{"id":"29e3d91c-9d1f-4366-9b5f-8ac0a18d094c","arxiv_id":"2607.18156","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A contact-based finite element updating method reconstructs spatially varying stiffness parameters in hyperelastic solids and shells from surface displacement and contact-force data.","lead":"Researchers built a computer method that infers how material stiffness varies inside a soft body by pressing it with a rigid probe and measuring the resulting surface motion and contact force. This could make tissue testing less invasive and aid in-vivo biomechanics, but realistic surface-only reconstruction is still limited in accuracy.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Surface-only 'sufficiency' claim rests on a regularization parameter tuned using the full-field inversion (Fig. 23, Eq. 65), unavailable in the intended surface-only application.","rationale":"I read the paper in good faith. The numerical experiments are extensive: analytical sensitivities are derived, FE-convergence studies are reported, synthetic data are generated on finer meshes with different penalty parameters, and noise is studied with statistical repeats. There is no sign of fabrication or circularity beyond the acknowledged inverse crime in Sec. 4.1. The reader's weakest-assumption choice—the fixed-active-set derivative in Eq. (68)—is a genuine theoretical limitation, and Remark 3 and Sec. 4.1 honestly display the resulting non-differentiability. However, the paper's examples, especially indentation with a prescribed probe, make active-set sensitivity to material parameters low, and the paper provides some evidence that the optimizer still works. I therefore do not treat that as the most load-bearing concern. The stronger issue is the regularization selection in the surface-only experiment. The central 'sufficiency' claim is operationalized through a regularizer whose strength is fixed using information from the corresponding full-field inversion. In a real surface-only clinical or laboratory scenario, that information is not available. The paper acknowledges the ill-conditioning and the need for regularization, but it does not offer a practical way to set α from surface data alone. This does not invalidate the method as a numerical feasibility study; it means the practical claim is conditional on a protocol that is not yet demonstrated. Thus the conditional verdict stands, but the condition should explicitly include a data-only regularization-selection rule or a clear statement that sufficiency has only been shown when the regularization target is known from a full-field experiment.","tokens_in":33313,"tokens_out":7055,"duration_ms":80048,"concrete_test":"Repeat the Sec. 4.3.2 surface-only inversion with α selected by a purely data-driven rule—e.g., generalized cross-validation or the Morozov discrepancy principle using the known noise level γ = 0.001L—and recompute the DSC values for Λ and µ. If the selected α leads to substantially lower DSC than in Figs. 25–26, or if no α chosen by a data-only criterion gives stable inclusion detection, then the 'sufficiency' claim indeed depends on the full-field-based tuning in Fig. 23. A complementary test: split the synthetic surface data into a tuning set and a validation set; choose α on the tuning set and evaluate on the validation set, comparing with the full-field-tuned result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The key finding—'full-field surface displacements and resultant contact forces are sufficient to identify the fields of Neo-Hookean model parameters in the bulk of the solid'—is supported in Sec. 4.3.2 only under a regularization parameter chosen by an oracle-like procedure. Without regularization, the surface-only problem fails (γK ≈ 1e16). The authors add R(α,q) = α²||q−q0,HP||² (Eq. 65) and select α in Fig. 23 to 'match the same level of γK as the corresponding full-field cases from Tab. 5 without noise.' That target requires already having solved the full-field inverse problem, which is precisely the information a surface-only, non-destructive measurement scenario is supposed to avoid. The L-curve is dismissed as overregularizing, but no data-only selection rule is supplied. Therefore the reported DSC values (Λ: 48–71%, µ: ~70%) and the 'sufficient' conclusion are conditional on a regularization parameter that cannot be set from surface data alone. This is not a claim of fraud or internal inconsistency; it is a gap between the demonstrated numerical experiment and the practical claim. The method may still be viable, but the current evidence does not establish sufficiency for the intended surface-only use.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a contact-based isogeometric Finite Element Model Updating (FEMU) procedure for reconstructing spatially heterogeneous hyperelastic parameters of 3D solids and thin shells. The forward model uses large-deformation hyperelasticity, Kirchhoff–Love/Canham/Koiter shell theories, and frictionless penalty contact; the inverse problem minimizes a normalized least-squares objective of full-field displacements and resultant contact forces under box constraints using trust-region reflective lsqnonlin with analytically derived Jacobians. Unknown parameter fields are represented on independent low-order Lagrange material meshes, and a material-continuation strategy reuses the previous equilibrium. Numerical examples cover a 1D shell strip on a rigid foundation, indentation of an abdominal wall Koiter-shell model, and probing of a Neo-Hookean block with hard/soft inclusions. Synthetic data are generated on finer meshes with different penalty parameters and Gaussian displacement noise. The paper concludes that contact probing can outperform pressure-based FEMU, that Λ is more noise-sensitive than µ, and that full-field surface displacements plus resultant contact forces are sufficient to reconstruct bulk Neo-Hookean fields, provided Tikhonov regularization is used.","tokens_in":33600,"tokens_out":6595,"duration_ms":76970,"significance":"If the claims hold, this is a useful extension of FEMU to contact-dominated soft-tissue identification, with practical relevance to in-vivo probing. The paper is careful about synthetic data: independent FE meshes, penalty mismatch, convergence studies, statistical repeats under noise, and sensitivity/collinearity diagnostics. The explicit analytical sensitivities and Jacobian-vector products are a notable contribution. However, the strongest conclusion—surface-only sufficiency—is not fully established because the regularization parameter is selected using full-field information, and the quantitative reconstruction in that setting remains weak. The comparison with pressure-based FEMU is also confounded by unequal data volume and mesh resolution.","major_comments":[{"comment":"The key claim that surface displacements plus contact forces are 'sufficient' is supported only under a regularization parameter α that is chosen by matching the collinearity index γK of the full-field solutions from Tab. 5. In the intended surface-only measurement scenario the full-field inverse solution is unavailable, so this is an oracle selection. Without regularization the surface-only problem fails (γK ≈ 10^16), and the L-curve is dismissed without a data-only alternative. The reported DSC values (Λ: 48–71%, µ: ~70%) and the noise runs inherit this choice. Please provide a selection rule based solely on surface data (e.g., discrepancy principle, GCV, or a stability test on α) or explicitly downgrade the conclusion to feasibility under oracle regularization.","section":"Sec. 4.3.2, Eq. (65), Fig. 23"},{"comment":"The conclusion that CBIA 'outperforms' PBIA for noisy data compares 30 contact load cases on a 56×56 mesh (Case 2.1all) with 4 pressure load cases on a 28×28 mesh (Cases 2.p1–2.p4). The improvement in ∆δL2 could be due to the sevenfold increase in data volume, the finer mesh, or both, rather than to the contact modality itself. A controlled comparison with matched number of load cases and mesh resolution (or a matched information metric) is needed before claiming contact-specific superiority.","section":"Sec. 4.2, Tabs. 2 and 4"},{"comment":"The analytic Jacobians are derived under a fixed active set A, while the objective is non-differentiable at active-set changes, as shown in Sec. 4.1. For unknown heterogeneous stiffness fields the active set can depend on q, so the TRR algorithm may use an inaccurate descent direction near folds. The paper's statement that the risk is 'very low' is plausible for stable indentation but is not demonstrated. Please quantify active-set stability along the optimization path (e.g., fraction of iterations with active-set changes) or adopt a smooth contact formulation; otherwise the claimed analytical-gradient advantage is not guaranteed.","section":"Eq. (68) and Remark 3"}],"minor_comments":[{"comment":"The example is an admitted inverse crime: the material mesh matches the reference distribution exactly, and Case 1.5 uses identical FE meshes for data generation and inversion. This is acceptable for a toy illustration, but the statement that the objective 'seems to have a unique minimum' should be explicitly limited to this noise-free, model-matched setting.","section":"Sec. 4.1"},{"comment":"The symbol f is used both for the equilibrium residual (Eq. (41)) and for the objective function (Eq. (47)). This is confusing; consider using different notation, e.g., R(u,q) for the residual.","section":"Notation"},{"comment":"The weights wU and wF are selected heuristically and their sensitivity is not studied. A brief discussion of how the reported results depend on this choice would strengthen the guidance for users.","section":"Eq. (47)"},{"comment":"The term 'full-field' in the block example refers to the full 2D cross-section, not 3D volume data. This should be stated explicitly to avoid overgeneralizing the identifiability conclusions.","section":"Sec. 4.3.1"},{"comment":"Some author names are corrupted by line breaks (e.g., 'A vril'), which should be cleaned in the final version.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a solid FEMU extension and the numerical work is generally careful, but the central 'surface-only sufficiency' claim rests on an oracle-chosen regularization parameter. I would encourage the editor to request a data-only selection strategy or a clearly weakened claim before acceptance. The comparison between CBIA and PBIA also needs a controlled experimental design."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a credible extension of Borzeszkowski et al.'s FEMU machinery to contact loading, with analytical sensitivities and an independent material mesh. The three synthetic examples are worked carefully: convergence studies, noise repeats, sensitivity and collinearity diagnostics. The abdominal-wall comparison against pressure-based FEMU is genuinely informative, and the full-field reconstructions of the Neo-Hookean block look good.\n\nWhere it gets soft: in Sec. 4.3.2, the paper claims that full-field surface displacements plus resultant contact forces are sufficient to identify the bulk material fields. But the surface-only problem without regularization fails (collinearity index ~1e16), and the Tikhonov parameter alpha is chosen in Fig. 23 to match the collinearity level of the full-field solutions from Tab. 5. That target requires already having solved the full-field problem, which is exactly the information the surface-only, non-destructive scenario is supposed to avoid. The L-curve is dismissed as overregularizing, but no data-only selection rule is supplied. So the reported Dice values (Lambda 48–71%, mu ~70%) are conditional on an oracle, and the abstract's non-destructive promise is stronger than the demonstrated evidence. This is not a fatal flaw—the method may still be viable—but the 'sufficient' finding needs rewording or a real selection rule.\n\nOther concerns are smaller. Example 4.1 is an acknowledged inverse crime; acceptable as a visualization of the non-differentiable landscape, not as validation. The analytic derivatives assume a fixed contact active set, and the paper honestly notes the objective is non-differentiable; the claim that practical risk is low is plausible but not deeply verified. All tests are synthetic with a known constitutive law, so in-vivo claims remain prospective. These are proportionate caveats, not signs of sloppiness.\n\nCitation pattern looks fine: the core FEMU machinery is credited to Borzeszkowski et al., and the self-citations are to related work that this paper genuinely extends. No circularity concerns.\n\nWould I engage? Yes—send this to review. The method is a real step beyond pressure-based FEMU, the derivations are careful, and the paper is unusually candid about its limitations. The referee should focus on the regularization protocol and on whether 'sufficient' can be supported without full-field information. With that fixed or qualified, it becomes a solid contribution.","headline":"Credible FEMU extension with real novelty in contact-based identification, but the surface-only 'sufficiency' claim rests on a regularization parameter chosen by an oracle — referee it, and push on that.","tokens_in":34069,"tokens_out":2116,"would_cite":true,"duration_ms":22691,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74B20","74M15","74K25","65N21","74S05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Pressing a rigid probe against a soft, heterogeneous solid can reveal how its stiffness varies inside, if surface displacements and the resultant contact force are measured.","keywords":["inverse problems","parameter identification","full-field measurements","heterogeneous materials","isogeometric analysis","mechanical contact","finite element model updating","hyperelasticity"],"falsifier":"Construct a two-material block whose inclusion stiffness is chosen so that a tiny parameter perturbation changes whether a surface point is in contact, run the inverse routine on synthetic data, and check whether the optimizer stalls or the reconstruction diverges exactly at the active-set boundary; if it does, the fixed-active-set gradient assumption is the failing point.","tokens_in":33172,"feed_emoji":"🩺","tokens_out":4017,"duration_ms":76991,"temperature":0.7,"pith_summary":"The paper tries to establish that mechanical contact probing—pushing a rigid indenter against a body, much like palpation—can serve as the loading modality for reconstructing spatially varying material properties in nonlinear solids and thin shells. It builds a finite element model updating (FEMU) scheme in which unknown material fields are adjusted until the model reproduces full-field measured displacements and the resultant contact force. Evidence comes from three synthetic experiments: a bending shell strip on a rigid foundation, an abdominal-wall shell model, and a Neo-Hookean block with a stiffness inclusion. The central finding is that full-field surface displacements plus contact forces identify the bulk material fields of Neo-Hookean parameters, while surface-only data can reveal an inclusion but not its peak value without regularization.","feed_headline":"Probing maps hidden stiffness inside soft solids","feed_subtitle":"Measured surface motion plus contact force is enough to reconstruct internal material fields, three numerical tests show.","key_machinery":"The engine is the FEMU least-squares objective that compares measured displacements and resultant contact forces to their finite element counterparts. Its Jacobian is computed analytically via the sensitivity matrix S = ∂f_int/∂q and the tangent stiffness K through K ∂u/∂q = −S, avoiding the cost of finite differences. A critical assumption is that the contact active set A is held fixed when differentiating, so changes in which surface points are in contact are ignored; the paper acknowledges this makes the objective non-differentiable. The unknown material fields live on a separate low-order Lagrange material mesh, and a material continuation strategy updates the parameters between optimiza","core_discovery":"The paper claims that a contact-based inverse analysis framework—combining isogeometric finite elements, an independent low-order Lagrange mesh for the material fields, analytically derived objective derivatives, and a trust-region reflective optimizer—can reconstruct spatially varying constitutive parameters from displacement measurements on at least the free surface plus the resultant contact force. In the Neo-Hookean block example, full-field surface displacements and contact forces are sufficient to identify the fields of the Lamé parameters Λ and μ in the bulk of the solid. In the abdominal-wall shell model, nine probe positions with several indentation depths recover the Young's modulu","pith_inferences":["The paper's evidence that Λ is poorly identified from surface indentation suggests a practical protocol: if compressibility is the target, add loading modes that excite volume change (e.g., pressure or volumetric strain measurements) rather than relying on contact alone.","The visible folds in the objective landscape point to a testable extension: replacing the discrete active-set contact with a smooth contact formulation could remove the artificial non-differentiabilities and make the optimizer more robust to parameter-dependent contact changes.","The framework assumes isotropic hyperelasticity and frictionless, adhesionless contact; extending it to anisotropic or viscoelastic tissue models and to frictional contact are natural next steps, and the sensitivity/collinearity machinery should carry over directly.","The finding that multiple probe locations greatly reduce noise sensitivity suggests that sparse probing at many locations with moderate indentation is more informative than a few deep indentations, which is directly testable in experimental design."],"forward_implications":["If correct, a single non-destructive contact test—indentation with a rigid probe—can reconstruct heterogeneous stiffness fields in soft tissues and laboratory specimens without cutting or invasive access.","The abdominal-wall example implies that contact-based analysis can match pressure-based methods in accuracy while being non-invasive and giving many more loading configurations (multiple probe locations, depths, and directions).","The block example implies that full-field surface displacements plus resultant contact forces are sufficient for bulk Neo-Hookean parameter fields, but surface-only measurements are severely ill-conditioned and demand regularization.","The sensitivity and collinearity analyses imply that compressibility-related parameters such as Λ are intrinsically harder to identify from indentation than shear-related parameters such as μ, so identified compressibility values should be treated cautiously.","The analytical derivatives and material continuation imply that inverse problems with hundreds to thousands of material unknowns—up to 6498 in the abdominal-wall mesh refinement—are computationally feasible with a local gradient-based optimizer."],"fun_headline_variants":["Contact forces map hidden stiffness inside soft solids","Surface motion plus contact force reveals internal stiffness fields","Probing soft solids with touch uncovers hidden material maps","Inverse contact analysis reconstructs stiffness from surface data","Touch-based inversion finds hidden stiffness in soft solids"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing assumption is that the contact region is stable enough during optimization that the analytic gradient, derived with the contact active set held fixed, remains a reliable search direction; if small material changes flip contact points on or off, the objective is non-differentiable and the optimizer can stall.","fun_headline_variants_meta":{"raw":{"variants":["Contact forces map hidden stiffness inside soft solids","Surface motion plus contact force reveals internal stiffness fields","Probing soft solids with touch uncovers hidden material maps","Inverse contact analysis reconstructs stiffness from surface data","Touch-based inversion finds hidden stiffness in soft solids"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000487,"raw_usage":{"total_tokens":2246,"prompt_tokens":760,"completion_tokens":1486,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":1413}},"tokens_in":504,"tokens_out":1486,"duration_ms":10892,"temperature":1.0,"reasoning_tokens":1413,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T15:48:13.522682+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a two-material block whose inclusion stiffness is chosen so that a tiny parameter perturbation changes whether a surface point is in contact, run the inverse routine on synthetic data, and check whether the optimizer stalls or the reconstruction diverges exactly at the active-set boundary; if it does, the fixed-active-set gradient assumption is the failing point.","supporting_citations":[],"review_version":1}