Title: Grids generation methods and adaptive meshes
1Grids generation methods and adaptive meshes
Pawel Cybulka Finite element method
2Plan presentation
- Grid generation
- mesh types,
- grids generation methods.
- Adaptive finite element method
- phadaptivity,
- error estymator,
- hierarchical grids.
3Mesh types
- Mesh types are varied as the numerical
methodologies they support, and can be classified
according to - conformality
- surface or body alignment
- topology
- element type.
4Conformality
- Conformal meshes are characterized by a perfect
match of edges and faces between neighbouring
elements. - Non-conforming meshes exhibit edges and faces
that do not match perfectly between neighbouring
elements, giving rise to so-called hanging nodes
or overlapped zones. - Figure 1. a) conforming mesh, b) non-conforming
mesh.
5Surface or body alignment
- Surface or body alignment is achieved in those
meshes whose boundary faces match the surface of
the domain to be gridded perfectly. If faces are
crossed by the surface, the mesh is denote as
being non-aligned. - Figure 2. a)surface aligned, b)non-surface aligned
6Mesh topology
- Mesh topology denote the structure or order of
the elements. There are three possibities - Micro-structured, each points has the same number
of neighbours . - Micro-unstrutured, each point can have arbitrary
number of neighboures - Macro-unstrutured, micro-structured, where the
mesh is assembled from groups of micro-structured
subgrids
7Element type
- Typical element types for 2D domains are
triangles and quads, and tetrahedra, prisms and
brick for 3D domains. - Figure 3. Element types
8Description of the domain to be gridded
- There are two possible ways of describing the
surface of a computational domain. - Using analytical functions. This is the preferred
choice if a CAD-CAM database exists for the
description of the domain. Typical data types
splines, B-splines, non-unifom rational B-splines
(NURBS) surfaces. Important characteristic of
this approach is that the surface is continuous,
there are no holes in the information. - Via discrete data. When we get a cloud of points
or an already existing surface triangulation
describes the surface of the computational
domain. Examples are remote sensing data, medical
imaging data, data sets from computer games. -
9Typical grid generation methods
- Structured mesh
- simple mappings
- multiblock.
- Unstructured mesh
- quadtree(2D) and octree(3D)
- the advancing front technique (AFT)
- Delaunay triangulation.
10Simple mappings
- The computational domain can be mapped into the
unit square or cube. The distribution of points
and elements in space is controlled either by an
algebraic function, or by solution of a partial
differential equation in the transformed space. -
- Figure 4. Structured meshes built on the base of
various coordinate system a) Cartesian
coordinate system, b) cylindrical system, c)
combination of various coordinate system
11Multiblock grid
- Multiblock grid is based on division of
difficult to discretization area into several
areas which are simpler to discretization and,
then proper connection of these areas. There are
some variations of this strategy including
overset method, patched multiblock, composite
multiblock. These methods differ depending on the
way of connection of the subareas into the whole. - Figure 5. Grids generated by various multiblock
methods a) overset mesh combination b) composite
mesh combination.
12Quadtree and octree mesh methods
- Quadtree and octree are a simple method where all
domain is mapping by quads(2D) or bricks(3D). In
the next step all quads containing the boundary
points are divided into four parts whereas bricks
are divided into eight parts. This process is
repeated by the moment when all boundary points
are closed in the least quads or bricks. The size
of the least quads is given by a user. In the
last step all quads are transformed into
triangles. - Figure 6. Scheme of grid generation by quadtree
method, a)quadtree grid after thicken, b) quatree
grid after division of quads into triangles.
13The advancing front technique (AFT)
- The principle of this method is based on the
so-called front created with the points located
on discretized boundary of domain. Properly
connected points form sides so that continuous
area boundary is replaced by a set of sides ( the
line segments in the case of 2D and triangles in
the case of 3D) creating a closed loop. Then,
elements are built in accordance with the
established direction in loop on the basis of
existing set of sides and possibly added points.
How will create another element (figure 7)
depends on a angle between two following sides
from the front.
14The advancing front technique (AFT)
- How to combine elements depending on a angle
- a lt 90 a new element is built (created from
existing points), - 90 lt a lt 120 a new point is added and two
elements are created, - 120 lt a a new point is added and one element is
created. - Figura 7. The principle of conduct during the
construction of the grid by AFM
15Delaunay triangulation
- Delaunay algorithm for triangulation starts by
forming the super triangle enclosing all the
points from set V that has to be triangulated.
Then, incrementally, a process of inserting the
points p into the set V is performed. After every
insertion step a search is made to find the
triangles whose circumcircles enclose p.
Identified triangles are then deleted from the
set. As a result, an insertion polygon containing
p is created. Edges between the vertices of the
insertion polygon and p are inserted and form the
new triangulation. - Figure 8. The Delaunay triangulation technique
16Convergence of FEM
- Using the FEM computation approximate results are
received. The accuracy of the approximation can
be computed using the formula - u ua lt Chp u
- u accurate solution,
- ua FEM solution ( approximate),
- h the size of elements,
- p the degree of approximation.
- Therefore the accuracy of the FEM solution
depends on the - size of elements,
- the degree of the function approximation.
17Influence of the number of elements and the
degree of approximation of functions on the
accuracy of the FEM calculations
18Adaptive finite element method
- The idea behind AFEM is to make local
hp-adaptivity based on local error analysis. - The aim is to obtain sufficient accuracy of the
result at the smallest computation cost.
19H - adaptivity
- H - adaptation is the process of changing the
concentration of elements in the area calculation
in order to change the accuracy of the
computation carried out there. - Typically, h-adaptation is associated with
thickening of areas of high variability of the
analyzed qualities by what the calculation error
is minimized in this area. - Figure 9. H-adaptation elements on the grain
boundaries.
20P - adaptivity
- The accuracy of the results obtained in the
domain increases or decreases by increasing or
decreasing the degree of approximation of
function of shape in the elements. - In the case of p-adaptivity we need to draw
attention to the proper connection of elements
with higher degree of approximation with
neighbouring elements with a lower degree of
approximation. Provided the correct computation
is the continuity of approximation. Therefore,
approximation of the function corresponding to
the side of element adjacent to the element with
a higher degree of approximation should be raised
to the same degree.
21Error estimation
- Error estimation is a way to evaluate the error
occurs in a given computation domain. - Error estimation includes a criterion defining
the degree of adaptivity that must be used in
order to obtain the assumed accuracy of
computation. - The criterion of adaptation Ei may be defined as
the second derivative normalized after medium
gradient test variable value. -
- Ui - test variable value
- cn depends on chosen algorithm for the
solution of the physical problem -
-
22Error estimation
- One of the simpler and more frequently used error
estimations is estimation as shown in Figure 10.
In the first step of the algorithm gradients
value in each element is computed. Then we
compare the values of the adjacent elements. If
the difference between neighbouring elements
exceeds the determined threshold the elements are
divided. - Figure 10. Schematics of a simple error estimator
-
23Hierarchical grids
- Hierarchical grids were formed for adaptive
finite element methods. Their structure
corresponds to all needs associated with
hp-adaptivity. - The algorithm of hierarchical grids reminds
quadtree (2D) and octree (3D) methods. - The starting point for hierarchical grids are
grids created by the grid generator. The elements
of this grid are called parents. Each of the
parent can be divided into proper number of
identical in terms of the shape children. Each
child can be a parent. Thus we have possibility
of any compacting the grid. - Each element in the hierarchical grid knows its
parent and its children by what the grid can be
easily thicken and thin.
24Hierarchical grid
- Figure 11. The schema of the hierarchical
structure of an adaptive numerical grid
25Summary
- The grids are a very important element in the
computation using finite element method. They
determine the accuracy of the computations
carried out and the time of their performance. - Adaptive finite element method based on
hierarchical meshes and local error estimation
enables to carry out of an approximation only in
these areas where it is required. Thanks to it
accurate results are obtainedat the smallest
cost computation.
26LITERATURE
- Löhner R., 2008, Applied Computational Fluid
Dynamics Techniques An Introduction Based on
Finite Element Methods, Second Edition, John
Wiley Sons, Ltd, Chichester. - Joe F. Thompson, Bharat K. Soni, Nigel P.
Weatherill, Handbook of Grid Generation, CRC
Press, 1998. - Banas K. Metoda Elementów Skonczonych, Seminarium
BJT CM UJ, 2006.
27Thanks for your attention