{"id":"c1af0a25-c316-4c63-9803-de98fead26a8","arxiv_id":"2605.26358","paper_version":2,"verdict":"UNVERDICTED","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"DARSM embeds a neural network inside an implicit algebraic Reynolds stress model derived from transport equations, trains it end-to-end via adjoint PDE optimization, and reports 2-4x average velocity error reduction plus generalization from attached to separated flows on duct and hill benchmarks.","lead":"The paper introduces DARSM, a neural network that predicts parameters inside a physics-derived algebraic equation for turbulence stresses in RANS simulations. If the reported error reductions and cross-regime generalization hold, the approach could make routine engineering flow predictions more reliable without requiring full high-fidelity simulations.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Validity of weak-equilibrium assumption in separated periodic-hill flows remains unverified and is central to the cross-regime generalization claim","rationale":"The reader's weakest_assumption directly isolates the load-bearing conceptual risk for the strongest_claim (2-4× error reduction and cross-regime generalization). The abstract's own wording on regime change reinforces that this is the point least secured by the reported evidence. No other internal inconsistency is visible from the abstract alone.","tokens_in":1848,"tokens_out":356,"duration_ms":17776,"concrete_test":"On the periodic-hill DNS at Re=10595, extract the anisotropy tensor in the separation region; insert the trained NN outputs into the implicit algebraic equation and solve for the predicted stresses; compute the L2 deviation from DNS stresses and compare to the deviation obtained from the baseline RANS closure in the same region. A deviation >30% larger than baseline would indicate the assumption fails to support the generalization claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The DARSM reduces the Reynolds stress transport equations to an implicit algebraic form via the weak-equilibrium assumption, then trains a NN on invariants from attached square-duct flows only. The headline result asserts that this model generalizes without retraining to separated periodic-hill flows despite an explicit \"regime change in the underlying physics.\" If the weak-equilibrium approximation (neglect of convective and diffusive transport relative to production and pressure-strain) introduces large, regime-dependent errors in the separation bubble, the algebraic closure itself becomes inaccurate there; any observed velocity-error reduction would then not be attributable to the physics-structured ML model. The abstract provides no separate validation that the algebraic stresses match full transport or DNS stresses in the separated regime.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript introduces the Deep Algebraic Reynolds Stress Model (DARSM), which embeds a neural network inside an implicit algebraic Reynolds stress closure derived from the Reynolds stress transport equations under the weak-equilibrium assumption. A NN maps invariants to closure coefficients; the model is trained end-to-end by minimizing the RANS residual on square-duct data and is reported to generalize without retraining to periodic-hill flows, yielding 2–4× average velocity-error reductions (peak 12×) relative to baseline RANS while outperforming five other ML closures.","tokens_in":2032,"tokens_out":580,"duration_ms":19883,"significance":"If the cross-regime generalization holds, the work would be significant for turbulence modeling: it supplies a physics-structured algebraic form that mitigates distribution shift, demonstrates training on small attached-flow datasets, and introduces an adjoint formulation that exploits the implicit-explicit solver structure for efficient optimization. These elements address long-standing obstacles in ML-RANS closures.","major_comments":[{"comment":"The headline generalization claim (abstract) that a model trained only on attached square-duct flows accurately predicts separated periodic-hill flows rests on the weak-equilibrium assumption remaining valid in the separation bubble. No direct comparison of the resulting algebraic Reynolds stresses against either full Reynolds-stress transport solutions or DNS data is supplied for the hill geometry, leaving open the possibility that observed velocity improvements arise from other factors rather than the structured ML closure.","section":"abstract and benchmark results"},{"comment":"Quantitative performance statements (2–4× average error reduction, 12× peak) are given without reported error bars, explicit train/test splits, baseline implementation details, or sensitivity tests to the weak-equilibrium assumption (abstract). These omissions make it impossible to judge whether the reported gains are statistically robust or sensitive to the regime change.","section":"abstract and evaluation sections"},{"comment":"The derivation of the implicit algebraic form (via weak equilibrium) is presented as the key physics constraint, yet the manuscript does not quantify the magnitude of the neglected convective and diffusive terms in the periodic-hill separation region; if those terms are O(1) relative to production and pressure-strain, the algebraic closure itself becomes inconsistent with the underlying transport equations in the test regime.","section":"model derivation"}],"minor_comments":[{"comment":"Notation for the invariants and the neural-network output coefficients should be unified between the derivation and the results tables to avoid reader confusion.","section":"model formulation"},{"comment":"Figure captions for the velocity and stress profiles should explicitly state the Reynolds numbers and the precise definition of the reported error norms.","section":"figures"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the detailed and constructive report. The comments raise important points on validation of the weak-equilibrium assumption, statistical robustness, and consistency of the algebraic closure. We respond point-by-point below and will make revisions to address the concerns.","responses":[{"response":"We agree that a direct comparison of the modeled algebraic Reynolds stresses to DNS data on the periodic hill would strengthen the evidence that improvements stem from the structured closure rather than other factors. While mean-velocity accuracy is the primary RANS objective and our cross-method comparisons support the claim, we will add a new subsection and figure comparing predicted Reynolds-stress components against available DNS data in the separation region.","revision_made":"yes","referee_comment":"[abstract and benchmark results] The headline generalization claim (abstract) that a model trained only on attached square-duct flows accurately predicts separated periodic-hill flows rests on the weak-equilibrium assumption remaining valid in the separation bubble. No direct comparison of the resulting algebraic Reynolds stresses against either full Reynolds-stress transport solutions or DNS data is supplied for the hill geometry, leaving open the possibility that observed velocity improvements arise from other factors rather than the structured ML closure."},{"response":"The reported benchmarks are deterministic fixed-mesh simulations, so ensemble-based error bars do not apply. We will expand the evaluation section to explicitly document the train/test splits (square-duct cases for training, periodic-hill cases for testing), provide full baseline implementation details and references, and include a sensitivity analysis to the weak-equilibrium assumption by comparing results with and without selected neglected terms.","revision_made":"yes","referee_comment":"[abstract and evaluation sections] Quantitative performance statements (2–4× average error reduction, 12× peak) are given without reported error bars, explicit train/test splits, baseline implementation details, or sensitivity tests to the weak-equilibrium assumption (abstract). These omissions make it impossible to judge whether the reported gains are statistically robust or sensitive to the regime change."},{"response":"We concur that quantifying the neglected terms is necessary to evaluate consistency of the algebraic form in the separated regime. Using the DNS data already employed for the periodic-hill benchmark, we will compute and report the relative magnitudes of convective, diffusive, and production/pressure-strain terms inside the separation bubble and discuss implications for the weak-equilibrium assumption.","revision_made":"yes","referee_comment":"[model derivation] The derivation of the implicit algebraic form (via weak equilibrium) is presented as the key physics constraint, yet the manuscript does not quantify the magnitude of the neglected convective and diffusive terms in the periodic-hill separation region; if those terms are O(1) relative to production and pressure-strain, the algebraic closure itself becomes inconsistent with the underlying transport equations in the test regime."}],"tokens_in":1587,"tokens_out":597,"duration_ms":28083,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing to know is that this paper derives an implicit algebraic Reynolds stress equation from the transport equations under the weak-equilibrium assumption, lets a neural net supply its coefficients from invariants, and optimizes the whole thing through the RANS solver using a custom adjoint. They report 2-4x lower average velocity error than baseline RANS on square-duct and periodic-hill benchmarks, with the duct-trained model generalizing to the hills without retraining, and it beats the five ML methods listed.\n\nThe combination of the algebraic structure, invariant-based NN, and adjoint handling of the stiff implicit solver is the concrete novelty relative to the baselines. End-to-end training through the governing equations is a practical way to reduce distribution shift, and the quantitative results on named cases are stated clearly.\n\nThe soft spot is the weak-equilibrium assumption itself. Training happens only on attached, anisotropy-dominated duct flows, yet the headline result is generalization to separated hill flows where the neglected convective and diffusive terms could be larger. The abstract gives no separate check that the algebraic stresses match full transport or DNS data inside the separation bubble, so it is not yet clear how much of the error reduction comes from the structured model versus how much might be limited by the assumption breaking. Data-split and error-bar details are also missing, which makes the numbers harder to weigh immediately.\n\nThis is for CFD and turbulence-modeling groups who already run RANS and want to test structured ML closures. A reader focused on engineering accuracy rather than new theory would get the most from it. The work is grounded enough and the application area important enough that it deserves a serious referee, even with the assumption needing more direct validation.","headline":"DARSM puts a NN inside a weak-equilibrium algebraic Reynolds stress model and trains end-to-end with adjoints, delivering 2-4x error drops on duct and hill cases with claimed cross-regime generalization, but the assumption's validity in separation is unverified.","tokens_in":2532,"tokens_out":436,"would_cite":false,"duration_ms":30204,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"A neural network embedded in an algebraic Reynolds stress equation improves RANS velocity predictions by 2-4 times on average and generalizes from attached to separated flows without retraining.","keywords":["Reynolds-averaged Navier-Stokes","turbulence modeling","algebraic Reynolds stress model","deep learning closure","machine learning for CFD","square duct flow","periodic hill flow","weak equilibrium assumption"],"falsifier":"Measure whether DARSM still reduces velocity error by at least 2 times relative to baseline RANS when applied to a new separated flow geometry whose mean-flow statistics visibly violate the weak-equilibrium assumption.","tokens_in":2750,"feed_emoji":"🌊","tokens_out":813,"duration_ms":18043,"temperature":0.7,"pith_summary":"The paper introduces the Deep Algebraic Reynolds Stress Model (DARSM), which embeds a neural network inside an implicit algebraic closure derived from the Reynolds stress transport equations under the weak-equilibrium assumption. The network learns mappings from flow invariants to closure parameters, and the entire system is optimized end-to-end through the governing PDEs using derived adjoint equations that respect the solver's implicit-explicit structure. On square-duct and periodic-hill test cases, DARSM lowers average velocity error relative to standard RANS by factors of 2-4 across Reynolds numbers and geometries, with some cases improving by 12 times. The model trained only on attached, anisotropy-dominated duct flows transfers directly to separated hill flows, a change in underlying physics regime. It also outperforms five other machine-learning closure approaches that either train offline or bypass the equations entirely.","feed_headline":"Neural closure cuts RANS velocity error by 2-4 times on average","feed_subtitle":"Model trained on square ducts generalizes without retraining to separated flows in periodic hills.","key_machinery":"The Deep Algebraic Reynolds Stress Model (DARSM), a neural network that supplies parameters to an implicit algebraic Reynolds stress equation derived under the weak-equilibrium assumption, optimized through the governing PDEs with adjoint equations that exploit the solver's implicit-explicit structure.","core_discovery":"DARSM trains a neural network to supply empirical coefficients inside an implicit algebraic Reynolds stress relation obtained from the full transport equations under the weak-equilibrium assumption; end-to-end differentiation through the coupled RANS solver via custom adjoint equations removes distribution shift, yielding 2-4 times lower average velocity error than baseline RANS on both square-duct and periodic-hill benchmarks while generalizing from attached to separated regimes without retraining.","pith_inferences":["The same hybrid structure could be tested on other RANS closures that admit algebraic reductions, such as scalar-flux or heat-transfer models.","If the weak-equilibrium assumption holds across a wider class of engineering flows, DARSM-style models might reduce the need for full Reynolds-stress transport simulations in many industrial cases.","The adjoint derivation technique may apply to other stiff implicit-explicit solvers in computational fluid dynamics where direct differentiation fails."],"forward_implications":["RANS simulations of attached and mildly separated turbulent flows can achieve 2-4 times lower average velocity error with a single trained DARSM model.","A model trained exclusively on square-duct data transfers directly to periodic-hill flows without retraining, indicating regime-level generalization.","End-to-end optimization through the coupled implicit closure eliminates the distribution shift that affects offline-trained ML closures.","DARSM outperforms offline training, tensor-basis networks, field-inversion methods, DeepONets, and physics-informed networks on the same benchmarks.","The approach requires only small high-fidelity datasets because the algebraic structure supplies most of the physics."],"fun_headline_variants":["DARSM reduces RANS velocity error 2-4x across regimes","Algebraic model generalizes RANS to new flow regimes","End-to-end training via adjoints fixes ML-RANS shift","DARSM outperforms five ML methods on square duct and hills"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The weak-equilibrium assumption used to collapse the Reynolds stress transport equations into an implicit algebraic form stays accurate enough for the attached duct flows in training and the separated hill flows in testing.","fun_headline_variants_meta":{"raw":{"variants":["DARSM reduces RANS velocity error 2-4x across regimes","Algebraic model generalizes RANS to new flow regimes","End-to-end training via adjoints fixes ML-RANS shift","DARSM outperforms five ML methods on square duct and hills"]},"model":"grok-4.3","cost_usd":0.010351,"raw_usage":{"total_tokens":4631,"prompt_tokens":766,"num_sources_used":0,"completion_tokens":69,"cost_in_usd_ticks":103512000,"prompt_tokens_details":{"text_tokens":766,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3796,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":766,"tokens_out":69,"duration_ms":30278,"temperature":1.0,"reasoning_tokens":3796,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-29T20:05:53.760307+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Measure whether DARSM still reduces velocity error by at least 2 times relative to baseline RANS when applied to a new separated flow geometry whose mean-flow statistics visibly violate the weak-equilibrium assumption.","supporting_citations":[],"review_version":1}