{"id":"b3ca7f6d-e2d5-42c6-8446-c42488bd3303","arxiv_id":"2608.08114","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Combining finite basis PINNs with hard boundary constraints keeps Stokes-flow errors near one percent as the number of perforations grows to 100.","lead":"Hard-constrained finite basis PINNs, which split the domain into overlapping subdomains and enforce no-slip conditions by construction, solve Stokes flow around many small obstacles with roughly one percent error in two-dimensional tests. The result matters because it suggests neural solvers can scale to the highly perforated microstructures used in composites manufacturing.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Composite hard-constraint ansatz is never stated; as written, wall lifting leaves a ~3e-4 no-slip residual on ∂Ω_p, so exact boundary enforcement underlying the theory is not demonstrated.","rationale":"The reader's conditional verdict is appropriate. The most load-bearing gap is not in the numerical experiments themselves, which show consistent sub-1% L2 errors across 16-100 perforations and multi-seed robustness for the 64-perforation case, but in the unstated exact form of the hard-constrained ansatz. The theoretical apparatus (Proposition 3.1, Theorem 3.1, Eq. (9)) is internally consistent, and the Fourier analysis is a valid qualitative account of spectral bias mitigation. The compatibility issue flagged in Section 3.2 is real: the wall lifting term from (44) does not vanish on ∂Ω_p for finite a_w, so the no-slip condition is enforced only up to residuals of order 3e-4 for the closest perforations in the 64-periodic geometry. This is small enough not to invalidate the 1% error claims, but it means the implemented method is not strictly in the exactly-constrained class NN^† used for the error bound (9), and the stiffness analysis of l_{∂Ω_p} alone does not cover the composed ansatz. A precise statement of the ansatz and a boundary-residual measurement would settle this. If the residual is as estimated, the central numerical claim survives and the paper can be accepted after adding the missing formula and a sentence quantifying the boundary error; hence no change from the reader's conditional verdict.","tokens_in":30387,"tokens_out":20160,"duration_ms":191539,"concrete_test":"Extract the exact velocity ansatz from the released code (or request it from the authors) and evaluate max_{x∈∂Ω_p}|uθ(x)| and max_{x∈∂Ω_w}|uθ(x)-g(x)| for the 64-periodic case at initialization and after 100k iterations. If the no-slip boundary residual exceeds 1e-3 relative to |g|, the hard-constraint guarantee (8)-(9) does not apply and the method reduces to a soft constraint with a small penalty; if it is below 1e-3, the concern becomes purely a missing formula, and the reported convergence claim stands. Report which ordering of C_w and l_{∂Ω_p} was actually used.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central theoretical premise is that the hard-constrained ansatz (8) enforces u=g on ∂Ω_w and u=0 on ∂Ω_p exactly, yielding the stronger error bound (9). The experimental construction, however, is never written down. Section 3.2 first presents the general form uθ=C_w[Σ ω_i C_p^i[z_i]] (33), then switches to applying 'both the wall and no-slip boundary hard constraints globally as in (26)' without giving the composite formula. The wall operator (44) is C_w[v]=tanh(a_w l_{∂Ω_w})v + g[1-tanh(a_w l_{∂Ω_w})]. If the velocity is formed as C_w[l_{∂Ω_p}\\bar u], the no-slip value on ∂Ω_p equals the lifting term g[1-tanh(a_w l_{∂Ω_w})], which is nonzero: for the 64-periodic case with nearest perforation at x1≈0.12875, l_{∂Ω_w}≈0.448 and the residual is ≈3×10^{-4} relative to |g|. If the other ordering is used, the wall condition fails. Thus no stated formula satisfies both boundary conditions exactly, and the compatibility warning in Section 3.2 is not resolved. Moreover, the curvature/stiffness analysis of Δl_{∂Ω_p} in (28) and Fig. 3 analyzes only the no-slip factor in isolation, not its composition with C_w. This gap is load-bearing because the method's name, the error estimate (9), and the claimed stiffness reduction all depend on exact boundary enforcement; the numerical 1% accuracy is nonetheless plausible because the violation is below the reported error.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes hard-constrained finite basis physics-informed neural networks (FBPINNs) for solving the Stokes equations in two-dimensional domains with many circular perforations. The method combines overlapping domain decomposition with FBPINN window functions, distance-function-based hard constraints for the no-slip condition on the perforations and for the wall boundary, non-dimensionalisation, adaptive loss weighting, and residual-based adaptive collocation refinement. The authors provide a universal approximation proof for FBPINNs, a Fourier analysis of the ansatz, and numerical experiments for periodic (16, 36, 64 perforations) and random (36, 64, 100 perforations) arrangements, comparing against soft- and hard-constrained PINNs and against Taylor–Hood finite element reference solutions.","tokens_in":30712,"tokens_out":3941,"duration_ms":40151,"significance":"If the main claims hold, this work provides a practical and scalable PINN framework for Stokes flow in highly perforated microstructures, a problem of direct relevance to composite manufacturing and microfluidics. The paper has clear strengths: the numerical validation uses external finite element references rather than constructed analytic solutions; the 64-periodic sensitivity study reports seed-averaged errors over five random seeds; and the theoretical sections give a proof of universal approximation and an explicit Fourier-space expression for the FBPINN ansatz. These are substantial contributions. However, the central claim that the implemented ansatz enforces both wall and no-slip boundary conditions exactly is not demonstrated, and the presented experiments do not fully support the headline claim of convergence being only weakly affected by the number of perforations, since most error curves are single runs.","major_comments":[{"comment":"The actual composite hard-constraint ansatz used in all numerical experiments is never written down. The text says that “both the wall and no-slip boundary hard constraints” are applied globally “as in (26)”, but Eq. (26) is only a single-constraint illustration. If the velocity is formed as C_w[l_{∂Ωp} \\bar u] with C_w defined in (44), then on ∂Ωp the lifting term g[1 - tanh(a_w l_{∂Ωw})] is nonzero: for the 64-periodic case it is about 3e-4 relative to |g|. If the ordering is reversed, the wall condition fails. Thus no stated formula satisfies both boundary conditions exactly, and the class (8) — and therefore the error estimate (9) — does not apply to the implemented method. The compatibility warning in §3.2 is not resolved in the numerical section. The authors need to state the exact composite formula, verify analytically or numerically that both boundary conditions are enforced, and quantify the boundary residual if it is not exactly zero.","section":"§3.2, Eqs. (26), (33), (44)"},{"comment":"The central convergence curves that support the claim of convergence “only weakly affected by the number of perforations” are single training runs without error bars or seed statistics. Seed-averaged results are reported only for the 64-periodic case (Tables 1–5), not for 16 and 36 periodic perforations or for 36, 64, and 100 random perforations. For a claim that is explicitly about robustness across problem sizes, the headline comparisons should include at least a few seeds per case or an explicit statement that the plotted curves are representative single runs.","section":"§4, Figs. 6–8 and Fig. 15"},{"comment":"The curvature and stiffness analysis of Δl_{∂Ωp} analyzes only the no-slip factor in isolation. The implemented ansatz composes l_{∂Ωp} with the wall operator C_w defined in (44), so the Laplacian of the constrained velocity also contains derivatives of tanh(a_w l_{∂Ωw}) and of the lifting term g[1 - tanh(a_w l_{∂Ωw})]. The claimed stiffness reduction from hard constraints is therefore not demonstrated for the actual formula used in the experiments. This issue is directly connected to the missing composite formula in the first comment; the analysis should be carried out for the full ansatz.","section":"§3.2, Eq. (28) and Fig. 3"}],"minor_comments":[{"comment":"In Eq. (2), writing the compatibility condition as ∫_{∂Ωw} g·η_w ds = ∫Ω div u dx = 0 is redundant and slightly confusing; the middle equality follows from the divergence theorem and the zero-divergence condition, but the notation suggests an independent condition. Please simplify.","section":"§2.1, Eq. (2)"},{"comment":"The sentence “we apply both the wall and no-slip boundary hard constraints globally as in (26)” is difficult to interpret because Eq. (26) contains only one distance function and one boundary datum. Please give the explicit global formula and explain how the two constraints are combined.","section":"§3.2, after Eq. (36)"},{"comment":"The notation “λ_k ← Gradient-based weight scaling (∇θ,ψLλk(θk,ψk), λ_k)” is ambiguous; specify whether all weights (λ_div, λ_b) are updated together and how the moving average of Eq. (11) is applied.","section":"Algorithm 1, line 13"},{"comment":"The caption of Fig. 14 does not identify which curves in panels (b) and (c) correspond to which model variant; please add a legend or refer to the colors/line styles described in the text.","section":"Figure 14"},{"comment":"The data availability statement says code will be made available upon publication; for a numerical methods paper, providing the code or a public repository at submission would strengthen reproducibility.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal and addresses a timely problem. The main technical gap — the unspecified composite hard-constraint formula — is load-bearing for the paper's central claims, but it appears fixable by rewriting the ansatz or by choosing a lifting function that vanishes on ∂Ωp and verifying the boundary residual. I do not see a reason to reject the paper on the basis of the current numerical evidence, which is otherwise well grounded in external FEM comparisons."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is worth a look: it combines finite-basis PINNs with global hard constraints for Stokes flow in perforated domains, and shows convincing numerical results up to 100 perforations. The Fourier analysis of FBPINNs (Theorem 3.1) and the universal approximation proposition are correct and useful. The real contribution is the empirical demonstration that hard constraints plus domain decomposition beat soft-constrained PINNs by a wide margin, with seed-averaged sensitivity tables that lend some robustness.\n\nThe main soft spot is in the hard-constraint design. The paper never states the composite ansatz that enforces both the wall condition and the no-slip condition. The general form (33) is given, but the actual implementation is described only as 'applying both globally as in (26)', which is not the same thing. If you form uθ = C_w[l_{∂Ω_p}\\bar u], then on ∂Ω_p the wall lifting term g[1−tanh(a_w l_{∂Ω_w})] leaves a residual that is small but nonzero. The stress-test note computes ~3e-4 for the 64-periodic case, which is below the reported errors, so this does not overturn the numerical results. But it does break the paper's claim of exact boundary enforcement, and it means the error bound (9), which relies on uθ ∈ NN†_Θ, does not strictly apply. The authors need to either write down the exact composite formula and prove it satisfies both conditions, or soften the language to 'approximate hard constraints' and adjust the theory accordingly.\n\nOther issues are minor: the central error curves (Figures 6–8, 15) are single runs without error bars, though the sensitivity tables partially compensate; no code is provided; and the comparison with regular PINNs uses different hyperparameters, so the speed advantage is not fully controlled. The literature review is honest and the math is in good shape.\n\nOverall, this is a solid paper with a genuine contribution. The stress-test concern is real but not fatal. I'd send it to a serious referee, expecting major revision to clarify the constraint construction and release code. If you work on SciML for porous media, it's worth your time.","headline":"A solid FBPINN+hard-constraint paper with unstated composite ansatz; the method works numerically, but the exactness claim needs fixing.","tokens_in":31271,"tokens_out":4768,"would_cite":true,"duration_ms":43732,"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":"This paper claims that hard-constrained finite basis PINNs—overlapping localized networks with exact boundary encoding—solve Stokes flow in highly perforated domains with relative errors near 1% at 100 perforations and convergence nearly…","keywords":["physics-informed neural networks","finite basis PINNs","hard boundary constraints","Stokes flow","perforated domains","spectral bias","domain decomposition","distance functions"],"falsifier":"Run the 100-random-perforation benchmark with the wall lifting function $\\bar g$ chosen so that it has nonzero support on one perforation boundary; the paper's compatibility warning predicts degraded accuracy or violated no-slip, so observing clean no-slip plus under-1% error would contradict its stated mechanism. Alternatively, measure the maximum of $|\\Delta l_{\\partial\\Omega_p}|$ and train with $m=12,a=25$: the paper predicts errors grow because of curvature stiffness, so flat errors there would falsify that part of the claim.","tokens_in":30160,"feed_emoji":"🌊","tokens_out":7718,"duration_ms":77051,"temperature":0.7,"pith_summary":"This paper sets out to establish that a neural-network PDE solver can remain accurate for Stokes flow in domains with many circular obstacles, a setting where ordinary PINNs degrade sharply. The proposed combination is finite basis PINNs—domain decomposition into overlapping, normalized subdomains blended by window functions—with hard constraints that encode the no-slip condition on every perforation boundary exactly, removing the boundary penalty term. The paper argues this mitigates spectral bias, reduces gradient stiffness, and makes convergence only weakly dependent on the number of perforations. A sympathetic reader would care because flow around dense fibre-like obstructions is exactly the multi-scale, high-frequency regime where soft-penalty PINNs struggle, and applications such as resin flow through fibre bundles require many holes. The numerical evidence includes relative $L^2$ errors below 1% for random arrangements of 100 perforations after 100,000 iterations.","feed_headline":"Under 1% error: neural solver handles 100-hole Stokes flow","feed_subtitle":"Hard boundary constraints and localized subnetworks keep convergence nearly independent of the number of perforations.","key_machinery":"The load-bearing construction is the hard-constrained FBPINN ansatz $$u_\\$\\theta$(x)=C_w\\left(\\sum_{i=1}^N \\omega_i(x)\\, \\mathrm{unnorm}_u\\circ C_p^i[NN_i\\circ \\mathrm{norm}_i](x)\\right),$$ where the $\\omega_i$ are overlapping cosine partition-of-unity windows and each local network sees coordinates normalized to a parent subdomain. The no-slip constraint uses the aggregated inverse distance function $$l_{\\partial\\Omega_p}(x)=\\tanh\\!\\left(a\\left(\\sum_{k=1}^K \\phi_{\\mathrm{disk}}(x,c_k)^{-m}\\right)^{-1/m}\\right),$$ which vanishes on every perforation boundary, while $C_w$ is an affine lifting operator that imposes the wall profile. Because $\\Delta(l\\,\\bar u)=l\\Delta \\bar u+2\\nabla l\\cdot\\nabla \\bar u+\\bar u\\,\\Delta l$, the distance function injects the perforation geometry into the Stokes residual as an inductive bias, and the paper studies how the localisation parameters $m,a$ control the curvature $\\Delta l_{\\partial\\Omega_p}$ and hence the stiffness of training.","core_discovery":"The central discovery is that soft boundary enforcement is the main bottleneck: once no-slip is imposed by multiplying the network output by a smoothed distance-to-perforations function, and the approximation is localized through the FBPINN ansatz, the Stokes residual can be trained to below 1% relative error with dozens to 100 perforations. The paper reports that hard constraints alone are not enough—hard-constrained single-network PINNs still worsen as perforation count grows—and that domain decomposition, input normalization, non-dimensionalisation, and residual-based adaptive collocation must act together. It also provides a Fourier decomposition of the FBPINN ansatz showing two opposing mechanisms: subdomain rescaling lowers the effective frequencies seen by local networks, while window localization broadens the spectral coupling kernel and introduces convolution-induced spectral leakage. The accompanying theory states a universal approximation result for FBPINNs and explains why hard constraints recover stronger $H^1$ velocity and $L^2$ pressure error norms compared with soft penalties.","pith_inferences":["If the mechanism is as described, the method should extend to three-dimensional fibre arrangements and to more than 100 perforations by keeping subdomain size tied to obstacle spacing rather than domain size; the paper does not run such cases.","The Fourier analysis suggests a quantitative design rule the authors do not state explicitly: overlap should be large enough that the window spectrum stays narrow relative to the rescaled target frequencies, so one could test whether the optimal overlap ratio tracks the spectral width of the window.","A natural stress test is replacing circular holes with sharp-cornered obstacles; the paper's regularity assumptions and its curvature-stiffness argument imply accuracy would degrade, and the architecture's handling of corner singularities is untested."],"forward_implications":["Under the paper's protocol, increasing periodic perforations from 16 to 64 does not produce the accuracy collapse seen for soft-constrained or single-network hard-constrained PINNs; the 64-hole case reaches relative $L^2$ errors around half a percent after 100,000 iterations.","Hard-constrained FBPINNs also handle random arrangements, with five different 100-perforation geometries all ending below 1% relative $L^2$ error after 100,000 iterations.","In the reported 64-perforation test, the FBPINN is faster in wall-clock time than the regular PINN baseline, about 20 versus 42 minutes on an A100 after 100,000 iterations.","Increasing the subdomain overlap from 2.0 to 2.4 reduces errors by about 21% but raises compute time by about 58%, so the paper recommends an overlap ratio of 2.0 as a balance.","Gradient alignment stays positive early for the hard-constrained FBPINN, whereas the soft-constrained FBPINN remains near $-0.7$, linking hard constraints to reduced boundary-induced gradient conflict."],"supporting_citations":[{"why":"Supplies the FBPINN ansatz, overlapping window functions, and domain-decomposition training structure that the method builds on.","marker":"[46]"},{"why":"Supplies the aggregated inverse distance function and hyperbolic tangent modulation used to impose boundary conditions exactly.","marker":"[35]"},{"why":"Provides the error estimates showing hard constraints give $H^1$ velocity and $L^2$ pressure approximation, motivating the hard-constrained ansatz.","marker":"[49]"},{"why":"Provides the regular PINN baseline with Fourier features and loss balancing against which the FBPINN results are compared.","marker":"[29]"},{"why":"Reports the earlier PINN treatment of Stokes flow in perforated domains that this paper identifies as poorly scalable.","marker":"[4]"},{"why":"Supplies the stiff gradient-flow and Hessian ill-conditioning analysis used to explain why soft boundary penalties fail.","marker":"[32]"},{"why":"Supplies the residual-based adaptive collocation point resampling strategy used in the training protocol.","marker":"[71]"}],"fun_headline_variants":["FBPINNs with hard constraints tame perforated Stokes flow","Localized PINNs keep error low despite many perforations","Under 1% error: FBPINNs handle 100-hole flow","Hard constraints plus localization solve 100-hole Stokes flow","Neural flow solver hits under 1% error with 100 holes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the assumption that the hand-built distance-to-the-perforations function can be multiplied into the network output and composed with the wall lifting so that both boundary conditions hold exactly while the curvature it injects into the Stokes residual stays mild enough to train; this is not a theorem, and the paper itself warns that wall-lifting support can break compatibility.","fun_headline_variants_meta":{"raw":{"variants":["FBPINNs with hard constraints tame perforated Stokes flow","Localized PINNs keep error low despite many perforations","Under 1% error: FBPINNs handle 100-hole flow","Hard constraints plus localization solve 100-hole Stokes flow","Neural flow solver hits under 1% error with 100 holes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001091,"raw_usage":{"total_tokens":4572,"prompt_tokens":974,"completion_tokens":3598,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":3510}},"tokens_in":590,"tokens_out":3598,"duration_ms":41442,"temperature":1.0,"reasoning_tokens":3510,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:24:10.140549+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the 100-random-perforation benchmark with the wall lifting function $\\bar g$ chosen so that it has nonzero support on one perforation boundary; the paper's compatibility warning predicts degraded accuracy or violated no-slip, so observing clean no-slip plus under-1% error would contradict its stated mechanism. Alternatively, measure the maximum of $|\\Delta l_{\\partial\\Omega_p}|$ and train with $m=12,a=25$: the paper predicts errors grow because of curvature stiffness, so flat errors there would falsify that part of the claim.","supporting_citations":[{"cited_title":"Moseley, A","cited_arxiv_id":null,"evidence_quote":"Supplies the FBPINN ansatz, overlapping window functions, and domain-decomposition training structure that the method builds on."},{"cited_title":"Zeinhofer, R","cited_arxiv_id":null,"evidence_quote":"Provides the error estimates showing hard constraints give $H^1$ velocity and $L^2$ pressure approximation, motivating the hard-constrained ansatz."},{"cited_title":"Korolev, T","cited_arxiv_id":null,"evidence_quote":"Reports the earlier PINN treatment of Stokes flow in perforated domains that this paper identifies as poorly scalable."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the residual-based adaptive collocation point resampling strategy used in the training protocol."}],"review_version":1}