Context and chronology Finite Element and Finite Volume. PDE. 70s: - - PowerPoint PPT Presentation

EPI Gamma 1996-2000-2004-2008 AND Gamma3 PROPOSAL

P.L. George & al.

Evaluation, 17-18-19 March 2009

Context and chronology. Gamma: 1996-2000-2004-2008. Gamma3 proposal. A few images.

Context and chronology

Finite Element and Finite Volume. PDE. 70s: First automatic mesh generation algorithms for arbitrary domains. 2D ADV (J.A. George, PhD Stanford). 1980: F . Hermeline PhD, P6. Delaunay 2D and 3D. 80s: First automatic tet meshers for arbitrary domains. “Octree” (RPI, Troy, NY), Delaunay based (MSU, MS and Princeton, NJ) and ADV (Imperial College and Swansea, UK). 1985: Thompson Warsi Mastin , “Numerical grid generation” (Elsevier). 1985: Shamos Preparata, “Computational Geometry” (Springer-Verlag). 1987: J. Peraire & al. “3D adaptive remeshing” (notion of a metric). 1989: R. L¨

• hner “3D adaptive remeshing” (quality meshing).

INRIA, EPI Modulef and Menusin, with P6 : Modulef library. 1990: P .L. G., “Automatic mesh generation” (Wiley). 1990: M.G. Vallet PhD, INRIA-P6. Anisotropic mesh generation 2D. 90s: INRIA designed (DRET grant) a Delaunay based tet mesher for arbitrary domains where the given surface mesh is preserved. 96: The Gamma EPI is created.

2 EPI Gamma

The Gamma EPI: 1996-2008. Main lines

About 10 persons including 5 permanent positions. Automatic mesh generation methods (3D tet-hex, 2D1/2 tri-quad, 2D tri-quad). Mesh optimisation, mesh modification, mesh visualization, definition of data structures, ... First periods, topics are 100% devoted to mesh generation algorithms and methods and related software components. > Emphasis on 3D domains (blocking point at this time). > Then surface mesh generation methods (under estimated topics in general). > Still 2D methods for advanced cases. All of that results in a suite of software components. Diffusion via INRIA (and UTT) then Simulog and nowadays Dist` ene.

3 EPI Gamma

The Gamma EPI: 1996-2008. Main lines

>>> 1996-2004 The know-how allows for numerical simulations within the EPI, other EPIs or made by industrial partners. This results in new challenges. >>> 2004-.... Indeed, it is no longer possible to work on mesh generation methods without considering the solver side and the concrete applications. Algorithms needed for mesh adaptation and adaptive computations require combining meshing methods, error estimators and solvers. This leads us to introduce a more important part in the solver side. It is no longer possible to look at the meshing side only, the risk being to miss the real life problems, to consider problems that are not really posed or purely academic cases.

3 EPI Gamma

The Gamma EPI: 1996-2008. Algorithms and theories

Main keywords : 2D anisotropy ==> surface (anisotropy cames from curvatures), 3D meshing (tets and hexes), 3D anisotropy (tets), Nanostructured materials, geometrical modelization, CAD surface modelization, topology, ridges, ... Metric and error estimate, Governed methods (size map or discrete metric field) (therefore ==> mesh adaptation), Adaptive computation schemes, CFD solvers (Euler). Main choices for basic algorithms : Speed, robustness, validation by means of classical or orthogonal choices: Simple (subtle) data structures, bucket, random, non-random, cache default, no C++ nor Matlab, no extended arithmetic, no external ressources, ...

4 EPI Gamma

The Gamma EPI: 1996-2008. Algorithms and theories

Theoretical frame : the notion of a “unit mesh”, related to the fondamental notion of a

• metric. An abstract definition of the “mesh quality” related to the

definition of adequate and application dependent metrics. error estimate based on the interpolation error (for various norms) and field interpolation with adequate properties. a theory about continuous meshes in multi-scale adaptation or goal

• riented adaptation (objective functional).

an octree dual based approach for hex meshing. extraction of CAD surface topology (skeleton). extraction of geometrical characteristics (lines) for parametric surface. a method for the geometric modelization for specific surfaces (nanostructures, molecules, ...).

4 EPI Gamma

The Gamma EPI: 1996-2008. Algorithms and theories

Validation, a primordial concern of Gamma : blind validation using our own Data Base (≈ 100 Kdata), real life problems on industrial cases or via diverse contracts (EEC, competitivity clusters, ....), via master, PhD or post-doc. Collaborations : EPIs Smash and Tropics at Sophia-Antipolis, EPI Macs at Rocquencourt, laboratoire JL2 at UPMC (P6), universit´ e de technologie de Troyes (UTT), universit´ e d’Ottawa, ´ ecole polytechnique f´ ed´ erale de Lausanne (EPFL). ONERA, Dassault Aviation (DA), Lectra, Lemma, Dist` ene and clusters or EEC contrats.

4 EPI Gamma

Software components, a key point of the EPI

GHS3D and sons, classical or governed (iso or aniso) tet mesher1, Hexotic, full hex mesher, BLSURF, parametric surface mesher2, BLMOL, surface mesher for molecules, BL2D, 2D mesher3, Wolf, Euler and Navier-Stokes solver, SonicBoom, sonic boom solver, Metrix, construction of metric fields from E.E. and operations about metrics, Interpol, field transfer from mesh to mesh, Shrimp, domain partitionner, specific renumbering (vertices, elements, ...), the suite PPxxx, diverse tools for domain partitionning , mesh intersection, mesh correction, medit, mesh and solution visualization. Transfers via Dist` ene or directly (INRIA and/or UTT). 1Abaqus, Akka, Alcan, Armines, Cea, Cetim, Cnrs, Cocreate, Coretech, Cst, Edf, Eta, Kias, Plassotech, Pointwise, Robobat, Samtech, Sharc, Honda, Simpoe, Simulation works, Snecma, Soliworks, Synopsis, Technostar, Transvalor, Wias, Williams, ... 2Cea, Cnrs, DA, Edf, Emw, Lectra, Onera, Simpoe, ... 3Alcatel, Safran/Snecma, Nippon Steel! ... 5 EPI Gamma

Main achievements

Books. 1996: P .L. G. and H. B. “Delaunay triangulation and meshing” (Hermes), 2000: P .F . and P .L. G. “Mesh generation” (Hermes), 2001: P .L. G. (eds), “Maillage et adaptation” (Hermes), 2008: P . F. and P .L. G. “Mesh generation” 2nd edition (ISTE and Wiley). PhD. 2002-2004: C. D. (PhD), 3D adaptation by means of local remeshing, 2004-2007: B. C. (PhD), parametric surface, cad, ridges extraction, 2005-2008: A. B. (PhD), nanostructured materials, geometric modelization, 2005-2008: A. L. (PhD), multi-scale adaptation and continuous meshes, 2006-....: E. R. (PhD), parametric surface, cad, topology, optimization, 2007-....: G. O. (PhD), ALE method. 6 EPI Gamma

Main achievements

Papers. 1997: H.B. & al., “Delaunay mesh generation governed by metric specifications”, 1 and 2, FEAD. 1997: P .L. G., “Improvement on Delaunay based 3D automatic mesh generator”, FEAD. 2000: H.B., P .L. G. and P . L., “Parametric surface meshing using a combined advancing front generalized Delaunay approach”, IJNME. 2002: P .L. G. and H.B., “Ultimate robustness in meshing an arbitrary polyhedron”, IJNME. 2008: Y. Bourgault, M. Picasso, F. Alauzet and A. Loseille, On the use of anisotropic error estimators for the adaptative solution of 3-D inviscid compressible flows, IJNMF . 2008: F. Alauzet, High-Order method and mesh adaptation for Euler equation, IJNMF . 2008: F. Alauzet, S. Borel-Sandou, L. Daumas, A. Dervieux, Q. Dinh, S. Kleinveld, Loseille, Y. Mesri and G. Rog´ e, Multi-model and multi-scale optimization strategies. Application to sonic boom reduction, EJCM. 2008: C. Bennis, H. B. and N. Flandrin, 3D conforming power diagrams for radial LGR in CPG reservoir grids, EWC. 2008: B. Cl´ emenc ¸on, H. Borouchaki and P . Laug, Ridge extraction and its application to surface meshing, EWC. 2008: P . Laug, H. Borouchaki, A. Benabbou and J. Lu, Mod´ elisation g´ eom´ etrique de structures granulaires, CRAS. 2008: T. Grosges, H. Borouchaki and D. Barchiesi, New adaptive mesh development for accurate near-field enhancement computation, JM. 2007: P .L. George, H. Borouchaki, P .J. Frey, P . Laug and E. Saltel, Encyclopedia of Computational Mechanics, ch. 17, Mesh Generation and mesh adaptivity. 2007: F. Alauzet, P . Frey, P .L. George and B. Mohammadi, 3D transient fixed point mesh adaptation for time-dependent problems: Application to CFD simulations, JCP . 2007: A. Dervieux, Y. Mesri, F. Alauzet, A. Loseille, L. Hascoet and B. Koobus, Continuous mesh adaptation models for CFD, CFDJ. 7 EPI Gamma

Main achievements

Papers (continued). 2006: F. Alauzet, X. Li, E. Seegyoung Seol and M.S. Shephard, Parallel anisotropic 3D mesh adaptation by mesh modification, EWC 2006: A. Cherouat, H. Borouchaki, K. Saanouni and P . Laug, Numerical methodology for metal forming processes using elastoplastic model with damage occurrence, JMST. 2005: P . Frey and F. Alauzet, Anisotropic mesh adaptation for CFD computations, CMAME. 2005: H. Borouchaki, P . Laug, A. Cherouat, K. Saanouni, Adaptive remeshing in large plastic strain with damage, IJNME. 2005: H. Borouchaki, J. Villard, P . Laug and P .L. George, Surface mesh enhancement with geometric singularities identification, CMAME. 2005: A. Cherouat, K. Saanouni, H. Borouchaki and P . Laug, Virtual metal forming with damage occurrence using adaptive remeshing, IJFP . 2004: P . Laug and H. Borouchaki, Curve linearization and discretization for meshing composite parametric surfaces, CNME. 2004: H. Borouchaki and P . Laug, Simplification of Composite Parametric Surface Meshes,EWC. 8 EPI Gamma

The Gamma3 proposal: the staff

Fr´ ed´ eric Alauzet, cr1, CFD, solvers, EE, metrics, ... Dominique Barchiesi (3), pr1, physics, electromagnetism, ... Houman Borouchaki (3), pr1, tetra, surface, ... Abel Cherouat (3), pr2, solid mecanics (structure), ... Maryse Desnous, ass., shared by other epis. Paul Louis George, dr1, tetra, 3D reconstruction, ... Laurence Giraud (3), mdc, optimization, ... Thomas Grosges (3), mdc, solver, electromagnetism, ... Patrick Laug, dr2, surface, nano, deformable meshes, ... Lo¨ ıc Mar´ echal, ing., hexa, // , multicores, ... Dominique Moreau (3), ing., hexa, visu, ... Thibaud Mouton, doct., reservoir, ... G´ eraldine Olivier, doct., ale, ... Longmin Ran, doc., bassin, ... Erwan Renaut, doct., surface, cad, ...

9 EPI Gamma

The Gamma3 proposal: topics

a touch of continuity (necessarily) qualitative and quantitative gaps completely new things quite (too?) ambitious the claim to give answers (or significative advances) a minima, one enrolment at INRIA and another one at UTT

10 EPI Gamma

The Gamma3 proposal: topics

Generic mesh algorithms. Geometric modelization. Large size meshes and related computer science methods. Adaptive calculation schemes and real life applications. strong anisotropy in 3D and related algorithms, difficult meshing problems for surfaces, geometry and topology, large size meshes and related algorithms (parallelism, use of the multicore architecture, ...) a major concern about nanostructures, geometric modelization and effective simulations, a more important implication about solvers, in CFD, structural mecanics, electromagnetism, energy and safety, in CFD, difficult cases with boundary layers, turbulence, multi-scale phenomena, ... problems with moving or deformable meshes.

10 EPI Gamma

The Gamma3 proposal: collaborations, ”competitors” and money

Collaborations. continue the present collaborations EPI Argsh, INRIA Bordeaux. EPI Nachos, INRIA Sophia. Montr´ eal Hong Kong ”Competitors” Meshing side. GMU (MSU, RPI). Solver side. None. Both sides. Cemef, GMU and RPI. ”Ressources (money)” Dist` ene DA, Andra, Cea, Ifp, Snecma EHPOC, HISAC2

11 EPI Gamma

Hex meshing.

• Gamma. Full hex
• Gamma3. Sharp regions, hex dominant + adaptation.

12 EPI Gamma

Skeleton of a CAD surface made up of ≈ 1000 patches and corresponding mesh.

• Gamma. Skeleton + BLSURF
• Gamma3. BLSURF in parallel

13 EPI Gamma

Uniform mesh of a wing taking care of a (ridge) crest line.

• Gamma. Ridge + BLSURF
• Gamma3. C2 modelization

14 EPI Gamma

• Gamma. 3D reconstruction.
• Gamma3. singularities (mechanical parts are not bunnies).

15 EPI Gamma

Nanostructured materials (Solid Mechanics).

• Gamma. Geometrical modelization (2D and 3D), meshing (2D).
• Gamma3. Meshing (3D), domain decomposition, diagrams, parallel and effective. simulation.

16 EPI Gamma

ALE method (CFD).

• Gamma. 2D moving meshes.
• Gamma3. 3D

17 EPI Gamma

Cylinder crushing (Solid Mechanics).

• Gamma. 2D BL2D deformable meshes.
• Gamma3. 3D complex geometry.

18 EPI Gamma

Part cuting (Solid Mechanics).

• Gamma. Preliminary tests for simple 3D deformable meshes.
• Gamma3. 3D complex.

19 EPI Gamma

Transsonic flow . Vortex capture up to 500 metres thanks to multi-scale adaptation. multi-´ echelle (CFD).

• Gamma. Euler adaptive (aniso) meshes
• Gamma3. N.S.

20 EPI Gamma

Supersonic flow. Accurate capture of various shocks up to 2 km. thanks to multi-scale adaptation (CFD).

• Gamma. Euler adaptive meshes.
• Gamma3. N.S. (metric gradation, boundary layers, multi-scale phenomena).

21 EPI Gamma

A few comparizons (Top left, funny CG 2D anisotropy!). 22 EPI Gamma

Comparing multi-scale adaptation and goal oriented adaptation (CFD).

• Gamma. Euler adaptive meshes (using the adjoint).
• Gamma3. N.S.

23 EPI Gamma

Nuclear wastes (safety).

• Gamma. Initial 3D surface and volume meshes.
• Gamma3. 3D adaptation.

24 EPI Gamma

Vertical illumination of a golden ball (cancer, electromagnetism, energy).

• Gamma. Preliminary adaptive tet + Finiet Element (vs F. Diff.)
• Gamma3. Complex geometry.

25 EPI Gamma