Acknowledgement: Simulation in Computer Graphics University of Freiburg

Acknowledgement
Computational Fluid Dynamics

An Introduction
This slide set is based on:

n
Simulation in Computer Graphics

University of Freiburg
John D. Anderson, Jr., Computational Fluid Dynamics

The Basics with Applications, McGraw-Hill, Inc.,
New York, ISBN 0-07-001685-2
WS 04/05
University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory
Motivation
Outline
n
n
n
introduction
pre-requisites
governing equations
n
n
n
D. Enright, S. Marschner, R. Fedkiw,

Animation and Rendering of
Complex Water Surfaces,
Siggraph 2002, ACM TOG,
vol. 21, pp. 736-744, 2002.
solution techniques
n
n
n
continuity equation
momentum equation
summary
Lax-Wendroff
MacCormack
numerical aspects
recent research in graphics

Introduction
n
physical aspects of fluid flow are governed by

three principles
n
n
n
mass is conserved
force = mass acceleration (Newtons second law)
energy is conserved (not considered in this lecture)
principles can be described with integral equations or

partial differential equations
in CFD, these governing equations are replaced by
discretized algebraic forms
algebraic forms provide quantities at discrete points
in time and space, no closed-form analytical solution
Continuous Quantities
n
n
Governing
Equations
Discretization
Navier-Stokes
Euler
Finite
Difference
System of
Algebraic
Equations
Discrete
Nodal
Values
Equation
Solver
Approx.
Solution
Lax-Wendroff
MacCormack
Problem
x, y, z 3D coordinate system
t time
( x,y,z,t ) density
v ( x,y,z,t ) velocity
v ( x,y,z,t ) = (u ( x,y,z,t ), v ( x,y,z,t ), w ( x,y,z,t ) )T
Numerical Solution - Overview
v (x,y,z,t )
(x,y,z,t )
v (x,y,z,t+t )
(x,y,z,t+t )
Substantial Derivative of
Outline
n
n
n
introduction
pre-requisites
governing equations
n
n
n
continuity equation
momentum equation
summary
solution techniques
n
n
n
Lax-Wendroff
MacCormack
numerical aspects
Substantial Derivative of
n
Taylor series at point 1

2 = 1 +
( x 2 x1 ) +
x 1
y

v +
w +
z 1
t 1
1
n
n
n
1 ( x1,y1,z1,t1 )
2 ( x2,y2,z2,t2 )
substantial derivative of
D

u
+v
+w
+
x
y
z
Dt
t
substantial derivative = local derivative + convective derivative
+ (v )
t
Dt
t2 t1
2 1

=
u +
t2 t1
x 1
y
v1(x1,y1,z1,t1) = (u (x1,y1,z1,t1 ),v (x1,y1,z1,t1 ),w (x1,y1,z1,t1 ))T

v2(x2,y2,z2,t2) = (u (x2,y2,z2,t2 ),v (x2,y2,z2,t2 ),w (x2,y2,z2,t2 ))T
Substantial Derivative

( y 2 y 1 ) +
(z 2 z 1 ) +
(t2 t1 )
z 1
t 1
1
y 2 y1 z 2 z 1
+
+
z 1 t 2 t1 t 1
1 t 2 t1
( x1,y1,z1 ) position at time t1

( x2,y2,z2 ) position at time t2
2 1 x 2 x1
=
+
t2 t1
x 1 t 2 t1 y
infinitesimally small fluid element moving with the flow

=
,
,
x y z
local derivative time rate of change at a fixed location

convective derivative time rate of change due to fluid flow
subst. derivative = total derivative with respect to time
Divergence of v
n
Outline
divergence of velocity v = time rate of change of the

volume V of a moving fluid element per unit volume
v =
u v w
1 D ( V
+
+
=
x y
z
V
Dt
n
n
n
introduction
pre-requisites
governing equations
n
n
n
solution techniques
n
n
n
Continuity Equation
n
n
n
time rate of change of the mass of the moving fluid

element is zero
D m
D ( V )
D
D V
=
= V
+
=0
Dt
Dt
Dt
Dt
Lax-Wendroff
MacCormack
numerical aspects

Continuity Equation
mass is conserved
infinitesimally small fluid element moving with the flow
fixed mass m, variable volume V, variable density
m = V
D
1 D V
+
=0
Dt
V Dt
continuity equation
momentum equation
summary
D
+ v = 0
Dt
substantial derivative
time rate of change of
density of a moving
fluid element
divergence of velocity time rate of change of volume

of a moving fluid element per
volume
D
+ v = 0
Dt
Continuity Equation
n
non-conservation form (considers moving element)
Continuity Equation
n
D
+ v = 0
Dt
manipulation
D
+ v =
+ v + v =
+ ( v )
Dt
t
t
conservation form (considers element at fixed location)
motivation for conservation form

n
n
n
n
+ ( v ) = 0
t
+ ( v ) = 0
t
time rate of mass

increase per volume
Outline
n
n
n
introduction
pre-requisites
governing equations
n
n
n
solution techniques
n
n
n
continuity equation
momentum equation
summary
Lax-Wendroff
MacCormack
numerical aspects
infinitesimally small element at a fixed location

mass flux through element
difference of mass inflow and outflow = net mass flow
net mass flow = time rate of mass increase
Momentum Equation
n
n
n
consider a moving fluid element

physical principle: F = ma (Newtons second law)
two sources of forces
n
body forces
n gravity
surface forces
n based on pressure distribution on the surface
n based on shear and normal stress distribution on the surface
due to deformation of the fluid element
friction
fluid
element
gravity
net mass flow

per volume
pressure
normal
stress
shear
stress
velocity
velocity
Pressure Force
Body Force
n
n
f = body force per unit mass

gravity: f = mg / m = g
body force on fluid element
f (dx dy dz )
n
n
consider the x component

pressure force
acts orthogonal
p dy dz
to surface into
the fluid element
dx
net pressure force in x direction
Stress
n
n
dy
dx dy dz
p p + dx dy dz =
x
x
dz
( p + p/ x dx ) dy dz
dx
dy
fluid
element
dz
Stress
normal stress related to the time rate of change of volume

shear stress related to the time rate of change
of the shearing deformation
jk stress in k direction applied to a surface perpendicular
to the j axis
xx normal stress in x direction
zx , yx shear stresses in x directions
yx
n
n
n
normal stress is related to pressure orthogonal to surface

shear stress is related to friction parallel to surface
friction and pressure are related to the velocity gradient
n
n
friction (shear stress) models viscosity (viscous flow)

in contrast to inviscid flow, where friction is not considered
friction
pressure
normal
stress
shear
stress
velocity
velocity
xx
x
Surface Forces in x
Momentum Equation in x
(yx + yx / y dy) dx dz
m a x = Fx
dz
p dy dz
( p + p/ x dx ) dy dz
xx dy dz
(xx + xx / x dx) dy dz
dy
dx
yx dx dz
Fx = p p + dx dy dz + xx + xx dx xx dy dz
x
x
zx not illustrated
dx dy dz
mass
+ yx + yx dy yx dx dz + zx + zx dz zx dy dz
z
y
p xx yx zx
+
dx dy dz
Fx =
+
+
y
z
x
x
n
n
viscous flow, non-conversation form

Navier-Stokes equations
yx zx
Du
p
= + xx +
+
+ fx
Dt
x x
y
z
Dv
p xy yy zy
= +
+
+
+ fy
Dt
y x
y
z
yz zz
Dw
p
= + xz +
+
+ fz
Dt
z
x
y
z
acceleration
time rate of change
of velocity of a
moving fluid element
surface force
pressure
normal stress
shear stress
body force
gravity
yx zx
p
Du
= + xx +
+
+ fx
x
x
y
z
Dt
Momentum Equation
Du p xx yx zx
dx dy dz + f x dx dy dz
+
= +
+
Dt x x
y
z
n
obtaining the conservation form of Navier-Stokes
Du
u
=
+ v u
Dt
t
( u)
u
=
+u
t
t
t
u
( u)
=u
t
t
t
( u v ) = u ( v) + ( v ) u
( v ) u = ( u v) u ( v)
Du ( u)
=
u + ( v) + ( u v )
Dt
t
t
continuity equation
Du ( u )
=
+ ( u v )
Dt
t
Momentum Equation
n
Navier-Stokes equations in conservation form

yx zx
( u )
p
+
+ fx
+ ( u v ) = + xx +
y
z
t
x x
( v )
p xy yy zy
+ fy
+
+
+ ( v v) = +
z
y
t
y x
Newtonian Fluids
n
xx = ( v ) + 2
n
n
n
n
n
n
intermolecular forces are negligible

R specific gas constant
T temperature
termal equation of state p = R T
u
x
v u
xy = yx = +
x y
yz zz
( w )
p
+
+ fz
+ ( w v ) = + xz +
y
z
t
z
x
Perfect Gas
Newton: shear stress in a fluid is proportional

to velocity gradients
most fluids can be assumed to be newtonian,
but blood is a popular non-newtonian fluid
Stokes:
yy = ( v ) + 2
v
y
u w
xz = zx = +
z x
zz = ( v ) + 2
w
z
w v
yz = zy = +
y z
is the molecular viscosity coefficient

is the second viscosity coefficient with = -2/3
yx zx
( u )
p
+ ( u v ) = + xx +
+
+ fx
t
x x
y
z
n
now,
p can be expressed using density and
can be expressed using velocity gradients,
fx is user-defined (e. g. gravity)
Governing Equations
Viscous Compressible Flow
n
n
n
introduction
pre-requisites
governing equations
n
n
n
n
n
continuity equation
momentum equation
summary
( v )
p xy yy zy
+ ( v v) =
+
+
+
+ fy
t
y
x
y
z
Lax-Wendroff
MacCormack
numerical aspects
yz zz
( w )
p
+
+ fz
+ ( w v ) = + xz +
y
z
t
z
x
Governing Equations
Inviscid Compressible Flow
Euler equations
Comments
on the Governing Equations
n
continuity equation
( u )
p
+ ( u v ) = + f x
t
x
( v )
p
+ ( v v ) = + f y
t
y
( w )
p
+ ( w v ) = + f z
t
z
n
n
momentum equation
+ ( v ) = 0
t
continuity equation
yx zx
( u )
p
+
+ fx
+ ( u v ) = + xx +
y
z
t
x x
thus, four equations and four unknowns , u, v, w
+ ( v ) = 0
t
solution techniques
n
momentum equation
Outline
coupled system of nonlinear partial differential equations

normal and shear stress terms are functions of the
velocity gradients
pressure is a function of the density
only momentum equations are Navier-Stokes equations,
however the name is commonly used for the full set
of governing equations (plus energy equation)
Governing Equations in
Conservation Form
n
can be obtained directly from a control volume fixed in

space
consider flux of mass, momentum (and energy) into and
out of the volume
have a divergence term which involves the flux of
mass (v), momentum in x, y, z (u v, v v, w v)
can be expressed in a generic form
Generic Form
viscous flow
?v
u
?w

u u + p xx
? u w t zx
u
? u v t yx
? fx
U=
F=
G =
H =
J
=
? f
v u xy
? v w + p t zy
v
? v v + p t yy
y
w
w u
? ww + p t
? f
? wv t
yz
xz
zz
inviscid flow

u
?v
u
uu + p
?u v
U=
F
=
G
=
vu
?v v + p
v
w
wu
?w v
U F G H
+
+
+
=J
t x y z
Generic Form
n
n
n
n
n
U solution vector (to be solved for)

F, G, H flux vectors
J source vector (external forces, energy changes)
U
F G H
=J
+
+
t
x y z
dependent flow field variables can be solved progressively
in steps of time (time-marching )
spatial derivatives are know from previous time steps
numbers can be obtained for density and flux variables
u, v, w
Motivation for Different Forms

of the Governing Equations
in many real-world problems, discontinuous changes in
the flow-field variables , v occur (shocks, shock waves)
problem: differential form of the governing equations
assumes differentiable (continuous) flow properties
simple 1-D shock wave: 1 u1 = 2 u2
?w
0
?u w
? fx
H =
J
=
? f
?v w + p
y
?w w + p
? f
conservation form solves for u, sees no discontinuity

positive effect on numerical accuracy and stability
10
Outline
n
n
n
introduction
pre-requisites
governing equations
n
n
n
continuity equation
momentum equation
summary
solution techniques
n
n
n
Lax-Wendroff Technique
Lax-Wendroff
MacCormack
numerical aspects
n
n

Lax-Wendroff
Example
n
flow-field variables , u, v, w are known at each discrete

spatial position ( x, y, z ) at time t
Lax-Wendroff computes all information at time t +t
second-order accuracy in space and time
explicit, finite-difference method
particularly suited to marching solutions
marching suited to hyperbolic, parabolic PDEs
unsteady (time-dependent), compressible flow
is governed by hyperbolic PDEs
Taylor Series Expansions
2D, inviscid flow, no body force, in non-conservation form
n
n
ti,j density at position (i, j ) and time t

ti,j , u ti,j , v ti,j are known
v

=
+u
+
+v
y
y
2

it,+j t = it, j +
t +
2
t i , j
t
u
u
u R T
= u
+v
+
t
y
x
x
2u
u
u it ,+j t = u it, j +
t + 2
t i , j
t
v
v
v R T
= u
+v
+
t
x
y
y
( t )2
+ ...
i, j 2
t
( t )
+ ...
i , j 2
t
2 v ( t )2
v
v it,+j t = v it, j +
+ ...
t + 2
t i , j
t i, j 2
t
11
Taylor Expansion - Density

2

it,+j t = it, j +
t + 2
t i, j
t
t
( t )
+ ...
i, j 2
t
/t can be replaced by spatial derivatives

given in the governing equations
=
+u
+
+v
t
x
y
y
x
2u
2
u
2 u
2v
v
2 v
=
+
+u
+
+
+
+v
+
t 2
x t x t
y t y t
y t y t
x t x t
using second-order central differences
t
t
vt
t
t u it+ 1, j u it 1, j
it 1, j
vit 1, j
it 1, j

+ u it, j i + 1, j
+ it, j i + 1, j
+ v it, j i + 1, j
= i , j
2 x
2 x
2 y
2 y
t i , j
2/t 2 can be obtained by differentiating the governing

equation with respect to time
mixed second derivatives can be obtained by differentiating

the governing equations with respect to a spatial variable,
e. g.
2
2
2
2
2
2
u u
u
R T
u
u v R T
= u
+
+v
+
2
+
x 2 x
x t
x
y
y
x
x
x 2
brackets missing on pp. 219, 220 in Andersons book!

n
2/t 2 can be computed using central differences for

spatial derivatives
higher-order terms are required, e. g.
t
u
u
2 =
i, j
2
t
i +1, j
2u + u
t
i, j
2
(x )
Outline
n
n
n
t
i 1, j
ut
+ ut
ut
ut
2u
= i+1, j +1 i 1, j1 i 1, j +1 i +1, j 1
4xy
xy i, j
introduction
pre-requisites
governing equations
n
n
solution techniques
n
n
n
... u, v are computed in the same way

continuity equation
momentum equation
summary
Lax-Wendroff
MacCormack
numerical aspects

12
MacCormack Technique
n
n
n
n
n
same characteristics like Lax-Wendroff

second-order accuracy in space and time
requires only first time derivative
predictor corrector
illustrated for density
predictor step for density
t
t
vt
t
t u it +1 , j u it1 , j
it1, j
v it 1, j
it 1, j

+ u it, j i + 1, j
+ it, j i + 1, j
+ v it, j i + 1, j
= i , j
2
x
2
x
2
y
2
y
t i, j

it,+j t = it, j +
t
t i , j
t
av

it,+j t = it, j +
t
t i, j
u, v are predicted the same way

u
u it, +j t = u ti , j +
t
t i , j
t
Comments
corrector step for density

t + t
i , j
u
= it,+j t
t + t
i +1 , j
2 x
t + t
i 1 , j
+ u it,+j t
t + t
i +1 , j
2x
t+ t
i 1 , j
+ it,+j t
av
t

1

t
it,+j t = it, j +
t = i, j +
+
t
t
i, j
i , j t

t+ t
i +1 , j
2 y
t + t
i 1 , j
+ v it,+j t
t + t
i +1 , j
2 y
t + t
i 1, j
t + t
i , j
corresponds to Heun scheme for ODEs

other higher-order schemes, e. g. Runge-Kutta 2 or 4,
could be used as well
Lax-Wendroff and MacCormack can be used for

unsteady flow
n
n
n
n
n
n
v
v it,+j t = vit, j +
t
t i, j
non-conservation form
conservation form
viscous flow
inviscid flow
higher-order accuracy required to avoid numerical

dissipation, artificial viscosity, numerical dispersion
note, that viscosity is represented by second partial
derivatives in the governing equations
13
Outline
n
n
n
introduction
pre-requisites
governing equations
n
n
n
solution techniques
n
n
n
continuity equation
momentum equation
summary
Lax-Wendroff
MacCormack
numerical aspects
Numerical Effects
n
n
n
n
truncation error causes dissipation and dispersion

numerical dissipation is caused by even-order terms
of the truncation error
numerical dispersion is caused by odd-order terms
leading term in the truncation error dominates the
behavior
original function
Numerical Effects
n
artificial viscosity compromises the accuracy, but

improves the stability of the solution
adding artificial viscosity increases the probability of
making the solution less accurate, but improves
the stability
similar to iterative solution schemes for implicit integration
schemes, where a smaller number of iterations introduces
artificial viscosity, but improves the stability
x
dissipation
x
dispersion
Summary
n
governing equations for compressible flow

n
n
n
continuity equation, momentum equation

Navier-Stokes (viscous flow) and Euler (inviscid flow)
discussion of conservation and non-conservation form
explicit time-marching techniques

n
n
n
Lax-Wendroff
MacCormack
numerical aspects
14
Recent Research
Recent Research
M. Carlson, P. J. Mucha, G. Turk, Rigid Fluid: Animating the

Interplay Between Rigid Bodies and Fluid, Siggraph 2004.
T. G. Goktekin, A. W. Bargteil, J. F. OBrien,

A Method for Animating Viscoelastic Fluids, Siggraph 2004.
15

Acknowledgement: Simulation in Computer Graphics University of Freiburg

Uploaded by

Document Information

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

Acknowledgement: Simulation in Computer Graphics University of Freiburg

Uploaded by

Copyright:

Available Formats

Acknowledgement

Computational Fluid Dynamics

This slide set is based on:

Simulation in Computer Graphics

John D. Anderson, Jr., Computational Fluid Dynamics

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

D. Enright, S. Marschner, R. Fedkiw,

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

recent research in graphics

physical aspects of fluid flow are governed by

principles can be described with integral equations or

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

v ( x,y,z,t ) = (u ( x,y,z,t ), v ( x,y,z,t ), w ( x,y,z,t ) )T

Numerical Solution - Overview

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

recent research in graphics

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

Taylor series at point 1

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

v1(x1,y1,z1,t1) = (u (x1,y1,z1,t1 ),v (x1,y1,z1,t1 ),w (x1,y1,z1,t1 ))T

( x1,y1,z1 ) position at time t1

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

infinitesimally small fluid element moving with the flow

local derivative time rate of change at a fixed location

divergence of velocity v = time rate of change of the

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

time rate of change of the mass of the moving fluid

recent research in graphics

divergence of velocity time rate of change of volume

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

non-conservation form (considers moving element)

motivation for conservation form

time rate of mass

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

recent research in graphics

infinitesimally small element at a fixed location

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

consider a moving fluid element

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

net mass flow

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

f = body force per unit mass

consider the x component

net pressure force in x direction

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

normal stress related to the time rate of change of volume

normal stress is related to pressure orthogonal to surface

friction (shear stress) models viscosity (viscous flow)

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

viscous flow, non-conversation form

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

obtaining the conservation form of Navier-Stokes

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

Navier-Stokes equations in conservation form

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

intermolecular forces are negligible

University of Freiburg - Institute of Computer Science - Computer Graphics Laboratory

Newton: shear stress in a fluid is proportional

is the molecular viscosity coefficient