{"id":"bd74b68e-54f2-45a5-a3de-823703265fa6","arxiv_id":"2607.20347","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A high-order flux-reconstruction actuator-line framework predicts vertical-axis wind turbine wakes and power coefficients on fixed Cartesian grids, using a streamtube correction to recover inflow induction.","lead":"This paper couples a high-order flux-reconstruction flow solver with an actuator-line model to simulate vertical-axis wind turbine blades without resolving their geometry. The method reproduces wake profiles and power coefficients close to higher-fidelity simulations near optimal operation, at much lower computational cost.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central CP claim is carried by the externally imposed DMST correction, not by the FR/CPR-ALM coupling; the solver's resolved induction feedback is negligible, so the 6% LES-ALM agreement is not a validation of the high-order solver.","rationale":"The paper is honest about the negligible induction feedback (Sec. III C and III D 1). The central claim in the abstract and conclusion is that the FR/CPR-ALM framework provides accurate power prediction. The concern is not that the method is unconventional (DMST is standard for VAWT engineering models), but that the validation does not actually test the coupled solver's ability to predict performance. The quantities that are tested—wake profiles at one tip-speed ratio—are not the quantities in the headline claim. A single self-consistent fine-mesh case would either confirm that the resolved flow can produce the induction needed for the correct CP, or demonstrate that the current coarse-mesh approach relies entirely on the external DMST correction. This is a tractable computational check, since the fine mesh used elsewhere in the paper (279×279) is already available. The reader's CONDITIONAL verdict seems appropriate; the concern does not change that verdict.","tokens_in":15412,"tokens_out":3931,"duration_ms":32286,"concrete_test":"Run one self-consistent simulation at λ=4.0 on a mesh satisfying Eq. (42) (e.g., h≤0.032 m) with the DMST correction disabled and blade forces sampled directly from the resolved velocity. Compare the time-averaged CP against (a) the coarse-mesh zero-induction result, (b) the coarse-mesh DMST-corrected result, and (c) the 3D LES-ALM reference value. If the self-consistent CP is not within ~6% of the LES-ALM value, or if it differs from the DMST-corrected CP by more than, say, 10%, then the reported accuracy is attributable to the DMST model rather than the FR/CPR-ALM coupling.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Sec. III D 1 the authors report that direct sampling of the resolved velocity at the blade yields a power curve 'nearly coincident with the zero-induction result,' and Sec. III C shows that on the production coarse mesh ε=0.043 m exceeds the chord floor, smearing forces over ~0.17 m and diluting induced velocity. The subsequent DMST correction (Sec. III D 2) replaces the blade-inflow velocity with U∞(1-a_u) / U∞(1-2a_u) in the force computation, but the resolved FR/CPR flow field is not correspondingly decelerated at the actuator; the momentum source terms are therefore evaluated with a velocity field inconsistent with the forces they impose. Consequently the predicted CP(λ) curve is essentially a DMST/BEMT result with airfoil polars and Boeing-Vertol corrections, and the stated 'within 6%' agreement with 3D LES-ALM validates that engineering model, not the coupled FR/CPR-ALM solver. The wake validation is a legitimate positive result, but it does not rescue the performance claim, since the wake dynamics are generated by the same DMST-modified forces and the near-wake comparison is only at λ=1.9, far from the optimal regime. The central claim should be conditioned on the DMST induction model being valid for the configuration.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a two-dimensional FR/CPR (flux reconstruction / correction procedure via reconstruction) solver coupled to a rotating actuator-line model (ALM) with isotropic Gaussian force projection for vertical-axis wind turbine (VAWT) aerodynamics on fixed Cartesian grids. A modified Boeing-Vertol dynamic stall correction is used, and a Double Multiple Streamtube (DMST) induction model is imposed when the resolved flow provides negligible induction feedback. The manuscript reports a near-wake validation against Bachant & Wosnik and Hezaveh et al., and a power-coefficient curve CP(λ) that agrees with 3D LES-ALM data to within about 6% near the optimal tip-speed ratio. It also documents azimuthal loading, lift hysteresis, wake structure, and computational cost. The main claim is that the FR/CPR-ALM framework is an accurate and efficient geometry-free tool for VAWT analysis.","tokens_in":15737,"tokens_out":5278,"duration_ms":47192,"significance":"The topic is timely and the paper is unusually candid about a key limitation: the manuscript itself states in Sec. III D 1 that direct sampling of the resolved velocity yields a power curve nearly coincident with the zero-induction result, and in Sec. III C that on the production mesh the Gaussian kernel is mesh-controlled, diluting induced velocity. This confirms the central concern that the reported CP(λ) curve is effectively a DMST/BEMT prediction using tabulated static polars and an empirical dynamic-stall correction, not a prediction produced by the coupled FR/CPR-ALM solution. The wake comparison at x/D=1 and λ=1.9 is a legitimate positive result with quantitative data, and the kernel-resolution criterion is useful for practitioners, but the wake validation alone does not support the headline 6% performance claim. The paper also compares against LES-ALM results that use the same actuator-line methodology and static polar source, so the performance comparison is not independent of the engineering model being implicitly tested. With a reframing of the claims and additional supporting analysis, the contribution could be a useful demonstration of high-order ALM wake simulation wi","major_comments":[{"comment":"The central CP(λ) claim is not a direct product of the coupled FR/CPR-ALM solver. The manuscript states in Sec. III D 1 that direct sampling of the resolved velocity at the blade yields a power curve 'nearly coincident with the zero-induction result,' and Sec. III C shows that the production mesh operates in the mesh-controlled regime with ε=0.043 m, smearing the force over ~0.17 m. The DMST correction in Eqs. (43)-(44) then replaces the inflow velocity with U∞(1-au) or U∞(1-2au) in the force evaluation (Eq. (27)), while the resolved flow field is not decelerated to match this velocity. The momentum source in Eq. (39) is therefore evaluated with a blade-inflow velocity inconsistent with the resolved field it acts on. Consequently the 'within 6%' agreement with LES-ALM in Sec. III D 3 validates the DMST/BEMT model with Sheldahl-Klimas polars and the Boeing-Vertol correction, not the high-","section":"Sec. III D 1, III D 2 and III C"},{"comment":"The near-wake validation is a genuine positive result, but it is limited and not independent enough to carry the performance claim. The comparison is made only at λ=1.9, far below the optimal regime (λ≈4-4.5), and in the deep-stall regime where 2D simulations are known to behave poorly. The manuscript itself notes that the 2D formulation overpredicts the deficit depth relative to experiment and LES. The reference 'ALM-LES' of Hezaveh et al. shares the same actuator-line modeling approach and, like the present work, uses Sheldahl-Klimas static airfoil data, so the agreement partly reflects a common modeling base rather than an independent validation of the FR/CPR discretization. Quantified error measures (e.g., velocity-deficit error norms) and a comparison at a near-optimal tip-speed ratio would substantially strengthen the validation. As written, the wake result supports the claim that","section":"Sec. III B, Fig. 5"},{"comment":"The 'within 6%' agreement is asserted only around the optimum, while the off-design behavior is much worse and is not discussed as a limitation. Table II shows that at λ=2.0 the present method gives ⟨CP⟩=0.075 while the 3D LES-ALM reference is −0.05, a discrepancy of order 0.125 in CP; at λ=1.5 the present value is 0.027 and the reference is not reported. Even near the optimum, the λ=4.5 point differs from the reference by approximately 6.6% (0.439 vs 0.47), slightly above the stated 6%. The selected range λ=3.5-5.5 brackets the peak and gives good agreement, but the paper should state precisely which points are included in the 'within 6%' claim and explicitly acknowledge the large low-λ discrepancies. Without this, the abstract's 'matches high-fidelity LES-ALM data to within 6%' is misleading as a global validation statement.","section":"Sec. III D 3, Table II"},{"comment":"The results depend on several user-set parameters for which no sensitivity study is reported: the chord-fraction constant κ_c=4.3 in Eq. (38), the kernel truncation radius r_c=4ε, the DMST under-relaxation factor 0.9/0.1 in Eq. (46), and the unspecified bypass threshold in Eq. (33) that removes the singularity as α*_L→α0. These parameters directly affect the smeared force distribution and the induction update, and therefore the reported CP and wake fields. In particular, κ_c controls whether the kernel is chord- or mesh-controlled, and the choice κ_c=4.3 is not justified beyond a single value. A parameter sensitivity study, or at least a statement of the threshold value and its influence on the dynamic-stall correction, is needed to establish that the headline results are robust rather than tuned.","section":"Sec. II B, Eqs. (38), (45)-(46)"}],"minor_comments":[{"comment":"Typos: 'one of the most fastest-growing' should be 'one of the fastest-growing'; 'V AWTs' appears with inconsistent spacing in several places.","section":"Abstract and Introduction"},{"comment":"Reference 37 is listed as 'unpublished' and reference 33 is listed as a preprint 'submitted/accepted'; such references should be updated or marked as 'in preparation' with a DOI if available.","section":"References"},{"comment":"The figure captions and axes would benefit from explicit labels for the static and Boeing-Vertol cases in the legend, and from stating the azimuthal averaging procedure in the caption or text.","section":"Fig. 7 and Fig. 8"},{"comment":"The phrase 'high-fidelity three-dimensional LES-ALM' is potentially misleading: the reference is an actuator-line LES, not a blade-resolved simulation. The text should say 'LES with an actuator-line model' to avoid implying geometric resolution of the blades.","section":"Sec. III D 3"}],"recommendation":"major_revision","confidential_remarks":"The paper's main problem is not that the DMST correction is invalid per se, but that the title, abstract, and conclusion claim the FR/CPR-ALM framework 'predicts' the power curve to 6%, when the paper's own evidence shows the resolved induction feedback is negligible and the performance curve is essentially produced by the external DMST model. This is fixable by substantial reframing: separate the directly sampled zero-induction result from the DMST-corrected result, characterize the inconsistency between prescribed and resolved inflow, present the CP comparison as a validation of the DMST-corrected engineering model embedded in the solver, and add a sensitivity study for the free parameters. The wake validation is a useful contribution and the kernel-resolution criterion is practical, so the paper is not a reject candidate. I would support reconsideration after major revision if the authors are willing to limit the claims to what the coupled solver actually demonstrates."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: this is a genuinely new and mostly honest engineering paper, but the headline claim—that the FR/CPR-ALM framework predicts VAWT power within 6% of 3D LES-ALM—is not supported as a validation of the coupled solver. The paper's own Sec. III D 1 shows the resolved induction feedback is negligible, and the DMST correction then supplies the inflow velocity that sets the forces. So the CP curve is essentially a DMST/BEMT calculation. The wake validation is real, but it doesn't rescue the performance claim.\n\nWhat is new: rotating ALM inside a high-order FR/CPR solver for VAWT on fixed Cartesian grids, with Gaussian force projection and a modified Boeing-Vertol dynamic stall model. The kernel resolution criterion h < cp/(2κ_c) is a small reusable result. The near-wake comparison against Bachant & Wosnik is a legitimate positive: the solver transports the momentum deficit reasonably, and the grid study is honest about the coarse-mesh regime. The cost number (~37 min on 48 cores for a 30 s run) makes the tool attractive for parametric sweeps.\n\nThe soft spots: the central performance claim is carried by the external DMST model, not by the FR/CPR-ALM coupling. In Sec. III D 1, direct sampling of the resolved velocity produces a power curve 'nearly coincident with the zero-induction result'; then Sec. III D 2 sets the blade inflow to U∞(1−a_u)/U∞(1−2a_u) while the resolved flow is not correspondingly decelerated. The momentum source terms are evaluated with a velocity field inconsistent with the forces they impose. That is a double-counting/mis-correction risk, and the authors do not address it. Therefore the 6% agreement validates the engineering model (DMST + static polars + dynamic stall correction) more than the high-order solver. The wake validation at λ=1.9 only, with forces generated by the same DMST-modified inflow, does not close that gap. Also, the comparison to the Shamsoddin & Porté-Agel LES data needs a clear statement that the rotor geometry, Reynolds number, and airfoil polars match; otherwise 'within 6%' is not a well-defined metric. Missing: one grid-resolved self-consistent run at an operating point, uncertainty/error bars on the wake comparison, and code or data release.\n\nOverall: the paper is worth reading and refereeing. The method combination is new, the kernel criterion is useful, and the authors are transparent about the induction feedback issue—more than many papers would be. But the abstract and conclusion overstate what the framework itself predicts. A serious referee should ask for the CP claim to be reframed as a DMST-corrected ALM result, and ideally for one high-resolution self-consistent case.","headline":"Worth a serious look for the wake validation and the kernel criterion, but treat the 6% CP claim as a DMST result until the solver is shown to produce its own induction.","tokens_in":16247,"tokens_out":3824,"would_cite":true,"duration_ms":33592,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A high-order actuator-line framework predicts VAWT power within 6% of high-fidelity data.","keywords":["vertical-axis wind turbine","actuator line model","flux reconstruction","power coefficient","dynamic stall","wake prediction","Gaussian force projection","DMST induction"],"falsifier":"Run the same FR/CPR-ALM case without DMST on a mesh fine enough for the kernel to be chord-controlled (h < cp/(2κc) ≈ 0.049 m for this geometry), sample the resolved velocity at the actuator point, and compute CP(λ). If the resulting curve deviates substantially from the DMST-corrected coarse curve, then the 6% agreement is not produced by the coupled solver. A complementary check: on the fine mesh, measure the induced velocity at the blade and compare it to the DMST values uu and ud; a large mismatch would indicate the correction is misrepresenting induction.","tokens_in":15258,"feed_emoji":"🌬️","tokens_out":6058,"duration_ms":43857,"temperature":0.7,"pith_summary":"This paper is trying to establish that a high-order flux reconstruction solver, coupled with a rotating actuator-line model, can simulate vertical-axis wind turbine aerodynamics on fixed Cartesian grids without resolving blade geometry. The central quantitative claim is that the predicted power-coefficient curve matches three-dimensional LES-ALM reference data to within 6% around the optimal tip-speed ratio. To achieve that on affordable coarse meshes, the authors add an explicit Double Multiple Streamtube induction correction, because the Gaussian-smeared blade forces on their production mesh produce negligible resolved-flow induction feedback. The framework also reproduces regime-dependent dynamic stall, lift hysteresis, and characteristic near- and far-wake structures. If correct, this offers a computationally cheap, geometry-free route for parametric studies and wind-farm-scale analysis.","feed_headline":"Vertical-axis turbine model matches power curve within 6%","feed_subtitle":"High-order actuator-line solver reproduces dynamic stall and wakes on plain grids, enabling fast parametric studies.","key_machinery":"The load-bearing mechanism is the actuator-line source term: each blade is represented as a rotating point force, projected onto the fixed Cartesian grid through an isotropic Gaussian kernel of width ε = max(2h/p, c/κc), where h is mesh size, p the polynomial degree, c the chord, and κc=4.3. This kernel converts blade-element lift and drag into volumetric momentum and energy sources added to the compressible Navier-Stokes equations solved by the FR/CPR discretization (a high-order 'flux reconstruction/correction procedure via reconstruction' scheme). The angle of attack comes from the local velocity triangle in a four-quadrant formulation, with a modified Boeing-Vertol model adding rate-depe","core_discovery":"On the paper's own terms, the discovery is that a two-dimensional FR/CPR-ALM formulation with a single actuator point per blade, an isotropic Gaussian kernel for force projection, and a Boeing-Vertol dynamic-stall correction captures the essential physics of a straight-bladed vertical-axis turbine. The predicted power coefficient rises from 0.027 at λ=1.5 to a peak of 0.443 at λ=4.0, within 5.7% of the 3D LES-ALM peak, then declines to 0.229 at λ=7.0, and the agreement with the reference curve is within 6% over the operating window bracketing the optimum. The authors derive a mesh-resolution criterion h < cp/(2κc) for the Gaussian kernel and show that on the coarse production mesh the kernel","pith_inferences":["Editorial inference: the same coarse-mesh-plus-external-DMST recipe could in principle be carried over to horizontal-axis rotors or actuator-disk farm layouts, but the paper only demonstrates a two-dimensional VAWT, so that transfer is untested.","Editorial inference: because the DMST correction effectively prescribes the rotor's induction, the coupled solver's main role is transporting the wake; a natural test is to run a fully resolved, fine-mesh case with self-consistent induction feedback and compare the two power curves to quantify how much physics the correction is replacing.","Editorial inference: the 2D single-actuator-point-per-blade representation is spanwise-infinite; extending to 3D with multiple spanwise actuator elements would be needed to resolve tip vortices and spanwise load variation, which the paper notes as straightforward but does not simulate.","Editorial inference: the large low-λ overprediction (C_P=0.075 at λ=2.0 vs -0.05 reference) suggests the framework's quantitative utility is concentrated near the optimum, not in deep-stall performance, which matters for design loads."],"forward_implications":["The power-coefficient curve CP(λ) matches 3D LES-ALM reference data within 6% around the optimal tip-speed ratios, with the peak within 5.7%.","The mean near-wake velocity profile agrees with experimental and LES-ALM measurements at x/D=1, reproducing the deficit magnitude and lateral asymmetry.","The Boeing-Vertol dynamic-stall correction is regime-dependent: +23% CP at λ=1.5, -10.7% at λ=3.5, +6.9% at λ=5.5.","A single 30-second simulation at λ=3.5 takes about 37 minutes on 48 CPU cores, making parametric tip-speed-ratio sweeps practical.","The derived kernel-resolution criterion h < cp/(2κc) tells when the Gaussian width is chord-controlled rather than mesh-controlled, providing a rule for other meshes and airfoils."],"fun_headline_variants":["High-order actuator-line model nails VAWT power curve","Geometry-free solver predicts VAWT power within 6%","Actuator-line framework hits 6% power accuracy for VAWTs","New solver captures VAWT dynamic stall on plain grids","High-order ALM framework matches VAWT power to 6%"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The central performance claim rests on the assumption that on the coarse production mesh the Gaussian-smeared actuator force produces negligible resolved-flow induction feedback, so that an externally imposed DMST correction recovers the physical inflow; if the smearing over-softens the force or the correction double-counts induction, the reported power curve is essentially a DMST/BEMT result rather than a product of the coupled FR/CPR-ALM solver.","fun_headline_variants_meta":{"raw":{"variants":["High-order actuator-line model nails VAWT power curve","Geometry-free solver predicts VAWT power within 6%","Actuator-line framework hits 6% power accuracy for VAWTs","New solver captures VAWT dynamic stall on plain grids","High-order ALM framework matches VAWT power to 6%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1258,"prompt_tokens":856,"completion_tokens":402,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":317}},"tokens_in":600,"tokens_out":402,"duration_ms":4147,"temperature":1.0,"reasoning_tokens":317,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:04:00.481519+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same FR/CPR-ALM case without DMST on a mesh fine enough for the kernel to be chord-controlled (h < cp/(2κc) ≈ 0.049 m for this geometry), sample the resolved velocity at the actuator point, and compute CP(λ). If the resulting curve deviates substantially from the DMST-corrected coarse curve, then the 6% agreement is not produced by the coupled solver. A complementary check: on the fine mesh, measure the induced velocity at the blade and compare it to the DMST values uu and ud; a large mismatch would indicate the correction is misrepresenting induction.","supporting_citations":[],"review_version":1}