Computational Fluid Dynamics

CFD

1
Something to discuss:
Why do we need CFD?
Are there alternative ways?
 Two views on CFD

We computer advocates dont CFD is like a dog walking on its hind

think wind tunnels will become legs; its doing it badly, but its
useless. They will always be amazing that it can do it at all Anon
great places to store computer
printout in D.R Chapman,
1991
 Examples
Diesel engine simulation

Model assumptions:
- Primary atomization not modeled
- Scania D12 mod. optical engine
- Particle size distribution
- 2500bar injection pressure
- Constant wall temperature
- 6bar IMPEg (20% load)
- Top-hat rail-pressure profile
- N-heptane fuel
- Valve motions not included
- 45, 90 and 135 inter-jet angle
- 2.18 Swirl ratio
- 15.1:1 Compression ratio
 Examples
Diesel engine simulation
Sample Results
Instantaneous axial velocity field

Traces from blades

Wake downstream
the tower
Staggered configuration

H
 Staggered configuration

Mean velocity Rms of fluctuation

DY=0 D

DY=0.36 D

DY=0.71 D

DY=1.43 D
 Relation to Hemodynamics

Basically local in character

Intermediate arteries
i.e. Re-dependent

Near bifurcations

Blood flow rate (Re)
Wall shear-stresses
- Magnitude
- Spatial- and temporal-fluctuations
Velocity Field
WSS-Magnitude
 Closed building

4 wind directions
60 m/s
k-e RNG
770000 cells
0.5 m node distance close to
surface
Domain size 1000X1000X500
m
 Closed building

Path lines

0 deg. 45 deg. 90 deg.

 Open building

2 blade configurations
4 m/s
k-e RNG
800000 cells
0.5 m node distance close to
surface
Domain size 1000X1000X500 m
 Open building

Case 1 Case 2
 Axial Compressor
Aeroelastic simulation of compressorblades using ANSYS 12
 Axial Compressor

Mean Mach number

 Axial Compressor
Aeroelastic simulation of compressorblades using ANSYS 12
Axial Compressor
Deformation of the blade tip
 Mathematical
description Results
ui

Numerical
xi
0
For example speed,
ui ui u j
t

x j

1 p
xi

u i ij

x j x j x j
Fi pressure, temperature
methods

Geometry

Models for
turbulence,
combustion etc.
Mathematical
description of
physical reality

FV, FD, FE?

Governing equations System of

Discretisation, Equation
and boundary algebraic
choice of grid system solver
conditions equations

Approximate
All these steps introduces errors! solution

How can we guarantee that the approximate

solution is close to the exact one and close to
reality
 Governing equations

The governing equations will depend on what assumptions can

be made regarding the flow. For example is it incompressible or
compressible?

The flow situation will determine the character of the system of

equations. This will in turn influence the choice of numerical
method.
 Discretisation and grid

Questions:
How complex is the geometry?
What accuracy is required? Grid quality?
What about stability?
Grid refinement?
Examples
Prismatic airfoil
 Examples
Prismatic airfoil
Inviscid flow
Grid refinement

Coarse: Cdp=0.0500
R1: Cdp=0.0518
R2: Cdp=0.0530

Coarse grid Refinement 1 Refinement 2

 Examples
Prismatic airfoil
Inviscid vs viscous

Inviscid Viscous
Cdp=0.0500 Cdp=0.0500
Cd,tot=0.0539
 Examples
Prismatic airfoil
Wall refinement

Coarse: Cdp=0.0500, Cd,tot=0.0539

Refined:Cdp=0.0531, Cd,tot=0.0569

refined
coarse
 Examples
Prismatic airfoil
Wall refinement

Converged: Cdp=0.0500, Cd,tot=0.0539

Non-converged:Cdp=0.0495, Cd,tot=0.0533

Residual 10-8 Residual 10-3

 System of equation solver

Questions:
Speed of the computation?
What accuracy is required?
What about stability?
 Approximate solution

Questions:
Is the solution physically reasonable? Conservation?
How to determine the accuracy of the solution?

Simulation Measurement
Governing
equations
 Governing equations
Differentialrelationer

System of equations:

Mass V 0
t
DV
Momentum f p
Dt
De
Energy q kT V V p p V f V
Dt
 Governing equations

System of equations:

ui
0
Mass
t xi
ui ui u j p ij
Momentum f i
t x j x j x j
E ui E q T u j ij ui p
Energy k ui f i
t xi t xi xi xi xi
 Governing equations
Non-conserved forms

ui
ui 0
Mass
t xi xi
ui ui p ij
Momentum u j f i
t x j x j x j
E E q T u j ij ui p
Energy ui k ui f i
t xi t xi xi xi xi
 Governing equations
Conserved form

ui
0
Mass
t xi
ui ui u j p ij
Momentum f i
t x j x j x j
E ui E q T u j ij ui p
Energy k ui f i
t xi t xi xi xi xi
 Governing equations
Conserved form
U Fi
Mass J
t xi U ui
E
U i Fij
Momentum Ji
t x j
ui
U F j
Energy J F ui u j ij p ij
t x j
T
Eu pu k uk jk
x j
j j
0

J f i
q
uk f k t
Classification of
PDEs
 Classification of PDEs
t Charateristic curves
A comment on characteristic curves

Example, an initial value problem.

u u Solution: u x, t u0 ( x ct )
c 0
t x
u ( x,0) u0 ( x) x
Characteristic curves are curves along which signals
are propagated.
By using the chain rule one
can find:
du dx
0 along lines defined by c The constancy of u is the
dt dt signal carried along the
or, characteristic curves
u constant on x ct
Classification of PDEs

u u v v
a1 b1 c1 d1 f1
x y x y
u u v v
a2 b2 c2 d 2 f 2
x y x y

Differentials:

u u
du dx dy
x y
P dy
v v
dx dv dx dy
x y
 Classification of PDEs

u
x
1
a b1 c1 d1 u f1 f1 b1 c1 d1
a b c d f f b c d
2 2 y B 2 2 2 2
2 2
2
dx dy 0 0 v du dx dy 0 0
0 0 dx dy x dv 0 0 dx dy

v
A
y

u B A 0
Characteristic curves correspond to
x A

A a1c2 a2c1 dy 2 a1d 2 a2 d1 b1c2 b2c1 dxdy b1d 2 b2 d1 dx 2

 Classification of PDEs
A a1c2 a2c1 dy 2 a1d 2 a2 d1 b1c2 b2c1 dxdy b1d 2 b2 d1 dx 2

2
dy dy
a1c2 a2c1 a1d 2 a2 d1 b1c2 b2c1 b1d 2 b2 d1 0
dx dx
2
dy dy
a b c 0
dx dx
dy b b 2 4ac

dx 2a
b 4ac 0
Three situations: 2
Hyperbolic, two real characteristics

b 2 4ac 0 Parabolic, one real characteristic

b 2 4ac 0 Elliptic, no real characteristic

 Classification of PDEs
Example: 2D inviscid steady flow of a compressible gas under small perturbations.

1 Ma 2
u ' v'

x y
0 u U u'
v v'
u ' v'
0
a1 1 Ma 2 ; b1 c1 0; d1 1
y v a2 0; b2 1; c2 1; d 2 0

Characteristic equation:

2
dy dy

a1c2 a2c1 a1d 2 a2 d1 b2c1 b1d 2 b2 d1 0
dx dx
a b c
 Classification of PDEs
Characteristic equation:

2
dy dy

a1c2 a2c1 a1d 2 a2 d1 b2c1 b1d 2 b2 d1 0
dx dx
a b c

a 1 Ma 2

2
dy
b0 1 Ma 2 1 0
dx
c 1

dy

1

4 Ma 2 1 dy b b 2 4ac
dx Ma 1
2
2 Ma 2 1 Compare:
dx

2a
Ma 2 1 elliptic
Ma 2 1 hyperbolic
 Classification of PDEs
Hyperbolic
y

Domain of
P
dependence Region of influence

Charateristic lines

x
Examples: Inviscid supersonic flow, unsteady inviscid flow
 Classification of PDEs
Parabolic
y
Known boundary conditions

Domain of
P
dependence Region of influence

Known boundary conditions x

Examples: Steady boundary layer flow, unsteady heat conduction
 Classification of PDEs
Elliptic
y

Every point influences all

other points

x
Examples: Steady subsonic inviscid flow, incompressible inviscid flow