the deformation modes UT, BT, and PS. 13 [PITH_FULL_IMAGE:figu… view at source ↗
Figure 4
Figure 4. Figure 4: Results for the spline surface Wsurf(ξ, η) in normalized coordinates (ξ, η). (a) Calibrated spline surface together with the logarithmic relative parameter activation. The inverse mapping of the spline grid from (ξ, η) to (I¯1, I¯2) defines the same surface on the admissible invariant domain. (b) Comparison of model predictions based on (a) with experimental Treloar data for the deformation modes UT, BT, a… view at source ↗
Figure 5
Figure 5. Figure 5: Quantitative evaluation of model accuracy and calibration time. (a) Mean squared error (MSE) for uniaxial tension, [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Parameter sensitivity surfaces of the tensor-product spline representation. (a) Interpolation grid in the mapped invariant [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Automatic selection of the curvature regularization parameter [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
read the original abstract

Invariant-based models for incompressible isotropic hyperelasticity are typically formulated as functions of the first and second invariants, $W = W(\bar{I}_1, \bar{I}_2)$. A widely used class of models employs separable representations of the form $W(\bar{I}_1, \bar{I}_2) = W_1(\bar{I}_1) + W_2(\bar{I}_2)$, which enable efficient calibration and straightforward enforcement of modeling constraints. However, this decomposition implicitly restricts the coupling between the invariants and may limit the achievable accuracy for complex material responses. Fully coupled data-driven approaches overcome this limitation but often require nonlinear optimization and large parameter sets. In this contribution, we propose a compact alternative: a bivariate B-spline surface defined directly on the physically admissible invariant domain. By aligning the approximation space with physically realizable states, all model parameters contribute meaningfully to the constitutive response. We utilize homogeneous deformation modes to perform a calibration directly from analytical stress relations, eliminating the need for finite element model updating. Owing to the linear dependence of the spline representation on its coefficients, the resulting parameter identification problem reduces to a constrained linear least-squares problem. This enables fast, robust, and initialization-independent calibration, which makes parameter identification practically instantaneous. The results demonstrate that the proposed model improves accuracy compared to separable approaches while requiring only mild regularization in weakly sampled regions. The combination of computational efficiency and the linear structure of a highly expressive spline surface makes the approach particularly attractive for applications requiring repeated calibration, such as uncertainty quantification or interactive material characterization.

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

1 major / 2 minor

Summary. The manuscript proposes a bivariate B-spline surface for the incompressible isotropic hyperelastic strain-energy function W(Ī₁, Ī₂) that is non-separable. Calibration is performed directly from analytical stress expressions obtained under homogeneous deformation modes (uniaxial, equibiaxial, pure shear, etc.), reducing the identification problem to a constrained linear least-squares problem whose solution is claimed to be fast, initialization-independent, and accurate with only mild regularization in sparsely sampled regions of the invariant domain. Numerical comparisons are presented against separable models, and the approach is positioned as attractive for repeated-calibration tasks such as uncertainty quantification.

Significance. If the central claims are substantiated, the method supplies a practical compromise between the expressiveness of fully coupled data-driven models and the computational tractability of separable analytic forms. The reduction to linear least squares and the alignment of the basis with physically admissible states are genuine strengths that could facilitate rapid recalibration in engineering workflows.

major comments (1)
  1. The calibration data consist exclusively of one-dimensional loci in the (Ī₁, Ī₂) plane generated by homogeneous deformations. The manuscript does not report any verification that the resulting spline surface produces physically admissible stresses or energies for deformation paths lying off these loci (e.g., non-homogeneous or multi-axial loading). This directly affects the claim that mild regularization suffices in weakly sampled regions and that accuracy gains are realized for general deformations.
minor comments (2)
  1. The abstract states that 'the results demonstrate' improved accuracy, yet the quantitative error tables or figures comparing the spline model against separable baselines should be referenced explicitly in the abstract or introduction for immediate clarity.
  2. Notation for the reduced invariants (Ī₁, Ī₂) and the spline knot placement should be introduced once and used consistently; occasional reversion to unbarred I₁, I₂ creates ambiguity.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address the major comment below, agreeing where appropriate and outlining planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: The calibration data consist exclusively of one-dimensional loci in the (Ī₁, Ī₂) plane generated by homogeneous deformations. The manuscript does not report any verification that the resulting spline surface produces physically admissible stresses or energies for deformation paths lying off these loci (e.g., non-homogeneous or multi-axial loading). This directly affects the claim that mild regularization suffices in weakly sampled regions and that accuracy gains are realized for general deformations.

    Authors: We agree that the manuscript would benefit from explicit verification on deformation paths off the homogeneous loci. The calibration procedure guarantees exact reproduction of stresses along the one-dimensional paths corresponding to uniaxial, equibiaxial, and pure-shear modes. Because the bivariate B-spline basis is defined over the full admissible invariant domain and the regularization penalizes unphysical curvature, the resulting surface is constructed to remain admissible everywhere; however, this property has not been demonstrated numerically for non-homogeneous or multi-axial states. To address the concern directly, we will add a dedicated verification subsection containing finite-element simulations of non-homogeneous problems (e.g., indentation and torsion of a cylinder). In these examples the local deformation histories traverse regions of the (Ī₁, Ī₂) plane away from the calibration loci, allowing us to confirm that stresses and energies remain physically admissible and that the accuracy advantage over separable models is retained. The same examples will also illustrate the sufficiency of the mild regularization in sparsely sampled areas. revision: yes

Circularity Check

0 steps flagged

No circularity: calibration uses external analytical stresses from homogeneous modes via standard linear least-squares.

full rationale

The paper constructs a bivariate B-spline surface for W(I1_bar, I2_bar) on the admissible invariant domain and obtains its coefficients by solving a constrained linear least-squares problem that matches the derived stresses to closed-form analytical stress expressions evaluated on homogeneous deformation paths (uniaxial, equibiaxial, pure shear, etc.). This is ordinary data fitting to independent external relations rather than any self-definitional loop, fitted input renamed as prediction, or load-bearing self-citation. The linearity is exploited only for computational efficiency of the fit; no uniqueness theorem, ansatz, or prior result from the same authors is invoked to force the form or the outcome. Claims of improved accuracy and mild regularization needs are presented as empirical outcomes of this procedure, not tautological consequences of the construction itself.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach rests on standard B-spline approximation theory and linear algebra; no new physical entities are introduced.

free parameters (1)
  • B-spline coefficients
    The coefficients of the bivariate surface are the unknowns solved by constrained linear least squares from stress data.
axioms (1)
  • standard math B-spline basis functions are non-negative, form a partition of unity, and have local support
    These properties are invoked to ensure the surface is well-behaved and all coefficients contribute on the invariant domain.

pith-pipeline@v0.9.0 · 5584 in / 1327 out tokens · 51496 ms · 2026-05-10T16:10:57.565756+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

36 extracted references · 36 canonical work pages

  1. [1]

    A Continuum Approach for Engineering(John Wiley & Sons, 2001), second print edn

    Holzapfel, G.Nonlinear solid mechanics. A Continuum Approach for Engineering(John Wiley & Sons, 2001), second print edn

    work page 2001

  2. [2]

    & Possart, G

    Steinmann, P., Hossain, M. & Possart, G. Hyperelastic models for rubber-like materials: consistent tangent operators and suitability for Treloar’s data.Archive of Applied Mechanics82, 1183–1217 (2012)

    work page 2012

  3. [3]

    & Wriggers, P

    Ricker, A. & Wriggers, P. Systematic Fitting and Comparison of Hyperelastic Continuum Models for Elastomers. Archives of Computational Methods in Engineering30(2023)

    work page 2023

  4. [4]

    A theory of large elastic deformation.Journal of applied physics11, 582–592 (1940)

    Mooney, M. A theory of large elastic deformation.Journal of applied physics11, 582–592 (1940)

    work page 1940

  5. [5]

    Characterization of elastic properties of carbon-black-filled rubber vulcanizates.Rubber chemistry and technology63, 792–805 (1990)

    Yeoh, O. Characterization of elastic properties of carbon-black-filled rubber vulcanizates.Rubber chemistry and technology63, 792–805 (1990)

    work page 1990

  6. [6]

    Computat

    Fuhg, J. N.et al.A Review on Data-Driven Constitutive Laws for Solids.Archives of Computational Methods in Engineering32, 1841–1883 (2025). URLhttps://doi.org/10.1007/s11831-024-10196-2

    work page doi:10.1007/s11831-024-10196-2 2025

  7. [7]

    A., Linden, L., Brummund, J., Metsch, P

    Kalina, K. A., Linden, L., Brummund, J., Metsch, P. & Kästner, M. Automated constitutive modeling of isotropic hyperelasticity based on artificial neural networks.Computational Mechanics69, 1–20 (2022)

    work page 2022

  8. [8]

    K., Fernández, M., Martin, R

    Klein, D. K., Fernández, M., Martin, R. J., Neff, P. & Weeger, O. Polyconvex anisotropic hyperelasticity with neural networks.Journal of the Mechanics and Physics of Solids159, 104703 (2022)

    work page 2022

  9. [9]

    K.et al.Neural networks meet hyperelasticity: A monotonic approach (2025)

    Klein, D. K.et al.Neural networks meet hyperelasticity: A monotonic approach (2025). URL http://arxiv. org/abs/2501.02670. ArXiv:2501.02670 [cs]

    work page arXiv 2025

  10. [10]

    URL https://linkinghub.elsevier.com/retrieve/pii/ S0022509623001679

    Linden, L.et al.Neural networks meet hyperelasticity: A guide to enforcing physics.Journal of the Mechanics and Physics of Solids179, 105363 (2023). URL https://linkinghub.elsevier.com/retrieve/pii/ S0022509623001679

    work page 2023

  11. [11]

    L., Jones, R

    Frankel, A. L., Jones, R. E. & Swiler, L. P. Tensor Basis Gaussian Process Models of Hyperelastic Materials. ArXivabs/1912.10872(2019). URLhttps://api.semanticscholar.org/CorpusID:209444982

    work page arXiv 1912

  12. [12]

    & Itskov, M

    Abdusalamov, R., Hillgärtner, M. & Itskov, M. Automatic generation of interpretable hyperelastic material models by symbolic regression.International Journal for Numerical Methods in Engineering124, 2093–2104 (2023). URLhttps://onlinelibrary.wiley.com/doi/10.1002/nme.7203

    work page doi:10.1002/nme.7203 2093

  13. [13]

    K., Sahli Costabal, F

    Taç, V ., Rausch, M. K., Sahli Costabal, F. & Tepole, A. B. Data-driven anisotropic finite viscoelasticity using neural ordinary differential equations.Computer Methods in Applied Mechanics and Engineering411, 116046 (2023). URLhttps://linkinghub.elsevier.com/retrieve/pii/S0045782523001706

    work page 2023

  14. [14]

    URLhttps://linkinghub.elsevier.com/retrieve/pii/S0045782525003615

    Thakolkaran, P.et al.Can KAN CANs? Input-convex Kolmogorov-Arnold Networks (KANs) as hyperelastic constitutive artificial neural networks (CANs).Computer Methods in Applied Mechanics and Engineering443, 118089 (2025). URLhttps://linkinghub.elsevier.com/retrieve/pii/S0045782525003615

    work page 2025

  15. [15]

    P., Aydin, R

    Abdolazizi, K. P., Aydin, R. C., Cyron, C. J. & Linka, K. Constitutive Kolmogorov–Arnold Networks (CK- ANs): Combining accuracy and interpretability in data-driven material modeling.Journal of the Mechanics and Physics of Solids203, 106212 (2025). URL https://linkinghub.elsevier.com/retrieve/pii/ S0022509625001887

    work page 2025

  16. [16]

    A., Wiesheier, S., Steinmann, P

    Flaschel, M., Moreno-Mateos, M. A., Wiesheier, S., Steinmann, P. & Kuhl, E. Unsupervised Material Fin- gerprinting: Ultra-fast hyperelastic model discovery from full-field experimental measurements (2026). URL https://arxiv.org/abs/2601.14965. Version Number: 1

    work page arXiv 2026

  17. [17]

    & Bathe, K.-J

    Sussman, T. & Bathe, K.-J. A model of incompressible isotropic hyperelastic material behavior using spline interpolations of tension–compression test data.Communications in numerical methods in engineering25, 53–63 (2009)

    work page 2009

  18. [18]

    A., Açan, A

    Dal, H., Denli, F. A., Açan, A. K. & Kaliske, M. Data-driven hyperelasticity, Part I: A canonical isotropic formulation for rubberlike materials.Journal of the Mechanics and Physics of Solids179, 105381 (2023). URL https://linkinghub.elsevier.com/retrieve/pii/S0022509623001850

    work page 2023

  19. [19]

    Z., Açan, A

    Tikeno˘gulları, O. Z., Açan, A. K., Kuhl, E. & Dal, H. Data-driven hyperelasticity, Part II: A canonical framework for anisotropic soft biological tissues.Journal of the Mechanics and Physics of Solids181, 105453 (2023). URL https://linkinghub.elsevier.com/retrieve/pii/S0022509623002570

    work page 2023

  20. [20]

    & Steinmann, P

    Wiesheier, S., Mergheim, J. & Steinmann, P. Discrete data-adaptive approximation of hyperelastic energy functions.Computer Methods in Applied Mechanics and Engineering416, 116366 (2023). URL https: //linkinghub.elsevier.com/retrieve/pii/S0045782523004905. 25 Data-adaptive spline surfaces for non-separable hyperelastic energy functionsA PREPRINT

    work page 2023

  21. [21]

    Wiesheier, S., Moreno-Mateos, M. A. & Steinmann, P. Versatile data-adaptive hyperelastic energy functions for soft materials.Computer Methods in Applied Mechanics and Engineering430, 117208 (2024). URL https://linkinghub.elsevier.com/retrieve/pii/S004578252400464X

    work page 2024

  22. [22]

    A., Wiesheier, S., Esmaeili, A., Hossain, M

    Moreno-Mateos, M. A., Wiesheier, S., Esmaeili, A., Hossain, M. & Steinmann, P. Biaxial characterization of soft elastomers: experiments and data-adaptive configurational forces for fracture (2025). URL http://arxiv.org/ abs/2505.20244. ArXiv:2505.20244 [cond-mat]

    work page arXiv 2025

  23. [23]

    Wiesheier, S., Moreno-Mateos, M. A. & Steinmann, P. Data-adaptive spline-based viscoelasticity for soft solids. Computer Methods in Applied Mechanics and Engineering451, 118705 (2026). URL https://linkinghub. elsevier.com/retrieve/pii/S0045782525009776

    work page 2026

  24. [24]

    & Tiller, W.The NURBS Book(Springer-Verlag, New York, NY , USA, 1996), second edn

    Piegl, L. & Tiller, W.The NURBS Book(Springer-Verlag, New York, NY , USA, 1996), second edn

    work page 1996

  25. [25]

    Boor, C.A Practical Guide to Splines(Springer Verlag, 1978)

    d. Boor, C.A Practical Guide to Splines(Springer Verlag, 1978)

    work page 1978

  26. [26]

    & Neff, P

    Hartmann, S. & Neff, P. Polyconvexity of generalized polynomial-type hyperelastic strain energy functions for near-incompressibility.International Journal of Solids and Structures40, 2767–2791 (2003). URL https: //linkinghub.elsevier.com/retrieve/pii/S0020768303000866

    work page 2003

  27. [27]

    & Neff, P

    Schröder, J. & Neff, P. Invariant formulation of hyperelastic transverse isotropy based on polyconvex free energy functions.International Journal of Solids and Structures40, 401–445 (2003). URL https://linkinghub. elsevier.com/retrieve/pii/S0020768302004584

    work page 2003

  28. [28]

    Thermodynamic relations for high elastic materials.Transactions of the Faraday Society57, 829–838 (1961)

    Flory, P. Thermodynamic relations for high elastic materials.Transactions of the Faraday Society57, 829–838 (1961)

    work page 1961

  29. [29]

    Sawyers, K. N. On the possible values of the strain invariants for isochoric deformations.Journal of Elasticity7, 99–102 (1977). URLhttp://link.springer.com/10.1007/BF00041134

    work page doi:10.1007/bf00041134 1977

  30. [30]

    Dammaß, F., Kalina, K. A. & Kästner, M. When invariants matter: The role of I1 and I2 in neural network models of incompressible hyperelasticity.Mechanics of Materials210, 105443 (2025). URL https://linkinghub. elsevier.com/retrieve/pii/S0167663625002054

    work page 2025

  31. [31]

    Convexity conditions and existence theorems in nonlinear elasticity.Archive for rational mechanics and Analysis63, 337–403 (1976)

    Ball, J. Convexity conditions and existence theorems in nonlinear elasticity.Archive for rational mechanics and Analysis63, 337–403 (1976)

    work page 1976

  32. [32]

    Butterfield, K. R. The Computation of all the Derivatives of a B-spline Basis.IMA Journal of Applied Mathematics 17, 15–25 (1976). URL https://academic.oup.com/imamat/article-lookup/doi/10.1093/imamat/ 17.1.15

    work page doi:10.1093/imamat/ 1976

  33. [33]

    C., Jensen, T

    Hansen, P. C., Jensen, T. K. & Rodriguez, G. An adaptive pruning algorithm for the discrete L-curve crite- rion.Journal of Computational and Applied Mathematics198, 483–492 (2007). URL https://linkinghub. elsevier.com/retrieve/pii/S0377042705007284

    work page 2007

  34. [34]

    Hansen, P. C. & O’Leary, D. P. The Use of the L-Curve in the Regularization of Discrete Ill-Posed Problems. SIAM Journal on Scientific Computing14, 1487–1503 (1993). URL http://epubs.siam.org/doi/10.1137/ 0914086

    work page 1993

  35. [35]

    & Callegaro, L

    Cultrera, A. & Callegaro, L. A simple algorithm to find the L-curve corner in the regularisation of ill-posed inverse problems.IOP SciNotes1, 025004 (2020). URL https://iopscience.iop.org/article/10.1088/ 2633-1357/abad0d

    work page 2020

  36. [36]

    Treloar, L. R. G. Stress-strain data for vulcanised rubber under various types of deformation.Transactions of the Faraday Society40, 59 (1944). URLhttps://xlink.rsc.org/?DOI=tf9444000059. 26

    work page 1944