You are on page 1of 37

MECH3780

Fluid Mechanics 2 and CFD

Computation Fluid
Dynamics (CFD)
Lecture 3 - Grid Generation
Reference: CFD Lecture Notes by Assoc Prof Dr Kamarul Arifin Ahmad
Objectives

• To introduce grid generation techniques


• To give overview about geometry generation
techniques for CFD analysis
• To give some practical advice upon grid
generation for CFD analysis
Grid Generation
• The governing equations of fluid flow are formulated on
the basis of the continuum hypothesis
• It is impossible to consider all points in the continuum
and hence we have to select a subset of points at which
flow quantities can be calculated
• The combination of the selected points and connections
between points defines a mesh or grid on which
numerical solutions of the flow equations is obtained
General requirement
• The computational grid must discretise the
domain and all boundaries. This closely links
the grid to the geometry. The distribution of
grid points must adequately represent the
geometry of the region.
• The distribution of the grid points should
allow for an appropriate resolution of features
in the domain.
CFD Process
Traditional CFD analysis cycle time

40% 80% 90% 100%

Geometry Acquisition

Grid generation

Solver setting

Start to compute
Geometry Data
The data describing the geometry of the body about which
the flow is to be computed can be provided
in a variety of means:
-Coordinates of points
-Blueprint mechanical drawings
-CAD files
CFD analysis mainly needs the surfaces
that “see” the flow.
These surfaces need to be extracted from
the geometry data in some manner.
Geometry Data

 Need to be reasonable with the


level of geometric detail to be
included.

 Greater geometric detail leads


to more grid points and longer
computations.

 A sequential approach is best


in which the detail of the
geometry is increased as the
CFD analyses proceed.
Cleaning CAD drawing

 Eliminate duplicated entities


 Eliminate small details
 Eliminate components not exposed to
the flow
 Rebuild geometrical connectivity
between parts
Example of cleaning CAD drawing
CAD MESH
 Realistic geometry are TOO complicated to be
generated from “simple” shapes
 Engineering design is based on CAD system

Gambit Can import files from CAD


•ACIS (Andy, Charles, and Ian’s software
System) •Direct imported file such as from
•Normally used for modelling CATIA, Solid Edge
software technology •Using IGES, STEP, STL

Translation between CAD and CFD systems is a critical issue


Geometry Data
Availability of CAD files for defining the geometry provides needed
information in a ready form; however, issues are:
 CAD files usually contain much more information than is needed by the CFD analysis
(mechanical details). Can take considerable effort to extract out the curves and surfaces
that are needed.
 Many CAD packages are solid modelers. They can output surfaces; however, the “quilt” of
surfaces can have some strange trimmed surfaces that have gaps in their edges.
 CAD files are usually generated by mechanical designers that are mindful of tolerances and
clearances needed in the mechanical design; however, CFD desires no clearances or gaps.
CFD analyst may need to spend time to “close up” the surfaces.
 CAD file format. IGES file format is commonly used since it does well representing surfaces
and curves.
 Mechanical drawings contain all the fine features; however, CFD may not wish to capture
such geometric resolution. CFD analyst has to remove such features.
Mesh Families
Two basic mesh families:

Cartesian

Conformal
Types of conformal mesh

Three basic types of conformal mesh:

Structured (quadrilateral or hexahedral)

Unstructured (triangular or tetrahedral)

Hybrid (mixture of above)


Example of Conformal Mesh

Structured Mesh Hybrid Mesh


Structured/Unstructured Mesh

i,j i+1,j i,j i+1,j

1
3 Element Nodes
1
6 2 2 3 4 1 1 2 3
4 5
8
5
2 1 7 2
7 7 6

3 2 4 3
10 11
9
4 2 7 6
Mesh quality
Cell aspect ratio

Cell skewness

Cell growth rate

Alignment to the flow


Mesh quality
• Density : High density gives more accuracy but
computation time is higher
• Smoothness: Large variations in grid
density/shape can cause numerical diffusion
and dispersion/refreaction waves
• Shape: Boundary layers in fluid flow
The Ideal Mesh/Generator
Simple to use Can be used by a novice, but has the functionality to
please an experienced user.

Fast to generate CPU-time should be small.

Good boundary discretisation Accurate representation of the


geometry.

Good interior discretisation The element shape should be as regular


as possible and the element quality should be high. The
element size should vary smoothly.

Control over the mesh size It should be possible to easily control the
mesh size in the various regions of the domain.

Fast to compute on Solution convergence rates can depend on mesh


quality and type.
Mesh comparison

Structured Unstructured Hybrid


Time High Low Medium
Fast to High Slow and may not Medium
generate always work
Good Yes Possible but may be Excellent
boundary and not predictable
interior
discretisation
Fast to Fast Slow Medium
compute on
Structured Mesh
• Direct (Algebraic) and Indirect (PDEs)
•Fast & economical •Time consuming
•Precise Spacing Control •Distribution loss
•Interactive user interface •Inherent smoothness
•High degree of •Elliptical & Hyperbolic
understanding
•Orthogonality &
Smoothness
Example of Structured Mesh

Algebraic scheme Elliptical scheme


H-type of Structured Grid
C-type of Structured Grid

f g d e f g

a
c b

e d a b c b a

Physical Space Computational Space


O-type of Structured Grid

c d c

d b a c

a b a

Physical Space Computational Space


Exercise

Try to mesh the backward step model below:


Unstructured mesh
• Usually related to FEM. Requirement for UM:
– The mesh must achieve the standard of the mesh validity. For
example, the mesh must not contain any holes, without self-
intersection, and etc.
– The mesh must also conform to the boundary of the domain.
– The mesh’s density must be controllable in order to allow the trade-off
between accuracy and solution time.
– The grid density should be allocated accordingly to the local accuracy
requirements.
– The shape of the elements must be in the good quality criteria, as the
quality of the grid has the major influence in the convergence
characteristics of CFD methods.
Unstructured Mesh
• Some of the unstructured mesh method are the quadtree and octree
based grid generations, the advancing-front method, the Delaunay point-
insertion method, and other unstructured grid generation methods
techniques based on the advancing front and Delaunay triangulation.
Example of Quadtree method
Hybrid Mesh

• Structured mesh may be accurate but for


complex geometry, it is time-consuming or
even impossible.
• But unstructured mesh sometimes give
unacceptable accuracy & computing time is
very high.
• Solution: Hybrid (mixture of structured &
unstructured)
Hybrid Mesh
Multi block Grids
• Within each sub-region, generate a coordinate
system
• The overall coordinate system is forming by
joining these sub-systems together
• Very time-consuming. Can be 1 man-month
for complex configurations (full configuration
aircraft)
Example of multi block grid
Example of multi block grid
Preliminary Steps

Know your flow : Inviscid, Laminar, turbulence


Define the geometry and limits of the domain :
Simplification
Set the mesh resolution: do you wish to capture wakes,
boundary layers, shock waves, vortices, etc.?
Grid Adaptation

You might also like