Академический Документы
Профессиональный Документы
Культура Документы
dynamics
Applications to human thermodynamics
Prof. Dr.-Ing. habil. Nikolai Kornev; Prof. Dr.-Ing. habil. Irina
Cherunova
Prof. Dr.-Ing. habil. Nikolai Kornev & Prof. Dr.-Ing. habil. Irina Cherunova
Contents
Contents
List of Tables
10
List of Figures
11
Preface
15
16
17
1.1
17
1.2
25
27
2.1
27
2.2
30
2.3
31
2.4 Exercises
31
DISCOVER
OUR WORLD
Some people know precisely where they want to go. Others seek the adventure of
discovering uncharted territory. Whatever you want your professional journey to be,
youll nd what youre looking for at Oliver Wyman.
Contents
33
3.1
Artificial viscosity
33
3.2
35
3.3 Exercise
4
37
38
4.1 Theory
38
4.2
Mixed schemes
39
4.3
Staggered grid
4.4
Approximation of
4.5
Approximation of
4.6
47
4.7
47
4.8
47
4.9
48
4.10
48
41
un un
xi j j 44
n
ui
xj xj
47
Contents
49
5.1
49
5.2
51
5.3
53
55
6.1
55
6.2 Sample
56
6.3
Explicit scheme
59
6.4
Implicit scheme
61
6.5
61
6.6
64
6.7
SIMPLE method
65
70
7.1
SIMPLE algorithm
70
7.2
PISO algorithm
70
7.3
SIMPLEC algorithm
73
89,000 km
Thats more than twice around the world.
careers.slb.com
Based on Fortune 500 ranking 2011. Copyright 2015 Schlumberger. All rights reserved.
Contents
Computational grids
74
8.1
Grid types
74
8.2
75
8.3
Morphing grids
75
II
77
Physics of turbulence
78
9.1
78
9.2
Vortex dynamics
78
9.3
Experimental observations
86
10
99
10.1
Reynolds averaging
99
10.2
100
10.3
100
10.4
Structure functions
109
Hellmanns is one of Unilevers oldest brands having been popular for over 100 years.
If you too share a passion for discovery and innovation we will give you the tools and
opportunities to provide you with a challenging career. Are you a great scientist who
would like to be at the forefront of scientic innovations and developments? Then you will
enjoy a career within Unilever Research & Development. For challenging job opportunities,
please visit www.unilever.com/rdjobs.
Contents
11
114
11.1
Physical background
114
11.2
Dissipation rate
116
11.3
Kolmogorov hypotheses
117
11.4
120
11.5
123
11.6
123
12
128
13
134
13.1
134
14
139
14.1
139
14.2
141
14.3
144
15
147
15.1
LES filtering
147
15.2
LES equations
149
15.3 Smagorinskymodel
150
15.4
153
15.5
155
15.6
156
15.7
158
16
160
16.1 Introduction
160
16.2
161
16.3
16.4
164
167
Contents
III
17
171
172
17.1 Designations
172
17.2 Introduction
173
17.3
174
17.4 Mathematicalmodel
175
17.5 Results
180
17.6 Discussion
185
18
19
187
CFD application for design of cloth for protection from low temperatures
under wind conditions. Influence of the wind on the cloth deformation
and heat transfer from the body
190
19.1
190
19.2
191
19.3
192
19.4
193
20
196
Bibliography
199
Index
203
List of Tables
List of Tables
4.1 Limiters function for TVD schemes
6.1
40
57
58
58
6.4
58
58
59
149
152
16.1 Results of the resistance prediction using different methods. CR is the resistance
coefficient, CP is the pressure resistance and CF is the friction resistance
170
175
177
17.3 Radii of layers used in simulations in fraction of the skin thickness 178
17.4 Thermodynamic coefficients of the body layers used in simulations
181
17.5 Coefficient K depending on the test person feelings and energy expenditure
E = M/A .(W/m2). A is the body surface (m2) and M is the work (W) 185
18.1 Heat flux from the diver depending on the cloth contamination
188
195
198
10
List of Figures
List of Figures
1.1
18
1.2
19
1.3
2.1
27
2.2
30
4.1
42
4.2
42
4.3
43
6.1
55
6.2
SIMPLE algorithm
65
6.3
68
8.1 Samples of a) structured grid for an airfoil, b) block structured grid for
cylinder in channel and c) unstructured grid for an airfoil
74
8.2
74
9.1
80
9.2 Tornado
81
9.3
81
82
9.5
83
9.6
84
9.7
84
9.8
86
9.9
87
9.10
88
88
9.12
90
9.13
90
9.14
90
9.15
91
9.16
91
9.17
92
9.18 Fine vortex structures in a confined jet mixer flow. PLIF measurements by
Valery Zhdanov (LTT Rostock). Spatial resolution is 31m 93
9.19
11
94
9.20
List of Figures
94
9.21 Conceptual model of the organization of the turbulence close to the wall
proposed by Adrian et al. (2000)
95
boundary layer
95
97
U C D ux =u 98
101
10.2 Distribution of the integral length of the scalar field along the jet
10.3
mixer centerline
103
104
105
106
10.6 Kurtosis of the structure function for the concentration of the scalar field
obtained in the jet mixer
110
116
116
11.4 Distribution of the Kolmogorov scale along the centerline of the jet mixer
and free jet. The dissipation rate " is calculated from the k " model and
11.5
120
11.6 Three typical ranges of the energy density spectrum in the turbulent flow at
high Reynolds number. 1- energy containing range, 2- inertial subrange, 3dissipation range
121
122
122
123
12
124
List of Figures
125
15.1
148
15.2
155
16.1
161
16.2
164
16.3
164
16.4 The division of the computational domain into the URANS (dark) and
LES (light) regions at one time instant for hybrid calculation of tanker
165
16.5
170
17.1
173
176
176
182
17.5 Temperature distributions around the body with continuous ice distribution
and with ice briquettes. Results of numerical simulations after 60 minutes
182
17.6 Development of the averaged temperature in the air gap between the
underwear and the ice protection on the human chest. Comparison
between the measurement (solid line) and the numerical simulations
(dotted line)
183
17.7 Test person weared overheating protection jacket (left) and distribution
17.8
184
186
18.1
Left: Heat transfer coecient at air speed of 1m/s. Right: Whole body
convective heat transfer coefficient hc from various published works.
18.2
The figure is taken from [2]. Blue crosses show results of the present work
188
189
19.1 Human body model in wind tunnel of the Rostock university (left).
Positions of measurement points (right)
190
13
193
19.3
List of Figures
194
195
20.1
197
20.2
197
20.3
198
14
Preface
Preface
The present book is used for lecture courses Computational heat and mass
transfer, Mathematical models of turbulence and Design of special cloth given
by the authors at the University of Rostock, Germany and Don State Technical University, Russia. Each of lecture courses contains about 14 lectures.
The lecture course Compuational heat and mass transfer was written proceeding from the idea to present the complex material as easy as possible.
We considered derivation of numerical methods, particularly of the finite volume method, in details up to final expressions which can be programmed.
Turbulence is a big and a very complicated topic which is difficult to cover
within 14 lectures. We selected the material combining the main physical concepts of the turbulence with basic mathematical models necessary to
solve practical engineering problems. The course Design of special cloth uses
the material of two parts of this book partially. The material for the third
part was gathered from research projects done by the authors of this book
within some industrial projects and research works supported by different
foundations. We express our gratitude to Andreas Gross, Gunnar Jacobi
and Stefan Knochenhauer who carried out CFD calculations for the third
part of this book.
360
thinking
360
thinking
360
thinking
Dis
Part I
Introduction into
computational methods for
solution of transport equations
16
15
Chapter 1
Chapter
1
Chapter
1
Main equations
of the
Main
of
Computational
Main equations
equations Heat
of the
theand Mass
Computational
Transfer
Computational Heat
Heat and
and Mass
Mass
Transfer
Transfer
1.1
1.1
mechanics
1.1.1 Fluid
Continuity
equationequations
1.1 Fluid
mechanics
equations
We
consider
the case of uniform
density distribution D const. The con1.1.1
Continuity
equation
tinuity
has the following
physical meaning: The amount of liquid
1.1.1 equation
Continuity
equation
We consider
thevolume
case ofUuniform
D const.
conflowing
into the
with thedensity
surfacedistribution
S is equal to the
amountThe
of liquid
We
consider
the
case
of
uniform
density
distribution
D
const.
The
continuity equation
has the following
meaning:
The amount of liquid
flowing
out. Mathematically
it can physical
be expressed
in form:
tinuity equation
has theUfollowing
physical meaning: The amount of liquid
flowing
into the volume
with the
Z surface S is equal to the amount of liquid
flowing
into
the
volume
U
with
the
S is equal
to the amount of liquid
flowing out. Mathematically it can surface
be expressed
in form:
u
E
n
E
ds
D
0
(1.1)
flowing out. Mathematically it can be expressed in form:
Z
S
Z
uE nE ds D 0
(1.1)
uE nE ds D components
0
(1.1)
Expressing the scalar product uE nE through
S
S
Z
Expressing the scalar product uE nE through components
Cu
uEynE cos.ny/
uz cos.nz/ ds D 0:
ux cos.nx/
Expressing the scalar
product
throughCcomponents
Z
S
u cos.nx/ C u cos.ny/ C u cos.nz/ds D 0:
Z
x
y
z
and using the Gauss
theorem
get
Cwe
uy cos.ny/
C uz cos.nz/ ds D 0:
ux cos.nx/
S
S
Z
@uy
@uz
@ux we get
and using the Gauss theorem
C get C
dU D 0
and using the Gauss theorem
we
@y
@z
Z @x
U
Z @ux C @uy C @uz d U D 0
@uz
@u
y
@xxUCis@u
@y
Since the integration volume
arbitrary,
C @z the
d Uintegral
D 0 is zero only if
U
@x
@y
@z
U
@uz the integral is zero only if
@uxU is@u
y
Since the integration volume
17
C arbitrary,
C
D0
(1.2)
Since the integration volume
arbitrary,
@xU is @y
@z the integral is zero only if
17
In the tensor form the continuity equation
reads:
17
@ui
D0
@xi
1.1.2
(1.3)
The inner forces acting in a fluid are subdivided into the body forces and
surface forces (Fig. 1.1).
Download free eBooks at bookboon.com
17
1.1.2
Figure 1.1: Body and surface forces acting on the liquid element.
1.1.2.1
Body forces
Let fE be a total body force acting on the volume U . Let us introduce
the strength of the body force as limit of the ratio of the force to the volume:
fE
U !0 U
FE D lim
(1.4)
1
which has the unit kgs 2m m
D ms 218
. Typical body forces are gravitational,
kg m3
electrostatic or electromagnetic forces. For instance, we have the following
relations for the gravitational forces:
3
fE D gU kE
(1.5)
gU kE
FE D lim .
/ D g kE
U !0
U
(1.6)
where fE is the gravitational force acting on a particle with volume U . The
strength of the gravitational force is equal to the gravitational acceleration:
The body forces are acting at each point of fluid in the whole domain.
1.1.2.2
Surface forces
The surface forces are acting at each point at the boundary of the fluid
element. Usually they are shear and normal stresses. The strength of surface
forces is determined as
Download free eBooks at bookboon.com
PEn
18
S !0 S
pEn D lim
(1.7)
kg
with the unit kgs 2m m12 D ms
2 . A substantial feature of the surface force is the
dependence of pEn on the orientation of the surface S.
The surface forces are very important because they act on the body from
1.1.2.2
Surface forces
The surface forces are acting at each point at the boundary of the fluid
Main equations of the Computational
element. Usually they are shear and normal stresses. The strength ofHeat
surface
Lectures on computational fluid dynamics
and Mass Transfer
forces is determined as
PEn
S !0 S
pEn D lim
(1.7)
kg
with the unit kgs 2m m12 D ms
2 . A substantial feature of the surface force is the
dependence of pEn on the orientation of the surface S.
The surface forces are very important because they act on the body from
the side of liquid and determine the forces RE arising on bodies moving in the
fluid:
RE D
E D
M
S
Z
pEn dS
(1.8)
.Er pEn /dS
1.1.2.3
d uE
D U FE C pEn S pEx Sx pEy Sy pEz Sz
dt
(1.9)
19
Download
freethe
eBooks
bookboon.com
Let
us find
limitat of
(1.10) at S ! 0:
U
S19x
D 0; lim
D cos.nx/;
S !0 S
S !0 S
lim
lim
Sy
D cos.ny/; lim
Sz
D cos.nz/
(1.11)
(1.12)
Dividing r.h.s and l.h.s. by the surface of inclined face S results in:
Sx
Sy
Sz
U d uE
FE D pEn pEx
pEy
pEz
(1.10)
S dt
S
S
S
U
Sx
D 0; lim
D cos.nx/;
S !0 S
S !0 S
lim
(1.11)
Sy
Sz
D cos.ny/; lim
D cos.nz/
(1.12)
S !0 S
S !0 S
Substitution of (1.11) and (1.12) into (1.10) results in the following relation
between pEn and pEx , pEy , pEz :
lim
(1.13)
20
Here ij are shear stress (for instance 12 D xy ), whereas pi i are normal
stress (for instance p11 D pxx ). From moment equations (see Fig. 1.3) one
can obtain the symmetry condition for shear stresses: zy a yz a D 0 )
zy D yz and generally:
ij D j i
(1.14)
1.1.3
(1.15)
Applying the Newton second law to the small fluid element d U with the
surface dS and using the body and surface forces we get:
Z
d uE
d U D
dt
FE d U C
pEn dS
(1.16)
The property of the surface force can be rewritten with the Gauss theorem
in the following form:
Z
Z
pEn dS D
pEx cos.nx/ C21pEy cos.ny/ C pEz cos.nz/ dS
S
Z
S
@pEx
@pEy
@pEz
C
C
@x
@y
@z
dU
d uE
d U D
dt
FE d U C
Z
21
@pEx
@pEy
@pEz
C
C
@x
@y
@z
dU
pEn dS D
S
pEx cos.nx/ C pEy cos.ny/ C pEz cos.nz/ dS
Z
S
@pEx
@pEy
@pEz
D fluid dynamics
C
C
Lectures on computational
@x
@y
@z
U
dU
Since the volume d U is arbitrary, the l.h.s. in the last formulae is zero only
if:
1
@
p
E
@
p
E
@
p
E
d uE
x
y
z
D FE C
C
C
(1.17)
dt
@x
@y
@z
The stresses in (1.17) are not known. They can be found from the generalized
Newton hypothesis
1
1
0
0
p 0 0
pxx xy xz
@ xy pyy yz A D @ 0 p 0 A C 2Sij
(1.18)
0 0 p
xz yz pzz
where p is the pressure,
S11
S21
S31
@uy
@uz
@ux
1 @ux
1 @ux
I S12 D Sxy D
C
I S13 D Sxz D
C
D Sxx D
@x
2 @y
@x
2 @z
@x
@uz
@uy
1 @uy
; S23 D Syz D
C
D S12 ; S22 D Syy D
@y
2 @z
@y
@uz
D S13 ; S32 D S23 ; S33 D Szz D
@z
@ux
@uy
@uz
; pyy D p C 2
; pzz D p C 2
@x
@y
@z
The sum of three normal stresses doesnt depend on the choice of the coordinate system and is equal to the pressure taken with sign minus:
pxx C pyy C pzz
D p
3
(1.19)
The last expression is the definition of the pressure in the viscous flow: The
pressure is the sum of three normal stresses taken with the sign minus. Substitution of the Newton hypothesis (1.18) into (1.17) gives (using the first
Download free
at bookboon.com
equation
as aeBooks
sample):
dux
@
D Fx C
dt
@x
22
@ux
p C 2
@x
@
@uy
@ux
C
C
@y
@x
@y
The sum of three normal stresses doesnt depend on the choice of the coordinate system and is equal to the pressure taken with sign minus:
Lectures on computational fluid dynamics
pxx C pyy C pzz
D p
(1.19)
The last expression is the definition of the pressure in the viscous flow: The
pressure is the sum of three normal stresses taken with the sign minus. Substitution of the Newton hypothesis (1.18) into (1.17) gives (using the first
equation as a sample):
dux
@ux
@
@uy
@ux
@
D Fx C
p C 2
C
C
dt
@x
@x
@y
@x
@y
@
@ux
@uz
C
C
D
@z
@z
@x
2
@ ux
@p
@2 ux
@2 ux
D Fx
C
C
C
C
@x
@x 2
@y 2
@z 2
@ @ux
@uy
@uz
C
C
C
@x @x
@y
@z
The last term in the last formula is zero because of the continuity equation.
Doing similar transformation with resting two equations in y and z directions, one can obtain the following equation, referred to as the Navier-Stokes
equation:
d uE
1
D FE rp C E
u
dt
(1.20)
The full or material substantial derivative of the velocity vector ddtuE is the
acceleration of the fluid particle. It consists of two parts: local acceleration
and convective acceleration:
The Wake
d uE
dt
@E
u
@t
@E
u
@x
@E
u
@y
D
C uto
C uy behind
C uz
the only emission
we want
x leave
local acceleration
@E
u
@z
convective acceleration
23
.QYURGGF'PIKPGU/GFKWOURGGF'PIKPGU6WTDQEJCTIGTU2TQRGNNGTU2TQRWNUKQP2CEMCIGU2TKOG5GTX
6JGFGUKIPQHGEQHTKGPFN[OCTKPGRQYGTCPFRTQRWNUKQPUQNWVKQPUKUETWEKCNHQT/#0&KGUGN6WTDQ
2QYGTEQORGVGPEKGUCTGQHHGTGFYKVJVJGYQTNFoUNCTIGUVGPIKPGRTQITCOOGsJCXKPIQWVRWVUURCPPKPI
HTQOVQM9RGTGPIKPG)GVWRHTQPV
(KPFQWVOQTGCVYYYOCPFKGUGNVWTDQEQO
23
tions, one can obtain the following equation, referred to as the Navier-Stokes
equation:
Main equations of the Computational
Heat
and Mass Transfer
(1.20)
d uE
1
Lectures on computational fluid dynamics
D FE rp C E
u
dt
The full or material substantial derivative of the velocity vector ddtuE is the
acceleration of the fluid particle. It consists of two parts: local acceleration
and convective acceleration:
d uE
D
dt
@E
u
@t
local acceleration
@E
u
@E
u
@E
u
C ux
C uy
C uz
@x
@y
@z
convective acceleration
@
ui
@xj
(1.23)
The Navier Stokes equation together with the continuity equation (1.3) is
the closed system of partial differential equations. Four unknowns velocity
components ux ; uy ; uz and pressure p are found from four equations. The
equation due to presence of the term @x@j .ui uj / is nonlinear.
The boundary conditions are enforced for velocity components and pressure
at the boundary of the computational domain. The no slip condition ux D
uy D uz D 0 is enforced at the solid body boundary. The boundary condition
for the pressure at the body surface can directly be derived from the Navier
Stokes equation. For instance, if y D 0 corresponds to the wall, the Navier
Stokes Equation takes the form at the boundary:
@p
@ 2 ux
D Fx C 2
@x
@y
@p
@2 uy
D Fy C 2
@y
@y
@2 uz
@p
D Fz C 2
@z
@y
Very often the last term in the last formulae is neglected because second
spatial derivatives of the velocity are not known at the wall boundary.
Till now, the existence of the solution of Navier Stokes has been not proven by
mathematicians. Also, it is not clear whether the solution is smooth or allows
Download free The
eBooks
at bookboon.com
singularity.
Clay
Mathematics Institute has called the NavierStokes
existence and smoothness problems one of the seven most important open
24
24
@x
@y
@p
@2 uy
D Fy C 2
@y
@y
@2 uz
D Fz C 2
@y
@z
Very often the last term in the last formulae is neglected because second
spatial derivatives of the velocity are not known at the wall boundary.
Till now, the existence of the solution of Navier Stokes has been not proven by
mathematicians. Also, it is not clear whether the solution is smooth or allows
singularity. The Clay Mathematics Institute has called the NavierStokes
existence and smoothness problems one of the seven most important open
problems in mathematics and has offered one million dollar prize for its
solution.
24
1.2
Let q.x; t / be the heat flux vector, U is the volume of fluid or solid body, S is
its surface and n is the unit normal vector to S. Flux of the inner energy
into the volume U at any point x 2 U is
q.x; t/ n.x/
(1.24)
(1.25)
(1.26)
From
R the other side the change of the inner energy in the volume U is equal
to U cp @t@ T .x; t/d U , where T is the temperature, cp is the specific heat
capacity and is the density. Equating this change to (1.26) we get:
Z
@
cp T .x; t/d U D
@t
r q.x; t/d U C
f .x; t/d U
(1.27)
(1.28)
Z
!
25
25
(1.29)
The following boundary conditions are applied for the heat conduction equation (1.30):
Neumann condition:
rT .x; t/ n.x/ D F1 .x; t/;
x2S
(1.32)
Dirichlet condition:
T .x; t/ D F2 .x; t/;
x2S
(1.33)
GOT-THE-ENERGY-TO-LEAD.COM
We believe that energy suppliers should be renewable, too. We are therefore looking for enthusiastic
new colleagues with plenty of ideas who want to join RWE in changing the world. Visit us online to nd
out what we are offering and how we are working together to ensure the energy of the future.
26
26
Chapter 2
Chapter
2
Chapter
2
Finite difference
method
Finite
Finite difference
difference method
method
2.1
Let us consider
the finite difference case
method for the one dimensional case.
2.1
One dimensional
2.1
dimensional
case
Let '.x/One
is the function
defined in the
range 0; a along the x axis. The
Let
us
consider
the
finite
difference
method
thethe
one
dimensionaldistricase.
section 0; a is subdivided in a set of points xifor
. For
homogeneous
Let
us
consider
the
finite
difference
method
for
the
one
dimensional
Let '.x/
range
a Fig.
along2.1).
the x axis. case.
The
bution
xi is
D the
.i function
1/I i Ddefined
1; N , inD the
a=.N
1/0;
(see
Let
'.x/
is
the
function
defined
in
the
range
0;
a
along
the
x
axis.
The
section 0; a is subdivided in a set of points xi . For the homogeneous distrisection x0;
a is subdivided in a set of points xi . For
distribution
(see the
Fig.homogeneous
2.1).
i D .i 1/I i D 1; N , D a=.N 1/
bution xi D .i 1/I i D 1; N , D a=.N 1/ (see Fig. 2.1).
Chapter 2
Let us consider the finite difference method for the one dimensional case.
Let '.x/ is the function defined in the range 0; a along the x axis. The
section 0; a is subdivided in a set of points xi . For the homogeneous distribution xi D .i 1/I i D 1; N , D a=.N 1/ (see Fig. 2.1).
Figure 2.1: One dimensional case.
@' 1
Figure
2.1: One
dimensional
case.
Let us approximate the
derivative
. The Taylor
series of the function '
@x
Figure
2.1:
One
dimensional
case.
at points xi 1 and xi C1 are:
@' 1
Let us approximate the derivative
. The Taylor
of the function '
@x 1
@'
2 series
Letpoints
us approximate
derivative
.
The
Taylor
series
of the function '
'
@'
1
@
at
xi 1 and xthe
are:
C1
@x
'are:
C x 2
:::
(2.1)
'i 1xiD
i x
2
at points xi 1 and
i C1
@x i 2
@x
i
@'
1 2 @2 '
@22' only
1
D
'
x
C
::: one variable (2.1)
'
i
1
i
We use the partial derivative although
the function
@'
121 xdepends
@'
@ '2 i on
2
@x
@x
2
i
x
D
'
x
C
(2.1)
'
i
1
i
'i C1 D 'i C x @x C 2 x @x 22 C :::
:::
(2.2)
@x
2
@x
i
i
i
i
1
We use the partial derivativealthough
the
function
depends
only
on
one
variable
27
1
We use the partial derivative although
the function depends only on one variable
@'
Expressing the derivative @x from (2.1) we get the Backward Difference
27
i
Scheme (BDS):
Figure 2.1: One27
dimensional case.
1 @' 1
@'
'i .'The
D
C O x
(2.3)
Let us approximate the derivative
series of the function
'
i 1 Taylor
@x i
x @x
at points xi 1 and xi C1 are:
(2.2) weget
from
Expressing the derivative @'
@x
@'
1 2 @2 'the Forward Difference
C x
:::
(2.1)
'i 1 D 'i x i
@x i 2
@x 2 i
Scheme (FDS):
the function
depends
only
on one variable
although
1
We use the partial derivative
1
@'
D
'i C1 'i C O x
(2.4)
x
@x i
Download free eBooks at bookboon.com
27
Accuracy of both schemes is of the first order. Subtracting (2.1) from (2.2)
27
we get the Central Difference Scheme (CDS)
@'
1
2
D
'i C1 'i 1 C O x
(2.5)
@x i
2x
1
'i 'i 1 C O x
(2.3)
x
i
Expressing the derivative @'
from (2.2) we get the Forward Difference
@x
@'
@x
Scheme (FDS):
@'
@x
1
D
'i C1 'i C O x
x
(2.4)
Accuracy of both schemes is of the first order. Subtracting (2.1) from (2.2)
we get the Central Difference Scheme (CDS)
@'
1
2
D
(2.5)
'i C1 'i 1 C O x
@x i
2x
(2.6)
The accuracy of BDS, FDS and CDS can be improved using the polynomial
representation of the function '.x/. For instance, consider the approximation
'.x/ D ax 2 C bx C c
@'
@x
1
3
D
2'i C1 C 3'i 6'i 1 C 'i 2 C O x
6x
(2.7)
@'
@x
1
D
6x
28
3
C O x
(2.8)
1 fluid dynamics
Lectures on@'
computational
@x
6x
3
C O x
(2.7)
@'
@x
1
D
6x
3
C O x
(2.8)
@'
@x
1
D
12x
(2.9)
for the Central Difference Scheme. As seen the accuracy order is sufficiently
improved by consideration of more adjacent points.
The second derivatives are:
2
@ '
1
2
D
'i C1 2'i C 'i 1 C O x
(2.10)
@x 2 i
x 2
for the polynomial of the second order and
@2 '
@x 2
1
D
12x 2
4
C O x
(2.11)
29
CAREERKICKSTART
An app to keep you in the know
Whether youre a graduate, school leaver or student, its a difcult time to start your career.
So here at RBS, were providing a helping hand with our new Facebook app. Bringing together
the most relevant and useful careers information, weve created a one-stop shop designed
to help you get on the career ladder whatever your level of education, degree subject or
work experience.
And its not just nance-focused either. Thats because its not about us. Its about you.
So download the app and youll get everything you need to know to kickstart your career.
So what are you waiting for?
Click here to get started.
29
for the polynomial of the fourth order. The formula (2.10) can also be obtained using consequently CDS
2
1 @'
@'
@ '
D
(2.12)
@x 2 i
x @x i C1=2 @x i 1=2
where i C 1=2 and i 1=2 are intermediate points (see Fig. 2.1). Using again
the CDS for the derivatives at intermediate points:
@'
'i C1 'i
(2.13)
D
@x i C1=2
x
@'
'i 'i 1
(2.14)
D
@x i 1=2
x
we obtain (2.10).
2.2
In the two dimensional case the function ' is the function of two variables ' D
'.x; y/. A sample of non-uniform grid is given in Fig. (2.2). In next chapters
we will consider different grids and principles of their generation. In this
chapter we consider uniform two dimensional grids .xi ; yj / with equal spacing
in both x and y directions.
ij
'ij C1 'ij 1
@'
D
Download free eBooks at bookboon.com
2y
@y ij
2.3
@'
@x
30
@'
Lectures on computational fluid dynamics
whereas on y is:
2.3
@x
@'
@y
ij
ij
'ij C1 'ij 1
2y
(2.16)
(2.17)
(2.18)
2.4
Exercises
(2.19)
(2.20)
@'
.x D 0/ D C1
@x
@'
@2 '
C 2 D f .x; t/
@t
@x
Use the explicit method and the central difference scheme for spatial
derivatives.
32
32
3.1
Artificial viscosity
Let
the genericviscosity
linear equation:
3.1us consider
Artificial
3.1
Artificial viscosity
@
@ equation:
Let us consider the generic linear
C
u
D0
Let us consider the generic linear
@t equation:
@x
@
@
@ C u @ D 0
@t C u @x D 0
The numerical upwind scheme @t
(UDS) @x
is:
8 u n u n
i1
The numerical
(UDS)
i xscheme
u > 0is:
<upwind
TheinC1
numerical
upwind
scheme (UDS) is:
in
D 8 u n u n
t
in
i1
u>0
n u
n n
:
uiC1
< u
i
x i1
i u
<0
u>
inC1 in
x
<
x
n D
inC1
ti
n
Donly
: the
uiC1
uinu > 0:
We consider
case
t
n
u<0
: uiC1
uin
nC1x
u<0
x n
i u > 0:u in u in1
We consider only the
case
i
u>0
We consider only thet
case uD
>
0:
x
nC1
n
n
function
u .x;
inn t/uin time
Taylor expansionsinC1
of the
in
in1 and space gives
D
u>0
u u
i
t i D i x i 1
u>0
2 n
2
t
n x
@ .x;
t
@
Taylor expansions of
the function
t/
in
time
and
space gives
nC1
n
t C
C
Dfunction
i C .x;
:::
i the
Taylor expansions of
t/
in
time
and
2
@t in
@t
2 in space gives
@2 t 2
@
n
n n t C 2 n 2 n C
D
C
inC1
@
@ t 2 :::
@
n nC1nD in C @t t C @t 2 2 x
::: :::
i D
i CC
i
i i 1 C
i x C @x
@t 22 in1
2 i2
@x @t
in1i
@ n
@22 n x 22
n
n
in Dand
i 1 C @into
x Cresults
@ in x C :::
Substitution of (3.2)
2 C :::
i D in1(3.3)
C @x i 1(3.1)
x C @x 22 i 1
@x i 1 2
@xn i 1
n
@233
t
@
D
C
@t i
@t 2 i 2
33
n
x 2
@ n 33
@2
u
C
x C
D
x
@x i 1
@x 2 i 1 2
(3.1)
(3.1)
(3.1)
(3.2)
(3.2)
(3.2)
(3.3)
(3.3)
(3.3)
(3.4)
@3 x 2
@ n @2
@ n
:::;
x C
D
@x 3 i 2
@x 2 i
@x i
@x i 1
n
n
n
Download free eBooks
@2
@3
@2 at bookboon.com
D
x :::
@x 2
@x 2
@x 3
i 1
33
n
@ n
@2 t
(3.5)
C
D
@t 2 i 2
@t i
n
x 2
@2
u
@ n
(3.4)
x C
D
C
2
@x
2 articial viscosity of numerical methods
x dynamics
@x i1
Lectures on computational fluid
Stability
i 1 and
The derivatives at i 1 th point can be expressed through these at i th
point:
n
n
@3 x 2
@ n @2
@ n
:::;
x C
D
@x i1
@x i
@x 2 i
@x 3 i 2
n
n
n
@2
@3
@2
D
x :::
@x 2 i1
@x 2 i
@x 3 i
(3.5)
n
@ n
@2 t
D
C 2
@t i
@t i 2
u
D
x
n
n
2 n
2
@ n @2
@
@3
x
C ::: (3.6)
x x C
x
@x i @x 2 i
@x 2 i @x 3 i
2
Finally we have:
n
@2 t
@ n
D
C 2
@t i
@t i 2
n
@ n @2 x 2
u
x
C :::
D
x
@x i
@x 2 i 2
(3.7)
2 n
@22 n
2 @2 n
@ 2 D u2 @ 2
D34u @x 2 i
@t
@t 2 ii
@x i
(3.8)
(3.9)
(3.9)
2 n
u
@ nn
@22 n x 22 u
@ nn
2 @2 n
@ D u x @ C @ 2 x u u2 @ 2
@t D x x @x i C @x 2 i 2 x u @x 2 i
@t ii
x
@x i
@x i 2 x
@x i
Finally we have
Finally we have
!
!
@
@
ux
ut @2
2
D
u
C
1
C :::
2
@
@
@
@t D u @x
2
x
@x
C ux
1 ut
C :::
@t
@x
2
x
@x 2
Compare
noweBooks
with the
original equation:
Download free
bookboon.com
Compare
now with atthe
original equation:
@
@34
@ D u @
@t D u @x
@t
@x
The additional term
The additional term
!
t
t C :::
2 C :::
2
(3.10)
(3.10)
@
@t
@
u @x
ux
2
1
ut
x
@2
@x 2
C :::
Stability and articial viscosity of numerical methods
1
ut
x
@2
@x 2
3.2
(3.11)
35
Masters in Management
Designed for high-achieving graduates across all disciplines, London Business Schools Masters
in Management provides specific and tangible foundations for a successful career in business.
This 12-month, full-time programme is a business qualification with impact. In 2010, our MiM
employment rate was 95% within 3 months of graduation*; the majority of graduates choosing to
work in consulting or financial services.
As well as a renowned qualification from a world-class business school, you also gain access
to the Schools network of more than 34,000 global alumni a community that offers support and
opportunities throughout your career.
35
(3.12)
ut
:
x
(3.13)
n
n
inC1
C1 D i C1 .1 c/ C c i D c "
time instant n C 2:
2
2
inC2 D inC1 .1 c/ C c inC1
1 D ".1 c/ D ".1 c/
nC1
nC1
inC2
D c ".1 c/ C c ".1 c/ D 2c ".1 c/
C1 D i C1 .1 c/ C c i
nC1
nC1
2
inC2
C2 D i C2 .1 c/ C c i C1 D c "
time instant n C 3:
3
inC3 D inC2 .1 c/ C c inC2
1 D ".1 c/
nC2
nC2
inC3
D 2c ".1 c/2 C c ".1 c/2
C1 D i C1 .1 c/ C c i
nC2
nC2
2
2
inC3
C2 D i C2 .1 c/ C c i C1 D c ".1 c/ C c .2"/.1 c/
nC3
nC2
3
iC3
D inC2
C3 .1 c/ C c iC2 D c "
time instant n C N :
36
.inCN
CN /
D cN "
::::::::::::::::::
As follows from the last formula,the perturbation decays if
Download free eBooks at bookboon.com
c < 1 36
(3.14)
The condition (3.14) is the Courant Friedrich Levy criterion of the stability
of explicit numerical schemes. If the velocity is changed within the computational domain, the maximum velocity umax is taken instead of u in for-
time instant n C N :
N
.inCN
CN / D c "
Lectures on computational fluid dynamics
Stability and articial viscosity of numerical methods
::::::::::::::::::
c<1
The condition (3.14) is the Courant Friedrich Levy criterion of the stability
of explicit numerical schemes. If the velocity is changed within the computational domain, the maximum velocity umax is taken instead of u in fort
mula (3.14). Physically the condition umax
< 1 means that the maximum
x
displacement of the fluid particle within the time step t; t C t does not
exceed the cell size x. The CFL parameter c can be reduced by decrease
of t (not by increase of x!).
3.3
Exercise
37
37
Chapter 4
Chapter
4
Chapter
4
Simple explicit
time advance
Simple
time
advance
scheme
for solution
the
Simple explicit
explicit
timeof
advance
scheme
for
of
Navier
Equation
scheme Stokes
for solution
solution
of the
the
Navier
Navier Stokes
Stokes Equation
Equation
4.1
Theory
The
term of the Navier Stokes Equation
4.1unsteady
Theory
4.1
Theory
@ui uStokes
1 @p
i
j
The unsteady term of@uthe
Navier
Equation@ @ui
D
C
(4.1)
The unsteady term of@tthe Navier
Equation
@xj Stokes
@x
@xj @xj
i
@ui
@ui uj
1 @p
@ @ui
D
C
(4.1)
is written in explicit @u
form:
@u
1
@p
@ @x
u
@ui
@ti D @xi j j @xi C @x
j
j
(4.1)
j n n @xi
@t
@x
@xj @xj
n
n
ui uj
1 p
ui
is written in explicit
nC1 form:
n
(4.2)
C
D
u
C
t
u
i
i
is written in explicit form:
@xi
xj xj
x
nj n
n
n
ui
nC1
n
uin ujn 1 p n C
@
(4.2)
D
u
C
t
u
u
u
i
i
1
p
u
approximation
of the
. x
Let
usniapply the diverwhere xj is thenC1
i derivative
j
n
@x
x
@x
x
j
j
i
j
j
(4.2)
C
ui D ui C t
xj
@xi
xj xj
gence operator
:
xi
@
where xj is the approximation of the derivative @xj . Let us apply the diver@
of theu
derivative
. Let us
where xj isuthe
n n
nC1approximation
n
@xjn
u
uni the diverp
u
1
apply
gence operator
:
i
j
i
i xi
D
C t
C
(4.3)
gence operator
xi xi : xi
xi x
@x
x
x
j
i
j
j
unC1
uni
uni ujn n 1 p n
uni
inC1
n u
n
n
n
n
D
C
t
C
(4.3)
ui uj i D1 p
ux
u
uis
Let uni is the
field,
i.e.
0. The task
to find(4.3)
the
i divergence
i
i xj xi @xi C xj xji
D xii Cfree
t x
xi moment
@x
x
xi n C 1xwhich
x
velocity fieldx
ati the time
is
also
divergence
free
j n
i
j
j
ui
Let uni is the divergence free field, i.e. u
D
0.
The
task
is
to find the
n
xii
Let uni isfield
theatdivergence
free field,
i.e.
D
0.
The
task
is
to
unnC1
velocity
the time moment
C
1
which
is
also
divergence
freefind the
xi
i
Dwhich
0
velocity field at the time moment n
C
1
is
also
divergence
free (4.4)
xi
unC1
inC1
D0
(4.4)
Substituting (4.4) into (4.3) one u
obtains:
i i
x
D
0
(4.4)
n n
x
n
i n
u
u
(4.3)i one
1 p
j obtains:
Substituting (4.4) into
39 C ui D 0
obtains:
(4.5)
Substituting (4.4) x
into
(4.3)
one
xj
@xi
xj xj
i
39
Expressing (4.5) with respect to the pressure
results in the Poisson equation:
39
#
"
2 uni ujn
2 uni
2pn
(4.6)
D
C
xj xi
xi xj xj
@xi2
The algorithm for time-advancing is as follows:
i) The solution at time n is known and divergence free.
h 2 un un
2
ii) Calculation of the r.h.s. of (4.6) xj ixji C xi xj
iii) Calculation
pressure
Download
free eBooksof
at the
bookboon.com
un
i
xj
xi
xj
@xi
xj xj
Expressing (4.5) with respect to the pressure results in the Poisson equation:
# explicit time advance scheme for
"
Simple
2 n n
n
2 n
2
u
u
u
p
i
j
i
Lectures on computational fluid D
dynamics
solution of the Navier
Stokes Equation
(4.6)
C
xj xi
xi xj xj
@xi2
The algorithm for time-advancing is as follows:
i) The solution at time n is known and divergence free.
h 2 un un
2
ii) Calculation of the r.h.s. of (4.6) xj ixji C xi xj
un
i
xj
4.2
Mixed schemes
The high accuracy of the CDS schemes is their advantage. The disadvantage
of CDS schemes is their instability resulting in oscillating solutions. On the
contrary, the upwind difference schemes UDS possess a low accuracy and high
stability. The idea to use the combination of CDS and UDS to strengthen
their advantages and diminish their disadvantages. Let us consider a simple
transport equation for the quantity ':
@'
@'
Cu
D0
(4.7)
@t
@x
with u > 0. A simple explicit, forward time, central difference scheme for
this equation may be written as
1
1
'inC1 D 'in c.'in C .'inC1 'in /
'in1 C .'in 'in1 /
/ D
2
2
1 n
1
D 'in c.'in 'in1
C 'iC1
'in
'in 'in1
/
2
2
(4.8)
where c D ut
is the CFL parameter.40The term c'in 'in1 is the diffusive
x
1st order upwind contribution. The term c. 12 'inC1 'in 12 'in 'in1 / is the
anti-diffusive component. With TVD (total variation diminishing) schemes
the anti-diffusive component is limited in order to avoid instabilities and
maintain boundness 0 < ' < 1:
1
1
n
n
'inC1 D 'in c.'in 'in1 C 'inC1 'in east
'in 'in1 west
/ (4.9)
2
2
where are limiters. Limiters functions for TVD schemes are given in table 4.1.
Limiters
Download free Table
eBooks 4.1:
at bookboon.com
Scheme
central
upwind
1
0
1
1
n
n
'inC1 D 'in c.'in 'in1 C 'inC1 'in east
'in 'in1 west
/ (4.9)
2
2
Simple explicit
timein
advance
where are limiters. Limiters functions for TVD schemes
are given
ta- scheme for
Lectures
ble 4.1.on computational fluid dynamics
central
upwind
Roe minimod
Roe superbee
Van Leer
Branley and Jones
1
0
' n
i
Here r D . @'
/n =. @'
/n , . @'
/n D xiC1
. The mixed upwind and cen@x west
@x east
@x east
iC1 xi
tral difference scheme are used in Sec. 4.4 for approximation of the convective
terms with the limiter (4.14).
4.3
Staggered grid
The grids are subdivided into collocated and staggered ones. On collocated grids the unknown quantities are stored at centres of cells (points P in
Fig. 4.1). The equations are also satisfied at cell centres. For the simplicity,
we considered the case x and y are constant in the whole computational
domain. Use of collocated grids meets the problem of decoupling between
the velocity and pressure fields. Let us consider the Poisson equation (4.6)
with the r.h.s.
@Tx
@Ty
@Hx
@Hy
@Dx
@Dy
C
D
C
C
C
@x
@y
@x
@y
@x
@y
where
41
40
'n
' n
i
/n =. @'
/n , . @'
/n D xiC1
. The mixed upwind and cenHere r D . @'
@x west
@x east
@x east
iC1 xi
tral difference scheme are used in Sec. 4.4 for approximation
the convective
Simple of
explicit
time advance scheme for
terms with
the limiterfluid
(4.14).
Lectures
on computational
dynamics
solution of the Navier Stokes Equation
4.3
Staggered grid
The grids are subdivided into collocated and staggered ones. On collocated grids the unknown quantities are stored at centres of cells (points P in
Fig. 4.1). The equations are also satisfied at cell centres. For the simplicity,
we considered the case x and y are constant in the whole computational
domain. Use of collocated grids meets the problem of decoupling between
the velocity and pressure fields. Let us consider the Poisson equation (4.6)
with the r.h.s.
@Tx
@Ty
@Hx
@Hy
@Dx
@Dy
C
D
C
C
C
@x
@y
@x
@y
@x
@y
where
@ 2 ux
@2 ux
C
41
@x 2
@y 2
@2 uy
@2 uy
Dy D 2 C 2
@x
@y
@ux uy
@ux ux
Hx D
@x
@y
@ux uy
@uy uy
Hy D
@x
@y
Dx D
(4.10)
n
n
n
n
/ . @p
/
. @p
Ty;N
Ty;S
Tx;W
Tx;E
. @p
/ . @p
/
@y N
@y S
@x E
@x W
C
D
C
2x
2y
2x
2y
n
or
n
n
pEE
pP
2x
n
n
pP
pW
W
2x
2x
n
n
pN
N pP
2y
n
n
pP
pSS
2y
2y
n
n
Tx;W
Tx;E
2x
n
n
Ty;N
Ty;S
2y
D QPH
(4.11)
where
This equation (4.11) has involves nodes which are 2 apart (see also [3])!
It is a discretized Poisson equation on a grid twice as coarse as the basic
Download
freeequations
eBooks at bookboon.com
one
but the
split into four unconnected systems, one with i and j
both even, one with i even and j odd, one
41 with i odd and j even, and
one with both odd. Each of these systems gives a different solution. For
a flow with a uniform pressure field, the checkerboard pressure distribution
shown in Fig. 4.2 satisfies these equations and could be produced. However,
where
p
p
1
explicit
time advance
scheme for
l D EE; W W; N N; S S , ApEE D ApW W D .21x /2 , ASimple
N N D AS S D .2y /2
P p fluid dynamics
Lecturespon computational
solution of the Navier Stokes Equation
and AP D l Al
This equation (4.11) has involves nodes which are 2 apart (see also [3])!
It is a discretized Poisson equation on a grid twice as coarse as the basic
one but the equations split into four unconnected systems, one with i and j
both even, one with i even and j odd, one with i odd and j even, and
one with both odd. Each of these systems gives a different solution. For
a flow with a uniform pressure field, the checkerboard pressure distribution
shown in Fig. 4.2 satisfies these equations and could be produced. However,
the pressure gradient is not affected and the velocity field may be smooth.
42
42
(4.12)
The uy velocities are stored at points staggered by y =2 in y-direction (circles). At these points the second Navier-Stokes equation is satisfied:
@uy
@uy uj
1 @p
@ @uy
D
C
@t
@xj
@xj @xj
@y
(4.13)
4.4
Approximation of
uni ujn
xj
The approximation of the convective term is a very critical point. For faster
flows or larger time steps, the discretization shall be closer to an upwinding
approach [4]. Following to [4] we implement a smooth transition between
centered differencing and upwinding using a parameter 2 0; 1. It is defined
as
Download free eBooks at bookboon.com
(4.14)
Click on the ad to read more
The value of gamma is the maximum fraction of a cells which information can
travel in one time step, multiplied by 1:2, and capped by 1. The factor of 1:2
(4.12)
The uy velocities are stored at points staggered by y =2 in y-direction (circles). At these points the second Navier-Stokes equation is satisfied:
@uy
@uy uj
@ @uy
1 @p
D
C
@t
@xj
@y
@xj @xj
(4.13)
4.4
Approximation of
uni ujn
xj
The approximation of the convective term is a very critical point. For faster
flows or larger time steps, the discretization shall be closer to an upwinding
approach [4]. Following to [4] we implement a smooth transition between
centered differencing and upwinding using a parameter 2 0; 1. It is defined
as
D mi n.1:2 t max.jux .i; j /j; juy .i; j /j/; 1/
(4.14)
The value of gamma is the maximum fraction of a cells which information can
travel in one time step, multiplied by 1:2, and capped by 1. The factor of 1:2
is taken from the experience that often times tending a bit more towards
upwinding can be advantageous for accuracy [5].
D 0 corresponds to the central difference scheme (CDS) whereas D 1
results in the upwind difference scheme (UDS).
4.4.1
@ux ux
Approximation of 44
@x
@ux uy
@y
x
x
D ux @u
uy @u
@x
@y
1=2; j / D .ufluid
j / C ux .i 1; j //=2:0
ux .ioncomputational
Lectures
x .i;dynamics
If ux .i 1=2; j / 0;
C1
1
ux .i; j / C
ux .i 1; j //
2
2
C1
1
ux .i; j / C
ux .i 1; j //
D ux .i 1=2; j /.
2
2
4.4.2
@u u
x y
Approximation of 45@x
@uy uy
@y
D ux
@uy
@x
uy
@uy
@y
If uy .i; j 1=2/ 0;
1
C1
uy .i; j / C
uy .i; j 1//
Simple
2
2 explicit time advance scheme for
Lectures on computational fluid dynamics
of the Navier Stokes Equation
C1
1 solution
else uy uy ji;j 1=2 D uy .i; j 1=2/.
uy .i; j / C
uy .i; j 1//
2
2
Point .i; j C 1=2/ for x-component of velocity:
then uy uy ji;j 1=2 D uy .i; j 1=2/.
STUDY. PLAY.
NORWAY.
YOUR IDEAL STUDY DESTINATION.
WWW.STUDYINNORWAY.NO
FACEBOOK.COM/STUDYINNORWAY
46
of
n
u
of ix
xjj
4.5
Approximation
4.5
Approximation
4.5 Approximation of xj xj
unni
ui
xj
xj
The second
secondderivative
derivative
is calculated
the Central
Scheme
The
is calculated
using using
the Central
DifferenceDifference
Scheme
(CDS):
(CDS):
(CDS):
C 1;
/ x .i;
2u .i;
/ C u .i u1;
j /j Cu1/
.i;
j xC.i;1/
2u
j / C u .i; j 1/
ux .i 1;
/ j
C uj
2u
j /
uxxx.i;.i;
x .i;
.i Cj1;
j 2u
/ 2ujxx/.i;
jx/.iCu1;xx j.i/ C
1;
j / C uxx .i;
j C 1/
C2u
.i;j j/1/C uxx .i; j 1/
Dxx.i;
.i;jj/ /DDuxu.ixC
D
2
Dx .i; j / D
C
y22
2x
y2
2x
x
uyu
.iy C
/ j 2u
C uj /.iCu1; j.i/ u1;
2u
j /
C2u
u .i;.i;
j j/1/
.i 1;
Cj1;
/
2ujy/.i;
j /j Cu1/
.i;
j C.i;1/
C u .i; j 1/
y .i;
y .i;
D
.i; jy/ C uyy .i C
1;
j / C uyy .i; j yC21/ 2uyyy .i; j / C uyy .i; j 1/
Dyy.i;
.i;jj/ /DD uy .i C 1; j / 2u
2 y 2
2
Dy .i; j / D
C
x
y
2xx
4.6
yy2
4.6
Calculation
of the r.h.s. for the Poisson
4.6 equation
Calculation
(4.6) of the r.h.s. for the Poisson
equation (4.6)
(4.6)
The r.h.s. equation
of (4.6) is
The r.h.s. of2 u(4.6)
is 2
n n
The r.h.s. of (4.6)
is uni
i uj
@Hy
@Dx
@Dy
@Hx
C
C
C
C
D
n
n
xj 2x
x
x
x
@x
@y
@x
@y
n
2
i
j
j
ui u
2 uni ujn
uni
u
@Hy
@H
@Hx
@H
@Dx
@D
@Dy
@D
x C
y C
x C
y
at ithej point
C (i,j) of the ipressure
D
The derivatives
as X in
xj xi C x
C @y(designated
C @x C
D @xstorage
@y
i xj xj
xj xi usingxCDS
@x
@y
@x
@y
Fig. 4.3) are calculated
i xj xj
x
x
@x
j/
1/
xD
D
1/x .i 1; j /
HH
@H
@Hyx
j /Hy .i;
Hjx .i
1;@Djy/ @DxDy .i; j /D
.i;
j /j
y .i;
y .i;
x .i;
@HxjijjijDD H
D x jij D Dx .i; j / Dx .i 1; j / ;
x .i; j / Hx .i ;1; j /j;ij@D
y
y
@y
@y
j
;
j
;
D
D
x
x
@x ij
@x ij
@x
x
@x
x
Hy .i; j / Hy .i; j 1/ @Dy
Dy .i; j / Dy .i; j 1/
j / the
Hy .i;
j 1/ ; @Dequation
j / Dy .i; j 1/
@HyySolution
4.7 @H
Poisson
y jij D Dy .i; (4.6)
jij D Hy .i; of
j
;
j
D
D
y
y
@y ij
@y ij
@y
y
@y
y
The numerical solution of the Poisson equation is discussed in [3].
4.7
Solution
of the
equation
4.7 Update
Solution
the Poisson
Poisson
equation (4.6)
(4.6)
4.8
the of
velocity
field
The numerical solution of the Poisson equation is discussed in [3].
Thevelocity
numerical
the Poisson
equation is discussed in [3].
The
field solution
is updatedofaccording
to formula
4.8
4.8
Update
Update the
the velocity
velocity field
field
unC1
.i; j / D unx .i; j /C t .Hxn .i; j /CDxn .i; j /.p n .i C1; j /p n .i; j //=x /
x
nC1
n
n
.i;
j / D uynfield
.i; j / C
C Dyn .i; j /
.i; j C 1/ p n .i; j //=y /
u
t .Hy .i; j /according
The
velocity
isupdated
to.pformula
y
unC1
.i; j / D unnx .i; j /C t .Hxnn .i; j /CDxnn .i; j /.p nn .i C1; j /p nn .i; j //=x /
x
unC1
.i; j / D ux .i; j /C t .Hx .i; j /CDx .i; j /.p .i C1; j /p .i; j //=x /
x
47
4.9
At this stage, the boundary conditions (BC) for the velocity field should be
4.9
Boundary conditions for the velocities
taken into account. The nodes at which the BC are enforced are shown by
grey
symbols
in boundary
Fig. 4.3. conditions
Enforcement
thetheBC
is easy,
if the
grey
At
this
stage, the
(BC)offor
velocity
field
should
be point
lies exactly
at the The
boundary
of which
the computational
domain.
not, by
two cases
taken
into account.
nodes at
the BC are enforced
are If
shown
should
be considered.
If the Neumann
D
C , the
grey
symbols
in Fig. 4.3. Enforcement
of the condition
BC is easy,isif enforced
the grey@u
point
@n
lies
exactly
at the boundary
of the
computational
If not, two cases
velocity
component
outside
of the
boundary domain.
u.0/ is calculated
through the
should
considered.
the Neumann condition is enforced @u
D
C
,
the
interiorbequantity
u.1/If from
@n
velocity component outside of the boundary u.0/ is calculated through the
interior quantity u.1/ from
u.1/ u.0/
DC
n
u.1/ u.0/
DC
If the Dirichlet condition u D C
nis enforced and the point 0 is outside of the
computational
domain,u the
u.0/ and
is calculated
the extrapolation
If
the Dirichlet condition
D Cvalue
is enforced
the point 0from
is outside
of the
procedure:
computational domain, the value u.0/ is calculated from the extrapolation
procedure:
u.0/ C u.1/
DC
u.0/ C u.1/
2 DC
2
4.10 Calculation
Calculation
vorticity
4.10
of of
thethe
vorticity
@u
@ux
y
@uy
The calculation
calculationof of
vorticity
!z @uDx
performed
is performed
as follows:
The
vorticity
!z D
as follows:
@y @x is@x
@y
1 1
u
1=2/
D D.ux.u
.i;xj.i;
/ Cj /uxC
.i;ujx .i;
C 1/
j /C1;
uxj.i/ C
1;
uxx.i.i 1=2;
1=2;j jCC
1=2/
jC
C u1/x .iCu1;
uxj.iC1//;
1; j C 1//;
x .i
4 4
1
u
1=2/
D D.u1x.u
.i; j / C ux .i 1; j / C ux .i; j 1/ C ux .i 1; j 1//;
uxx.i.i 1=2;
1=2;j j
1=2/
4 4 x .i; j / C ux .i 1; j / C ux .i; j 1/ C ux .i 1; j 1//;
1
uy .i C 1=2; j 1=2/ D .u1y .i; j / C uy .i C 1; j / C uy .i; j 1/ C uy .i C 1; j 1//;
uy .i C 1=2; j 1=2/ D
4 .uy .i; j / C uy .i C 1; j / C uy .i; j 1/ C uy .i C 1; j 1//;
1 4
uy .i 1=2; j 1=2/ D .u1y .i; j / C uy .i 1; j / C uy .i; j 1/ C uy .i 1; j 1//;
4 .uy .i; j / C uy .i 1; j / C uy .i; j 1/ C uy .i 1; j 1//;
uy .i 1=2; j 1=2/ D
ux .i 1=2; j C41=2/ ux .i 1=2; j 1=2/
!z .i; j / D u .i 1=2; j C 1=2/ u .i 1=2; j 1=2/
x
x
y
!z .i; j / D
yj 1=2/
uy .i C 1=2; j 1=2/ uy .i 1=2;
uy .i C 1=2; j 1=2/
x uy .i 1=2; j 1=2/
x
48
48
48
Chapter 5
Chapter
5
Chapter
Splitting 5schemes for solution
Splitting
schemes
solution
of
multidimensional
Splitting
schemes for
forproblems
solution
of
of multidimensional
multidimensional problems
problems
5.1
@unsteady
@ heat conduction equation:
@
Let us consider the two dimensional
(5.1)
D unsteady
C 2heat conduction equation:
2
Let us consider the two dimensional
@t
@y
@x
@2
@
@2
Crank
2
2 Nicolson and CDS for spatial
We use implicit scheme proposed
by
and
(5.1)
C
D
@
@t D @@x2 C @@y2
(5.1)
derivatives:
@x 2 and
@y 2 Nicolson and CDS for spatial
@t
We use implicit scheme proposed
by Crank
We
use implicit schemeproposed by Crank
and Nicolson and CDS
derivatives:
for spatial
nC1
n
2 n
2 n
2 nC1
2 nC1
@
@
@
@
derivatives:
C
C
C
(5.2)
D
2
2
2
2
t
2
@x
@y
@x
@y
nC1 n
@2 n @2 nC1
@2 nC1
@2 n
nC1
n D
2 n C 2 n n
2
nC1
C
C
(5.2)
n 2 n
t
@@y
@2@y
nC1
2 @n 2
2
2
2@i;j
iC1;j
@xC i 1;j
C
C
(5.2)
D 2@ @x 2 C
2
(5.3)
t
2 @x 2@x D @y 2
@x
@y 2
2
.x/
2 n i;j
n
n
C in1;j
@2 n D inC1;j 2i;j
n
n
(5.3)
ii;j
2i;j2C i;j
@@x2
C1;j
i 1;j
C1 .x/
1
i;j D
Technical
training
on
(5.3)
(5.4)
2
2
.x/
@y
.y/
@x
i;j
n
n
n
2 n
C i;j 1 you need it
i;jneed,
WHAT
C1 2i;jWHEN
@ n you
n
n
n approximations: (5.4)
D
In what follows we use@2the
designations
for
derivative
2
C i;j
2
i;j
C1
i;j2our
1
.y/
At IDC@y
Technologies
we
can
tailor
technical
and engineering
i;j D
(5.4)
2
2
.y/
n
n
n
training@y
workshops
to
suit
your
needs.
We
have
extensive
OIL & GAS
2
i;j
2
C
In what followsexperience
we use the
designations
fori;j
derivative
i C1;j
i 1;j approximations:
n
ENGINEERING
in
training
technical
and
engineering
staff
and
D
2 designations .x/
2
In what followshave
we trained
usexthe
for derivative
approximations: (5.5)
people in
such
n organisations
n
n as General
2
C i 1;j to name a few.
i C1;j 2
ELECTRONICS
nSiemens,
i;jHoneywell
Motors, Shell,
and
n
n
2 D n BHP
(5.5)
2
C
2 n
ii;j
C1;j
i 1;j
C1 .x/
i;j2
i;j
1
xtraining
Our onsite
is cost effective,
convenient and completely
D
(5.5)AUTOMATION &
(5.6)
2
2
x
.x/
y2 to the technical
.y/
PROCESS CONTROL
customisable
and
engineering
areas
you
want
n
n
n
comprehensive
2i;j
C i;j
i;j
workshops
C1 all
1
covered. Our
are
hands-on
learning
n
n
n
n
D
(5.6) MECHANICAL
2to
i;j
22 with
experiences
time
given
and
2C i;j sessions
C149
i;jpractical
1
n ample
y
.y/
D
(5.6) ENGINEERING
demonstrations.
We communicate well
2
2 to ensure that workshop content
y
.y/
and timing match the knowledge, skills, and abilities of the participants.
INDUSTRIAL
49
We run onsite training all year round and hold the workshops on
DATA
COMMS
49
your premises or a venue of your choice for your convenience.
ELECTRICAL
POWER
49
We use implicit scheme proposed by Crank and Nicolson and CDS for spatial
derivatives:
2 nC1
schemes for solution of
Splitting
nC1 n
@2 n
@2 nC1
@
@2 n
Lectures on computational
multidimensional
problems
C
C
C
(5.2)
D fluid dynamics
t
2
@x 2
@y 2
@x 2
@y 2
@2 n
@x 2
@2 n
@y 2
i;j
i;j
n
C in1;j
inC1;j 2i;j
(5.3)
n
n
n
i;j
C1 2i;j C i;j 1
(5.4)
.x/2
.y/2
(5.5)
n
n
n
i;j
2 n
C1 2i;j C i;j 1
D
y 2
.y/2
(5.6)
t 2
t 2
t 2
t 2
nC1
1
D 1C
1
1C
nC
2 x 2
2 y 2
2 x 2
2 y 2
.t /2 2 2
nC1
n
C
4 x 2 y 2
(5.7)
t 2
t 2
1
D 1C
n
2 x 2
2 y 2
(5.8)
t 2
t 2
nC1
1
D 1C
2 y 2
2 x 2
(5.9)
It can be shown that the resulting method is of the second order of accuracy
and unconditionally stable.
5.2
Within the fractional step method the original equation is split according
to physical processes. Splitting according to physical processes is used for
unsteady
problems.
general idea is illustrated for the transport equation:
Download free
eBooks atThe
bookboon.com
50
@u
D C u C Du C P
@t
Here
(5.10)
1
y 2
nC1 D 1 C
x 2
(5.9)
5.2
Within the fractional step method the original equation is split according
to physical processes. Splitting according to physical processes is used for
unsteady problems. The general idea is illustrated for the transport equation:
@u
D C u C Du C P
@t
(5.10)
Here
C is the convection operator,
D is the diffusion operator,
50
(5.11)
(5.12)
solving sequentially. The sense of this splitting is that the numerical solution
of partial problems (*)-(***) is simpler and more stable than that of the
whole problem. The disadvantage of this procedure is that it is applicable to
only unsteady problem formulation. Another disadvantage is the low order
of accuracy with respect to time derivative approximation. The order of time
derivative approximations can be derived using the sample with two physical
processes described by operators L1 and L2 :
@u
D L1 .u/ C L2 .u/
@t
The splitting of (5.10) results in two steps procedure:
@u
D L1 .u /; u j t Dtn D un
@t
@unC1
D L2 .unC1 /; unC1 j t Dtn D u
@t
(5.13)
(5.14)
where
Downloadfree eBooks
at bookboon.com
n
n
u D u C tL1 .u / C O.t 2 /;
(5.15)
@u
D L1 .u /; u j t Dtn D un
@t
nC1
@ufluid
Lectures on computational
dynamics nC1
D L2 .u /; unC1 j t Dtn D u
@t
Splitting schemes
(5.14) for solution of
multidimensional problems
where
u D un C tL1 .un / C O.t 2 /;
(5.15)
The accuracy of the final solution unC1 is of the first order in time.
Very often the diffusion step is treated implicitly. This is done to diminish
51 process. Otherwise, the stability
the time step restriction for the diffusion
requires t to be proportional to the spacial discretization squared, if a pure
explicit scheme is applied. The semi implicit scheme for the two dimensional
Navier Stokes equation reads:
Convection step is treated explicitly:
ux unx
unx unx unx uyn
D
t
x
y
uy uyn
t
D
uyn unx
x
uyn uyn
y
(5.16)
(5.17)
The solutions ux;y are used then for the diffusion process which is
treated implicitly:
2
ZZZVWXG\DWWXGHOIWQO
u
2 u
ux
x ux
x
C
(5.18)
D
t
x 2
y 2
5DQNHGWKLQWKHZRUOG
2
7+(67HFKQRORJ\UDQNLQJ
uy
2 uy
uy uy
(5.19)
D
C
$OPRVW\HDUVRISUREOHPVROYLQJ
t
x 2
y 2
H[SHULHQFH
([FHOOHQW6SRUWV &XOWXUHIDFLOLWLHV
The
solutions u
x;y are used then for the next process which is treated
explicitly:
&KHFNRXWZKDWDQGKRZZHWHDFKDW
nC1
p nC1
ZZZRFZWXGHOIWQO ux ux
t
uynC1 uy
t
D
D
p nC1
y
(5.20)
(5.21)
where the pressure p nC1 should be pre computed from the continuity
equation demanding the velocity at n C 1 time slice is divergency free:
uynC1
unC1
x
C
D0
x
y
(5.22)
Applying the operator r to the equations (5.20) and (5.21) we get the Poisson
equation
for the pressure:
Download free eBooks at bookboon.com
uy
2 p nC1
2 p nC1
1 52u
x
C
D
C
t x
x 2
y 2
y
52
Click on(5.23)
the ad to read more
ux unx
un un ux uy
D x x
t
x
y
uy uyn
uyn unx
t
(5.16)
Splitting schemes for solution of
multidimensional
problems
(5.17)
uyn uyn
y
The solutions ux;y are used then for the diffusion process which is
treated implicitly:
2
u
2 u
ux
x ux
x
C
(5.18)
D
t
x 2
y 2
uy uy
t
D
2 uy
x 2
2 uy
y 2
(5.19)
D
p nC1
y
(5.21)
where the pressure p nC1 should be pre computed from the continuity
equation demanding the velocity at n C 1 time slice is divergency free:
uynC1
unC1
x
C
D0
x
y
(5.22)
Applying the operator r to the equations (5.20) and (5.21) we get the Poisson
equation for the pressure:
uy
2 p nC1
2 p nC1
1 u
x
C
D
C
x 2
y 2
t x
y
5.3
(5.23)
52
Increase of the accuracy
of time derivatives approximation using the Lax-Wendroff
scheme
(5.24)
@F
@u
F t D Fu u t D Fu Fx AFx
(5.25)
Substitution of these results into the time Taylor series gives the Lax-Wendroff
scheme which is of the second order in time:
2
u t t .t/ C O. 3 /
2 53
2
D u.t/ Fx .t/ C .A.t/Fx .t//x C O. 3 /
2
(5.26)
AD
@F
@u
u t t D Fxt D F tx ;
F t D Fu u t D Fu Fx AFx
(5.25)
Substitution of these results into the time Taylor series gives the Lax-Wendroff
scheme which is of the second order in time:
2
u t t .t/ C O. 3 /
2
2
D u.t/ Fx .t/ C .A.t/Fx .t//x C O. 3 /
2
(5.26)
53
54
Chapter 6
Chapter 6
Chapter
6
Finite Volume
Method
Finite Volume Method
Finite Volume Method
6.1
The Navier Stokes
@.ui uj /
1 @p
@
@
@ui equation
C
D Fi
C
ui
The Navier Stokes
@xj
@t equation
@xi
@xj @xj
@.ui uj /
1 @p
@ @
@ui
C
D
F
C
uthe
i
i
u
/
@.u
1
@p
@j U)@x
@in
@u
is fulfilled within@teach
mesh
volume
integral
i
i j j element (finite
@x
@x
@x
i
j
C
D Fi
C
ui
For that it is integrated
over
the
volume
U
:
@xj
@t
@xi
@xj @xj
is fulfilled within each mesh element (finite volume U) in the integral
is fulfilled
within each mesh element
(finite
in the integral
For
U : volume U)
it is integrated over the
Zthat
Z volume
For that@u
it iis integrated
:
@
@.ui uj / over the volume1 U@p
@
dU D
Fi
C
C
ui d U
@xj
@xi
@xj @xj
Z @t
Z
U
U
Z @ui C @.ui uj / d U D Z Fi 1 @p C @ @ ui d U
@u
@
/
1 @pi
@
@xi u
@t i C @.u
jj
dU
D U results
Fi @x
C @xj @xj ui d U
Application
of
the
Gauss
theorem
in
U
@xj
@t
@xi
@xj @xj
U
(6.1)
(6.1)
sense.
(6.1)
sense.
sense.
(6.2)
(6.2)
(6.2)
Application
of the
results Zin
Z Gauss theorem
Z
Z
Z
@
Application
of the Gauss theorem results1 in
ui d U C ui uE nE dS D Fi d U
p eEi nE dS C grad ui nE dS
Z
@t Z
Z
Z
Z
S
U
S
S
1
@ UZ
Z ui uE nE dS D Z Fi d U Z p eEi nE dS C Z grad ui nE dS
u
d
U
C
i
1
@
@t
ui d procedure
U C S ui uEapplied
nE dS D Uto F
S p eEequation
E dS C gives
grad ui nE dS
idU
in
The
same
the
continuity
U
S
@t
U
S
U
S
S
Z
The same procedure applied to the continuity equation gives
E dS D 0 equation gives
The same procedure applied to theuE ncontinuity
Z
S
Z uE nE dS D 0
uE nE dS D 0
S
55
(6.3)
(6.3)
(6.3)
(6.4)
(6.4)
(6.4)
55
55
Download freeFigure
eBooks at
bookboon.com
6.1:
Staggered
6.2
6.2
Sample
Sample
55
Let us consider the two dimensional transport equation without the diffusion
term
Let us consider the two dimensional transport equation without the diffusion
Lectures on computational
dynamics arrangement of finite volumes.
Figure 6.1:fluid
Staggered
6.2
Sample
Let us consider the two dimensional transport equation without the diffusion
term
8
@ui
@p
@ui uj
< @t C @x D @x
j
i
@uj
:
D0
@xj
(6.5)
ui d U C
ui uE nE dS D
p eEi nE dS
(6.6)
We use the staggered grid (Fig. 13.2). The pressure is stored at the volume
centers. The ux velocity is stored at the centers of vertical faces, whereas
the velocity uy component at centers of horizontal faces. The x- equation
is satisfied for volumes displaced in x-direction, whereas the y-equation for
these displaced in y-direction.
Below we consider approximations of different terms:
6.2.1
(6.7)
ux d U D 2
n
unC1
xij uxij
(6.8)
t
D
p eE2 nE dS pn Sn ps Ss
D pij C1 pij
(6.9)
uy d U D 2
nC1
n
uyij
uyij
t
(6.10)
6.2.2
nE uE
ui
(6.10)
6.2.2
nE uE
uxe
uxw
uy n
uys
ui
uxe
uxw
ux n
uxs
6.2.3
Visit us at www.dtu.dk
57
approximation
uxe D 12 .uxij C uxi C1j /
uxw D 12 .uxij C uxi 1j /
ux n D 12 .uxij C uxij C1 /
uxs D 12 .uxij C uxij 1 /
uy n D 12 .uyij C uyi C1j /
uys D 12 .uyij 1 C uyi C1j 1 /
flux
.uxij C uxi C1j /2
4
4 .uxij C uxi 1j /2
.uxij C uxij C1 /.uyij C uyi C1j /
4
4 .uxij C uxij 1 /.uyi C1j 1 C uyij 1 /
nE uE
uxe
uxw
uy n
uys
ui
uye
uyw
uy n
uys
6.2.4
approximation
uxe D 12 .uxij C uxij C1 /
uxw D 21 .uxi 1j C uxi 1j C1 /
uy n D 12 .uyij C uyij C1 /
uys D 12 .uyij C uyij 1 /
uye D 12 .uyij C uyi C1j /
uyw D 12 .uyij C uyi 1j /
X-equation approximation
2
n
unC1
xij uxij
t
58
58
.uxij
4
6.2.5
flux
.uxij C uxij C1 /.uyij C uyi C1j /
4
4 .uxi 1j C uxi 1j C1 /.uyij C uyi 1j /
.uyij C uyij C1 /2
4
4 .uyij C uyij 1 /2
.uxij C uxiC1j /2 .uxij C uxi 1j /2 C
4
4
C uxij C1 /.uyij C uyiC1j / .uxij C uxij 1 /.uyij 1 C uyi C1j 1 /
4
C piC1j pij D 0
Y-equation approximation
6.3
nC1
n
uyij
uyij
C
t
C uxij C1 /.uyij C uyiC1j / .uxi 1j C uxi 1j C1 /.uyij C uyi 1j /C
4
.uyij C uyij C1 /2 .uyij C uyij 1 /2 C
4
4
C pij C1 pij D 0
.uxij
4
Explicit scheme
The next task is to specify the upper index in X and Y equations. If the
index is n we get fully explicit scheme which is similar to that derived above
for finite difference method
2
n
unC1
xij uxij
nC1
n
uyij
uyij
C
t
C unxiC1j /2 .unxij C unxi 1j /2 C
4
n
n
n
C uyiC1j
/ .unxij C unxij 1 /.uyij
1 C uyi C1j 1 /
4
n
n
C pi C1j 59pij D 0
(6.11)
n
.u
4 xij
n
n
.u C unxij C1 /.uyij
4 xij
C
t
n
n
n
n
n
n
.uyij C uyiC1j
/.unxij C unxij C1 / .uyij
C uyi
1j /.uxi 1j C uxi 1j C1 /C
4
4
Download free eBooks at bookboon.com
n
n
n
2
n
2
.uyij C uyij
.uyij C uyij
C1 /
1 / C
59
4
4
n
n
C pij C1 pij D 0
(6.12)
nC1
n
uyij
uyij
C
t
Lectures
fluid dynamics
Finite Volume Method
n
non computational
n
n
n
n
n
n
.uyij C uyi
.uyij C uyi
C1j /.uxij C uxij C1 /
1j /.uxi 1j C uxi 1j C1 /C
4
4
n
n
n
2
n
2
.uyij C uyij
.u C uyij
C1 /
1 / C
4
4 yij
n
n
C pij C1 pij D 0
(6.12)
The Poisson equation for pressure is derived in the same manner as above
for finite difference method. For that the equations (6.11) is differentiated
on x, whereas the equation (6.12) is differentiated on y. Then both results
are summed under assumptions that both unC1
and unij are divergence free:
ij
unC1
xij
x
nC1
uyij
D 0;
unxij
x
n
uyij
D0
This equation is coupled with equations (6.11) and (6.12). The explicit
scheme has advantage that the solution at the time instant n C 1 is explicitly expressed through the solution at time instant n. The solution of
linear algebraic equations which is the most laborious numerical procedure
is necessary only for the solution of the Poisson equation. The momentum
nC1
equations (6.11) and (6.12) are solved explicitly. Velocities unC1
xij and uyij
are computed then from simple algebraic formula (6.11) and (6.12). The big
disadvantage of the explicit method is the limitation forced by the Courant
Friedrich Levy criterion. The time step t should be very small to secure the
numerical stability. This disadvantage can be overcome within the implicit
schemes.
6.4
Implicit scheme
n
xij
u
with
u
Increase your impact
C MSM Executive Education
2
t
nC1
nC1
2
2
.u
C unC1
.uxij C unC1
xi C1j /
xi 1j / C
4 xij
4
60
For almost 60 years Maastricht School of Management has been enhancing the management capacity
of professionals and organizations around the world through state-of-the-art management education.
Our broad range of Open Enrollment Executive Programs offers you a unique interactive, stimulating and
multicultural learning experience.
Be prepared for tomorrows management challenges and apply today.
For more information, visit www.msm.nl or contact us at +31 43 38 70 808 or via admissions@msm.nl
18-08-11 15:13
60
6.4
Implicit scheme
n
unC1
xij uxij
C
t
nC1
nC1
2
2
.uxij C unC1
.u
C unC1
xiC1j /
xi 1j / C
4
4 xij
nC1
nC1
nC1
60 .unC1 C unC1 /.unC1 C unC1
.uxij C unC1
xij C1 /.uyij C uyi C1j /
xij 1
yij 1
yi C1j 1 /
4
4 xij
nC1
nC1
C pi C1j pij
D0
(6.13)
2
nC1
n
uyij
uyij
C
t
nC1
nC1
nC1
nC1
nC1
nC1
nC1
nC1
.uyij C uyi
.u
C uyi
C1j /.uxij C uxij C1 /
1j /.uxi 1j C uxi 1j C1 /C
4
4 yij
nC1
nC1
nC1
nC1 2
2
.uyij C uyij
.u
C uyij
C1 /
1 / C
4
4 yij
nC1
nC1
C pij C1 pij
D0
(6.14)
The Poisson equation for pressure is derived in the same manner as above
for finite difference method. For that the equations (6.13) is differentiated
on x, whereas the equation (6.14) is differentiated on y. Then both results
are summed under assumptions that both unC1
and unij are divergence free:
ij
unC1
xij
x
nC1
uyij
D 0;
unxij
x
n
uyij
D0
The resulting Poisson equation can not be solved because both the r.h.s.
(velocities) and the l.h.s (pressure) depend on n C 1. The term on r.h.s.
cannot be computed until the computation of velocity field at time n C 1
is completed and vice versa. Other problem is that the equations (6.13)
and (6.14) are non linear.
6.5
To solve the nonlinear system and the whole system of equations we use the
iterative procedure. Let m be an iteration number. The nonlinear term is
represented in form:
.m1/
@u.m/
@ui uj
i uj
D
@xj
@xj
(6.15)
The velocity uj is taken from the previous iteration .m1/. The system (6.11)
61
and (6.12) is rewritten in the form
61
.m1/
@u.m/
i uj
@xj
@xj
Finite
Volume Method
(6.15)
The velocity uj is taken from the previous iteration .m1/. The system (6.11)
and (6.12) is rewritten in the form
2
u.m/
unxij
xij 61
t
.m/
.m/
.m1/
.m1/
.uxij C u.m/
.u C u.m/
C u.m1/
C u.m1/
xi C1j /.uxij
xi C1j /
xi 1j /.uxij
xi 1j /C
4
4 xij
.m/
.m/
.m1/
.m1/
.m1/
.m1/
.uxij C u.m/
.u C u.m/
C uyi
xij C1 /.uyij
C1j /
xij 1 /.uyij 1 C uyi C1j 1 /
4
4 xij
.m/
.m/
C pi C1j pij D 0
2
.m/
n
uyij
uyij
t
.m/
.m/
.m/
.m1/
.m/
.m1/
.m1/
.uyij C uyi
.uyij C uyi
C u.m1/
C1j /.uxij
xij C1 /
1j /.uxi 1j C uxi 1j C1 /C
4
4
.m/
.m/
.m/
.m1/
.m1/
.m/
.m1/
.m1/
.uyij C uyij
.uyij C uyij
C uyij
C uyij
C1 /.uyij
C1 /
1 /.uyij
1 /C
4
4
.m/
.m/
C pij C1 pij D 0
or
.m/
.m/
.m/
.m/
qxi 1j u.m/
xi 1j Cqxij uxij Cqxi C1j uxi C1j Cqxij 1 uxij 1 Cqxij C1 uxij C1 C (6.16)
pi.m/
C1j
pij.m/
D rxij
.m/
.m/
.m/
.m/
.m/
qyi 1j uyi1j
Cqyij uyij
Cqyi C1j uyi
C1j Cqyij 1 uyij 1 Cqyij C1 uyij C1 C (6.17)
.m/
.m/
C pij C1 pij D ryij
where
62
62
.m/
.m/
.m/
.m/
.m/
qyi 1j uyi1j
Cqyij uyij
Cqyi C1j uyi
C1j Cqyij 1 uyij 1 Cqyij C1 uyij C1 C (6.17)
.m/
.m/
Lectures on computational fluidC
dynamics
pij C1 pij D ryij
where
.m1/
.m1/
C u.m1/
C u.m1/
.u
xi C1j / .uxij
xi 1j /C
4 xij62
.m1/
.m1/
.m1/
.m1/
C uyi
C .uyij
C1j / .uyij 1 C uyi C1j 1 /
.m1/
.m1/
C u.m1/
/;
qxi C1j D .u.m1/
xij
xi C1j /; qxi 1j D .uxi 1j C uxij
4
4
.m1/
.m1/
.m1/
.m1/
qxij C1 D .uyij
C uyi
C1j /; qxij 1 D .uyij 1 C uyi C1j 1 /
4
4
.m1/
.m1/
2
.m1/
.m1/
qyij D
C
C
u
u
uyij C uyij
C1
yij
yij 1 C
t
4
.m1/
.m1/
.m1/
C
u
u
C uxij C u.m1/
xij C1
xi 1j
xi 1j C1
.m1/
qyi 1j D uxi 1j C u.m1/
C u.m1/
qyi C1j D
u.m1/
xi 1j C1 ;
xij
xij C1
4
4
.m1/
.m1/
.m1/
.m1/
C uyij
qyij 1 D uyij C uyij
qyij C1 D
uyij
1 ;
C1
4
4
n
n
uxij
uyij
rxij D 2
; ryij D 2
t
t
qxij D 2 =t C
.m/
.m/
.m/
.m/
axi 1j u.m/
xi 1j C uxij C axi C1j uxi C1j C axij 1 uxij 1 C axij C1 uxij C1 C (6.18)
pi.m/
C1j
pij.m/
=qxij D Rxij
.m/
.m/
.m/
.m/
Challenge the .m/
way we run
ayi 1j uyi
1j C uyij C ayi C1j uyi C1j C ayij 1 uyij 1 C ayij C1 uyij C1 C (6.19)
.m/
.m/
C pij C1 pij =qyij D Ryij
EXPERIENCE
THEand
POWER
OFrx;ykl =qx;yij . In what follows we
where ax;ykl
D qx;ykl =qx;yij
Rx;ykl D
use the operator
form of equations (6.18) and (6.19):
FULL ENGAGEMENT
u D Au C Bp C C
RUN FASTER.
63
RUN LONGER..
RUN EASIER
(6.20)
1349906_A6_4+0.indd 1
Download
free eBooks at bookboon.com
22-08-2014 12:56:57
63
xi1j C1
xij C1
4 xi1j
4 xij
.m1/
.m1/
.m1/
.m1/
C uyij
qyij 1 D uyij
qyij C1 D
uyij C uyij
1 ;
C1
4
4
n
n
uxij
uyij
rxijdynamics
; ryij D 2
D 2
Lectures on computational fluid
Finite Volume Method
t
t
.m/
.m/
.m/
.m/
axi1j u.m/
xi1j C uxij C axiC1j uxiC1j C axij 1 uxij 1 C axij C1 uxij C1 C (6.18)
.m/
.m/
C pi C1j pij =qxij D Rxij
.m/
.m/
.m/
.m/
.m/
ayi1j uyi1j
C uyij
C ayiC1j uyiC1j
C ayij 1 uyij
1 C ayij C1 uyij C1 C (6.19)
.m/
.m/
C pij C1 pij =qyij D Ryij
where ax;ykl D qx;ykl =qx;yij and Rx;ykl D rx;ykl =qx;yij . In what follows we
use the operator form of equations (6.18) and (6.19):
u D Au C Bp C C
6.6
(6.20)
Pressure correction
method
63
The velocity field satisfying the equation (6.20) is the solution of the linearized Navier Stokes equation. It doesnt fulfill the continuity equation.
The iterative solution satisfying the whole system of equations is computed
using the pressure correction method.
The iterative scheme consists of following steps. First, the intermediate solution is calculated with pressure taken from the previous iteration:
u D Au C Bp .m1/ C C
(6.21)
p .m/ D p .m1/ C p 0
(6.22)
are computed within next steps. Substitution of (6.22) into (6.20) gives
.u C u0 / D A.u C u0 / C B.p .m1/ C p 0 / C C
(6.23)
u0 D Au0 C Bp 0
(6.24)
Since u satisfies the equation (6.21) the equation for the velocity correction
reads
(6.25)
ru
what results in
.mC1/
D0
(6.26)
u0 D Au0 C Bp 0
(6.24)
u.m/ D u C Au0 C Bp 0
(6.25)
(6.26)
(6.27)
what results in
6.7
SIMPLE method
A very popular pressure correction method is the SIMPLE method. The main
assumption of this method is neglect of the term rAu0 in (6.27) and (6.24):
rBp 0 D ru
(6.28)
64
At each time instant the inner loop iterations are performed until residuals
are getting smaller than some threshold
.m1/
maxju.m/
x;yij ux;yij j < "u ;
As soon as the inner loop iterations are converged the solution at time instant n C 1 is equaling to the solution from the last iteration and the next
time
instant
computed.
The structure of the inner loop is shown in Fig. 6.2.
Download
freeis
eBooks
at bookboon.com
6.7.1
65
Let us consider the pressure correction equation (6.28) in details.This equation is solved for the control volume shown in Fig. 6.3. The divergency
.m1/
maxju.m/
x;yij ux;yij j < "u ;
As soon as the inner loop iterations are converged the solution at time instant n C 1 is equaling to the solution from the last iteration and the next
Lectures
on computational
fluid dynamics
Finite
Volume Method
time instant
is computed.
The structure of the inner loop is shown in Fig.
6.2.
6.7.1
Let us consider the pressure correction equation (6.28) in details.This equation is solved for the control volume shown in Fig. 6.3. The divergency
y
x
operator rf D @f
C @f
is represented within Finite Volume Method as
@x
@y
Z
Z
Z
follows
@fx @fy
Z rf d U D Z .
C
/d U D Z fndS D .fxe fxw Cfy n fys / (6.29)
@f
@f
@yy
U
U @xx
C
/d U D S65
rf d U D .
fndS D .fxe fxw Cfy n fys / (6.29)
@x
@y
U
U
S
Therefore, the right hand side of the equation (6.28) takes the form
Therefore, the right hand side of the equation (6.28) takes the form
Z
Z
Z ru d U D Z u ndS D .uxij uxi 1j C uyij
uyij
(6.30)
1 /
U
S
ru d U D
u ndS D .uxij uxi 1j C uyij uyij 1 /
(6.30)
U
S
0
As follows from (6.18) and (6.19) the operator Bp has the following values
at faces
of the
As
follows
fromcontrol
(6.18) volume
and (6.19) the operator Bp 0 has the following values
at faces of the control volume
0
0
0
0
0
0
.Bp /xij D pi C1j pij =qxij ; .Bp /xi 1j D pij pi 1j =qxi 1j
.Bp 0 /xij D pi0 C1j pij0 =qxij ; .Bp 0 /xi 1j D pij0 pi0 1j =qxi 1j
0
0
0
0
0
.Bp /yij D pij C1 pij =qyij ; .Bp /yij 1 D pij pij 1 =qyij 1 :
.Bp 0 /yij D pij0 C1 pij0 =qyij ; .Bp 0 /yij 1 D pij0 pij0 1 =qyij 1 :
(6.31)
Substitution of (6.30) and (6.31) into (6.28) results in
(6.31)
Substitution of (6.30) and (6.31) into (6.28) results in
pi0 C1j pij0 =qxij pij0 pi0 1j =qxi 1j C pij0 C1 pij0 =qyij
pi0 C1j pij0 =qxij pij0 pi0 1j =qxi 1j C pij0 C1 pij0 =qyij
0
or
or
0
0
pij pij 1 =qyij 1 D .uxij uxi 1j C uyij
uyij
1 /
0
0
pij pij 1 =qyij 1 D .uxij uxi 1j C uyij uyij 1 /
(6.32)
(6.32)
i C1j pi0 C1j C ij pij0 C i 1j pi0 1j C ij C1 pij0 C1 C ij 1 pij0 1 D cij (6.33)
i C1j pi0 C1j C ij pij0 C i 1j pi0 1j C ij C1 pij0 C1 C ij 1 pij0 1 D cij (6.33)
where
where
1
1
1
1
;
C
C
C
cij D .uxij uxi 1j Cuyij uyij 1 /; ij D
1
1
qxi11j qyij
qyij1 1
qxij
C
C
;
cij D .uxij uxi 1j Cuyij uyij 1 /; ij D GETCTHERE FASTER
qxi
qwhere
qSome
know
precisely
want
go. Others seek the adventure of
xij people
1j
yij theyq
yijto1
discovering
1
1
1 uncharted territory. Whatever
1 you want your professional journey to be,
; i 1j D
; ij C1 D youll nd; what
ijyoure
D for at Oliver Wyman.
:
i C1j D
1 looking
1
1
1
q
q
qyij the world of Oliver Wyman
q 1 oliverwyman.com/careers
; ij 1 D yijat1
:
i C1j D xij ; i 1j D xi 1j ; ij C1 D Discover
qxij
qxi 1j
qyij
qyij 1
DISCOVER
OUR WORLD
66
66
66
i C1j
xij
xi 1j
xij
ij
.Bpon
/yij
D pij0 C1
pij0dynamics
=qyij ; .Bp 0 /yij 1
Lectures
computational
fluid
ij
i 1j
pij0
pij0 1
xi 1j
=qyijFinite
1 : Volume Method
(6.31)
pi0 C1j
pij0
pij0
pij0 1
=qxij
pij0
pi0 1j
=qxi 1j C
pij0 C1
pij0
=qyij
=qyij 1 D .uxij uxi 1j C uyij
uyij
1 /
(6.32)
or
i C1j pi0 C1j C ij pij0 C i 1j pi0 1j C ij C1 pij0 C1 C ij 1 pij0 1 D cij
(6.33)
where
cij D
.uxij
uxi 1j
i C1j D
6.7.2
Cuyij
1
qxij
uyij
1 /; ij
; i 1j D
1
qxi 1j
D
1
qxij
; ij C1 D
1
qyij
1
qxi 1j
; ij 1 D
1
qyij
1
qyij 1
1
qyij 1
ayNy uyN
C uy
C ayCNy uyCN
C ay1 uy1
C ayC1 uyC1
C
y
y
.m1/
p.m1/ =qy D Ry
C pC1
Download
free eBooksof
at bookboon.com
ii) Calculation
the pressure
algebraic equations:
0
0
0
0
CNy pCN
C p0 C Ny pN
C C1 pC1
C 1 p1
D c
y
y
0
p.m1/
C pCNy
=qx D Rx
ayNy uyN
C uy
C ayCNy uyCN
C ay1 uy1
C ayC1 uyC1
C
y
y
.m1/
.m1/
=qy D Ry
C pC1 p
D
0
pCN
y
D
0
pC1
p0
p0
=qx
=qy
p.m/ D p.m1/ C p0
67
Figure 6.3: Control volume used for the pressure correction equation.
v) Check the difference between two iterations:
.m1/
maxju.m/
x;y ux;y j < "u ;
68
u.m1/
x;y
.m1/
u.m/
x;y ; p
D p.m/
and go to the step i). Otherwise the calculation at the time moment
Lectures
computational
fluid dynamics
Figureon6.3:
Control volume
used for the pressure correction equation.Finite Volume Method
and go to the step i). Otherwise the calculation at the time moment
n C 1 is completed
.m/
nC1
unC1
D p.m/
x;y D ux;y ; p
68
69
7.1
SIMPLE algorithm
7.1linearized
SIMPLE
algorithm
The
Navier Stokes
equation written in operator form is
7.1
SIMPLE algorithm
(7.1)
(7.2)
(7.3)
(7.2)
(7.2)
(7.3)
(7.3)
(7.4)
(7.4)
(7.4)
(7.5)
(7.5)
(7.5)
7.2.1
First iteration
The term Au0 is neglected, i.e. Au0 D 0. The pressure correction is found
from the Poisson equation
rBp 0 D ru
(7.6)
70
u0 D Bp 0
(7.7)
7.2.1
First iteration
The term Au0 is neglected, i.e. Au0 D 0. The pressure correction is found
from the Poisson equation
Lectures on computational fluid dynamics
7.2.2
rBp D ru
(7.6)
u0 D Bp 0
(7.7)
Second iteration
The pressure correction within the second iteration is found from the Poisson
equation
rBp 00 D rAu0
(7.8)
u00 D Au0 C Bp 00
(7.9)
7.2.3
Correction
(7.10)
u D Au C Bp
.m1/
(7.11)
70
C C; u0 D Bp 0 ; u00 D Au0 C Bp 00
the equation (7.11) is not satisfied. The residual is Au00 . The residual can be
reduced within next iterations. However, usually, PISO algorithm uses only
two iterations.
7.2.4
Summary
(7.12)
rBp 0 D ru
71
(7.13)
Calculation
Calculation of
of the
the auxiliary
auxiliary velocity
velocity
Calculation of the auxiliary velocity
.m1/
(7.12)
u
D Au
Au C
C Bp
Bp .m1/ C
CC
C
(7.12)
u D
.m1/
D
Au
C
Bp
C
C
u
Lectures on computational fluid dynamics
Overview of pressure(7.12)
correction methods
0
Calculation
Calculation of
of the
the pressure
pressure correction
correction p
p 0 ::
Calculation of the pressure correction p 0 :
0
rBp
D ru
ru
rBp 0 D
rBp 0 D ru
(7.13)
(7.13)
(7.13)
0
Calculation
Calculation of
of the
the velocity
velocity correction
correction u
u0
Calculation of the velocity correction u0
0
0
u
u0 D
D Bp
Bp 0
u0 D Bp 0
(7.14)
(7.14)
(7.14)
00
Calculation
Calculation of
of the
the pressure
pressure correction
correction p
p 00 ::
Calculation of the pressure correction p 00 :
00
0
rBp
D rAu
rAu0
rBp 00 D
rBp 00 D rAu0
(7.15)
(7.15)
(7.15)
00
Calculation
Calculation of
of the
the velocity
velocity correction
correction u
u00
Calculation of the velocity correction u00
00
0
00
u
u00 D
D Au
Au0 C
C Bp
Bp 00
u00 D Au0 C Bp 00
Correction
Correction
Correction
(7.16)
(7.16)
(7.16)
0
00
.m/
.m1/
0
00
.m/ D u C u0 C u00 ; p .m/ D p .m1/ C p 0 C p 00
u.m/
D u C u C u ;p
Dp
Cp Cp
u
u.m/ D u C u0 C u00 ; p .m/ D p .m1/ C p 0 C p 00
(7.17)
(7.17)
(7.17)
Both
Both algorithms
algorithms PISO
PISO and
and SIMPLE
SIMPLE are
are widely
widely used
used in
in CFD
CFD codes.
codes.
Both algorithms PISO and SIMPLE are widely used in CFD codes.
71
71
71
In the past four years we have drilled
89,000 km
Thats more than twice around the world.
careers.slb.com
Based on Fortune 500 ranking 2011. Copyright 2015 Schlumberger. All rights reserved.
72
7.3
SIMPLEC algorithm
Another way to hold the term rAu0 (see 6.25) is implemented in the SIMPLEC algorithm. The velocity correction at th control volume u0 can
be calculated using the interpolation over N adjacent control volumes:
N
X
D1
A u0 u0
N
X
(7.18)
D1
where is the number of adjacent control volumes around the control volume
with the number .
u0
PN
0
D1 A u
PN
D1 A
(7.19)
u0 D Au0 C Bp 0
(7.20)
or
u0
N
X
D1
A u0 C Bp 0
(7.21)
u0
N
X
D1
A C Bp 0
(7.22)
and
u0 D
1
Bp 0
PN
(7.23)
D1 A
rBp D ru rA
or
r B CA
1
B
PN72
1
D1 A
Bp 0
PN
D1
p 0 D ru
(7.24)
The computational steps are the same as these in SIMPLE algorithm with
only difference that the equation (7.24) is solved instead of (7.3).
Download free eBooks at bookboon.com
73
Chapter 8
Chapter
8
Chapter
8
Computational
grids
Computational
Computational grids
grids
8.1
Computational grids
Grid types
The
grids are subdivided into:
8.1computational
Grid types
8.1
Grid types
structured grids
(seeare
Fig.
8.1a), into:
The computational
grids
subdivided
The computational grids are subdivided into:
block
structured
gridsFig.
(see8.1a),
Fig. 8.1b),
structured
grids (see
structured grids (see Fig. 8.1a),
unstructured
grids
(see(see
Fig.Fig.
8.1c).
block structured
grids
8.1b),
block structured grids (see Fig. 8.1b),
Disadvantage
of thegrids
structured
grid
is shown in Fig. 8.2. Refinement of the
unstructured
(see Fig.
8.1c).
gridclose
to the wallgrids
results
inFig.
the refinement
in areas where this refinement is
unstructured
(see
8.1c).
Disadvantage
thedisadvantage
structured grid
in Fig.
8.2. ofRefinement
of the
not
necessary. of
This
canisbeshown
overcome
by use
block-structured
Disadvantage
ofwall
the results
structured
grid
is shown
Fig. where
8.2. Refinement
of the
grid
close
toand
the
in grid
the
refinement
this refinement
is
(Fig.
8.1b)
unstructured
(Fig.
8.1c).ininareas
grid
close
to
the
wall
results
in
the
refinement
in
areas
where
this
refinement
is
not necessary. This disadvantage can be overcome by use of block-structured
not necessary.
disadvantage
be 8.1c).
overcome by use of block-structured
(Fig.
8.1b) andThis
unstructured
gridcan
(Fig.
(Fig. 8.1b) and unstructured grid (Fig. 8.1c).
8.2
Overset or Chimera74grids
For complicated objects one uses overset or Chimera grids. The idea of
chimera or overset grids is to generate the grids separately around each geo-
grid resolution should be high especially in areas of boundary layers and close
to the free surface. For this sake the special refinement is used in these areas.
To increase the accuracy of the computations in boundary layers one uses
specialongrid
boundaryfluid
layers
close to walls.
Lectures
computational
dynamics
Computational grids
8.2
For complicated objects one uses overset or Chimera grids. The idea of
chimera or overset grids is to generate the grids separately around each geometrical entity in the computational domain. After that the grids are combined together in such a way that they overlap each other where they meet.
The crucial operation is an accurate transfer of quantities between the different grids at the overlapping region. The most important advantage of the
overset or Chimera grid is the possibility to generate high quality structured
particular grids separately for different body elements completely independent of each other, without having to take care of the interface between
grids.
8.3
Morphing grids
Very efficient way of CFD body simulation is the use of moving or morphing
grids [6]. The idea is the computational grid is moved in accordance with the
displacement of the body by using an analytical weighted regridding which
is a type of extrapolation of rigid transformation. The possible problem of
morphing grid is poor quality caused by its motion. Consequently if the
mesh surrounding the body is allowed to deform the elements around the
body deform. This can quickly lead to poor quality elements if care is not
taken. An alternative method is to replicate the motion of the body with the
fluid domain split into an inner and outer region. The outer domain remains
fixed in space while the inner domain containing the body moves laterally to
replicate the motion. The mesh in the inner sub domain remains locked in
position relative to the lateral motion of the body. This prevents deformation
of the detailed mesh around the body.
76 The outer mesh is deformed due to
the motion of the inner region.
If moving grids are used the Navier Stokes should be transformed to take the
velocity of grid faces UEg into account,
o
1
@E
u n
C uE UEg r uE D fE rp C E
u
@t
(8.1)
Thomas and Lombard have shown that the function UEg can not be arbitrary rather than they have to be found from the Geometric Conservation
Law
Z
Z
@
d U UEg nE dS D 0
(8.2)
@t
U
Where U and S are respectively volume and surface of cells. The equation (8.2) is derived from the condition that the computation of the control
volumes or of the grid velocities must be performed in a such a way that
the resulting numerical scheme preserves the state of the uniform flow, independently
of theatdeformation
Download free eBooks
bookboon.com of the grid. The equation (8.2) is satisfied
automatically if the control volumes dont change their shape. The Geomet75
ric conservation law (8.2) should solve coupled with other fluid flow equations
using the same discretizations schemes.
More detailed information about grid generation can be found in [3], [7]
trary rather than they have to be found from the Geometric Conservation
Law
Z
Z
@
Lectures on computational fluid dynamics
Computational
grids
d U UEg nE dS D 0
(8.2)
@t
U
Where U and S are respectively volume and surface of cells. The equation (8.2) is derived from the condition that the computation of the control
volumes or of the grid velocities must be performed in a such a way that
the resulting numerical scheme preserves the state of the uniform flow, independently of the deformation of the grid. The equation (8.2) is satisfied
automatically if the control volumes dont change their shape. The Geometric conservation law (8.2) should solve coupled with other fluid flow equations
using the same discretizations schemes.
More detailed information about grid generation can be found in [3], [7]
and [8].
Hellmanns is one of Unilevers oldest brands having been popular for over 100 years.
If you too share a passion for discovery and innovation we will give you the tools and
opportunities to provide you with a challenging career. Are you a great scientist who
would like to be at the forefront of scientic innovations and developments? Then you will
enjoy a career within Unilever Research & Development. For challenging job opportunities,
please visit www.unilever.com/rdjobs.
76
Part II
Mathematical modelling of
turbulent flows
77
79
Chapter 9
Chapter
9
Chapter
9 turbulence
Physics of
Physics
Physics of
of turbulence
turbulence
9.1
Physics of turbulence
Flow motions
are subdivided
into laminar
flows and turbulent ones. The
9.1
Definition
of the
turbulence
9.1
Definition
ofmeans
thelayer.
turbulence
word Laminar
in Greek
The fluid particles move orderly
Flow
motions
are intense
subdivided
intomixing.
laminarThe
flows
and turbulent
ones.
The
in
layers
without
lateral
disruption
between
layers
is
Flow
motions
are
subdivided
into
laminar
flows
and
turbulent
ones.
The
word
Laminar
in
Greek
means
layer.
The
fluid
particles
move
orderly
absent. On the contrary the turbulent flow is very chaotic with strong eddies
word
Laminar
inacross
Greek
layer. The
The disruption
fluid particles
movelayers
orderly
in layers
without
lateral
mixing.
between
is
and
intense
mixingintense
themeans
flow.
in
layers
without
intense
lateral
mixing.
The
disruption
between
layers
is
absent.
On
the
contrary
the
turbulent
flow
is
very
chaotic
with
strong
eddies
Turbulent motion is the three dimensional unsteady flow motion with
absent.
On
the
contrary
the
turbulent
flow
is
very
chaotic
with
strong
eddies
and intense mixing across the flow.
and
mixing
the
flow.
Turbulent
motion
isacross
the three
dimensional
intense
chaotical
trajectories
of fluid
particles,unsteady flow motion with
Turbulent motion is the three dimensional unsteady flow motion with
fluctuations
of the velocity
and
chaotical trajectories
of fluid
particles,
chaotical trajectories of fluid particles,
strong
mixing
fluctuations
of the velocity and
fluctuations of the velocity and
strong
mixing
arisen
at large
Re numbers due to unstable vortex dynamics.
strong mixing
arisen at large Re numbers due to unstable vortex dynamics.
arisen
large Re numbers
due to unstable vortex dynamics.
9.2 atVortex
dynamics
The
dynamics dynamics
is the key to understand what happens in the turbulent
9.2vortex
Vortex
9.2
Vortex dynamics
flow.
The vortex dynamics is the key to understand what happens in the turbulent
The
flow.vortex dynamics is the key to understand what happens in the turbulent
9.2.1
Vorticity transport equation
flow.
The
vectorVorticity
calculus relation
reads: equation
9.2.1
transport
9.2.1
1relation reads:
r.A A/ D A .r A/ C .Ar/A
2relation reads:
1
.r A/ C .Ar/A
Taking u D A we get:12 r.A A/ D A 81
r.A A/ D A .r A/ C .Ar/A
2
1
r.u u/ D u81 ! C .ur/u
81
2
where ! D r u is the vorticity.
Application of the curl operator to (9.2) results in
The vector calculus
The vector calculus
(9.1)
(9.1)
(9.1)
(9.2)
(9.4)
@u
1
C .ur/u/ D r . rp C u/
(9.5)
Physics of turbulence
. 12 r.u
u// D 0.
Here we used the identity r
Both vectors u and ! satisfy the continuity equation, i.e. r! D 0 and
ru D 0:
r ..ur/u/ D .!r/u C .ur/!
(9.4)
1
@u
C .ur/u/ D r . rp C u/
@t
(9.5)
@!
C r ..ur/u/ D .r u/ D !
@t
Substituting (9.4) into (9.6) results in
(9.6)
@!
C .ur/! D .!r/u C !
@t
360
thinking
D!
D .!r/u C !
Dt
The equation (9.8) is the vorticity transport equation.
9.2.2
(9.7)
(9.8)
The vortices are main players in turbulent flows. Here we would like to
emphasize the difference between the vorticity and vortices. The vorticity
is the curl of the velocity ! D r u. The vorticity is usually not zero in
viscous flows especially in areas close to the walls. Speaking about vortices
we bear in mind the concentrated structures of the vorticity field ! D r u.
The difference between the vorticity and vortices is illustrated in Fig. 9.1.
The boundary layer is the flow area with strong but smoothly distributed
vorticity (Fig. 9.1a). Due to instabilities, that will be discussed later, the
concentrated vortex structures arise in the smooth vorticity field (Fig. 9.1b).
360
thinking
82
360
thinking
Dis
@t
D!
D .!r/u C !
Dt
Lectures
on computational
The equation
(9.8) is fluid
the dynamics
vorticity transport equation.
9.2.2
(9.8)
Physics of turbulence
The vortices are main players in turbulent flows. Here we would like to
emphasize the difference between the vorticity and vortices. The vorticity
is the curl of the velocity ! D r u. The vorticity is usually not zero in
viscous flows especially in areas close to the walls. Speaking about vortices
we bear in mind the concentrated structures of the vorticity field ! D r u.
The difference between the vorticity and vortices is illustrated in Fig. 9.1.
The boundary layer is the flow area with strong but smoothly distributed
vorticity (Fig. 9.1a). Due to instabilities, that will be discussed later, the
concentrated vortex structures arise in the smooth vorticity field (Fig. 9.1b).
A famous sample of concentrated vortex structures is the tornado (Fig. 9.2).
82
The vorticity is solenoidal:
r! D r.r u/ D 0
(9.9)
9.2.3
Vortex amplification as an important mechanism
Download free eBooks at bookboon.com
of the turbulence generation
80
As mentioned above the vortices are main players during the laminar- turbulent transition. The vorticity can be essentially intensified (amplified) due
to action of neighboring vortices or even due to self induction.
9.2.3
Physics of turbulence
As mentioned above the vortices are main players during the laminar- turbulent transition. The vorticity can be essentially intensified (amplified) due
to action of neighboring vortices or even due to self induction.
83
81
The r.h.s. of (9.11) contains two terms. The first term .!r/u is responsible
for the rotation of the vorticity vector ! and enlargement or reduction of its
magnitude j!j. The second diffusion term results in spreading of the vorticity in the space. The term .!r/u is responsible for the amplification of the
Lectures on computational fluid dynamics
Physics of turbulence
vorticity.
The effect of the amplification can easily be understood if we consider the
vortex with vector aligned along the x-axis !x > 0. If such a vortex is in the
x
x
fluid stretching area @u
> 0, the term !x @u
> 0 is positive. As a result,
@x
@x
D!x
> 0 is positive, what leads to the increase of the vorticity !x . As shown
Dt
analytically by Novikov [9] for a simple model problem, the vortex strength
84
this termfree
willeBooks
be introduced
in the next chapter
Download
at bookboon.com
85
82
Lectures
computational
fluid dynamics
vorticesoninduce
self induced
velocities.
Physics of turbulence
85
83
nario (see Fig. 9.6). Let the vorticity at a certain point of the flow grows.
In the real physical process folding prevents the growth of the kinetic energy
and increases the canceling effect of viscosity. Then the vortex structures
breaks on
down
into small
structures
Lectures
computational
fluid
dynamics due to reconnection mechanism described
Physics of turbulence
in the next subsection.
9.2.4
Vortex reconnection
Let us consider the vortex ring (see Fig. 9.7). Due to convective instability
or influence of neighboring vortices the vortex ring is deformed. Due to self
86
Figurefluid
9.7:
Scenario of vortex reconnection.
Lectures on computational
dynamics
Physics of turbulence
induction two opposite sides of the ring are merged. As soon as two elements
with opposite vorticity sign are approaching each to other, they start to
cancel each other by mutual diffusion. The vortices disappear in the area of
the contact. Two small vortices are created from one big vortex. Each small
vortex ring breaks then down into smaller vortices and so on. The energy
of small vortices is equal to the energy of the big vortex with a small loss
caused by the dissipation. This fact is formulated in the sentence, which is
common in the turbulence theory: The energy is transferred from large scale
vortices to small scales vortices. The reconnection process can be observed
on macroscales. The decay, break up of tip vortices behind the airplane
proceeds according to the same scenario (see Fig. 9.8). The reconnection
process is, perhaps, the main mechanism of vortex cascade in the turbulent
flows, i.e. transformation of big vortices into small ones. The reconnection
can also lead to enlargement of small vortices if two rings approach each to
other as shown in Fig. 9.7 (see red circle). In this case the energy of small
vortices turns into the energy of the big vortex. This process is referred to
as the energy back scattering. Statistically, the direct energy flux sufficiently
surpasses the backward one.
9.2.5
The vortex turbulence cascade means that large eddies break down to form
small eddies as turbulence cascades from large scales to small ones. This idea
was formulated in the famous poem by Richardson (1922):
Big whorls have little whorls,
The Wake
87
.QYURGGF'PIKPGU/GFKWOURGGF'PIKPGU6WTDQEJCTIGTU2TQRGNNGTU2TQRWNUKQP2CEMCIGU2TKOG5GTX
6JGFGUKIPQHGEQHTKGPFN[OCTKPGRQYGTCPFRTQRWNUKQPUQNWVKQPUKUETWEKCNHQT/#0&KGUGN6WTDQ
2QYGTEQORGVGPEKGUCTGQHHGTGFYKVJVJGYQTNFoUNCTIGUVGPIKPGRTQITCOOGsJCXKPIQWVRWVUURCPPKPI
HTQOVQM9RGTGPIKPG)GVWRHTQPV
(KPFQWVOQTGCVYYYOCPFKGUGNVWTDQEQO
85
other as shown in Fig. 9.7 (see red circle). In this case the energy of small
vortices turns into the energy of the big vortex. This process is referred to
as the energy back scattering. Statistically, the direct energy flux sufficiently
surpasses
the backward
Lectures
on computational
fluidone.
dynamics
Physics of turbulence
9.2.5
The vortex turbulence cascade means that large eddies break down to form
small eddies as turbulence cascades from large scales to small ones. This idea
was formulated in the famous poem by Richardson (1922):
Big whorls have little whorls,
87
Figure 9.8: Sample of the vortex reconnection of tip vortices behind an airFigure
plane. 9.8: Sample of the vortex reconnection of tip vortices behind an airplane.
Figure 9.8: Sample of the vortex reconnection of tip vortices behind an airplane.
Which feed on their velocity;
feed onhave
theirlessor
velocity;
AndWhich
little whorls
whorls,
Which
feed
on
their
velocity;
And
little
whorls
have
lessor
whorls,
And so on to viscosity (in the molecular sense).
And
whorls (in
havethe
lessor
whorls,sense).
And so
on little
to viscosity
molecular
And so on to viscosity (in the molecular sense).
9.2.6 Summary
9.2.6 Summary
Vortex arise
in the fluid due to viscosity effects. They experience instability
9.2.6
Summary
Vortex
arise in the fluid
due to
viscosity effects.
They experience
and amplification.
Diffusion
counteracts
the amplification.
If theinstability
Reynolds
and
amplification.
Diffusion
counteracts
the
amplification.
If
the
Reynolds
Vortex
arise
in
the
fluid
due
to
viscosity
effects.
They
experience
instability
number is large, the vortex structures are strong and concentrated.
The
number
is
large,
the
vortex
structures
are
strong
and
concentrated.
The
and
amplification.
Diffusion
counteracts
the
amplification.
If
the
Reynolds
amplification can dominate at some fluid region over the diffusion.
The
amplification
can the
at
some
fluid
diffusion.
The
number
is large,
vortex
structures
are region
strong
and
concentrated.
vortex instability
isdominate
not
damped
by viscosity.
Theover
flow the
becomes
stochastic
vortex
instability
isdominate
not damped
by viscosity.
The
flow
becomes
stochastic
amplification
can
at some
fluid region
the
diffusion.
The
due to mutual
interaction
of unstable
vortices.
Theover
big vortices
break
down
due
to
mutual
interaction
of
unstable
vortices.
The
big
vortices
break
down
vortex
instability
is
not
damped
by
viscosity.
The
flow
becomes
stochastic
into small ones by means of vortex reconnection. The vortex instability
into
small
onesinteraction
by as
means
ofunstable
vortex reconnection.
The
vortexbreak
instability
due
to
mutual
vortices. The big
vortices
down
process
is identified
the of
turbulence.
process
is
identified
as
the
turbulence.
into small ones by means of vortex reconnection. The vortex instability
process is identified as the turbulence.
Different
regimes
of the
fluid
motion da
were
revealed
very early,
perhaps,
in
antique times.
Much
later,
Leonardo
Vinci
recognized
two states
of the
antique
times.
Much
later,
Leonardo
Vinci
recognized
two states
of the
Different
regimes
of
the
fluid
motion
were
revealed
very early,
perhaps,
in
fluid motion
and
introduced
the
termda
la
turbolenza.
Arkady
Tsinober
fluid
motion
and
introduced
the
term
la
turbolenza.
Arkady
Tsinober
antique
times.
Much
later,
Leonardo
da
Vinci
recognized
two
states
of
the
in his book An informal introduction to turbulence [1] presented most
in
hismotion
book results
An
introduction
tointurbulence
[1]Arkady
presented
most
fluid
and informal
introduced
the research
term la
turbolenza.
Tsinober
outstanding
in turbulence
chronological
order
(Fig.
9.9).
outstanding
results
in
turbulence
research
in
chronological
order
(Fig.
9.9).
in
his book An informal introduction to turbulence [1] presented most
Download free eBooks at bookboon.com
outstanding results in turbulence research
in chronological order (Fig. 9.9).
88
86
88
88
Physics of turbulence
9.3.1
Quantitative study of turbulence was started by Osborne Reynolds (18421912) who performed in 1883 very famous experiment shown in Fig. 9.10.
The water flows from the vessel A to the pipe B. The ink injected into the
pipe B with the local flow velocity is not mixed in transversal direction and
keeps its identity if the flow velocity is small (Fig. 9.10, right). The flow
under such a condition is laminar. As soon as the flow velocity increases due
to water level raise in the vessel A, the ink jet loses the stability and is mixed
with surrounding water (Fig. 9.10, right). The flow becomes turbulent. Ink
jet development at different flow velocities in the circular pipe is shown in
Fig. 9.11.
The great merit of Reynolds lies in the fact, that he in contrast to his predecessors quantified the laminar turbulent transition. He showed that the
transition in pipes occurs if the Reynolds number Re D Ub D= exceeds the
threshold around 2400. Here Ub is the bulk velocity in pipe determined as
the ratio of the flow rate to the pipe cross section, i.e. Ub D Q=.D 2 =4/, D is
the pipe diameter and is the kinematic viscosity coefficient ( 106 m2 s 1
for water and 106 m2 s 1 for air). Later, it was shown that the transition
strongly depends on the perturbations presented in the flow. The experimental setup of Reynolds has been preserved at the University of Manchester in
89
87
Physics of turbulence
Figure 9.11: Development of instability during the laminar- turbulent transition in the circular pipe (taken from [1]).
UK. The experiments done nowadays shown that the laminar- turbulent transition Reynolds number is less than that documented originally by Reynolds.
The reason is, presumably, the building vibration and noise caused by traffic
which was not in time of Reynolds. If the perturbations are eliminated the
transition can be delayed up to Re 40000:::50000.
90
9.3.2
The jet flows experiences also laminar turbulent transition shown in Fig. 9.12.
Download free
eBooks
bookboon.com
Obviously,
the
flowatcan
be fully or only partially turbulent. Close to the nozzle the flow is laminar. The instability is developed
downstream. The shear
88
flow at the jet boundary is the area of rapid velocity change from the jet
velocity to zero outside of the jet. The shear flow experiences the so called
Kelvin Helmholtz instability (Fig. 9.13) resulting in formation of concen-
Lectures
on computational
fluid dynamics
transition
can be delayed
up to Re 40000:::50000.
9.3.2
Physics of turbulence
The jet flows experiences also laminar turbulent transition shown in Fig. 9.12.
Obviously, the flow can be fully or only partially turbulent. Close to the nozzle the flow is laminar. The instability is developed downstream. The shear
flow at the jet boundary is the area of rapid velocity change from the jet
velocity to zero outside of the jet. The shear flow experiences the so called
Kelvin Helmholtz instability (Fig. 9.13) resulting in formation of concentrated vortices which are approximately circular. It happens close to jet
nozzle at x=D < 1 (see Fig. 9.14), where x is the distance from the nozzle
and D is the nozzle diameter.
The Kelvin Helmholtz vortices experience then the pairing (see Fig. 9.14).
One vortex overtakes the neighbor vortex creating a pair. This process
has an inviscid convective nature and can be explained thinking back to
the famous leapfrog motion of two vortex rings. In the inviscid flow the
leapfrog motion is running as long as the convective instability destroys
the vortices. One vortex runs the next down, its radius decreases whereas
the speed increases. The radius of the next vortex increases, the speed
decreases. The first ring moves through the second one. The process is
then repeated. The movie illustrating this process can be downloaded from
http://www.lemos.uni-rostock.de/en/gallery/.
The paired vortices experience the azimuthal instability and takes the crude
ring form. Later they are destroyed downstream in the region 1 < x=D < 6.
In far field at large x=D the vortex structures look like a tree with branches
oriented against the main flow direction (see Fig. 9.15).
89
the vortices. One vortex runs the next down, its radius decreases whereas
the speed increases. The radius of the next vortex increases, the speed
decreases. The first ring moves through the second one. The process is
then repeated. The movie illustrating this process can be downloaded from
Lectures on computational fluid dynamics
Physics of turbulence
http://www.lemos.uni-rostock.de/en/gallery/.
The paired vortices experience the azimuthal instability and takes the crude
ring form. Later they are destroyed downstream in the region 1 < x=D < 6.
In far field at large x=D the vortex structures look like a tree with branches
oriented against the main flow direction (see Fig. 9.15).
Physics of turbulence
Creation of vortex rings is the reason for the jet noise. The noise produced,
for instance, by jet propulsors of airplanes is the action of these vortices.
The vortices play a significant positive role in jet mixers widely used in food
industry, chemical engineering, etc. That is why, one of the most perspective
way to reduce the noise or to increase mixing is the manipulation of vortices arising behind the jet nozzle. To increase the mixing it is necessary to
strengthen the Kelvin Helmholtz vortices. To decrease the noise the vortices
should be broken down into small ones. Fig. 9.16 shows the effect of the
acoustic impact on jet. The original vortices (lower picture) are split into
small ones (upper picture).
High resolution laser diagnostics methods LIF and PIV allow to get a deep
inside into the structure of the turbulent flow. Fig. 9.17 shows the structure
of the confined jet mixer flow displayed by Planar Laser Induced Fluorescence (PLIF) Method. The macrostructure obtained with low resolution is
shown in the upper Figure. A small window with sizes 2:08mm 2:72mm
92
Figure
structures
Download
free 9.16:
eBooksVortex
at bookboon.com
91
was selected for high resolution PLIF measurements. The vortex microstrucwas selected for high resolution PLIF measurements. The vortex microstrucwas
for highinresolution
measurements.
The vortex
microstructuresselected
are presented
the lowerPLIF
Figure.
A very important
observation
is the
tures are presented in the lower Figure. A very important observation is the
Physics of turbulence
9.3.3
The boundary layer on a plate is the thin layer of rapid change of the velocity
from zero to 99:5% of the incident flow. A possible scenario of the laminar
turbulent transition in the boundary layer on a flat plate is shown in Fig. 9.19
First, the transversal vortices are generated in the boundary layer due to
the so called Tollmien- Schlichting instability [12]. They experience the secondary instability and form downstream the lambda structures. The latter
interact each with other and experience the tertiary instability. They loss
original regular form and become stochastic. An important feature of the
turbulent boundary layer is the presence of streaks (strips) of the low velocity fluid regions (see Fig. 9.20). They arise due to induction of lambda
94
92
Physics of turbulence
Figure 9.18: Fine vortex structures in a confined jet mixer flow. PLIF measurements by Valery Zhdanov (LTT Rostock). Spatial resolution is 31m.
structures schematically shown in Fig. 9.21.
9.3.4
CAREERKICKSTART
u y
in the know
Whether youre a graduate, school leaver or student, its a difcult time to start your career.
So hereinatwall
RBS,turbulence
were providing
helping
hand with our
app.laminar
Bringing together
The stress
flowacan
be considered
asnew
theFacebook
sum of the
the most relevant and useful careers information, weve created a one-stop shop designed
and turbulent stresses:
to help you get on the career ladder whatever your level of education, degree subject or
work experience.
D l C t
(9.12)
And its not just nance-focused either. Thats because its not about us. Its about you.
So download the app and youll get everything you need to know to kickstart your career.
So what are you waiting for?
95
93
Figure 9.18: Fine vortex structures in a confined jet mixer flow. PLIF measurements by Valery Zhdanov (LTT Rostock). Spatial resolution is 31m.
Lectures
on computational
fluidshown
dynamics
structures
schematically
in Fig. 9.21.
9.3.4
Physics of turbulence
u y
The stress in wall turbulence flow can be considered as the sum of the laminar
and turbulent stresses:
D l C t
(9.12)
95
w
(9.13)
Applying the Newton hypothesis (1.18) to the two dimensional wall bounded
(9.13)
Applying the Newton hypothesis (1.18) to the two dimensional wall bounded
flow, one gets from (9.13)
w D
dux
dy
(9.14)
or
ux
D yC C C
u
(9.15)
From the condition at the wall ux D 0 the unknown constant C is zero, i.e.
96
Figure
9.22:
Vertical
distribution of the velocity ux at three different time
Download
free eBooks
at bookboon.com
Figure
9.22:
Vertical
distribution of the velocity ux at three different time
instants in boundary layer.
95
instants in boundary layer.
ux
C
ux D y C
u D y
(9.16)
(9.16)
Physics of turbulence
(9.17)
ux D lim
T !1
ZT
(9.18)
ux .t/dt
(9.19)
ux0 2 D
98
d ux96
lx
dy
Click on(9.20)
the ad to read more
D u0x uy0
(9.19)
Physics of turbulence
ux0 2 D
d ux
lx
dy
(9.20)
98
Root square of the averaged squared pulsation
in vertical direction is written
in a similar form:
q
d ux
ly
uy0 2 D
(9.21)
dy
q q
Introducing the correlation coefficient Rxy D u0x uy0 = uy0 2 uy0 2 and using
(9.20) and (9.21) one gets:
j12 j D j t j D Rxy lx ly
d ux
dy
2
D l
d ux
dy
2
(9.22)
(9.23)
where A is the Van Driest constant, which is equal to 26 or 27. In shear flows
97
l D C onst .x/, where is the shear layer thickness.
We consider again the thin boundary layer, i.e. the stress is approximately
equal to the wall stress w D 12 . Using (9.22) and (9.23) we get
d ux
dy
2
w
u
D
y
(9.25)
(9.26)
(9.27)
98
Chapter 10
Chapter
10
Chapter
10
Basic definitions
of the
Basic
statistical
theory of
of the
turbulence
Basic definitions
definitions
of
the
statistical
statistical theory
theory of
of turbulence
turbulence
10.1
Reynolds averaging
Reynolds proposed
to represent
any stochastic quantity in turbulent flow as
10.1
Reynolds
averaging
10.1
Reynolds
averaging
the sum of its
averaged part and
fluctuation. For instance, this representation
Reynoldsforproposed
represent any
applied
velocity to
components
readsstochastic quantity in turbulent flow as
Reynolds
to represent
any stochastic
turbulent
flow as
the
sum ofproposed
its averaged
part and fluctuation.
Forquantity
instance,inthis
representation
the
sum
of
its
averaged
part
and
fluctuation.
For
instance,
this
representation
0
0
0
applied for velocity
ux D ucomponents
N x C ux I uyreads
D uN y C uy I uz D uN z C uz I
(10.1)
applied for velocity components
reads
The Reynolds uaveraged
N x Cvelocities
u0x I uy are
D uN y C uy0 I uz D uN z C u0z I
(10.1)
x D u
0
ux D uN x C ux I uy D uN y C uy0 I uz D uN z C u0z9I
(10.1)
ZT
ZT
The Reynolds averaged velocities
are
>
>
1
1
>
The Reynolds uNaveraged
velocities
are
>
ux dt
I uN y D lim
uy dtI>
x D lim
9
>
>
T !1 T ZT
T !1 T ZT
=
>
9
>
0
0
1 ZT
1 ZT
>
(10.2)
>
uN x D lim 1 ZT ux dt I uN y D lim 1
uy dtI>
>
>
T !1 T
T !1 T
=
>
uN x D lim 1 0 ux dt I uN y D lim
uy dtI>
>
T !1 T 0
!1 T
uz dt
uN z D Tlim
>
=
(10.2)
>
T !1 T Z
;
0T
0
>
>
(10.2)
0
>
1 T
>
>
uN z D lim 1 Z uz dt
>
>
!1 T
;
>
The averaged fluctuation
is zero
uz dt
uN z D Tlim
>
>
0
T !1 T
;
0
The averaged fluctuation is zero u0
D0
The averaged fluctuation is zero x;y;z
The root of the averaged square uof0 fluctuations
is called root mean square,
x;y;z D 0
or r.m.s.. The quantity averagedu0x;y;z
twiceDis0 equal to quantity averaged once
The
root of the averaged square of fluctuations is called root mean square,
u D u.
The
root
ofThe
thequantity
averaged
square
oftwice
fluctuations
istocalled
rootr.m.s
mean
square,
orthe
r.m.s..
is equal(for
quantity
averaged
once
If
turbulence
process isaveraged
statistically
unsteady
instance
is changed
or
r.m.s..
The
quantity
averaged
twice
is
equal
to
quantity
averaged
uD
u. the definition of the Reynolds averaging (10.2) is not applicableonce
in
time),
and
u
D
If theu.turbulence process is statistically unsteady (for instance r.m.s is changed
If
turbulence
processofisthe
statistically
instance
is changed
in the
time),
the definition
Reynolds
averaging(for
(10.2)
is notr.m.s
applicable
and
101unsteady
in time), the definition of the Reynolds averaging (10.2) is not applicable and
101
101
Designed for high-achieving graduates across all disciplines, London Business Schools Masters
Masters in Management
in Management provides specific and tangible foundations for a successful career in business.
This 12-month, full-time programme is a business qualification with impact. In 2010, our MiM
employment rate was 95% within 3 months of graduation*; the majority of graduates choosing to
work in consulting or financial services.
As well as a renowned qualification from a world-class business school, you also gain access
to the Schools network of more than 34,000 global alumni a community that offers support and
opportunities throughout your career.
99
uN x D lim
T !1
ux dt I
uN y D lim
T !1
ZT
1
D
lim
uz dt
u
N
z
Lectures on computationalTfluid
!1 dynamics
T
0
uy dtI>
>
>
=
(10.2)
>
>
>
Basic denitions
of the statistical theory of
>
>
>
>
turbulence
;
u0x;y;z D 0
The root of the averaged square of fluctuations is called root mean square,
or r.m.s.. The quantity averaged twice is equal to quantity averaged once
u D u.
If the turbulence process is statistically unsteady (for instance r.m.s is changed
in time), the definition of the Reynolds averaging (10.2) is not applicable and
should be extended using the concept of ensemble averaging. Within the ensemble averaging the stochastic process
101 is repeated N times from the initial
state. The turbulent quantity u is measured at a certain time t N times.
The ensemble averaged quantity is then:
N
1 X
u.t / D lim
ui .t /
N !1 N
i D1
10.2
The turbulence is isotropic if r.m.s of all three velocity fluctuations are equal
02
02
u02
x D uy D uz
(10.3)
The turbulence parameters are invariant with respect to the rotation of the
reference system. The turbulence is homogeneous in some fluid volume if
all statistical parameters are the same for all points in this volume, i.e.
u02
E D u02
E C rE/. This equality can be written for all statistical
x;y;z .x/
x;y;z .x
moments. The turbulence parameters are invariant with respect to the translation of the reference system.
10.3
The product of two fluctuations is the correlation. The product of two fluctuations at two point separated by the distance rE is the correlation function:
Rij .x;
E j0 .xE C rE/
E rE/ D u0i .x/u
(10.4)
(10.5)
E 0i .xE C rE/
u0i .x/u
u02
E
i .x/
(10.6)
100
A sample of the autocorrealtion function coefficient
for scalar fluctuation f 0
102
Rij .x;
E rE/ D ui .x/u
E j .xE C rE/
(10.4)
turbulence
(10.5)
u0i .x/u
E 0i .xE C rE/
u02
E
i .x/
(10.6)
f .x;
E rE/ D
f 0 .x/f
E 0 .xE C rE/
f 02 .x/
E
(10.7)
at three different points A; B and C across the jet mixer is shown in Fig. 10.1.
In measurements presented in Fig. 10.1 the scalar f is the concentration of
the dye injected from the nozzle (see Fig. 10.1, low picture, right). The
change of the function has a certain physical meaning. Let us consider the
autocorrelation function with respect to the point C (blue line):
f .rC ; r/ D
f 0 .rC /f 0 .rC C r/
f 02 .rC /
(10.8)
where r is the radial coordinate across the pipe. The f .rC ; rA / is negative. It
means the increase of the quantity f at the point C (f 0 .rC / > 0) is followed
by the decrease of this quantity at the point A (f 0 .rA / < 0) . This is true in
statistical sense, i.e. the most probable consequence of the increase f .rC / is
the decrease of f .rA /.
The correlation function and autocorrelation function can be written not
only for spatial separation but also for separation in time. For example, the
autocorrelation
temporal
function of the ui fluctuation is
Download free eBooks
at bookboon.com
103
101
where r is the radial coordinate across the pipe. The f .rC ; rA / is negative. It
means the increase of the quantity f at the point C (f 0 .rC / > 0) is followed
denitions
statistical
theory of
by the decrease of this quantity at the point A (f 0 .rABasic
/ < 0)
. Thisofisthe
true
in
Lectures
on computational
fluidmost
dynamics
statistical
sense, i.e. the
probable consequence of the increase f .rC / is turbulence
the decrease of f .rA /.
The correlation function and autocorrelation function can be written not
only for spatial separation but also for separation in time. For example, the
autocorrelation temporal function of the ui fluctuation is
i i .x;
E / D
u0i103
.x;
E t/u0i .x;
E t C /
(10.9)
u02
E t/
i .x;
Z1
i i .x;
E xj /dxj
(10.10)
is the integral length. The integral lengths are estimations of the size of the
largest vortex in flow. A sample of the integral length of the scalar field f
along the jet mixer centerline (Fig. 10.1, right)
Lf .x/ D
ZD=2
(10.11)
f .r/dr
D=2
is shown in fig. 10.2, where f is the autocorrelation function across the jet
mixer, d is the nozzle diameter, D is the diameter of the closing pipe. Lf
is the estimation of the largest structure of the scalar field (in this case, the
size of the spot of colored liquid injected from the nozzle).
The integral of the temporal autocorrelation functions
Ti .x/
E D
Z1
i i .x;
E /d
(10.12)
10.3.1
.x/further
D u0tinformation
.x/u0t .x/ contact:
D ::: the autocorIn the isotropic turbulence u02 D u0l .x/u0lFor
T: +44
(0)1483
681681
relation function can be represented in the
form
[13]
E: pg-enquiries@surrey.ac.uk
www.surrey.ac.uk/downloads
104
Download free eBooks at bookboon.com
102
Lf .x/ D
Lectures on computational fluid dynamics
ZD=2
f .r/dr
D=2
is shown in fig. 10.2, where f is the autocorrelation function across the jet
mixer, d is the nozzle diameter, D is the diameter of the closing pipe. Lf
is the estimation of the largest structure of the scalar field (in this case, the
size of the spot of colored liquid injected from the nozzle).
The integral of the temporal autocorrelation functions
Ti .x/
E D
Z1
i i .x;
E /d
(10.12)
10.3.1
In the isotropic turbulence u02 D u0l .x/u0l .x/ D u0t .x/u0t .x/ D ::: the autocorrelation function can be represented in the form [13]
104
Figure 10.2: Distribution of the integral length of the scalar field along the
jet mixer centerline. [47].
103
Rij D u
f g
ri rj C gij
r2
(10.13)
where
f .r/ D
u0l 105
.x/u0l .x C r/
u0l .x/u0l .x/
(10.14)
u0x .x/u0x .x C r/
u0x .x/u0x .x/
(10.15)
uy0 .y/uy0 .y C r/
uy0 .y/uy0 .y/
(10.16)
The function f .r/ is the same in both cases (10.15) and (10.16). The autocorrelaton function g.r/ is calculated for transversal velocities along any
direction
u0 .x/u0t .x C r/
g.r/ D t
(10.17)
u0t .x/u0t .x/
For instance, the autocorrelation function of the ux fluctuation calculated in
y direction,
u0 .y/u0x .y C r/
;
(10.18)
g.r/ D x
u0x .y/u0x .y/
or the autocorrelation function of the uy fluctuation calculated in x direction:
g.r/ D
uy0 .x/uy0 .x C r/
uy0 .x/uy0 .x/
(10.19)
The function g.r/ is the same in both cases (10.18) and (10.19). The velocities
components used in the previous definitions are illustrated in Fig.
Download free eBooks at bookboon.com
10.4. ui is the velocity pulsation vector, ui l is its projection on the direction
104
connecting two points (i.e. longitudinal direction),
ui t is its projection on
the transversal direction. Products like ui t ui t are the correlations between
points 1 and 2.
uy0 .x/uy0 .x C r/
uy0 .x/uy0 .x/
The function g.r/ is the same in both cases (10.18) and (10.19). The velocities components used in the previous definitions are illustrated in Fig.
10.4. ui is the velocity pulsation vector, ui l is its projection on the direction
connecting two points (i.e. longitudinal direction), ui t is its projection on
the transversal direction. Products like ui t ui t are the correlations between
points 1 and 2.
106
The following relations are valid between g and f in the isotropic homogeneous turbulence (see [13]):
1 @f
gDf C r
(10.20)
2 @r
Typical form of f and g is shown in Fig. 10.5. The change of the sign of g
function is due to the continuity equation of the velocity field. The integral
length calculated using f is twice as large as that calculated using g.
10.3.2
Taylor microscale
Until Kolmogorov derived his estimations for vortices in 1941, it has been
thought that the minimum vortices arising in the turbulent flow have sizes
estimated by Taylor. Let us consider the parabola fitted to the autocorrelation function at point r D 0. The parabola intersects the horizontal axis
at
a certain
The
coordinate of this point is the scale introduced by
Download
free point.
eBooks at
bookboon.com
107
105
10.3.2
Taylor microscale
Until Kolmogorov derived his estimations for vortices in 1941, it has been
thought that the minimum vortices arising in the turbulent flow have sizes
estimated by Taylor. Let us consider the parabola fitted to the autocorrelation function at point r D 0. The parabola intersects the horizontal axis
at a certain point. The coordinate of this point is the scale introduced by
Taylor and called as the Taylor microscale. The Taylor microscale can be
calculated through the second derivative
107 of the autocorrelation function at
r D 0. The Taylor series of f in the vicinity of the point r D 0 is
1 @2 f
.0/r 2 C O.r 4 /
(10.21)
2
2 @r
The parabola fitted to the curve f .r/ at r D 0 intersects the horizontal axis
at the point:
f .r/ D 1 C
f D
2
@2 f
.0/
@r 2
(10.22)
2
@2 g
.0/
@r 2
(10.23)
Re D u0 =
(10.24)
@r 2
2
@2 g
.0/
@r 2
(10.24)
10.3.3
Z1
E t/e i kErE d kE
fO.k;
(10.25)
1
Z1
1
f .Er ; t/e i k rE d rE
(10.26)
The Fourier transformation can be also written for the correlation function:
Rij .Er / D
Z1
E
ij .k/e
i kErE
108
E D 1
ij .k/
8 3
E
d k;
1
Z1
1
Rij .Er /e i k rE d rE
(10.27)
Z1
1
Z1 Z1
ij .k1 ; k2 ; k3 /d k2 d k3
1 1
(10.28)
Proof of the formula (10.28) The inverse Fourier transform of the function
Rij .r1 ; 0; 0/:
1
ij .k1 ; 0; 0/ D
2
The inverse transformation reads:
Rij .r1 ; 0; 0/ D
The general definition is
Z1
(10.29)
1
Z1
(10.30)
1
Rij .r1 ; r2 ; r3 / D
Z1 Z1 Z1
1 1 1
107
ij .k1 ; k2 ; k3 /e i k rE d k1 d k2 d k3
(10.31)
1
Z1
Rij .r
0/ D
Lectures on computational
fluid
dynamics
1 ; 0;
1
Rij .r1 ; r2 ; r3 / D
Z1 Z1 Z1
ij .k1 ; k2 ; k3 /e i k rE d k1 d k2 d k3
(10.31)
1 1 1
Rij .r1 ; 0; 0/ D
Z1 Z1 Z1
1
ij .k1 ; k2 ; k3 /d k2 d k3 e i k1 r1 d k1
1 1
(10.32)
10.3.4
Z1 Z1
ij .k1 ; k2 ; k3 /d k2 d k3
(10.33)
1 1
109
According to the definition of the correlation
function
E j0 .xE C rE/
Rij .Er / D u0i .x/u
(10.34)
Z1
1
E kE D
i i .k/d
Z1
E kE
i i .k/d
(10.35)
The quantity
E.k/ D
E kE
i i .k/d
(10.36)
E
jkj
Z1
E.k/d k
(10.37)
10.4
10.4.1
Structure functions
Probability density function
10.3.4
E j0 .xE C rE/
Rij .Er / D u0i .x/u
turbulence
(10.34)
Z1
1
E kE D
i i .k/d
Z1
E kE
i i .k/d
(10.35)
The quantity
E.k/ D
E kE
i i .k/d
(10.36)
E
jkj
Z1
E.k/d k
(10.37)
10.4
10.4.1
Structure functions
Probability density function
10.4.2
Structure function
109
Figure 10.6: Kurtosis of the structure function for the concentration of the
scalar field obtained in the jet mixer. [47].
(10.38)
(10.39)
where is the mean value of the stochastic value, the turbulence is Gaussian.
In reality, the most of the turbulence parameters are not Gaussian. The
deviations from the Gaussian turbulence is characterized by the kurtosis
Kurt and skewness S k. The kurtosis
Kurt D
h.u2l u1l /4 i
.h.u2l u1l /2 i/2
(10.40)
is three for the Gaussian turbulence. Big values of the kurtosis means that
the p.d.f. distribution of the structure function S1 .l/ D hu2l u1l i is very
flat. The kurtosis is also often called as flatness. If kurtosis for small l 0
is big, it means that the field of the stochastic field is very intermittent. Big
differences are possible even if the separation between two points l is small.
The p.d.f. function has long tails in this case. A sample of kurtosis for the
scalar structure function S.x/ D f .x C r/ f .x/ is given in Fig.10.6.
The skewness
h.u2l u1l /3 i
111
(10.41)
Sk D
.h.u2l u1l /2 i/3=2
is zero for the Gaussian process. For the isotropic turbulence the skewness
of the derivative
l 3
i
h @u
@l
Sk D h
i
l 2 3=2
110
h @u
i
@l
(10.42)
l
is equal to 0:5. Physically it means that negative values of the derivative @u
@l
are more probable than positive ones. Please prove that the skewness (10.42)
h.u2l u1l / i
(10.41)
S k D h.u2l u1l2/33 i3=2
(10.41)
S k D .h.u
u1l
u1l/2/i/i3=2
h.u
2l2l
(10.41)
S k D .h.u2l u1l /2 i/3=2
u1lisotropic
/ i/
is zero for the Gaussian process..h.u
For
turbulence the skewness
2l the
is zero for the Gaussian process. For the isotropic turbulence
theofskewness
Basic denitions
the statistical theory of
of
the
derivative
is
zero
for
the
Gaussian
process.
For
the
isotropic
turbulence
the skewness turbulence
Lectures
on computational fluid dynamics
of the derivative
of the derivative
l 3
3 i
h @u
@u
@ll i
S k D h h @u
(10.42)
@ll 3i3=2
2 ii3=2
S k D h h@u@l
(10.42)
l
2
h
i
(10.42)
S k D hh @u
l
i
@l i 3=2
@u
@ll 2
h @l i
l
is equal to 0:5. Physically it means that
negative values of the derivative @u
@ll
is equal to 0:5. Physically it means that negative values of the derivative @u
@ll
@u
areequal
more probable
than positive
ones.that
Please
provevalues
that the
skewness
(10.42)
is
0:5. Physically
it means
negative
of the
derivative
@l
are moretoprobable
than positive
ones. Please
prove that the
skewness
(10.42)
is
the
skewness
of the
structure
functions
of the
firstthat
order
S1skewness
.l/ D hu2l(10.42)
u1l i
are
more
probable
than
positive
ones.
Please
prove
the
is the skewness of the structure functions of the first order S1 .l/ D hu2l u1l i
calculated
at l !
0. structure functions of the first order S1 .l/ D hu2l u1l i
is the skewness
of the
calculated
at l !
0.
calculated at l ! 0.
Exercise 1.
Exercise 1.
Calculate
the
averaged values of the time dependent signals
Exercise
1. Reynolds
Calculate the
Reynolds averaged values of the time dependent signals
Calculate the Reynolds averaged values of the time dependent signals
Solution:
Solution:
Solution:
u.t/ D cos2 t;
u.t/ D cos22 t;
cos t;t;
u.t/ D sin
u.t/ D sin
t;
1t; t;
u.t/ D sin
u.t/ D 1 t;
u.t/ D 1
0; t;
0;
u.t/ D
0;
t
t
ttt
t
1;
1;
> 1;
1:
> 1:
> 1:
1
u.t/ D 1 ;
u.t/ D 21 ;
u.t/ D 2 ;
u.t/ D 0;
2
u.t/ D 0;
u.t/
D
0;
u.t/ D 0:
u.t/ D 0:
u.t/ D 0:
STUDY. PLAY.
Exercise 2.
Exercise 2.
Find
the Reynolds
stresses for the isotropic turbulence.
Exercise
2.
Find
the Reynolds
stresses for the isotropic turbulence.
1
Find the Reynolds stresses for0
isotropic turbulence.
0the
2 r12 r13 1
r living
r13 1
0 a2good
The stuff you'll need to make
The stuff that makes life worth living
@
A
rr12
r23
22
12 r
13 A
@rr221
r
r
r22
r23
@rr21
A
32
21 r
22 r33
23
r31
31 r32 r33
r31 r32 r33
Solution:
1
0
Solution:
02 0 01
Solution:
0
A
@202 020 0001
@0 2 0A
@00 02 20A
0 0 2
0 0 2
112
112
112
NORWAY.
YOUR IDEAL STUDY DESTINATION.
WWW.STUDYINNORWAY.NO
FACEBOOK.COM/STUDYINNORWAY
111
1
u.t/ D ;
2
u.t/ D 0;
u.t/ D 0:
Exercise 2.
Find the Reynolds stresses for the isotropic turbulence.
0
1
2 r12 r13
@r21 r22 r23 A
r31 r32 r33
1
0
2 0 0
@0 2 0A
0 0 2
Solution:
Exercise 3.
Calculate
Exercise the
3. turbulent kinetic energy
112of the isotropic turbulence, if r33 D 1
Calculate the turbulent kinetic energy of the isotropic turbulence, if r33 D 1
Solution:
k D 3=2
Solution:
k D 3=2
Exercise 4.
Relation
longitudinal autocorrelation function
Exercisebetween
4.
Relation between longitudinal autocorrelation function
ul .x/ul .x C r/
f .r/ D
ul .x/u
.x C r/
u2ll.x/
f .r/ D
u2l .x/[13]
and energy density E.k/is given by formula
and energy density E.k/is given
by formula [13]
Z1
sin kr
2
cos kr/d k
(10.43)
f .r/ D 2 Z1 E.k/k 2 r 2 .
kr
u2l
sin
kr
cos kr/d k
(10.43)
f .r/ D 2 0 E.k/k 2 r 2 .
kr
ul
0
Energy density of the isotropic
decaying turbulence is described as E.k/ D
4
2
k
Expk
.
Calculate
Energy density of the isotropic decaying turbulence is described as E.k/ D
k 4 Expk 2 . Calculate
longitudinal autocorrelation function f ,
longitudinal autocorrelation function f ,
integral length and
integral length and
Taylor microscale
Taylor microscale
for the isotropic decaying turbulence.
for the isotropic decaying turbulence.
Solution:
Solution:
2
u2l
Z1
Z1
sin kr
2
cos kr/d k D
f .r/ D 2 Z1 E.k/k 2 r 2 .
kr
u2l
2 2 sin kr
cos kr/d k D
f .r/ D 2 0 E.k/k r .
kr
ul
u2l
0
0
(10.44)
sin kr
1p
cos kr/d k D
Expr 2 =4
kr
4
p
sin kr
1
Expk 2 k 2 r 2 .
cos112
kr/d k D
Expr 2 =4
kr
4
113
2 2
1 Expk k r
D free
.
Download
2 ZeBooks at bookboon.com
(10.44)
113
Solution:
Z1
sin kr
2
cos Basic
kr/ddenitions
kD
(10.44)
f .r/ D 2 E.k/k 2 r 2 .
of the
statistical theory of
kr
u
l
Lectures on computational fluid dynamics
turbulence
0
2
D 2
ul
Z1
Expk 2 k 2 r 2 .
1p
sin kr
cos kr/d k D
Expr 2 =4
kr
4
LD
Z1
113
f .r/dr D
4
D
4
1=4
Exercise 5.
Calculate the probability density function of the time dependent signal
time
signal
0
0
1
0.1
2
0.9
3
0.5
4
0.7
5
0.5
6
0.05
7
0.65
8
0.2
9
0.78
10
0.43
11
0.98
12
0.67
13
0.92
14
0.55
0-0.2
3/15
0.2-0.4
2/15
0.4-0.6
3/15
0.6-0.8
4/15
0.8-1.0
3/15
Exercise 6.
The velocity u was measured at N points across the pipe: uki , where i D 1; N
and k D 1; K is the time step number. Write a program to calculate the autocorrelation function of u with respect to point N=2.
Solution:
k
uk;0
i D ui ui
K
1 X k
ui D
ui
K
kD1
R.i; N=2/ D
K
P
kD1
k;0
uk;0
i uN=2
K
P
kD1
uk;02
N=2
114
Download free eBooks at bookboon.com
113
Chapter 11
Chapter 11
Chapter
11 theory K41
Kolmogorov
Kolmogorov theory K41
Kolmogorov theory K41
11.1
Physical background
11.1
background
One of thePhysical
most outstanding
results in turbulence theory was obtained by
11.1 Physical
Kolmogorov
(s. Fig. 11.1)background
in 1941. The Kolmogorov theory known as K41
is
based
on the
of local
isotropy
of the turbulent
motion
at small
One
of the
mosthypothesis
outstanding
results
in turbulence
theory was
obtained
by
One
of
the
most
outstanding
results
in
turbulence
theory
was
obtained
by
scales.
The (s.
physical
model
the Kolmogorov
Kolmogorov theory
theoryknown
is the as
vortex
Kolmogorov
Fig. 11.1)
in behind
1941. The
K41
Kolmogorov
(s. hypothesis
Fig.in11.1)
in
1941.
The
Kolmogorov
theory
knownatassmall
K41
cascado
Fig. of
11.2.
Big
vortices
scales
L motion
(corresponds
to
is
based illustrated
on the
local
isotropy
of with
the turbulent
is based
on
the
hypothesis
of local
isotropy
of break
the turbulent
motion
small
the
waveThe
numbers
=Lmodel
in the
Fourier
space)
up theory
to small
which
scales.
physical
behind
the
Kolmogorov
is ones,
theatvortex
scales.
The
physical
model
behind
the
Kolmogorov
theory
is thewith
vortex
in
turn split
into
even
smaller
andBig
so vortices
on up
towith
the smalles
the
cascado
illustrated
in Fig.
11.2.
scales
Lvortices
(corresponds
to
cascado
illustrated
in Fig.
11.2.Fourier
Big
vortices
with
scales
(corresponds
to
scale
. One
of the =L
most
important
vortex
break
up mechanisms
isones,
the vortex
the
wave
numbers
in
the
space)
break
up to Lsmall
which
theturn
wavesplit
numbers
=Lsmaller
in the and
Fourier
space)
break
up tofrom
small
which
reconnection
described
above.
The
energy
transferred
bigones,
vortices
to
in
into
even
so
on
upis to
the smalles
vortices
with
the
in turn
split
into
smaller
so The
on up
to the
smalles
vortices
with
the
small
without
theand
loss.vortex
massive
"istakes
place
scale
.ones
Onealmost
of
theeven
most
important
break
up dissipation
mechanisms
the
vortex
scale
. One
of
thereferred
mostabove.
important
break
upthe
mechanisms
the
vortex
at small
vortices
to asThe
thevortex
dissipative
or
Kolmogorov
vortices.
reconnection
described
energy
is transferred
from bigis vortices
to
reconnection
described
above.
The
energy
is these
transferred
from by
big
vortices
to
The real
vortices
are
similar
calculated
Isazawa
et
small
onesturbulent
almost
without
the
loss.
Thetomassive
dissipation
" takes
place
small
ones
almost
without
The
massive
dissipation
" takes
place
al.
(Fig.
11.3).
Vortices
are
displayed
at three
different
time instants.
The
at small
vortices
referred
tothe
as loss.
the dissipative
or
the
Kolmogorov
vortices.
at small
vortices
to are
as the
dissipative
or by
thefiltering
Kolmogorov
vortices.
upper
pictures
arereferred
obtained
from
the
lower
ones
outIsazawa
the
hight
The
real
turbulent
vortices
similar
to these
calculated
by
et
The(Fig.
real 11.3).
turbulent
are similar
these
calculated
IsazawaThe
et
frequencies.
As Vortices
seenvortices
bigare
vortices
are revealed
in
low
frequency
simulation.
al.
displayed
at to
three
different
time by
instants.
al. the
(Fig.
11.3). are
Vortices
are displayed
at
three
different
time
instants.
The
If
resolution
is obtained
increased,
more
more
small
scale
vortex
filaments
upper
pictures
from
theand
lower
ones
by filtering
out the
hight
upper
pictures
are
obtained
from
the
lower
ones
by
filtering
out
the
hight
appears
on
the
place
of
big
smooth
vortices.
Thus,
the
most
important
frequencies. As seen big vortices are revealed in low frequency simulation.
frequencies.
As seen
big vortices
arebreak
revealed
in
low scale
frequency
physical
processes
the vortex
up are:
If
the resolution
isduring
increased,
more
and
more
small
vortexsimulation.
filaments
If the resolution
is increased,
more and
more small
filaments
appears
on the place
of big smooth
vortices.
Thus,scale
the vortex
most important
appears processes
on the place
of big
smoothbreak
vortices.
Thus, the most important
physical
during
the vortex
up are:
Transfer
energy
fromthe
large
scales
to small
ones and
physical
processes
during
vortex
break
up are:
Dissipation
Transfer energy
from
largeinscales
small ones and
of the
energy
smallto
vortices.
Transfer energy from large scales to small ones and
Dissipation of the energy in small vortices.
Dissipation of the energy in small vortices.
115
115
115
114
Technical training on
WHAT you need, WHEN you need it
At IDC Technologies we can tailor our technical and engineering
training workshops to suit your needs. We have extensive
experience in training technical and engineering staff and
have trained people in organisations such as General
Motors, Shell, Siemens, BHP and Honeywell to name a few.
AUTOMATION &
Our onsite training is cost effective, convenient and completely
PROCESS CONTROL
customisable to the technical and engineering areas you want
covered. Our workshops are all comprehensive hands-on learning
MECHANICAL
experiences with ample time given to practical sessions and
ENGINEERING
demonstrations. We communicate well to ensure that workshop content
and timing match the knowledge, skills, and abilities of the participants.
INDUSTRIAL
We run onsite training all year round and hold the workshops on
DATA COMMS
your premises or a venue of your choice for your convenience.
ELECTRICAL
Figure 11.2: Illustration of the vortex cascado.
For a no obligation proposal, contact us today
POWER
116
115
116
11.2
Dissipation rate
Two parameters, which are of importance during the cascado process, are
the kinematic viscosity and the dissipation rate ". The energy dissipation
rate per unit mass of a turbulent fluid is given by
@u0
@uj0 2
X @u0i
"D
C
D 2sij sij
2 i;j @xj
@xi
@uj0
@xi
(11.1)
1 @ui
@uj
1 @ui
@uj
C
/ .
C
/
sij D Sij S ij D .
2 @xj
@xi
2 @xj
@xi
(11.2)
The dissipation " is a random function of the coordinates and time, which
Download free eBooks at bookboon.com
fluctuates
together with the field u.x; t/. In what follows we consider the
mean dissipation rate " designating it as 116
". The energy dissipated by the
small vortices is generated by large scale vortices. The energy production is
defined as
2
@u0
@uj0
@xi
i;j
@xj
@xi
1 @ui
Lectures on computational fluid dynamics
@uj
1 @ui
@uj
C
/ .
C
/
sij D Sij S ij D .
@xi
@xi
2 @xj
2 @xj
(11.2)
The dissipation " is a random function of the coordinates and time, which
fluctuates together with the field u.x; t/. In what follows we consider the
mean dissipation rate " designating it as ". The energy dissipated by the
small vortices is generated by large scale vortices. The energy production is
defined as
P D u0i uj0
@ui
@xj
(11.3)
(11.4)
The formula (11.4) is valid for very high Reynolds numbers for the turbulence
being in the equilibrium, i.e. the production of the turbulence is compensated
by its dissipation, i.e. P D ".
11.3
Kolmogorov hypotheses
The basis of the Kolmogorov theory are three hypotheses, whichpare supposed
p
to be valid for high Reynolds numbers Re t D vL
,
where
v
D
TKE
D
k
is the characteristic fluctuation velocity.
The Kolmogorov hypothesis of local isotropy reads:
At sufficiently high Reynolds number Re t , the small-scale turbulent motions
(l lE I ) are statistically isotropic
Here lE I is the lengthscale as the demarcation between the anisotropic large
eddies and the isotropic small eddies. Kolmogorov argued that all information about the geometry of the large eddies - determined by the mean flow
field and boundary conditions - is also lost. Directional information at small
scales is lost. With the other words, the direction of the vorticity vector !
of small turbulent vortices is uniformly distributed over the sphere. As a
consequence, the statistics of the small-scale motions are in a sense universal
- similar in every high-Reynolds-number turbulent flow (see [14]).
The Kolmogorov first similarity hypothesis reads:
In every turbulent flow at sufficiently high Reynolds number Re t , the statistics
of the small-scale motions (l < lE I ) have a universal form that is uniquely
determined by and ".
For this range of scales we can introduce characteristic size , characteristic
velocity u and characteristic time which depend only on two parameters
and ":
Download free eBooks at bookboon.com
D " ;
u
u D u "117
; D "
(11.5)
u D u "u ;
D "
(11.5)
(11.6)
D .="/1=2
u =u D .Ret /1=4 ;
(11.7)
(11.8)
=T D .Ret /1=2
ZZZVWXG\DWWXGHOIWQO
Very remarkable
is the first formula defining the ratio between the smallest
5DQNHGWKLQWKHZRUOG
and largest vortices in the flow. If L is, say one meter, and the fluctua 7+(67HFKQRORJ\UDQNLQJ
tion 1m=s, the turbulent Reynolds number in water is Re t D 106 . The
$OPRVW\HDUVRISUREOHPVROYLQJ
Kolmogorov scale is in this case 32000 as less as the flow macroscale L. Es H[SHULHQFH
timations of the Kolmogorov scale in the jet mixer with nozzle diameter of
([FHOOHQW6SRUWV &XOWXUHIDFLOLWLHV
d D 1cm and closing pipe of D D 5cm diameter is shown in Fig. 11.4.
&KHFNRXWZKDWDQGKRZZHWHDFKDW
The Kolmogorov
second similarity hypothesis reads:
ZZZRFZWXGHOIWQO
In every turbulent flow at sufficiently high Reynolds number, the statistics of
the motions of scale l in the range L l have a universal form that is
uniquely determined by ", independent of .
This range is called as the inertial subrange. Since the vortices of this range
are much larger than Kolmogorov vortices, we can assume that their Reynolds
numbers lul = are large and their motion is little affected by the viscosity.
The energy density depends on the wave number k and the dissipation rate
"
E.k/ D " k
(11.9)
119
118
u k1=2
"
L
some useful estimations can be derived from (11.6):
=L .Ret /3=4 ;
u =u D .Ret /1=4 ;
(11.7)
(11.8)
=T D .Ret /1=2
Very remarkable is the first formula defining the ratio between the smallest
and largest vortices in the flow. If L is, say one meter, and the fluctuation 1m=s, the turbulent Reynolds number in water is Re t D 106 . The
Kolmogorov scale is in this case 32000 as less as the flow macroscale L. Estimations of the Kolmogorov scale in the jet mixer with nozzle diameter of
d D 1cm and closing pipe of D D 5cm diameter is shown in Fig. 11.4.
The Kolmogorov second similarity hypothesis reads:
In every turbulent flow at sufficiently high Reynolds number, the statistics of
the motions of scale l in the range L l have a universal form that is
uniquely determined by ", independent of .
This range is called as the inertial subrange. Since the vortices of this range
are much larger than Kolmogorov vortices, we can assume that their Reynolds
numbers lul = are large and their motion is little affected by the viscosity.
The energy density depends on the wave number k and the dissipation rate
"
E.k/ D " k
(11.9)
119
Figure
of the
scale along
the centerline
the jet mixer
Figure11.4:
11.4:Distribution
Distribution
ofKolmogrov
the Kolmogorov
scale
along theofcenterline
of
and
jet [47].and
Thefree
dissipation
ratedissipation
is calculated
k model
and the
the
the free
jet mixer
jet. The
ratefrom
" isthecalculated
from
3
k " model estimation
and the experimental
estimatin
of Miller
Dimotakis
experimental
of the Miller and
Dimotakis
(1991) and
= 48(U
/d)((x x(1991)
)/d)4.
0
d
The
analysis of dimension leads to the Kolmogorov law
Download free eBooks at bookboon.com
k 5=3
E.k/ D "2=3119
(11.10)
(11.10)
11.4
Figure 11.5: Three typical scale ranges in the turbulent flow at high Reynolds
numbers.
120
Figure 11.5: Three typical scale ranges in the turbulent flow at high Reynolds
numbers.
Lectures on computational fluid dynamics
Figure 11.6: Three typical ranges of the energy density spectrum in the
turbulent flow at high Reynolds number. 1- energy containing range, 2inertial subrange, 3- dissipation range.
121
121
Figure 11.7: Experimental confirmation of the Kolmogorov law. The compensated energy spectrum for different flows. [14].
122
Figure 11.8: Experimental confirmation of the Kolmogorov law for the concentration fluctuations in the jet mixer. Measurements of the LTT Rostock.[47].
11.5
The energy spectrum Fig. 11.6 is used to classify three main methods of tur122 general strategy is the Direct
bulent flows modelling (Fig. 11.9). The most
Numerical Simulation (DNS). Within the DNS the whole spectrum of turbulent structures is modelled starting from the biggest vortices of the energy
containing range till the smallest dissipative Kolmogorov vortices. The Large
Figure 11.8: Experimental confirmation of the Kolmogorov law for the conKolmogorov theory K41
centration fluctuations in the jet mixer. Measurements of the LTT Rostock.
11.5
The energy spectrum Fig. 11.6 is used to classify three main methods of turbulent flows modelling (Fig. 11.9). The most general strategy is the Direct
Numerical Simulation (DNS). Within the DNS the whole spectrum of turbulent structures is modelled starting from the biggest vortices of the energy
containing range till the smallest dissipative Kolmogorov vortices. The Large
Eddy Simulation (LES) models the energy containing vortices and a fraction
of vortices corresponding to the inertial subrange. The effect of remaining vortices is considered using different approximation models. Since small
vortices are universal, the models are also supposed to be universal. The
Reynolds averaged Navier Stokes (RANS) models are dealing with the large
vortices corresponding to the energy containing range. The effect of other
vortices is taken by different semi- empiric models which are not universal.
11.6
123
(11.11)
(11.12)
Fig. 11.11 shows that the predictions of Kolmogorov and Obukhov deviate
from measurement data. The reason of the discrepancy is the physical phe124
Visit us at www.dtu.dk
124
(11.11)
(11.12)
Fig. 11.11 shows that the predictions of Kolmogorov and Obukhov deviate
from measurement data. The reason of the discrepancy is the physical phenomenon called the intermittency. The intermittency is caused by the presence of laminar spots in every turbulent
124 flows even at very high Reynolds
numbers.
(11.13)
Assumption 2:
.ln"a/2
2 2
l
P ."l / D ce
q D q.1 q/
2
a D lnN"
l2 D A C ln.L= l/
(11.14)
q
&q D C q.3 q/
< "2l > l
3
18
11.6.1
Exercises
Solution:
1
u.t/ D ;125
2
u.t/ D 0;
u.t/ D 0:
u.t/ D sin t;
Lectures on computational fluid dynamics 1 t; t 1;
u.t/ D
0; t > 1:
Solution:
1
u.t/ D ;
2
u.t/ D 0;
u.t/ D 0:
1
2 0 0
@0 2 0A
0 0 2
0
Solution:
Exercise 3. Calculate the turbulent kinetic energy of the isotropic turbulence, if r33 D 1
Solution:
k D 3=2
Exercise 4. Relation between longitudinal autocorrelation function
f .r/ D
ul .x/ul .x C r/
u2l .x/
Z1
E.k/k 2 r 2 .
sin kr
cos kr/d k
kr
(11.15)
2
D 2
ul
Z1
Expk 2 k 2 r 2 .
(11.16)
126
sin kr
1p
cos kr/d k D
Expr 2 =4
kr
4
Z1
Z1
E.k/k 2 r 2 .
sin kr
cos kr/d k D
kr
(11.16)
Expk 2 k 2 r 2 .
1p
sin kr
cos kr/d k D
Expr 2 =4
kr
4
Z1
LD
f .r/dr D
4
1=4
Exercise 5. Calculate the probability density function of the time dependent signal
time
signal
0
0
time
signal
8
0.2
1
0.1
2
0.9
9
0.78
3
0.5
10
0.43
4
0.7
5
0.5
6
0.05
11
0.98
12
0.67
7
0.65
13
0.92
14
0.55
0-0.2
3/15
0.2-0.4
2/15
0.4-0.6
3/15
0.6-0.8
4/15
0.8-1.0
3/15
Exercise 6. The velocity u was measured at N points across the pipe: uki ,
where i D 1; N and k D 1; K is the time step number. Write a program to
calculate the autocorrelation function of u with respect to point N=2.
Solution:
127k
uk;0
i D ui ui
K
1 X k
ui D
ui
K
kD1
R.i; N=2/ D
K
P
kD1
k;0
uk;0
i uN=2
K
P
kD1
uk;02
N=2
127
Our broad range of Open Enrollment Executive Programs offers you a unique interactive, stimulating and
multicultural learning experience.
Be prepared for tomorrows management challenges and apply today.
For more information, visit www.msm.nl or contact us at +31 43 38 70 808 or via admissions@msm.nl
18-08-11 15:13
128
fN 0 D 0
(12.3)
(12.4)
(12.5)
@fN
@f
D
@x
@x
(12.6)
(12.7)
(12.8)
The starting point of the derivation of the RANS equation is the original
Navier Stokes (NS) equation:
@ui
@ui
1 @j i
C uj
D Fi C
@t
@xj
@xj
(12.9)
@ui
@ui
@ui
@ui
D u1
C u2
C u3
@xj
@x1
@x2
@x3
(12.10)
The NS equation is supplied with the continuity equation, which for the case
of incompressible flow takes the form:
@ui
@u1
@u2
@u3
D
C
C
D0
@xi
@x1
@x2
@x3
(12.11)
Using the continuity equation the convective term is written in the conservative form:
uj
@ui
@ui
@uj
@.ui uj /
D uj
C ui
D
@xj
@xj
@xj
@xj
(12.12)
(12.13)
(12.14)
uj
@xj
D uj
@xj
C ui
@xj
@xj
(12.12)
@xj
D Fi C
@xj
(12.13)
(12.14)
(12.15)
(12.15)
@.uN i uNj C u0 u0 /
1 @N
@uN
@uN ii C @.uN i uNj C u0ii ujj0 / D FNNi C 1 @Njj ii
(12.16)
D Fi C @xj
(12.16)
@t C
@x
@xjj
@t
@xj
Writing the term u0 u0 on the r.h.s we obtain the Reynolds averaged Navier
Writing the term u0ii ujj0 on the r.h.s we obtain the Reynolds averaged Navier
Stokes equation (RANS). Its unsteady version is called as the unsteady
Stokes equation (RANS). Its unsteady version is called as the unsteady
Reynolds averaged NS equation (URANS):
Reynolds averaged NS equation (URANS):
@uN i
@uN i uNj
@
@uN i C @uN i uNj D FNNi C @ .Nj i %u00i uj00 /
(12.17)
@t C @xj D Fi C @xj .Nj i %ui uj /
(12.17)
@t
@xj
@xj
There are two important features of URANS in comparison with NS:
There are two important features of URANS in comparison with NS:
The URANS is written for the averaged quantities, whereas the NS for
The URANS is written for the averaged quantities, whereas the NS for
instantaneous ones,
instantaneous ones,
The URANS has additional term on the r.h.s. %u0 u0 which is called
The URANS has additional term on the r.h.s. %u0ii ujj0 which is called
the Reynolds stress R .
the Reynolds stress Rijij .
Generally the Reynolds stress is the matrix with nine terms:
Generally the Reynolds stress is the matrix with nine terms:
131
131
1
2
@uN j
@xi
@uN i
@xj
pothesis with the only difference that instead of the kinematic viscosity the
turbulent viscosity t is used
@uN i
@uNj
2
0 0
C
(12.20)
ui uj D t
ij k
3
@xi
@xj
In the simplest form for flow along the plate with ux .y/ and uy D uz D 0
the formula (12.20) reads
u0x uy0 D t
d uN x
dy
(12.21)
The last term in (12.20) is introduced to keep the consistency. Indeed the
sum of three diagonal terms of the Reynolds matrix is equal to the turbulent
kinetic energy k D 12 u0i u0i . Without this term the sum of r.h.s of (12.20)
would result in the sum of the diagonal terms of the strain rate matrix
S11 C S22 C S33 which is zero due to the continuity equation. It would
be wrongEXPERIENCE
result because THE
k 0.POWER
The turbulent
OF closures (12.20) are referred
to as the isotropic because the coefficient t is equal for all matrix elements
FULL ENGAGEMENT
Rij .
While the kinematic viscosity depends on the liquid, the turbulent kinematic
viscosity depends on the turbulent state of the flow. According to estimation
of Landau the ratio of the turbulent kinematic viscosity to the kinematic one
RUN
FASTER.
is proportional
to the
ratio of the Reynolds number to that corresponding to
the transition for
thisLONGER..
type of flow
RUN
READ MORE & PRE-ORDER TODAY
RUN EASIER
WWW.GAITEYE.COM
t = Re=Recri t
(12.22)
The URANS closure models are subdivided into algebraic and differential
1349906_A6_4+0.indd
1
ones.
The
most
prominent
model amount the algebraic models is the Prandtl 22-08-2014
Download
free
eBooks
at bookboon.com
132
131
12:56:57
1
2
@uN j
@xi
@uN i
@xj
pothesis
the onlyfluid
difference
the kinematic
the
Lectures
onwith
computational
dynamicsthat instead of
Reynolds
Averaged viscosity
Navier Stokes
Equation (RANS)
turbulent viscosity t is used
u0i uj0
D t
@uN i
@uNj
C
@xi
@xj
2
ij k
3
(12.20)
In the simplest form for flow along the plate with ux .y/ and uy D uz D 0
the formula (12.20) reads
u0x uy0 D t
d uN x
dy
(12.21)
The last term in (12.20) is introduced to keep the consistency. Indeed the
sum of three diagonal terms of the Reynolds matrix is equal to the turbulent
kinetic energy k D 12 u0i u0i . Without this term the sum of r.h.s of (12.20)
would result in the sum of the diagonal terms of the strain rate matrix
S11 C S22 C S33 which is zero due to the continuity equation. It would
be wrong result because k 0. The turbulent closures (12.20) are referred
to as the isotropic because the coefficient t is equal for all matrix elements
Rij .
While the kinematic viscosity depends on the liquid, the turbulent kinematic
viscosity depends on the turbulent state of the flow. According to estimation
of Landau the ratio of the turbulent kinematic viscosity to the kinematic one
is proportional to the ratio of the Reynolds number to that corresponding to
the transition for this type of flow
t = Re=Recri t
(12.22)
The URANS closure models are subdivided into algebraic and differential
ones. The most prominent model amount the algebraic models is the Prandtl
model described above. The disadvantages of the algebraic models are:
132
they are good only for the simplest flow,
not suitable for 3D flows,
not suitable for separation flows,
turbulent viscosity depends on averaged values of velocities,
do not consider the flow history.
These disadvantages can be overcome using differential models which are
subdividided into one, two and multi equation models. One and two equation models are usually isotropic based on the Boussinesq approach (12.20).
Among the one equation models the most modern and efficient is the model
of Spalart Allmares (SA model) written for the modified kinematic turbulent
viscosity . The equation for reads
2
@
@
1 @
@
Cb2 d d
C uNj
D Cb1 Cw1 fw
C
. C /
C
@t
@xj
d
@xk
@xk
dxk dxk
(12.23)
Download free eBooks at bookboon.com
where
132
C1 D 7; 1;
D 2=3;
(12.24)
f1
3
D 3
;
3
C C1
D ;
f2
g D r C Cw2 .r r/;
D S C 2 2 f2 ;
k d
6
1 C Cw3
fw D g 6
6
g C Cw3
rD 2 2
k d
D
;
1 C f1
p
S D 2ij ij ;
1=6
(12.24)
1 @uN i @uNj
ij D
2 @xj
@xi
(12.25)
(12.26)
Within the more advanced k " model the turbulent kinetic energy and the
dissipation rate are calculated from the transport equations:
9
@k
@k
@uN i
@
t @k
>
>
133 C ij
>
C uNj
D
C
"
=
@t
@xj
@xj
k @xj
@xj
(12.27)
@"
t @"
C"1 " @uN i C"2 "2 >
@"
@
>
>
C
C
C uNj
ij
D
;
@t
@xj
@xj
" @xj
k
@xj
k
If k and " are known the turbulent viscosity t can be found from the dimension analysis, applied to the dissipation rate
" k 3=2 =L
(12.28)
p
t D C kL
(12.29)
k2
(12.30)
t D C kL D C
"
If t is known, the Reynolds stresses can be determined from the Boussinesq
approach (12.20).
133
13.1
13.1
of the RSM Equations
13.1.1 Derivation
Step 1
13.1 Derivation of the RSM Equations
The
k t h Step
Navier-Stokes
equation
13.1.1
1
13.1.1
Step 1
@
1 @p
1 @
@uk
The k t h Navier-Stokes
equation
C
.uj uk / D
C
j k
The k t h Navier-Stokes
equation
@t
@x
@xk
@xj
j
@
1 @p
1 @
@uk
C @ .ucomponent
C 1 @ j k
j uk / D 1 u@p
is multiplied with @u
thek velocity
i
@t C @xj .u u / D @xk C @xj
j k
jk
@t
@xj
@xk
@xj
@uthe
1
@
@p
@
1
is multiplied with
velocity
component
u
k
i
C
ui
.uj uk / D ui
C
j k
is multiplied with
component
@xj
@xj
@tthe velocity
ui @xk
@uk
1 @p
@
1 @
ui @u C
.uj uk /
D uki th1velocity
C 1component:
j k
The i t h equation
is
multiplied
with
@p
@
@
k
@x
@t
@x
@x
k C
j
C j .uj uk / D ui
ui
jk
@t
@xk
@xj
@xj
@u
@
@p
@
1
1
The i t h equationi is multiplied with k th velocity component:
uk
C
.uj ui / D uk
C
j k
The i t h equation
thvelocity
@xj
@xi
component:
@xj
@t is multiplied
with k
@ui
@
uk 1 @p C 1 @ j k
uk @u (13.2)
C @ and
.uj u(13.3)
i / D are
Resulting equations
then1 sumed:
@p
1 @
i
uk @t C @xj .uj ui / D uk @xi C @xj j k
@xi
@t
@xj
@xj
Resulting equations (13.2) and (13.3) are then sumed:
@.u(13.2)
@p are then
@p sumed:
@j k
@j i
@.ui uequations
k/
i uk uj /and (13.3)
Resulting
D ui
uk
C ui
C uk
C
@t
@xj
@xk
@xi
@xj
@xj
@.ui uk / @.ui uk uj /
@p
@p
@j k
@j i
C @.u u u / D ui @p uk @p C ui @ C uk @
@.u
i jk jdecomposition
jk
ji
Substitution
@ti uk / of
@x
D ui @xk uk @xi C ui @xj C uk @xj
C Reynolds
@t
@xj
@xk
@xi
@xj
@xj
0
0
0
Substitution of u
Reynolds
decomposition
N i C ui ; p D pN C p ; j i D Nj i C j i
i D u
Substitution of Reynolds decomposition
into the equation
inD pN C p 0 ; D N C 0
D uN i Cresults
u0i ; p
ui (13.4)
ji
ji
ji
0
ui D uN i C ui ; p D pN C p 0 ; j i D Nj i C j0 i
into the equation (13.4) results in 135
into the equation (13.4) results in
135
135
(13.1)
(13.1)
(13.1)
(13.2)
(13.2)
(13.2)
(13.3)
(13.3)
(13.3)
(13.4)
(13.4)
(13.4)
(13.5)
(13.5)
(13.5)
DISCOVER
OUR WORLD
Some people know precisely where they want to go. Others seek the adventure of
discovering uncharted territory. Whatever you want your professional journey to be,
youll nd what youre looking for at Oliver Wyman.
134
@t
@xk
@xj
@xj
@u
1
@p
@
1
@
i
Lectures on computational
fluid dynamics
Stress Model (RSM)
uk
C
.uj ui / D uk
C
j k Reynolds(13.3)
@xi
@t
@xj
@xj
Resulting equations (13.2) and (13.3) are then sumed:
@p
@p
@j k
@j i
@.ui uk / @.ui uk uj /
D ui
uk
C ui
C uk
C
@t
@xj
@xk
@xi
@xj
@xj
(13.4)
p D pN C p 0 ;
j i D Nj i C j0 i
(13.5)
@
@
@
.uN i uN k / C .u0i u0k / C
uN i uNj uN k C uN i uj0 u0k C
135
@t
@t
@xj
@pN
@p 0
CuNj u0i u0k C uN k u0i uj0 C u0i uj0 u0k D uN i
u0i
@xk
@xk
@j0 k
@j0 i
@pN
@p 0
@j k
@j i
uN k
u0k
C uN i
C u0i
C uN k
C u0k
@xi
@xi
@xj
@xj
@xj
@xj
13.1.2
(13.6)
Step 2
@pN
@
@
@uN k
0 0
C
.uNj uN k / D uN i
C
.Nj k uj uk /
@t
@xj
@xk
@xj
(13.7)
(13.8)
Again the i th Reynolds averaged Navier Stokes equation is multiplied with
the k t h component of averaged velocity
uN k
@
@
@uN i
@pN
0 0
C
.uNj uN i / D uN k
C
.Nj k uj ui /
@t
@xj
@xi
@xj
(13.9)
13.1.3
Step 3
@j0 i
@uN135
@uN k
i
C u0k
uj0 u0k
uj0 u0i
C u0i
@xj
@xj
@xj
@xj
136
(13.11)
13.1.3
Step 3
@j0 i
@uN i
@uN k
C u0k
uj0 u0k
uj0 u0i
C u0i
@xj
@xj
@xj
@xj
(13.11)
u0i
0
@u0k
@p 0
@p 0
@
@
@ui
0
0 0
0 0
0
uk
Dp
C
.u p / C ij
.u p /
j k
@xk
@xi
@xk
@xi
@xj i
@xj k
(13.12)
@j0 k
@xj
Cu0k
@2 u0k
@2 u0i
@u0i @u0k
@2 0 0
0
0
D ui
Cuk 2 D 2 ui uk 2
(13.13)
@xj
@xj @xj
@xj2
@xj
@xj
@j0 i
@u0i @u0k
@2 0 0
u
u
2
@xj @xj
@xj2 i k
@uN i
@uN k
uj0 u0i
@xj
@xj
which can be written in a compact form
uj0 u0k
@
@ 0 0
@
.ui uk / C uNj
.u0i u0k / D
Di k C Ri k C Pi k "i k
@t
@xj
@xj
(13.14)
Di k D
Ri k
@.u0i u0k /
1
u0i uj0 u0k C .j k u0i C ij u0k /p 0
@xj
@u0k
1 @u0i
p0
D
C
@xk
@xi
Pi k D uj0 u0k
Diffusion (13.15)
"i k D 2
@u0i @u0k
@xj @xj
137
! Generation
!136
Dissipation
(13.17)
(13.18)
Ri k
@u0k
1 @u0i
p0
D
C
@xi
@xk
@uN i
Lectures on computational fluid0 dynamics
0
Pi k D uj uk
@xj
"i k D 2
13.1.4
uj0 u0i
@uN k
@xj
@u0i @u0k
@xj @xj
! Generation
! Dissipation
(13.17)
(13.18)
@.u0i u0k /
1
D
u0i uj0 u0k C .j k u0i C ij u0k /p 0
@xj
(13.19)
is due to
molecular diffusion, described by the term:
@.u0i u0k /
@xj
(13.20)
(13.21)
@u0k
1 @u0i
p0
C
D
@xi
@xk
(13.23)
describes the redistribution of the energy between different tensor components u0i u0k caused by correlation between the stresses and pressure fluctuations.
138
Download free eBooks at bookboon.com
137
@xj
(13.20)
(13.21)
@u0k
1 @u0i
p0
C
D
@xk
@xi
(13.23)
describes the redistribution of the energy between different tensor components u0i u0k caused by correlation between the stresses and pressure fluctuations.
The term
The term
138
@uN i
uj00 u0k0 @uN i
uj uk @xj
@xj
@uN k
uj00 u0i0 @uN k
(13.24)
Pi k D
uj ui @xj
(13.24)
Pi k D
@xj
is responsible for the energy generation, i.e. the transport of the energy
is
responsible
for the energy
i.e. theflow
transport
of the energy
transfer
from averaged
(mean)generation,
flow to oscillating
(fluctuations).
And,
transfer
from
averaged
(mean)
flow
to
oscillating
flow
(fluctuations).
And,
finally,
finally,
@u0 @u0
"i k D 2 @ui0i @u0kk
(13.25)
"i k D 2 @xj @xj
(13.25)
@xj @xj
is the dissipation. This unclosed term is responsible for the transformation
is
This energy
unclosed
term
responsible
forthe
theflow.
transformation
of the
the dissipation.
turbulent kinetic
into
theisinner
energy of
of
the
turbulent
kinetic
energy
into
the
inner
energy
of
the
flow.
RSM model based on equations (13.14) is used to determine the Reynolds
RSM
model
on equations
(13.14) It
is used
determine
theBoussinesq
Reynolds
stresses
frombased
the transport
equations.
is nottobased
on the
stresses
from
the
transport
equations.
It
is
not
based
on
the
Boussinesq
approach and takes the anisotropy of stresses into account. This model is
approach
andamong
takes RANS
the anisotropy
the best one
models. of stresses into account. This model is
the best one among RANS models.
138
Chapter 14
Equations of the k Model
Chapter
14
Chapter
Equations14of the k - " Model
Equations
Equations of
of the
the k
k -- "" Model
Model
14.1
According
to definition
14.1 Derivation
14.1
of the k-Equation
Derivation of the k-Equation
According to definition
k D u0k u0k =2
(14.1)
According to definition
Assuming i D k in the Reynolds stress0 model
equations (13.14)
k D uk u0k =2
(14.1)
0 0
k D uk uk =2
(14.1)
@
@
Assuming@ i.uD0 uk0 /inCthe
Reynolds
0 0 stress model equations (13.14)
uNj
.ui uk / D
Di k C Ri k C Pi k "i k
(14.2)
Assuming
Reynolds
stress
@t i Di kk in the@x
@xj model equations (13.14)
j
@ 0 0
@
@
0 0
.ui uequations
/ C uNj @for.uk=1,2
uk / Dand
Ri k Cthe
Pi ktransport
"i k
(14.2)
ans summing
obtain
equations
ik C
@
@ 3Dwe
i
k
@t .u0 u0 / C uNj @xj .u0 u0 / D @x
j Di k C Ri k C Pi k "i k
(14.2)
for the total
energy
i k
i k
@t kinetic
@xj k:
@xj
ans summing equations for k=1,2 and 3 we obtain the transport equations
@kfor k:
@k and @3 we obtain the transport equations
ans the
summing
equations
k=1,2
for
total kinetic
energy
C uNj
D
Ds C P "S
(14.3)
for the total kinetic energy
@t k: @xj
@xj
@k
@k
@
C uNj @k D @ Ds C P "S
(14.3)
where
@k
@t C uNj @xj D @xj Ds C P "S
(14.3)
@t
@xj
@xj
where
whereD D @k C 1 u0 p 0 u0 k 0 ; k 0 D u0 u0 =2
Diffusion
(14.4)
S
jk k
k k
j
@xj
@k
1
Diffusion
(14.4)
DS D @k C 1 j k u0k p 0 uj0 k 0 ; k 0 D u0k u0k =2
Diffusion
(14.4)
DS D @xj C j k u0k p 0 0 uj00 k@0u;N k k 0 D u0k u0k =2
Generation
(14.5)
@xj
P D uj uk
@xj
@uN k
0 0
Generation
(14.5)
0P D
0 uj uk @u
@ukP@uDk u0 u0 @xN jk
Generation
(14.5)
j Dissipation
k
"S D
( Pseudodissipation)
(14.6)
@xj
@xj @xj
@u0 @u0k
D @uk0the
Dissipation
( Pseudodissipation)
(14.6)
The relation"Sbetween
pseudodissipation
is
@u0true and
"S D @xjk @xjk
Dissipation ( Pseudodissipation)
(14.6)
@x
j @xj0
0 2
The relation between
and
@u true @u
the
@
@ is
141
j pseudodissipation
" D the ktrue
(14.7)
C and pseudodissipation
"S C
isuj0 u0k
The relation between
2 @xj
@xk
@xj @xk
141
For large Reynolds numbers the true141
dissipation and the pseudodissipation
are equal.
" "S
(14.8)
(14.9)
where
Download free eBooks at bookboon.com
Re t D
p
kL=
(14.10)
139
" "S
(14.8)
(14.9)
Equations
of the k Model
where
Re t D
p
kL=
(14.10)
@k
1
C j k u0k p 0 uj0 k 0
@xj
(14.11)
(14.12)
(14.13)
D t
@uN i
@uNj
C
@xi
@xj
2
ij k
3
(14.14)
89,000 km
142
careers.slb.com
Based on Fortune 500 ranking 2011. Copyright 2015 Schlumberger. All rights reserved.
140
"s D CD k
(14.13)
=L
@uN i
@uNj
0 dynamics
0
Lectures on computational
fluid
u
C
i uj D t
@xi
@xj
2
ij k
3
(14.14)
14.2
@uNj @uN i @uNj
k 3=2
t @k
(14.15)
C
C t
C
CD
k @xj
@xi @xj @xi
L
142
(14.16)
@u0
(14.17)
@
@ui @u0i
@
@uN i @u0i
@
@uN i @u0i
0
D
C
C
uj
uNj
u
@xk
@xj @xk
@xk
@xj @xk
@xk j @xj @xk
@u0i @u0i
@u0i @u0i
@
@
0
C
uNj
u
C
@xk
@xj @xk
@xk j @xj @xk
(14.18)
(14.19)
@u0i @u0i
@2 u0i @u0i
@
@uNj @u0i @u0i
uNj
D
C uNj
D
@xk
@xj @xk
@xk @xj @xk
@xj @xk @xk
D
(14.20)
@uj0 @u0i @u0i
@u0i @u0i
@2 u0i @u0i
@
0
D
C uj0
D
uj
@xk
@xj @xk
@xk @xj @xk
@xj @xk @xk
D
@u0i
1 @ 0
D
C
u "s
@xk @xj @xk
2 @xj j
141
i
143
(14.21)
(14.20)
@uj0 @u0i @u0i
@u0i @u0i
@2 u0i @u0i
@
0
uj
D
C uj0
D
@xk
@xj @xk
@xk @xj @xk
@xj @xk @xk
(14.21)
143
@
@uN i @u0i
@uN i @uj0 @u0i
@2 uN i 0 @u0i
0
u
uj
D
C
@xk
@xj @xk
@xj @xk @xk
@xk @xj j @xk
1 @p @u0i
1 @p 0 @u0i
D
% @xi @xk
% @xi @xk
0
0
1 @
@
1
@
@u
@
@ui
i
D
j0 i
jN i C j0 i
% @xk @xj
@xk
% @xk @xj
@xk
0
@uj0
@ 0
@ui
@2 u0i
@
D
j i D
C
@xj
@xj
@xj
@xi
@xj @xj
(14.22)
(14.23)
(14.24)
2 0 0
0 0
1 @
@ 0 @u0i
@ ui
@ui
@ui @ui
@
@
j i
D
D
D
% @xk @xj
@xk
@xk @xj @xj @xk
@xj @xj @xk @xk
@2 u0i @2 u0i
@ 1 @ @u0i @u0i
D
@xj 2 @xj @xk @xk
@xj @xk @xj @xk
(14.25)
This gives:
@"s
@
@"s
C uNj
D
D" C P" ""
@t
@xj
@xj
(14.26)
where
D" D
P" D
2uj0
@"s
@u0i @p 0
uj0 "0s 2
@xj
@xk @xi
0 0
@u0i @2 uN i
@ui @uj @uN i
@u0i @u0i @uN i
2
C
@xk @xj @xk
@xk @xk @xj
@xj @xk @xk
@u0i @uj0 @u0i
2
@xj @xk @xk
"" D 2 2
@2 u0i @2 u0i
;
@xj @xk @xj @xk
"0s D
@u0i @u0i
@xk @xk
(14.27)
(14.28)
(14.29)
The terms on the r.h.s. were approximated according to the following forDownload free eBooks at bookboon.com
mula:
142
144
@
@
D" D
@xj
@xj
T
C
"
@"
C"1 " @uN i
C"2 "2
I P" D
ij
I "" D
@xj
k
@xj
k
(14.30)
C"2 D 1:92;
k D 1;
" D 1:3
(14.31)
(14.32)
Under assumption that the generation of the turbulent energy equals to the
its dissipation (the turbulence is in equilibrium, turbulent scales are in the
inertial range) Kolmogorov and Prandtl derived the relation between the
kinetic energy, the dissipation rate and the integral lengths L:
"
k 3=2
L
(14.33)
D Cnew
kLthings
Could you think of t101
to do with eggs and oil?
follows
p
k2
t D C kL D C
"
(14.34)
(14.35)
As soon as k and " are known the turbulent kinematic viscosity t is computed from (14.35) and Reynolds stresses can be calculated from the Boussinesq hypothesis and then substituted into the Reynolds averaged Navier
Stokes equations. The problem is mathematically closed.
The k " model is the classical approach, which is very accurate at large Re
numbers. At small Re number, for instance close to the wall, the approximations used in derivation of k " model equations are not valid. To overcome
this disadvantage
various low Reynolds k " models were proposed.
Hellmanns is one of Unilevers oldest brands having been popular for over 100 years.
If you too share a passion for discovery and innovation we will give you the tools and
opportunities to provide you with a challenging career. Are you a great scientist who
would like to be at the forefront of scientic innovations and developments? Then you will
enjoy a career within Unilever Research & Development. For challenging job opportunities,
please visit www.unilever.com/rdjobs.
145
143
inertial range) Kolmogorov and Prandtl derived the relation between the
kinetic energy, the dissipation rate and the integral lengths L:
Lectures on computational fluid dynamics
"
k 3=2
L
(14.33)
(14.34)
follows
t D C
k2
kL D C
"
(14.35)
As soon as k and " are known the turbulent kinematic viscosity t is computed from (14.35) and Reynolds stresses can be calculated from the Boussinesq hypothesis and then substituted into the Reynolds averaged Navier
Stokes equations. The problem is mathematically closed.
The k " model is the classical approach, which is very accurate at large Re
numbers. At small Re number, for instance close to the wall, the approximations used in derivation of k " model equations are not valid. To overcome
this disadvantage various low Reynolds k " models were proposed.
14.3
@uN k
D"
@xj
@uN x
@y
u0x uy0
@uN x 2
@uN x
D t
D"
@y
@y
(14.36)
t D l
@uN x
@y
@ux
@y
D t = l 2
t D c k 1=2 L
144
(14.37)
(14.38)
c k 2
(14.39)
The derivative
@uN x
@y
Prandtl:
Nx
Lectures on computational fluid dynamics
2 @u
t D l
@y
@ux
@y
D t = l 2
t D c k 1=2 L
(14.37)
(14.38)
c k 2
"
(14.39)
"
t
1=2
"2
c k 2
1=2
"
c1=2 k
@uN x
"
D % t 1=2 D
@y
c k
c k 2 "
D %c1=2 k
" c1=2 k
w
%c1=2
Let yp be the ordinate of the first grid node near the wall.
Using approximation of Prandtl for l we obtain the dissipation rate at yp :
"p D c3=4 kp3=2 = yp
(14.40)
kp D
w
%c1=2
w can be found from the assumption that the first node is in the logarithmic
range:
ux
1
D ln Ey C
u
x
x uN xp
u D
ln EypC
Download free eBooks at bookboon.com
r
r
w
w
x uN xp
145 k 1=2 c 1=4
D
p
%
%
ln EypC
%x uN xp kp1=2 c1=4
w D
ln EypC
(14.41)
"p D c3=4 kp3=2 = yp
(14.40)
w
%c1=2
w can be found from the assumption that the first node is in the logarithmic
range:
ux
1
D ln Ey C
u
x
x uN xp
u D
ln EypC
r
r
w
w
x uN xp
D
kp1=2 c1=4
C
%
%
ln Eyp
(14.41)
%x uN xp kp1=2 c1=4
w D
ln EypC
Values "p (14.40) and w (14.41) are used within finite volume method when
the k " equations are written for the volumes adjacent to the wall. For
instance the generation term is calculated as
Z
P dxdy D
360
thinking
Z
147
@u
@u
ij dxdy D w dxdy D w xuxp
@y
@y
360
thinking
(14.42)
360
thinking
Dis
15.1
LES filtering
'QQ ';
Q
147
e
@'
@'Q
D
@xj
@xj
e e e
Q 0 C ' 0 0
Q 0 C '
'f
D 'f
Q Q C '
In the case of Reynolds averaging only the first and the last terms remain.
The properties of large and small scale motions are shown in the table 15.1.
Large scale motion
Generated by mean flow
Download free eBooks at bookboon.com
Depends on the flow geometry
Regular
Deterministic description
Heterogeneous
Anisotrop
e e e
e
e e
e e
e
In
case of Reynolds
averaging
only motions
the first are
andshown
the last
terms
remain.
Thethe
properties
of large and
small scale
in the
table
15.1.
In
case
of Reynolds
averaging
only motions
the first are
andshown
the last
terms
remain.
Lectures
on
computational
fluid
dynamics
Large
Eddy
Simulation (LES)
Thethe
properties
of large
and
small scale
in the
table
15.1.
TheLarge
properties
of motion
large and small scale
motions
aremotion
shown in the table 15.1.
scale
Small
scale
Large
scale
motion
Small
scale
motion
Generated
by motion
mean flow
Generated
bymotion
large scale structures
Large
scale
Small
scale
Generated
mean
Depends onbythe
flowflow
geometry Generated
Universal by large scale structures
Generated
by
mean
flow
Generated
by large scale structures
Depends
Regular on the flow geometry Universal
Stochastic
Depends
on the flow geometry Universal
Regular
Stochastic description
Deterministic description
Regular
Stochastic
Deterministic
Stochastic
description
Heterogeneousdescription
Homogeneous
Deterministic
description
Stochastic
description
Heterogeneous
Homogeneous
Anisotrop
Isotrop
Heterogeneous
Homogeneous
Anisotrop
Isotrop
Exists long time
Exists short time
Anisotrop
Isotrop
Exists
long time
Exists
short time
Diffusive
Dissipative
Exists
long
time
Exists
short
time
Diffusive
Dissipative
Modelling is complicated
Easy to model
Diffusive
Dissipative
Modelling is complicated
Easy to model
Modelling
is 15.1:
complicated
Easy and
to model
Table
Properties of large
small scale motions
Table 15.1: Properties of large and small scale motions
Table 15.1: Properties of large and small scale motions
A very important conclusion from this table is the fact that the small scale
A
very important
thisexpect
table isthat
thethe
factmodels
that the
small scale
motion
is universal.conclusion
Thereforefrom
one can
describing
the
A
very
important
conclusion
from
this
table
is
the
fact
that
the
small
scale
motion
is
universal.
Therefore
one
can
expect
that
the
models
describing
the
small scale motion in contrast to RANS models are also universal.
motion
is
universal.
Therefore
one
can
expect
that
the
models
describing
the
small scale motion in contrast to RANS models are also universal.
small scale motion in contrast to RANS models are also universal.
15.2
LES equations
15.2
LES equations
15.2
LES
equations
The governing
equations
of LES are derived from the Navier Stokes equation
The governing equations of LES are derived from the Navier Stokes equation
The governing equations of LES are
from the
derived
Navier Stokes equation
@
@
@ui
@p
@
@uj
.ui / C @ .ui uj / D @ @u C @u @p C gi
(15.1)
@
@t .u / C @xj .u u / D @xj @xji C @xji
@xi C g
(15.1)
@
@
@
@ui
@uj
@p
i
i j
i
@t .ui / C @xj .ui uj / D @xj @xi C @xj
(15.1)
@xi C gi
@xj
@t
@xj
@xj 151@xi
@xi
151
151
149
@uj
@p
@
@
@
@ui
.ui / C
.ui uj / D
C
C gi
@t
@xj
@xj
@xi
@xj
@xi
B E F e e
(15.2)
ijS GS D ui uj uQ i uQj
the equation (15.2) is rewritten in the final form
@
@
@
@uQ i
@pQ
S GS
.uQ i / C
.uQ i uQj / D
ij
C gi
@t
@xj
@xj
@xj
@xi
The term ijS GS D ui uj uQ i uQj is the subgrid stress (SGS) which considers
the effect of small fine vortices on large scale motion directly resolved on the
grid.
15.3
Smagorinsky model
Note that the fine scale vortices are not resolved. They filtered out by the
filtering operation. The effect of these vortices is taken by the term ijS GS
into account. Since the small vortices are not modeled, the subgrid stress are
calculated using phenomenological models. The most recent phenomenological model was proposed by Smagorinsky in 1963. The Smagorinsky model
is just the extension of the Boussinesq approach
1
ij kk ij 2 t SNij ;
3
Smagorinsky introduced the subgrid viscosity S GS instead of the turbulent
kinematic viscosity
1 S GS
ijS GS kk
ij 2S GS SQij ;
3
Expression for the subgrid viscosity was obtained by Smagorinsky with the
use of idea taken from the Prandtl mixing length theory. According to
Prandtl, the turbulent kinematic viscosity is proportional to the mixing
length squared and the velocity gradient
152 close to the wall
t D l 2j
d uN x
j
dy
Where
t D l 2j
d uN x
j
dy
According to Smagorinsky, the subgrid viscosity is proportional to the magnitude of the strain rate tensor Sij and to a certain length lS squared
q
2 Q
Q
jSij j D 2SQij SQij
S GS D lS jSij j;
Where
@
u
Q
1
@
u
Q
i
j
C
SQij D
2 @xj
@xi
The length lS is assumed to be proportional to the mesh size
lS D CS
where CS is the constant of Smagorinsky.
The Smagorinsky constant was estimated first by Lilly. The main assumption
of the Lilly analysis is the balance between generation
Pr D ij SQij D 2 t SQij SQij D t jSQij j2
and dissipation of the turbulent kinetic energy
" D PN D t jSQij j2 D lS2 jSQij j3
(15.3)
Lilly estimated the strain rate tensor magnitude for Kolmogorov spectrum
SQ 2 7C "2=3 4=3
Substitution of the last formula into (15.3) results in:
Q 2 3=2 1=2
S
lS D
3=4
.7C /
SQ 3
3=2
1
lS
D
0:17
.7C /3=4
Advantages
Simple
Low computational costs
Stable
Good accuracy in ideal
conditions
Disadvantages
Laminar flow is not modelled
Constant of Smagorinsky is constant in time and
space
Actually, the constant is chosen arbitrarily
depending on the problem under consideration
Sensible to grid
Purely dissipativ
Damping of pulsation is too strong
15.4The
Model
of Germano ( Dynamic SmagorinWake
sky Model)
the only emission we want to leave behind
.QYURGGF'PIKPGU/GFKWOURGGF'PIKPGU6WTDQEJCTIGTU2TQRGNNGTU2TQRWNUKQP2CEMCIGU2TKOG5GTX
6JGFGUKIPQHGEQHTKGPFN[OCTKPGRQYGTCPFRTQRWNUKQPUQNWVKQPUKUETWEKCNHQT/#0&KGUGN6WTDQ
2QYGTEQORGVGPEKGUCTGQHHGTGFYKVJVJGYQTNFoUNCTIGUVGPIKPGRTQITCOOGsJCXKPIQWVRWVUURCPPKPI
HTQOVQM9RGTGPIKPG)GVWRHTQPV
(KPFQWVOQTGCVYYYOCPFKGUGNVWTDQEQO
152
15.4
ijS GS D ui uj uQ i uQj
(15.4)
b
e
(15.5)
b
e b
(15.6)
(15.7)
_
Tijt est D ui uj ui uj D ui uj uOQ i uOQj
We suppose that the double filter width is small. Therefore the Smagorinsky
model is valid for both stresses ijS GS and Tijt est :
2
1 S GS
ijS GS kk
ij D 2 Cs jSQij j SQij D 2C mSijGS ;
3
2
1 t est
Tijt est Tkk
ij D 2 Cs O jSOQij j SOQij D 2C mtijest ;
3
(15.8)
1 S GS
OijS GS Okk
ij D 2C m
O SijGS
(15.9)
3
where C D CS2 . Here we supposed that the filtered product of the constant
C with mij is equal to the product of filtered mij with the same constant
Cm Cm
O
We introduce the tensor Lij which is equal to the difference between test
filter and once filtered original SGS stress:
Download free eBooks at bookboon.com
155
(15.10)
Oij
Okk ij D 2C m
O ij
(15.9)
3
where C D CS2 . Here we supposed that the filtered product of the constant
C with mij is equal to the product of filtered mij with the same constant
Lectures on computational fluid dynamics
Cm Cm
O
We introduce the tensor Lij which is equal to the difference between test
filter and once filtered original SGS stress:
(15.10)
(15.11)
where
Mij D mtijest m
O SijGS
The system (15.11) is overdefined (six equations for one unknown coefficient
C ). To get an unique solution we multiply both the l.h.s. and r.h.s of (15.11)
with the tensor Sij . The final result for C is
C D
Lij SQij
2Mij SQij
(15.12)
Use of (15.14) is problematic since the denumerator Mij SQij can become zero.
To overcome this difficulty Lilly proposed to determine the constant from the
condition of the minimum residual of the equation (15.11):
2
1
Q D Lij Lkk ij 2CMij
! min
(15.13)
3
The minimum is attained at the point with zero derivative of the functional
Q on the parameter C :
@Q
1
D 4Mij Lij Lkk ij 2CMij D 0
(15.14)
@C
3
It follows directly from (15.14):
C D
(15.15)
156
154
15.5
(15.18)
GOT-THE-ENERGY-TO-LEAD.COM
We believe that energy suppliers should be renewable, too. We are therefore looking for enthusiastic
new colleagues with plenty of ideas who want to join RWE in changing the world. Visit us online to nd
out what we are offering and how we are working together to ensure the energy of the future.
157
155
Despite the fact that diffusion of the classic Smagorinsky model was substantially reduced by the dynamic choice of the Smagorinsky constant, the
Dynamic Smagorinsky model remains very diffusive. This disadvantage was
overcome within the similarity models. The main point of the similarity
model is the assumption that the statistical properties of the once filtered
field uQ i are identical to these of the double filtered field uQQ i . It is the case
if the filter width is small. The difference between once and double filtered
velocities is negligible, i.e. different scale motions are similar.
Let us consider uQj as the original (unfiltered) field. uQQj is the filtered field
and uQj uQQj is the pulsation (see Fig. 15.2). Then from the definition of the
subgrid stress one obtains
(15.18)
15.6
The experience shows that diffusion produced by the scale similarity model (15.18) is too low. The numerical calculations are often unstable using
this model. Taking the fact into account, that the diffusion of the Smagorinsky model is too high, it was decided to combine the Samgorinsky and scale
similarity models to get the proper diffusion. The advantages and disadvantages of both models are summarized as follows
Dynamic Smagorinsky Model (DSM): energy dissipation is overestimated (drawback), energy backscattering is not reproduced (drawback),
Scale similarity model: energy backscattering is reproduced (advantage), energy dissipation is underestimated (drawback).
The idea combine models to strengthen the advantages and to overcome disadvantages of both models. The hybrid model called as the mixed similarity
model is written as
2
Q SQij
ijr D uQ i uQj uQQ i uQQj 2 CS jSj
(15.19)
The mixed model can be derived in a more formal way. For that the velocity
decomposition into filtered and pulsation parts:
e e e e
ui uj D .uQ i C u0i /.uQj C uj0 / D uQ i uQj C u0i uQj C uj0 uQ i C u0i uj0
(15.20)
D ui uj uQ i uQj
(15.21)
(15.22)
The mixed model can be derived in a more formal way. For that the velocity
decomposition into filtered and pulsation parts:
e e e e
(15.20)
ijS GS D ui uj uQ i uQj
(15.21)
(15.22)
where
Lij D uQ i uQj uQ i uQj is the Leonard stress
Cij D uQ i uj0 C uQj u0i is the Cross stress
Rij D u0i uj0
is the Reynolds stress
e
e e
e
The sum of the cross and Reynolds stresses is calculated via the Smagorinsky
model with the dynamically determined constant CS
2
Q SQij
Cij C Rij D 158
2 CS jSj
(15.23)
(15.24)
The SGS stress does not depend on the reference system speed:
Wi Wj WQ i WQj D
G
GG
e CC e
(15.25)
(15.26)
(15.27)
e
e e
e
(15.28)
(15.26)
(15.27)
e
e e
e
(15.28)
15.7
Two tests are used to verify LES models. The comon way is the a-posteriori
test. The LES simulation is performed and then the flow parameters obtained
from the simulation are compared with these from measurement. Depending
on comparison results the conclusion about quality of LES models is drawn.
The disadvantage of such approach is that the LES results are affected by
modelling errors, errors of approximation of differential operators and rounding errors. In a-posteriori test they can not be separated.
Direct test of quality of subgrid stresses is the a-priori test. First, the subgrid
stress is calculated at each time instant from the definition
(15.29)
Then the subgrid stress is computed again at each time instant from any
model, say Smagorinsky one
1 S GS mod
ijS GS mod kk
ij D 2.CS /2 jSQij jSQij
3
(15.30)
The subgrid stresses ijS GSdef and ijS GS mod averaged in time are compared
each with other. If
ijS GSdef ijS GS mod
the SGS model is accurate.
A big difficulty of a-priori tests is the determination of velocities uQ i . For
that it is necessary first to obtain the unfiltered velocities ui with spatial and
temporal resolutions compared with the Kolmogorov scales. At present this
free eBooksto
at measure
bookboon.com
isDownload
a big challenge
three components of velocity in a volume with
high spatial and temporal resolutions. The158
Particle Image Velocimetry (PIV)
measurements are mostly planar measurements within a two dimensional
window. Direct Numerical Simulation data are often used as the source for
a-priori tests. Three components of velocity in a volume, obtained from DNS,
ij
ij
CAREERKICKSTART
An app to keep you in the know
Whether youre a graduate, school leaver or student, its a difcult time to start your career.
So here at RBS, were providing a helping hand with our new Facebook app. Bringing together
the most relevant and useful careers information, weve created a one-stop shop designed
to help you get on the career ladder whatever your level of education, degree subject or
work experience.
And its not just nance-focused either. Thats because its not about us. Its about you.
So download the app and youll get everything you need to know to kickstart your career.
So what are you waiting for?
Click here to get started.
159
161
Chapter
Chapter 16
16
Chapter 16
Chapter
16
Hybrid
Hybrid URANS-LES
URANS-LES
Hybrid URANS-LES
Hybrid
URANS-LES
16.1 Introduction
16.1
methods
methods
methods
methods
Introduction
As
discussed
above, the most promising approach to resolve the flow un16.1
Introduction
As discussed
above, the most promising approach to resolve the flow unsteadiness is the Large Eddy Simulation (LES), which is already widely used
16.1
steadiness Introduction
is the Large Eddy Simulation (LES), which is already widely used
for research purposes. Typical Reynolds numbers in engineering are very
for
research purposes.
Reynolds approach
numbers in
are very
As discussed
above, theTypical
most promising
to engineering
resolve the flow
unlarge even at model scales. The grid resolution necessary for a pure LES
large
even
at
model
scales.
The
grid
resolution
necessary
for
a
pure
LES
steadiness
is
the
Large
Eddy
Simulation
(LES),
which
is
already
widely
used
As discussed above, the most promising approach to resolve the flow unis so huge that it makes the direct application of LES impossible (see Sec.
is
soresearch
huge is
that
itLarge
makes
the Simulation
direct
application
of LES
impossible
(see
Sec.
for
purposes.
Typical
Reynolds
numbers
inisengineering
are used
very
steadiness
the
Eddy
(LES),
which
already widely
16.4). A practical solution of this problem is the use of hybrid URANS-LES
16.4).
A practical
solution
of The
thisReynolds
problem
is the use
of engineering
hybrid
large
even
atpurposes.
model
scales.
grid resolution
necessary
forURANS-LES
a pure
LES
for research
Typical
numbers
in
are very
methods, where the near body flow region is treated using URANS and far
methods,
near
body
flow
region
is treated
using
URANS
andLES
far
is
so huge
that
itthe
makes
the
direct
application
ofnecessary
LES
impossible
(see
Sec.
large
evenwhere
at model
scales.
The
grid
resolution
for a pure
flow regions are treated with LES.
flow
are it
treated
16.4).
A practical
solution
of LES.
this problem
is theofuse
of hybrid
URANS-LES
is so regions
huge
that
makeswith
the
direct
application
LES
impossible
(see Sec.
methods,
where
the
near
body
flow
region
is
treated
using
URANS
and far
16.4). A practical solution of this problem is the use of hybrid URANS-LES
According
to
Peng
[17]
the
hybrid
techniques
can
be
subdivided
into
flow
flow
regions
treated
LES.
methods,
where
the [17]
nearwith
body
flow techniques
region is treated
URANSinto
andflow
far
According
toare
Peng
the
hybrid
can beusing
subdivided
matching and turbulence matching methods. Within the flow matching
flow regions
areturbulence
treated with
LES. methods. Within the flow matching
matching
and
matching
methods the interface between URANS and LES is explicitly defined. LES
methods
URANS
and LES
explicitly
defined.
Accordingthe
to interface
Peng [17]between
the hybrid
techniques
canis be
subdivided
into LES
flow
filtered equations are solved in the LES region, whereas URANS equations
filtered
equations
are
solved
in
the
LES
region,
whereas
URANS
equations
matching
and
turbulence
matching
methods.
Within
the
flow
matching
According to Peng [17] the hybrid techniques can be subdivided into flow
are solved in the URANS domain. The flow parameters (velocities, kinetic
are
solvedthe
in interface
the
URANS
domain.
The
flow LES
parameters
(velocities,
kinetic
methods
between
URANS
and
is explicitly
defined.
LES
matching
and
turbulence
matching
methods.
Within
the
flow
matching
energy) are matched at the interface between the URANS and the LES reenergy)
are
the interface
between
thewhereas
anddefined.
theequations
LESLES
refiltered
are at
solved
in URANS
the LES
region,
URANS
methodsequations
thematched
interface
between
and
LES
isURANS
explicitly
gions. Among the most important contributions to the development of flow
gions.
Among
the
most
important
contributions
to
the
development
of
flow
are
solved
in
the
URANS
domain.
The
flow
parameters
(velocities,
kinetic
filtered equations are solved in the LES region, whereas URANS equations
matching methods we mention the works of (Davidson, Dalstroem [18]; Termatching
methods
weatmention
the works
of (Davidson,
Dalstroem
Terenergy)
are
thedomain.
interface
between
the URANS
and the[18];
LES
reare solved
inmatched
the URANS
The
flow
parameters
(velocities,
kinetic
racol [19]; Jakirlic et. al. [20]; Temmerman et. al. [21]) and others. A
racol
[19];
Jakirlic
et.
al.
[20];
Temmerman
et.
al.
[21])
and
others.
A
gions.
Among
the
most
important
contributions
to
the
development
of
flow
energy) are matched at the interface between the URANS and the LES reserious weakness of this approach is the development of robust procedures to
serious
weakness
thismention
approach
the development
of robust
procedures
to
matching
methods
we
theisworks
of (Davidson,
Dalstroem
[18];
gions. Among
theofmost
important
contributions
to the
development
of Terflow
set the URANS-LES interface for complicated flow geometries. Within the
set
the[19];
URANS-LES
for
complicated
flowal.
geometries.
the
racol
Jakirlic we
et.interface
al. [20];
et.
[21])
and Within
others.
A
matching
methods
mention
theTemmerman
works of (Davidson,
Dalstroem
[18]; Terframework of the turbulence matching method an universal transport equaframework
of
the
turbulence
matching
method
an
universal
transport
equaserious
weakness
of
this
approach
is
the
development
of
robust
procedures
to
racol [19]; Jakirlic et. al. [20]; Temmerman et. al. [21]) and others. A
tion is solved in the whole computational domain. The stress terms in this
tion
is solved
in the
whole
computational
domain.
stress procedures
terms
in this
set
the
URANS-LES
interface
foriscomplicated
flow The
geometries.
Within
the
serious
weakness
of this
approach
the development
of robust
to
equation are treated in different ways in LES and URANS domains. There
equation
are
treated
in
different
ways
in
LES
and
URANS
domains.
There
framework
of the turbulence
method flow
an universal
transport
equaset the URANS-LES
interfacematching
for complicated
geometries.
Within
the
are various procedures to distinguish between LES and URANS cells. The
are
various
procedures
to distinguish
LES
and stress
URANS
cells.inequaThe
tion
is solved
in the
whole
computational
domain.
The
terms
this
framework
of
the
turbulence
matchingbetween
method
an universal
transport
most popular hybrid method is Detached Eddy Simulation (DES) proposed
most
hybrid
method
is Detached
Eddyand
Simulation
(DES)
proposed
equation
are treated
in
different
ways in LES
URANS
domains.
There
tion(Spalart
ispopular
solved
the whole
The stress
terms
in
this
by
et in
al.[22]).
Thecomputational
original versiondomain.
of this method
is based
on
the
are
various
procedures
to
distinguish
between
LES
and
URANS
cells.
The
equation
are
treated
in
different
ways
in
LES
and
URANS
domains.
There
classic Smagorinsky LES model and 163
the Spalart-Allmaras (SA) URANS apmost
popular
hybrid method
is Detached
Eddy Simulation
(DES)cells.
proposed
163
are
various
procedures
to
distinguish
URANS
proach. SA is used close to the wall, between
whereas LES
LES and
in the
rest part of The
the
most popular
hybrid method
Detached
Eddy Simulation
proposed
flow.
The switching
between isthe
two techniques
is smooth(DES)
and occurs
in
163
a gray subdomain. There are two major improvements of DES, developed recently. The first one, DDES163
(Delayed DES), has been proposed to
detect the boundary layers and to prolong the RANS mode, even if the wallparallel grid spacing would normally activate the DES limiter (Spalart [23]).
The second one, IDDES (Improved DDES), allows one to solve the problems with modelled-stress depletion and log-layer mismatch. For the details
see the review (Spalart [23]). In spite of a wide application area DES has
serious principle limitations thoroughly analyzed by (Menter, Egorov [24]).
Other versions of the turbulence matching methods using different blending functions to switch the solution between LES and URANS modes were
proposed by (Peng [17]; Davidson, Billson [25]; Abe, Miyata [26]) and others.
Download free eBooks at bookboon.com
160
A very critical point of the turbulence matching
methods is the transition
from the time (or ensemble) averaged smooth URANS flow to the oscillating LES flow, see (Menter, Egorov, 2005). The oscillations have to appear
within a short flow domain in a gray zone between LES and URANS. Ex-
16.2
16.2
The most popular hybrid method -detached eddy simulation- was proposed
The
most
hybrid
method
-detachedofeddy
proposed
in
1997
by popular
Spalart et
al. [22].
The principle
DES simulationis illustratedwas
in Fig.
16.1.
in 1997tobythe
Spalart
al.solution
[22]. The
of using
DES isthe
illustrated
Fig. 16.1.
Close
body etthe
is principle
calculated
URANSinmode.
Far
Closethe
to wall
the body
the equations
solution isare
calculated
using
thezone
URANS
mode.
Far
from
the LES
solved. The
grey
between
URANS
fromLES
the wall
LES solution.
equations are solved. The grey zone between URANS
and
is thethe
mixed
and
is the
mixedofsolution.
The LES
classical
version
the DES approach is based on the Spalart Almaras
The classical
version of with
the DES
approach
is based on
the Spalart
Almaras
(SA)
model formulated
respect
to the modified
turbulent
viscosity
Q D
with respect
the modified turbulent viscosity Q D
(SA)
. The formulated
transport equation
for Q to
reads
t =f1model
t =f1 . The transport equation for Q reads
164
164
2
@Q
@Q
Q
C vNj
D Cb1 SQ Q Cw1 fw
C
@t
@xj
d
Generation
Destruction
@Q
Cb2 @Q @Q
1 @
. C /
Q
C
C
@x
@xk
@xk @xk
k
Diffusion
where
161
Cb2 D 0:622;
C1 D 7:1;
Cb1
1 C Cb2
;
D 2=3; Cw1 D 2 C
C D 0:3; C D 2:0; D 0:41;
Cb1 D 0:1355;
(16.1)
@t
@xj
Destruction
@Q
Cb2 @Q @Q
1 @
. C /
Q
C
C
@xk
@xk @xk
@xk
Lectures on computational fluid
dynamics
Generation
(16.1)
Hybrid URANS-LES methods
Diffusion
where
Cb1 D 0:1355;
Q
SQ D S C 2 2 f2 ; S D
d
1 @N i
@Nj
ij D
2 @xj
@xi
p
2ij ij ;
165
162
Here d is the distance from the wall. The physical sense of different terms
is illustrated in (16.1). Far from the wall the generation and the distruction
terms are approaching each to other and the turbulence attains the equilibrium state:
2
Q
Cb1 SQ Q Cw1 fw
0
d
Generation
Destruction
Cb1 Q 2
Sd
Cw1
lS2 jSQij j;
jSQij j D
q
2SQij SQij
lS D CS
DES inventors proposed to use the following expression for d :
d D minfd; CDE S g;
where CDE S 1:3 is the DES constant. Now the main idea of the DES
becomes obvious:
At small wall distance d < CDE S the Spalart Almaras URANS model is
active
At large wall distance d > CDE S the Spalart Almaras URANS model is
smoothly passed into the Smagorinsky model.
Samples of DES applications are presented in Fig. 16.2 and 16.3.
Despite of the wide application Detached Eddy Simulation technique is not
free of disadvantages. Menter [24] notes: The essential concern with DES
is that it does not continuously change from RANS to LES under grid refinement. In order for LES structures to appear, the grid spacing and time
step have to be refined beyond a case-dependent critical limit. In addition, a
sufficiently large instability mechanism has to be present to allow the rapid
formation of turbulent structures in regions where the DES limiter is activated. If one of the two, or both requirements are violated, the resulting
model is undefined and the outcome is largely unpredictable.
166
163
Figure 16.2: Flow around combat aircraft. (Squires K.D., Detached-eddy simulation:
current
status and perspectives)
perspectives.
Figure
around
combat
aircraft. (Squires
K.D., Detached-eddy
simulation:
Figure16.3:
16.3:Flow
Squires
K.D.,
Detached-eddy
simulation:
current status
and
current
status and perspectives)
perspectives.
16.3
A simple hybrid model proposed in [29] is based on the observation that the
basic transport equations have the same form in LES and RANS
@.ijl C ijt /
@ui
@.ui uj /
@p
C
D
C
;
@t
@xj
@xi
@xj
(16.2)
but the interpretation of the overline differs. In LES it means filtering, but
in RANS it stands for the Reynolds, or ensemble, averaging. Here we used
the standard notation of p for the pseudo-pressure, and ijl and ijt for the
167
164
Figure 16.4: The division of the computational domain into the URANS
(dark) and LES (light) regions at one time instant for hybrid calculation of
tanker.
laminar and turbulent stresses respectively. Note that the turbulent stresses
are calculated in different ways in LES and URANS regions.
The computational domain in our model is dynamically (i.e. at each time
step) divided into the LES and URANS regions. A cell of the mesh belongs
to one or the other region depending on the relation between the integral
length scale L and the extended LES filter according to the following rule:
if L > then the cell is in the LES region;
if L < then the cell is in the URANS region:
(16.3)
The integral length scale is calculated from the known formula of Kolmogorov
and Prandtl with the correction factor 0:168 taken from [12]
k
The next stepL Dfor
C
;
(16.4)
"
top-performing
where
k is the turbulent kinetic energy and " is the dissipation rate. The
constant C is C 0:168 close to the wall y= < 0:2, C 0:35 at 0:2 < y= <
graduates
0:7 and C 1:0 in the outer area of the bounder layer y= > 0:7, where is
3=2
the boundary layer thickness. L varies from one time step to another, which
results in varying decomposition of the computational domain into the LES
and URANS regions. The extended LES filter is computed as
q
2
D dmax
C 2;
(16.5)
Designed
high-achieving
graduates
acrossD
all disciplines,
whereMasters
dmax in
is Management
the maximal length
of for
the
cell edges
dmax
max.dxLondon
; dy ;Business
dz / Schools Masters
in Management provides specific and tangible foundations for a successful career in business.
1=3
and D .the cell volume/
is the common filter width used in LES. This
This 12-month, full-time programme is a business qualification with impact. In 2010, our MiM
choice ensures that very flat cells
in the
boundary
which
of0graduates choosing to
employment
rate was
95% within 3 layer
months of(for
graduation*;
the majority
work in consulting or financial services.
but dmax > 0) are treated correctly. depends only on the mesh and it is
As well as a renowned qualification from a world-class business school, you also gain access
to the Schools network of more than 34,000 global alumni a community that offers support and
opportunities throughout your career.
168
165
length scale L and the extended LES filter according to the following rule:
if L > then the cell is in the LES region;
if L < then the cell is in the URANS region:
Hybrid(16.3)
URANS-LES methods
The integral length scale is calculated from the known formula of Kolmogorov
and Prandtl with the correction factor 0:168 taken from [12]
k 3=2
;
(16.4)
"
where k is the turbulent kinetic energy and " is the dissipation rate. The
constant C is C 0:168 close to the wall y= < 0:2, C 0:35 at 0:2 < y= <
0:7 and C 1:0 in the outer area of the bounder layer y= > 0:7, where is
the boundary layer thickness. L varies from one time step to another, which
results in varying decomposition of the computational domain into the LES
and URANS regions. The extended LES filter is computed as
q
2
D dmax
C 2;
(16.5)
LDC
where dmax is the maximal length of the cell edges dmax D max.dx ; dy ; dz /
and D .the cell volume/1=3 is the common filter width used in LES. This
choice ensures that very flat cells in the boundary layer (for which 0
but dmax > 0) are treated correctly. depends only on the mesh and it is
precomputed only once before the main computation.
168
As a sample LES and URANS regions are shown in Fig. 16.4 for flow around
a tanker. The URANS region is located close to the ship surface and plays
the role of a dynamic wall function. In areas of bilge vortices formation, the
boundary layer is shedding from the hull and penetrates into the outer flow
part. Since the boundary layer is a fine scale flow the procedure (16.3) recognizes the bilge vortex formation zones as URANS ones. There is a technical
issue concerning the cells which are far from the ship hull and where both
k and " are small, so large numerical errors are introduced into the integral
length scale computed according to Eq. (16.4). To avoid an irregular distribution of URANS and LES zones, the general rule (16.3) of the domain
decomposition is corrected in such a way that the LES region is switched to
URANS one if k is getting less than some threshold. This procedure has no
influence on the ship flow parameters since it is used far from the area of the
primary interest.
We have performed several calculations with different combinations of LES
and URANS models to find the most efficient one for the problem under
consideration. Among the models we used in our computations are the linear
and nonlinear k-", k-! SST and k"v 2 f URANS models combined with the
simple and dynamic Smagorinsky as well as with the dynamic mixed LES
closure models. The experience shows that the most satisfactory results are
obtained using the URANS approach based on the k"v 2 f turbulent model
of [30] and LES approach based on the Smagorinsky dynamic model. The
turbulent stresses ijt are calculated from the Boussinesq approximation using
the concept of the turbulent viscosity. The only difference between LES
and URANS is the definition of the kinematic viscosity. Within LES it is
considered as the subgrid viscosity and calculated according to the dynamic
model of Smagorinsky:
Download free eBooks at bookboon.com
@ui
1 @uj
166
2
C
SGS D cD jSij j; Sij D
;
(16.6)
2 @xi
@xj
where Sij is the strain velocity tensor and cD is the dynamic constant. In
the URANS region the viscosity is calculated from the turbulent model of
turbulent stresses ijt are calculated from the Boussinesq approximation using
the concept of the turbulent viscosity. The only difference between LES
and URANS is the definition of the kinematic viscosity. Within LES it is
considered
as the subgrid
viscosity and calculated according to the
dynamic
Lectures
on computational
fluid dynamics
Hybrid
URANS-LES methods
model of Smagorinsky:
1 @uj
@ui
2
SGS D cD jSij j; Sij D
C
;
(16.6)
2 @xi
@xj
where Sij is the strain velocity tensor and cD is the dynamic constant. In
the URANS region the viscosity is calculated from the turbulent model of
[30]:
k2
2
t D min 0:09 ; 0:22v T t ;
"
(16.7)
where v 2 is the wall normal component of the stresses and T t is the turbulent
p
time T t D max.k="; 6 ="/.
169
16.4
In the CFD community one can observe tendency to use pure LES without
paying any attention to resolution problems. Very often LES is running on
typical RANS grids. In fact, such computations can give correct results if the
flow structures to be captured are large enough and exist for a long time. In
some cases modeling of such structures does not require detailed resolution
of boundary layers and a thorough treatment of separation regions. As an
example one can mention flows around bluff bodies with predefined separation lines like ship superstructures. Application of underresolved LES for
well streamlined hulls should be considered with a great care. First of all,
one should not forget that the basic LES subgrid models are derived under
the assumption that at least the inertial turbulent subrange is resolved. Second, underresolution of wall region leads to a very inaccurate modeling of
the boundary layer, prediction of the separation and overall ship resistance.
It is clearly illustrated in the Table 16.1. The ship resistance obtained from
underresolved LES using the wall function of (Werner, Wengle [31]) is less
than half of the measured one and that obtained from RANS. Obviously, the
application of modern turbulence LES models, more advanced than RANS
models, does not improve but even makes the results much worse with the
same space resolution. The change from RANS to LES should definitely be
followed by the increase of the resolution which results in a drastic increase
of the computational costs. These facts underline necessity of further development towards hybrid methodology. Although in (Alin et al. [32]) it has
been shown that the accuracy of the resistance prediction using pure LES at
a very moderate resolution with y C 30 can be improved using special wall
functions, the most universal way for the present, to our opinion, is application of hybrid methods. The impossibility of pure LES is illustrated below
for flow around the KVLCC2 tanker.
The precise determination of the necessary LES resolution is quite difficult.
Estimations presented below are based on the idea that about 80% of the
Download free eBooks at bookboon.com
turbulent
kinetic energy should be directly resolved and the rest is modeled
in a properly resolved LES simulation. Implementation
of this idea implies
167
the knowledge of the Kolmogorov and the integral length L scales which are
used to draw the typical spectra of the full developed turbulence E.k/. The
The maximum possible cell size is then max D 2=k . The scales L and
are found from the known expression D . 3 ="/1=4 and Eq. (16.4), where
the kinetic energy k and the dissipation rate " are taken from RANS simulations using k-" linear model. The ratio D max = is then used as the scale
parameter for grid generation. Both lengths vary in space which makes the
grid generation procedure very complicated. To roughly estimate the size of
the grid we assume that is constant. We performed different calculations
determining at the two following points: i) the point where L= is maximal
in the ship boundary layer and ii) the point in the propeller disk where the
vorticity !E is maximal (region of the concentrated vortex structure). The latter is dictated by the wish to resolve the most intensive vortex flow structures
which have the strongest influence on the propeller operation. Since LES application is required in the ship stern area only this part of the computational
volume has been meshed. It covers the boundary layer of the stern region
starting from the end of the parallel midship section. The thickness of the
meshed region has been constant and equal to the maximum boundary layer
thickness at the stern BL . The grid for a pure LES is generated using the
following algorithm. The minimum Kolmogorov length min is determined
in the near wall region. The cell sizes in x and z directions along the wall
are calculated by multiplication of min with the scale parameter . These
sizesMA
remain
constant for
all cells row inwith
y direction
which isBusiness
normal to the
Intercultural
Communication
International
shipMA
surface
(see Fig. 16.5).
The
cells have at Marketing
least two equal sizes which is
Communication
and
International
desirable from the point of view of LES accuracy. The choice of the size in y
direction is dictated by proper resolutionMAofIntercultural
the boundary
layer. Close to the
Communication with International Business
wall this size is chosen from the condition
wyouDwithmin.y
; min /. Since
yw is in contemporary
Provides
a criticalw
understanding
of communication
contexts
combining in
linguistic,
cultural/media studies and
chosen as the ordinate where y C D 1 thesocio-cultural
first nodes
laybydeeply
the viscous
international business and will prepare you for a wide range of careers.
sublayer. The size in y direction at the upper border of the boundary layer
MA Communication and International Marketing
is equal to 1 D , where is the Kolmogorov scale at y D BL . A
Equips you with a detailed understanding of communication in contemporary
simple grading is used in y direction between
w
and
international
marketing
contexts
1 . to enable you to address the market needs of
further
information
Results of the estimations are as follows:Forthe
required
gridcontact:
size ranges from
(0)1483
5 M to 25 M for Re D 2:8 106T:, +44
and
from681681
7 M to 60 M for
E: pg-enquiries@surrey.ac.uk
Re D 5:8 106 . The results vary depending
on the value of in use, so
www.surrey.ac.uk/downloads
they should be considered as very rough estimations. Together with similar estimations for the nonlinear k-" model these results show that the LES
171
168
E.k/d k
Lectures on computational fluid dynamics
Rk
1
0
E.k/d k
0:2:
Hybrid (16.8)
URANS-LES methods
The maximum possible cell size is then max D 2=k . The scales L and
are found from the known expression D . 3 ="/1=4 and Eq. (16.4), where
the kinetic energy k and the dissipation rate " are taken from RANS simulations using k-" linear model. The ratio D max = is then used as the scale
parameter for grid generation. Both lengths vary in space which makes the
grid generation procedure very complicated. To roughly estimate the size of
the grid we assume that is constant. We performed different calculations
determining at the two following points: i) the point where L= is maximal
in the ship boundary layer and ii) the point in the propeller disk where the
vorticity !E is maximal (region of the concentrated vortex structure). The latter is dictated by the wish to resolve the most intensive vortex flow structures
which have the strongest influence on the propeller operation. Since LES application is required in the ship stern area only this part of the computational
volume has been meshed. It covers the boundary layer of the stern region
starting from the end of the parallel midship section. The thickness of the
meshed region has been constant and equal to the maximum boundary layer
thickness at the stern BL . The grid for a pure LES is generated using the
following algorithm. The minimum Kolmogorov length min is determined
in the near wall region. The cell sizes in x and z directions along the wall
are calculated by multiplication of min with the scale parameter . These
sizes remain constant for all cells row in y direction which is normal to the
ship surface (see Fig. 16.5). The cells have at least two equal sizes which is
desirable from the point of view of LES accuracy. The choice of the size in y
direction is dictated by proper resolution of the boundary layer. Close to the
wall this size is chosen from the condition w D min.yw ; min /. Since yw is
chosen as the ordinate where y C D 1 the first nodes lay deeply in the viscous
sublayer. The size in y direction at the upper border of the boundary layer
is equal to 1 D , where is the Kolmogorov scale at y D BL . A
simple grading is used in y direction between w and 1 .
Results of the estimations are as follows: the required grid size ranges from
5 M to 25 M for Re D 2:8 106 , and from 7 M to 60 M for
Re D 5:8 106 . The results vary depending on the value of in use, so
they should be considered as very rough estimations. Together with similar estimations for the nonlinear k-" model these results show that the LES
171
169
min
min
min
=11
yy =
min
min
=1
++
min
z
x
z
z y
y
y
x
x
Figure 16.5: The cell parameters.
Figure
Figure 16.5:
16.5: The
The cell
cell parameters.
parameters.
grid should have the order of tens of millions of nodes. Nowadays, the computations with
hundred
millions
andmillions
even with nodes.
a few billions of nodes
are
grid
grid should
should have
have the
the order
order of
of tens
tens of
of millions of
of nodes. Nowadays,
Nowadays, the
the comcombecoming available
in the researchand
community.
However,
a numerical
study
putations
putations with
with hundred
hundred millions
millions and even
even with
with aa few
few billions
billions of
of nodes
nodes are
are
of engineering
problems
implies
usually
many computations
which
havestudy
to be
becoming
available
in
the
research
community.
However,
a
numerical
becoming available in the research community. However, a numerical study
performed within
a reasonable time
with moderate
computational have
resources.
of
of engineering
engineering problems
problems implies
implies usually
usually many
many computations
computations which
which have to
to be
be
In
this
sense,
the
results
of
the
present
subsection
clearly
demonstrate
that
performed
within
a
reasonable
time
with
moderate
computational
resources.
performed within a reasonable time with moderate computational resources.
the puresense,
LES is impossible
for ship applications
so far. To
verify that the
In
In this
this sense, the
the results
results of
of the
the present
present subsection
subsection clearly
clearly demonstrate
demonstrate that
that
resolution
estimation
procedure
we
used
gives meaningful
results,
it has
been
the
pure
LES
is
impossible
for
ship
applications
so
far.
To
verify
that
the pure LES is impossible for ship applications so far. To verify that the
the
applied forestimation
turbulent boundary
layer
(TBL)
benchmark.
We
foundit from
meresolution
procedure
we
used
gives
meaningful
results,
has
been
resolution estimation procedure we used gives meaningful results, it has been
thodical calculations
that the pure LES
with 1M cells isWe
quite accurate
for
applied
applied for
for turbulent
turbulent boundary
boundary layer
layer (TBL)
(TBL) benchmark.
benchmark. We found
found from
from memeprediction of the velocity
distribution,
TBL thickness,
TBL displacement
thodical
thodical calculations
calculations that
that the
the pure
pure LES
LES with
with 1M
1M cells
cells is
is quite
quite accurate
accurate for
for
thickness
and
the
wall
shear
stress.
The
estimation
procedure
presented
prediction
of
the
velocity
distribution,
TBL
thickness,
TBL
displacement
prediction of the velocity distribution, TBL thickness, TBL displacement
above predictedthe
the necessary
resolution around
0.5M. Therefore,presented
the estithickness
thickness and
and the wall
wall shear
shear stress.
stress. The
The estimation
estimation procedure
procedure presented
mations presented
for a ship model
are rather lower
bound for the resolution
above
above predicted
predicted the
the necessary
necessary resolution
resolution around
around 0.5M.
0.5M. Therefore,
Therefore, the
the estiestirequired
for
a
pure
LES.
mations
presented
for
a
ship
model
are
rather
lower
bound
for
the
resolution
mations presented for a ship model are rather lower bound for the resolution
required
required for
for aa pure
pure LES.
LES.
CR
CP
CF
3
KRISO Exp.
4:11 10
15% 85%
2
3
RANS k"v f
4:00 10
16% 84%
k-! SST SAS
3:80 103 18% 82%
Underresolved LES 1:70 103 81% 19%
Hybrid RANS LES 4:07 103 17% 83%
Table 16.1: Results of the resistance prediction using different methods. CR is
the resistance coefficient, CP is the pressure resistance and CF is the friction
resistance
172
172
172
170
Part III
CFD applications to human
thermodynamics
171
175
Chapter 17
Chapter 17
Chapter
17
Mathematical
model of the ice
Mathematical
the ice
protection
of amodel
humanof
at
Mathematical
model
ofbody
the ice
protection
of a human
at
high
temperatures
of body
protection
of a human
body at
high
temperatures
of
surrounding
medium
high temperatures
of
surrounding medium
surrounding medium
17.1
Designations
17.1 Designations
17.1.1
List of symbols
17.1
Designations
17.1.3
Superscripts
A area (m2 )
17.1.1
List of symbols
c
specific
heatofcapacity
(J/(kgK))
p
17.1.1 List
symbols
2
A
(m of) heat release(W=m3 )
F area
density
2
) capacity
cA
specific
heat
(J/(kgK))
hp area
sector(mheight
(m)
3
cF
specific
heat
capacity
(J/(kgK))
of
heat
release(W=m
) temperature
Kp density
correlation
factor
of the
rectal
3
F
density
of
heat
release(W=m
)
h
sector
heightconductivity
(m)
k
the
thermal
(W/(mK))
h
sectordone
height
(m)
K
correlation
factor
of the
M
work
by
person
(W)rectal temperature
K body
correlation
factor
of the rectal
temperature
k
the
thermal
conductivity
(W/(mK))
m
mass (kg)
k
the
thermal
conductivity
(W/(mK))
M
work
done
by
person
(W)
P body height (cm)
M internal
work done
by
person
m
body
mass
(kg)
Q
energy
(J) (W)
m
body
mass
(kg)
P
P heat flux
height
Q
(W)(cm)
P
body
height
Q
internal
energy
(J)
R sector radius(cm)
(m)
P
internal
(J)
Q temperature
heat
fluxenergy
(W)(deg)
T
P
Q
heat
flux
(W)
R radial
sector coordinate
radius (m) along the sector (m)
x
R
sector
(m)
T
temperature
centralradius
sector(deg)
angle (deg)
T
temperature
(deg)
x
radial
coordinate
skin
thickness
(m)along the sector (m)
2 the sector (m)
x
radial
coordinate
along
central sector
angle
(deg)
thermal
diffusivity
(m
=s)
3
central
sector (kg=m
angle
skin
(m) (deg)
massthickness
density
)
2
skin
(m) (m
thermal
diffusivity
=s)
initialthickness
temperature
(deg)
3
thermal
diffusivity
mass density
(kg=m(m)2 =s)
178
3
mass
density
(kg=m
)
initial
temperature
(deg)
17.1.2 Subscripts
degree of Celsius
17.2
Introduction
172
j
number of point on the human body
med
surrounding medium
melt melting ice
outer outer boundary of clothes
Lectures
computational
skin onhuman
bodyfluid
skindynamics
17.1.3
0
Superscripts
degree of Celsius
17.2
Introduction
The heat transfer inside the human body is an extremely complicated problem which is difficult to formulate properly. The body generates the heat
within certain organs which is then transferred by the thermal diffusion
through the body substance possessing very non uniform properties. However, the thermal diffusion being complicated is not the main difficulty of
modeling. A large fraction of the heat generated by internal organs is transported to the body periphery by a complicated net of blood vessels. This
process has still not been modeled with a desirable accuracy. While a full
detailed model of the human thermodynamics is still remaining the problem
of the future, the practical design of protection clothes demands the reliable
models already now. Such models are necessary to reduce the time and costs
consuming measurements and to avoid dangerous experiments with people
under emergency conditions. The models can be only of a semi empiric character and must relay on measurements data to diminish the modelling errors.
Especially interesting is the case of human thermodynamics at high temperatures T of surrounding medium. At T > 350 the ability of human
thermoregulation is very restricted [33]. The acceptable body temperature
variation can be only one degree of Celsius. If the body temperature exceeds
this threshold the protecting mechanisms regulating the excessive heat in the
organism through the breathing, humidity diffusion of the skin, radiation and
transpiration dont work [34], [35]. The most effective principle of the protection is the mechanism of the heat absorption through, for instance, the ice
embedded into the clothes. This approach being environmentally friendly,
179
Figure 17.1: Sketch of the human body used in simulations. A- heart, BDownload free eBooks at bookboon.com
liver,
C - kidney.
Figure
17.1: Sketch of the human body used in simulations. A- heart, Bliver, C - kidney.
173
technologically easy and renewable can be used for protection clothes of res0
andhigh
renewable
can be used
for 50
protection
clothes of rescuetechnologically
teams workingeasy
under
temperatures
up to
and more.
cue teams working under high temperatures up to 500 and more.
17.3
The human body model used for simulation is shown in Fig. (17.1). The
main sizes of the body which are important for the discussion of results are
presented in the table 17.1.
Commonly the human body is represented as a set of geometric elements according to the original idea proposed by Stolwijk [36]. The body is assumed
to be slender with domination of the heat transfer in horizontal planes. On
the contrary to common models the horizontal cross section of the human
body is represented in our model as a nearly elliptical section with the displaced center. The cross section of the body is subdivided into the five
layers [33]: skin, fat layer, muscles and two core layers (see Fig. (17.2)). The
core represents all human organs and blood vessels. The cross section of
the human body is then represented as a set of sectors subdivided along the
radius into five layers with constant thermal properties (see Fig. 17.2). The
180
174
Size in meter
0.127
0.127
0.234
0.090
0.215
radii of each sector R can be found from the table 17.2. The upper elliptical
cylinder is from the neck to the waist, whereas the lower one from the waist
to the thing (Fig. 17.1). The skin thickness is approximately calculated as
the radius R divided by 60:25 [33]. The sizes of different layers inside of the
cross section are presented in the table 17.3. At the high temperatures the
main part of the heat production occurs in certain organs. The knowledge
about the heat distribution inside the body allows one to choose the most
efficient design of the ice protection which typical construction is shown in
Fig. 17.3. In the mathematical model the ice and the clothes layer are modelled by additional layers covering the human body (Fig. 17.2). The human
body model was developed on the base of data taken from [36], [37], [38]
and [39].
17.4
Mathematical model
@T
@T
dQ1 D k.x C dx/S.x C dx/
k.x/S.x/
@x
@x
181
175
176
lower elliptical
cylinder, m
0.109
0.112
0.121
0.135
0.156
0.176
0.190
0.173
0.145
0.121
0.106
0.980
0.940
Table 17.3: Radii of layers used in simulations in fraction of the skin thickness
Elliptical
cylinder
Element
length, m
upper
0.476
lower
0.317
Area
m2
Layer
Core
0.489 Muscle
Fat layer
Skin
Core
0.293 Muscle
Fat layer
Skin
Radius
36:15
53:70
59:25
60:25
36:15
53:70
59:25
60:25
of the sector. Within the material we have additionally the heat release with
the heat density F (heat per unit volume):
dQ2 D SF dxdt
(17.2)
183
177
OP
0.106
0.940
Elliptical
cylinder
Element
length, m
upper
0.476
lower
0.317
Area
m2
Layer
Core
0.489 Muscle
Fat layer
Skin
Core
0.293 Muscle
Fat layer
Skin
Radius
36:15
53:70
59:25
60:25
36:15
53:70
59:25
60:25
of the sector. Within the material we have additionally the heat release with
the heat density F (heat per unit volume):
dQ2 D SF dxdt
(17.2)
(17.4)
where .x/ D k=.cp / is the thermal diffusivity and f .x/ D F .x/=.cp /.
The equation (17.4) is solved at the initial condition
T .x; 0/ D '.x/
(17.5)
(17.6)
(17.7)
The temperature of the surrounding medium was Tmed D 500 . It was assumed
that the core center (at x D 0) temperature was constant in time
Download free eBooks at bookboon.com
Tcore D 36:70 .
178
The heat transfer process is subdivided into three steps. Within the first
step the ice is warmed up to the melting temperature. The melting process
is followed by the absorption of the heat coming from both the surrounding
T .0; t/ D Tcore
and at the outer boundary of the clothes
Lectures on computational fluid dynamics
T .xout er ; t/ D Tmed
(17.6)
Mathematical model of the ice protection
of a human body at high temperatures
of surrounding medium
(17.7)
The temperature of the surrounding medium was Tmed D 500 . It was assumed that the core center (at x D 0) temperature was constant in time
Tcore D 36:70 .
The heat transfer process is subdivided into three steps. Within the first
step the ice is warmed up to the melting temperature. The melting process
is followed by the absorption of the heat coming from both the surrounding
medium Qmed
@T
QP med D med
Amed
@x
and the human body Qhum caused by metabolism
@T
QP hum D ski n Aski n
@x
(17.8)
(17.9)
(17.10)
Here Qmelt is the heat necessary to melt the whole ice. After that, within
the third step the water from the melted ice is warming up due to heat fluxes
(17.8) and (17.9) continuing in time. The calculations within the third step
are running as long as the comfort conditions are fulfilled.
The equation (17.4) is solved using the finite differential method on the
uniform grid x D const with the constant time step t D const. The
numerical implementation utilizes the central differential scheme for space
derivatives and Crank Nicolson implicit representation of the unsteady term.
An inhouse code was developed for this purposes.
17.5
Results
17.5.1
The calculations were performed for five elements of the human body shown
in Fig.17.1. Thermodynamic properties of the body layers are given in the
table
17.4. Usually, the ice protection construction consists of the ice layer
Download free eBooks at bookboon.com
separated from the surrounding medium by several layers. In the present
179 with thickness of 0:001m, isowork we used three layers: outer clothing layer
lating layer of 0:009m and the air layer of 0:003m. All layers have the initial
temperature Ti0 D 190 . The air gap between the ice and the underwear of
the thickness 0:0025 m was 0:003 m thick. The human body was 180 cm in
uniform grid x D const with the constant time step t D const. The
numerical implementation utilizes the central differential
scheme for space
Mathematical model of the ice protection
derivatives and Crank Nicolson implicit representation of
unsteady
of athe
human
body atterm.
high temperatures
An
inhouse
code
was
developed
for
this
purposes.
Lectures on computational fluid dynamics
of surrounding medium
17.5
Results
17.5.1
The calculations were performed for five elements of the human body shown
in Fig.17.1. Thermodynamic properties of the body layers are given in the
table 17.4. Usually, the ice protection construction consists of the ice layer
separated from the surrounding medium by several layers. In the present
work we used three layers: outer clothing layer with thickness of 0:001m, isolating layer of 0:009m and the air layer of 0:003m. All layers have the initial
temperature Ti0 D 190 . The air gap between the ice and the underwear of
the thickness 0:0025 m was 0:003 m thick. The human body was 180 cm in
height with the mass of 80 kg. The full description of the input data can be
found in [41]. It was assumed that the test person does the work at a rate of
about M D 420 watts. Additional heat release due to this work was taken
into account during numerical simulations in the heat production rate f .x/
(see Eq. (17.4)).
185
heat
thermal
density
capacity, conductivity,
KJ/K
W/(mK)
kg m3
core
39.97
0.4186
1050
muscles
74.68
0.3959
1050
The stuff you'll need to make a good living
fat
17.73
0.3348
850
skin
5.07
0.0807
1000
initial
heat
temperature, production,
Deg. Celsius
W
36.89
76.24-80.00
36.28
8.30-9.00
The stuff that makes life worth living
34.53
3.57-4.00
33.62
0.66-0.7
STUDY. PLAY.
The numerical simulations are used to find the distribution of the ice thickness necessary to keep the body core (both inner an outer) temperature at less
than 36:70 10 within one hour at the temperature of surrounding medium of
500 . The numerical results were utilized to design a special protection jacket
for rescue team working under emergency conditions in the mining industry.
Since the continuous distribution of the ice protection
is difficult from the
NORWAY.
technology point of view and undesirable from ergonomics
considerations
the
YOUR IDEAL
STUDY DESTINATION.
jacket protection was designed using the discrete distribution of the ice in
form of briquettes embedded into the jacket (see WWW.STUDYINNORWAY.NO
Fig. 17.4). The further
FACEBOOK.COM/STUDYINNORWAY
simulations have been performed to prove the ability of the new designed
jacket to satisfy the protection requirements. Fig. 17.5 illustrates the temperature distribution around the human body in centers of the upper and the
lower elliptical cylinders shown in Fig. 17.1. The discrepancy between the
desirable temperature 36:70 corresponding to the continuous ice distribution
and the actual temperature corresponding to the discrete ice distribution can
Download free eBooks at bookboon.com
be
considered as acceptable. Therefore the jacket designed on the base of
numerical simulations was manufactured and
experimental
180 tested in further Click
on the ad to read more
investigations.
17.5.2
heat
capacity,
KJ/K
core
39.97
muscles
74.68
fat
17.73
skin
5.07
thermal
density
conductivity,
W/(mK)
kg m3
0.4186
1050
0.3959
1050
0.3348
850
0.0807
1000
initial
temperature,
Deg. Celsius
36.89
36.28
34.53
33.62
heat
production,
W
76.24-80.00
8.30-9.00
3.57-4.00
0.66-0.7
The numerical simulations are used to find the distribution of the ice thickness necessary to keep the body core (both inner an outer) temperature at less
than 36:70 10 within one hour at the temperature of surrounding medium of
500 . The numerical results were utilized to design a special protection jacket
for rescue team working under emergency conditions in the mining industry.
Since the continuous distribution of the ice protection is difficult from the
technology point of view and undesirable from ergonomics considerations the
jacket protection was designed using the discrete distribution of the ice in
form of briquettes embedded into the jacket (see Fig. 17.4). The further
simulations have been performed to prove the ability of the new designed
jacket to satisfy the protection requirements. Fig. 17.5 illustrates the temperature distribution around the human body in centers of the upper and the
lower elliptical cylinders shown in Fig. 17.1. The discrepancy between the
desirable temperature 36:70 corresponding to the continuous ice distribution
and the actual temperature corresponding to the discrete ice distribution can
be considered as acceptable. Therefore the jacket designed on the base of
numerical simulations was manufactured and tested in further experimental
investigations.
17.5.2
The numerical prediction and the jacket designed on basis of numerical simulations was proved in various tests. The first series of 48 measurements
was carried out in a thermal chamber. The measurements were performed
with healthy candidates at the age between 30 and 45 years, P D 168 188
186
181
Figure 17.5: Temperature distributions around the body with continuous ice
distribution and with ice briquettes. Results of numerical simulations after
60 minutes.
187
182
Figure 17.6: Development of the averaged temperature in the air gap between
the underwear and the ice protection on the human chest. Comparison between the measurement (solid line) and the numerical simulations (dotted
Figure 17.6: Development of the averaged temperature in the air gap between
line).
the underwear and the ice protection on the human chest. Comparison between the measurement (solid line) and the numerical simulations (dotted
line).
cm in height and normal weight m. The body surface A can be calculated
from the empirical formula A D 0:24m0:425 P 0:4 [41]. The candidates weared
the jackets (see Fig. 17.4)were doing the work at a rate of about M D 420
cm in height and normal weight m. The body surface A can be calculated
watts which corresponds
to a hard work
typical
for the mining industry. The
0:425
Technical
from the empirical formula
A D 0:24mtraining
P 0:4 [41].on
The candidates weared
temperature of the surrounding medium was T 500 . The temperature has
the jackets (see Fig. 17.4)were
doingneed,
the work WHEN
at a rate of about need
M D 420
been measured in WHAT
the air gapyou
between
the underwear andyou
the ice at sevenit
watts which corresponds
to
a
hard
work
typical
for
the
mining
industry.
The
IDC Technologies
can tailor
technical
and engineering
points around theAtchest
(section 2 inweFig.
17.1).our
The
0 temperature averaged
temperature of the
surrounding
medium
was needs.
T 50
. have
The extensive
temperature has
training
workshops
suit 17.6.
your
OIL & GAS
over these points
is presented
in to
Fig.
The We
discrepancy
between the
training
technical
engineering
been measured experience
in the air ingap
between
the and
underwear
andstaff
the and
ice at seven ENGINEERING
measurement and
the
numerical
simulations
does notasexceed
14 percent afhave
trained
people 2
inin
organisations
General
points around the
chest
(section
Fig. 17.1). such
The temperature
averaged
ter 60 minutes Motors,
of the real
time.
ThisBHP
agreement
can betoconsidered
Shell,
Siemens,
and Honeywell
name a few. as quite ELECTRONICS
over these points is presented in Fig. 17.6. The discrepancy between the
satisfactory taking
the simplicity
ofeffective,
the used
modelandand
complexity of theAUTOMATION &
Ourthe
onsite
training is cost
convenient
completely
measurement and
numerical
simulations
does not
exceed
14 percent afPROCESS CONTROL
problem into account.
customisable to the technical and engineering areas you want
ter 60 minutes ofcovered.
the real
This
canhands-on
be considered
Our time.
workshops
are agreement
all comprehensive
learning as quite
experiences
with ampleof
time
given
to practical
and
satisfactory taking
the simplicity
the
used
modelsessions
and complexity
of the MECHANICAL
ENGINEERING
demonstrations.
We
communicate
well
to
ensure
that
workshop
content
problem into account.
and timing match the knowledge,
skills, and abilities of the participants.
188
INDUSTRIAL
We run onsite training all year round and hold the workshops on
your premises or a venue of your choice for your convenience.
188
DATA COMMS
ELECTRICAL
POWER
183
the underwear and the ice protection on the human chest. Comparison between the measurement (solid line) and the numerical
simulations
(dotted
Mathematical
model
of the ice protection
line).
of a human body at high temperatures
of surrounding medium
188
Figure 17.7: Test person weared overheating protection jacket (left) and
distribution of the temperature sensors on the human body (right).
The task of this study is the determination of the temperature in the body
core. It is a difficult problem since the direct measurement is impossible.
The experience gathered in physiology [42] shows that the core temperature
can reliably be determined if the temperature Tj at five characteristic points
(forehead, chest, hand, thigh and shin) is known (see Fig.17.7). The sensors
were mounted directly on the human skin with the rate of press not exceeding
0:2 0:25 Pa. The accuracy of measurements is estimated as 0:10 C . These
temperatures are summed up with weighting coefficients wj . Each temperature is nearly constant within a certain area Aj . The weighting coefficients
are calculated as the ratio of Aj to the total body surface A, i.e. wj D Aj =S.
The averaged temperature is calculated then from the formula [41]
T D 0:07T1 C 0:5T2 C 0:05T3 C 0:18T4 C 0:2T5
(17.11)
Download
eBooks
at bookboon.com
where
Tj free
sensors
readings
according to the numeration shown in Fig. 17.7.
The prediction accuracy can be sufficiently184
increased when the core temperature is calculated with account for the test person feeling and the rectal
temperature Trect al according to the formula [41]
(17.11)
17.6
17.6 Discussion
Discussion
17.6
Discussion
This chapter
presents relatively simple and efficient model of the heat trans-
This chapter presents relatively simple and efficient model of the heat transfer within the human body at high temperatures of the surrounding medium.
This
chapter
relatively
simple
and efficient
model
of the heat
transfer within
thepresents
human body
at high
temperatures
of the
surrounding
medium.
The human body cross section is represented as an ellipse with a few layfer within
thebody
human
body
at high
theellipse
surrounding
The
human
cross
section
is temperatures
represented asofan
with a medium.
few layers, modelling the internal organs. The heat generation by internal organs
The
human
body
cross
section
is
represented
as
an
ellipse
with
a
layers, modelling the internal organs. The heat generation by internalfew
organs
is taken into account. The ice protection and the clothes are modelled as
ers,taken
modelling
the internal
The heat
byare
internal
organs
is
into account.
Theorgans.
ice protection
andgeneration
the clothes
modelled
as
additional layers covering the human body. Neglecting the heat transfer in
is taken into
account.
Thetheicehuman
protection
the clothes
additional
layers
covering
body.and
Neglecting
the are
heatmodelled
transfer as
in
vertical and circumferential directions the problem is reduced to the solution
additional
layers
covering
the
human
body.
Neglecting
the
heat
transfer
in
vertical and circumferential directions the problem is reduced to the solution
of the one dimensional heat conduction equation with variable heat diffuvertical
anddimensional
circumferential
the equation
problem is
reduced
to the
solution
of
the one
heatdirections
conduction
with
variable
heat
diffusivity. The simulations were performed using the finite differential method.
of the one
heat performed
conductionusing
equation
with differential
variable heat
diffusivity.
The dimensional
simulations were
the finite
method.
Results of simulation were used for calculation of the ice layer thickness necsivity.
The
simulations
were
performed
using
the
finite
differential
method.
Results of simulation were used for calculation of the ice layer thickness necessary to prevent the body overheating and to keep the temperature of the
Results
simulation
were
for calculation
the the
ice layer
thicknessofnecessary
tooffree
prevent
bodyused
overheating
and toofkeep
temperature
the
Download
eBooksthe
at bookboon.com
essary to prevent the body overheating and to keep the temperature of the
190 185
190
190
ZZZVWXG\DWWXGHOIWQO
5DQNHGWKLQWKHZRUOG
7+(67HFKQRORJ\UDQNLQJ
$OPRVW\HDUVRISUREOHPVROYLQJ
H[SHULHQFH
([FHOOHQW6SRUWV &XOWXUHIDFLOLWLHV
&KHFNRXWZKDWDQGKRZZHWHDFKDW
ZZZRFZWXGHOIWQO
191
191
186
Chapter 18
Chapter 18
Chapter 18
CFD Design of cloth for
CFD
Design
clothatfor
protection
of of
divers
low
CFD Design
of
cloth for
protection
of divers
low
temperatures
under at
current
protection of divers
at
low
temperatures
under current
conditions
temperatures under current
conditions
conditions
Study was performed for the model Tin Man presented in Fig. 18.1. Inner
thermodynamics in the body was not modelled. The temperature of the huStudy
was was
performed
fortothe
Tinand
Man
presented
in the
Fig.temperature
18.1. Inner
man body
assumed
be model
constant
equal
to 330 at
Study
was
performed
for
the
model
Tin
Man
presented
in
Fig.
18.1.the
Inner
thermodynamics
in the
body of
was10not
modelled.
The temperature
huof
the surrounding
medium
degrees
of Celsius.
Thus, theofDirichlet
0
thermodynamics
in the body
was
not modelled.
The
temperature
of the human
body
was
assumed
to
be
constant
and
equal
to
33
at
the
temperature
boundary conditions were enforced for the temperature
whereas the heat
man body
was assumed to beofconstant
and of
equal
to 330 Thus,
at the the
temperature
of
surrounding
10 degrees
Celsius.
Dirichlet
fluxthe
from
the body medium
was calculated.
Heat exchange
by shivering,
breathing
of
the
surrounding
medium
of
10
degrees
of
Celsius.
Thus,
the
Dirichlet
boundary
conditions
were enforced
forexchange
the temperature
whereas
and radiation
was neglected.
The heat
between the
body the
and heat
surboundary
conditions
were
enforced for
the
temperature
whereas breathing
the heat
flux
from
the
body
was
calculated.
Heat
exchange
by
shivering,
rounding medium is mostly determined by the convection due to wind. The
flux radiation
from the bodyneglected.
was calculated.
Heat
exchange
by shivering,
and
The heat
exchange
between
the bodybreathing
and surincompressiblewas
flow was calculated
using
steady RANS
(Reynolds
Averaged
and
radiation
was
neglected.
The
heat
exchange
between
the
body
and The
surrounding
medium
is mostly using
determined
by!the
convection
to wind.
Navier- Stokes
Equations)
the k
SST
model. due
Temperature
was
rounding medium
is was
mostly
determined
the convection
due to wind.
The
incompressible
calculated
usingbysteady
(Reynolds
Averaged
considered as aflow
passive
scalar determined
from RANS
the temperature
transport
incompressible
flow
was calculated
using
steady
(Reynolds
Averaged
NavierEquations)
using the
k
! utilized
SSTRANS
model.
Temperature
was
equation.Stokes
The framework
OpenFOAM
was
for the
numerical soluNavier- Stokes
Equations)
using
the k !from
SSTthe
model.
Temperature
was
considered
as
a
passive
scalar
determined
temperature
transport
tion of this problem.
considered The
as a passive scalar
determined
the for
temperature
transport
equation.
OpenFOAM
wasfrom
utilized
the numerical
First, CFD was framework
validated for
the case of the
air
flow. Distribution
of the soluheat
equation.
The
framework
OpenFOAM
was
utilized
for
the
numerical
solution
of
this
problem.
transfer coefficient along the Tin Man body is shown in Fig. 18.1, left. The
tion ofCFD
this was
problem.
First,
validated
for the
caseseparation
of the air flow.
Distribution
thevelocheat
minimal heat
flux
takes place
in the
area with
reduced of
flow
First,
CFD
was
validated
for
the
case
of
the
air
flow.
Distribution
of
the
transfer
along
Tinand
Manhead,
bodyasiswell
shown
in the
Fig.stagnation
18.1, left. heat
The
ities, i.e.coefficient
behind the
bodythe
arms
as in
area
transfer coefficient
alongplace
the Tin
Man
body is area
shown
in Fig.
18.1,flow
left.velocThe
minimal
heat
flux
takes
in
the
separation
with
reduced
in the front part of the body. Figure 18.1 illustrates the integral coefficient
minimal
heat
fluxthe
takes
place
in the
separation
areaaswith
reduced
flow velocities,
behind
body
arms
and
as well
the
stagnation
of thei.e.
heat
transfer
hc obtained
by head,
integration
overinthe
whole
body area
surities,
i.e.
behind
the
body
arms
and
head,
as
well
as
in
the
stagnation
area
in
the
front
part
of
the
body.
Figure
18.1
illustrates
the
integral
coefficient
face. Results of authors marked by crosses are compared with experimental
in the front
part
of the body.
Figure
18.1 illustrates
coefficient
of
transfer
obtained
by integration
overthe
theintegral
whole
body
surdata ofheat
de Dear
et al.hc(triangles)
and
different calculations.
Shadowed
grey
of
theResults
heat transfer
hc marked
obtainedbybycrosses
integration
over thewith
whole
body surface.
of scattering
authors
compared
experimental
area shows
the
of data otbainedare
using
various approaches.
Big
face.
Results
of
authors
marked
by
crosses
are
compared
with
experimental
data of de Dear et al. (triangles) and different calculations. Shadowed grey
data shows
of de Dear
et al. (triangles)
different
calculations.
Shadowed grey
area
the scattering
of dataand
otbained
using
various approaches.
Big
193
area shows the scattering of data otbained using various approaches. Big
193
193
Figure 18.1: Left: Heat transfer coefficient at air speed of 1m=s. Right:
Whole body convective heat transfer coefficient hc from various published
works. The figure is taken from [2]. Blue crosses
187 show results of the present
work.
Figure
Whole
works.
work.
188
2:5
4:5
6:5
5:27 106
7:74 106
1:01 105
4:46
6:33
8:08Design of cloth for protection of divers at
CFD
195
189
Chapter 19
Chapter 19
Chapter 19
CFD application for design of
CFD
application
for from
design
of
cloth
for
protection
low
CFD application for design of
cloth
for protection
temperatures
under from
wind low
cloth for protection
from
low
temperatures
under wind
conditions.
Influence
of the
temperatures under wind
conditions.
of the
wind
on theInfluence
cloth deformation
conditions.
Influence
of the
wind
on the
cloth from
deformation
and
heat
transfer
the
wind on the cloth deformation
and
heat transfer from the
body.
and heat transfer from the
body.
body.
Lebedeva and Brink [44] have shown the influence of wind on heat transfer
due to cloth deformation caused by wind induced pressures. Experimental
Lebedeva
Brink [44]
havetype
shown
thetunnel
influence
windtest
on heat
transfer
study was and
performed
in open
wind
withof
closed
section
up to
Lebedeva
anddeformation
Brink [44] have
shown
the
influence
of
wind
on
heat
transfer
due
to
cloth
caused
by
wind
induced
pressures.
Experimental
5
Reynolds number of 3:6 10 based on the air speed and diameter of cloth
due towas
cloth
deformation
caused
by wind
induced
Experimental
study
in open
type
withpressures.
closed
section
up to
packages
inperformed
form of cylinders.
Thewind
aim tunnel
of the present
worktest
is to
get similar
5 type wind tunnel with closed test section up to
study
was
performed
in
open
Reynolds number
3:6 10body
based
on the air speed and diameter of cloth
estimations
for realofhuman
form.
Reynolds in
number
3:6 105 based
on the
airpresent
speed and
cloth
packages
form ofofcylinders.
The aim
of the
workdiameter
is to getofsimilar
packages in for
form
of human
cylinders.
The
aim of the present work is to get similar
estimations
real
body
form.
estimations for real human body form.
19.1
19.1
19.1
The
conducted
in the
wind
of square
the Chair
of section
Ocean
type experiment
has the testwas
section
of lengths
of 2:8
m tunnel
with the
cross
The experiment
was
conducted
in
the
wind
tunnel
of
the
Chair
of
Ocean
Engineering
at
the
Rostock
University.
The
wind
tunnel
of
the
G
o
ttingen
of 1:4 1:4 m2 . The model of the height of 700 mm was manufactured by
Engineering
thesection
Rostock
The
of the
Gottingen
type
has theat
test
of University.
lengths of 2:8
m wind
with tunnel
the square
cross
section
2
type
themtest
section
of lengths
of 2:8ofm700
withmm
thewas
square
cross section
of
1:4has
1:4
. The
model
of the height
manufactured
by
197
of 1:4 1:4 m2 . The model of the height of 700 mm was manufactured by
197
197
Figure 19.1: Human body model in wind tunnel of the Rostock university
(left). Positions of measurement points (right).
Download free eBooks at bookboon.com
downsizing the real body with the scale factor of 0:39. Within the open test
section the size of the air jet with uniform velocity distribution is estimated
as one meter at least. Since it is one and half times as large as the model
height, the influence of the jet boundaries on flow around the body can be
considered as negligible. Two measurement series were performed. In the
first one the pressure was measured at nine points distributed at three cross
sections z D 0:329; 0:418 and 0:476 (see Fig. 19.1 and 19.3) using inclined
manometers which are very accurate for low pressures measurements. Due
to high inertia of inclined manometers the unsteady pressure oscillations
are not captured. Position of measurement points is shown in Fig. 19.1
(points 1; 10 ; 100 , 5; 50 ; 500 and 9; 90 ; 900 ). The measurements were performed
with the air speed of 10m=s at 14 degrees of Celsius. Reynolds number
based on the maximum transversal body size is around 1:33 105 . In the
second series the measurements were performed at nine points 1; :::; 9 shown
in Fig. 19.3(right) only at z D 0:418 (waist). This more detailed investigation
was carried out for two air speeds of 10 and 15 m=s.
19.2
Numerical simulations were performed using the commercial software package STAR CCM+ for the air speed of 10 m/s. The boundary layer on the
windward side is supposed to be in transitional state. Due to strong separation the flow on the leeward side of the model and in the wake can be
considered as a turbulent one. The computational domain has the cross section of 2:1 2:1 m2 . A length of the domain is 2:1 m in front of the human
198 it. The grid consists of 1:2 106
body (0:7 m height) and 4:2 m behind
polyhedron cells with maximal y C value of 40. On the model sides the y C
value is varied between 20 and 25. To exclude ambiguity connected to the
flow character, simulations were performed both for pure turbulent and pure
laminar flows with and without roughness.
19.3
Comparison with measurements for the first series is presented in Fig. 19.2.
The results of measurements and simulations agree well in the front part of
the body and in the separation area. Both simulations and measurements
predict the increase of the pressure coefficient in the section z D 0:476 m
at angle 90 degrees caused by the stagnation effect of the shoulder. On
the side of the model in cross sections z D 0:329 and 0:418 the agreement is
not satisfactory. To clarify this problem, the more detailed second measureDownload
free was
eBooks
at bookboon.com
ment
series
performed.
All results obtained at z D 0:418 are presented
in Fig. 19.3. The big difference takes place around the angle of approx. 90
191
degrees in the area of the separation (point 5 in Fig. 19.1). The experiment
doesnt predict the strong under pressure region on the side of the model.
The minimum experimental pressure coefficient is around 1 whereas the
polyhedron cells with maximal y C value of 40. On the model sides the y C
value is varied between 20 and 25. To exclude ambiguity connected to the
application
forpure
design of cloth for
flow character, simulations were performed both for pureCFD
turbulent
and
protection
from
low
temperatures
under wind
laminar flows with and without roughness.
19.3
Comparison with measurements for the first series is presented in Fig. 19.2.
The results of measurements and simulations agree well in the front part of
the body and in the separation area. Both simulations and measurements
predict the increase of the pressure coefficient in the section z D 0:476 m
at angle 90 degrees caused by the stagnation effect of the shoulder. On
the side of the model in cross sections z D 0:329 and 0:418 the agreement is
not satisfactory. To clarify this problem, the more detailed second measurement series was performed. All results obtained at z D 0:418 are presented
in Fig. 19.3. The big difference takes place around the angle of approx. 90
degrees in the area of the separation (point 5 in Fig. 19.1). The experiment
doesnt predict the strong under pressure region on the side of the model.
The minimum experimental pressure coefficient is around 1 whereas the
simulation predicts 2:2 both for laminar and turbulent cases. The difference between the laminar and turbulent solutions is negligible. This might
be due to two facts. First, the cross section is elliptical (Fig. 19.2, left)
and the averaged position of flow separation points is fixed at the widest
axis and doesnt depend on the flow character. Second, strong unsteady
flow oscillation makes the difference between laminar and turbulent pressure
distributions negligible.
The experimental results raised many questions. Without big error, the
flow can be considered as quasi two dimensional in cross sections along the
human body. Then Cpmi n is minus three for the case of two dimensional
Study at one of Europes
cylinder flow. Due to three dimensional effects Cpmi n can be smaller, but
notleading
three times universities
smaller like in measurements. The numerical simulation
reveals no separation up to angle of 90 degrees. Within the separation zone
at angles larger than 120 degrees the numerical pressure coefficient is nearly
constant. These facts are in a good agreement with classic knowledge about
bluff bodies flows. In experiment, Cp slightly increases in the separation
area at angles larger than 90 degrees. Also, a relatively big discrepancy
199
DTU, Technical University of Denmark, is ranked as one
of the best technical universities in Europe, and offers
internationally recognised Master of Science degrees in
39 English-taught programmes.
DTU offers a unique environment where students have
hands-on access to cutting edge facilities and work
Visit us at www.dtu.dk
192
ence between the laminar and turbulent solutions is negligible. This might
be due to two facts. First, the cross section is elliptical (Fig. 19.2, left)
and the averaged position of flow separation points isCFD
fixed
at the for
widest
application
design of cloth for
low temperatures
axis and doesnt depend on the flow character.protection
Second,from
strong
unsteady under wind
Lectures
on computational
conditions.
Infl uence of
the wind on the cloth
flow oscillation
makesfluid
thedynamics
difference between laminar
and turbulent
pressure
deformation and heat transfer from the body.
distributions negligible.
The experimental results raised many questions. Without big error, the
flow can be considered as quasi two dimensional in cross sections along the
human body. Then Cpmi n is minus three for the case of two dimensional
cylinder flow. Due to three dimensional effects Cpmi n can be smaller, but
not three times smaller like in measurements. The numerical simulation
reveals no separation up to angle of 90 degrees. Within the separation zone
at angles larger than 120 degrees the numerical pressure coefficient is nearly
constant. These facts are in a good agreement with classic knowledge about
bluff bodies flows. In experiment, Cp slightly increases in the separation
area at angles larger than 90 degrees. Also, a relatively big discrepancy
199
19.4
Estimations of pressure influence on cloth thermal conductivity was performed experimentally in the Thermal laboratory of the Don State Technical
University. The cloth of 30 mm thickness was manufactured from the Saviour
with Flamestat Cotton (upper sheet), the insulation Thinsulate and Taffeta
as a lining. The cloth is used in oil industry for work at very low temperatures
under oil contamination conditions.
free eBooks
at bookboon.com
ItDownload
was assumed
that
the pressure acts only in normal direction and shear
stresses are neglected. The cloth has such a structure that it can be pressed
193
but not stretched. With the other words, the cloth has no deformation if
the pressure difference p D p pat m is negative. Bearing this in mind, the
cloth deformation calculated using numerical pressure data can be considered
Estimations of pressure influence on cloth thermal conductivity was performed experimentally in the Thermal laboratory of the Don State Technical
University. The cloth of 30 mm thickness was manufactured
from the Saviour
CFD application
for design of cloth for
with Flamestat Cotton (upper sheet), the insulation
Thinsulate
and
Taffeta under wind
protection
from low
temperatures
Lectures
on computational
dynamics
conditions.
of the wind on the cloth
as a lining.
The cloth fluid
is used
in oil industry for work
at very Infl
lowuence
temperatures
deformation
and
heat
transfer from the body.
under oil contamination conditions.
It was assumed that the pressure acts only in normal direction and shear
stresses are neglected. The cloth has such a structure that it can be pressed
but not stretched. With the other words, the cloth has no deformation if
the pressure difference p D p pat m is negative. Bearing this in mind, the
cloth deformation calculated using numerical pressure data can be considered
200
Figure 19.3: Left: Pressure distribution p on the body obtained using StarCCM+ commercial software. Contours of three cross sections at
z D 0:329; 0:418 and 0:476 m are marked by black lines. Right: Pressure
coefficient Cp distribution around the body at z D 0:418. Points position
1; ::; 9 is shown in Fig. 19.1. Grey zone is the area of unsteady pressure
coefficient oscillations in the laminar solution. Vertical lines indicate the
scattering of experimental data at points 5, 6 and 7.
as reliable one since the area of positive Cp is approximately the same in
experiments and simulations. The cloth samples were subjected to pressure
and then the thermal conductivity was measured using steady state method.
Approximation of measurement points results in the following interpolation
formula:
k D 0:22 C c1 p C c2 p 2
(19.1)
where # and r are zylindrical coordinates in a cross section. The results are
summarized in the table 19.1. At strong wind of 20 m/s or 72 km/h the
integral
of at
thermal
conductivity can be up to seventy percent in the
Downloadincrease
free eBooks
bookboon.com
chest cross section. At moderate wind of 10 m/s the maximum influence is
194
less than ten percent.
Table 19.1: Thermal conductivity factor f .z/ integrated in circumferential
chest up to four and half times for strong wind of 20 m/s. To estimate the
integral wind influence f the thermal conductivity referred to that without
CFD application for design of cloth for
wind was integrated over the body at five sections:
protection from low temperatures under wind
I dynamics
Lectures on computational fluid
conditions. Infl uence of the wind on the cloth
f .z/ D k.wi nd 0/=k.wi nd deformation
D 0/r d# and heat transfer
(19.2)from the body.
where # and r are zylindrical coordinates in a cross section. The results are
summarized in the table 19.1. At strong wind of 20 m/s or 72 km/h the
integral increase of thermal conductivity can be up to seventy percent in the
chest cross section. At moderate wind of 10 m/s the maximum influence is
less than ten percent.
Table 19.1: Thermal conductivity factor f .z/ integrated in circumferential
direction
Body part
Wind 10 m/s 20 m/s
Chest (z D 0:476 m)
1:096
1:729
Waist (z D 0:418 m)
1:025
1:199
Hip (z D 0:329 m)
1:085
1:668
Upper leg (z D 0:25 m)
1:025
1:194
Lower leg (z D 0:12 m)
1:018
1:141
202
203
195
The comfort conditions of drivers and passengers is one of the most important
research topics in car industry. In this chapter we present a fragment of the
The comfort conditions of drivers and passengers is one of the most important
research
conducted
in this
field at
the chair of modelling
and
simulation
of
The
comfort
conditions
of drivers
is one
of the
most
important
research
topics
in car industry.
Inand
thispassengers
chapter we
present
a fragment
of the
the
University of
Rostock.
The calculations
presented
were carried
research
in car
presentbelow
a fragment
of the
research topics
conducted
in industry.
this field In
at this
the chapter
chair of we
modelling
and
simulation
of
out
by
Stefan
Knochenhauer.
A
compact
car
was
chosen
as
the
object
of
research
conducted
in this field
the chair ofpresented
modellingbelow
and simulation
of
the
University
of Rostock.
Theatcalculations
were carried
investigations.
The
cabin geometry
is represented in Fig.
20.1.
The
inlet
the
of
Rostock.
TheA calculations
below
were
carried
out University
by Stefan Knochenhauer.
compact carpresented
was chosen
as the
object
of
of
the
air
stream
coming from A
thecompact
ventilation
channel
is marked
by green
out
by
Stefan
Knochenhauer.
car
was
chosen
as
the
object
of
investigations. The cabin geometry is represented in Fig. 20.1. The inlet
color
whereas
the
air
outlet
by
red
one.
The
grid
with
6.5
million
of
cells
investigations.
The
cabin from
geometry
is represented
in Fig.
20.1. The
inlet
of
the air stream
coming
the ventilation
channel
is marked
by green
is
shown
instream
Fig. 20.2.
The
mathematical
model
is based
on the by
URANS
of
the
air
coming
from
the
ventilation
channel
is
marked
green
color whereas the air outlet by red one. The grid with 6.5 million of cells
equation
(12.17)
equation
(1.30).
The turbulent
color
whereas
theand
air temperature
outlet
redtransport
one. The
gridiswith
6.5on
million
of cells
is
shown
in Fig.
20.2.
The by
mathematical
model
based
the URANS
character
of the
temperature
transport
is taken
intoisaccount
by introduction
is
shown
in
Fig.
20.2.
The
mathematical
model
based
on
the
URANS
equation (12.17) and temperature transport equation (1.30). The turbulent
of
the additionaland
turbulent
heat conduction.
It means
thatThe
instead
of
equation
temperature
transport
equation
(1.30).
turbulent
character (12.17)
of the temperature
transport
is taken
into account
by introduction
in
the
equation
(1.30)
we
use
the
sum
C
,
where
is
the
turbulent
t by introduction
character
of the temperature
transport
is taken tinto
account
of
the additional
turbulent heat
conduction.
It means
that instead of
heat
conduction
coefficient.
is
expressed
through
the
turbulent
kinematic
t
of
the
additional
turbulent
heat
conduction.
It
means
instead
of
in the equation (1.30) we use the sum C t , where tthat
is the
turbulent
viscosity
and
the
turbulent
Prandtl
number
P
r
t
t
in theconduction
equation coefficient.
(1.30) we use
the
sum C
t , where
t is the kinematic
turbulent
heat
t is
expressed
through
the turbulent
heat
conduction
coefficient.
is
expressed
through
the
turbulent
kinematic
viscosity t and the turbulentt Prandtl
number
P rt
t
t
D number P r t
(20.1)
viscosity t and the turbulent Prandtl
c
Pr t
tp
t
D t
ist computed
The turbulent kinematic viscosityc
from the URANS closure(20.1)
modP
r
p D
t
(20.1)
els (see Chapters 12, 13 and 14)cwhereas
the
turbulent
Prandtl
number
is
P
r
t
The turbulent kinematic viscosity ispcomputed
from the URANS closure modassumed
to bekinematic
constant P
r t 0:7.
The URANS
closure
modelclosure
used inmodthis
The
turbulent
computed
the URANS
els (see
Chapters
12, 13 viscosity
and
14) iswhereas
thefrom
turbulent
Prandtl
number is
work
is
the
k
!
SST
model
[45].
Numerical
simulations
were
performed
els (see Chapters
12, 13Pand
the turbulent
number
is
assumed
to be constant
r t 14)
0:7.whereas
The URANS
closure Prandtl
model used
in this
using thetofinite
volume method
(seeThe
Chapter
6). closure
Boundary
conditions
are
assumed
be
constant
P
r
0:7.
URANS
model
used
in
this
t
work is the k ! SST model
[45]. Numerical simulations were performed
given is
in the
the ktable
20.1.
work
!
SST
model
simulations
performed
using the finite volume method[45].
(see Numerical
Chapter 6).
Boundarywere
conditions
are
A
few
selected
results
of
calculations
are
illustrated
in
Fig.
20.3.
The
distriusing in
thethe
finite
volume
given
table
20.1. method (see Chapter 6). Boundary conditions are
butions
of
velocity
and temperature inside of the car cabin are strongly
given
theairtable
20.1.of
A few in
selected
results
calculations are illustrated in Fig. 20.3. The distriA few selected
results of
calculations
are inside
illustrated
Fig.cabin
20.3.are
The
distributions
of air velocity
and
temperature
of theincar
strongly
205 inside of the car cabin are strongly
butions of air velocity and temperature
For almost 60 years Maastricht School of Management has been enhancing the management capacity
205
of professionals and organizations around the world through state-of-the-art management education.
205
Our broad range of Open Enrollment Executive Programs offers you a unique interactive, stimulating and
multicultural learning experience.
Be prepared for tomorrows management challenges and apply today.
For more information, visit www.msm.nl or contact us at +31 43 38 70 808 or via admissions@msm.nl
18-08-11 15:13
196
(20.1)
The turbulent kinematic viscosity is computed from the URANS closure models (see Chapters 12, 13 and 14) whereas the turbulent Prandtl number is
assumed to be constant P r t 0:7. The URANS closure model used in this
work is the k ! SST model [45]. Numerical simulations were performed
using the finite volume method (see Chapter 6). Boundary conditions are
given in the table 20.1.
A few selected results of calculations are illustrated in Fig. 20.3. The distributions of air velocity and temperature inside of the car cabin are strongly
inhomogeneous. In regions of strong velocity the convective heat transfer
dominates whereas in stagnation areas
205 the heat exchange is mostly due to
the heat conduction.
Figure 20.2: Grids with 6.5 million of cells generated with snappyHexMesh.
197
Surface
physical quantity
Condition
ux uy uz
Velocity
grad p D 0
Pressure
Inlet
T D konst:
Temperature
k D konst:
! D konst:
grad U D 0
Velocity
pD0
Pressure
Outlet
grad T D 0
Temperature
grad k D 0
grad ! D 0
U D0
Velocity
grad p D 0
Pressure
T D konst: or
grad T D 0 or
coupled boundary condition
198
Bibliography
Bibliography
Bibliography
[1] Tsinober A. An informal introduction to turbulence. Kluwer Academic
Publisher, 2004.
[1] Tsinober A. An informal introduction to turbulence. Kluwer Academic
[2] Publisher,
Novieto D.2004.
Zhang Y. and Yingchun J. Human envoronmental heat transfer simulation with cfd- the advances and challenges.building simulation
[2] Novieto
D. Zhang
Y. andEleventh
Yingchun
J. Human envoronmental
heat trans2009. pages
21622168.
International
IBPSA Conference,
Glasfer
simulation
the 2009
advances
gow,
Scotland,with
Julycfd2730,
2009.and challenges.building simulation
2009. pages 21622168. Eleventh International IBPSA Conference, GlasScotland,
2009 2009.
[3] gow,
Ferziger
J. and July
Peric2730,
M. Computational
Methods for Fluid Dynamics.
Springer, 2002.
[3] Ferziger J. and Peric M. Computational Methods for Fluid Dynamics.
[4] Springer,
Seibold B.2002.
A compact and fast matlab code solving the incompressible navier-stokes equations on rectangular domains. Applied Mathemat[4] ics,
Seibold
B. A compact
and offast
matlab code
solving the incompressMassachusetts
Institute
Technology,
www-math.mit.edu/
seibold,
ible
navier-stokes
equations
on
rectangular
domains.
Applied
Mathemat2008.
ics, Massachusetts Institute of Technology, www-math.mit.edu/ seibold,
[5] 2008.
Dornseifer T. Griebel M. and Neunhoeffer T. Numerical simulation in
fluid dynamics: A practical introduction. Society for Industrial and
[5] Applied
Dornseifer
T. Griebel M.
and Neunhoeffer
T. 1998.
Numerical simulation in
Mathematics,
Philadelphia,
PA, USA,
fluid dynamics: A practical introduction. Society for Industrial and
Applied Mathematics,
PA, M.
USA,
1998.simulations of static
[6] Phillips
A.B. Turnock Philadelphia,
S.R. and Furlong
Urans
drift and dynamic manouveres of the KVLCC2 tanker. pages F63F68.
[6] Proceedings
Phillips A.B.ofTurnock
S.R. and
Furlong 2008.
M. Urans simulations of static
the SIMMAN
workshop,
drift and dynamic manouveres of the KVLCC2 tanker. pages F63F68.
Proceedings
of the
SIMMAN Methods.
workshop,Springer,
2008. 2010. 390p.
[7] Liseikin
V. Grid
Generation
[7] Liseikin
V. Grid
Generation
Springer,
2010. 390p.grid genera[8]
Warsi Z.U.A
Thompson
I.F. Methods.
and Mastin
C.W. Numerical
tion. Foundations and applications, 1997.
[8] Warsi Z.U.A Thompson I.F. and Mastin C.W. Numerical grid generaChallenge
the way
we run 1997.
tion. Foundations
and applications,
[9] Novikov
E. Generalized
dynamics
of three-dimensional singularities
(vortons). Sov. JETP, 84(3):975 981, 1984.
[9] Novikov E. Generalized dynamics of three-dimensional singularities
(vortons).
Sov. evolution
JETP, 84(3):975
981, vortex.
1984. Comm. in Mathematical
[10] Chorin
A. The
of a turbulent
Physics, 83:517 535, 1982.
[10] Chorin A. The evolution of a turbulent vortex. Comm. in Mathematical
Physics,
83:517 535,
1982.
EXPERIENCE
THE
POWER
209 OF
FULL ENGAGEMENT
209
RUN FASTER.
RUN LONGER..
RUN EASIER
1349906_A6_4+0.indd 1
Download
free eBooks at bookboon.com
22-08-2014 12:56:57
199
946.
Proc.
equation. pages 941 946. Proc. of
of
Flow
Flow Phenomena,
Phenomena, 2005.
2005.
models
models based
based on
on the
the length-scale
length-scale
4th
Int.
Symp.
Turb.
4th Int. Symp. Turb. and
and Shear
Shear
[25] Davidson
L. and
and Billson
Billson M.
M. Hybrid
Hybrid les-rans
les-rans using
using synthesized
synthesized turbulent
turbulent
[25]
Davidson L.
fluctuations
for
forcing
in
the
interface
region.
Int.
J.
of
Heat
and
fluctuations for forcing in the interface region. Int. J. of Heat and Fluid,
Fluid,
27:1028
1042,
2006.
27:1028 1042, 2006.
[26]
[26] Abe
Abe K.
K. and
and Miyata
Miyata Y.
Y. An
An investigation
investigation of
of hybrid
hybrid les/rans
les/rans models
models for
for
predicting
flow
fields
with
separation.
pages
1153
1158.
Proc.
of
4th
predicting flow fields with separation. pages 1153 1158. Proc. of 4th
Int.
Int. Symp.
Symp. Turb.
Turb. and
and Shear
Shear Flow
Flow Phenomena,
Phenomena, 2005.
2005.
[27] Pitsch
H. Schlueter
Schlueter J.
J. U.
U. and
and Moin
Moin P.
P. onsistent
onsistent boundary
boundary conditions
conditions for
for
[27]
Pitsch H.
integrated
les/rans
simulations:
Les
outflow
conditions.
volume
3121,
integrated les/rans simulations: Les outflow conditions. volume 3121,
pages
pages 11 8.
8. AIAA,
AIAA, 2002.
2002.
[28]
[28] Ducros
Ducros F.
F. Benerafa
Benerafa Y.
Y. and
and Sagaut
Sagaut P.
P. Rans/les
Rans/les coupling
coupling using
using aa forcing
forcing
term approach.
approach. pages
pages 1141
1141 1146.
1146. 4th
4th Int.
Int. Symp.
Symp. Turb.
Turb. and
and Shear
Shear Flow
Flow
term
Phenomena,
2005.
Phenomena, 2005.
[29] Shchukin
Kornev N.,E.
of
[29]
Shchukin
E.Taranov
KornevA.,
N.,Shchukin
TaranovE.,
A. and
and Kleinsorge
KleinsorgeL.L.
L.Development
Development
[29]
Kornev
N.,
Taranov
A.
and
Kleinsorge
Development
of
hybrid
urans-les
methods
forsimulations
flow simulations
simulations
in area.
the ship
ship
stern
area.
hybrid
urans-les
methods
for flow
in the ship
Oceanstern
Engineering,
of
hybrid
urans-les
methods
for
flow
in
the
area.
Ocean
Engineering,
38(16):1831 1838,
1838, 2011.
2011.
Ocean
Engineering,
38(16):1831
38(16):18311838,
20111.
[30]
[30] Durbin
Durbin P.
P. Near-wall
Near-wall turbulence
turbulence closure
closure modeling
modeling without
without damping
damping funcfunctions.
tions. J.
J. Theor.
Theor. Comput.
Comput. Fluid
Fluid Dyn,
Dyn, 1991.
1991.
[31]
[31] Werner
Werner H.
H. and
and Wengle
Wengle H.
H. Large-eddy
Large-eddy simulation
simulation of
of turbulent
turbulent flow
flow over
over
and
around
a
cube
in
a
plate
channel.
8th
Symposium
on
Turbulent
and around a cube in a plate channel. 8th Symposium on Turbulent
Shear
Shear Flows,
Flows, 1991.
1991.
[32]
[32] Fureby
Fureby C.
C. Huuva
Huuva T.
T. Alin
Alin N.,
N.,
capabilities
of
des
and
les
capabilities of des and les for
for
Research,
Research, 54:184
54:184 196,
196, 2010.
2010.
Bensow
Bensow R.
R.
submarines
submarines
and
and Svennberg
Svennberg U.
U.
at
straight
course.
at straight course.
Current
Current
J.
J. Ship
Ship
[33]
[33] Guminer
Guminer P.K.
P.K. Thermoregulation
Thermoregulation studying
studying in
in hygiene
hygiene and
and work
work physiolphysiology.
Medgiz,
Moscow,
page
147,
1962.
ogy. Medgiz, Moscow, page 147, 1962.
[34]
Novieto
D.
Y.
Download
free eBooks
at bookboon.com
[34]
Novieto
D. Zhang
Zhang
Y. and
and Yingchun
Yingchun J.
J. Human
Human environmental
environmental heat
heat transtransfer
simulation
with
cfd
the
advances
and
challenges.
pages
2162
2168.
fer simulation with cfd the advances and
201 challenges. pages 2162 2168.
Eleventh
Eleventh International
International IBPSA
IBPSA Conference,
Conference, Glasgow,
Glasgow, Scotland,
Scotland, 2009.
2009.
[35]
[35] Cherunova
Cherunova I.V.
I.V. New
New technologies
technologies of
of heat
heat protection
protection wear
wear design,
design, (in
(in
russian).
russian). Technologies
Technologies of
of textile
textile industry,
industry, pages
pages 51
51 54,
54, 2009.
2009.
212
202
Index
Index
"-Equation, 141
continuity equation, 17
convective acceleration, 24
"-Equation, 143
a-posteriori
test,implicit,
160 49
alternating
direction
a-priori test,
160
amplification
of the
vorticity, 81
alternating
direction
implicit, 49
artificial
viscosity,
33
amplification of the vorticity, 84
artificial viscosity, 33
autocorrelation function, 102
density, 25
backward
scheme, 28
Biot-Savart
law,difference
80
Biot-Savart
block
structured law,
grids, 83
74
central
difference
scheme,
28
central
difference
scheme,
28
collocated grid, 41
collocated grid, 41
constant of Karman, 99
constant of Karman, 97
continuity equation, 17
convective acceleration, 23
correlation function, 102
Courant Friedrich Levy criterion, 37
Crank-Nicolson scheme, 31
cross stress, 158, 159
density, 25
Detached Eddy Simulation, 163, 164
Direct Numerical Simulation, 123
Dirichlet condition, 26, 48
dissipation, 139
dissipation range, 120
dissipation rate, 117
Dynamic Smagorinsky Model, 154
heat conduction
coefficient,
26
energy production,
117
heat conduction
equation,
25,
ensemble averaged quantity, 10026
heat flux,
25 of Reynolds, 87
experiment
heat sources, 25
homogeneous turbulence, 102
hybrid URANS-LES methods, 163
implicit scheme, 31
inertial subrange, 119, 120
inner energy balance equation, 25
integral length, 102
isotropic turbulence, 102
k - " Model, 141
k-Equation, 141
Kelvin Helmholtz instability, 91
Kolmogorov first similarity hypothesis, 118
213
GET THERE FASTER
DISCOVER
OUR WORLD
Some people know precisely where they want to go. Others seek the adventure of
discovering uncharted territory. Whatever you want your professional journey to be,
youll nd what youre looking for at Oliver Wyman.
203
Index
explicit scheme, 31
Lax-Wendroff scheme, 53
Leonard stress, 157, 158
folding, 83
limiters function, 40
Fourier law, 25
local acceleration, 24
logarithmic region, 98
friction velocity, 94
Friedman-Keller problem, 138
Galilean invariance, 157
heat conduction coefficient, 26
heat conduction equation, 25, 26
heat flux, 25
heat sources, 25
homogeneous turbulence, 101
hybrid URANS-LES methods, 160
implicit scheme, 31
inertial subrange, 119, 120
inner energy balance equation, 25
integral length, 102
isotropic turbulence, 103
k - " Model, 139
k-Equation, 139
Kelvin Helmholtz instability, 89
Kolmogorov first similarity hypothesis, 117
Kolmogorov hypothesis of local isotropy, 118
Kolmogorov law, 120
pseudodissipation, 139
kurtosis, 110
lambda structures, 92
laminar, 81
Richardson poem, 86
204
Index
time-advancing, 40
shear stress, 21
turbulence, 81
SIMPLE method, 65
SIMPLEC algorithm, 73
skewness, 110
Smagorinsky model, 152
Spalart Allmares (SA model), 132
specific heat capacity, 25
splitting according to physical processes, 51
staggered grid, 41
streaks, 92, 94
structure function, 109
structured grids, 74
TVD schemes, 41
unstructured grids, 75
upwind difference scheme, 28
Van Driest constant, 98
viscous sublayer, 96
vortex cascado, 116
vortex reconnection, 86
vorticity, 82
surface forces, 19
wake region, 98
205