Pith. sign in

REVIEW 4 major objections 8 minor 26 references

Naturally curved quadrilateral mesh generation using an adaptive spectral element solver

T0 review · 4 major / 8 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read An adaptive spectral element Laplace solve makes naturally curved quadrilateral mesh generation automatic.

desk verdict Useful incremental extension of the authors' guiding-field mesher, with an unproven efficiency claim and a sensor caveat they honestly document. read the letter →

arxiv 1908.04272 v2 pith:DHWGK2NT submitted 2019-08-12 math.NA cs.CGcs.NA

Julian Marcon , David A. Kopriva , Spencer J. Sherwin , Joaquim Peiró This is my paper · ORCID
classification math.NAcs.CGcs.NA MSC 65N3065N3565N50
keywords crossfieldguidingquadrilateralmeshgenerationspectralelementmethodp-adaptationhigh-orderPoincareindexLaplaceequation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that field-guided quadrilateral mesh generation for two-dimensional domains can be made automatic by computing a smooth guiding field with an adaptively refined spectral element Laplace solve, then reading the block decomposition directly from that field's critical points and streamlines. The new ingredient is p-adaptation: a cheap modal sensor, the relative energy of the last polynomial mode, raises or lowers the local approximation order until an error tolerance is met, removing the ad hoc choice of a uniform high order. If the argument is right, users obtain valid naturally curved quadrilateral blocks, with few irregular nodes, for multiply connected domains and sharp corners, and those coarse blocks can be refined into finer high-order meshes while preserving validity.

What carries the argument

The carrying object is the guiding field, a pair of harmonic scalar fields $(u,v)$ whose boundary data encode the fourfold rotational symmetry of the boundary tangent directions; the phase function $\psi = \tfrac14 \operatorname{atan2}(v,u)$ and its jump lines carry the topological information. Three formulas do the work: the modal energy sensor $S_e$ that drives p-adaptation, the contour integral $I_c$ that converts the winding of $\psi$ into a valence, and the streamline ODE $d\mathbf{x}/dt = \tilde{v}(\psi'(\mathbf{x}))$ with the direction adjusted across jump lines. Supporting operations include a reference-space Newton search for zeros of $(u,v)$ to locate critical points, streamline integration with synchronous merging of opposing fronts, and an isoparametric splitting that refines coarse curved blocks into finer high-order meshes while retaining validity.

What would settle it

Take a domain with an analytic harmonic guiding field, such as an annulus or a rectangle with known boundary data, and compare the p-adapted solution's critical point positions and separatrix graph with the exact field as the thresholds $\epsilon_u$ and $\epsilon_l$ vary. Finding a case where tightening the tolerance moves a critical point by more than the integration step, or where the $v$-based sensor drives the order higher while the field is already exact, would show that the sensor is not tracking the error that matters for mesh validity.

Watch

Extended reading notes

Core claim

The central claim is that the entire pipeline of field-guided quadrilateral meshing—locating irregular nodes, computing valences, tracing separatrices, and splitting a CAD model into curved quadrilateral blocks—can run on a guiding field $(u,v)$ computed as the solution of Laplace's equation $\nabla^2 \mathbf{v} = 0$, with boundary values $(u_b,v_b) = (\cos 4\theta_b, \sin 4\theta_b)$ set by the boundary tangent angle. The phase $\psi = \tfrac14 \operatorname{atan2}(v,u)$ is tangent or orthogonal to the field's streamlines; its $\pi/2$ jumps mark jump lines, and its zeros coincide with critical points. From the Poincar\'e index $I_c = \frac{1}{\pi/2}\oint_c \frac{d\psi}{d\theta}\,d\theta$ the paper derives the valence $V = 4 - I_c$ for interior nodes and a corresponding open-contour formula for boundary corners, then integrates streamlines from these nodes using an adjusted direction that respects the $\pi/2$ rotational symmetry. Although motivated by cross fields, the method never constructs crosses: all analysis is performed on the guiding field itself. The new ingredient is an elementwise p-adaptive spectral element solve driven by the modal error sensor $S_e = \|u_P - u_{P-1}\|^2_{2,e}/\|u_P\|^2_{2,e}$, with a discontinuous Galerkin discretization available where corners make the boundary data discontinuous. The paper argues that this produces a high-resolution field—and therefore accurate critical points and separatrices—at lower cost than a uniform high-order solve, and demonstrates on five geometries that the resulting coarse curved blocks are valid, preserve symmetries, avoid limit cycles, and can be refined further with an isoparametric splitting.

Load-bearing premise

The modal energy ratio $S_e = \|u_P - u_{P-1}\|^2 / \|u_P\|^2$ is assumed to mark the elements where error in the guiding field would actually spoil critical-point location or streamline integration; if that link fails, the adaptively chosen orders could refine in the wrong places.

Editorial extensions

If this is right

  • Users no longer need to choose a uniform polynomial order for the guiding-field solve; the adaptive loop meets an error tolerance with locally chosen orders, reducing cost on smooth regions.
  • Block decompositions inherit the smoothness of a harmonic field, which the paper argues prevents limit cycles and preserves symmetries and patterns in the tested geometries.
  • Corners whose angles are not multiples of $\pi/2$ can be handled through a discontinuous Galerkin formulation rather than ad hoc smoothing of the boundary data.
  • The coarse naturally curved blocks can be split using isoparametric refinement into finer high-order meshes while preserving element validity.
  • For the five geometries tested, the whole mesh generation process takes on the order of a hundred seconds, with streamline integration accounting for roughly half to ninety percent of that time.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A practical robustness fix suggested by the paper's own Fig. 14 would be to run the sensor on both components of the guiding field and take the maximum, or to use a rotationally invariant combination; as written, the $v$-based sensor over-refines on Cartesian-aligned geometry because a nearly constant $v$ component carries no energy in its low modes.
  • The modal sensor is heuristic; replacing it with a certified a posteriori error estimator for the Laplace solve would turn p-adaptation from a cost-saving device into a guarantee about critical-point location, which is the quantity that actually drives mesh validity.
  • If the cost scaling holds, the same adaptive strategy should be even more valuable in three-dimensional hexahedral block decomposition, where uniform high-order solves are expensive and the field-topology analysis is more delicate.
  • A testable consequence is that on a domain with an analytic guiding field, the separatrix graph produced by the p-adaptive pipeline should match the exact graph once the sensor threshold is tightened, and the degrees of freedom needed should stay below the uniform-order count.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 8 minor

Summary. The manuscript describes an adaptive version of a field-guided method for generating naturally curved quadrilateral block decompositions and meshes on multiply connected two-dimensional domains. A guiding field is obtained by solving Laplace equations for two components u and v with boundary conditions derived from the boundary tangent angle, using a continuous or discontinuous Galerkin spectral element method. Critical points are located as zeros of the field, valences are computed from a Poincaré-index integral, and streamlines are integrated and merged to form a separatrix graph that is then used to split the CAD model into quadrilateral blocks. The p-adaptation strategy of Algorithm 1 adjusts the local polynomial order element by element using a modal-energy-ratio sensor defined in Eq. (19). Five example geometries demonstrate valid coarse and refined high-order quadrilateral meshes, and the paper reports that all computations take on the order of 100 seconds in the Nektar++/NekMesh framework.

Significance. If the claims hold, the paper offers a practical, open-source pipeline for generating high-order curved quadrilateral meshes with a small number of irregular nodes, while avoiding the explicit construction of cross fields and handling non-orthogonal corners through a DG discretization. The underlying Laplace model, Poincaré-index relations, and streamline-integration construction are standard and correctly assembled, and the examples provide credible visual evidence of valid block decompositions on benchmark geometries from the cross-field literature. However, the central benefit claim concerning p-adaptation, namely that it achieves faster convergence of the solution with respect to computational cost, is not supported by quantitative evidence in the manuscript. The proposed sensor is not shown to track the actual error in the guiding-field components, and the observed geometry-dependence of the sensor-component choice (Sec. 4, Fig. 14) reintroduces an ad hoc element into the procedure. The paper is therefore an interesting proof of concept, but the adaptation advantage needs stronger validation before the main efficiency claim can be accepted.

major comments (4)
  1. [Sec. 2.2.1, Eq. (19), Algorithm 1] The modal-energy sensor Se = ||u_P - u_{P-1}||^2 / ||u_P||^2 is introduced as an error indicator, but no evidence is provided that this smoothness indicator, borrowed from shock-capturing and compressible-flow adaptation ([16], [8,9]), bounds or reliably tracks the discretization error in the Laplace guiding-field components. The thresholds epsilon_u, epsilon_l, the bounds Pmax and Pmin, and the convergence/stopping behavior of the while-loop in Algorithm 1 are never specified. Since the paper's abstract and introduction claim that p-adaptation yields faster convergence with respect to computational cost, this unsupported link between Se and the field accuracy relevant to critical-point location and streamline integration is load-bearing and needs to be established with numerical experiments.
  2. [Sec. 4, Fig. 14] The paper itself demonstrates that the sensor behaves poorly for the v-component on Cartesian-aligned geometries: because v is nearly constant in large parts of Geometry I, the low-order modes carry no energy, and any high-mode energy spuriously triggers high polynomial orders. The suggested remedy, to base the sensor on u for such geometries, is a rule of thumb with no stated scope or criterion. This means the method replaces the ad hoc choice of polynomial order, which the introduction criticizes, with an ad hoc choice of which solution component to use in the sensor. A general, geometry-independent sensor selection rule, or a robust combination of both components, is required to support the claim that p-adaptation removes the ad hoc nature of the approximation-order choice.
  3. [Sec. 4, abstract] The central efficiency claim that p-adaptation achieves faster convergence of the solution with respect to computational cost is not quantified anywhere in the paper. The only timing information is the global statement that examples take O(100) seconds and that streamline integration consumes 50%-90% of that time. There is no comparison of the adaptive Laplace solve against a uniform high-order solve in terms of degrees of freedom, wall-clock time, or accuracy, and no convergence study showing that the adaptive strategy reduces cost for a given accuracy in critical-point positions, valences, or separatrix locations. Without such a comparison, the claimed advantage of p-adaptation over a uniform-order solve is unsupported.
  4. [Sec. 3.2.2] There is an internal sign inconsistency in the valence calculation. The text states that a positive jump in psi (from -pi/4 to pi/4) leads to a negative index Ic=-1, and that the two reference critical points are characterized by a single negative jump line and therefore have valence V=3. However, Eq. (9) gives V = 4 - Ic, and the cited result [1, Lemma 5.1] says Ic=+1 gives V=3 and Ic=-1 gives V=5. A positive jump from -pi/4 to pi/4 increases psi by pi/2, so Eq. (8) yields Ic=+1, not -1. The text and the formula therefore disagree about the sign convention, which is critical for reproducing the valence computation. Please correct the sign description or the formulas and re-verify the stated valences in Fig. 1.
minor comments (8)
  1. [Sec. 2.2.1] There is a typo in the sentence following Eq. (19): "phyically" should be "physically".
  2. [Sec. 3.2.1] The text contains "Netwon's method" and should read "Newton's method."
  3. [Fig. 5 caption] The caption contains "Higher polonomial orders" and should read "Higher polynomial orders."
  4. [Sec. 3.2.3] The phrase "the high order interpolation fo ⃗v" should be "the high order interpolation of ⃗v."
  5. [Sec. 2.1.1, Eq. (11)] The expression "Delta theta / pi/2" is ambiguous; it should be written as Delta theta / (pi/2) or, more clearly, as 2 Delta theta / pi.
  6. [Sec. 2.1.1] The term "open line circle integral" is unusual and should be replaced by "open line integral" or "integral over a partial contour."
  7. [Sec. 3.1] The condition for continuous boundary conditions is phrased as "the curve at each point is C1-continuous"; this should be reworded to say the boundary curve is C1 at every point.
  8. [References] Reference [6] is listed as "Journal of Computational Physics, 2019. In press" without volume or article number; please update it with the final publication data if available.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core guiding-field derivation, valence formulas, and streamline integration are parameter-free and benchmarked against external cross-field literature; p-adaptation is an explicitly imported heuristic rather than a fitted prediction.

full rationale

The paper's central chain is: boundary conditions (4) define the Laplace BVP (5), which is solved with the spectral element scheme (14)-(17); critical points are located from zero contours of the solution field (Sec. 3.2.1); valences are computed from the index integrals (8)-(11); and streamlines are integrated from (12). None of these steps is defined in terms of the final mesh, and no parameter is fitted to the quantities being predicted. The valence/index relations are cited to external cross-field literature ([1], [14]), and the benchmark geometries are standard external examples ([1], [10], [14]). Self-citations to [6], [8], and [9] appear, but they inherit implementation details and the p-adaptation loop; they are not used to justify the core mathematical claims, which are presented and derived in this paper. The p-adaptation sensor Se in Eq. (19) is a heuristic from shock capturing ([16]) and prior adaptation work ([8], [9]), and the paper itself acknowledges in Sec. 4 / Fig. 14 that the v-based sensor misfires on Cartesian-aligned geometries because low-order modes carry no energy. This is a correctness/robustness limitation of a tuning heuristic, not a circular reduction: Se is not fitted to the reported mesh outputs, no predicted quantity is defined as the result of the sensor, and the paper does not claim a rigorous error-estimation theorem for it. Thus no load-bearing circularity is present; the derivation is self-contained relative to its external benchmarks.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

No new physical or mathematical entities are introduced; the guiding field, jump lines, and valence relations come from the authors' prior work [6] and the cross-field literature. Free parameters are algorithmic tolerances and radii chosen by the user, not parameters fitted to data. The central assumptions are the PDE model for the guiding field and the reliability of the modal-energy sensor as an error indicator.

free parameters (5)
  • Error thresholds epsilon_u and epsilon_l = not specified in paper
    Algorithm 1 raises the polynomial order when Se > epsilon_u and lowers it when Se < epsilon_l. The values define the adapted solution quality but are never given; they are user choices (Sec. 2.2.1).
  • Contour radius c for valence integral = distance to nearest critical point, unspecified numeric value
    Radius used in Eq. (8) around each critical point; must enclose exactly one critical point. Chosen heuristically (Sec. 2.1.1, Sec. 3.2.2).
  • Streamline merging distance = integration step size, or 5x step for aggressive merging
    Front points closer than this threshold and pointing in opposite directions are merged, changing the separatrix graph topology (Sec. 3.2.3, Fig. 6).
  • Polynomial order bounds Pmax and Pmin = not specified
    Limits on the adapted order in Algorithm 1; presumably problem-dependent user choices.
  • Adams-Bashforth integration step size = not specified
    Step size for the 4th order multi-step integrator of Eq. (12); affects merging decisions and accuracy of the separatrices (Sec. 3.2.3).
assumptions (5)
  • domain assumption Boundary values vb = (cos 4*theta_b, sin 4*theta_b) correctly encode boundary alignment for quadrilateral meshing.
    Entered in Sec. 2.1, Eq. (4). The entire guiding-field method depends on this encoding, inherited from [6] and the cross-field literature; its validity is not derived in this paper.
  • domain assumption The Laplace equation (5) propagates boundary alignment smoothly into the interior, and its solutions have the critical points presupposed.
    Entered in Sec. 2.1, Eq. (5). Harmonic interpolation of crosses is standard in the cross-field literature [1,14] but is a modeling assumption, not a theorem proved here.
  • standard math Valence-index relations V = 4 - I_c (Eq. 9) and Eq. (11) correctly determine node valence from the guiding field.
    Adapted from the Poincare index and cited to [1, Lemma 5.1] and [14] in Sec. 2.1.1. Assumes a circle can be chosen around each isolated critical point.
  • ad hoc to paper The modal-energy sensor (19) is a reliable error indicator for the Laplace guiding-field components.
    The sensor is taken from shock-capturing and compressible-flow contexts [8,9,16] and applied here to elliptic Laplace solves without validation; the paper concedes anomalous behavior for the v-component (Sec. 4, Fig. 14).
  • domain assumption Spectral element solutions on the coarse triangular mesh represent the guiding field accurately enough for critical point location and streamline tracing.
    Spectral accuracy holds for smooth fields [3], but corner singularities in the boundary data (discontinuous BCs) require the DG handling whose details are deferred in Sec. 3.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Naturally curved quadrilateral mesh generation using an adaptive spectral element solver." pith.science (2026). https://pith.science/paper/DHWGK2NT

@misc{pith2026190804272,
  author       = {Pith},
  title        = {Pith review of: Naturally curved quadrilateral mesh generation using an adaptive spectral element solver},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DHWGK2NT}},
  note         = {Machine review of arXiv:1908.04272}
}
abstract

We describe an adaptive version of a method for generating valid naturally curved quadrilateral meshes. The method uses a guiding field, derived from the concept of a cross field, to create block decompositions of multiply connected two dimensional domains. The a priori curved quadrilateral blocks can be further split into a finer high-order mesh as needed. The guiding field is computed by a Laplace equation solver using a continuous Galerkin or discontinuous Galerkin spectral element formulation. This operation is aided by using $p$-adaptation to achieve faster convergence of the solution with respect to the computational cost. From the guiding field, irregular nodes and separatrices can be accurately located. A first version of the code is implemented in the open source spectral element framework Nektar++ and its dedicated high order mesh generation platform NekMesh.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

  1. [6]

    A High Resolution PDE Approach to Quadrilateral Mesh Generation

    Marcon J., Kopriva D., Sherwin S., Peir´ o J. “A High Resolution PDE Approach to Quadrilateral Mesh Generation.” Journal of Computational Physics, 2019. In press

    work page 2019

  2. [16]

    Sub-Cell Shock Captur- ing for Discontinuous Galerkin Methods

    Persson P.O., Peraire J. “Sub-Cell Shock Captur- ing for Discontinuous Galerkin Methods.” 44th AIAA Aerospace Sciences Meeting and Exhibit . American Institute of Aeronautics and Astronau- tics, Reston, Virigina, jan 2006

    work page 2006

  3. [1]

    An Approach to Quad Meshing Based on Harmonic Cross-Valued Maps and the Ginzburg-Landau Theory

    Viertel R., Osting B. “An Approach to Quad Meshing Based on Harmonic Cross-Valued Maps and the Ginzburg-Landau Theory.” 2017

    work page 2017

  4. [2]

    Simula- tion of Dynamic Earthquake Ruptures in Com- plex Geometries Using High-Order Finite Differ- ence Methods

    Kozdon J., Dunham E., Nordstr¨ om J. “Simula- tion of Dynamic Earthquake Ruptures in Com- plex Geometries Using High-Order Finite Differ- ence Methods.” Journal of Scientific Computing , vol. 55, no. 1, 92–124, 2013

    work page 2013

  5. [3]

    Spectral/hp Ele- ment Methods for Computational Fluid Dynamics (Second Edition)

    Karniadakis G.E., Sherwin S.J. Spectral/hp Ele- ment Methods for Computational Fluid Dynamics (Second Edition). Oxford University Press, 2005

    work page 2005

  6. [4]

    Implementing Spectral Methods for Partial Differential Equations

    Kopriva D.A. Implementing Spectral Methods for Partial Differential Equations. Scientific Compu- tation. Springer, May 2009

    work page 2009

  7. [5]

    Quad-Mesh Generation and Processing: A Survey

    Bommes D., L´ evy B., Pietroni N., Puppo E., Silva C., Tarini M., Zorin D. “Quad-Mesh Generation and Processing: A Survey.” Computer Graphics Forum, vol. 32, no. 6, 51–76, 2013

    work page 2013

  8. [7]

    Compari- son of Adaptive h and p Refinements for Spectral Difference Methods

    Li Y., Premasuthan S., Jameson A. “Compari- son of Adaptive h and p Refinements for Spectral Difference Methods.” 40th Fluid Dynamics Con- ference and Exhibit . American Institute of Aero- nautics and Astronautics, Reston, Virigina, jun 2010

    work page 2010

Show all 26 references
  1. [8]

    A p-adaptation method for compressible flow prob- lems using a goal-based error indicator

    Ekelschot D., Moxey D., Sherwin S., Peir´ o J. “A p-adaptation method for compressible flow prob- lems using a goal-based error indicator.” Com- puters & Structures , apr 2016

    2016

  2. [9]

    Towards p-Adaptive Spectral/hp Element Methods for Modelling Industrial Flows

    Moxey D., Cantwell C., Mengaldo G., Serson D., Ekelschot D., Peir´ o J., Sherwin S., Kirby R.M. “Towards p-Adaptive Spectral/hp Element Methods for Modelling Industrial Flows.” Spec- tral and High Order Methods for Partial Differ- ential Equations ICOSAHOM 2016 , pp. 63–79. Spr...

    2016

  3. [10]

    Toward a paver replacement

    Viertel R., Staten M., Osting B. “Toward a paver replacement.” 26th International Meshing Roundtable, IMR26. 2017

    2017

  4. [11]

    A Continuum Theory for Unstruc- tured Mesh Generation in Two Dimensions

    Bunin G. “A Continuum Theory for Unstruc- tured Mesh Generation in Two Dimensions.” Tech. rep., sep 2006

    2006

  5. [12]

    Towards Unstructured Mesh Genera- tion Using the Inverse Poisson Problem

    Bunin G. “Towards Unstructured Mesh Genera- tion Using the Inverse Poisson Problem.” ArXiv e-prints, Feb. 2008

    2008

  6. [13]

    Con- tinuous Topology Simplification of Planar Vector Fields

    Tricoche X., Scheuermann G., Hagen H. “Con- tinuous Topology Simplification of Planar Vector Fields.” Proceedings of the Conference on Visu- alization ’01 , VIS ’01, pp. 159–166. IEEE Com- puter Society, Washington, DC, USA, 2001

    2001

  7. [14]

    A PDE based approach to multi-domain partitioning and quadrilateral meshing

    Kowalski N., Ledoux F., Frey P. “A PDE based approach to multi-domain partitioning and quadrilateral meshing.” 21th Int. Meshing Roundtable. 2012

    2012

  8. [15]

    Nektar++: An open-source spectral/hp element framework

    Cantwell C., Moxey D., Comerford A., Bolis A., Rocco G., Mengaldo G., De Grazia D., Yakovlev S., Lombard J.E., Ekelschot D., Jordi B., Xu H., Mohamied Y., Eskilsson C., Nelson B., Vos P., Biotto C., Kirby R., Sherwin S. “Nektar++: An open-source spectral/hp element framework.”...

    2015

  9. [17]

    Nektar++: enhancing the capability and appli- cation of high-fidelity spectral/hp element meth- ods

    Moxey D., Cantwell C., Bao Y., Cassinelli A., Castiglioni G., Chun S., Juda E., Kazemi E., Lackhove K., Marcon J., Mengaldo G., Serson D., Turner M., Xu H., Peir´ o J., Kirby R., Sherwin S. “Nektar++: enhancing the capability and appli- cation of high-fidelity spectral/hp eleme...

    2019

  10. [18]

    Mesh generation in curvi- linear domains using high-order elements

    Sherwin S., Peir´ o J. “Mesh generation in curvi- linear domains using high-order elements.” Inter- national Journal for Numerical Methods in Engi- neering, vol. 53, no. 1, 207–223, jan 2002

    2002

  11. [19]

    Nonlinear Par- ticle Tracking for High-Order Elements

    Coppola G., Sherwin S., Peir´ o J. “Nonlinear Par- ticle Tracking for High-Order Elements.” Journal of Computational Physics , vol. 172, no. 1, 356– 386, sep 2001

    2001

  12. [20]

    Hexahedral meshing using midpoint subdivision and integer programming

    Li T., McKeag R., Armstrong C. “Hexahedral meshing using midpoint subdivision and integer programming.” Computer Methods in Applied Mechanics and Engineering , vol. 124, no. 1-2, 171–193, jun 1995

    1995

  13. [21]

    Open Cascade

    Open Cascade SAS. “Open Cascade.”, 2018

    2018

  14. [22]

    An isoparametric approach to high-order curvilin- ear boundary-layer meshing

    Moxey D., Green M., Sherwin S., Peir´ o J. “An isoparametric approach to high-order curvilin- ear boundary-layer meshing.” Computer Methods in Applied Mechanics and Engineering , vol. 283, 636–650, jan 2015

    2015

  15. [23]

    High-order curvilinear hybrid mesh generation for CFD sim- ulations

    Marcon J., Turner M., Peir´ o J., Moxey D., Pol- lard C., Bucklow H., Gammon M. “High-order curvilinear hybrid mesh generation for CFD sim- ulations.” 2018 AIAA Aerospace Sciences Meet- ing, January. American Institute of Aeronautics and Astronautics, Reston, Virginia, jan 2018

    2018

  16. [24]

    A semi-structured approach to curvilinear mesh generation around streamlined bodies

    Marcon J., Peir´ o J., Moxey D., Bergemann N., Bucklow H., Gammon M. “A semi-structured approach to curvilinear mesh generation around streamlined bodies.” AIAA Scitech 2019 Forum . American Institute of Aeronautics and Astronau- tics, Reston, Virginia, jan 2019

    2019

  17. [25]

    Practical 3D frame field generation

    Ray N., Sokolov D., L´ evy B. “Practical 3D frame field generation.” ACM Transactions on Graph- ics, vol. 35, no. 6, 1–9, nov 2016

    2016

  18. [26]

    Multi-Block Decomposition Using Cross- Fields

    Fogg H.J., Armstrong C.G., Robinson T.T. “Multi-Block Decomposition Using Cross- Fields.” J.P.M. de Almeida, P. D´ ıez, C. Tiago, N. Pares, editors, VI International Conference on Adaptive Modeling and Simulation . 2013

    2013

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.