Pith. sign in

REVIEW 3 major objections 6 minor 41 references

The paper introduces 3-refinement templates that turn adaptive 27-tree grids into conforming all-hex meshes under a moderately-balanced condition, cutting element counts by several-fold to hundreds-fold versus earlier 3-refinement methods.

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 · deepseek-v4-flash

2026-08-03 15:56 UTC pith:F7MEKOVE

load-bearing objection A genuinely useful 3-refinement template system with a real but unverified completeness claim; the 2-refinement comparison is tuned. the 3 major comments →

arxiv 2512.14862 v5 pith:F7MEKOVE submitted 2025-12-16 cs.CG

Element-Saving Hexahedral 3-Refinement Templates

Hua Tong , Yongjie Jessica Zhang This is my paper
classification cs.CG MSC 65M5068U05
keywords hexahedral mesh3-refinement27-treeconforming meshtemplate-based refinementmoderately-balanced conditionadaptive gridhanging node removal
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.

This paper aims to solve the hanging-node problem in adaptive hexahedral meshing: when a grid is refined locally, the transition cells have hanging nodes that must be removed to produce a conforming mesh. The authors propose a 3-refinement template system that works under a new moderately-balanced condition, which permits adjacent cells sharing a face or edge to differ by at most one tree level. The key move is recursive local refinement: complex vertex and edge configurations are decomposed into simpler sub-cells that match a small set of fundamental templates, each replaced by a predefined hex subdivision. On a benchmark of 202 models, the two variants keep element growth within a factor of about five, cutting element counts by several-fold to over a hundred-fold relative to earlier 3-refinement methods, while matching or slightly beating a leading 2-refinement method on element count and Hausdorff ratio. If the template set is complete and conforming, this gives a practical, template-based route to all-hex meshes with coplanar faces and stable Jacobian values.

Core claim

The paper's central claim is that any 27-tree grid satisfying the moderately-balanced condition can be converted into a conforming all-hex mesh via cell-by-cell template substitution. The vertex-based variant enumerates all 256 binary vertex configurations (22 up to symmetry) and covers them with ten fundamental and twelve composite templates; the edge-based variant enumerates all 4,096 binary edge configurations (144 up to symmetry) and covers them with 31 specialized templates plus a universal [X][X][X] template. Composite cases are handled by splitting the cell into sub-cells whose patterns match fundamental templates, so conformity follows from the fundamental templates' face and edge sc

What carries the argument

The central machinery is the template lookup: a mapping from a cell's vertex or edge subdivision pattern (black dots on vertices, or trisected edges) to a predefined hexahedral decomposition. The recursion principle is the load-bearing idea: any pattern not directly in the fundamental set is first split into sub-cells whose patterns are simpler, and those sub-cells are replaced by fundamental templates. The moderately-balanced condition (level difference ≤1 for cells sharing a face or edge) guarantees the interface patterns — 1→3 edge transitions and 1×1→3×3 face transitions — stay within the small set that the templates are designed to handle.

Load-bearing premise

The claim stands or falls on the completeness and conformity of the template sets: every one of the 22 vertex configurations and 4,096 edge configurations must be covered by a template whose hex subdivision has no hanging nodes, no non-convex cells, and consistent face/edge interface patterns — and the paper asserts this is 'readily verified' rather than providing a formal proof or machine-checked verification for all cases.

What would settle it

An automated checker could enumerate all 4,096 edge configurations and all 22 vertex configurations, run the published template lookup, and test each generated hex mesh for hanging nodes and positive Jacobian; a single non-conforming output would refute the completeness claim. A second test: apply the open-source implementation to a deliberately adversarial grid whose refinement pattern alternates 1 and 0 edges in a checkerboard arrangement and inspect the resulting mesh for face/edge mismatches.

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

If this is right

  • Any 27-tree satisfying the moderately-balanced condition can be converted into a conforming all-hex mesh without extra global refinement, making the condition a practical replacement for stronger 3-refinement conditions.
  • Element growth ratios drop to roughly 1–5× the initial grid, versus 10–100× for earlier 3-refinement methods, so 3-refinement becomes viable for simulation workflows where element count dominates cost.
  • Primal 3-refinement preserves information stored on grid cells (e.g., signed distance fields or triangle indices), which dual 2-refinement methods lose, and produces planar quad faces with min SJ around 0.22–0.23.
  • The edge-based variant with the greedy edge-marking optimizer yields marginally fewer elements than the vertex-based variant in 92% of tested models, at about five times the runtime.

Where Pith is reading between the lines

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

  • The moderately-balanced condition is a middle ground; if even weaker conditions (e.g., weakly-balanced with 1→9 edge transitions) could be tamed by new templates, element counts might approach 2-refinement levels, but the paper regards 1→9 transitions as too irregular.
  • The greedy optimizer is order-dependent, so a global optimization (e.g., integer linear programming over edge-marking decisions) could produce order-independent, closer-to-optimal element counts; the paper flags this as suboptimal future work.
  • Because the method is purely topological and template-based, it could be adapted to other graded tree structures or to non-cubic cell shapes, provided the interface patterns remain within a manageable template family.
  • The Hausdorff-ratio comparison against 2-refinement is based on five models with a specific tuning protocol; a broader comparison across the full 202-model benchmark would test whether the apparent surface-fitting advantage holds generally.

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 / 6 minor

Summary. This paper presents two template-based methods for converting adaptive 3-refinement (27-tree) grids into conforming all-hex meshes under a new 'moderately-balanced' condition, intermediate between weakly- and strongly-balanced conditions. The first method uses vertex-based templates for all 22 canonical vertex configurations; the second uses edge-based templates claiming to cover all 4,096 edge configurations via 31 specialized cases plus a universal [X][X][X] template, with recursive local refinement reducing complex patterns to fundamental templates. The authors report experiments on 202 models from [9], showing large element-count reductions relative to prior 3-refinement methods, and a preliminary comparison with the 2-refinement method [22] on five models, where they claim slightly fewer hex elements and lower Hausdorff ratios. An open-source C++ implementation is provided.

Significance. If the template sets are complete and conforming as asserted, the paper makes a significant advance: it relaxes 3-refinement from the strongly-balanced-plus-pairing conditions to a moderately-balanced condition, yielding element counts several times lower than previous 3-refinement methods while preserving planar faces and convex hexes. The recursive-decomposition idea is elegant and practically motivated. The paper's strengths include a full implementation attached, a large-scale empirical study on 202 models, and a clear comparison with three prior 3-refinement baselines. The main unresolved risk is that completeness/conformity of the template sets is not demonstrated mechanically or formally, and the central claim hinges on this. The comparison with 2-refinement is also not yet on equal footing.

major comments (3)
  1. [Section 3] The paper claims that the vertex-based templates cover all 22 canonical configurations (256 total), but §3 does not provide a complete enumeration or formal proof. It states that the ten fundamental templates are 'readily verified' and that the twelve composite templates are 'subdivided into five or six sub-cells' whose patterns match fundamental templates or 11100000, but no table or listing shows which of the 22 configurations map to which template, nor is the recursive decomposition for each composite pattern spelled out beyond the single example 11111110. Because the universal conversion theorem depends on every possible input cell being covered, this is a load-bearing gap. The attached code can settle the issue; please include an exhaustive machine check (e.g., a script that instantiates all 22 classes, applies the template lookup, and asserts no hanging nodes, convexity, and matchi
  2. [Section 4] The edge-based method's coverage of all 4,096 configurations is asserted but not verified. The universal [X][X][X] template (Figure 4(B)) is justified by saying a cell 'can always be partitioned via local refinement into seven sub-cells' with certain encodings whose 'last eight bits are invariably zero', but no proof is given that this partition is always possible under the moderately-balanced condition or that the set of seven encodings covers every configuration not matched by the first 31 templates. Likewise, the statement that applying the six face templates 'produces six conforming templates' is an inspection claim. Please supply an exhaustive verification over all 4,096 edge configurations (or 144 symmetry classes) using the attached code and report the verification result in the paper or supplement. Without this, the claim that the edge-based template set handles every possible ed
  3. [Section 5 / Table 2] The comparison with the 2-refinement state-of-the-art [22] is not a fair test of the claimed advantage. The text states 'tuning parameters such that the proposed method produces a slightly lower element count', so the lower count is by construction. The Hausdorff ratio result is also not uniformly in favor: for the elk model, the proposed methods report HR 2.0 and 1.9 versus 1.6 for [22] (Table 2). The comparison is limited to five models and includes additional projection/optimization steps for both pipelines, so it does not isolate the template contribution. The abstract's claim of 'lower Hausdorff ratio using slightly fewer elements' should be restricted to the tested cases and parameter choices, or the experiment should be redone with a pre-specified, untuned threshold.
minor comments (6)
  1. [Section 4] 'After exhaustive enumeration' — the reader cannot verify this without a supplementary listing of the 144 symmetry classes or a script; please include an enumeration table or code output in an appendix.
  2. [Section 5] For the 202-model comparison, please state the policy for choosing the shape diameter function threshold (fixed, per-model, or tuned to match [22]) and whether the same threshold is used for all compared 3-refinement methods.
  3. [Section 6] The greedy edge-based algorithm's order-dependence is acknowledged, but no deterministic tie-breaking rule is specified. Please document a canonical edge traversal order so the published element counts are reproducible.
  4. [Section 2 / Figure 1] The caption defines 'moderately-balanced' with a figure, but the notation for marking schemes (black dots, 0/1) is first used in Figure 2; a first-time reader may benefit from a one-sentence explanation in the text before Figure 2.
  5. [Section 3] Typo: 'This method require' in the conclusion should be 'This method requires'.
  6. [Abstract] The abbreviation 'min SJ' is used without definition; please define it as minimum scaled Jacobian at first occurrence.

Circularity Check

1 steps flagged

Element-saving comparison to 2-refinement is partly constructed by tuning the shape-diameter threshold; the template derivation itself is self-contained.

specific steps
  1. fitted input called prediction [Section 5, Table 2 caption and the 2-refinement comparison paragraph]
    "The hex element counts for the proposed methods are adjusted via tuning the shape diameter function threshold to be slightly smaller than [22]. ... For each test model, hex meshes are generated using both methods, tuning parameters such that the proposed method produces a slightly lower element count."

    The headline comparative claim that the proposed method uses 'slightly fewer elements' than the state-of-the-art 2-refinement method is not a free prediction: the shape-diameter threshold is explicitly tuned so that the proposed method ends up with fewer elements. The reported element-count advantage is therefore an input to the comparison, not an output of the template method per se. The Hausdorff-ratio comparison is then made under this tuned condition, so any conclusion combining 'lower Hausdorff ratio' with 'slightly fewer elements' is partly engineered by parameter choice. This does not make the template construction itself circular, but it makes the central evaluation claim statistically forced.

full rationale

The core technical content of the paper is a template construction: a moderately-balanced 27-tree is converted into a conforming hex mesh by vertex-based and edge-based template substitution, with recursive local refinement reducing complex configurations to fundamental templates. This derivation is self-contained and does not fit parameters to the target element counts or Hausdorff ratios. The fundamental templates are adapted from prior work [14], not from the paper's own conclusions, and the moderately-balanced condition is a new input condition rather than a consequence of the results. No load-bearing uniqueness theorem is imported from the authors' prior work, and the self-citations ([32], [33], [34]) are used only for post-processing and optimization steps, not to establish the central conversion claim. The completeness/conformity of the template sets is asserted as 'readily verified' rather than machine-checked, but that is a correctness-risk issue, not circularity. The one genuine circularity is in the evaluation design: the comparison against the 2-refinement baseline [22] explicitly tunes the shape-diameter threshold so that the proposed method produces a slightly lower element count, and then reports that it 'achieve[s] a lower Hausdorff ratio using slightly fewer elements.' The 'slightly fewer' part is an input chosen to make the comparison favorable, so the headline comparative claim is partially constructed by parameter selection rather than discovered. This affects the evaluation claim, not the internal logic of the templates, warranting a score of 3 rather than a higher circularity score.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 0 invented entities

The key assumptions are the completeness and face-compatibility of the template sets, which are asserted with 'readily verified' and visual reasoning rather than formal proof. The moderately-balanced condition is a design choice; its sufficiency is demonstrated empirically on the benchmark set but not proven in general.

free parameters (1)
  • shape diameter function threshold = tuned per model
    In Section 5 (Figure 8/Table 2), the SDF threshold is tuned so the proposed method yields slightly fewer hex elements than [22] before comparing Hausdorff ratios. This is a per-case calibration that directly influences the comparative claim.
axioms (4)
  • ad hoc to paper Every one of the 22 vertex configuration classes (after symmetry) has a valid conforming hex template.
    Section 3 asserts ten fundamental and twelve composite templates cover all 22 cases, with composite cases recursively decomposed. The verification is visual/asserted, not machine-checked.
  • ad hoc to paper The 31 specialized templates plus the universal [X][X][X] scheme cover all 4,096 edge configurations.
    Section 4 argues via Pólya counting and a universal fallback, but does not provide an exhaustive proof for all 144 symmetry-distinct patterns.
  • domain assumption The moderately-balanced condition is sufficient for the template set to produce a conforming mesh.
    The paper excludes weakly-balanced conditions due to 1→9 edge transitions being 'too irregular' without formal proof. Sufficiency is demonstrated empirically, not proven in general.
  • ad hoc to paper Face transition schemes in Figures 3(C) and 4(C) are pairwise compatible on shared faces.
    The paper states these are 'readily verified' but does not provide a formal proof of mutual compatibility across all combinations.

pith-pipeline@v1.3.0-alltime-deepseek · 13699 in / 12443 out tokens · 119097 ms · 2026-08-03T15:56:19.210542+00:00 · methodology

0 comments
read the original abstract

Conforming hex meshes are widely regarded as an effective computational domain for simulation because of their nice numerical properties, yet automatically decomposing a general 3D volume into a conforming hex mesh remains a formidable challenge. Among existing approaches, methods that construct an adaptive Cartesian grid and subsequently convert it into a conforming mesh stand out for their robustness. However, topological conversion schemes require strict compatibility conditions that inevitably increase element count. State-of-the-art 2-refinement octree methods employ weakly-balanced and generalized pairing conditions to yield low element counts, but suffer from critical limitations: primal cell information is lost after dualization, and resulting dual cells often exhibit non-planar quad faces. Alternatively, 3-refinement 27-tree methods directly generate conforming hex meshes through template-based replacement, producing higher-quality elements with planar faces, but previous techniques impose far stricter conditions, severely over-refining grids by factors of ten to one hundred. This article introduces a novel 3-refinement approach using a moderately-balanced condition, slightly stronger than weakly-balanced but substantially more relaxed than prior 3-refinement requirements. The key insight is that recursively applying local refinements can isolate and reduce complex configurations to simpler cases covered by a fundamental template set. Two open-sourced variants are provided: one optimized for speed, and another trading some computational cost for marginally reduced element counts. Compared to previous 3-refinement methods, they significantly reduce final hex element counts while preserving min SJ values and guaranteeing convex polyhedral cells; relative to 2-refinement state-of-the-art, they also achieve a lower Hausdorff ratio using slightly fewer elements.

Figures

Figures reproduced from arXiv: 2512.14862 by Hua Tong, Yongjie Jessica Zhang.

Figure 1
Figure 1. Figure 1: Grid cells refined under unbalanced, weakly-balanced, moderately-balanced, and strongly-balanced [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Transition templates for a grid cell (left) and a face (right) from prior work: (A) [25]; (B) [14]; and (C) [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Transition templates for (A, B) a grid cell and (C) a face in vertex-based templates. Vertices adjacent [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (A) Transition templates for a grid cell in edge-based templates. Edges adjacent to deeper-level cells are [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Performance comparison on 202 models in the dataset of [9]. All panels except the upper right show [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Four representative cases (selected from 202 models in [9]) where previous 3-refinement methods [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Adaptive refinement to conformal hex meshes. (A) Initial adaptively refined grid. (B-D) Previous [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Evaluation on five randomly selected models (rocker, elk, gargoyle, thai [PITH_FULL_IMAGE:figures/full_fig_p013_8.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

41 extracted references

  1. [1]

    Benzley, E

    S. Benzley, E. Perry, K. Merkley, B. Clark, and G. Sjaardema , A comparison of all hexagonal and all tetrahedral finite element meshes for elastic and elasto-plastic analysis , Proceedings, 4th International Meshing Roundtable, 17 (1995)

    1995

  2. [2]

    T. Blacker , Meeting the challenge for automated conformal hexahedral meshing , in 9th International Meshing Roundtable, Albuquerque, NM, USA, 2000, Citeseer, Sandia National Laboratories, pp. 11--20

    2000

  3. [3]

    A. O. Cifuentes and A. Kalbag , A performance study of tetrahedral and hexahedral elements in 3- D finite element structural analysis , Finite Elements in Analysis and Design, 12 (1992), pp. 313--318

    1992

  4. [4]

    https://coreform.com/products/coreform-cubit/government/, 2025

    Coreform , Coreform cubit . https://coreform.com/products/coreform-cubit/government/, 2025

    2025

  5. [5]

    N. G. De Bruijn , Polya's theory of counting , in Applied Combinatorial Mathematics, John Wiley & Sons, 1964, pp. 144--184

    1964

  6. [6]

    M. S. Ebeida, A. Patney, J. D. Owens, and E. Mestreau , Isotropic conforming refinement of quadrilateral and hexahedral meshes using two-refinement templates , International Journal for Numerical Methods in Engineering, 88 (2011), pp. 974--985

    2011

  7. [7]

    A. H. Elsheikh and M. Elsheikh , A consistent octree hanging node elimination algorithm for hexahedral mesh generation , Advances in Engineering Software, 75 (2014), pp. 86--100

    2014

  8. [8]

    Fabri and S

    A. Fabri and S. Pion , C GAL : The computational geometry algorithms library , in Proceedings of the 17th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2009, pp. 538--539

    2009

  9. [9]

    X. Gao, H. Shen, and D. Panozzo , Feature preserving octree-based hexahedral meshing , Computer Graphics Forum, 38 (2019), pp. 135--149

    2019

  10. [10]

    Garanzha, I

    V. Garanzha, I. Kaporin, L. Kudryavtseva, F. Protais, N. Ray, and D. Sokolov , Foldover-free maps in 50 lines of code , ACM Transactions on Graphics (TOG), 40 (2021), pp. 1--16

    2021

  11. [11]

    K. Hu, J. Qian, and Y. Zhang , Adaptive all-hexahedral mesh generation based on a hybrid octree and bubble packing. 22nd I nternational M eshing R oundtable , 2013

    2013

  12. [12]

    Hu and Y

    K. Hu and Y. J. Zhang , Centroidal V oronoi T essellation based polycube construction for adaptive all-hexahedral mesh generation , Computer Methods in Applied Mechanics and Engineering, 305 (2016), pp. 405--421

    2016

  13. [13]

    https://intrinsim.com/MeshGems.html, 2025

    IntrinSIM , Mesh G ems: The standard for meshing technology . https://intrinsim.com/MeshGems.html, 2025

    2025

  14. [14]

    Y. Ito, A. M. Shih, and B. K. Soni , Octree-based reasonable-quality hexahedral mesh generation using a new set of refinement templates , International Journal for Numerical Methods in Engineering, 77 (2009), pp. 1809--1833

    2009

  15. [15]

    Livesu, L

    M. Livesu, L. Pitzalis, and G. Cherchi , Optimal dual schemes for adaptive grid based hexmeshing , ACM Transactions on Graphics (TOG), 41 (2021), pp. 1--14

    2021

  16. [16]

    Maggioli, S

    F. Maggioli, S. Melzi, and M. Livesu , Volumetric functional maps , 2025

    2025

  17. [17]

    L. Mar \'e chal , Advances in octree-based all-hexahedral mesh generation: Handling sharp features , in Proceedings of the 18th International Meshing Roundtable, Springer, Berlin, Heidelberg, 2009, pp. 65--84

    2009

  18. [18]

    S. A. Mitchell , A characterization of the quadrilateral meshes of a surface which admit a compatible hexahedral mesh of the enclosed volume , in Annual Symposium on Theoretical Aspects of Computer Science, 1996, pp. 465--476

    1996

  19. [19]

    S. J. Owen , A survey of unstructured mesh generation technology , 7th International Meshing Roundtable, 3 (2000)

    2000

  20. [20]

    S. J. Owen, R. M. Shih, and C. D. Ernst , A template-based approach for parallel hexahedral two-refinement , Computer-Aided Design, 85 (2017), pp. 34--52

    2017

  21. [21]

    Pietroni, M

    N. Pietroni, M. Campen, A. Sheffer, G. Cherchi, D. Bommes, X. Gao, R. Scateni, F. Ledoux, J. Remacle, and M. Livesu , Hex-mesh generation and processing: A survey , ACM Transactions on Graphics (TOG), 42 (2022), pp. 1--44

    2022

  22. [22]

    Pitzalis, M

    L. Pitzalis, M. Livesu, G. Cherchi, E. Gobbetti, and R. Scateni , Generalized adaptive refinement for grid-based hexahedral meshing , ACM Transactions on Graphics (TOG), 40 (2021), pp. 1--13

    2021

  23. [23]

    Qian and Y

    J. Qian and Y. Zhang , Automatic unstructured all-hexahedral mesh generation from B - R eps for non-manifold CAD assemblies , Engineering with Computers, 28 (2012), pp. 345--359

    2012

  24. [24]

    Schneider, Y

    T. Schneider, Y. Hu, X. Gao, J. Dumas, D. Zorin, and D. Panozzo , A large-scale comparison of tetrahedral and hexahedral elements for solving elliptic PDE s with the finite element method , ACM Transactions on Graphics (TOG), 41 (2022), pp. 1--14

    2022

  25. [25]

    Schneiders , Refining quadrilateral and hexahedral element meshes , Fifth International Meshing Roundtable, 1 (1996)

    R. Schneiders , Refining quadrilateral and hexahedral element meshes , Fifth International Meshing Roundtable, 1 (1996)

    1996

  26. [26]

    383--398

    height 2pt depth -1.6pt width 23pt, Octree-based hexahedral mesh generation , International Journal of Computational Geometry & Applications, 10 (2000), pp. 383--398

    2000

  27. [27]

    Schneiders, R

    R. Schneiders, R. Schindler, and F. Weiler , Octree-based generation of hexahedral element meshes , in Proceedings of the 5th International Meshing Roundtable, Pittsburgh, PA, USA, 1996, Sandia National Laboratories, pp. 205--215

    1996

  28. [28]

    Shapira, A

    L. Shapira, A. Shamir, and D. Cohen-Or , Consistent mesh partitioning and skeletonisation using the shape diameter function , The Visual Computer, 24 (2008), pp. 249--259

    2008

  29. [29]

    J. F. Shepherd and C. R. Johnson , Hexahedral mesh generation constraints , Engineering with Computers, 24 (2008), pp. 195--213

    2008

  30. [30]

    L. Sun, G. Zhao, and X. Ma , Adaptive generation and local refinement methods of three-dimensional hexahedral element mesh , Finite Elements in Analysis and Design, 50 (2012), pp. 184--200

    2012

  31. [31]

    T. J. Tautges , The generation of hexahedral meshes for assembly geometry: Survey and progress , International Journal for Numerical Methods in Engineering, 50 (2001), pp. 2617--2642

    2001

  32. [32]

    H. Tong, E. Halilaj, and Y. J. Zhang , Hybridoctree\_hex: Hybrid octree-based adaptive all-hexahedral mesh generation with J acobian control , Journal of Computational Science, 78 (2024), p. 102278

    2024

  33. [33]

    Tong and Y

    H. Tong and Y. J. Zhang , Fast and robust hexahedral mesh optimization via augmented L agrangian, L-BFGS , and line search , in SIAM International Meshing Roundtable, Philadelphia, PA, 2025, Society for Industrial and Applied Mathematics, pp. 24--33

    2025

  34. [34]

    Tong and Y

    H. Tong and Y. J. Zhang , M CH ex: Marching cubes based adaptive hexahedral mesh generation with guaranteed positive J acobian , 2025

    2025

  35. [35]

    E. Wang, T. Nelson, and R. Rauch , Back to elements - tetrahedra vs. hexahedra. P roceedings of the 2004 I nternational ANSYS C onference , 01 2004

    2004

  36. [36]

    W. Wang, Y. Cao, and T. Okaze , Comparison of hexahedral, tetrahedral and polyhedral cells for reproducing the wind field around an isolated building by LES , Building and Environment, 195 (2021), p. 107717

    2021

  37. [37]

    Y. Zhang , Challenges and advances in image-based geometric modeling and mesh generation , in Image-Based Geometric Modeling and Mesh Generation, Springer, Dordrecht, The Netherlands, 2013, pp. 1--10

    2013

  38. [38]

    Zhang and C

    Y. Zhang and C. Bajaj , Adaptive and quality quadrilateral/hexahedral meshing from volumetric data , Computer Methods in Applied Mechanics and Engineering, 195 (2006), pp. 942--960

    2006

  39. [39]

    Zhang, T

    Y. Zhang, T. J. Hughes, and C. L. Bajaj , An automatic 3 D mesh generation method for domains with multiple materials , Computer Methods in Applied Mechanics and Engineering, 199 (2010), pp. 405--415

    2010

  40. [40]

    Zhang, X

    Y. Zhang, X. Liang, and G. Xu , A robust 2-refinement algorithm in octree or rhombic dodecahedral tree based all-hexahedral mesh generation , Computer Methods in Applied Mechanics and Engineering, 256 (2013), pp. 88--100

    2013

  41. [41]

    Y. J. Zhang , Geometric Modeling and Mesh Generation from Scanned Images , Chapman & Hall/CRC, Boca Raton, FL, 2016

    2016