{"id":"f32c3f39-268c-47c6-bb8e-ae86a2b47611","arxiv_id":"2412.08656","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":16,"one_line_summary":"For AS4/PEEK, associated flow in pressure-dependent composite plasticity models produces spurious tensile transverse strains under shear-dominated biaxial loads; non-associated deviatoric flow corrects the direction but overpredicts shear-with-compression response.","lead":"This paper tests three Drucker-Prager-type plastic models for fiber composites, with and without non-associated flow, against biaxial experiments on AS4/PEEK. It finds that associated flow predicts non-physical tensile transverse strains under shear-dominated loading, while a deviatoric non-associated flow fixes the direction but overestimates compression-shear coupling.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central conclusion is conditional on an untested zero-dilatancy assumption: κ=0 in the flow potentials (Eqs. 23, 33, 41) enforces the corrected flow direction, and Section 5.2.4 concedes real crazing/dilatancy would invalidate it.","rationale":"The reader's CONDITIONAL verdict is based on exactly the assumption I find most load-bearing: the non-associated potentials are taken to be pressure-independent (κ=0 in Eqs. 23, 33, 41), which forces the corrected plastic flow direction. The paper is honest about this in Section 5.2.4, mentioning crazing and the need for a pressure-dependent potential if dilatancy is present, but this makes the central claim conditional rather than established. I did not find an internal inconsistency severe enough to justify rejection; the tensor algebra around Eq. (18) is difficult to verify from the OCR text and is not needed for this concern. A controlled numerical comparison keeping the yield parameters fixed between associated and non-associated versions would also help separate recalibration from flow-rule effects, but the existing conditional verdict already covers the necessary checks. No change to the reader's verdict is recommended.","tokens_in":19570,"tokens_out":15736,"duration_ms":151107,"concrete_test":"Use the raw strain histories from Vogler and Kyriakides (1999) for load paths 01–03 and 05–08. Subtract the elastic strains computed from E1, E2, ν12 and the current stress state to obtain the plastic strain path; then form the incremental plastic volume change J̇p = tr(ε̇p) over the shear-dominated segments. If J̇p is significantly non-zero beyond experimental scatter, Eqs. (23)/(33)/(41) are falsified and the non-associated correction is an artifact of the assumed pressure independence; if J̇p ≈ 0, the central claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing step is the definition of the non-associated plastic potentials as the yield function with κ set to zero (Eqs. 23, 33, 41). This choice is what removes the pressure term from ε̇p and rotates the predicted plastic strain direction under pure shear back to the vertical axis in Fig. 4. It is a modeling assumption, not a consequence of the experimental data. Vogler and Kyriakides [3] provide stress–strain curves, but no direct measurement of the plastic strain direction or of the inelastic volume change is reported. Section 5.2.4 explicitly acknowledges that polymer matrices can dilate through crazing and that in that case the flow potential would need a non-zero pressure term. If the true matrix dilatancy is non-zero, the claimed correction is partly an artifact of the chosen potential, and the residual overprediction of shear-with-compression (Figs. 5b, 6a) may be a symptom of this misspecification rather than an intrinsic limitation of non-associated flow. Additionally, no error metrics or scatter bands are provided, so the word \"quantitative\" cannot separate this assumption from small model misspecification.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops three Drucker-Prager-type elastoplastic constitutive models for unidirectional fibre composites — a modified isotropic Drucker-Prager model (Model-I), a mapped-tensor model following Car, Oller and Oñate (Model-II), and an anisotropic representation-theorem extension (Model-III) — and implements each in both associated and non-associated flow variants. The models are calibrated against pure in-plane shear and transverse compression of AS4/PEEK and then compared with experimental biaxial shear–compression load paths. The central finding is that associated flow produces physically inconsistent tensile transverse strains under pure shear and shear-dominated combined loads, whereas non-associated flow with a pressure-independent plastic potential removes this inconsistency but still overpredicts the shear response under transverse compression preloads. The authors acknowledge the residual deviation and attribute it to the assumed pressure-independent flow potential and to the lack of dilatancy data.","tokens_in":20054,"tokens_out":6260,"duration_ms":75616,"significance":"The paper's main claim is practically relevant: if confirmed, it shows that associated Drucker-Prager-type plasticity should not be used for shear-dominated combined loading of polymeric composites, and that a non-associated formulation is a simple remedy. A clear strength is that the biaxial load paths used for evaluation were not used for calibration; only pure shear and transverse compression data enter the parameter identification, so the comparisons in Figs. 5 and 6 are genuine predictions. The unified implementation and comparison of three model structures in one numerical framework is also useful. The authors are transparent about the residual deviations and about the need for additional experimental data. However, the quantitative claim is weakened by the absence of error metrics and by the untested zero-dilatancy assumption in the flow potentials, so the significance is conditional on strengthening those points.","major_comments":[{"comment":"The central conclusion that non-associated flow circumvents the inconsistencies of associated flow rests on the specific choice of a pressure-independent plastic potential, κ̃ = 0 in Eqs. (23), (33) and (41). This is a constitutive assumption, not a measured property; the experiments of Ref. [3] are not used to determine the plastic strain direction or the inelastic volume change. Section 5.2.4 itself concedes that crazing or other dilatancy in the polymer matrix would require a pressure-dependent flow potential, and the Conclusions state that the choice of flow rule is unclear without additional biaxial data. The residual overprediction of shear-with-compression (Figs. 5b and 6a) may therefore be an artifact of the assumed potential rather than an intrinsic limitation of non-associated flow. I ask the authors to either determine κ̃ (or, equivalently, the plastic strain direction) from an independent measurement, or to report a sensitivity study over κ̃ in the range from 0 to κ and show that the qualitative correction in Fig. 4 and the deviations in Figs. 5 and 6 are robust within that range.","section":"Section 5.2 and Figs. 5-6"},{"comment":"The manuscript repeatedly describes the comparison with experiments as 'quantitative' (abstract, Section 5) but nowhere reports error metrics, scatter bands, or uncertainty quantification. Judgments such as 'significant deviations' (Section 5.2.3) and 'slightly overpredicted' (Section 5.2.3) are based on visual inspection. Because the main conclusion distinguishes between acceptable and unacceptable model behavior, the absence of quantitative agreement measures is load-bearing. Please include a deviation measure per load path (e.g., RMSE or maximum relative error in stress and strain) and, if available, experimental scatter from Ref. [3]. This would also make it possible to judge whether the residual overprediction is small model misspecification or evidence against the non-associated structure.","section":"Section 5.2 and Figs. 5-6"}],"minor_comments":[{"comment":"The caption states that the corrected response is shown in panels (b), (d) and (e), but panel (e) already displays Model-III-a; the panel for the corrected Model-III-b response is presumably (f). Please correct the panel references.","section":"Figure 4 caption"},{"comment":"The opening sentence of the Discussion, 'the predictions are in excellent agreement with the experimental results whilst using an associative flow rule', conflicts with the earlier statements of 'erroneous predictions' and 'not satisfactory' in Sections 5.2.1 and 5.2.2. Please clarify which models and load paths are meant.","section":"Section 5.2.4"},{"comment":"The convexity condition 'a1−2 ≥ 0' appears to be a typographical error or is at least unclear, since a1 has units of inverse stress squared from Eq. (39). Please state the intended condition on a1 and a2 in a dimensionally consistent form.","section":"Section 4.2.3"},{"comment":"For Model-II, the text notes that the non-associative predictions are 'largely similar' to the associative ones. Since a change of flow potential would normally affect the plastic strain direction significantly, a brief explanation of why the transformation mapping suppresses this difference would help the reader reproduce and trust the result.","section":"Section 5.2.3"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope and the computational comparison is transparent. The main risk is that the non-associated flow result is tied to the ad hoc κ=0 assumption; I would not reject on that basis because the authors acknowledge it, but the revision should either substantiate the assumption or bound its influence. I would also ask the editor to consider whether the 'quantitative' claim can be supported without error metrics."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the paper does something useful—it takes three known Drucker-Prager-type models, calibrates them on pure shear and transverse compression, and then compares associated and non-associated flow variants on biaxial paths. The models themselves are mostly assembled from prior work, so the novelty is the comparison, not the formulation. That comparison produces a clear finding: with associated flow, pressure-dependent models predict spurious transverse tensile strains under shear-dominated combined loads, and switching to a purely deviatoric non-associated potential fixes the flow direction but leaves an overprediction of shear response in compression. The plots support that reading, and the fact that the biaxial paths were not used to fit parameters is a real strength. The paper also earns credit for being honest about its own gaps—Section 5.2.4 says outright that crazing/dilatancy in the polymer matrix would require a pressure-dependent flow potential and that the issue is left open.\n\nThe soft spots are real but not fatal. The largest is the way non-associated flow is implemented: the plastic potential is simply the yield function with κ=0, so the correction to the plastic strain direction is built in by assumption rather than inferred from data. If the matrix actually dilates, the corrected direction and the residual overprediction of shear-with-compression could both be artifacts of that choice. The paper acknowledges this but does not quantify it. There are also no error metrics or scatter bands—\"quantitative\" means visual comparison to plotted curves—and the simulations use a single C3D8 element for one material. No code or data are shipped, so reproduction would take real effort. These are limitations, not disqualifiers; the central qualitative conclusion is consistent with earlier work and the paper doesn't oversell the residuals.\n\nWho is this for? Someone doing model selection for pressure-dependent plasticity in UD composites, especially if they care about flow-rule choices and spurious dilatancy. It is not a landmark, but it is a solid comparative study with a clear takeaway.\n\nMy recommendation: send it to peer review. A serious referee should ask for quantitative error measures, a direct treatment of the κ=0 assumption (or at least a sensitivity study), and ideally a data/code release. But the paper deserves referee time, not a desk reject.","headline":"A useful, honest side-by-side comparison of three Drucker-Prager-type composite plasticity models with associated vs. deviatoric non-associated flow; the main conclusion is real but conditional on the assumed zero-dilatancy flow potential.","tokens_in":20492,"tokens_out":2755,"would_cite":true,"duration_ms":27533,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Pressure-dependent yield functions with an associated flow rule predict physically wrong tensile transverse strains under shear-dominated combined loading of fibre composites; replacing the flow rule with a non-associative deviatoric…","keywords":["anisotropic plasticity","fibre-reinforced composites","Drucker-Prager yield function","associated flow rule","non-associative flow rule","pressure-dependent plasticity","transverse isotropy","AS4/PEEK"],"falsifier":"Measure the plastic transverse strain increment during pure in-plane shear of the AS4/PEEK composite: the non-associated models predict zero transverse plastic strain because the plastic potential is purely deviatoric, so any measured transverse plastic strain or measurable dilatancy under shear with compression would refute the assumed pressure-independent flow potential.","tokens_in":19400,"feed_emoji":"⚙️","tokens_out":13582,"duration_ms":108721,"temperature":0.7,"pith_summary":"The paper asks whether the choice of flow rule inside pressure-dependent plasticity models of unidirectional fibre composites changes the predicted inelastic response, and shows that it does. Three Drucker-Prager-type models—an isotropic one, a mapped-tensor one, and an anisotropically extended one—are calibrated against pure-shear and transverse-compression data for a carbon-fibre/PEEK composite and then tested on biaxial load paths. With the associated (normality) flow rule, the models produce physically unrealistic transverse strains under shear-dominated combined stress states: tensile strains for Models I and III, and a spurious compressive strain for Model II under pure shear. Switching to a non-associative flow rule with a purely deviatoric plastic potential corrects the flow direction, but the shear response under transverse compression remains overestimated. The conclusion is that pressure sensitivity of the yield function is necessary while the associated flow rule is not usable for shear-dominated loads, and that choosing the right flow rule needs additional biaxial yield-surface data.","feed_headline":"Non-associative flow fixes shear strain direction in composites","feed_subtitle":"Associated flow predicts tensile strains under shear; the fix overestimates shear under compression.","key_machinery":"The load-bearing object is the transversely isotropic Drucker-Prager-type yield function $\\chi = \\kappa p + \\|\\Sigma\\|_{\\mathbb{C}_8} - \\sqrt{2/3}[y_0 - \\beta] + \\frac{b}{2}\\|\\beta\\|^2_{\\mathbb{C}_9}$, where $p$ is the hydrostatic pressure in the effective stress, $\\mathbb{C}_8$ is a fourth-order deviatoric projection tensor that selects shear modes while leaving the fibre direction elastic, and $m = a \\otimes a$ encodes the fibre orientation. The associated flow rule generates plastic strain rates from $\\partial_\\sigma \\chi$ and therefore carries the pressure term $\\frac{\\kappa}{3}(1-m)$ into the flow direction; this is the term that produces tensile transverse strain under shear. The non-associative machinery is a second plastic potential $\\phi = \\chi|_{\\kappa=0}$ that is purely deviatoric, so the plastic strain rate is governed only by $\\mathbb{C}_8 : \\Sigma / \\|\\Sigma\\|_{\\mathbb{C}_8}$. Comparing Models I, II, and III isolates what the choice of isotropic versus anisotropic dissipative response adds on top of the flow-rule choice.","core_discovery":"The central claim is that the negative slope of Drucker-Prager-type yield surfaces in the pressure–shear plane breaks the normality assumption: an associated flow rule turns pure shear and shear-dominated combined stress states into non-physical transverse strain, with Models I and III reporting tensile transverse strain where experiments show compression and Model II reporting a small spurious compressive strain under pure shear. The authors demonstrate this defect in all three model formulations and then show that a non-associative flow rule built from the same yield function with the pressure coefficient set to zero in the plastic potential ($\\kappa=0$ in Eqs. (23), (33), (41)) restores the experimentally observed flow direction under pure shear. However, on the compression-preload biaxial path the non-associated models still overestimate the shear response, which the authors attribute to the pressure-independent flow rule and possibly to matrix dilatation due to crazing. The discovery is therefore a separation of effects: the flow-rule choice controls the direction of plastic strain, while the pressure coupling controls the magnitude of the shear–compression interaction, and neither the associative nor the non-associative form alone reproduces the full biaxial response.","pith_inferences":["A natural extension the paper leaves implicit is a family of plastic potentials with a pressure coefficient between zero and the associative $\\kappa$; the observed overprediction brackets the admissible value from above, so a small positive pressure coupling in the flow potential may preserve the corrected flow direction while reducing the shear-under-compression error.","The associated-flow inconsistency is a property of the negative yield-surface slope in the pressure–shear plane, so the same problem should appear in other cohesive pressure-dependent materials such as glassy polymers, foams, and granular solids under combined shear and compression.","A testable rate-dependent extension is to replace the rate-independent setting with a Perzyna-type viscous regularization using a low pressure slope, as the paper suggests, and check whether the overestimated shear-under-compression response relaxes towards the experiments at finite strain rates.","The paper's comparison implies that a purely isotropic pressure-dependent yield function cannot reproduce the biaxial response; an anisotropic plastic potential built from the same projection tensors as the yield function is the most direct cure, and its pressure coefficient should be fit to the missing $\\tau_{12}\\to-\\sigma_{22}$ data."],"forward_implications":["Whenever a Drucker-Prager-type yield function with a negative pressure slope is used, an associated flow rule is unreliable for shear-dominated combined loads because it generates non-physical tensile transverse strains.","The non-associative deviatoric plastic potential corrects the plastic strain direction under pure shear and shear-preload paths while preserving the calibrated pure-shear and transverse-compression responses.","The remaining overprediction of shear response under transverse compression is shared by all three model structures, so a pressure-independent plastic flow alone cannot close the gap.","Anisotropic plastic response functions track the experimental biaxial response more closely than isotropic ones, but they still overpredict at the highest preloads.","Additional experimental data on the $\\tau_{12}\\to-\\sigma_{22}$ biaxial path, which maps the evolution of the yield surface and plastic flow potential, are needed to decide between associative and non-associative formulations."],"supporting_citations":[{"why":"supplies the AS4/PEEK experimental data for pure shear, transverse compression, and the biaxial load paths used for calibration and validation.","marker":"[3]"},{"why":"companion modelling study that first showed pressure-dependent models with associative flow predict a decrease in transverse strain for increasing shear, the trend this paper corrects.","marker":"[8]"},{"why":"source of the anisotropic Model-III formulation and the calibration procedure for its plastic parameters.","marker":"[16]"},{"why":"origin of the mapped-tensor Model-II formulation based on a fictitious isotropic space.","marker":"[18]"},{"why":"basis of the modified isotropic Drucker-Prager Model-I used as the simplest response function.","marker":"[22]"},{"why":"motivates the non-associative flow rule by showing associative flow predicts tensile rather than compressive transverse strain in off-axis tests.","marker":"[30]"},{"why":"provides the deviatoric plastic-potential construction with zero pressure coupling used to define the non-associative flow rules.","marker":"[31]"}],"fun_headline_variants":["Non-associated flow sets shear strain direction in composites","Flow rule controls strain direction, not just magnitude, in composites","Shear direction in composites depends on flow rule associativity","Composite shear strain direction fixed by non-associative flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The non-associated models remove pressure effects from the plastic flow direction entirely, assuming the polymer matrix never changes volume while it flows; if real PEEK expands under shear with compression, for instance through crazing, the remaining overprediction of the shear response may be an artifact of that choice.","fun_headline_variants_meta":{"raw":{"variants":["Non-associated flow sets shear strain direction in composites","Flow rule controls strain direction, not just magnitude, in composites","Shear direction in composites depends on flow rule associativity","Composite shear strain direction fixed by non-associative flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000523,"raw_usage":{"total_tokens":2495,"prompt_tokens":876,"completion_tokens":1619,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":1554}},"tokens_in":492,"tokens_out":1619,"duration_ms":12524,"temperature":1.0,"reasoning_tokens":1554,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:20:37.975618+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the plastic transverse strain increment during pure in-plane shear of the AS4/PEEK composite: the non-associated models predict zero transverse plastic strain because the plastic potential is purely deviatoric, so any measured transverse plastic strain or measurable dilatancy under shear with compression would refute the assumed pressure-independent flow potential.","supporting_citations":[{"cited_title":"Inelastic behavior of an AS 4/PEEK com- posite under combined transverse compression and shear. Part I : experiments","cited_arxiv_id":null,"evidence_quote":"supplies the AS4/PEEK experimental data for pure shear, transverse compression, and the biaxial load paths used for calibration and validation."},{"cited_title":"Inelastic behav ior of an AS4/PEEK composite under combined transverse compression and shear. Part II: modeling","cited_arxiv_id":null,"evidence_quote":"companion modelling study that first showed pressure-dependent models with associative flow predict a decrease in transverse strain for increasing shear, the trend this paper corrects."},{"cited_title":"Constitut ive modeling of anisotropic plasticity with application to fiber-reinforced composite s","cited_arxiv_id":null,"evidence_quote":"source of the anisotropic Model-III formulation and the calibration procedure for its plastic parameters."},{"cited_title":"An anisotropic elastopla stic constitutive model for large strain analysis of fiber reinforced composite mater ials","cited_arxiv_id":null,"evidence_quote":"origin of the mapped-tensor Model-II formulation based on a fictitious isotropic space."},{"cited_title":"A plasticity model for unidirection al composite materials and its applications in modeling composites testing","cited_arxiv_id":null,"evidence_quote":"basis of the modified isotropic Drucker-Prager Model-I used as the simplest response function."},{"cited_title":"A simple nonlinear constitutive model based on non-associative plasticity fo r UD composites: Development and calibration using a Modified Arcan Fixture","cited_arxiv_id":null,"evidence_quote":"motivates the non-associative flow rule by showing associative flow predicts tensile rather than compressive transverse strain in off-axis tests."},{"cited_title":"Towards variational const itutive updates for non-associative plasticity models at finite strain: Models based on a v olumetric- deviatoric split","cited_arxiv_id":null,"evidence_quote":"provides the deviatoric plastic-potential construction with zero pressure coupling used to define the non-associative flow rules."}],"review_version":1}