CFD Notes

Computational uid dynamics
Martin Kronbichler
[email protected]
Applied Scientic Computing (Tillmpad berkningsvetenskap)
February 11, 2010
Martin Kronbichler (TDB) FVM for CFD February 11, 2010 1 / 72
Introduction
The NavierStokes equations
Conservation of mass
Conservation of momentum
Constitutive & kinematic relations
Conservation of energy
Equations of state & compressible NavierStokes equations
Incompressible NavierStokes equations
Discretization
Overview of spatial discretizations
The Finite Volume Method
Time discretization
Spatial discretization
Turbulence and its modeling
Introduction
What is CFD?
Fluid mechanics deals with the motion of uids (liquids and gases),
induced by external forces.
Fluid ow is modeled by partial dierential equations (PDE),

describing the conservation of mass, momentum, and energy.
Computational Fluid Dynamics (CFD) is the discipline of

discretizing these PDE and solving them using computers.
Introduction
In which application elds is CFD used?
Aerospace and aeronautical applications (airplanes, water and space

vehicles)
Mechanical applications (gas turbines, heat exchange, explosions,

combustion, architecture)
Biological applications (blood ow, breathing, drinking)
Meteorological applications (weather prediction)
Environmental applications (air and water pollution)
and many more . . .

Introduction
Why to use CFD?
Pre-design of components: simulation vs. experiment (simulations

are cheaper, faster, and safer, but not always reliable)
vehicles with lower fuel consumption, quieter, heavier loads
combustion engines
oil recovery
water and gas turbines (eectiveness)
stress minimization
Detection and prediction
hurricanes, storms, tsunamis
pollution transport
diseases
forces, stresses
Introduction
Why to use CFD?

combustion engines
oil recovery
stress minimization
pollution transport
diseases
forces, stresses
Introduction
Why to use CFD?

combustion engines
oil recovery
stress minimization
pollution transport
diseases
forces, stresses
Introduction
Visualization of CFD results
ONERA M6 wing optimization
O. Amoignon, M. Berggren
Introduction
Visualization of CFD results
Pipe ow, computer lab
Introduction
Requirements for industrial CFD
Robustness give a solution (for as many input cases as possible)
Reliability give a good solution
Performance give a good solution fast
Geometries give a good solution fast for real problems
Automatic tool chain reduce requirements of user interaction

Knowledgeable user controls and evaluates simulation outcomes!
Introduction
CFD Solution Tool Chain
CAD description of geometry
Grid generation from CAD model (usually the most time consuming
part)
Problem set up - knowledgeable user needed
Mathematical model to use
Spatial and temporal discretization
Available resources
Much more . . .
Preprocessing - parse conguration and prepare solver
Solving - run the solver
Post processing - extract and compute information of interest
Visualization - often most important
Interpretation - physics, mathematics, numerics, experiments

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
part)
Available resources
Much more . . .

Introduction
CFD software
Commercial: Fluent, Comsol, CFX, Star-CD
In-house codes: Edge (FOI), DLR-Tau (German Aerospace Center),

Fun3D (NASA), Sierra/Premo (American Aerospace)
Open Source: OpenFOAM, FEniCS, OpenFlower

Introduction
CFD links
http://www.cfd-online.com
http://www.uent.com
http://www.openfoam.org
A mathematical model for uid ow
NavierStokes equations an overview
The governing equations of uid dynamics are the conservation laws

of mass, momentum, and energy.
In CFD, this set of conservation laws are called the

NavierStokes equations.
Derive compressible NavierStokes equations rst
Simplify these equations to get the incompressible NavierStokes

equations
Notation
density
u velocity, u = (u
1
, u
2
, u
3
)
p pressure
T temperature
dynamic viscosity
kinematic viscosity, = /
nabla operator, =
_

x
1
,

x
2
,

x
3
_
inner product, a b = a
1
b
1
+a
2
b
2
+a
3
b
3
General form of conservation laws
Monitor the ow characteristics of a xed control volume .
Rate of total change in the control volume:
change in the interior,
ow over boundary of control volume,

expressed as
D
Dt
_
(t)
q d =
_
q
t
d +
_
qu nds =
_
(t)
S d
for some quantity q and a source term S (generation or elimination).
n denotes the outer normal on . This is the Reynolds transport
theorem.
Note: (t) is an ensemble of molecules, is a xed control volume.
Dierential form:
Dq
Dt
=
q
t
+u q = S.

expressed as
D
Dt
_
(t)
q d =
_
q
t
d +
_
qu nds =
_
(t)
S d
theorem.
Dierential form:
Dq
Dt
=
q
t
+u q = S.

expressed as
D
Dt
_
(t)
q d =
_
q
t
d +
_
qu nds =
_
(t)
S d
theorem.
Dierential form:
Dq
Dt
=
q
t
+u q = S.
The NavierStokes equations Conservation of mass
Continuity equation I
The continuity equation describes the conservation of mass .
The mass M of the material in (t) is constant (molecules cannot be
created/destroyed), that is M(t) = M(t +t).
DM
Dt
=
D
Dt
_
(t)
d
. .
M
= 0
(t) denotes the uid density.
The NavierStokes equations Conservation of mass
Continuity equation II
Rewritten for a stationary control volume by using the Reynolds transport
theorem:
rate of mass change in + mass ow over = 0,
_
t
d +
_
u nds = 0.
The NavierStokes equations Conservation of momentum
Momentum equation I
The momentum equation describes the conservation of momentum .
Newtons second law of motion: the total rate of momentum m change in
(t) is equal to the sum of acting forces K, i.e.,
Dm
Dt
=
D
Dt
_
(t)
ud = K,
(interpretation: mass acceleration = force).
Momentum equation II
Rewritten for a stationary control volume :
_
u
t
+
_
(u u) nds = K,
where u u is a rank-2 tensor with entries u
i
u
j
.
Decompose force into surface forces and volume forces:
rate of momentum change in + momentum ow over
= surface forces on + volume forces on ,
_
u
t
dV+
_
(uu)nds =
_
pnds +
_
nds
. .
surface forces
+
_
f d.
Momentum equation II
Rewritten for a stationary control volume :
_
u
t
+
_
(u u) nds = K,
where u u is a rank-2 tensor with entries u
i
u
j
.
Decompose force into surface forces and volume forces:
rate of momentum change in + momentum ow over
= surface forces on + volume forces on ,
_
u
t
dV+
_
(uu)nds =
_
pnds +
_
nds
. .
surface forces
+
_
f d.
Momentum equation III
Surface forces acting on ds of :
Pressure force p(n) ds (analog to isotropic stress in structural

mechanics),
Viscous force nds, is the Cauchy stress tensor (analog to

deviatoric stress in structural mechanics).
Volume forces f d acting on small interior volume d:
Gravity force g d, e.g., f = g,
Other types of forces: Coriolis, centrifugal, electromagnetic, buoyancy

Momentum equation III
Surface forces acting on ds of :
Pressure force p(n) ds (analog to isotropic stress in structural

mechanics),
Viscous force nds, is the Cauchy stress tensor (analog to

deviatoric stress in structural mechanics).
Volume forces f d acting on small interior volume d:
Gravity force g d, e.g., f = g,
Other types of forces: Coriolis, centrifugal, electromagnetic, buoyancy

Newtonian uid
Model that relates the stress tensor to the velocity u:
We consider so-called Newtonian uids, where the viscous stress is
linearly related to strain rate, that is,
= 2(u)
2
3
I tr (u) =
_
u +(u)
T
_
2
3
( u)I,
(compare with constitutive and kinematic relations in elasticity theory).
For Cartesian coordinates
ij
=
_
u
j
x
i
+
u
i
x
j
_
2
3
k=1
u
k
x
k
.
The dynamic viscosity is assumed constant
= (T, p) constant.
Newtonian uid
Model that relates the stress tensor to the velocity u:
We consider so-called Newtonian uids, where the viscous stress is
linearly related to strain rate, that is,
= 2(u)
2
3
I tr (u) =
_
u +(u)
T
_
2
3
( u)I,
(compare with constitutive and kinematic relations in elasticity theory).
For Cartesian coordinates
ij
=
_
u
j
x
i
+
u
i
x
j
_
2
3
k=1
u
k
x
k
.
The dynamic viscosity is assumed constant
= (T, p) constant.
The NavierStokes equations Conservation of energy
Energy equation I
The energy equation describes the conservation of energy .
The rst law of thermodynamics:
The total rate of total energy E changes in (t) is equal to the rate of
work L done on the uid by the acting forces K plus the rate of heat added
W, that is
DE
Dt
=
D
Dt
_
(t)
E d = L +W
where E is the total energy per unit volume.
Energy Equation II
Rewritten for a stationary control volume , explicitly specifying the source
terms:
Rate of total energy change in + Total energy ow over =
Rate of work of pressure and viscous forces on +
Rate of work of forces on + Rate of heat added over
_
E
t
d +
_
Eu nds =
_
pu nds +
_
( u) nds
+
_
f ud +
_
kT nds
for the uid temperature T and thermal conductivity k.
Involved variables and equations
We have
seven variables (, u
1
, u
2
, u
3
, p, T, E)
ve equations (continuity, 3momentum, energy)

We need two more equations to close the system, the so-called equations
of state.
The NavierStokes equations Equations of state & compressible NavierStokes equations
Equations of state
Properties of a perfect gas:
p =( 1)
internal energy
..
_
E
1
2
|u|
2
_
T =
1
c
v
_
E
1
2
|u|
2
_
where =
c
p
c
v
is the ratio of specic heats, and c
p
and c
v
are the specic
heats at constant pressure and volume.
A typical value for air at sea level pressure is = 1.4.
The compressible NavierStokes equations in integral form
The equations of continuity, momentum, and energy can be combined into
one system of equations.
Dene compound variable U = (, u

1
, u
2
, u
3
, E) (called
conservative variables)
Dene ux vectors
F n =
_
_
u n
(u u) n +pI n
(E +p)u n
_
_
. .
inviscid/convective
_
_
0
n
( u) n +k(T) n
_
_
. .
viscous
and an external strength vector F
e
= (0, f, f u)
T
Compressible NavierStokes equations in integral form
_
U
t
d +
_
F nds =
_
F
e
d
The compressible NavierStokes equations in integral form
The equations of continuity, momentum, and energy can be combined into
one system of equations.
Dene compound variable U = (, u

1
, u
2
, u
3
, E) (called
conservative variables)
Dene ux vectors
F n =
_
_
u n
(u u) n +pI n
(E +p)u n
_
_
. .
inviscid/convective
_
_
0
n
( u) n +k(T) n
_
_
. .
viscous
and an external strength vector F
e
= (0, f, f u)
T
Compressible NavierStokes equations in integral form
_
U
t
d +
_
F nds =
_
F
e
d
Compressible NavierStokes equations, dierential form
Assume: ux tensor F is dierentiable
Apply the Gauss theorem to the integral form and get
_
_
U
t
+ F F
e
_
d = 0.
Since the integral is zero for an arbitrary control volume , we obtain the
dierential form of the
Compressible NavierStokes equations
U
t
+ F = F
e
.
Compressible NavierStokes equations, dierential form
Assume: ux tensor F is dierentiable
Apply the Gauss theorem to the integral form and get
_
_
U
t
+ F F
e
_
d = 0.
Since the integral is zero for an arbitrary control volume , we obtain the
dierential form of the
Compressible NavierStokes equations
U
t
+ F = F
e
.
The NavierStokes equations Incompressible NavierStokes equations
The incompressible NavierStokes equations
Motivation: we can use simpler equations if we have more information
Incompressible uid: density does not change with pressure, i.e.,

= const.
Energy equation decouples from the rest of the system; continuity and
momentum equations can be simplied.
Many practically relevant ows are incompressible (e.g. air at speeds

up to 100 m/s) predominant CFD model
Continuity equation
The continuity equation,
t
+ (u) = 0,
becomes for constant simply
u = 0, i.e.,
3
i =1
u
i
x
i
= 0.
Momentum equation
The momentum equation,
u
t
+ (u u) = p + +f,
can be simplied by using u = 0,
(u u) = (u )u +( u)u = (u )u
=
_
u +(u)
T
_
2
3
( u)I =
_
u +(u)
T
_
=
_
u +(u)
T
_
=
_
2
u +( u)
_
=
2
u.
This gives the incompressible version of the momentum equation
u
t
+(u )u =
1
p +

2
u +f
Momentum equation
u
t
+ (u u) = p + +f,
(u u) = (u )u +( u)u = (u )u
=
_
u +(u)
T
_
2
3
( u)I =
_
u +(u)
T
_
=
_
u +(u)
T
_
=
_
2
u +( u)
_
=
2
u.
u
t
+(u )u =
1
p +

2
u +f
Momentum equation
u
t
+ (u u) = p + +f,
(u u) = (u )u +( u)u = (u )u
=
_
u +(u)
T
_
2
3
( u)I =
_
u +(u)
T
_
=
_
u +(u)
T
_
=
_
2
u +( u)
_
=
2
u.
u
t
+(u )u =
1
p +

2
u +f
The incompressible NavierStokes equations II
Since the energy equation does not enter the momentum and continuity
equation, we have a closed system of four equations:
Incompressible NavierStokes equations
_
_
_
u = 0,
u
t
+(u )u =
1
p +
2
u +f,
where =

is the uid kinematic viscosity.

Equation in temperature
The energy equation can be expressed in terms of temperature by using the
equation of state. Transforming the integral form into dierential form as
done before
1
yields
c
p
_
T
t
+ (Tu)
_
= (kT) + : u,
where
: u =
3
i ,j =1
ij
u
j
x
i
=
3
i ,j =1
1
2
_
u
i
x
j
+
u
j
x
i
_
2
,
and is called a dissipative function.
1
for further reading, see e.g. J. Blazek: Computational Fluid Dynamics, Elsevier,
Amsterdam
Incompressible NavierStokes equations III
Energy equation is decoupled from the continuity and momentum

equations
rst solve the continuity and momentum equations to get the velocity
and pressure.
temperature (energy) equation can be solved for temperature using the

already computed velocity.
Pressure level only dened up to a constant for incompressible

ow
pressure level has to be xed at one point in the ow.
p yields then the relative pressure dierence with respect to that

pressure level.
No time derivative for the pressure
mathematical diculty of the incompressible NavierStokes equations.
p is indirectly determined by the condition u = 0 (p has to be such

that the resulting velocity is divergence free, Lagrange multiplier).

equations
and pressure.


ow

pressure level.


equations
and pressure.


ow

pressure level.

Additional conditions
To complete the physical problem setup after having derived the
appropriate PDE, we have to dene the
domain
initial conditions (velocity, temperature, etc at t = 0)
boundary conditions (inow, outow, wall, interface, . . .)
material properties
Discretization
Discretization
Numerical solution strategies for the
ow equations
Discretization
Solution Strategies Outline
Some points we will cover regarding solver strategies of CFD
Pros and cons of various discretization methods
Finite Volume Method (FVM)

Discretization Overview of spatial discretizations
Overview of Methods
Discretization methods for the CFD equations:
Finite Dierence Method (FDM)

+ eciency, + theory, geometries
Finite Element Method (FEM)

+ theory, + geometries, shocks, uses no directional information
Spectral Methods (Collocation, Galerkin, . . .)

+ accuracy, + theory, geometries
Finite Volume Methods (FVM)

+ robustness, + geometries, accuracy
Discontinuous Galerkin Methods (DGM)

+ geometries, + shocks, eciency (e.g. smooth solutions)
Hybrid Methods
+ versatile, complex (not automatable)
Overview of Methods





Hybrid Methods
Overview of Methods





Hybrid Methods
Discretization The Finite Volume Method
Introduction
The Finite Volume Method (FVM) is based on the integral form of the
governing equations.
The integral conservation is enforced in so-called control volumes
( d ) dened by the computational mesh.
The type of FVM is specied by
the type of control volume
the type of evaluation of integrals and uxes

FVM derivation, introduction
Consider a general scalar hyperbolic conservation law:
u
t
+ F(u) = 0 in , (1)
with appropriate boundary and initial conditions.
Compared to the compressible NavierStokes equations, we use a problem
where
u replaces the vector of conservative variables (, u, E),
F replaces the ux tensor as dened earlier,
we have a scalar problem instead of a system.

FVM derivation, introduction
Consider a general scalar hyperbolic conservation law:
u
t
+ F(u) = 0 in , (1)
with appropriate boundary and initial conditions.
Compared to the compressible NavierStokes equations, we use a problem
where
u replaces the vector of conservative variables (, u, E),
F replaces the ux tensor as dened earlier,
we have a scalar problem instead of a system.

FVM derivation, control volumes
Consider the equation in two dimensions, and we divide the domain into
M non-overlapping control volumes K
m
.
K
m
vertex-centered nite volume method
FVM derivation, discretized solution
In each control volume, K
m
, we store one value u
m
, which is the average
of u in K
m
.
u
m
=
1
|K
m
|
_
K
m
u d
FVM derivation, integral form I
The FVM is based on the integral formulation of the equations.
We start by integrating (1) over one of the control volumes,
_
K
m
u
t
d +
_
K
m
F(u) d = 0. (2)
The rst integral can be simplied by changing order of time derivative and
integration to get
_
K
m
u
t
d =
d
dt
_
K
m
u d = |K
m
|
du
m
dt
. (3)
Inserting (3) back into (2) yields
du
m
dt
=
1
|K
m
|
_
K
m
F(u) d.
FVM derivation, integral form I
The FVM is based on the integral formulation of the equations.
We start by integrating (1) over one of the control volumes,
_
K
m
u
t
d +
_
K
m
F(u) d = 0. (2)
The rst integral can be simplied by changing order of time derivative and
integration to get
_
K
m
u
t
d =
d
dt
_
K
m
u d = |K
m
|
du
m
dt
. (3)
Inserting (3) back into (2) yields
du
m
dt
=
1
|K
m
|
_
K
m
F(u) d.
FVM derivation, time evolution
We need to propagate the solution u
m
in time.
There are many possibilities for choosing a time integration method, for
example RungeKutta, multistep methods etc.
See some Ordinary Dierential Equation (ODE) textbook for more
information.
2
We will use forward Euler in all our examples, i.e.,
du
m
dt
=
u
n+1
m
u
n
m
t
,
where u
n
m
= u
m
(t
n
).
2
e.g., E. Hairer, S.P. Nrsett, G. Wanner: Solving Ordinary Dierential Equations. I:
Nonsti Problems. Springer-Verlag, Berlin, 1993.
We need to propagate the solution u
m
in time.
There are many possibilities for choosing a time integration method, for
example RungeKutta, multistep methods etc.
See some Ordinary Dierential Equation (ODE) textbook for more
information.
2
We will use forward Euler in all our examples, i.e.,
du
m
dt
=
u
n+1
m
u
n
m
t
,
where u
n
m
= u
m
(t
n
).
2
e.g., E. Hairer, S.P. Nrsett, G. Wanner: Solving Ordinary Dierential Equations. I:
Nonsti Problems. Springer-Verlag, Berlin, 1993.
Inserting the forward Euler time discretization into our system gives
u
n+1
m
= u
n
m
t
|K
m
|
_
K
m
F(u) d.
FVM derivation, integral form II
By integration by parts (Gauss divergence theorem) of the ux integral we
get the basic FVM formulation
u
n+1
m
= u
n
m
t
|K
m
|
_
K
m
F(u) nds. (4)
The boundary integral describes the ux of u over the boundary K
m
of
the control volume.
Note: Transforming the volume to a surface integral gets us back to the
form used for the derivation of the NavierStokes equations.
Dierent ways of evaluating the ux integral specify the type of FVM
(together with the control volume type).
FVM derivation, integral form II
By integration by parts (Gauss divergence theorem) of the ux integral we
get the basic FVM formulation
u
n+1
m
= u
n
m
t
|K
m
|
_
K
m
F(u) nds. (4)
The boundary integral describes the ux of u over the boundary K
m
of
the control volume.
Note: Transforming the volume to a surface integral gets us back to the
form used for the derivation of the NavierStokes equations.
Dierent ways of evaluating the ux integral specify the type of FVM
(together with the control volume type).
FVM derivation, numerical ux
The integral in (4) is evaluated using a so-called numerical ux function.
There are a lot of dierent numerical uxes available for the integral
evaluation, each of them having its pros and cons.
Terminology: Riemann solvers (they are usually designed to solve the
Riemann problem which is a conservation law with constant data and
a discontinuity).
Both exact and approximate solvers (i.e., ways to evaluate the ux

integral (4)) have been developed, usually the latter class is used.
E.g. exact solver for Euler

3
equations developed by Godunov.
Widely used approximate solvers: Roe, HLLC, central ux.

3
Contains only the inviscid terms from the compressible NavierStokes equations
FVM derivation, numerical ux
The integral in (4) is evaluated using a so-called numerical ux function.
There are a lot of dierent numerical uxes available for the integral
evaluation, each of them having its pros and cons.
Terminology: Riemann solvers (they are usually designed to solve the
Riemann problem which is a conservation law with constant data and
a discontinuity).
Both exact and approximate solvers (i.e., ways to evaluate the ux

integral (4)) have been developed, usually the latter class is used.
E.g. exact solver for Euler

3
equations developed by Godunov.
Widely used approximate solvers: Roe, HLLC, central ux.

3
Contains only the inviscid terms from the compressible NavierStokes equations
FVM derivation, numerical ux II
The numerical ux functions are usually denoted by F
(u
L
, u
R
, n) where u
L
is the left (local) state and u
R
is the right (remote) state of u at the
boundary K
m
,
_
K
m
F(u) nds =
j
_
K
j
m
F(u) n
j
ds
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
and j denotes the index of each subboundary of K
m
with a given normal
n
j
.
FVM derivation, numerical ux II
The numerical ux functions are usually denoted by F
(u
L
, u
R
, n) where u
L
is the left (local) state and u
R
is the right (remote) state of u at the
boundary K
m
,
_
K
m
F(u) nds =
j
_
K
j
m
F(u) n
j
ds
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
and j denotes the index of each subboundary of K
m
with a given normal
n
j
.
FVM derivation, full discretization
The full discretization for control volume K
m
is hence
u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
In one dimension, we have
u
n+1
m
= u
n
m
t
x
_
K
m
F(u) nds
= u
n
m
t
x
_
F(u
n
m+1/2
) F(u
n
m1/2
)
_
= u
n
m
t
x
_
F
(u
n
m
, u
n
m1
, 1) +F
(u
n
m
, u
n
m+1
, 1)
_
m
is hence
u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
u
n+1
m
= u
n
m
t
x
_
K
m
F(u) nds
= u
n
m
t
x
_
F(u
n
m+1/2
) F(u
n
m1/2
)
_
= u
n
m
t
x
_
F
(u
n
m
, u
n
m1
, 1) +F
(u
n
m
, u
n
m+1
, 1)
_
m
is hence
u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
u
n+1
m
= u
n
m
t
x
_
K
m
F(u) nds
= u
n
m
t
x
_
F(u
n
m+1/2
) F(u
n
m1/2
)
_
= u
n
m
t
x
_
F
(u
n
m
, u
n
m1
, 1) +F
(u
n
m
, u
n
m+1
, 1)
_
FVM derivation, upwind scheme I
Get the information upwind, where the information comes from.
Image from http://en.wikipedia.org
FVM derivation, upwind scheme II
Example in one dimension
F(u) = u,
where is the direction of the ow. So in each control volume we compute
_
K
m
un ds.
Case > 0 (information ows from left to right), equidistant mesh:
u
m1/2
= u
m1
, u
m+1/2
= u
m
.
Hence, we get
u
n+1
m
= u
n
m
t
x
_
F(u
n
m+1/2
)n
m+1/2
+F(u
n
m1/2
)n
m1/2
_
= u
n
m
t
x
_
u
n
m+1/2
u
n
m1/2
_
= u
n
m
t
x
_
u
n
m
u
n
m1
_
.
F(u) = u,
_
K
m
un ds.
u
m1/2
= u
m1
, u
m+1/2
= u
m
.
Hence, we get
u
n+1
m
= u
n
m
t
x
_
F(u
n
m+1/2
)n
m+1/2
+F(u
n
m1/2
)n
m1/2
_
= u
n
m
t
x
_
u
n
m+1/2
u
n
m1/2
_
= u
n
m
t
x
_
u
n
m
u
n
m1
_
.
F(u) = u,
_
K
m
un ds.
u
m1/2
= u
m1
, u
m+1/2
= u
m
.
Hence, we get
u
n+1
m
= u
n
m
t
x
_
F(u
n
m+1/2
)n
m+1/2
+F(u
n
m1/2
)n
m1/2
_
= u
n
m
t
x
_
u
n
m+1/2
u
n
m1/2
_
= u
n
m
t
x
_
u
n
m
u
n
m1
_
.
FVM derivation, upwind scheme (Roe ux)
Practical implementation of upwinding: construct it by the Roe Flux
F
(u
L
, u
R
, n) =
n F(u
L
) +n F(u
R
)
2
+
n F
( u)
u
L
u
R
2
.
Here, u satises the mean value theorem,
n F(u
L
) = n F(u
R
) +n F
( u) (u
L
u
R
) .
and is called the Roe average (problem dependent).
Upwinding is rst order accurate.
FVM derivation, upwind scheme (Roe ux)
Practical implementation of upwinding: construct it by the Roe Flux
F
(u
L
, u
R
, n) =
n F(u
L
) +n F(u
R
)
2
+
n F
( u)
u
L
u
R
2
.
Here, u satises the mean value theorem,
n F(u
L
) = n F(u
R
) +n F
( u) (u
L
u
R
) .
and is called the Roe average (problem dependent).
Upwinding is rst order accurate.
FVM derivation, central scheme I
For F(u) = u on a equidistant mesh, the central scheme approximation is
u
m1/2
=
u
m1
+u
m
2
, u
m1/2
=
u
m
+u
m+1
2
.
Hence,
u
n+1
m
= u
n
m
t
x
_
F(u
n
m+1/2
)n
m+1/2
+F(u
n
m1/2
)n
m1/2
_
= u
n
m
t
x
_
u
n
m+1/2
u
n
m1/2
_
= u
n
m
t
x
_
u
n
m
+u
n
m+1
2

u
n
m1
+u
n
m
2
_
= u
n
m
t
x
_
u
n
m+1
u
n
m1
2
_
.
FVM derivation, central scheme I
For F(u) = u on a equidistant mesh, the central scheme approximation is
u
m1/2
=
u
m1
+u
m
2
, u
m1/2
=
u
m
+u
m+1
2
.
Hence,
u
n+1
m
= u
n
m
t
x
_
F(u
n
m+1/2
)n
m+1/2
+F(u
n
m1/2
)n
m1/2
_
= u
n
m
t
x
_
u
n
m+1/2
u
n
m1/2
_
= u
n
m
t
x
_
u
n
m
+u
n
m+1
2

u
n
m1
+u
n
m
2
_
= u
n
m
t
x
_
u
n
m+1
u
n
m1
2
_
.
FVM derivation, central scheme II
Possibilities for evaluating the central ux:
1. F
(u
L
, u
R
, n) = n F
_
u
L
+u
R
2
_
d,
2. F
(u
L
, u
R
, n) =
n F(u
L
) +n F(u
R
)
2
d.
The central scheme is second order accurate.
But:
the central scheme gives rise to unphysical oscillations around steep

gradients (shocks)
the central scheme ignores the direction of the ow
choice of articial dissipation d can reduce accuracy

FVM derivation, central scheme II
Possibilities for evaluating the central ux:
1. F
(u
L
, u
R
, n) = n F
_
u
L
+u
R
2
_
d,
2. F
(u
L
, u
R
, n) =
n F(u
L
) +n F(u
R
)
2
d.
The central scheme is second order accurate.
But:
the central scheme gives rise to unphysical oscillations around steep

gradients (shocks)
the central scheme ignores the direction of the ow
choice of articial dissipation d can reduce accuracy

FVM derivation, second derivatives
Flux tensor F contains terms with rst derivatives in u for the
NavierStokes equations (corresponding to second derivatives in the
dierential form of the equations).
Need to approximate these terms before evaluating the ux function.
Usually, apply central dierences of the kind
_
u
x
_
m1/2
=
u
m
u
m1
x
.
FVM derivation, second derivatives
Flux tensor F contains terms with rst derivatives in u for the
NavierStokes equations (corresponding to second derivatives in the
dierential form of the equations).
Need to approximate these terms before evaluating the ux function.
Usually, apply central dierences of the kind
_
u
x
_
m1/2
=
u
m
u
m1
x
.
Higher order discretization schemes for FVM
There are a number of higher order schemes for orders of accuracy 2,
like
MUSCL (Monotone Upstream-centered Schemes for Conservation

Laws)
ENO (Essentially Non-Oscillatory)
WENO (Weighted ENO)
RDS (Residual Distribution Scheme)

They combine dierent techniques to attain a high-order solution without
excessive oscillations, for example so called ux limiters (reducing articial
oscillations), wider stencils over several control volumes, dierent
weightings, reconstruction, and adding degrees of freedom in each control
volume.
Higher order discretization schemes for FVM
There are a number of higher order schemes for orders of accuracy 2,
like
MUSCL (Monotone Upstream-centered Schemes for Conservation

Laws)
ENO (Essentially Non-Oscillatory)
WENO (Weighted ENO)
RDS (Residual Distribution Scheme)

They combine dierent techniques to attain a high-order solution without
excessive oscillations, for example so called ux limiters (reducing articial
oscillations), wider stencils over several control volumes, dierent
weightings, reconstruction, and adding degrees of freedom in each control
volume.
Choosing the discretization parameters
A necessary condition for stability of the time discretization is the CFL
condition (depends on the combination of space and time discretization).
For the one dimensional upwind discretization
u
n+1
m
= u
n
m
t
x
_
u
n
m
u
n
m1
_
,
we require

t
x
1.
Interpretation: The dierence approximation (which only uses information
from one grid point to the left/right) can only represent variations in the
solution up to /x from one time step to the next time step size has
to be smaller than that limit
Steady state calculations
For certain computations, the time-dependance is of no importance.
Sought: steady state solution. when the solution has stabilized, the
change in time will go to zero,
u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
. .
residual
= u
n
m
, m.
Measure whether steady state has been achieved: residual gets small.
To reach steady state, many dierent acceleration techniques can be used,
for example local time stepping and multigrid.
Two examples of steady state:
computation of the drag and lift around an airfoil,
mixing problem considered in the computer lab.

u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
. .
residual
= u
n
m
, m.

u
n+1
m
= u
n
m
t
|K
m
|
j
|K
j
m
|F
(u
L
, u
R
, n
j
)
. .
residual
= u
n
m
, m.

FVM for the incompressible NavierStokes equations
The incompressible NavierStokes equations are a system in ve

equations (u, p, T), so go through them one by one according to the
above procedure
Need to take special care of p, since it does not involve a time

derivative
Computational domain and material parameters , specied from

application
Initial and boundary conditions complete the formulation

Initial condition
The state of all variables at time t
0
has to be dened in order to initiate
the solution process.
For example
u|
t=0
= u
0
.
Boundary conditions, inow
The inow is the part of the domain where u n < 0.
Setting an inow Dirichlet boundary condition, that is u = u
in
on K
in
:
strongly imposed, set the inow boundary node values to the

respective value
U
m
= U
in
for all faces K
m
on K
in
.
weakly imposed, use the numerical ux,

F
(U
m
, U
in
, n).
The inow data can be viewed as given in a ghost point; boundary
treated as the interior.
in
on K
in
:

respective value
U
m
= U
in
for all faces K
m
on K
in
.

F
(U
m
, U
in
, n).
in
on K
in
:

respective value
U
m
= U
in
for all faces K
m
on K
in
.

F
(U
m
, U
in
, n).
Boundary conditions, outow
The outow is the part of the domain where u n > 0.
Treating outow boundaries is more complicated than inow parts.
Usually one does a weak formulation such that
1. F
(U
L
, U
L
, n) is used as ux function,
2. F
(U
L
, U
, n) is used as ux function with U
a far-eld
velocity/pressure/temperature.
There are also strong formulations.
Beware of articial back ows!
A common trick is to make the domain big enough to avoid distorting the
solution in the domain of interest.
Boundary conditions, outow
The outow is the part of the domain where u n > 0.
Treating outow boundaries is more complicated than inow parts.
Usually one does a weak formulation such that
1. F
(U
L
, U
L
, n) is used as ux function,
2. F
(U
L
, U
, n) is used as ux function with U
a far-eld
velocity/pressure/temperature.
There are also strong formulations.
Beware of articial back ows!
A common trick is to make the domain big enough to avoid distorting the
solution in the domain of interest.
Boundary conditions, solid walls
The solid wall is the part of the domain where u n = 0.
For viscous (NavierStokes) ow we apply a no-slip condition
u = 0.
These boundary conditions are usually imposed strongly.
Heat transfer
When considering heat transfer for the incompressible NavierStokes
equations, we also need to assign a BC for the temperature T. Two
mechanisms can be applied:
specify temperature with a Dirichlet BC, T = T

w
, or
specify heat ux with a Neumann BC, T/n = f

q
/.
Turbulence
Turbulence
Introduction, dimensionless form
In CFD, a scaled dimensionless form of the NavierStokes equation is often
used.
We introduce new variables
x
i
=
x
i
L
, u
i
=
u
i
U
, p
=
p
U
2
,

x
i
=
1
L
i
,

t
=
U
L
,
in which the incompressible NavierStokes equations read
= 0,
u
+(u
)u
+

UL
..
Re
1
2
u
.
Reynolds number and turbulence
The Reynolds number is dened as
Re =
UL
=
Inertial forces
Viscous forces
,
where U, L, are the characteristic velocity, characteristic length, and the
kinematic viscosity, respectively.
Fluids with the same Reynolds number behave the same way.
When the Reynolds number becomes larger than a critical value, the
formerly laminar ow changes into turbulent ow, for example at
Re 2300 for pipe ows.
Reynolds number and turbulence
The Reynolds number is dened as
Re =
UL
=
Inertial forces
Viscous forces
,
where U, L, are the characteristic velocity, characteristic length, and the
kinematic viscosity, respectively.
Fluids with the same Reynolds number behave the same way.
When the Reynolds number becomes larger than a critical value, the
formerly laminar ow changes into turbulent ow, for example at
Re 2300 for pipe ows.
Characterization of turbulence
Turbulent ows are
time dependent
three dimensional
irregular
vortical ( = u)
Image from http://en.wikipedia.org
Describing the turbulence can be done in many ways, and the choice of the
method depends on the application at hand and on computational
resources.
Direct Numerical Simulation (DNS)
All scales in space and time are resolved.
No modeling of the turbulence.
Limited to small Reynolds numbers, because extremely ne grids and

time steps are required, O(Re
3
) spatial and temporal degrees of
freedom.
Applications of DNS:
Turbulence research.
Reference results to verify other turbulence models.

Large Eddie Simulation (LES)
Only the large scales are resolved, and small scales are modeled.
Active research.
LES quite popular in certain industries, but still very costly.
Reynolds-Averaged NavierStokes (RANS) I
Concept introduced by Reynolds in 1895. The most commonly used
approach in industry. Only time averages are considered, nonlinear eects
of the uctuations are modeled.
Use the decomposition
u(x, t) = u(x, t)
. .
time average
+ u
(x, t)
. .
uctuation
,
where the time average is
u(x, t) =
1
_
t+
t
u(x, ) d,
(spatial averaging or ensemble averaging can otherwise be used).
Reynolds-Averaged NavierStokes (RANS) II
Continuity equation for averaged quantities:
u = 0.
Momentum equations:
u
t
+(u )u +(u
)u
=
1
p +
2
u +f,
or, equivalently,
u
t
+(u )u =
_
pI +u u
_
+f,
Reynolds-Averaged NavierStokes (RANS) III
u
t
+(u )u =
_
pI +u u
_
+f,
u = 0.
The Reynolds stress term u
contains uctuations, and its eect

needs to be modeled.
Examples are Spalart-Allmaras, K , K , and SST.

CFD Notes

Uploaded by

Copyright:

Available Formats

CFD Notes

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CFD Notes

Uploaded by

Copyright:

Available Formats

Computational uid dynamics

Fluid ow is modeled by partial dierential equations (PDE),

Computational Fluid Dynamics (CFD) is the discipline of

Aerospace and aeronautical applications (airplanes, water and space

Mechanical applications (gas turbines, heat exchange, explosions,

Biological applications (blood ow, breathing, drinking)

Meteorological applications (weather prediction)

Environmental applications (air and water pollution)

and many more . . .

Pre-design of components: simulation vs. experiment (simulations

vehicles with lower fuel consumption, quieter, heavier loads

water and gas turbines (eectiveness)

Detection and prediction

hurricanes, storms, tsunamis

Pre-design of components: simulation vs. experiment (simulations

vehicles with lower fuel consumption, quieter, heavier loads

water and gas turbines (eectiveness)

Detection and prediction

hurricanes, storms, tsunamis

Pre-design of components: simulation vs. experiment (simulations

vehicles with lower fuel consumption, quieter, heavier loads

water and gas turbines (eectiveness)

Detection and prediction

hurricanes, storms, tsunamis

Robustness give a solution (for as many input cases as possible)

Reliability give a good solution

Performance give a good solution fast

Geometries give a good solution fast for real problems

Automatic tool chain reduce requirements of user interaction

CAD description of geometry

Problem set up - knowledgeable user needed

Mathematical model to use

Spatial and temporal discretization

Preprocessing - parse conguration and prepare solver

Solving - run the solver

Post processing - extract and compute information of interest

Visualization - often most important

Interpretation - physics, mathematics, numerics, experiments

CAD description of geometry

Problem set up - knowledgeable user needed

Mathematical model to use

Spatial and temporal discretization

Preprocessing - parse conguration and prepare solver

Solving - run the solver

Post processing - extract and compute information of interest

Visualization - often most important

Interpretation - physics, mathematics, numerics, experiments

CAD description of geometry

Problem set up - knowledgeable user needed

Mathematical model to use

Spatial and temporal discretization

Preprocessing - parse conguration and prepare solver

Solving - run the solver

Post processing - extract and compute information of interest

Visualization - often most important

Interpretation - physics, mathematics, numerics, experiments

CAD description of geometry

Problem set up - knowledgeable user needed