{"id":"e7927dee-07fb-4db6-8512-55846d40514c","arxiv_id":"2505.04833","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An adjoint-derived sign field around a Flettner rotor predicted, in the four tested placements, whether a container stack improves or worsens lift and drag.","lead":"This paper maps, around a spinning Flettner rotor, the regions where adding ship structures would help or hurt its lift and drag, using a sensitivity field from adjoint CFD. The map is checked by placing container stacks in predicted good and bad spots and recomputing the forces.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing assumption is Eq. (20)'s infinitesimal-porosity derivative predicting the sign of a finite solid stack; only four hand-picked cases, with drag tested only in the beneficial sign, support it.","rationale":"The reader's weakest_assumption matches my concern. The derivation itself is standard, and the four reported sign agreements are genuine supporting evidence, especially the lift increase/decrease pair. However, the central advertised use is sign-based placement of a real solid structure, and the only link between the computed derivative and that use is the assumption that a first-order local response survives finite perturbation. This is not proven and is known to be fragile in separated high-Reynolds-number flows. The missing detrimental-drag case is a concrete hole: if no positive (drag-increasing) placement is tested, the sign map's discriminating power for drag is unverified. The paper's own conclusion flags the need for finite-difference validation and notes that only the sign was checked, which further supports treating this as a limitation rather than a contradiction. I agree with the reader's CONDITIONAL verdict; no new failure is demonstrated, so the verdict should remain CONDITIONAL. The proposed \\alpha-sweep and solid-stack comparison would either strengthen or refute the transfer assumption.","tokens_in":12106,"tokens_out":4471,"duration_ms":52214,"concrete_test":"Repeat the validation with an \\alpha-sweep at the same [x1/H, x2/H] locations used in Sec. 3.3 and at one location where the drag sensitivity is negative: introduce a localized porous region, set \\alpha to 0.1, 1, 10, 100 kg/(s m^3), converge each case, and compare \\Delta J to sign(s). Also replace the porous region with the actual solid stack and check that the sign of \\Delta J agrees with the \\alpha\\to\\infty limit. If the sign flips between small and large \\alpha, Eq. (20) is not a reliable sign map.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (20) gives s = \\hat{v}_i v_i = dJ/d\\alpha at \\alpha=0 for a virtual Darcy momentum sink, but Sec. 3.3 validates this by inserting an impermeable six-TEU container stack, i.e., a finite, non-porous body. The sign of an infinitesimal local derivative is not guaranteed to match the sign of a finite geometry change, especially at Re_D=2e6 where separation and blockage can produce nonlocal, nonlinear responses. The frozen-turbulence approximation in Eq. (9) further weakens the derivative, as the paper itself acknowledges. The four validation runs cover the lift map in both signs and the drag map only in the beneficial (drag-reducing) direction; no predicted-detrimental drag placement, no neutral/counterfactual placement, and no uncertainty quantification are reported, so the claim that the sign map identifies beneficial and detrimental locations rests on a small, favorable sample. The conclusion explicitly calls for finite-difference validation and notes that only the sign was checked. Without a test of sign transfer beyond the selected locations, the central design-guidance claim is conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a continuous-adjoint, topology-optimization-inspired sensitivity diagnostic for the flow around a Flettner rotor. A virtual Darcy-type porosity term is included in the momentum equations only for the purpose of differentiation; the actual primal simulation has zero porosity. The resulting topological sensitivity field is given by the inner product of the adjoint and primal velocity in Eq. (20). The sign of this field is interpreted as indicating where adding solid material (e.g., a container stack) would improve or deteriorate a chosen aerodynamic objective: drag, lift, or a combined drift objective. The method is applied at Re_D = 2e6 and spinning ratio k = 3. Validation is performed by inserting a six-TEU20 container stack at four hand-picked locations: two for the lift/drift map (one positive-sign, one negative-sign region) and two for the drag map (both in positive-sign, drag-beneficial regions). The measured force changes match the predicted sign in all four cases. The paper frames the method as a non-intrusive post-processing design-support tool and explicitly notes in the conclusion that only the sign was checked and future finite-difference validation is needed.","tokens_in":1782,"tokens_out":6600,"duration_ms":86333,"significance":"If the central claim holds, the method would be a practically useful early-design screening tool: it would let a designer read, directly from a scalar field, where placing structures helps or hurts a rotor's lift or drag without running an optimization loop. The continuous-adjoint derivation is standard and the sensitivity expression is not fitted to the validation data, which is a genuine strength. The paper is also honest about its limitations: it acknowledges the frozen-turbulence assumption as delicate and states in the conclusion that only the sign of the sensitivity was checked. The main gap is that the load-bearing extension from an infinitesimal virtual-porosity derivative to a finite, impermeable container stack is supported by only a small, favorable sample of validation cases, and the drag direction is tested only in the beneficial sign. No error bars, grid-dependence study, or averaging-convergence evidence is reported. These are fixable shortcomings rather than fundamental flaws, but they need to be addressed before the claimed predictive capability is established.","major_comments":[{"comment":"The validation never tests the sign transfer for a predicted-detrimental drag location. Both drag arrangements (c) and (d) are placed in positively predicted (beneficial) regions, and both reduce drag. The central claim that the sign map identifies 'beneficial and detrimental locations' is therefore demonstrated only for the lift/drift functional, not for drag. Please add at least one drag case placed in a blue (negative-sensitivity) region, and ideally a neutral or counterfactual placement, to show that the sign map is specific rather than only identifying some region that improves drag.","section":"Sec. 3.3 and Eq. (20)"},{"comment":"Eq. (20) gives the derivative of the cost functional with respect to an infinitesimal porosity parameter at alpha = 0, but the validation inserts a finite, impermeable six-TEU container stack. The paper does not justify why the sign of a local, infinitesimal derivative should persist for a large solid body at Re_D = 2e6, where blockage and separation produce strongly nonlinear, nonlocal responses. A direct supporting test would be a finite-difference check of dJ/dalpha for a small virtual-porosity bump at several candidate locations, or a series of stack sizes/porosities showing that the sign remains stable. As written, the extrapolation from the derivative to the finite solid body is an unsupported load-bearing step.","section":"Sec. 2, Eq. (20) versus Sec. 3.3"},{"comment":"The paper acknowledges that the frozen-turbulence assumption is 'a delicate assumption' at the Reynolds numbers considered, but then states that 'later applications will show that satisfactory sensitivity derivatives can still be predicted.' This is a promise, not evidence. Since the entire diagnostic reduces to the sign of the sensitivity field, the paper should provide at least a quantitative check of the omitted terms, e.g., by comparing with a formulation that includes turbulent-viscosity variations, or by demonstrating sign stability under a perturbed turbulence model. A sign inversion under this assumption would directly undermine the design-guidance claim.","section":"Sec. 2, Eq. (9)"},{"comment":"No grid-dependence study, statistical-convergence study, or uncertainty quantification is reported for either the primal or adjoint solutions. The validation figure (Fig. 7) reports relative changes without error bars, so it is unclear whether the force differences are converged in averaging time or mesh resolution. Since the central claim is about sign accuracy, at least one mesh-refinement check and an averaging-convergence estimate for the reported drag and lift changes are needed.","section":"Sec. 3.1 and Sec. 3.3"}],"minor_comments":[{"comment":"In the Introduction, 'an thus, the CFD solver' should read 'and thus the CFD solver.'","section":"Sec. 1"},{"comment":"'is an delicate assumption' should read 'is a delicate assumption.'","section":"Sec. 2, after Eq. (9)"},{"comment":"The sentence on the differences between the lift and drift functionals contains a typo: 'the lift ((b),(e),(h)) and the lift ((c),(f),(i))' should read 'the lift ((b),(e),(h)) and the drift ((c),(f),(i)).'","section":"Sec. 3.2"},{"comment":"The text and caption for Fig. 2(a) should be consistent: the text says slices at x3/H = [0, 1/3, 2/3, 1], while the caption uses the notation [0, 1, 2, 3]/3; please unify the notation.","section":"Sec. 3.1"},{"comment":"The abbreviation TEU20 is used as 'six TEU20 containers' but the expansion in the text is 'six Twenty-foot Equivalent container Units (TEU20)'; please define the abbreviation once and use it consistently.","section":"Sec. 3.3"},{"comment":"The force-change plot would benefit from explicit statements about convergence of the averaging window and, if available, estimated statistical error bars. At minimum, please state whether the four simulations were run to the same convergence criterion as the reference case.","section":"Fig. 7"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of physics.flu-dyn and the core adjoint derivation is standard and sound. The main concern for me as reviewer is not the mathematics but the weight given to a four-case, sign-only validation in support of a design-guidance claim. The author's own conclusion already concedes that only the sign was checked and calls for finite-difference validation; I think the manuscript needs at least one detrimental-drag case and one direct check of the infinitesimal-to-finite transfer before the claim is publishable in its current strong form. The reliance on the author's prior validation of the FreSCo+ solver is acceptable as evidence dependency, but an independent check on the new quantity (topological sensitivity sign) would strengthen the paper considerably."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Straight to it: the paper is a competent application of a known adjoint porosity-sensitivity formula, and the genuinely new bit is reading the sign of the sensitivity field as a design-support map instead of running a topology-optimization loop. That is a useful idea for early-stage ship design, and the Flettner rotor example is well chosen. The validation, though, is thinner than the abstract implies.\n\nWhat the paper does well: the continuous adjoint derivation in Section 2 is standard but clean, and the sensitivity expression (the inner product of primal and adjoint velocity) is correct for the virtual Darcy term. The forward simulations are described in enough detail (3.8M cells, 200 flow-through times, resolved TKE ratio above 80%) to look credible, and the author is candid about the frozen-turbulence assumption being delicate at this Reynolds number. The four container placements all behave as the sign maps predict, which is a real check.\n\nWhere it gets shaky: the sensitivity is a derivative with respect to an infinitesimal porosity change at alpha = 0, but the validation inserts a solid, finite six-TEU stack. Nothing in the paper shows that the sign of that local derivative survives the jump to a finite impermeable body, and at this Reynolds number separation and blockage can make the response nonlocal and nonlinear. The author acknowledges this in the conclusion, but the abstract and validation section call it a validation without that caveat. More importantly, the drag map is tested only in the beneficial direction: both drag cases reduce drag. A map that claims to identify detrimental locations should be tested on at least one predicted-detrimental drag placement. There are also no error bars, no grid-dependence study, and no finite-difference or linearized benchmark at any location, so the evidence base is four hand-picked points with no negative control.\n\nNone of this breaks the central idea. The method is sound, the paper is honest about its limits, and the sign-map interpretation is a legitimate contribution. It just is not proven beyond a small favorable sample. A referee should ask for one detrimental drag case, a finite-difference or linearized sensitivity check at a single location, and either a grid study or an uncertainty estimate, or for the claims to be softened to match the evidence.\n\nThis is a paper for ship aerodynamics and adjoint-CFD people, and a serious referee would get value from it. I would send it to peer review and engage with it. My recommendation: major revision, with the additional validation as the condition for acceptance.","headline":"Sound derivations, useful design-support idea, but the validation is too thin to support the strong claims.","tokens_in":12811,"tokens_out":4226,"would_cite":false,"duration_ms":39127,"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":"The sign of one scalar field, $s = \\hat{v}_i v_i$, computed from primal and adjoint flow fields, tells where structures around a Flettner rotor will help or hurt the chosen force objective.","keywords":["Computational Fluid Dynamics","Continuous Adjoint Sensitivity Analysis","Topology Optimization","Flettner Rotor","Maritime Aerodynamics","virtual porosity","deck cargo placement","drag and lift sensitivity"],"falsifier":"Run the same rotor case and compare the sign map with a genuine porosity perturbation: take the converged flow, add a small but finite $\\alpha$ in a spherical probe region at several points, and measure the objective change; if the sign of measured change disagrees with the sign of $s$ at that point, or if the disagreement appears only for larger container blocks, then the sign map's predictive range is bounded. A wind-tunnel or higher-fidelity test placing a container stack exactly at a predicted-beneficial location and measuring drag and lift would settle the practical claim.","tokens_in":11844,"feed_emoji":"🚢","tokens_out":6485,"duration_ms":53941,"temperature":0.7,"pith_summary":"The paper claims that a single scalar field, obtained by multiplying a converged flow solution with a continuous-adjoint companion solution, maps out where adding a solid structure near a Flettner rotor improves or worsens rotor drag and lift. The field is the topological sensitivity with respect to a virtual porosity penalty, $s = \\hat{v}_i v_i$, and only its sign is used. This matters because a designer can screen many deck arrangements by reading colored slices of the field, without running an optimization loop or modifying a CFD solver. The paper supports the claim by placing a six-container stack at positions the sign map marks beneficial or detrimental and confirming that lift and drag change in the predicted directions. The demonstration is for one full-scale rotor at $Re_D = 2\\times10^6$ and spin ratio $k=3$.","feed_headline":"Sign map predicts where deck cargo helps or hurts a Flettner rotor","feed_subtitle":"Validated by moving a six-container stack: lift and drag change in directions the sensitivity field predicts.","key_machinery":"The load-bearing object is the topological sensitivity $s = \\hat{v}_i v_i$ arising from a continuous-adjoint treatment of a virtual porosity source term $\\alpha(v_i - v_i^{\\mathrm{tar}})$ with $\\alpha\\to0$. The porosity field is never introduced into the primal computation or updated; instead, the adjoint solution is post-multiplied by the averaged primal velocity to produce a scalar map. The map is then read by sign, following the steepest-descent convention (the plotted quantity is $-s$). The derivation uses the frozen-turbulence assumption and neglects density variations, and the paper relies on the same adjoint solver infrastructure used for shape optimization, so the method is a post-processing step rather than a new solver.","core_discovery":"The central claim is that the sign of the topological sensitivity field $s = \\hat{v}_i v_i$ identifies where introducing material improves or deteriorates a selected aerodynamic objective around a Flettner rotor. Starting from a time-averaged, scale-resolving flow field, the paper derives adjoint equations for an incompressible flow with a virtual Darcy-like penalty term $\\alpha(v_i - v_i^{\\mathrm{tar}})$, with $\\alpha$ identically zero in the actual simulation. The sensitivity of the force objective to a porosity perturbation reduces to the inner product of the primal and adjoint velocity vectors; negating it gives the direction of steepest descent. The author interprets regions of negative sensitivity as beneficial and positive as detrimental for the chosen objective, for drag, lift, and a combined drift objective. Validation runs place six TEU20 containers in predicted positive and negative regions and report relative changes in lift and drag consistent with the sign map, e.g., a starboard stack increases lift while a port stack reduces it, and both tested drag-beneficial placements reduce drag by roughly half.","pith_inferences":["If the sign map is read as a derivative at zero porosity, the strongest untested extension is quantitative: finite solid bodies at the same location but different size or shape may not follow the infinitesimal prediction; a size-threshold study would show where the extrapolation breaks.","The same virtual-porosity sensitivity could be used to guide placement of other deck structures, superstructure elements, or windshields, and also underwater appendages, by defining the objective on a different surface and re-running the adjoint.","A natural test is to compare the sign map against a finite-difference porosity perturbation of small but finite strength; agreement for small strengths and divergence for large ones would calibrate how far the linearization can be trusted.","The author's suggested extension to true local topology optimization would reveal whether the sign map's zero contours coincide with the optimal material distribution once an optimization loop is run."],"forward_implications":["A designer can screen an arbitrary number of deck arrangements around a Flettner rotor by reading the sign of $s$, re-running the primal/adjoint pair only when the rotor geometry or operating condition changes.","For any force objective, the same machinery works by changing the adjoint boundary condition $r_i$ on the rotor, e.g., $\\delta_{i1}$ for drag, $\\delta_{i2}$ for lift, or a normalized combination for drift.","The validated outcome implies that drag reductions of around 50 percent are available by placing cargo in the predicted beneficial zone, while a misplaced stack can increase drag significantly.","Because no porosity design variable enters the flow solver, the method can be attached to existing adjoint-enabled CFD setups without altering the turbulence model or discretization.","The sign map is qualitative and does not claim optimality: it identifies improved, not optimal, placements, as the paper states."],"supporting_citations":[{"why":"Supplies the continuous-adjoint formulation for topological and surface sensitivities of ducted flows that the paper's virtual-porosity sensitivity derivation builds on.","marker":"Othmer (2008)"},{"why":"Source of the sensitivity expression $s=\\hat{v}_i v_i$ for porosity-type adjoint topology sensitivity.","marker":"Gerdes (2018)"},{"why":"Establishes the porous-media topology-optimization basis for fluid flow that motivates the virtual porosity penalty.","marker":"Borrvall and Petersson (2003)"},{"why":"Comparative review of topology optimization approaches used to position the method relative to classical porosity optimization.","marker":"Sigmund and Maute (2013)"},{"why":"Reference for topology optimization theory and the material-distribution concept underlying the approach.","marker":"Bendsoe and Sigmund (2013)"},{"why":"Provides the continuous adjoint methodology for turbulent flows applied to shape and topology optimization.","marker":"Papoutsis-Kiachagias and Giannakoglou (2016)"},{"why":"Defines the Improved Delayed DES turbulence model used in the primal flow solution.","marker":"Gritskevich et al. (2012)"},{"why":"One of the references flagging the frozen-turbulence assumption the paper adopts and acknowledges as delicate at high Reynolds numbers.","marker":"Löhner et al. (2003)"},{"why":"Supplies the adjoint-based design background for the continuous adjoint derivation.","marker":"Giles and Pierce (2000)"},{"why":"Used to demonstrate that the hybrid RANS/LES method captures maritime aerodynamic flows in earlier studies.","marker":"Angerbauer and Rung (2020)"}],"fun_headline_variants":["Topology-inspired sensitivity fields predict cargo effects on Flettner rotors","Sensitivity field identifies beneficial and detrimental spots for Flettner rotor cargo","Where to place deck cargo? Flettner rotor sensitivity map answers","Adjoint-based sensitivity field tells if deck cargo helps or hurts Flettner rotor","Container stacks validate sensitivity-map predictions for Flettner rotor cargo"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the sign of the sensitivity obtained from an infinitesimal virtual-porosity perturbation, computed with frozen turbulence, still predicts the effect of placing a large solid, non-porous container stack at the same location; the paper tests this with four configurations but does not derive the transfer.","fun_headline_variants_meta":{"raw":{"variants":["Topology-inspired sensitivity fields predict cargo effects on Flettner rotors","Sensitivity field identifies beneficial and detrimental spots for Flettner rotor cargo","Where to place deck cargo? Flettner rotor sensitivity map answers","Adjoint-based sensitivity field tells if deck cargo helps or hurts Flettner rotor","Container stacks validate sensitivity-map predictions for Flettner rotor cargo"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001484,"raw_usage":{"total_tokens":5986,"prompt_tokens":998,"completion_tokens":4988,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":4893}},"tokens_in":614,"tokens_out":4988,"duration_ms":32552,"temperature":1.0,"reasoning_tokens":4893,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:20:29.650488+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same rotor case and compare the sign map with a genuine porosity perturbation: take the converged flow, add a small but finite $\\alpha$ in a spherical probe region at several points, and measure the objective change; if the sign of measured change disagrees with the sign of $s$ at that point, or if the disagreement appears only for larger container blocks, then the sign map's predictive range is bounded. A wind-tunnel or higher-fidelity test placing a container stack exactly at a predicted-beneficial location and measuring drag and lift would settle the practical claim.","supporting_citations":[],"review_version":1}