Skip to content

rheoMesh — AMR literature survey (2026-10-28)

English | Japanese

« rheoMesh

Survey for the next steps (c) and (d) of 2026-10-28 on the development log.

For (c) and (d), papers, theses and public method descriptions on AMR were surveyed. No source code of other codes was read (clean room). Every reference was checked against its DOI or the publisher's record; the main ones are listed under References.

What the literature says

  1. Indicators that need no target value are mature. Examples:
  2. h|ω|/U (Popinet 2003; Kamkar et al. 2011);
  3. LES quality indices from the resolved energy fraction (Pope 2004, M; Celik et al. 2005, IQ);
  4. the small-scale energy fraction from the statistics of a preliminary run (Naddei et al. 2019; Antepara et al. 2015);
  5. dissipation from time-averaged fields (Daviller et al. 2017).

For LBM, the departure of the populations from equilibrium (Thorimbert et al. 2022) is reported to beat vorticity and Q. On the FV side the residual estimate (Jasak & Gosman 2001, 2003) is the best calibrated. 2. Imano et al., parts 2 and 3, agree with the later literature. The normalised threshold corresponds to Dörfler's (1996) bulk marking and to Löhner's (1987) normalised indicator with refine/coarsen hysteresis. The upstream stream tube of NUAMR is the same idea as Spalart's (2001) DES focus region: the set of points from which "a particle can return to very near the body". 3. The standard for LBM AMR is a 2:1-balanced, block-structured forest of octrees. Examples are p4est (Burstedde et al. 2011), Schornbaum & Rüde (2018) and, on GPUs, Jaber et al. (2025). Mapping the populations between levels is settled (Rohde et al. 2006; Chen et al. 2006). Noise when vortices cross level interfaces is still an open problem (Gendre et al. 2017; Astoul et al. 2020). 4. Quasi-static AMR is the norm for statistically stationary bluff-body wakes. The mesh is refined from the statistics of a short preliminary run and then frozen (Naddei et al. 2019; Daviller et al. 2017; Wackers et al. 2017). Per-step re-adaptation is used where the feature moves, such as free surfaces. 5. Stopping criteria. The literature stops on: - a cell budget; - a fixed point, where the mesh stops changing (Alauzet & Loseille 2016); - grid convergence of a few observables, with uncertainty estimates (Roache 1994; Eça & Hoekstra 2014).

These answer the future work stated in part 2: when to stop AMR from the estimated error. 6. No published work was found that uses an LBM-derived refinement map for an FV RANS mesh. The nearest is LES grid sizing from a RANS precursor (Addad et al. 2008). For DrivAer, Irigaray et al. (2024) is a solution-adaptive study. The AutoCFD 2 DrivAer summary (Hupertz et al. 2022) gives an external reference band.

Recommendations for rheoMesh

What Main sources
R1 In each uniform-start LBM cycle, take the window means of h|ω|/U, the small-scale energy fraction (or Pope's M) and the LBM non-equilibrium sensor. Normalise each by its mean and standard deviation, flag above ε = 1, and refine where any flag is set part 2; Popinet; Kamkar et al.; Naddei et al.; Pope; Thorimbert et al.
R2 Apply smoothing and buffers only inside the NUAMR stream tube. Set the downstream limit of the wake from Spalart's focus/departure split part 3; Spalart; Berger & Colella
R3 Use the normalised threshold as a floor, combined with Dörfler bulk marking (θ = 0.5). Apply two-threshold hysteresis between refinement and coarsening. Use the per-volume form of the indicator Dörfler; Löhner; Jasak & Gosman
R4 Do not refine further where the resolved energy fraction is above 0.8 (off-wall only) Pope; Celik et al.
R5 Run 2–3 quasi-static cycles, then freeze. Stop when the indicator and the map stop changing, and the change of Cd, recirculation length and spectra is below the wake-state scatter Naddei et al.; Wackers et al.; Alauzet & Loseille; Roache; Eça & Hoekstra
R6 Set the layers as follows, then correct per face with EREE: first layer from the y+ band; count from a growth ratio of 1.25 or less and the log-layer span; total height from the boundary-layer thickness estimate part 2; Spalart; Chapman; Choi & Moin; Jasak & Gosman
R7 In the frozen LBM mesh, move level interfaces one ring out of high-shear blocks Gendre et al.; Astoul et al.
R8 No directional (anisotropic) AMR for now. If ever needed, give the blocks, not the leaves, a level per direction Ham et al.; Keats & Lien; Wackers et al.
R9 Validate the LBM-to-FV transfer with the FV residual estimate, plus one coarsening pass after the transfer part 1; Addad et al.

Decided (2026-10-28)

All proposed settings were adopted: - statistics windows of 1.5 s for cycles 1–2 and 3 s for the last cycle; - θ = 0.5, ε = 1.0; - stream-tube target: the body plus the near wake up to one body height; - resolved-fraction cap of 0.8, applied at least two blocks from walls; - coarsening allowed from cycle 2; finest level L3; - stopping observables: Cd, recirculation length and probe spectra, with the scatter floor measured on the uniform start; - layer count from the LBM y+ in every cycle, corrected with EREE in the FV production step; - one FV coarsening pass with the residual estimate after the transfer; - external references: the experiment (0.243–0.252) and the AutoCFD 2 Cd band.

Both tracks are pursued: FVM-only AMR and AMR assisted by LBM. On the FVM-only track, a short URANS from the steady solution provides the unsteadiness, which is compared with the LBM u'rms. The iteration-to-iteration scatter of SIMPLE with large relaxation factors is compared as well. That scatter is not a physical-time fluctuation and also appears at poor cells, so it is used only as an auxiliary marker.

References

  • M. Imano, M. Kamata, Y. Sakamoto: Automatic Cartesian grid generation for CFD analysis of wind environment — comparison of error estimation methods in finite volume CFD analysis (part 1), J. Environ. Eng. AIJ 72 (612), 51-57, 2007 (in Japanese). J-STAGE
  • M. Imano, N. Ohnishi, M. Kamata, Y. Sakamoto: Automatic Cartesian grid generation for CFD analysis of wind environment — automatic generation of cost-effective grid using adaptive mesh refinement (part 2), J. Environ. Eng. AIJ 72 (617), 31-38, 2007 (in Japanese). J-STAGE
  • M. Imano, N. Ohnishi, M. Kamata, Y. Sakamoto: Automatic generation of orthogonal grids for CFD analysis of wind environment around buildings — development of a non-uniform type adaptive mesh refinement method (part 3), J. Environ. Eng. AIJ 73 (624), 207-214, 2008 (in Japanese). doi:10.3130/aije.73.207

  • M. Berger, P. Colella: Local adaptive mesh refinement for shock hydrodynamics, J. Comput. Phys. 82, 64–84, 1989. doi:10.1016/0021-9991(89)90035-1

  • R. Löhner: An adaptive finite element scheme for transient problems in CFD, Comput. Methods Appl. Mech. Engrg. 61, 323–338, 1987. doi:10.1016/0045-7825(87)90098-3
  • W. Dörfler: A convergent adaptive algorithm for Poisson's equation, SIAM J. Numer. Anal. 33, 1106–1124, 1996. doi:10.1137/0733054
  • H. Jasak, A. D. Gosman: Residual error estimate for the finite-volume method, Numer. Heat Transfer B 39, 1–19, 2001. doi:10.1080/104077901460650
  • S. Popinet: Gerris: a tree-based adaptive solver for the incompressible Euler equations in complex geometries, J. Comput. Phys. 190, 572–600, 2003.
  • S. J. Kamkar, A. M. Wissink, V. Sankaran, A. Jameson: Feature-driven Cartesian adaptive mesh refinement for vortex-dominated flows, J. Comput. Phys. 230, 6271–6298, 2011. doi:10.1016/j.jcp.2011.04.024
  • S. B. Pope: Ten questions concerning the large-eddy simulation of turbulent flows, New J. Phys. 6, 35, 2004. doi:10.1088/1367-2630/6/1/035
  • I. B. Celik, Z. N. Cehreli, I. Yavuz: Index of resolution quality for large eddy simulations, J. Fluids Eng. 127, 949–958, 2005. doi:10.1115/1.1990201
  • O. Antepara, O. Lehmkuhl, R. Borrell, J. Chiva, A. Oliva: Parallel adaptive mesh refinement for large-eddy simulations of turbulent flows, Comput. Fluids 110, 48–61, 2015. doi:10.1016/j.compfluid.2014.09.050
  • F. Naddei, M. de la Llave Plata, V. Couaillier, F. Coquel: A comparison of refinement indicators for p-adaptive simulations of steady and unsteady flows using discontinuous Galerkin methods, J. Comput. Phys. 376, 508–533, 2019.
  • G. Daviller, M. Brebion, P. Xavier, G. Staffelbach, J.-D. Müller, T. Poinsot: A mesh adaptation strategy to predict pressure losses in LES of swirled flows, Flow Turbul. Combust. 99, 93–118, 2017. doi:10.1007/s10494-017-9808-z
  • J. Wackers et al.: Can adaptive grid refinement produce grid-independent solutions for incompressible flows?, J. Comput. Phys. 344, 364–380, 2017. doi:10.1016/j.jcp.2017.04.077
  • F. Alauzet, A. Loseille: A decade of progress on anisotropic mesh adaptation for computational fluid dynamics, Computer-Aided Design 72, 13–39, 2016. doi:10.1016/j.cad.2015.09.005
  • P. J. Roache: Perspective: a method for uniform reporting of grid refinement studies, J. Fluids Eng. 116, 405–413, 1994. doi:10.1115/1.2910291
  • L. Eça, M. Hoekstra: A procedure for the estimation of the numerical uncertainty of CFD calculations based on grid refinement studies, J. Comput. Phys. 262, 104–130, 2014. doi:10.1016/j.jcp.2014.01.006
  • P. R. Spalart: Young-person's guide to detached-eddy simulation grids, NASA/CR-2001-211032, 2001.
  • C. Burstedde, L. C. Wilcox, O. Ghattas: p4est: scalable algorithms for parallel adaptive mesh refinement on forests of octrees, SIAM J. Sci. Comput. 33, 1103–1133, 2011. doi:10.1137/100791634
  • F. Schornbaum, U. Rüde: Extreme-scale block-structured adaptive mesh refinement, SIAM J. Sci. Comput. 40, C358–C387, 2018. doi:10.1137/17M1128411
  • K. Jaber, E. E. Essel, P. E. Sullivan: GPU-native adaptive mesh refinement with application to lattice Boltzmann simulations, Comput. Phys. Commun. 311, 109543, 2025. doi:10.1016/j.cpc.2025.109543
  • M. Rohde, D. Kandhai, J. J. Derksen, H. E. A. van den Akker: A generic, mass conservative local grid refinement technique for lattice-Boltzmann schemes, Int. J. Numer. Meth. Fluids 51, 439–468, 2006. doi:10.1002/fld.1140
  • H. Chen et al.: Grid refinement in lattice Boltzmann methods based on volumetric formulation, Physica A 362, 158–167, 2006. doi:10.1016/j.physa.2005.09.036
  • F. Gendre, D. Ricot, G. Fritz, P. Sagaut: Grid refinement for aeroacoustics in the lattice Boltzmann method: a directional splitting approach, Phys. Rev. E 96, 023311, 2017. doi:10.1103/PhysRevE.96.023311
  • T. Astoul, G. Wissocq, J.-F. Boussuge, A. Sengissen, P. Sagaut: Analysis and reduction of spurious noise generated at grid refinement interfaces with the lattice Boltzmann method, J. Comput. Phys. 418, 109645, 2020.
  • Y. Thorimbert et al.: Local mesh refinement sensor for the lattice Boltzmann method, J. Comput. Sci. 64, 101864, 2022. doi:10.1016/j.jocs.2022.101864
  • Y. Addad, U. Gaitonde, D. Laurence, S. Rolfo: Optimal unstructured meshing for large eddy simulations, Quality and Reliability of Large-Eddy Simulations, ERCOFTAC Series 12, Springer, 2008.
  • O. Irigaray et al.: Adaptive mesh refinement (AMR) criteria comparison for the DrivAer model, Heliyon 10, e31966, 2024. doi:10.1016/j.heliyon.2024.e31966
  • B. Hupertz, N. Lewington, C. Mockett, N. Ashton, B. Duncan: Towards a standardized assessment of automotive aerodynamic CFD prediction capability — AutoCFD 2: Ford DrivAer test case summary, SAE Technical Paper 2022-01-0886, 2022. doi:10.4271/2022-01-0886