Sie sind auf Seite 1von 48

| | 

  

Dr. K. P. Sinhamahapatra
Aerospace Engineering Department
IIT Kharagpur
 | 
° A mesh/grid is an artificial geometric
construction that facilitates the spatial
discretization of the governing equations to
be solved.

° The mesh determines the locations in the


field where the variables will be evaluated
and the stencil of the discrete equations.
  

° The final accuracy and efficiency of any


numerical solution are highly dependent on the
particular meshing strategy and mesh density
distribution employed.
° A good matching of the strengths and
weaknesses of the grid generation and flow
solution techniques and a strong and favourable
interplay between the two is the key to an
efficient overall numerical solution.
K   
°    Physical location of any mesh
point and the identity of its neighbours are known
implicitly. Physical locations may have to be stored.
°    Physical location of a mesh
point and the identity of its neighbours, i.e., the
connectivity of the mesh are to be determined
explicitly.
°   A combination of the two above.
° |   A set of disconnected points
distributed throughout the field.
  
° K  esh generation is trivial. The
grid points and their connectivity are known
implicitly. ethods can be extended to complex
geometries using cut-cell approach.
°  rid lines/surfaces conform
to the boundary lines/surfaces. A warped or
mapped Cartesian-type mesh where the
boundaries of the mesh coincide exactly with the
the boundaries of the physical domain. Physical
location of the mesh points must be stored but the
identity of the neighbours known implicitly.
    K
°   ultiple overlapping grids
to discretize the domain, the solver
interpolates values between the various grids
in the regions of overlap.

°     The domain is


decomposed into a number of topologically
simpler domains and each domain is meshed
independently with a structured grid.
Single block structured mesh about a wing configuration
An overset grid for a complex geometry
A multi-block structured grid
  | | 

°
   eometric data of the
Cartesian coordinates in the interior of a domain
are generated from specified values at the
boundaries through interpolations or specific
functions of the curvilinear coordinates.
°   apping by solving
PDEs with the dependent and independent
variables being the physical domain coordinates
and transformed computational domain
coordinates respectively.

  
°    utilize tensor
products of unidirectional FE interpolation
functions (Lagrangian, Hermite or Spline) for
two or three dimensions.

  rA 
p    r 
  r  , rp 1,2,3, ,, 1,2
  p 


p  rA, , p, rp 1,2,3, 1,....,8

   K

°     tensor


products of unidirectional interpolation
but with all sides of the boundaries
interpolated and matched. The corner
nodes are also matched.
°  (2-dimensions)
1. Pick four points on ´ which are identified
as the images of the four corners of Ë.
 K
2. The resulting four curve segments are identified as
the graphs of the four vector valued functions F(0,´),
F(1,´), F(´,0) and F(´,1) the 4 segments of the
physical boundary are images of the 4 sides of the
computational domain.
3. A bilinearly blended transfinite function U(´, ´) is
constructed using (Boolean sum projection) the four
F functions that maps the boundary of the
computational domain to that of the physical
domain.
4. Check for univalency criteria nonsingular Jacobian
 K
° The univalent function v : U U matches  on
the boundary of and interpolates to  at a
finite set of points.

  
rA r  rA r r rA rA r 
rA r 1 rA 1 r0, 2 rA 2 r1,
r rA 1 r 1 rA,0 2 r 2 rA,1
rA r  rA  r 
Physical domain

Transformed computational
domain

Transfinite interpolation
FN match the function at four corners but not
on all boundaries
Parameterization for 2D C-type structured grid
 
°  | |  Solution of a set of
elliptic PDE, (Laplace or Poisson equations)

r2
Ö  AA 2rA Ö A Ö AÖ
2
Ö r
2
A  ÖÖ  r
A Ö
2
A
2

r2
Ö Ö2 AA 2rA Ö A Ö AÖ r
2
A A2 ÖÖ  2 r
A Ö

Iterative solution in the computational domain to


determine the grid coordinates (x,y).
1. Smooth grid point distribution
2. Orthogonality at boundaries
3. Desired clustering using appropriate
control functions P and Q
4. Construction of the control functions
is often difficult
5. Larger computational time
6. ost widely used
Hyperbolic rid enerator
° Applicable to open domain problem
° Computationally efficient and less expensive
marching type solution
° Inability to match prescribed point distribution
on all boundaries
° Hyperbolic PDE for constraints of orthogonality
and cell volume/arc length
 A Ö  A Ö  0 
 p    A  Ö   A Ö   
2 2 2 2 2
  A  A  Ö   r 
Ö
Treatment of doubly and multiply connected domain for O type grid

Treatment of doubly and multiply connected domain for O-


type grid
     
|   
   | 
°  Union of ordered point sets that
define multiple cross-sections of the geometry.
Inaccurate and ambiguous form of surface
discretization. eometry details like small
gaps, slope and curvature continuity not
preserved.
°  eometry definition by a set of 3 or 4
sided curved surface patches and trimmed
surfaces.
Approximation of a surface
with hole by two patches and
by a single trimmed surface

° Surface Repair Removal of unrealistic
gaps, discontinuities and small overlaps
created by the CAD packages modified
input geometry.
° Projection Surface The surface grid is
constructed on a projection surface which is
then placed over the collection of surface
patches that defines the actual geometry.
     

° Physical space approach grid points must


coincide with the actual surface and need to be
determined from the actual surface geometry.
° Parametric space approach 2D meshing
problem. To be mapped back to physical space.
Possibility of invalid physical surface mesh for
highly warped surface or irregular
parameterization. lobal or quilted patches
solely for meshing.
  | 
° The governing equations are

1
r 22,11  
11 ,22 2 
12 ,12 
1 ,1 
2 ,2 r1
1  2
 
2

 A   A , 11  2
A  A
2
  2
A, 11  2
 2
  2
,
 
12  A   A   A  , ,1  A , A , A  , ,2   ,  ,  
  
,11  AA , AA , AA  , ,22   ,  ,  , ,12  A , A , A 
1 2
 ,   p p   
1 2

  | 
° Construction of curves on the surfaces and
surface patches using appropriate basis
polynomials and control vectors NURBS are
most widely used.
° Union of the patches is the global surface.
° For valid mesh the curves bordering each
patch are to be meshed the same way in all
patches containing them.
° esh each patch, parametric space preferred.
Structured surface grid on the top surface of a generic hyperplane
Structured surface grid on the bottom surface of the hyperplane
Surface patches created on a hypersonic vehicle for unstructured grid generation

  
° esh point movement or mesh redistribution
structure and connectivity preserved.
1. Spring analogy each edge a spring, stiffness depends
on the quantity to be minimized.
2. Variational principle minimization functional
containing various solution based criteria as well as
grid quality criteria simultaneously.
3. Control functions modified to produce clustering
based on solution gradients or truncation errors.
° esh enrichment addition of extra vertices,
structure and connectivity lost.
  | 
° Requirement of structure in the mesh removed
offering increased flexibility.
° Nodes numbered in any order, and have arbitrary
number of neighbours.
° Arbitrary but homogeneous connectivity Ú Single
data structure for the entire mesh unlike block
structured mesh.
° Adaptive meshing is easy to implement
° Algorithms closely tied to computational
geometry.
  | 
° Elements are generally triangles and
tetrahedrons but need not be.
° Two most prevalent mesh generation
approaches.

!
   
"     

   
° Initial Front union of the edges that discretize the
geometry boundary. This front advances out into the
field. A stack or priority queue.
° Selecting an edge from this list, a new point is created
based on specified criteria so that an optimal triangle
is formed.
° Updating the front by removing the current edge
and adding the two newly created edges depending on
their visibility.
° Process terminates when the stack (front) is empty.
Advancing Front Concept. Dotted line is
the initial front
 
O Field points are created to produce
triangles of optimal shape and size.
1. Specification of parameters
2. Field function or distribution function
3. Background grid
4. Interpolation
5. Cross-over (intersection) with other edges
6. Smaller edge or angle later
7. Smooth variation of triangle sizes
 #

1. Initial front is the surface grid (2D triangular


mesh on the boundary surfaces.
2. New points ahead of the front to form
tetrahedra.
3. Both edge-edge and edge-face intersection
check.

$   %& 


 ' (   
   
   
° Triangulation of a set of points using Delaunay
criterion *p    p  p 
 pp ppp ppp

O Unique triangulation (in 2D)


O ore efficient than AF
O Boundary integrity lost, boundary to be
recovered
O ax-min property maximizes the
smallest angle in the triangulation.
     

O Predetermined mesh points put in a list.


O Initial triangulation of just a few triangles
to completely cover the domain to be
gridded.
O esh points inserted sequentially into the
existing triangulation
     

O    into an existing


triangulation is locating and deleting all
existing triangles whose circumcircle
contain the inserted point. A new
triangulation is then constructed by
joining the new point to all boundary
vertices of the cavity created. (Bowyer-
Watson Algorithm)
Point insertion technique

 

° An initial coarse triangulation


° A priority queue based on some triangle
parameter.
° Field distribution of the parameter as desired.
° Triangles in the queue are sequentially examined
and if required a point is inserted at the
circumcentre
° New triangles are put in the queue if not
acceptable.
  
° Not guaranteed if the domain is concave.
Edges or faces that define the boundary do
not form a subset of the triangulation.
° Boundary to be recovered by local
transformation (edge and face swapping)
and modifying the boundary point
resolution.
° Constrained triangulation.
Edge swapping process

Edge-face swapping
Breakthrough of boundary in Delaunay triangulation
Example of a 2D unstructured grid
A tetrahedral unstructured grid for a 3D geometry

Das könnte Ihnen auch gefallen