Pith. sign in

REVIEW 3 major objections 5 minor 37 references

PDE-Constrained High-Order Mesh Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper proposes PDE-constrained r-adaptivity: moving high-order mesh nodes to minimize a weighted sum of mesh quality and a PDE solution error measure, using adjoint gradients, achieving up to 10x smaller errors on Poisson and linear…

desk verdict A genuinely useful integration of TMOP with adjoint-based r-adaptivity, but Eq. (17) has a real sign error in the central sensitivity derivation that must be fixed before the method is reproducible. read the letter →

arxiv 2507.01917 v1 pith:57BFO7DY submitted 2025-07-02 math.NA cs.MScs.NA

Tzanio Kolev , Boyan Lazarov , Ketan Mittal , Mathias Schmidt , Vladimir Tomov This is my paper
classification math.NAcs.MScs.NA MSC 65N3065N5065K10
keywords meshoptimizationr-adaptivityhigh-orderfiniteelementsPDE-constrainedadjointsensitivityanalysisTarget-MatrixParadigmerrormeasureslinearelasticity
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

This paper is trying to establish that r-adaptivity—moving nodes of an existing mesh without changing connectivity or polynomial degree—can be posed as a single optimization problem that couples mesh quality with the accuracy of a finite-element solution. The objective is a weighted sum of a mesh-quality metric and one of three PDE-error measures, and the gradient of that objective with respect to node positions is computed exactly through adjoint sensitivity analysis. If the approach is right, a user can tune one weight and a filter radius to move mesh resolution toward regions where the PDE solution is hard to approximate, recovering optimal convergence rates and cutting discretization error by about a factor of ten on smooth Poisson problems. The same mechanism also improves displacement error and load-functional accuracy for linear elasticity, with the strongest gains on lower-order meshes.

What carries the argument

The load-bearing device is the PDE-constrained optimization problem in (11) and its adjoint-based gradient. The state equation $R_P(u;x)=0$ ties the finite-element solution $u$ to the nodal coordinates $x$; the adjoint variable $\lambda$ solves $\left(\partial R_P/\partial u\right)^T\lambda=\partial F_p/\partial u$, and the implicit term $\partial F_p/\partial u\,\partial u/\partial x$ is replaced by $\lambda^T\partial R_P/\partial x$. Mesh quality enters through the Target-Matrix Optimization Paradigm (TMOP), a method that compares each element's Jacobian $A$ to a prescribed target matrix $W$—ideal square, cube, or equilateral simplex—via a metric such as shape or shape-plus-orientation. A Helmholtz-type convolution filter smooths the raw node displacements before they update the mesh, preventing localized gradients from distorting high-order elements.

What would settle it

Compute the true $L^2$ error on the smooth tan-inverse Poisson problem for meshes optimized by the gradient-continuity measure at several element counts; if any optimized mesh has larger $L^2$ error than the uniform mesh of the same order while $F_p$ decreased, the heuristic's claimed control of discretization error fails.

Watch

Extended reading notes

Core claim

The central claim is that moving mesh nodes according to the gradient of an objective $F(x)=\alpha F_p(u(x),x)+F_\mu(x)$, subject to the PDE residual $R_P(u;x)=0$, simultaneously controls element quality and discretization error. For elliptic problems, choosing $F_p$ as the negative load functional makes the reduction exact: by Galerkin orthogonality, the energy norm of the error equals $l(u)-l(u_h)$, so decreasing $-l(u_h)$ provably decreases the energy error. The other two measures—element-local variation and gradient continuity—are heuristic surrogates that concentrate nodes where the solution varies sharply or its gradient changes across element boundaries; the paper shows numerically that on smooth Poisson problems they recover $\mathcal{O}(h^{p+1})$ convergence and reduce the $L^2$ error by up to a factor of 10 relative to a uniform mesh.

Load-bearing premise

The load-bearing premise is that the chosen error measures are smooth functions of node positions whose minimization actually lowers the true discretization error; this is guaranteed for the negative load functional by Galerkin orthogonality but only demonstrated experimentally for the element-local-variation and gradient-continuity measures, and the paper shows that a poor choice of the weight $\alpha$ can make the optimized mesh worse than the uniform one.

Editorial extensions

If this is right

  • For smooth Poisson problems, optimized linear, quadratic, and cubic meshes recover the expected convergence order $\mathcal{O}(h^{p+1})$ while the $L^2$ error drops by roughly a factor of 10 relative to the initial uniform mesh.
  • The load-functional measure applies directly to vector-valued problems: on a cantilever beam and a shear wall with a hole, optimized meshes bring the maximum displacement and the load functional closer to the reference solution, with the largest improvement for linear elements.
  • The weight $\alpha$ is a genuine tuning knob: increasing it from zero first improves solution error and then degrades mesh quality, so the framework exposes the trade-off between error control and element shape explicitly.
  • The method is not tied to a specific element type or dimension: the same objective adapts quadrilateral, triangular, hexahedral, and tetrahedral meshes, including unstructured meshes and cases where the mesh order differs from the solution order.

Reading between the lines

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

  • A natural extension the paper leaves implicit is automatic selection of $\alpha$ and the filter radius by scanning the reported trade-off curve; the strong degradation at $\alpha=4\cdot10^6$ suggests a Pareto front exists between error reduction and mesh quality.
  • The exact equivalence of the load-functional measure relies on Galerkin orthogonality for self-adjoint elliptic problems; carrying the same pipeline to nonsymmetric, hyperbolic, or time-dependent PDEs would require a different provably consistent error surrogate, while the adjoint machinery itself should still apply.
  • Because the optimizer cannot alter connectivity or element count, this r-adaptivity step would pair naturally with h-adaptivity: using the same error measure to cluster existing nodes first could give a better starting point for subsequent local refinement, a combination the paper does not test.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript proposes a PDE-constrained r-adaptivity framework for high-order finite element meshes. Mesh nodes are moved by solving an optimization problem whose objective is a convex combination of a TMOP mesh-quality metric and a PDE-dependent error measure; the implicit sensitivity of the error measure with respect to nodal positions is computed by adjoint sensitivity analysis, with a Helmholtz-type filter for regularization and MMA as the optimizer. Three error measures are considered: element-local variation, the negative load functional, and gradient continuity. Numerical experiments on Poisson problems in 2D/3D and on linear elastostatic problems report up to about 10x reductions in L2 error relative to uniform meshes and improved load functionals relative to reference solutions.

Significance. The framework is timely and potentially useful because it couples mesh quality and PDE accuracy in a single differentiable objective. The load-functional measure rests on solid ground: Eq. (24) is an exact Galerkin-orthogonality identity, so minimizing -l(u_h) is equivalent to minimizing the energy norm of the error. The experiments compare against analytic or high-order reference solutions, and the use of automatic differentiation for TMOP metrics should make extension to new metrics straightforward. However, the central adjoint relation (17) is written with an internal sign inconsistency, and the formulation does not specify how boundary nodes are treated; as printed, the method is not reproducible. If these issues are corrected, the contribution is a worthwhile advance in high-order mesh optimization.

major comments (3)
  1. [Section 3, Eqs. (11) and (18)] The sign in Eq. (17) is wrong. Let A = ∂R_P/∂u, B = ∂R_P/∂x, and f = ∂F_P/∂u. Eq. (15) gives du/dx = -A^{-1}B, and Eq. (16) gives λ_u = A^{-T}f. The chain-rule term in Eq. (13) is f^T du/dx = -f^T A^{-1}B = -λ_u^T B. Eq. (17) first writes this term as λ_u^T ∂R_P/∂x and then asserts it is also equal to -((∂R_P/∂u)^{-T} ∂F_p/∂u)^T ∂R_P/∂x. These two expressions are negatives of each other, so the equation would force λ_u^T ∂R_P/∂x = 0 for every admissible state, which is not true in general. A literal implementation of Eq. (17) would compute either the wrong-sign gradient or a zero contribution, and MMA would not minimize the stated objective. The numerical results presumably come from the correct expression -λ_u^T ∂R_P/∂x, but the manuscript as written cannot be used to reproduce the method. This is the core sensitivity formula of the paper and must be corrected.
  2. [Section 4.3, Eq. (28)] The optimization variables and boundary treatment are not specified. Problem (11) is written as a minimization over all nodal coordinates x, and the filtered displacement field w~ used to update x is not constrained on ∂Ω. If boundary nodes are free to move, the optimized mesh no longer represents the fixed domain Ω on which the exact solutions (34) and (35) are defined, so the procedure would silently change the geometry rather than perform r-adaptivity. All reported examples are on fixed domains, and the figures appear to keep boundary nodes on the boundary, but the text never states whether w is zero on Γ, projected to be tangential, or otherwise constrained, nor how such a constraint enters the filter equation and the adjoint calculation. The revision must define the admissible set for w (e.g., homogeneous Dirichlet on the boundary, or tangential movement for curved boundaries) and give the corresponding modification of Eq. (18).
  3. The sensitivity analysis for the gradient-continuity measure is incomplete. The objective depends on u and on the smoothed gradient g defined by the projection constraint R_π(g; x, u) = 0, but the manuscript only states the extended chain rule (28). It does not derive the adjoint equation for g (or for the coupled (u, g) system), nor does it give the expression for ∂g/∂x needed in (28). Since Fig. 6(c) and the corresponding convergence results rely on this measure, the revision should include the additional adjoint solve, or cite a prior derivation with the exact equations. Without this, the gradient-continuity experiments cannot be reproduced from the paper.
minor comments (5)
  1. [Section 8.1 and Table 1] The 10x improvement is obtained after per-problem tuning of α, δ, and the iteration count, as the text acknowledges. This is acceptable for a demonstration, but a few sentences giving the chosen values and a short sensitivity study (or selection guidelines) would substantially strengthen the headline claim.
  2. [Section 6] The stopping criterion 'the ratio of the norm of the gradient of the objective with respect to the current and initial mesh decreases under a certain threshold' is not quantified; the threshold should be reported for reproducibility.
  3. [Section 3.2] The filter defined in Eqs. (18)-(20) is a Helmholtz-type PDE filter, not literally a convolution; the terminology is standard in topology optimization but should be clarified to avoid confusion.
  4. [Section 8.1 and Figure 6] After optimization the mesh is no longer uniform, so h is not well-defined for the optimized data points; the text should state explicitly that h denotes the element size of the initial uniform mesh before adaptation.
  5. [Section 4.2, Eq. (25)] The source term is denoted b in Section 2 but f in Eq. (25); the notation should be unified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivations are self-contained and the numerical claims are checked against independent exact or enriched-reference solutions.

full rationale

The paper's derivation chain does not reduce to its own inputs. The load-functional objective in Eq. (25) is justified by the exact Galerkin-orthogonality identity in Eq. (24), and the reported reductions in discretization error are measured against either the manufactured exact solution of Eq. (34) or an enriched-space reference solution in Section 8.3, not against the objective values used by the optimizer. The element-local variation and gradient-continuity error measures are heuristics, but they are not fitted to the reported errors; their optimization is validated by L2 and H1 convergence behavior in Figure 6. The TMOP mesh-quality term (30) and the Helmholtz-type convolution filter (18)-(20) are adopted from prior literature, including some self-citations, but they are auxiliary, externally established components and are not the load-bearing claim that the proposed r-adaptivity reduces PDE discretization error. The paper also explicitly acknowledges that alpha, delta, and iteration counts are tuned (Section 8.1), which is a robustness limitation rather than circularity. The sign inconsistency in Eq. (17) identified by the skeptic is a correctness and reproducibility defect in the printed adjoint sensitivity formula, not a circular reduction of a prediction to its input, so it does not change this circularity verdict.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters are optimization hyperparameters that are tuned per experiment, and the axioms are standard finite element, adjoint, filter, and mesh quality assumptions. The central claim rests on these user choices and standard theory, not on new postulated mechanisms.

free parameters (4)
  • alpha (penalization weight) = varied per experiment (0, 4e3, 4e4, 4e6, 1e6)
    Controls the tradeoff between the PDE error measure and the mesh quality term in Eq. (11); the reported 10x improvement depends on tuning alpha per problem.
  • delta (filter radius) = delta^2 = 0.005 in one example; otherwise tuned
    Controls the smoothness of the filtered displacement field in Eq. (18)-(19); affects optimization stability and mesh distortion.
  • max iterations and gradient ratio threshold = 300 iterations and unspecified ratio threshold
    Stopping criteria for the MMA optimizer described in Section 6; chosen by the user and not systematically studied.
  • gamma (Dirichlet penalty parameter) = 1e5 by default
    Penalty parameter for weakly enforced Dirichlet boundary conditions in Eq. (6); a user choice that affects the discrete problem.
assumptions (4)
  • standard math Galerkin orthogonality and the equivalence of maximizing the load functional to minimizing the energy norm of the error (Eq. 24).
    Used in Section 4.2 to justify the negative load functional as an exact error measure. Standard finite element theory, assuming exact discrete solution.
  • domain assumption The discrete residual R_P(u;x) is differentiable with respect to nodal positions x, and the FE solution u is a smooth function of x throughout the optimization.
    Required for adjoint sensitivity analysis in Eq. (14)-(17); assumes elements remain valid (det(A)>0) and the PDE solution depends smoothly on node locations.
  • domain assumption The Helmholtz-type convolution filter (Eq. 18-20) produces a regularized displacement field that prevents mesh distortion.
    Standard in topology optimization (Lazarov-Sigmund), adopted here for high-order meshes; the filter radius delta is user-tuned.
  • domain assumption The Target-Matrix Optimization Paradigm (TMOP) with prescribed target matrices provides sufficient control over element quality during r-adaptivity.
    Prior work by the same authors; the paper relies on TMOP shape metrics to maintain mesh validity, but the interaction with large PDE-driven displacements is not fully analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PDE-Constrained High-Order Mesh Optimization." pith.science (2026). https://pith.science/paper/57BFO7DY

@misc{pith2026250701917,
  author       = {Pith},
  title        = {Pith review of: PDE-Constrained High-Order Mesh Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/57BFO7DY}},
  note         = {Machine review of arXiv:2507.01917}
}
abstract

We present a novel framework for PDE-constrained $r$-adaptivity of high-order meshes. The proposed method formulates mesh movement as an optimization problem, with an objective function defined as a convex combination of a mesh quality metric and a measure of the accuracy of the PDE solution obtained via finite element discretization. The proposed formulation achieves optimized, well-defined high-order meshes by integrating mesh quality control, PDE solution accuracy, and robust gradient regularization. We adopt the Target-Matrix Optimization Paradigm to control geometric properties across the mesh, independent of the PDE of interest. To incorporate the accuracy of the PDE solution, we introduce error measures that control the finite element discretization error. The implicit dependence of these error measures on the mesh nodal positions is accurately captured by adjoint sensitivity analysis. Additionally, a convolution-based gradient regularization strategy is used to ensure stable and effective adaptation of high-order meshes. We demonstrate that the proposed framework can improve mesh quality and reduce the error by up to 10 times for the solution of Poisson and linear elasto-static problems. The approach is general with respect to the dimensionality, the order of the mesh, the types of mesh elements, and can be applied to any PDE that admits well-defined adjoint operators.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

37 extracted references · 29 canonical work pages

  1. [1]

    Cambridge University Press, ??? (2002)

    Deville, M., Fischer, P., Mund, E.: High-order Methods for Incompressible Fluid Flow. Cambridge University Press, ??? (2002)

    work page 2002

  2. [2]

    The International Journal of High Performance Computing Applications34(5), 562–586 (2020)

    Fischer, P., Min, M., Rathnayake, T., Dutta, S., Kolev, T., Dobrev, V., Camier, J.-S., Kronbichler, M., Warburton, T., ´Swirydowicz, K.,et al.: Scalability of high-performance PDE solvers. The International Journal of High Performance Computing Applications34(5), 562–586 (2020)

    work page 2020

  3. [3]

    Computers & Mathematics with Applications 81, 42–74 (2020) https://doi.org/10.1016/j.camwa.2020.06.009

    Anderson, R., Andrej, J., Barker, A., Bramwell, J., Camier, J.-S., Cerveny, J., Dobrev, V.A., Dudouit, Y., Fisher, A., Kolev, T.V., Pazner, W., Stowell, M., Tomov, V.Z., Akkerman, I., Dahm, J., Medina, D., Zampini, S.: MFEM: a modu- lar finite elements methods library. Computers & Mathematics with Applications 81, 42–74 (2020) https://doi.org/10.1016/j.ca...

    doi:10.1016/j.camwa.2020.06.009 2020

  4. [4]

    The International Journal of High Performance Computing Applications35(6), 527–552 (2021)

    Kolev, T., Fischer, P., Min, M., Dongarra, J., Brown, J., Dobrev, V., Warbur- ton, T., Tomov, S., Shephard, M.S., Abdelfattah, A.,et al.: Efficient exascale discretizations: High-order finite element methods. The International Journal of High Performance Computing Applications35(6), 527–552 (2021)

    work page 2021

  5. [5]

    Engineering with Computers 27(1), 41–50 (2011) 23

    Luo, X.-J., Shephard, M., Lee, L.-Q., Ge, L., Ng, C.: Moving curved mesh adap- tation for higher-order finite element simulations. Engineering with Computers 27(1), 41–50 (2011) 23

    work page 2011

  6. [6]

    SIAM Journal on Scientific Computing34(5), 606–641 (2012)

    Dobrev, V., Kolev, T., Rieben, R.: High-order curvilinear finite element methods for Lagrangian hydrodynamics. SIAM Journal on Scientific Computing34(5), 606–641 (2012)

    work page 2012

  7. [7]

    Computers & Fluids136, 48–66 (2016)

    Boscheri, W., Dumbser, M.: High order accurate direct Arbitrary-Lagrangian- Eulerian ADER-WENO finite volume schemes on moving curvilinear unstruc- tured meshes. Computers & Fluids136, 48–66 (2016)

    work page 2016

  8. [8]

    Computers & Fluids (2020)

    Dobrev, V.A., Knupp, P., Kolev, T.V., Mittal, K., Rieben, R.N., Tomov, V.Z.: Simulation-driven optimization of high-order meshes in ALE hydrodynamics. Computers & Fluids (2020)

    work page 2020

Show all 37 references
  1. [9]

    Journal of Computational and Applied Mathematics50(1), 67–83 (1994) https://doi.org/10.1016/0377-0427(94)90290-9

    Verf¨ urth, R.: A posteriori error estimation and adaptive mesh-refinement tech- niques. Journal of Computational and Applied Mathematics50(1), 67–83 (1994) https://doi.org/10.1016/0377-0427(94)90290-9

    1994 doi

  2. [10]

    Oxford University Press, ??? (1994)

    Baines, M.J.: Moving Finite Elements. Oxford University Press, ??? (1994)

    1994

  3. [11]

    Contemporary mathematics383(8), 141–173 (2005)

    Tang, T.: Moving mesh methods for computational fluid dynamics. Contemporary mathematics383(8), 141–173 (2005)

    2005

  4. [12]

    Huang, W., Russel, R.D.: Adaptive Moving Mesh Methods, p. 434. Springer, ??? (2011). https://doi.org/10.1007/978-1-4419-7916-2

    2011 doi

  5. [13]

    Dobrev, V.A., Knupp, P., Kolev, T.V., Mittal, K., Tomov, V.Z.: HR-adaptivity for nonconforming high-order meshes with the Target-Matrix Optimization Paradigm. Eng. Comput. (2021) https://doi.org/10.1007/s00366-021-01407-6

    2021 doi

  6. [14]

    International Journal for Numerical Methods in Engineering61(1), 1–21 (2004) https://doi.org/10.1002/nme.1052

    Thoutireddy, P., Ortiz, M.: A variational r-adaption and shape-optimization method for finite-deformation elasticity. International Journal for Numerical Methods in Engineering61(1), 1–21 (2004) https://doi.org/10.1002/nme.1052

    2004 doi

  7. [15]

    International Journal of Fracture 147(2007) https://doi.org/10.1007/s10704-007-9143-9

    Scherer, M., Denzer, R., Steinmann, P.: Energy-based r-adaptivity: a solution strategy and applications to fracture mechanics. International Journal of Fracture 147(2007) https://doi.org/10.1007/s10704-007-9143-9

    2007 doi

  8. [16]

    Journal of computational physics229(8), 2866–2897 (2010)

    Loseille, A., Dervieux, A., Alauzet, F.: Fully anisotropic goal-oriented mesh adap- tation for 3d steady euler equations. Journal of computational physics229(8), 2866–2897 (2010)

    2010

  9. [17]

    In: AIAA Scitech 2021 Forum, p

    Park, M.A., Balan, A., Clerici, F., Alauzet, F., Loseille, A., Kamenetskiy, D.S., Krakos, J.A., Michal, T.R., Galbraith, M.C.: Verification of viscous goal-based anisotropic mesh adaptation. In: AIAA Scitech 2021 Forum, p. 1362 (2021)

    2021

  10. [18]

    Computer-Aided Design145, 103187 (2022) 24

    Wallwork, J.G., Barral, N., Ham, D.A., Piggott, M.D.: Goal-oriented error esti- mation and mesh adaptation for tracer transport modelling. Computer-Aided Design145, 103187 (2022) 24

    2022

  11. [19]

    SIAM Journal of Scientific Computing41(1), 50–68 (2019)

    Dobrev, V.A., Knupp, P., Kolev, T.V., Mittal, K., Tomov, V.Z.: The Target- Matrix Optimization Paradigm for high-order meshes. SIAM Journal of Scientific Computing41(1), 50–68 (2019)

    2019

  12. [20]

    Engineering with Computers38(4), 3721–3737 (2022) https://doi.org/10.1007/ s00366-021-01407-6

    Dobrev, V., Knupp, P., Kolev, T., Mittal, K., Tomov, V.: hr-adaptivity for non- conforming high-order meshes with the target matrix optimization paradigm. Engineering with Computers38(4), 3721–3737 (2022) https://doi.org/10.1007/ s00366-021-01407-6

    2022

  13. [21]

    Journal of Computational Physics474, 111808 (2023)

    Camier, J.-S., Dobrev, V., Knupp, P., Kolev, T., Mittal, K., Rieben, R., Tomov, V.: Accelerating high-order mesh optimization using finite element partial assembly on GPUs. Journal of Computational Physics474, 111808 (2023)

    2023

  14. [22]

    Computers in Industry 25(1), 3–13 (1994) https://doi.org/0166-3615(94)90028-0

    Hsu, Y.-L.: A review of structural shape optimization. Computers in Industry 25(1), 3–13 (1994) https://doi.org/0166-3615(94)90028-0

    1994

  15. [23]

    Structural and Multidis- ciplinary Optimization66(2023) https://doi.org/10.1007/s00158-023-03684-9

    Swartz, K.E., Mittal, K., Schmidt, M., Barrera, J.L., Watts, S., Tortorelli, D.A.: Yet another parameter-free shape optimization method. Structural and Multidis- ciplinary Optimization66(2023) https://doi.org/10.1007/s00158-023-03684-9

    2023 doi

  16. [24]

    International Journal for Numerical Methods in Engineering86, 765–781 (2011) https://doi.org/10.1002/nme.3072

    Lazarov, B.S., Sigmund, O.: Filters in topology optimization based on Helmholtz- type differential equations. International Journal for Numerical Methods in Engineering86, 765–781 (2011) https://doi.org/10.1002/nme.3072

    2011 doi

  17. [25]

    The International Journal of High Performance Computing Applications38(5), 447–467 (2024)

    Andrej, J., Atallah, N., B¨ acker, J.-P., Camier, J.-S., Copeland, D., Dobrev, V., Dudouit, Y., Duswald, T., Keith, B., Kim, D.,et al.: High-performance finite elements with MFEM. The International Journal of High Performance Computing Applications38(5), 447–467 (2024)

    2024

  18. [26]

    Allaire, G., Jouve, F., Toader, A.M.: Structural Optimization Using Sensitivity Analysis and a Level-set Method vol. 194, pp. 363–393 (2004). https://doi.org/ 10.1016/j.jcp.2003.09.032

    2004 doi

  19. [27]

    Springer, Berlin (2004)

    Bendsoe, M.P., Sigmund, O.: Topology Optimization: Theory, Methods and Applications, 2nd edn. Springer, Berlin (2004)

    2004

  20. [28]

    In: 2018 AIAA Aerospace Sciences Meeting, p

    Zahr, M.J., Persson, P.-O.: An optimization based discontinuous Galerkin approach for high-order accurate shock tracking. In: 2018 AIAA Aerospace Sciences Meeting, p. 0063 (2018)

    2018

  21. [29]

    Cambridge University Press, ??? (2007)

    Braess, D.: Finite Elements: Theory, Fast Solvers, and Applications in Solid Mechanics. Cambridge University Press, ??? (2007)

    2007

  22. [30]

    Computer Methods in Applied Mechanics and Engineering 50(3), 231–261 (1985) https://doi.org/10.1016/0045-7825(85)90095-7 25

    Delfour, M., Payre, G., Zol´ esio, J.-P.: An optimal triangulation for second-order elliptic problems. Computer Methods in Applied Mechanics and Engineering 50(3), 231–261 (1985) https://doi.org/10.1016/0045-7825(85)90095-7 25

    1985 doi

  23. [31]

    International journal for numerical methods in engineering24(2), 337–357 (1987)

    Zienkiewicz, O.C., Zhu, J.Z.: A simple error estimator and adaptive procedure for practical engineering analysis. International journal for numerical methods in engineering24(2), 337–357 (1987)

    1987

  24. [32]

    Technical Report LLNL-TR-795097, Lawrence Livermore National Lab.(LLNL), Livermore, CA (2019)

    Knupp, P.: Target formulation and construction in mesh quality improvement. Technical Report LLNL-TR-795097, Lawrence Livermore National Lab.(LLNL), Livermore, CA (2019)

    2019

  25. [33]

    Technical Report LLNL-TR-817490, Lawrence Livermore National Lab.(LLNL), Livermore, CA (United States) (2020)

    Knupp, P.: Metric type in the target-matrix mesh optimization paradigm. Technical Report LLNL-TR-817490, Lawrence Livermore National Lab.(LLNL), Livermore, CA (United States) (2020)

    2020

  26. [34]

    Partial Differential Equations in Applied Mathematics, 100390 (2022)

    Knupp, P.: Geometric parameters in the target matrix mesh optimization paradigm. Partial Differential Equations in Applied Mathematics, 100390 (2022)

    2022

  27. [35]

    Computer-Aided Design 158, 103499 (2023)

    Barrera, J.-L., Kolev, T., Mittal, K., Tomov, V.: High-order mesh morphing for boundary and interface fitting to implicit geometries. Computer-Aided Design 158, 103499 (2023)

    2023

  28. [36]

    SIAM Journal on Optimization12, 555–573 (2002) https://doi.org/10.1137/S1052623499362822

    Svanberg, K.: A class of globally convergent optimization methods based on con- servative convex separable approximations. SIAM Journal on Optimization12, 555–573 (2002) https://doi.org/10.1137/S1052623499362822

    2002 doi

  29. [37]

    https://arxiv.org/abs/2506.00746 26

    Andrej, J., Kolev, T., Lazarov, B.: Scalable Analysis and Design Using Automatic Differentiation (2025). https://arxiv.org/abs/2506.00746 26

    2025 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.