{"id":"54d770b6-bbe3-4877-b2c4-94473ba484e5","arxiv_id":"2411.13992","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A smoothing-length-scaled transport-velocity correction with a fitted limiter prevents tensile instability and overcorrection in weakly-compressible SPH, including low-velocity and variable-resolution flows.","lead":"This paper modifies an existing SPH stabilization technique so it works for slow-moving fluids and for simulations with mixed particle sizes. The change is useful for fluid-structure and biomedical flow simulations, but the limiter constants are tuned to the test cases rather than derived.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Derivation mismatch in Eq. (14): α=7 and CFL_ad=0.25 yield ξ=-0.4375, not -0.2, so the central correction coefficient in Eq. (16) is not derived from the stated constants.","rationale":"I read the paper in good faith. The proposed limiter and smoothing-length-scaled correction are plausible, and the numerical tests in Sections 4.1-4.6 provide real empirical support: Taylor-Green matches the analytical decay, the cavity and FDA nozzle benchmark agree with reference data, and the multi-resolution cylinder converges. However, the derivation of Eq. (14) contains an internal algebraic inconsistency that is more specific than the fitted-limiter concern: the printed constants α=7.0 and CFL_ad=0.25 force ξ=-0.4375, not -0.2. This matters because the central claim is that Eq. (16) is a derived, algorithmically equivalent form of the Adami et al. transport velocity. If the printed derivation is what a reader must follow, the method is not reproducible from the paper alone. The reader flagged this same coefficient issue in passing but made the limiter constants the main weak assumption; I agree with the CONDITIONAL verdict and propose the coefficient as the load-bearing point, with a direct re-derivation and two-parameter rerun as the decisive check. I do not see grounds for rejection, because the numerical validations are substantial and a corrected derivation plus a sensitivity study on the limiter constants could resolve the issue.","tokens_in":11508,"tokens_out":7038,"duration_ms":70059,"concrete_test":"Recompute Eq. (14) symbolically with α=7.0, CFL_ad=0.25, p0=α ρ0 |v|_ad², and Δt_ad=CFL_ad h/|v|_ad, writing m_j=ρ_j V_j and ρ_i≈ρ0. If the result is -0.4375 h² Σ_j ∇W_ij V_j rather than -0.2 h² Σ_j ∇W_ij V_j, the derivation is wrong as printed. Then run the Taylor-Green case of Section 4.1 at dp=1/50 twice in SPHinXsys with the same limiter (15): once with ξ=-0.2 and once with ξ=-0.4375, and compare kinetic-energy decay against the analytical value -16π²/Re. If ξ=-0.4375 also matches the analytical curve and avoids overcorrection, the discrepancy is a typo and the method stands. If it reintroduces the overcorrection shown in Figure 3, then the -0.2 coefficient is load-bearing calibration not derivable from the stated constants.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claim is that Eq. (16) is the improved transport-velocity correction. That claim rests on Eq. (14), which supposedly derives ξ=-0.2 from the standard background-pressure formula (12). Substituting p0=α ρ0 |v|_ad² (Eq. 13) with α=7.0 and Δt_ad=CFL_ad h/|v|_ad with CFL_ad=0.25 into Eq. (12), and writing m_j=ρ_j V_j with ρ_i≈ρ0, gives Δe_r = (1/2)(-2) α CFL_ad² h² Σ_j ∇W_ij V_j = -α CFL_ad² h² Σ_j ∇W_ij V_j = -0.4375 h² Σ_j ∇W_ij V_j. The text instead states ξ=-0.2 (Eq. 14) with no alternative value of α or CFL_ad. Thus Eq. (16) is not algebraically equivalent to the Adami et al. transport-velocity formula under the stated constants; either the coefficient is an unexplained calibration or one of the printed constants is inconsistent with the implementation. Because every validation in Section 4 uses the -0.2 coefficient, an outside re-implementation following the printed derivation would apply a correction 2.19 times larger and cannot reproduce the paper. The limiter constants (5×10^-4 and C=10^3) are also stated to be chosen from numerical cases, but the coefficient mismatch is the more fundamental barrier: without a correct derivation, Eq. (16) is an ansatz rather than the promised extension of the transport-velocity formulation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an improved transport-velocity formulation for weakly-compressible SPH. In Section 3, the authors replace the background-pressure-dependent correction of Adami et al. with a smoothing-length-scaled correction, Eq. (16), and introduce a limiter, Eq. (15), that suppresses the correction for well-ordered particle distributions to avoid overcorrection in low-velocity flows. The formulation is claimed to extend the transport-velocity approach to variable-resolution simulations and to prevent particle clumping and void formation without the overcorrection of the original method. Section 4 validates the method on a Taylor-Green vortex, 2D and 3D lid-driven cavity flows, flow-induced vibration of an elastic beam, multi-resolution flow around a cylinder, and the 3D FDA nozzle, with comparisons to references and experiments.","tokens_in":11841,"tokens_out":6669,"duration_ms":55013,"significance":"The practical importance of a transport-velocity correction that scales with smoothing length rather than background pressure is clear: it could make the method applicable to multi-resolution SPH and address the known overcorrection at low velocities. The paper has several strengths: the benchmark suite is broad, including 2D and 3D cases, fluid-structure interaction, variable resolution, and an experimental device; the SPHinXsys code is publicly accessible, which supports reproducibility; and the proposed limiter targets a real limitation of the original formulation. However, the central coefficient in Eq. (16) is not derived from the stated constants in Eq. (14), and the limiter constants are calibrated from numerical cases without a sensitivity study. These issues place the core quantitative claim in need of revision before the result can be relied upon.","major_comments":[{"comment":"The coefficient ξ = -0.2 is not obtained from the stated formulas. Substituting Eq. (13) with α = 7.0 and Eq. (9) with CFL_ad = 0.25 into Eq. (12), and using m_j = ρ_j V_j with ρ_i ≈ ρ_0, gives Δe_r = -α CFL_ad^2 h^2 Σ_j ∇W_ij V_j = -7 × (0.25)^2 h^2 Σ_j ∇W_ij V_j = -0.4375 h^2 Σ_j ∇W_ij V_j. The printed value -0.2 is smaller by a factor of 2.1875. Since all validations in Section 4 use the coefficient -0.2 of Eq. (16), the derivation in Eq. (14) does not support the implemented correction; an independent re-implementation following the printed constants would apply a much larger correction and would not reproduce the reported results. The authors should either correct the derivation with consistent constants or explicitly state that ξ is a calibrated parameter. This issue is load-bearing because Eq. (16) is the central claim of the paper.","section":"Section 3, Eq. (14)"},{"comment":"The limiter constants C = 10^3 and the threshold 5 × 10^-4 are introduced 'according to the numerical cases'. Since the limiter is added specifically to remove the overcorrection observed with the original formulation, part of the reported improvement is built into these fitted parameters. The manuscript provides no sensitivity study with respect to C or the threshold, and no scale analysis showing that these values are robust across resolutions, Reynolds numbers, or variable-resolution ratios. Without such evidence, the general-accuracy claim for low-velocity and multi-resolution flows is not fully supported. At minimum, the authors should report results for nearby values of C and the threshold, or derive the constants from dimensional or error considerations.","section":"Section 3, Eq. (15)"},{"comment":"The statement that the lid-driven cavity results show 'approximate second-order convergence' is not backed by a convergence table or error norms. The figure shows velocity profiles at three resolutions, but no quantitative errors are reported. Please add a table of L1/L2 errors and observed orders for the velocity profiles, or soften the convergence claim.","section":"Section 4.2"},{"comment":"The formulation uses h_min in Eqs. (14)-(16) for variable-resolution flows, but the manuscript does not justify why the minimum smoothing length is the correct length scale for the correction across a resolution interface, nor does it test sensitivity to the resolution ratio or to the location of the interface. Since the central novelty includes variable-resolution applicability, this assumption should be either derived or tested, for example by varying the refinement ratio and measuring differences in the solution.","section":"Sections 3 and 4.4"}],"minor_comments":[{"comment":"The word 'stablity' in the introduction should be 'stability'.","section":"Section 1"},{"comment":"The text refers to ρ* as the density prior to reinitialization, but Eq. (8) does not define ρ* or show how it enters the summation; please clarify the notation.","section":"Section 2.3, Eq. (8)"},{"comment":"The notation p_L/p_R in Eq. (5) is inconsistent with p_l/p_r in Eq. (4); please unify.","section":"Section 2.2, Eqs. (4) and (5)"},{"comment":"The phrase 'as depicted in the left panel of Figure 12 (left panel)' repeats 'left panel'; remove the duplicate.","section":"Section 4.5"},{"comment":"The word 'sceneries' in the conclusion should be 'scenarios'.","section":"Section 5"},{"comment":"The column header 'Amplitude in y direction/(D)' is unclear; please state explicitly that the amplitude is normalized by D.","section":"Table 1"},{"comment":"The caption refers to panels (a) and (b), but the panels are not labeled in the figure; please add labels.","section":"Figure 13"}],"recommendation":"major_revision","confidential_remarks":"The coefficient mismatch in Eq. (14) is a load-bearing issue that must be resolved before publication; however, it is correctable in revision. The open-source code availability and the breadth of benchmarks count strongly in the paper's favor. I do not see grounds for rejection, but the central derivation needs to be fixed and the limiter calibration made transparent."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The practical SPH fix is plausible and the test suite is broad, but the printed derivation of the central coefficient doesn't close. Substituting α=7 and CFL_ad=0.25 into Eq. (14) gives ξ=-0.4375, not -0.2. That means Eq. (16) as derived is not what was simulated; the -0.2 coefficient is effectively calibrated. This is the main thing to know.\n\nWhat's new and good: the paper extends Adami et al.'s transport velocity by replacing background pressure with a smoothing-length-scaled correction, which is the right move for variable-resolution simulations, and adds a limiter based on local particle disorder to avoid overcorrection in low-speed flows. The limiter idea is physically reasonable: well-ordered particle distributions need little correction. The benchmark set is genuinely useful—Taylor-Green, 2D/3D lid-driven cavity, FSI beam behind a cylinder, multi-resolution cylinder, and the FDA nozzle. Results mostly match Ghia, LBM, and PIV data. The code is open in SPHinXsys, though not pinned to a commit.\n\nSoft spots: the coefficient mismatch is load-bearing. A re-implementation following the printed math applies a correction 2.19 times larger and won't reproduce Section 4. Either a constant is misprinted (α would have to be 3.2 to get -0.2) or ξ is an empirical fit; the paper needs to say which. The limiter constants, threshold 5e-4 and gain 1e3, are stated to be chosen from numerical cases, and there is no sensitivity study. That is a moderate concern: the method's advantage over the original may depend on these values. I'd also like error bars or convergence statistics for the benchmark quantities; the cavity shows convergence, but the FSI and cylinder comparisons are single-run values. These are fixable.\n\nThe citation pattern is fine. The paper is honest that Eq. (14) is algorithmically equivalent to Adami if the constants are right, and the open code helps reproducibility.\n\nVerdict: this deserves a serious referee, but with major revision. The derivation has to be corrected or re-labeled as calibration, and the limiter robustness needs a small parameter study. Who it's for: SPH practitioners working on low-speed or variable-resolution flows; they will get value from the limiter and the h_min scaling even before the derivation is cleaned up. I wouldn't cite it as a formal method until the coefficient issue is resolved.","headline":"Useful SPH limiter and broad benchmarks, but Eq. (14)'s coefficient is off by a factor of 2.19 and the central correction is effectively calibrated, not derived.","tokens_in":12389,"tokens_out":3035,"would_cite":false,"duration_ms":29228,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76M28","65M75"],"pacs":["47.11.-j"],"model":"deepseek-v4-flash","headline":"SPH clumping fixed by smoothing-length transport correction","keywords":["weakly-compressible SPH","tensile instability","transport velocity","smoothing length","variable-resolution flows","limiter","particle clumping","void regions"],"falsifier":"Run a low-velocity variable-resolution simulation, such as a lid-driven cavity with a fine region embedded in a coarse region and maximum velocity below $10^{-3}$, and check whether the limiter stays off in the uniform coarse region while particles still develop clumps or an energy plateau appears. If overcorrection or clumping persists outside the tested parameter range, the threshold and gain are not general.","tokens_in":11280,"feed_emoji":"💧","tokens_out":3813,"duration_ms":33152,"temperature":0.7,"pith_summary":"The paper proposes a modified transport-velocity correction for weakly-compressible SPH that is scaled to the smoothing length instead of the background pressure, making the correction suitable for variable-resolution simulations. To stop overcorrection in low-velocity flows, the correction is multiplied by a limiter that switches on only when a local disorder measure indicates a non-uniform particle distribution. The resulting displacement correction is $\\Delta \\mathbf{e}_r = -0.2\\, \\beta\\, h_{\\min}^2 \\sum_j \\nabla W_{ij} V_j$ with $\\beta$ given by the limiter. Numerical tests on the Taylor-Green vortex, lid-driven cavity, fluid-structure interaction, multi-resolution cylinder flow, and a 3D FDA nozzle show the method avoids overcorrection, prevents clumping and void regions, and matches reference solutions.","feed_headline":"SPH clumping fixed by smoothing-length transport correction","feed_subtitle":"A disorder-triggered limiter replaces background pressure, stopping overcorrection in slow flows.","key_machinery":"The central object is the limiter $\\beta$ of Eq. (15), a switch built on the scalar disorder measure $h^2 |\\sum_j \\nabla W_{ij} V_j|^2$, where $h=h_{\\min}$ in variable-resolution regions. When the measure is above the threshold $5\\times10^{-4}$, $\\beta = \\min(C \\cdot \\text{measure}, 1)$ with $C=10^3$; otherwise $\\beta=0$. This multiplier sits in front of the smoothing-length-scaled transport displacement $\\Delta \\mathbf{e}_r = -0.2\\, \\beta\\, h_{\\min}^2 \\sum_j \\nabla W_{ij} V_j$, so the same formula handles both uniform and disordered particle clouds without needing a background pressure.","core_discovery":"On the paper's own terms, the transport-velocity correction need not depend on a global background pressure. The authors derive, from the background-pressure formulation of Adami et al. under dual-criteria time stepping, that the correction per advection step is a displacement proportional to $h^2 \\sum_j \\nabla W_{ij} V_j$ with coefficient $-0.2$. They then bound this correction with a limiter $\\beta$ that is zero unless $h^2|\\sum_j \\nabla W_{ij} V_j|^2$ exceeds $5\\times10^{-4}$, and saturates at 1 with gain $C=10^3$. The central discovery is that this disorder-triggered, smoothing-length-scaled limiter removes the overcorrection visible in low-velocity Taylor-Green flow while preserving the stabilising effect in disordered or variable-resolution regions.","pith_inferences":["The disorder measure could be made dimensionless and adaptive, replacing the fixed $5\\times10^{-4}$ threshold with a value tied to kernel shape, which would remove the case-dependent tuning.","The limiter is likely sensitive to the Wendland kernel and $h=1.3dp$ used here; kernels with different derivative behaviour may need recalibration of $C$ and the threshold.","If the measure is local and cheap, it could also gate particle-shifting or artificial-viscosity corrections in other SPH variants, not just transport velocity.","A direct comparison against the particle-shifting technique on the same variable-resolution FSI cases would clarify which strategy better preserves free-surface volume."],"forward_implications":["Low-velocity flows no longer show the energy decay plateau caused by the original background-pressure formulation.","The correction scales with smoothing length, so variable-resolution SPH uses the same formula with $h=h_{\\min}$ in refined regions.","The limiter allows transport velocity to be used with dual-criteria time stepping without the sound-speed-dependent background pressure.","Accuracy benchmarks for cavity flow, FSI, and the FDA nozzle match reference data, so the change does not sacrifice solution quality.","Two- and three-dimensional tests show the same stabilising effect, so the method generalises beyond planar flows."],"supporting_citations":[{"why":"The original transport-velocity formulation that the paper extends and re-derives.","marker":"[1]"},{"why":"Generalized transport-velocity formulation with variable background pressure, the immediate predecessor being improved.","marker":"[26]"},{"why":"Dual-criteria time stepping that rewrites the background pressure in terms of advection speed, used in the derivation.","marker":"[27]"},{"why":"Free-stream boundary condition supplying the coefficient $\\alpha=7.0$ used in Eq. (13).","marker":"[32]"},{"why":"Riemann-based WCSPH discretization forming the governing equations the transport correction is added to.","marker":"[24]"}],"fun_headline_variants":["Smoothing-length limiter fixes SPH clumping","SPH transport velocity without background pressure","Disorder-triggered limiter stabilises SPH flows","New transport-velocity correction for SPH","Variable-resolution SPH stabilised by limiter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The limiter's fixed threshold and gain constants, tuned on the test cases, are assumed to separate reliably the particle distributions that need transport correction from those that do not, across all velocities and resolutions.","fun_headline_variants_meta":{"raw":{"variants":["Smoothing-length limiter fixes SPH clumping","SPH transport velocity without background pressure","Disorder-triggered limiter stabilises SPH flows","New transport-velocity correction for SPH","Variable-resolution SPH stabilised by limiter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000181,"raw_usage":{"total_tokens":1266,"prompt_tokens":861,"completion_tokens":405,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":331}},"tokens_in":477,"tokens_out":405,"duration_ms":3789,"temperature":1.0,"reasoning_tokens":331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:39:46.665376+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a low-velocity variable-resolution simulation, such as a lid-driven cavity with a fine region embedded in a coarse region and maximum velocity below $10^{-3}$, and check whether the limiter stays off in the uniform coarse region while particles still develop clumps or an energy plateau appears. If overcorrection or clumping persists outside the tested parameter range, the threshold and gain are not general.","supporting_citations":[{"cited_title":"Zhang, X","cited_arxiv_id":null,"evidence_quote":"Generalized transport-velocity formulation with variable background pressure, the immediate predecessor being improved."},{"cited_title":"Zhang, M","cited_arxiv_id":null,"evidence_quote":"Dual-criteria time stepping that rewrites the background pressure in terms of advection speed, used in the derivation."},{"cited_title":"Zhang, W","cited_arxiv_id":null,"evidence_quote":"Free-stream boundary condition supplying the coefficient $\\alpha=7.0$ used in Eq. (13)."},{"cited_title":"Zhang, X","cited_arxiv_id":null,"evidence_quote":"Riemann-based WCSPH discretization forming the governing equations the transport correction is added to."}],"review_version":1}