Structured Grid Generation for Turbo Machinery Applications using Topology Templates

Transcription

1 Structured Grid Generation for Turbo Machinery Applications using Topology Templates January 13th 2011 Martin Spel page 1

2 Agenda: R.Tech activities Grid Generation Techniques Structured Grid Generation Advantages and inconveniences Automating structured grids Topology templates Internal surfaces Topology design Conclusions page 2

3 R.Tech Engineering company specialized in CFD Based near Toulouse (2001) and near Amsterdam (2009) Specialized in Space and Aeronautical industry Need for high quality meshes for heat transfer applications Use of in-house developed CFD code for high Mach number chemical non-equilibrium flow and OpenFOAM(TM) for 'low speed' 2 computational clusters of 120 nodes each page 3

4 Grid Generation Techniques - Overview Cartesian octree Unstructured Tetrahedral Unstructured Hex Unstructured Polyhedral Block structured Structured Overset Hybrid: Unstructured Prisms + Tetrahedral Structured + Cartesian Cartesian octree + boundary layer Links to software are examples only, for a more complete survey : page 4

5 Grid Generation Techniques Cartesian octree Grid aligned with cartesian axis Near wall cut cells can be used to conform to boundaries Cells are refined by cutting in 8 (octree) based on refinement criteria (or 2/4 directional for more advances algorithms) Advantages: Fast grid generation (fully automatic) Easy computation of gradients Relatively easy to adapt to flow features Disadvantage: difficult to model boundary layer (large aspect ratio cells are needed to capture the monodirectional gradient) Not possible to align grid with flow features Hanging nodes page 5

6 Grid Generation Techniques Cartesian octree Coding : people.nas.nasa.g ov/~aftosmis/vki Example from our experimental tool Blizzard page 6

7 Grid Generation Techniques Unstructured tet Tetrahedral cells (Delaunay, advancing front) Quality triangles at the surface, needs some user interaction Advantages: Grid generation speed (but user interaction needed) Relatively easy to adapt to flow features Disadvantage: Difficult to model boundary layer (large aspect ratio cells are needed to capture the monodirectional gradient Cell count higher compared to hex Complex computation of gradients page 7

8 Grid Generation Techniques Unstructured tet Example of public domain tool is Netgen ( jku.at/netgen/) page 8

9 Grid Generation Techniques Unstructured poly General polyhedral cells (for example dual mesh of tet mesh) General polygons at the surface Advantages: Great flexibility in mesh handling Geometrical flexibility Could possibily resolve boundary layer issues Disadvantage: More difficult data structures, could be resolved using adapted libraries such as OpenFOAM (TM) Complex computation of gradients page 9

10 Grid Generation Techniques Unstructured hex All hexahedral cells Quality quad cells at the surface, needs user interaction Advantages: Higher quality Geometrical flexibility Capturing of boundary layer flows with high aspect ratio cells Disadvantage: Harder to generate than unstructured tet meshes or polyhedral meshes page 10

11 Grid Generation Structured Multiblock Hexahedral cells structured in IJK space Each block (IxJxK cells) can be seen as an unstructured cell in an unstructured hex mesh Advantages: Geometric flexibility due to unstructured approach on block level Orthogonal meshes able to capture boundary layer flows Computational codes can make use of the structure to solve the equations more efficiently Reduction in cell number Disadvantage: Relatively hard to generate the mesh page 11

12 Grid Generation Which approach?? The best approach depends on the level of physics to be modelled, so application dependent High level physics (boundary layer heat transfer, turbulence modelling such as LES) requires high quality meshes: Orthogonal at walls Orthogonal in space Smooth (no sudden jumps in cell size) Support for high aspect ratio cells to capture monodimensional gradients page 12

13 Automating structured grid generatioon Automation can be accomplished by 'recording' the user action during the building of the grid, but this option is not general enough We opt for the creation of an abstract topology model, which is independent of the underlying geometry The link between topology space and geometry is made by 'surface assignments' Create a topology once for a specific class of geometry, and apply it to different (but similar) geometry page 13

14 Automating structured grid generation Topological description using nodes and edges: page 14

15 Automating structured grid generation General position in space for topology description page 15

16 From topology space to 3D space (1) 4 loop corner makes a face: patch in 3D space loop of 6 faces makes a block: IJK domain in 3D space Surface assignments: One node assigned to surface node will be projected on surface in the final mesh Two nodes connected by edge assigned to surface edge will be projected to a surface curve Four nodes loop assigned to surface face will be projected to a surface patch Points are free to move anywhere they like on the surface important for optimization of the mesh page 16

17 From topology space to 3D space (2) Node assigned to one surface free to move on surface Node assigned to two surfaces free to move on the curve describing the intersection of the two surfaces Node assigned to three surfaces fixed in space page 17

18 Advantages One topology, different geometries page 18

19 Internal surfaces The grid smoothing process smooths volume and surface mesh simultaneously resulting in high quality mesh To control the positioning of face boundaries, internal surfaces are introduced Same assignment rules as regular surfaces Allows 'encapsulating' the grid in specific regions Allow the precise capturing of sharp edges (however additional work in the internal surface definition is needed, for which tools are currently under development) page 19

20 Internal surfaces page 20

21 Topology design Topology construction tools allow to experiment with topology design. Topology design: 'Wrapping' of surfaces to avoid propagation of boundary layer grid into flow field Traditional edge type topology GridPro topology page 21

22 Topology design Topological refinement (could be recursive) page 22

23 Topology design topological corner block on 180 boundary page 23

24 Topology design Surface singularities page 24

25 Topology design Surface singularities : removed using outside wrap page 25

26 Topology design 3 topological corners at 180 surface page 26

27 Topology design Surface singularities page 27

28 Topology design Surface singularities : removed using inside wrap page 28

29 Topology design Internal singularities : page 29

30 Topology design Internal singularities : removed using internal wrap page 30

31 Topology design periodic staggering page 31

32 Topology design w/o periodic staggering page 32

33 Topology design periodic staggering page 33

34 Topology construction tools recursive refinement on block level : automated tool to go from coarse to fine block density, or from fine to coarse page 34

35 Topology construction tools page 35

36 Topology construction tools page 36

37 Hybridization structured / cartesian cut cells Mixture of high level and low level physics Example for helicopter aerodynamics using actuator disk: Actuator disk requires modeling of high gradient region perpendicular to disk: structured high aspect ratio grid Effect of helicopter on ground infrastructure: Detailed modeling of helicopter boundary layer is not required: cartesian cut cells Detailed modeling of ground effects is required: structured grid near ground / infrastructure page 37

38 Hybridization structured / cartesian cut cells A high quality structured mesh is created first. The geometry needing less quality is then integrated by refining and cutting cells. page 38

39 Conclusion High quality block structured grid generation can be automated to a large extend using the topology templates Topology construction tools allow the rapid construction of the most appropriate topology design Hybridization of block structured grids and other approaches can be of interest for certain applications page 39