Вы находитесь на странице: 1из 67

Introduction to

Computational Fluid Dynamics

Experimental Vs

CFD

Measure the real world

Approximated solutions of
differential equations

May required complicated


instrumentations & rigs
Provides limited flow infor

Everything are done on a


computer
Provides details of flow
Motion manipulation device

kiel probes

Wind tunnel

Motivation
Modern engineers apply both experimental and CFD
analyses
the two complement each other
For example, engineers may obtain global properties, such
as lift, drag, pressure drop, or power, experimentally, but
use CFD to obtain details about the flow field, such as shear
stresses, velocity and pressure profiles, and flow
streamlines.
In addition, experimental data are often used to validate
CFD solutions by matching the computationally and
experimentally determined global quantities.
CFD is then employed to shorten the design cycle through
carefully controlled parametric studies, thereby reducing
the required amount of experimental testing.

Equations of Motion
Continuity equation
=0
NavierStokes equation
1
= + 2

The equations of motion to be solved by


CFD for the case of steady,
incompressible, laminar flow of a
Newtonian fluid with constant properties
and without free-surface effects.
A Cartesian coordinate system is used.
There are four equations and four
unknowns: , , , and .
4

first-order approximation

Equations of Motion

2 1

Continuity equation
=0
NavierStokes equation
1
= + 2

The equations of motion to be solved by


CFD for the case of steady,
incompressible, laminar flow of a
Newtonian fluid with constant properties
and without free-surface effects.
A Cartesian coordinate system is used.
There are four equations and four
unknowns: , , , and .
5

kiel probes

CFD Process
Geometry
& Domain

Meshing

Boundary
conditions
& Fluid
properties

Initial values,
discretization
methods,
& solution
algorithms

Solved
Iteratively

Post
processing

=0

=0

Computational domain is the region in


space in which the equations of motion
are solved by CFD
A cell is a small subset of the
computational domain

2-D domain and quadrilateral cell

3-D domain and hexahedral cell

Grid Generation
Structured grid
Unstructured grid

Node
9 nodes and 8 intervals on
the top & bottom edges
5 nodes and 4 intervals on
the left & right edges

Note: Same node distribution in the structured and unstructured grids

A quality grid is essential to


a quality CFD simulation

Equiangle skewness:
max equal equal min
EAS = MAX
,
180 equal
equal

Hybrid grid with the sharp


corner chopped off

EAS max = 0.53

Grid Independent
CFD solution is grid independent?

To test:
Repeat the simulation using finer grid
(a factor of 2 in all directions if feasible)
If the results do not change
appreciably, the original grid is
probably adequate
If it does change significantly, the
original grid resolution is inadequate
Use even finer grid(s) until the grid is
adequately resolved

Grid Independent
CFD solution is grid independent?

To test:
Repeat the simulation using finer grid
(a factor of 2 in all directions if feasible)
If the results do not change
appreciably, the original grid is
probably adequate
If it does change significantly, the
original grid resolution is inadequate
Use even finer grid(s) until the grid is
adequately resolved

Twice finer

23 = 8

e.g.
1 million cells 8 million cells 64 million cells

Boundary Conditions
Appropriate boundary conditions
are required to obtain an accurate
CFD solution

Wall Boundary Conditions


Fluid cannot pass through a wall
The normal component of velocity is set
to zero relative to the wall along a face
on which the wall boundary condition is
prescribed
Because of the no-slip condition, we
usually set the tangential component of
velocity at a stationary wall to zero as
well

= 0

Wall BC = 0 (no-slip wall)


= (Free-slip wall)
20

Boundary conditions
Side walls and ceiling:
Free slip

Free-slip wall BC

Vehicle surface:
Log-law BC
Ground:
Log-law Velocity BC or Log-law BC

21

Inflow/Outflow Boundary Conditions


There are several options at the
boundaries through which fluid enters
the computational domain (inflow) or
leaves the domain (outflow).
They are generally categorized as either:
Velocity-specified conditions
Pressure-specified conditions
At a velocity inlet, we specify the velocity
of the incoming flow along the inlet face
If energy and/or turbulence equations
are being solved, the temperature and/or
turbulence properties of the incoming
flow need to be specified as well

22

With pressure inlet and pressure outlet,


we specify the pressure, not the velocity
As the CFD solution converges, the
velocity adjusts itself such that the
prescribed pressure BC are satisfied

With Outflow BC, the gradient or slope of


velocity normal to the outflow face is
zero
Neither pressure nor velocity are
specified at the outflow boundary

23

Miscellaneous Boundary Conditions


Some boundaries are neither walls nor
inlets or outlets, but rather enforce some
kind of symmetry or periodicity
For example, the periodic boundary
condition is useful when the geometry
involves repetition

Periodic boundary conditions must be


specified as either translational
(periodicity applied to two parallel faces,
or rotational (periodicity applied to two
radially oriented faces).

24

The symmetry BC is imposed on a face so


that the flow across that face is a mirror
image of the calculated flow

25

Internal Boundary Conditions


The final classification of boundary conditions is
imposed on faces or edges that do not define a
boundary of the computational domain, but
rather exist inside the domain.
When an interior boundary condition is
specified on a face, flow crosses through the
face without any user-forced changes, just as it
would cross from one interior cell to another.
This boundary condition is necessary for
situations in which the computational domain is
divided into separate blocks or zones, and
enables communication between blocks.

26

Case 1:
LAMINAR CFD CALCULATIONS

Pipe Flow Entrance


Region at Re = 500

27

28

29

Case 2: Flow around a Circular Cylinder at Re = 150

30


Coarse

120

Cd
1.00

Medium 110

0.982

Fine

109

0.977

Exp.

82

1.1 to 1.4

31

32

33

34

35

36

TURBULENT CFD CALCULATIONS


CFD simulations of turbulent flow are
much more difficult than those of laminar
flow, even for cases in which the flow field
is steady in the mean
The reason is that the finer features of the
turbulent flow field are always unsteady
and 3-D random, swirling, vortical
structures called turbulent eddies of all
orientations arise in a turbulent flow

Laser-induced fluorescence image of an incompressible turbulent boundary layer, by C.


Delo . Flow is from left to right, the flow was visualized with disodium fluorescein dye in
water. Reynolds number based on momentum thickness is 700.

Types of CFD Method based on the


extend to which the turbulent eddies are being resolved

Direct numerical simulation (DNS)

Attempt is made to resolve the unsteady


motion of all the scales of the turbulent flow

Large eddy simulation (LES)


Only large eddies are resolved
Small eddies are modeled
Significantly reducing computer requirements

Reynolds-averaged Navier-Stokes (RANS)


All turbulent eddies are modeled
Only Reynolds-averaged flow properties are
calculated
Computer requirements are minimum

Specific Reynolds
stress tensor

In Reynolds-averaged NavierStokes (RANS) method, turbulence model is


needed due to the additional parameters introduced in the RANS equation

39

40

Case 3: Flow around a Circular Cylinder at Re = 10,000

41

Flow around a Circular Cylinder at Re = 107

42

Case 4: Design of
the Stator for a
Vane-Axial Flow Fan

43

Finer grid near walls

44

Targets:
avg > 45 deg.
No significant flow separation

45

46

47

48

154 CFD WITH HEAT TRANSFER


By coupling the differential form of the
energy equation with the equations of fluid
motion, we can use a computational fluid
dynamics code to calculate properties
associated with heat transfer (e.g.,
temperature distributions or rate of heat
transfer from a solid surface to a fluid).
Since the energy equation is a scalar
equation, only one extra transport equation
(typically for either temperature or
enthalpy) is required, and the
computational expense (CPU time and
RAM requirements) is not increased
significantly.
Heat transfer capability is built into most
commercially available CFD codes, since
many practical problems in engineering
involve both fluid flow and heat transfer. As
mentioned previously, additional boundary
conditions related to heat transfer need to
be specified.

49

Case 5: Temperature Rise through a Cross-Flow Heat Exchanger

Hot tube

50

Temperature Rise through a Cross-Flow Heat Exchanger


= 0
T increase = 5.51 K

= 10
T increase = 5.65 K
Improved by 2.5%

51

Temperature Rise through a Cross-Flow Heat Exchanger


= 0
Turbulent intensity = 10%
T increase = 5.51 K
Turbulent intensity = 25%
T increase = 5.87 K
Improved by 6.5%

= 10
T increase = 5.65 K
Improved by 2.5%

52

= 0
Tube wall is modelled as rough surface
Characteristic roughness height of 0.01 m (1% of cube diameter)
T increase = 14.48 K
Improved by 163%

53

Case 6: Cooling of an Array of Integrated Circuit Chips

54

55

56

57

58

155 COMPRESSIBLE FLOW CFD CALCULATIONS


When the flow is compressible, density is no longer a constant, but
becomes an additional variable in the equation set.

We limit our discussion here to ideal gases.


When we apply the ideal-gas law, we introduce yet another unknown,
namely, temperature T.
Hence, the energy equation must be solved along with the
compressible forms of the equations of conservation of mass and
conservation of momentum.
In addition, fluid properties, such as viscosity and thermal conductivity,
are no longer necessarily treated as constants, since they are functions
of temperature; thus, they appear inside the derivative operators in the
differential equations of Fig. 1574.
While the equation set looks ominous, many commercially available
CFD codes are able to handle compressible flow problems, including
shock waves.
59

60

Compressible Flow through a ConvergingDiverging


Nozzle

61

62

63

64

Oblique Shocks over a Wedge

65

66

Two-phase Flow Simulation


Flow over a Bump on the Bottom of a Channel

Fr = 1.81

Fr = 0.452
67

Flow through a Sluice Gate (Hydraulic Jump)

68

How to ensure good quality CFD results?


Grid quality (skewness of cells)? Errors, converged results

Grid adequately fine? Grid independent test


Domain large enough? Verify domain
Appropriate flow assumptions? Verify flow types
Suitable boundary conditions? Verify BCs
Suitable turbulence model? Verify turbulence models
Simulation results are physically correct? Validation
(compare to experimental or analytical results)