Pith. sign in

REVIEW 3 major objections 8 minor 20 references

Where you place passive flow-control devices around a spinning cylinder is set mainly by spin rate, not Reynolds number, across seven decades of Re.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-30 10:58 UTC pith:XC4HNIVY

load-bearing objection Useful spin-dominated sensitivity atlas for rotating cylinders, carefully validated in-point, but the robustness claim rests on visual sign maps and is undercut by the paper’s own failed cross-Re transfer. the 3 major comments →

arxiv 2607.23831 v1 pith:XC4HNIVY submitted 2026-07-26 physics.flu-dyn physics.comp-ph

Adjoint Sensitivity Maps for Passive Flow Control Around Rotating Circular Cylinders Across a Wide Operating Envelope

Niklas K\"uhl This is my paper
classification physics.flu-dyn physics.comp-ph PACS 47.85.L-47.11.-j47.32.Ef
keywords rotating circular cylinderpassive flow controladjoint methodstopology sensitivityFlettner rotorMagnus effectdrag-lift design maps
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Rotating cylinders (as in Flettner rotors) generate useful side force, but designers still guess where extra passive structures should sit. This paper builds a full atlas of adjoint topological sensitivities for drag, lift, and torque from Re = 10 to 10^7 and spinning ratios from 0 to 2π. The maps show that the large-scale pattern of favourable and unfavourable regions is controlled chiefly by the spinning ratio; Reynolds number only weakly rearranges them over much of the envelope. Combined drag–lift sign maps then mark zones where extracting momentum improves both objectives at once. Validated with Darcy sources and with actual add-on geometries, the atlas supplies practical placement rules and argues that passive devices can stay effective over moderate operating ranges rather than a single design point.

Core claim

Over Reynolds numbers from 10^1 to 10^7 and spinning ratios from 0 to 2π, the large-scale topology of the drag, lift, and torque sensitivity fields—and of the combined drag–lift sign maps—is governed primarily by spinning ratio, while Reynolds-number influence remains comparatively weak; therefore robust passive-control regions can exist over moderate operating ranges.

What carries the argument

Topology-based continuous adjoint sensitivity under a Darcy porous-medium model: the local field s = −v̂·v ranks where momentum extraction helps or hurts each objective; the combined sign(s_D)+sign(s_L) map is the design object that flags simultaneous drag reduction and lift gain.

Load-bearing premise

That two-dimensional frozen-turbulence adjoint maps built from a porous-medium model remain trustworthy first-order guides for real solid passive structures, especially in the turbulent regime where the rotor actually operates.

What would settle it

At a Flettner-relevant point (Re ~ 10^6, λ ~ π), build the add-on from the local combined-sign map, then measure forces while Re and λ vary moderately: if drag reduction and lift-to-drag gain vanish or reverse while the atlas still paints the occupied region favourable, the claim that the maps stay robust fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Passive devices should be sited on the decelerated side of the rotor within the green combined-sign regions of the atlas for typical ship-rotor Re and λ.
  • A structure designed at one spin rate can be reused across a band of nearby Re if the favourable patch stays aligned; it should not be copied across large jumps in λ without re-mapping.
  • Sign-based multi-objective maps can be extended to several operating points at once to hunt devices that stay beneficial over a prescribed envelope.
  • The same sensitivity fields can seed a later shape-optimization stage (topology first, shape second).
  • Reducing rotor drag via passive add-ons can widen the useful wind-speed range even when absolute side force is not increased.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Three-dimensional end plates and finite span will almost certainly warp the near-tip favourable patches; a 3D atlas is the natural next filter before hardware.
  • Because torque sensitivity is systematically weaker than lift or drag, passive devices aimed only at power draw may need larger volume or stronger porosity than lift–drag devices.
  • If the frozen-turbulence assumption under-predicts sensitivity magnitude in the drag-crisis band, designers may still trust the sign maps while needing larger safety margins on expected gain.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 8 minor

Summary. The manuscript presents a continuous-adjoint topology-sensitivity study of a rotating circular cylinder over Re_D = 10^1–10^7 and spinning ratios λ = 0–2π. Using a Darcy porous-medium formulation, local sensitivity fields s = −v̂·v are computed for drag, lift, and torque objectives from independent primal and adjoint solves (laminar below Re=10^5, SST-RANS with frozen turbulence above, steady adjoint on time-averaged fields). Validation proceeds in three layers at Re=10, λ=π: (i) signed Darcy source/sink forward runs recover the predicted diagonal response signs (Tab. 1); (ii) sink-only perturbations preserve the directions (Tab. 2); (iii) a body-fitted structure extracted from the combined drag–lift sign field reduces cylinder drag by ~66% (Tab. 3). A sensitivity atlas over the full envelope is presented as sign maps (Figs. 9–12) with integral sensitivity trends (Fig. 7) and six-level Eça–Hoekstra uncertainty quantification. At Re=10^6, λ=π, transferring the low-Re structure fails badly (Tab. 4) while an operating-point-specific structure succeeds on drag (Tab. 5). The authors conclude that sensitivity topology is governed primarily by λ with comparatively weak Re influence, and that robust passive-control regions may exist over moderate operating ranges.

Significance. If the robustness characterization is made quantitative, this is a useful reference: to my knowledge it is the first adjoint topology-sensitivity atlas for rotating cylinders spanning seven decades in Re and the full practical λ range, directly relevant to Flettner-rotor integration. Strengths worth naming: a large, systematically organized computational campaign (7 Re × 5 λ × 6 grid levels, plus three adjoints per point); a formal six-level Eça–Hoekstra uncertainty assessment at every operating point; and three complementary, genuinely falsifiable validation layers (signed Darcy, sink-only Darcy, and body-fitted structures), including an honestly reported negative control (the failed Re-transfer, Tab. 4). The combined drag–lift sign maps are an intuitive, transferable design artifact. The principal limitation is that the most novel claim — weak Reynolds-number dependence supporting cross-operating-point robustness — currently rests on qualitative sign-map inspection and is partly contradicted by the paper's own transfer test.

major comments (3)
  1. [§4, Figs. 9–12, Tab. 4] Abstract, §4, Figs. 9–12 vs. Tab. 4: the headline claim — Re influence 'remains comparatively weak over large parts of the envelope' so that 'robust solutions may exist over moderate operating ranges' — is supported only by visual inspection of binary sign maps, and the manuscript's single functional cross-Re test contradicts it at fixed λ: transferring the Re=10 structure to Re=10^6 at λ=π gives ΔcL/cL = −48.6% and Δ(cL/cD) = −95.7% (Tab. 4). Sign maps are magnitude-blind, so near-zero sensitivity regions can flip sign without appearing 'topologically stable'. The claim needs a quantitative similarity measure (e.g., a sign-agreement fraction or correlation of the sensitivity fields between adjacent Re rows at fixed λ, restricted to the regions where structures are actually placed), and the abstract/§5 wording should be reconciled with the failed transfer, which currently reads as an exc
  2. [§4.1, Tabs. 3 & 5] §3.2 (Tab. 3) and §4.1 (Tab. 5): the combined drag–lift sign maps (Fig. 12) are the paper's emphasized product, but the structural validations support the drag leg far more strongly than the lift leg. At Re=10^6 the cylinder-only lift change is +0.26% (Tab. 5) — plausibly below the Eça–Hoekstra discretization uncertainty, which §3.3 reports as growing beyond Re=10^5 — while combined-balance torque rises +15.8%. At Re=10 the combined balance loses half the lift (−50.5%, Tab. 3). Please report the estimated uncertainty on each validated coefficient change, state explicitly which changes exceed it, and temper the lift-guidance claims at high Re accordingly.
  3. [§2, §3.3] §2 (time-averaged adjoint, frozen turbulence) and §3.3: two fidelity questions bound the atlas's physical reach and deserve explicit treatment. (i) For the unsteady laminar rows (Re=10^2–10^4), sensitivities are computed from a steady adjoint on time-averaged fields; for shedding or near-shedding states the mean-flow sensitivity need not represent the limit-cycle response — a short justification or a spot check against an unsteady adjoint at one shedding point would resolve this. (ii) At Re≥10^5 the frozen-turbulence 2D-RANS forward runs already deviate noticeably from experiment (Fig. 6), so the Tab. 5 validation establishes adjoint–primal consistency within the model, not physical predictivity; please state this distinction and add a quantitative cL(λ)/cD(λ) comparison against published high-Re rotating-cylinder data (e.g., measurements compiled in Seifert 2012).
minor comments (8)
  1. [Tabs. 3–5] Tabs. 3–5: relative changes are misleading where the reference coefficient is near zero — §3.3 notes cT → 0 in the turbulent regime and negative cD at high λ, so Tab. 4's ΔcD/cD = +1089% and Tab. 5's torque percentages are partly denominator artifacts. Please add absolute coefficient changes (or the reference values) to each validation table.
  2. [§6/§9] Sections 6 and 9 are duplicate Declarations of Competing Interest; the AI-assistance acknowledgment also duplicates the acknowledgments structure. Remove the duplicates.
  3. [Figs. 3, 8, 13] Normalization is inconsistent between figures: Fig. 3 uses s/(V√(1+λ)) while Fig. 13 uses s/(V(1+λ)). Please harmonize or explain; the same applies to the velocity normalization v/(V√(1+λ)) in Figs. 3 and 8.
  4. [§2] §2: unsteady quantities are averaged 'over 100 cylinder revolutions', which is undefined for λ = 0 (n = 0). State the averaging interval in convective time units for the stationary cases.
  5. [§3.3, Fig. 7] Fig. 7 integral sensitivities depend on the arbitrary 5D×5D integration window; a brief note on the sensitivity of the Fig. 7 trends (including the reported sign changes) to window size would strengthen this section.
  6. [Fig. 2] Fig. 2 caption: the level numbering is confusing (panels show levels 1–4; 'coarsest (level 5) and finest (level 0)' are omitted, yet the text says results use a grid finer than Fig. 2(a), i.e. finer than level 1). Please clarify the indexing convention.
  7. [§4, keywords] Terminology: 'sensitivity spectra' (§4) suggests a decomposition; these are spatial sensitivity fields/sign maps. Consider 'sensitivity maps' or 'atlas fields'. Typo: 'Keywords:Rotating' missing space.
  8. [§10] Data availability 'upon reasonable request' is weak given the 840-simulation uncertainty campaign; depositing at least the atlas sign/sensitivity fields and the validation-table coefficients as supplementary data would materially improve reproducibility.

Circularity Check

1 steps flagged

No meaningful circularity: sensitivities and the atlas are independent primal/adjoint outputs; validations are separate forward predictions, not fits renamed as results.

specific steps
  1. self citation load bearing [§2 Numerical Method, opening paragraph; also adjoint BC pointer]
    "The employed methodology follows the adjoint-based topological sensitivity framework introduced in the authors’ previous work (Kühl [2025]) and is therefore only summarized briefly in the following. [...] The corresponding adjoint equations and boundary conditions follow directly from the continuous adjoint formulation and are not repeated here for brevity. Further details can be found in the authors’ previous work. e.g., Kühl et al. [2019, 2021], Kühl [2025]."

    Method and adjoint details are deferred to overlapping-author preprints rather than fully restated. This is ordinary framework reuse, not a uniqueness import or a fit that forces the sensitivity atlas: the (Re,λ) fields, sign maps, and forward validations are new computations in this paper. Not load-bearing for the spin-vs-Re topology claim; listed only as minor self-citation.

full rationale

The load-bearing chain is standard and non-circular: incompressible (U)RANS/laminar primals → continuous adjoint solves for drag/lift/torque → topological sensitivity s=−v̂·v from the Darcy porous-medium model → sign maps and integral measures over the (Re,λ) grid. That formula is the usual first variation of the objective w.r.t. the local Darcy coefficient; it is not defined from the later force deltas. Source/sink and geometric validations are independent forward simulations whose ΔcD, ΔcL, ΔcT are genuine checks of predicted directions, not quantities that were fitted into the maps. Self-citations (Kühl 2025; Kühl et al.) only import the adjoint/ FreSCo+ framework already used elsewhere; they do not supply a uniqueness theorem or ansatz that forces the atlas topology or the ‘spin-dominated, weak-Re’ claim. That claim is an a-posteriori reading of newly computed fields (Figs. 8–12), not a parameter fit or a renaming of prior data. Hand-extracting a structure from a same-OP sign map and verifying improvement at that OP is a legitimate first-order adjoint check, not circular construction. Overstatement of cross-OP robustness relative to the failed Re-transfer test is a correctness/evidence issue, not circularity. Score 1 only for routine method self-citation that is not load-bearing for the central atlas result.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

The central atlas claim rests on standard incompressible NS/(U)RANS modeling, continuous adjoint topology sensitivity via a Darcy term, frozen turbulence, 2D reduction, and several hand-chosen numerical/design parameters (step size β, exclusion zones, sensitivity contour level for CAD extraction, 5D×5D integration box). No new physical entities are postulated; novelty is in the computed maps, not new laws.

free parameters (5)
  • Darcy step-size β = 1e-3
    Chosen as 10^{-3} (with stated dimensions) to convert sensitivity into α=βs for validation; magnitude of force changes depends on this hand-set scale.
  • Near-cylinder Darcy deactivation distance = D/10
    Sink-only validation turns off α within D/10 of the wall to avoid boundary-layer surgery; affects how ‘passive-device-like’ the test is.
  • Structural exclusion zone and contour level = 1.1D exclusion; isosurface 1.1
    Geometry built from combined-sign interface with 1.1D exclusion (low-Re case) or lift-sensitivity isosurface s/(V(1+λ))=1.1 (high-Re case); hand-selected design knobs, not derived optima.
  • Integral sensitivity window 5D×5D = 5D × 5D
    ˆc_D, ˆc_L, ˆc_T integrate local s over a fixed 5D×5D box; window size is conventional and affects reported integral levels.
  • Inlet turbulence intensity and eddy viscosity (RANS) = Tu=1%, ν_t/ν=1
    1% intensity and ν_t/ν=1 at velocity boundaries for Re≥10^5; standard but free modeling choices that can shift separation and sensitivities.
axioms (5)
  • domain assumption Incompressible Navier–Stokes with optional SST-2003 RANS closure adequately represent the mean flow for sensitivity purposes over Re_D=10–10^7.
    §2 switches laminar/URANS at 10^5 and uses time averages over 100 revolutions for unsteady cases.
  • domain assumption Frozen-turbulence continuous adjoint (turbulence quantities not differentiated) yields usable topological sensitivities.
    Explicitly adopted in §2 with citations to Soto/Othmer/Stück; known approximation in turbulent adjoint design.
  • domain assumption Topological sensitivity reduces to s=−v̂_i v_i for a Darcy momentum sink αv in the primal momentum equation.
    Eq. (5) and porous-medium topology framework from Borrvall–Petersson/Othmer line; sign convention stated in §2.
  • domain assumption Two-dimensional flow (symmetry in span) is sufficient to map design regions for the intended Flettner-relevant conclusions.
    Fig. 1 and BC description; outlook admits 3D finite-span rotors remain future work.
  • ad hoc to paper Positive local sensitivity under distributed Darcy insertion indicates favourable locations for finite passive geometric structures.
    Bridging assumption tested in §3.2 and §4.1 but not generally proved; add-on loads lie outside the original surface objective.

pith-pipeline@v1.2.0-grok45-kimik3 · 19341 in / 3757 out tokens · 79433 ms · 2026-07-30T10:58:57.513353+00:00 · methodology

0 comments
read the original abstract

Rotating circular cylinders are employed in a variety of engineering applications, one prominent example being Flettner rotors for wind-assisted ship propulsion. Besides optimizing the aerodynamic performance of the cylinder itself, passive flow-control devices placed in its vicinity offer additional potential for manipulating the resulting aerodynamic forces. The present work introduces a topology-based adjoint sensitivity analysis for rotating circular cylinders over a wide operating envelope covering Reynolds numbers from 1E+01 to 1E+07 and spinning ratios between 0 and 2 pi. Local sensitivity fields associated with drag, lift, and torque are derived using a porous-medium formulation and validated by dedicated forward simulations employing both distributed Darcy-type source terms and a sensitivity-informed passive flow-control structure. Particular emphasis is placed on the combined sign distribution of the drag and lift sensitivities, yielding intuitive design maps that directly identify regions where local momentum extraction simultaneously improves or deteriorates both objectives. A systematic investigation of the resulting sensitivity spectra reveals that the large-scale topology of the sensitivity fields is governed primarily by the spinning ratio, whereas the influence of the Reynolds number remains comparatively weak over large parts of the investigated operating envelope. The resulting sensitivity atlas provides practical design guidance for passive flow-control concepts and demonstrates that robust solutions may exist over moderate operating ranges.

Figures

Figures reproduced from arXiv: 2607.23831 by Niklas K\"uhl.

Figure 1
Figure 1. Figure 1: Two-dimensional rotating-cylinder configuration and definition of the coordinate system. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Near-wall discretization in the vicinity of the cylinder for four representative members of the six-level [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Representative normalized primal, adjoint, and non-negative sensitivity fields for the validation case [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sign of the drag sensitivity, sign of the lift sensitivity, and resulting combined sign field for Re [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Structural validation case at ReD = 10 and λ = π: (a) body-fitted near-wall mesh around the sensitivity-informed structure, (b) streamwise velocity component v1, and (c) transverse velocity component v2 obtained from the corresponding forward simulation. The resulting aerodynamic coefficients are subsequently compared to the corresponding reference solution. Since the additional structure contributes to th… view at source ↗
Figure 6
Figure 6. Figure 6: Integral aerodynamic quantities as functions of Reynolds number and spinning ratio: (a) drag coeffi [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Integral adjoint quantities associated with the drag, lift, and torque objectives as functions of Reynolds [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Magnitude of the primal velocity field for all investigated Reynolds numbers and spinning ratios. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sign of the drag sensitivity for all investigated Reynolds numbers and spinning ratios. Black regions [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Sign of the lift sensitivity for all investigated Reynolds numbers and spinning ratios. Black regions [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Sign of the torque sensitivity for all investigated Reynolds numbers and spinning ratios. Black [PITH_FULL_IMAGE:figures/full_fig_p018_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Combined drag–lift sign fields for all investigated Reynolds numbers and spinning ratios. Green [PITH_FULL_IMAGE:figures/full_fig_p019_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Generation of the operating-point-specific passive flow-control structure at Re [PITH_FULL_IMAGE:figures/full_fig_p020_13.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 14 canonical work pages

  1. [13]

    doi:10.1017/S0022112002002938. C. Othmer. A Continuous Adjoint Formulation for the Computation of Topological and Surface Sensi- tivities of Ducted Flows. International Journal for Numerical Methods in Fluids, 58(8):861–877,

    work page doi:10.1017/s0022112002002938

  2. [14]

    doi:10.1002/fld.1770. T. Rung, K. W¨ ockner, M. Manzke, J. Brunswig, C. Ulrich, and A. St¨ uck. Challenges and Perspectives for Maritime CFD Applications. Jahrbuch der Schiffbautechnischen Gesellschaft, 103:127–39,

    work page doi:10.1002/fld.1770

  3. [19]

    doi:10.1016/j.apenergy.2013.07.026. T. Tsutsui and T. Igarashi. Drag Reduction of a Circular Cylinder in an Air-Stream. Journal of Wind Engineering and Industrial Aerodynamics, 90(4-5):527–541,

    work page doi:10.1016/j.apenergy.2013.07.026 2013

  4. [1988]

    doi:10.1007/BF01061285. S. Kang, H. Choi, and S. Lee. Laminar Flow Past a Rotating Circular Cylinder. Physics of Fluids, 11(11): 3312–3321,

    doi:10.1007/bf01061285

  5. [1990]

    doi:10.1017/S0022112090000933. A. St¨ uck.Adjoint Navier-Stokes Methods for Hydrodynamic Shape Optimisation. PhD thesis, Hamburg Uni- versity of Technology,

    work page doi:10.1017/s0022112090000933

  6. [1999]

    doi:10.1063/1.870190. N. K¨ uhl. On the Continuous Adjoint of Prominent Explicit Local Eddy Viscosity-based Large Eddy Sim- ulation Approaches for Incompressible Flows. Flow, Turbulence and Combustion, 113(2):293–330,

    work page doi:10.1063/1.870190

  7. [2000]

    doi:10.1023/A:1011430410075. A. Jameson. Aerodynamic Design via Control Theory. Journal of Scientific Computing, 3(3):233–260,

    doi:10.1023/a:1011430410075

  8. [2002]

    doi:10.1016/S0167-6105(01)00199-4. M.M. Zdravkovich. Flow around Circular Cylinders: Volume 2: Applications, volume

    work page doi:10.1016/s0167-6105(01)00199-4

  9. [2003]

    doi:10.1002/fld.426. H. Choi, W.P. Jeon, and J. Kim. Control of Flow Over a Bluff Body. Annual Review of Fluid Mechanics, 40 (1):113–139,

    doi:10.1002/fld.426

  10. [2004]

    doi:10.1108/09615530410544292. D. Stojkovi´ c, M. Breuer, and F. Durst. Effect of High Rotation Rates on the Laminar Flow around a Circular Cylinder. Physics of Fluids, 14(9):3160–3178,

    work page doi:10.1108/09615530410544292

  11. [2005]

    doi:10.1007/s00158-004-0508-7. M.B. Giles and N.A. Pierce. An Introduction to the Adjoint Approach to Design. Flow, Turbulence and Combustion, 65(3):393–415,

    work page doi:10.1007/s00158-004-0508-7

  12. [2008]

    doi:10.1146/annurev.fluid.39.050905.110149. C.B. Dilgen, S.B. Dilgen, D.R. Fuhrman, O. Sigmund, and B.S. Lazarov. Topology Optimization of Turbulent Flows. Computer Methods in Applied Mechanics and Engineering, 331:363–393,

    arXiv

  13. [2012]

    doi:10.1016/j.paerosci.2012.07.001. O. Soto, R. L¨ ohner, and C. Yang. An Adjoint-Based Design Methodology for CFD Problems. International Journal of Numerical Methods for Heat & Fluid Flow, 14(6):734–759,

    work page doi:10.1016/j.paerosci.2012.07.001 2012

  14. [2013]

    doi:10.1016/j.jcp.2013.01.002. M. Traut, P. Gilbert, C. Walsh, A. Bows, A. Filippone, P. Stansby, and R. Wood. Propulsive Power Contri- bution of a Kite and a Flettner Rotor on Selected Shipping Routes. Applied Energy, 113:362–372,

    work page doi:10.1016/j.jcp.2013.01.002 2013

  15. [2014]

    doi:10.1016/j.jcp.2014.01.006. A. Gersborg-Hansen, O. Sigmund, and R. B. Haber. Topology Optimization of Channel Flow Problems. Structural and Multidisciplinary Optimization, 30(3):181–192,

    work page doi:10.1016/j.jcp.2014.01.006 2014

  16. [2018]

    doi:10.1016/j.cma.2017.11.029. L. E¸ ca and N. Hoekstra. A Procedure for the Estimation of the Numerical Uncertainty of CFD Cal- culations Based on Grid Refinement Studies. Journal of Computational Physics, 262:104–130,

    work page doi:10.1016/j.cma.2017.11.029 2017

  17. [2019]

    doi:10.2514/1.J058376. N. K¨ uhl, P. M. M¨ uller, and T. Rung. Adjoint Complement to the Universal Momentum Law of the Wall.Flow, Turbulence and Combustion,

    work page doi:10.2514/1.j058376

  18. [2020]

    doi:10.1080/17445302.2019.1612544. M. Manzke. Development of a Scalable Method for the Efficient Simulation of Flows using Dynamic Goal-Orientated Local Grid-Adaption. PhD thesis, Hamburg University of Technology,

    arXiv 2019

  19. [2021]

    doi:10.1007/s10494-021-00286-7. R. Lu and J.W. Ringsberg. Ship Energy Performance Study of Three Wind-Assisted Ship Propulsion Tech- nologies Including a Parametric Study of the Flettner Rotor Technology. Ships and offshore structures, 15 (3):249–258,

    work page doi:10.1007/s10494-021-00286-7

  20. [2024]

    doi:10.1007/s10494-024-00543-5. 13 N. K¨ uhl. Adjoint-Assisted Topology-Optimization-Inspired Analysis of Pseudo-Porous Flow Fields: Application to a Flettner Rotor. arXiv preprint arXiv:2505.04833,

    Pith/arXiv arXiv