{"id":"8bd62bf0-7331-49cc-b84a-d471cb10e103","arxiv_id":"2607.27580","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A preprocessing tool refines hybrid unstructured meshes around immersed bodies so immersed-boundary CFD matches body-fitted and wind-tunnel results for airfoil flows.","lead":"This paper presents a mesh-refinement preprocessing tool that lets fluid-dynamics simulations keep a body-fitted grid for fixed parts and embed a changing part—like a flap—as an immersed body with local refinement. It is a practical advance for industrial CFD because it reduces costly remeshing when only one component moves.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central pipeline depends on CODA accepting hanging-node interfaces, which is never stated or tested.","rationale":"I read the paper as a mostly sound engineering contribution: the refinement algorithms are described in detail, and the reported cylinder, NACA0012, and MDA30P30N results are individually consistent with body-fitted and experimental data. The single most load-bearing weakness is the unstated reliance on CODA's support for nonconforming faces. The paper's own successful runs provide circumstantial evidence, but they do not document the solver-side handling of hanging nodes, and a silent misaccounting at these interfaces would corrupt the central accuracy claims. This concern is concrete, external to the presented algorithms, and testable, so the conditional verdict is appropriate. I do not see a more damaging internal inconsistency: the `nFacets > 1` pseudocode discrepancy in Algorithm 4 is likely a typo given the surrounding prose says 'greater than zero,' though it should be corrected in a revision.","tokens_in":28130,"tokens_out":10193,"duration_ms":95208,"concrete_test":"Run CODA on a minimal two-cell mesh with one QUAD4QUAD nonconforming interface (one hexahedron split into 8 children next to an unrefined hexahedron) for a scalar advection or manufactured Navier–Stokes problem. Compare the L2 error and the discrete conservation residual at the hanging-node face against a conforming 1:1 mesh. If CODA rejects the mesh, produces a flux imbalance across the interface, or gives different convergence behavior, the claimed HDF5 compatibility and the validity of the Section 5 results are not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The preprocessing tool's refinement deliberately creates hanging nodes and nonconforming faces between elements (Section 4.1, Figure 4), and Algorithm 1 exports 'the resulting mesh in HDF5 format, compatible with the CFD solver.' Nowhere in Section 4 or 5 is it stated or demonstrated that CODA's finite-volume or DG discretizations support nonconforming interfaces. If CODA requires conforming faces, or silently computes fluxes incorrectly across hanging-node faces, the entire pipeline fails: the exported mesh cannot be consumed as claimed. The reported successful CODA simulations are indirect evidence that some support may exist, but the manuscript does not describe any conforming step or any solver-side handling of hanging nodes, leaving the central claim of compatibility dependent on an unverified external capability. This is a load-bearing assumption because every validation in Section 5 relies on the refined nonconforming meshes being run in CODA exactly as exported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents a preprocessing tool that refines hybrid unstructured (tetrahedral, hexahedral, prismatic, pyramidal) body-fitted background meshes locally around immersed geometries for use with the CODA CFD solver and an immersed-boundary volume-penalization method. The tool constructs nonconforming hanging-node interfaces, balances refinement to a 2:1 level ratio, and exports an HDF5 mesh that is claimed to be directly usable by the solver. The authors validate the approach on 2D-extruded cases: laminar flow past a cylinder (Re=40), laminar flow past an NACA0012 airfoil (Re=5000), and RANS flow past an MDA30P30N multi-element airfoil (Re=9e6), reporting pressure coefficients and force coefficients in reasonable agreement with body-fitted and experimental references. A further example optimizes flap position in a two-element airfoil with GEMSEO. The central claim is that the refined meshes allow the IBM-equipped CODA solver to simulate flows around immersed components accurately and efficiently by avoiding full remeshing.","tokens_in":28352,"tokens_out":5682,"duration_ms":54893,"significance":"If substantiated, the tool would enable a practically useful workflow: a body-fitted, wall-resolved mesh for a fixed geometry, with changing components (flaps, control surfaces) treated as immersed and locally refined, avoiding costly re-meshing. The paper gives a detailed algorithmic description and validates against independent published data without curve-fitting; the reported coefficient matches (e.g., cylinder Cd≈1.53–1.55 vs. 1.56–1.57 in the literature; fine-mesh MDA30P30N Cl=2.776 vs. 2.876 experiment) are encouraging. However, the strongest claims extend to true three-dimensional hybrid meshes, while every validation case is extruded 2D. Moreover, the pipeline's compatibility with CODA hinges on an unstated assumption about nonconforming-interface support, and key numerical parameters (notably η) are not reported. These gaps weaken the paper in its current form but are addressable.","major_comments":[{"comment":"The refinement deliberately creates hanging nodes and nonconforming faces (Fig. 4, Section 4.1), and Algorithm 1 exports 'the resulting mesh in HDF5 format, compatible with the CFD solver.' Nowhere is it stated or demonstrated that CODA's finite-volume or discontinuous-Galerkin discretizations accept nonconforming interfaces, nor is any conforming step described before export. The reported successful runs are indirect evidence, but the central compatibility claim ('compatible with the CFD solver') is load-bearing and rests on an unverified external capability. Please state explicitly whether CODA supports hanging-node/nonconforming faces, or add a conforming/interface-treatment step, with a reference or a test.","section":"§4.1, §4.2, Algorithm 1"},{"comment":"The element-flagging logic is internally inconsistent. The text preceding Algorithm 4 says an element is flagged when the number of intersecting geometry facets is 'greater than zero,' but Algorithm 4 uses 'if Elem.nFacets > 1.' Algorithm 2 appears to query only a single nearest facet ('Facet←NearestNeighbors(BBox,FacetsBary)') and increments Elem.nFacets by 1, whereas the text says the number of intersecting facets is stored. If the threshold is truly >1, elements crossed by exactly one geometry facet would never be refined, creating gaps in the refined band; if the pseudocode is literal, many overlapping facets are missed. Please clarify the actual implementation and correct the pseudocode, since this is the core of the refinement algorithm.","section":"§4.2, Algorithms 2 and 4"},{"comment":"The penalization parameter η is central to the volume-penalization source term (Eq. 44) and is listed as a free parameter, but its numerical value is not reported for any test case (cylinder, NACA0012, MDA30P30N, or optimization). Without η (and the refinement box dimensions), the simulations cannot be reproduced. Please provide the η values and box-region sizes used in each case.","section":"§5 and Eq. (44)"},{"comment":"The abstract claims the meshes enable simulations 'in an accurate and efficient manner,' but no efficiency data are reported: no wall-clock times, CPU-hour counts, mesh-generation times, or comparison with the cost of a full body-fitted re-mesh. The fine MDA30P30N IBM mesh has 4,585,591 elements versus 736,472 for the body-fitted mesh, yet the paper does not discuss whether the IBM+refinement workflow is actually cheaper. Please add quantitative efficiency metrics or temper the efficiency claim.","section":"Abstract and §5.3"},{"comment":"All validation cases are extruded 2D: the computational domain is a disk extruded one element in the y-direction with symmetry boundary conditions on the front and back faces. The tool's stated capability for genuine three-dimensional hybrid meshes (tetrahedra, pyramids, etc.) is therefore not tested by any three-dimensional case. The conclusions generalize to 3D industrial geometries and mention extensions to 'a billion elements,' but the present results do not substantiate that. Please add at least one truly 3D demonstration or explicitly restrict the scope to 2D-extruded configurations.","section":"§5.1–§5.3 and Conclusions"}],"minor_comments":[{"comment":"Typo: 'vecinity' should be 'vicinity'.","section":"§5.1"},{"comment":"The turbulence model is repeatedly called 'Spallart–Allmaras'; the correct spelling is 'Spalart–Allmaras'.","section":"§2.2"},{"comment":"The legend labels contain garbled characters (e.g., '∝O⌋⊣↕2-BFM'); these should be rendered as 'FV-II-BFM', 'DG-III-IBM', etc.","section":"Figures 8, 10, 12"},{"comment":"The optimization result is presented as 'consistent with a similar optimization workflow with only body-fitted meshes' in the Conclusions, but no comparison data or reference to the companion paper is shown. Please provide evidence or qualify the statement.","section":"§5.4"},{"comment":"The loop upper bound 'L' is not defined. It should be the current maximum refinement level; please define it explicitly.","section":"Algorithm 6"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the journal's scope and the algorithmic contribution is potentially useful, but the central compatibility claim with CODA's handling of nonconforming faces is unverified, and the flagging logic in Algorithms 2/4 contains ambiguities that go beyond presentation. The lack of reported η values and efficiency data also needs attention. If the authors can clarify or demonstrate the CODA nonconforming-interface support, correct the algorithm pseudocode, and add the missing numerical parameters, the paper would be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper describes a preprocessing tool that refines hybrid unstructured meshes around immersed bodies for CFD with volume-penalization IBM. The genuinely new piece is the integration: a body-fitted background mesh with local refinement around an immersed geometry, with hanging-node balancing across four element types, followed by export to the CODA solver. Prior adaptive IBM work uses Cartesian meshes; this hybrid-mesh variant is not in the cited literature. That is a legitimate contribution. The validations—cylinder at Re=40, NACA0012 at Re=5000, and the MDA30P30N high-lift case—show force coefficients within a few percent of body-fitted and experimental values. The algorithms are described in enough detail to be reimplemented.\n\nThe main soft spot is the one the stress-test flags: the paper never states that CODA can consume meshes with hanging nodes. Section 4.1 explicitly says the refinement creates nonconforming faces, and Section 4.2 says the exported HDF5 is 'compatible with the CFD solver.' No conforming step is described, and the paper never mentions whether CODA's finite-volume or DG discretizations support nonconforming interfaces. The reported successful simulations are indirect evidence that CODA handles them, but the claim of compatibility is load-bearing and needs to be stated and demonstrated explicitly.\n\nTwo smaller issues: all test cases are extruded 2D, so the tetrahedron and pyramid splitting rules are never exercised in an actual numerical simulation. And the optimization example claims consistency with a body-fitted workflow, but no comparison data are shown. These are fixable. The paper does not fit any constants to the outputs; the free parameters are the penalization parameter and refinement settings, which is standard for this family of methods.\n\nOverall, this is a coherent engineering paper with a real, addressable gap. It deserves a serious referee. The authors should verify and state CODA's nonconforming support, add a genuinely 3D test or at least one that uses tetrahedra and pyramids, and either provide the missing body-fitted optimization comparison or drop that claim. I would not cite it in my own work until those are resolved, but I would bring it to a CFD reading group.","headline":"Useful integration paper with a real gap: CODA's handling of the exported hanging-node meshes is never stated.","tokens_in":28841,"tokens_out":3319,"would_cite":false,"duration_ms":30269,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A preprocessing tool locally refines hybrid unstructured meshes around immersed bodies so that immersed-boundary flow predictions match body-fitted computations.","keywords":["mesh adaptation","hybrid meshes","immersed boundaries","volume penalization","hanging nodes","computational fluid dynamics","lift optimization","unstructured mesh"],"falsifier":"Take a problem with a known smooth solution and run the same pipeline on a hybrid mesh with a deliberately introduced hanging-node face; if the solver does not converge to the analytical solution at the expected rate, or if a refined mesh with hanging nodes fails to reproduce the body-fitted force coefficient on the same geometry, the central claim fails.","tokens_in":28029,"feed_emoji":"🛠️","tokens_out":4833,"duration_ms":43374,"temperature":0.7,"pith_summary":"The paper describes a preprocessing tool that locally refines hybrid unstructured meshes—made of tetrahedra, hexahedra, prisms, and pyramids—around a geometry treated as immersed rather than body-fitted. The central idea is to split the full geometry into a fixed part that keeps a body-fitted mesh and a changing part, say a flap, that is immersed and locally refined. The authors establish that this hybrid body-fitted/immersed approach can reproduce force coefficients from full body-fitted simulations and experiments, as shown for flow past a cylinder, a NACA0012 airfoil, a multi-element airfoil, and a flap-position optimization. A sympathetic reader would care because it offers a path to reuse a background mesh while optimizing or varying only a component, avoiding full remeshing.","feed_headline":"Hybrid-mesh refinement closes the immersed-boundary accuracy gap","feed_subtitle":"Local refinement around an immersed flap reproduces body-fitted lift and drag on cylinder, airfoil, and multi-element wing.","key_machinery":"The load-bearing mechanism is a set of element-splitting conventions plus master–slave interface arrays (TRI4TRI for triangular faces, QUAD4QUAD for quadrilateral faces). Each tetrahedron splits into 8 tetrahedra, each hexahedron into 8 hexahedra, each prism into 8 prisms, and each pyramid into 6 pyramids plus 4 tetrahedra; hanging nodes created by neighboring elements at different refinement levels are encoded through the master/slave arrays. Hash tables and kd-trees make the geometry-overlap detection and face-matching efficient. These interfaces drive the 2:1 balancing step that keeps the mesh locally refined without producing more than one refinement level difference between neighbors. T","core_discovery":"On its own terms, the paper claims that the new mesh-refinement tool makes immersed-boundary volume penalization accurate and efficient on hybrid unstructured meshes. The tool takes a conforming hybrid background mesh, defines a refinement box around where the immersed body will sit, refines elements that overlap the body, and balances refinement so neighboring elements differ by at most a 2:1 level ratio. The resulting locally refined meshes, with hanging nodes and nonconforming faces, are exported for the CFD solver. Validation cases match body-fitted and experimental pressure and force coefficients when the mesh is sufficiently refined: the drag for cylinder and NACA0012 is within a few p","pith_inferences":["Beyond the paper: if the CFD solver truly accepts nonconforming faces, this pipeline could be applied to any hybrid mesh with local refinement, not just two-dimensional extrusions; the same splitting conventions should extend to 3-D refinement of complex aircraft components.","The 2:1 balancing rule is a design choice; one could test whether relaxing it to 4:1 or allowing graded refinement reduces element counts without hurting force predictions, though accuracy would need checking near the immersed surface.","The optimization example suggests a practical workflow: pre-refine a region large enough to contain all possible flap positions, then move the immersed geometry inside it without remeshing—an implicit cost-saving claim the paper does not quantify directly.","Because the immersed geometry is refined by surface-overlap, the tool's accuracy will depend on the facet size of the STL representation relative to the mesh; a sensitivity study on facet density would be a natural next test."],"forward_implications":["A single body-fitted background mesh can serve a family of configurations: only the changing part, such as a flap, is placed as an immersed body and locally refined.","Local refinement around immersed surfaces can replace uniform Cartesian refinement, keeping element counts and runtime practical for industrial cases.","The reported force coefficients (drag for cylinder and NACA0012, lift for the multi-element airfoil) approach full body-fitted and experimental values as the refinement level increases.","Optimization loops can reuse a pre-refined region that covers all possible positions of the moving component, avoiding a full remesh at each design evaluation.","Both finite volume and discontinuous Galerkin spatial discretizations can be coupled with the volume-penalization immersed boundary method on the refined hybrid meshes."],"fun_headline_variants":["Hybrid mesh refinement sharpens immersed-boundary CFD","Adaptive refinement for accurate immersed-boundary flow","Refining hybrid meshes boosts immersed-boundary accuracy","New tool refines hybrid meshes for immersed-boundary CFD"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The pipeline depends on the CFD solver correctly handling hanging-node, nonconforming faces between neighboring elements after refinement; the paper demonstrates refined meshes but does not show the solver's treatment of nonconforming interfaces.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid mesh refinement sharpens immersed-boundary CFD","Adaptive refinement for accurate immersed-boundary flow","Refining hybrid meshes boosts immersed-boundary accuracy","New tool refines hybrid meshes for immersed-boundary CFD"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001171,"raw_usage":{"total_tokens":4743,"prompt_tokens":868,"completion_tokens":3875,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":612,"completion_tokens_details":{"reasoning_tokens":3820}},"tokens_in":612,"tokens_out":3875,"duration_ms":22565,"temperature":1.0,"reasoning_tokens":3820,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T05:15:20.654130+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a problem with a known smooth solution and run the same pipeline on a hybrid mesh with a deliberately introduced hanging-node face; if the solver does not converge to the analytical solution at the expected rate, or if a refined mesh with hanging nodes fails to reproduce the body-fitted force coefficient on the same geometry, the central claim fails.","supporting_citations":[],"review_version":1}