Lectures On Computational Fluid Dynamics

Prof. Dr.-Ing. habil. Nikolai Kornev & Prof. Dr.-Ing. habil.
Irina Cherunova
Lectures on computational fluid dynamics

Applications to human thermodynamics
2
Download free eBooks at bookboon.com
Lectures on computational fluid dynamics: Applications to human thermodynamics
1st edition
© 2015 Prof. Dr.-Ing. habil. Nikolai Kornev & Prof. Dr.-Ing. habil. Irina Cherunova &
bookboon.com
ISBN 978-87-403-0815-0
3
Lectures on computational fluid dynamics Contents
Contents
List of Tables 10
List of Figures 11
Preface 15
I Introduction into computational methods for solution of transport equations 16
1 Main equations of the Computational Heat and Mass Transfer 17

1.1 Fluid mechanics equations 17
1.2 Heat conduction equation 25
2 Finite difference method 27

2.1 One dimensional case 27
2.2 Two dimensional case 30
2.3 Time derivatives. Explicit versus implicit 31
2.4 Exercises 31
�e Graduate Programme
I joined MITAS because for Engineers and Geoscientists
I wanted real responsibili� www.discovermitas.com
Maersk.com/Mitas �e G
I joined MITAS because for Engine
I wanted real responsibili� Ma
Month 16
I was a construction Mo
supervisor ina const
I was
the North Sea super
advising and the No
Real work he
helping foremen advis
International
al opportunities
Internationa
�ree wo
work
or placements ssolve problems
Real work he
helping fo
International
Internationaal opportunities
�ree wo
work
or placements ssolve pr
4 Click on the ad to read more

3 Stability and artificial viscosity of numerical methods 33

3.1 Artificial viscosity 33
3.2 Stability. Courant Friedrich Levy criterion (CFL) 35
3.3 Exercise 37
4 Simple explicit time advance scheme for solution of the Navier

Stokes Equation 38
4.1 Theory 38
4.2 Mixed schemes 39
4.3 Staggered grid 41
ıun un
4.4 Approximation of ıxi j j 44
n
ı ıui
4.5 Approximation of ıxj ıxj 47
4.6 Calculation of the r.h.s. for the Poisson equation (4.6) 47
4.7 Solution of the Poisson equation (4.6) 47
4.8 Update the velocity field 47
4.9 Boundary conditions for the velocities 48
4.10 Calculation of the vorticity 48
www.job.oticon.dk

5 Splitting schemes for solution of multidimensional problems 49

5.1 Splitting in spatial directions. Alternating direction implicit (ADI) approach 49
5.2 Splitting according to physical processes. Fractional step methods 51
5.3 Increase of the accuracy of time derivatives approximation using the
Lax-Wendroff scheme 53
6 Finite Volume Method 55

6.1 Transformation of the Navier-Stokes Equations in the Finite Volume Method 55
6.2 Sample 56
6.3 Explicit scheme 59
6.4 Implicit scheme 61
6.5 Iterative procedure for implicit scheme 61
6.6 Pressure correction method 64
6.7 SIMPLE method 65
7 Overview of pressure correction methods 70

7.1 SIMPLE algorithm 70
7.2 PISO algorithm 70
7.3 SIMPLEC algorithm 73
How will people travel in the future, and

how will goods be transported? What re-
sources will we use, and how many will
we need? The passenger and freight traf-
fic sector is developing rapidly, and we
provide the impetus for innovation and
movement. We develop components and
systems for internal combustion engines
that operate more cleanly and more ef-
ficiently than ever before. We are also
pushing forward technologies that are
bringing hybrid vehicles and alternative
drives into a new dimension – for private,
corporate, and public use. The challeng-
es are great. We deliver the solutions and
offer challenging jobs.
www.schaeffler.com/careers

8 Computational grids 74
8.1 Grid types 74
8.2 Overset or Chimera grids 75
8.3 Morphing grids 75
II Mathematical modelling of turbulent flows 77
9 Physics of turbulence 78
9.1 Definition of the turbulence 78
9.2 Vortex dynamics 78
9.3 Experimental observations 86
10 Basic definitions of the statistical theory of turbulence 99

10.1 Reynolds averaging 99
10.2 Isotropic and homogeneous turbulence 100
10.3 Correlation function. Integral length 100
10.4 Structure functions 109
In the past 5 years we have drilled around
95,000 km
—that’s more than twice around the world.
Who are we?

We are the world’s leading provider of reservoir characterization,
drilling, production, and processing technologies to the oil and
gas industry.
Who are we looking for?

We offer countless opportunities in the following domains:
n Operations
n Research, Engineering, and Manufacturing
n Geoscience and Petrotechnical
n Commercial and Business
We’re looking for high-energy, self-motivated graduates

with vision and integrity to join our team. What will you be?
careers.slb.com

11 Kolmogorov theory K41 114

11.1 Physical background 114
11.2 Dissipation rate 116
11.3 Kolmogorov hypotheses 117
11.4 Three different scale ranges of turbulent flow 120
11.5 Classification of methods for calculation of turbulent flows 123
11.6 Limitation of K-41. Kolmogorov theory K-62 123
12 Reynolds Averaged Navier Stokes Equation (RANS) 128
13 Reynolds Stress Model (RSM) 134

13.1 Derivation of the RSMEquations 134
14 Equations of the k - " Model 139

14.1 Derivation of the k-Equation 139
14.2 Derivation of the " -Equation 141
14.3 Method of wall functions 144
15 Large Eddy Simulation (LES) 147

15.1 LES filtering 147
15.2 LES equations 149
15.3 Smagorinskymodel 150
15.4 Model of Germano ( Dynamic Smagorinsky Model) 153
15.5 Scale similarity models 155
15.6 Mixed similarity models 156
15.7 A-posteriori and a-priori tests 158
16 Hybrid URANS-LES methods 160

16.1 Introduction 160
16.2 Detached Eddy Simulation (DES) 161
16.3 Hybrid model based on integral length as parameter switching between
LES and URANS 164
16.4 Estimations of the resolution necessary for a pure LES on the
example of ship flow 167
8
III CFD applications to human thermodynamics 171
17 Mathematical model of the ice protection of a human body at high

temperatures of surrounding medium 172
17.1 Designations 172
17.2 Introduction 173
17.3 Human body and ice protectionmodels 174
17.4 Mathematicalmodel 175
17.5 Results 180
17.6 Discussion 185
18 CFD Design of cloth for protection of divers at low temperatures under

current conditions 187
19 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 Wind tunnel measurements of pressure distribution 190
19.2 Numerical simulations of pressure distribution and comparison
with measurements 191
19.3 Comparison of CFD results withmeasurements 192
19.4 Change of thermal conductivity caused by wind induced pressures 193
20 Simulation of human comfort conditions in car cabins 196
Bibliography 199
Index 203
9
Lectures on computational fluid dynamics List of Tables
List of Tables
4.1 Limiters function for TVD schemes 40
6.1 nE uE and ui at different sides. x-equation 57
6.2 Velocities at different sides. x-equation 58
6.3 Convection flux. x-equation 58
6.4 nE uE and ui at different sides. y-equation 58
6.5 Velocities at different sides. y-equation 58
6.6 Convection flux. y-equation 59
15.1 Properties of large and small scale motions 149
15.2 Advantages and disadvantages of the Smagorinsky model 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
17.1 Sizes of the body used in simulations 175
17.2 Radii of the elliptical cross sections used in simulations 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
19.1 Thermal conductivity factor f (z) integrated in circumferential direction 195
20.1 Boundary conditions 198
10
Lectures on computational fluid dynamics List of Figures
List of Figures
1.1 Body and surface forces acting on the liquid element 18
1.2 Forces acting on the liquid element 19
1.3 Stresses acting on the liquid cube with sizes a 21
2.1 One dimensional case 27
2.2 A sample of non uniform grid around the profile 30
4.1 Sample of the collocated grid 42
4.2 Checkerboard pressure solution on the collocated grid 42
4.3 Grid points of staggered grid 43
6.1 Staggered arrangement of finite volumes 55
6.2 SIMPLE algorithm 65
6.3 Control volume used for the pressure correction equation 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 Illustration of structured grid disadvantage 74
9.1 Vorticity and vortices 80
9.2 Tornado 81
9.3 Vortices in two-dimensional and three dimensional cases 81
9.4 Velocities induced by vortices. Three dimensional curvilinear vortices
induce self induced velocities 82
9.5 Illustration of the vortex folding 83
9.6 Scenario of vortex amplication 84
9.7 Scenario of vortex reconnection 84
9.8 Sample of the vortex reconnection of tip vortices behind an airplane 86
9.9 Most outstanding results in turbulence research according to [1] 87
9.10 Sketch of the Reynolds experiment 88
9.11 Development of instability during the laminar- turbulent transition in the
circular pipe (taken from [1]) 88
9.12 Development of instability in the jet (taken from [1]) 90
9.13 Development of instability in the free jet 90
9.14 Development of instability in the free jet 90
9.15 Vortex structures in a free jet in a far field 91
9.16 Vortex structures in a free jet with acoustic impact 91
9.17 Vortex structures in a confined jet mixer flow 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 Scenario of laminar turbulent transition in the boundary layer on a flat plate 94
11
9.20 Streaks visualized by hydrogen bubbles in the boundary layer on a flat plate 94
9.21 Conceptual model of the organization of the turbulence close to the wall
proposed by Adrian et al. (2000) 95
9.22 Vertical distribution of the velocity ux at three dierent time instants in
boundary layer 95
9.23 Illustration of the Prandtl derivation 97
9.24 Structure of the velocity distribution in the turbulent boundary layer.
U C D ux =u 98
10.1 Autocorrelation function coecient for scalar fluctuation at three different
points A, B and C across the jet mixer 101
10.2 Distribution of the integral length of the scalar field along the jet
mixer centerline 103
10.3 Autocorrelation functions in free jet flow 104
10.4 Illustrations of velocities used in calculations of the longitudinal
f and transversal g autocorrelations 105
10.5 Illustration of the autocorrelation functions f and g and Taylor microscales 106
10.6 Kurtosis of the structure function for the concentration of the scalar field
obtained in the jet mixer 110
11.1 Andrey Kolmogorov was a mathematician, preeminent in the 20th century,
who advanced various scientific fields (among them probability theory, topology,
intuitionistic logic, turbulence, classical mechanics and computational
complexity) 115
11.2 Illustration of the vortex cascado 116
11.3 Turbulent vortices revealed in DNS calculations performed by
Isazawa et al. (2007) 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
the experimental estimatin of Miller and Dimotakis (1991)
" D 48.Ud3 =d /..x x0 /=d /4 119
11.5 Three typical scale ranges in the turbulent flow at high Reynolds numbers 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, 3-
dissipation range 121
11.7 Experimental confirmation of the Kolmogorov law. The compensated energy
spectrum for different flows. 122
11.8 Experimental confirmation of the Kolmogorov law for the concentration
fluctuations in the jet mixer. Measurements of the LTT Rostock 122
11.9 Three main methods of turbulent flows modelling 123
11.10 Vortex structures resolved by different models 124
12
11.11 Power of the structure function. Experiments versus prediction of

Kolmogorov and Obukhov 125
15.1 Different filtering functions used in LES 148
15.2 Illustrations for derivation of the scale similarity model 155
16.1 Zones of the Detached Eddy Simulation 161
16.2 Squires K.D., Detached-eddy simulation: current status and perspectives 164
16.3 Squires K.D., Detached-eddy simulation: current status and perspectives 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 The cell parameters 170
17.1 Sketch of the human body used in simulations. A- heart, B- liver, C - kidney 173
17.2 Horizontal cross section of the human body represented as an ellipse
with five layers: 1- inner core, 2- outer core, 3- muscles, 4- fat, 5- skin 176
17.3 Ice protection construction. 1- polyurethane foam, 2- ice briquette, 3- human
body, 4-special overheating protection clothes, 5- polyurethane net (air layer) 176
17.4 Overheating protection jacket designed on the base of simulations 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
of the temperature sensors on the human body (right) 184
17.8 Inner human body temperature versus time 186
Left: Heat transfer coecient at air speed of 1m/s. Right: Whole body
18.1
convective heat transfer coefficient hc from various published works.
The figure is taken from [2]. Blue crosses show results of the present work 188
18.2 Heat flux from the diver depending on the cloth contamination 189
19.1 Human body model in wind tunnel of the Rostock university (left).
Positions of measurement points (right) 190
19.2 Contrours of torso (left) and pressure coefficient Cp distribution around
the body at three different altitudes z = 0:329, 0:418 and 0:476m 193
13
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 coecient Cp
distribution around the body at z = 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 7 194
19.4 Change of thermal conductivity due to pressure induced by wind of 10 m/s
(left) and 20 m/s (right). Thermal conductivity without wind is 0:22 W/mK 195
20.1 Sketch of the car cabin studied numerically 197
20.2 Grids with 6.5 million of cells generated with snappyHexMesh 197
20.3 Results of numerical simulations 198
14
Lectures on computational fluid dynamics 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 Tech-
nical University, Russia. Each of lecture courses contains about 14 lectures.
The lecture course Compuational heat and mass transfer was written pro-
ceeding from the idea to present the complex material as easy as possible.
We considered derivation of numerical methods, particularly of the finite vol-
ume 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 physi-
cal 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.
Linköping University
– Innovative, well ranked, European
Interested in Engineering and its various
branches? Kick-start your career with an
English-taught master’s degree.
liu.se/master
Click here!

Part I
Introduction into
computational methods for
solution of transport equations
16
15 at bookboon.com
Download free eBooks
Main equations of the Computational
Chapter 1
Lectures on computational fluid dynamics Heat and Mass Transfer
Chapter
Chapter 1
1
Main equations of the
Main
Main equations
Computational of
of the
equations Heat
theand Mass
Computational
Computational Heat
Transfer Heat and
and Mass
Mass
Transfer
Transfer
1.1 Fluid mechanics equations
1.1
1.1.1 Fluid mechanics
Continuity equationequations
1.1 Fluid mechanics equations
We consider
1.1.1 the case of uniform
Continuity equation density distribution D const. The con-
1.1.1 equation
tinuity Continuity equation
has the following physical meaning: The amount of liquid
We consider
flowing thevolume
into the case ofUuniform
with thedensity
surfacedistribution D const.
S is equal to the amountThe con-
of liquid
We consider the
tinuity equation
flowing case of uniform
has the following
out. Mathematically density
it can physical distribution
meaning:
be expressed D const. The
The amount of liquid
in form: con-
tinuity equation
flowing has theUfollowing
into the volume with the physical meaning: The amount of liquid
Z surface S is equal to the amount of liquid
flowing
flowing out. Mathematically it can surface
into the volume U with the S is equal
be expressed to the amount of liquid
in form:
uE nE ds D 0
flowing out. Mathematically it can be expressed in form: (1.1)
Z
Z
S
uE nE ds D 0 (1.1)
uE nE ds D components
Expressing the scalar product uE nE through 0 (1.1)
S
Z S
Expressing the scalar product uE nE through components
ux cos.nx/
Expressing the scalar Cu
product throughCcomponents
uEynE cos.ny/ uz cos.nz/ ds D 0:
Z
Z
S
u cos.nx/ C u cos.ny/ C u cos.nz/ds D 0:
x y z
and using the Gauss theorem
ux cos.nx/ Cwe get
uy cos.ny/ C uz cos.nz/ ds D 0:
S
S Z
@ux we get
@uy @uz
and using the Gauss theorem C get C
we dU D 0
Z @x @y @z
Z @ux C @uy C @uz d U D 0
U
@xxUCis@u
@u @yy @uz
Since the integration volume C @z the
arbitrary, D 0 is zero only if
d Uintegral
U @x @y @z
U
@uxU is@u
Since the integration volume @uz the integral is zero only if
C arbitrary,
y
17
C D0 (1.2)
Since the integration volume
@xU is @y
arbitrary,
@z the integral is zero only if
17
In the tensor form the continuity equation reads:
17
@ui
D0 (1.3)
@xi
1.1.2 Classification of forces acting in a fluid

The inner forces acting in a fluid are subdivided into the body forces and
surface forces (Fig. 1.1).
17
1.1.2 Classification of forces acting in a fluid
The inner forces acting in a fluid are subdivided into the
Mainbody forces
equations andComputational
of the
Lectures
surfaceonforces
computational fluid dynamics
(Fig. 1.1). Heat and Mass Transfer
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
FE D lim (1.4)
U !0 U
D ms 218
3
which has the unit kgs 2m m 1
kg m3
. Typical body forces are gravitational,
electrostatic or electromagnetic forces. For instance, we have the following
relations for the gravitational forces:
fE D gU kE (1.5)
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:
gU kE
FE D lim . / D g kE (1.6)
U !0 U
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
PEn
pEn D lim (1.7)
18
S !0 S
with the unit kgs 2m m12 D ms

kg Download free eBooks at bookboon.com
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
element. Usually they are shear and normal stresses. The strength ofHeat
surface
and Mass Transfer
forces is determined as
PEn
pEn D lim (1.7)
S !0 S
with the unit kgs 2m m12 D ms

kg
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:
Z
RE D pEn dS
Z
S
(1.8)
E D
M .Er pEn /dS
S
1.1.2.3 Properties of surface forces

Let us consider a liquid element in form of the tetrahedron (Fig. 1.2).
Its motion is described by the 2nd law of Newton:
d uE
U D U FE C pEn S pEx Sx pEy Sy pEz Sz (1.9)
dt
19
Figure 1.2: Forces acting on the liquid element.
Dividing r.h.s and l.h.s. by the surface of inclined face S results in:

U d uE Sx Sy Sz
FE D pEn pEx pEy pEz (1.10)
S dt S S S
Let us find the limit of (1.10) at S ! 0:
U S19x
lim D 0; lim D cos.nx/; (1.11)
S !0 SDownload
S S at bookboon.com
!0eBooks
free
Sy Sz
lim D cos.ny/; lim D cos.nz/ (1.12)
Figure 1.2: Forces acting on the liquid element. Heat and Mass Transfer
Dividing r.h.s and l.h.s. by the surface of inclined face S results in:

U d uE Sx Sy Sz
FE D pEn pEx pEy pEz (1.10)
S dt S S S
Let us find the limit of (1.10) at S ! 0:
U Sx
lim D 0; lim D cos.nx/; (1.11)
S !0 S S !0 S
Sy Sz
lim 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 :
pEn D pEx cos.nx/ C pEy cos.ny/ C pEz cos.nz/ (1.13)

Let us write the surface forces through components:
E xz
pEx D Eipxx C jExy C k
E yz
pEy D Eiyx C jEpyy C k
E zz
pEz D Eizx C jEzy C kp
20

678'<)25<2850$67(5©6'(*5((
&KDOPHUV8QLYHUVLW\RI7HFKQRORJ\FRQGXFWVUHVHDUFKDQGHGXFDWLRQLQHQJLQHHU
LQJDQGQDWXUDOVFLHQFHVDUFKLWHFWXUHWHFKQRORJ\UHODWHGPDWKHPDWLFDOVFLHQFHV
DQGQDXWLFDOVFLHQFHV%HKLQGDOOWKDW&KDOPHUVDFFRPSOLVKHVWKHDLPSHUVLVWV
IRUFRQWULEXWLQJWRDVXVWDLQDEOHIXWXUH¤ERWKQDWLRQDOO\DQGJOREDOO\
9LVLWXVRQ&KDOPHUVVHRU1H[W6WRS&KDOPHUVRQIDFHERRN

Lectures on computational fluid dynamics Heat and Mass Transfer
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)
Figure 1.3: Stresses acting on the liquid cube with sizes a.
The stress matrix is symmetric and contains 6 unknown elements:

0 1
pxx xy xz
@ xy pyy yz A (1.15)
xz yz pzz
1.1.3 Navier Stokes Equations

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 Z Z
d uE
d U D FE d U C pEn dS (1.16)
dt
U U S
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
Z
S S
@pEx @pEy @pEz
D C C dU
@x @y @z
U
The second law (1.16) takes the form: 21
Z Z
Z eBooks
Download free at bookboon.com
d uE @pEx @pEy @pEz
d U D FE d U C C C dU
dt @x @y @z

pEn dS D pEx cos.nx/ C pEy cos.ny/ C pEz cos.nz/ dS
Z
S S
@pEx @pEy @pEz Main equations of the Computational
D fluid dynamics
Lectures on computational C C dU Heat and Mass Transfer
@x @y @z
U
The second law (1.16) takes the form:
Z Z Z
d uE E @pEx @pEy @pEz
d U D F d U C C C dU
dt @x @y @z
Z
U U U
d uE @ E
p @ E
p @ E
p
FE
x y z
C C dU D 0
dt @x @y @z
U
Since the volume d U is arbitrary, the l.h.s. in the last formulae is zero only
if:

d uE 1 @ E
p @ E
p @ E
p
D FE C
x y z
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
0 1 0 1
pxx xy xz p 0 0
@ xy pyy yz A D @ 0 p 0 A C 2Sij (1.18)
xz yz pzz 0 0 p
where p is the pressure,

@ux 1 @ux @uy 1 @ux @uz
S11 D Sxx D I S12 D Sxy D C I S13 D Sxz D C
@x 2 @y @x 2 @z @x

@uy 1 @uy @uz
S21 D S12 ; S22 D Syy D ; S23 D Syz D C
@y 2 @z @y
@uz
S31 D S13 ; S32 D S23 ; S33 D Szz D
@z
The liquids obeying (1.18) are referred to as the Newtonian liquids.
The normal stresses can be expressed through the pressure p:
22
@ux @uy @uz
pxx D p C 2 ; pyy D p C 2 ; pzz D p C 2
@x @y @z
The sum of three normal stresses doesn’t depend on the choice of the coor-
dinate system and is equal to the pressure taken with sign minus:
pxx C pyy C pzz

D p (1.19)
3
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. Sub-
stitution of the Newton hypothesis (1.18) into (1.17) gives (using the first
equation as a sample):
22

dux @ @ueBooks
Download free x at@bookboon.com
@uy @ux
D Fx C p C 2 C C
dt @x @x @y @x @y

The sum of three normal stresses doesn’t depend on the choice of the coor-
dinate system and is equal to the pressure taken with sign minus:
pxx C pyy C pzz Heat and Mass Transfer
D p (1.19)
3
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. Sub-
stitution 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
@p @ ux @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 direc-
tions, one can obtain the following equation, referred to as the Navier-Stokes
equation:
d uE 1
D FE rp C E
u (1.20)
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 @E
u @E
u @E
u @E
u
D C ux C uy C uz
Welcome to
dt „ƒ‚…
@t @x @y @z
our world „ ƒ‚ …
local acceleration
of teaching! convective acceleration
Innovation, flat hierarchies

and open-minded professors 23
Study in Sweden -
close collaboration
with future employers
Mälardalen university collaborates with
many employers such as ABB, volvo and
Ericsson
TAKE THE
debajyoti nag
Sweden, and particularly
MDH, has a very impres-
sive reputation in the field
RIGHT TRACK
of Embedded Systems Re-
search, and the course
design is very close to the
industry requirements.
He’ll tell you all about it and
answer your questions at
mdustudent.com
give your career a headstart at mälardalen university
www.mdh.se

tions, one can obtain the following equation, referred to as the Navier-Stokes
equation:
d uE
Lectures on computational fluid dynamics 1 Heat and Mass Transfer
D FE rp C E
u (1.20)
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 @E
u @E
u @E
u @E
u
D C ux C uy C uz
dt „ƒ‚…
@t
„
@x
ƒ‚
@y @z
…
local acceleration convective acceleration
The local acceleration is due to the23change of the velocity in time. The

convective acceleration is due to particle motion in a nonuniform velocity
field. The Navier-Stokes Equation in tensor form is:

@ui @ui 1 @p @ @
C uj D Fi C ui (1.21)
@t @xj @xi @xj @xj
Using the continuity equation (1.3) the convective term can be written in the
conservative form:

@ui @
uj D ui uj (1.22)
@xj @xj
Finally, the Navier Stokes in the tensor form is:

@ui @ 1 @p @ @
C .ui uj / D Fi C ui (1.23)
@t @xj @xi @xj @xj
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
@p @2 uz
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
singularity. The Clay Mathematics Institute has called the Navier–Stokes
existence and smoothness problems one of the seven most important open
24
24
@x @y
@p @2 uy
D Fy C 2
@y @y
@p @2 uz Heat and Mass Transfer
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
singularity. The Clay Mathematics Institute has called the Navier–Stokes
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 Heat conduction equation

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)

Integrating (1.24) over the surface S we obtain:
Z
q ndS (1.25)
S

Z Z
q.x; t/ n.x/dS D r q.x; t/d U (1.26)
S U
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 Z Z
@
cp T .x; t/d U D r q.x; t/d U C f .x; t/d U (1.27)
U @t U U
Here f is the heat sources within the volume U .

Fourier has proposed the following relation between the local heat flux and
temperature difference, known as the Fourier law:
q.x; t/ D rT .x; t/ (1.28)

where is the heat conduction coefficient.
Substitution of the Fourier law (1.28) into the inner energy balance equa-
tion (1.27) results in
Z Z
@
cp T .x; t/ r .rT .x; t// d U D f .x; t/d U (1.29)
! @t U
Since the volume U is arbitrary, (1.29) is reduced to
25 25
tion (1.27) results in
Z Z Main equations of the Computational

@
Lectures on computational
cp T .x;fluid
t/ dynamics
r .rT .x; t// d U D f .x; t/d U Heat and Mass Transfer
(1.29)
! @t U
Since the volume U is arbitrary, (1.29) is reduced to

@
cp T .x; t/ r .rT
25 .x; t// D f .x; t/ (1.30)
@t
The equation (1.30) is the heat conduction equation. The heat conduction
coefficient for anisotropic materials is the tensor
0 1
11 12 13
D @ 12 22 23 A (1.31)
13 23 33
The following boundary conditions are applied for the heat conduction equa-
tion (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)
LIFE SCIENCE IN UMEÅ, SWEDEN

– YOUR CHOICE!
• 32 000 students • world class research • top class teachers
• modern campus • ranked nr 1 in Sweden by international students
• study in English
– Bachelor’s programme in Life Science

– Master’s programme in Chemistry
– Master’s programme in Molecular Biology
Download
brochure
here!
26

Chapter 2
Lectures on computational fluid dynamics Finite dierence method
Chapter
Chapter 2
2
Finite difference method
Finite
Finite difference
difference method
method
2.1 One dimensional case
2.1 One dimensional
Let us consider the finite difference case
method for the one dimensional case.
2.1
Let '.x/One dimensional
is the function defined in thecase
range Œ0; a along the x axis. The
section Œ0; a is subdivided in a set of points xifor
Let us consider the finite difference method . Forthethe
one dimensionaldistri-
homogeneous case.
Chapter 2
Let us
Let '.x/
bution
Let '.x/
consider
xi is
D the
is the
the finite
.i function
function
difference
1/I i Ddefined
defined in
method
1; N , inD the range
a=.N
the
for
1/Œ0;
range Œ0;
the one
a Fig.
(see
a along
dimensional
the x axis. case.
along2.1).
the x axis.
The
The
section Œ0; a is subdivided in a set of points xi . For the homogeneous distri-
section xŒ0;
bution a is subdivided in a set of points xi . For
i D .i 1/I i D 1; N , D a=.N 1/ (see the
Fig.homogeneous
2.1). distri-
bution xi D .i 1/I i D 1; N , D a=.N 1/ (see Fig. 2.1).
Finite difference method
2.1 One dimensional case

Let us consider the finite difference method for the one dimensional case.
Let '.x/ is the function defined in the range Œ0; a along the x axis. The
section Œ0; a is subdivided in a set of points xi . For the homogeneous distri-
bution xi D .i 1/I i D 1; N , D a=.N 1/ (see Fig. 2.1).
Figure 2.1: One dimensional case.

@' 1
Let us approximate the Figure 2.1: One
derivative @x
dimensional
. The Taylor case.
series of the function '
Figure
at points xi 1 and xi C1 are: 2.1: One dimensional case.
@' 1
Let us approximate the derivative @'@x 1
. The Taylor 2 series
of the function '
Letpoints
at us approximate
xi 1 and xthe C1
derivative
are: @' . The
1 Taylor
@ ' series of the function '
'i 1xiD
at points xi 1 and i x
'are: @x
C x 2 ::: (2.1)
i C1 @x i 2 @x 2
i
@' 1 2 @2 '
1 ' i 1 D 'i x
We use the partial derivative although
@' C @22' only
121 xdepends
the function ::: one variable (2.1)
D @'
@x C 2 @ '2 i on
@x (2.1)
'i C1 D 'i C x @x C 2 x @x 22 C :::
' i 1 'i x i x 2
::: (2.2)
@x ii 2 @x i
i
1
We use the partial derivativealthough
the
27 function depends only on one variable
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
i 27
Scheme (BDS): Figure 2.1: One27 dimensional case.

@' 1 @' 1
Let us approximate the derivative D 'i .'The C O x
i 1 Taylor series of the function (2.3)
'
@x i x @x
at points xi 1 and xi C1 are:
Expressing the derivative @' from (2.2) weget
@x @' 1 2 @2 'the Forward Difference
'i 1 D 'i x i C x ::: (2.1)
Scheme (FDS): @x i 2 @x 2 i
1 although
We use the partial derivative
the function depends only on one variable
@' 1
D 'i C1 'i C O x (2.4)
@x i x
27
Accuracy of both schemes is of the first order. Subtracting (2.1) from (2.2)
we get the Central Difference Scheme (CDS) 27
free eBooksat bookboon.com
Download
@' 1 2
D 'i C1 'i 1 C O x (2.5)
@x i 2x
1 @'
D
'i 'i 1 C O x (2.3)
i x @x

Expressing the derivative @'@x
from (2.2) we get the Forward Difference
i Finite dierence method
Scheme (FDS):

@' 1
D 'i C1 'i C O x (2.4)
@x i x
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 'i C1 'i 1 C O x (2.5)
@x i 2x
which is of the second order accuracy.
For the approximation of derivatives ui @'
@x
where ui is the flow velocity
i
one uses the Upwind Difference Scheme (UDS):
(
@' BDS, if u > 0
D (2.6)
@x i FDS, if u < 0
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
within the section Œxi 1 ; xi C1 .
Without loss of generality we assume xi 1 D 0. The coefficient c can be
obtained from the condition:
'.0/ D 'i 1 D c
Other two coefficients a and b are determined from the conditions:
'i D ax 2 C bx C 'i 1

'i C1 D a4x 2 C b2x C 'i 1
28
'i C1 2'i C 'i 1
aD
2x 2
'i C1 C 4'i 3'i 1
bD
2x
The first derivative using CDS is then

@' 'i C1 'i 1
D 2a C b D
@x i 2x
the second derivative:
2
@ ' 'i C1 2'i C 'i 1
D 2a D
@x 2 i x 2
If the polynomial of the 3rd order '.x/ D ax 3 C bx 2 C cx C d is applied, we
get:

@' 1 3
D 2'i C1 C 3'i 6'i 1 C 'i 2 C O x (2.7)
@x i 6x
for the Backward Difference Scheme,
28

@' 1 3
D 'i C2 C 6'i C1 3'i 2'i 1 C O x (2.8)
@x i 6x
If the polynomial of the 3rd order '.x/ D ax 3 C bx 2 C cx C d is applied, we
get:

Lectures on@' 1 fluid dynamics
computational 3 Finite dierence method
D 2'i C1 C 3'i 6'i 1 C 'i 2 C O x (2.7)
@x i 6x
for the Backward Difference Scheme,

@' 1 3
D 'i C2 C 6'i C1 3'i 2'i 1 C O x (2.8)
@x i 6x
for the Forward Difference Scheme and

@' 1 4
D 'i C2 C 8'i C1 8'i 1 C 'i 2 C O x (2.9)
@x i 12x
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 ' 1 4
D 'i C2 C 16'i C1 30'i C 16'i 1 'i 2 C O x
@x 2 i 12x 2
(2.11)
29
Scholarships
Open your mind to

new opportunities
With 31,000 students, Linnaeus University is
one of the larger universities in Sweden. We
are a modern university, known for our strong
international profile. Every year more than Bachelor programmes in
1,600 international students from all over the Business & Economics | Computer Science/IT |
world choose to enjoy the friendly atmosphere Design | Mathematics
and active student life at Linnaeus University.
Master programmes in
Welcome to join us! Business & Economics | Behavioural Sciences | Computer
Science/IT | Cultural Studies & Social Sciences | Design |
Mathematics | Natural Sciences | Technology & Engineering
Summer Academy courses

Lectures on computational fluid dynamics Finite dierence method
for the polynomial of the fourth order. The formula (2.10) can also be ob-
tained 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
D (2.13)
@x i C1=2 x

@' 'i 'i 1
D (2.14)
@x i 1=2 x
we obtain (2.10).
2.2 Two dimensional case

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.
Figure 2.2: A sample of non uniform grid around the profile.
The function ' at a point .xi ; yj / is 'ij . The CDS approximation of the
derivative on x at this point is:

@' 'i C1j 'i 1j
D 30
@x ij 2x
whereas on y is:

@' 'ij C1 'ij 1
D
@y ij 2y
30
2.3 Time derivatives.
Download freeExplicit versus implicit
eBooks at bookboon.com
Let the unsteady partial differential equation is written in the form:

@'
Lectures on computational fluid dynamics 'i C1j 'i 1j Finite dierence method
D
@x ij 2x
whereas on y is:

@' 'ij C1 'ij 1
D
@y ij 2y
2.3 Time derivatives. Explicit versus implicit

Let the unsteady partial differential equation is written in the form:
@g
D G.g; t/ (2.15)
@t
The solution is known at the time instant n. The task is to find the solution
at n C 1 time instant. Using forward difference scheme we get:
g nC1 D g n C G.g; t/t (2.16)

Taking the r.h.s. of (2.15) from the n th time slice we obtain:
g nC1 D g n C G.g n ; t/t (2.17)

The scheme (2.17) is the so called explicit scheme (simple Euler approach).
Taking the r.h.s. of (2.15) from the n C 1 th time slice we obtain:
g nC1 D g n C G.g nC1 ; t/t (2.18)

The scheme (2.18) is the implicit scheme. The r.h.s. side of (2.18) depends on
the solution g nC1 . With the other words, the solution at the time slice n C 1,
g nC1 can not be expressed explicitly through the solutions known the from
previous time slices 1; 2; ::; n for nonlinear dependence G.g; t/.
Mix between explicit and implicit schemes is called the Crank-Nicolson Scheme:
1
g nC1 D g n C .G.g n ; t/ C G.g nC1 ; t//t
2
2.4 Exercises
1. Using the CDS find the derivative

@ @'
.x/ D ::: (2.19)
@x @x i
2. Using the CDS approximate the mixed derivative

31

@2 ' @ @'
D (2.20)
@x@y ij @x @y ij
3. Write the program on the language C to solve the following partial

differential equation:
@' @2 '
C ˛ 2 D f .x/
@x @x
with the following boundary conditions:
31
@' free eBooks at bookboon.com
Download
.x D 0/ D C1
@x
@'
Lectures on computational fluid dynamics @2 ' Finite dierence method
C ˛ 2 D f .x/
@x @x
@'
.x D 0/ D C1
@x
'.x D 0/ D C2
Use the central difference scheme.

@' @2 '
˛ C 2 D f .x; t/
@t @x
@'
.x D 0/ D C1
@x
'.x D 0/ D C2
and initial condition '.x; 0/ D F .x/.

Use the explicit method and the central difference scheme for spatial
derivatives.
32

Chapter 3
Stability and articial viscosity of numerical methods
Chapter
Chapter 3
Stability 3and artificial viscosity
Stability
of and
numerical
Stability artificial
artificial viscosity
and methods viscosity
of
of numerical
numerical methods
methods
3.1 Artificial viscosity
3.1us consider
Let Artificial
the genericviscosity
linear equation:
3.1 Artificial viscosity
Let us consider the generic linear @ equation: @
Let us consider the generic linear C u D0
@t equation: @x
@ @
@ C u @ D 0
@t C u @x D 0
The numerical upwind scheme @t (UDS) @x is:
8̂ u n u n
i xscheme
The numerical<upwind i1
(UDS)u > 0is:
TheinC1
numerical
in upwind scheme (UDS) is:
D 8̂ u n u n
t
8̂
:̂
< u
in
n u
uiC1
i1
n n u>0
i u
x i1 i
inC1 in < x u> <0
n D
x
inC1
ti
We consider
t :̂ the
Donly n
uiC1 case
uinu > 0:
:̂ uiC1 n
uin u<0
nC1x x n
u<0
We consider only the i i u > 0:u in u in1
case
We consider only thet case uD > 0: x
u>0 (3.1)
nC1 n
Taylor expansionsinC1 of the function
in inn t/uin time
u .x; n
in1 and space gives
i D u u u>0 (3.1)
t i D ˇ i x i 1 ˇ u>0 (3.1)
t ˇn x
@ .x; @ 2
t 2 ˇn
Taylor expansions of nC1 the function
n
Dfunction
i C .x; ˇ t C t/ in time andˇ Cspace gives
Taylor expansions of i the ˇ
@t ˇin t/ in@ttime
2 ˇ :::
2 ˇin space gives
and (3.2)
ˇ@ ˇˇ @2 t 2 ˇˇ
n ˇˇn t C 2 ˇn 2 ˇˇn C
inC1 ˇ @ ˇt ˇˇ 2 :::
n
D C
n nC1nD in C ˇ@t ˇˇ t C @t 2 ˇ 2 x
@ @ (3.2)
i D i i 1 C ˇi x C @x i CC
2 ˇi2
::: ::: (3.2)
@x ˇˇ@t @t 22 ˇˇin1
i (3.3)
in1i
@ ˇˇˇn @22 ˇˇˇn x 22
n
in Dand n
i 1 C @into ˇ x Cresults @ ˇ in x C :::
C @x ˇˇˇi 1(3.1) x C @x 22 ˇˇˇi 1
Substitution of (3.2) (3.3)
i D in1(3.3) 2 C ::: (3.3)
@xˇn i 1 ˇ @x i 1 2
@ ˇˇ ˇˇ t
n
@233
ˇ C D
@t i @t 2 ˇi 2
ˇ 33 ˇn
u @ ˇˇn 33 @2 ˇˇ x 2
D C x C
@x ˇi 1 @x 2 ˇi 1 2
(3.4)
x
The derivatives at i 1 th point can be expressed through these at i th

point:
ˇ ˇ ˇn ˇn
@ ˇˇn @ ˇˇn @2 ˇˇ @3 ˇˇ x 2
D x C :::;
@x ˇi 1 @x ˇi @x 2 ˇi @x 3 ˇi 2
ˇn ˇn ˇn (3.5)
@2 ˇˇ @2 ˇˇ @3 ˇˇ
D x :::
@x 2 ˇi 1 @x 2 ˇi @x 3 ˇi 33
The expressions (3.5) are then used in

Download free(3.4)
eBooks at bookboon.com
ˇ ˇn
@ ˇn @2 ˇ t
ˇ C ˇ D
@t ˇi @t 2 ˇi 2
ˇ ˇn
u @ ˇˇn @2 ˇˇ x 2
D C x C
@x ˇi1 2ˇ
(3.4)
x dynamics
Lectures on computational fluid @x
Stability 2 articial viscosity of numerical methods
i 1 and
The derivatives at i 1 th point can be expressed through these at i th

point:
ˇ ˇ ˇn ˇn
@ ˇˇn @ ˇˇn @2 ˇˇ @3 ˇˇ x 2
D x C :::;
@x ˇi1 @x ˇi @x 2 ˇi @x 3 ˇi 2
ˇn ˇn ˇn (3.5)
@2 ˇˇ @2 ˇˇ @3 ˇˇ
D x :::
@x 2 ˇi1 @x 2 ˇi @x 3 ˇi
The expressions (3.5) are then used in (3.4)

ˇ ˇn
@ ˇˇn @2 ˇˇ t
C 2ˇ D
@t ˇi @t i 2
ˇ ˇn 2 ˇn ˇn 2
u @ ˇˇn @2 ˇˇ @ ˇˇ @3 ˇˇ x
D x x C C ::: (3.6)
@x ˇi @x 2 ˇi @x 2 ˇi @x 3 ˇi
x
x 2
Finally we have:
ˇ ˇn
@ ˇˇn @2 ˇˇ t
C 2ˇ D
@t ˇi @t i 2
ˇ ˇn
u @ ˇˇn @2 ˇˇ x 2
D C :::
@x ˇi @x 2 ˇi 2
x (3.7)
x
Let us consider the left hand side of the equation (3.7).

Differentiating @
@t
@
D u @x on time results in

@2 @ @ 2
2@
D u D Cu (3.8)
@t 2 @x @t @x 2
@2
This allows one to find the derivative @@t22 :
This allows one to find the derivative @t 2 :
ˇn ˇ
@22 ˇˇn 2 ˇn
2 @2 ˇn
@ 2 ˇ D u2 @ 2 ˇ (3.9)
D34u @x 2 ˇi
@t 2 ˇii
@t (3.9)
@x i
Using the last result the equation (3.7) is rewritten in form:

Using the last result the equation (3.7) is rewritten in form:
ˇ ˇ ˇn ˇ
@ ˇˇnn u @ ˇˇnn @22 ˇˇn x 22 u 2 ˇn
2 @2 ˇn t
@ ˇ D u x @ ˇ C @ 2 ˇ x u u2 @ 2 ˇ t C ::: (3.10)
@t D x x @x ˇi C @x 2 ˇi 2 x u @x 2 ˇi 2 C :::
@t ˇii
(3.10)
x @x i @x i 2 x @x i 2
Finally we have
Finally we have
! !
@ @ ux ut @2
@ D u @ C 1 2
@ C :::
@t D u @x
@t @x
C ux
2
2
1 ut
x
x
@x 2
@x 2
C :::
Compare now with the original equation:

@ @34
@ D u @
@t D u @x at bookboon.com
Download
@t free eBooks
@x
The additional term
The additional term
!
!
@ @ ux ut @2
@t
D u @x C 2
1 x @x 2
C :::
Lectures on computational fluid dynamics Stability and articial viscosity of numerical methods
@ @
D u
@t @x
The additional term
!
ux ut @2
2
1 x @x 2
describes the error of numerical approximation of derivatives in the original

@2
equation (3.1). It looks like the term describing the physical diffusion @x 2,
where is the diffusion coefficient. Therefore, the error term can be in-
terpreted as the numerical or artificial diffusion with the diffusion coeffi-
cient ux
2
.1 ut
x
/ caused by errors of equation approximation. The pres-
ence of the artificial diffusion is a serious drawback of numerical methods. It
could be minimised by increase of the resolution x ! 0.
3.2 Stability. Courant Friedrich Levy crite-

rion (CFL)
Let us consider the partial differential equation:
@ @
Cu D0 (3.11)
@t @x
35
YOUR WORK AT TOMTOM WILL
BE TOUCHED BY MILLIONS.
AROUND THE WORLD. EVERYDAY.
Join us now on www.TomTom.jobs
follow us on
#ACHIEVEMORE

which is approximated using explicit method and upwind differential scheme

at u > 0:
inC1 in n in1

D u i (3.12)
t x
It follows from (3.12):
u t n
inC1 in D . in1 /
x i
u t u t n
inC1 D in .1 /C
x x i 1
ut
Let us introduce the Courant Friedrich Levy parameter c D x
:
inC1 D in .1 c/ C cin1 (3.13)
We consider the zero initial condition. At the time instant n we introduce

the perturbation " at the point i. The development of the perturbation is
considered below in time and in x direction:
time instant n:
in D "
time instant n C 1:
inC1 D in .1 c/ C c in1 D ".1 c/

inC1 n n
C1 D i C1 .1 c/ C c i D c "
time instant n C 2:
inC2 D inC1 .1 c/ C c inC1 2

1 D ".1 c/ D ".1 c/
2
inC2 nC1 nC1

C1 D i C1 .1 c/ C c i D c ".1 c/ C c ".1 c/ D 2c ".1 c/
inC2 nC1 nC1 2
C2 D i C2 .1 c/ C c i C1 D c "
time instant n C 3:
inC3 D inC2 .1 c/ C c inC2

1 D ".1 c/
3
inC3 nC2 nC2

C1 D i C1 .1 c/ C c i D 2c ".1 c/2 C c ".1 c/2
inC3 nC2 nC2 2 2
C2 D i C2 .1 c/ C c i C1 D c ".1 c/ C c .2"/.1 c/
nC3
iC3 D inC2 nC2 3
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
c < 1 36 (3.14)
The condition (3.14) is the Download
Courant free
Friedrich Levy
eBooks at criterion of the stability
bookboon.com
of explicit numerical schemes. If the velocity is changed within the compu-
tational domain, the maximum velocity umax is taken instead of u in for-
time instant n C N :
.inCN
CN / D c "
N
::::::::::::::::::
As follows from the last formula,the perturbation decays if
c<1 (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 compu-
tational domain, the maximum velocity umax is taken instead of u in for-
mula (3.14). Physically the condition umax
x
t
< 1 means that the maximum
displacement of the fluid particle within the time step Œt; 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
The field of the velocity component ux is given as ux;ij D exp...ix
0:5/2 C .jy 0:5/2 /.
Calculate uy;ij from continuity equation (1.3) and the t satisfying the CFL
criterion.
37

Simple explicit time advance scheme for
Chapter 4
Lectures on computational fluid dynamics solution of the Navier Stokes Equation
Chapter
Chapter 4
4
Simple explicit time advance
Simple
Simple explicit
scheme time
for solution
explicit advance
timeof the
advance
scheme
Navier for
for solution
scheme Stokes of
of the
Equation
solution the
Navier
Navier Stokes
Stokes Equation
Equation
4.1 Theory
4.1unsteady
The Theory
term of the Navier Stokes Equation
4.1 Theory
The unsteady term of@uthe i @ui uStokes
Navier j 1 @p Equation@ @ui
The unsteady term of@tthe Navier D C (4.1)
@xj Stokes @x Equation
i @xj @xj
@ui @ui uj 1 @p @ @ui
is written in explicit @u form:D @u u 1 @p C @ @x @ui (4.1)
@ti D @xi j j @xi C @x j j (4.1)
@t j n n @xi
@x n @xj @xj n
is written in explicitnC1 form: n
ıui uj 1 ıp ı ıui
u
is written in explicit form:
i D u i C t C (4.2)
ıx nj n
@xi
n
ıxj ıxj
n
ı u nC1
D u n
C t ıuin ujn 1 ıp n C @
ı ıui
(4.2)
where ıxj is thenC1 i approximation
i
n
ıu
of theıx
u
i derivative
j 1
ıp
@x ı ıu
. ıx
Let usniapply the diver-
ıx
ui ı D ui C t j i C
@xj j j (4.2)
gence operator : ıxj @xi ıxj ıxj
ı ıxi @
where ıxj is the approximation of the derivative @xj . Let us apply the diver-
where ıxıj isıuthe
nC1approximation
ı
ıu n
ıof theıu derivative
n n
u 1
@
@xjn
ıp . Let usı apply

ıuni the diver-
gence operator i ıxi : i i j
ıD C t C (4.3)
gence operator ıxi ıxi : ıxi ıxi ıx j @x i ıxj ıxj
ıunC1
inC1 ıuni ı ıuni ujn n 1 ıp n ı ıuni
ıuıx D ıu n C t ı n ıu n
ıui uj i D1 ıp n C ı ıuis n (4.3)
Let uni is the i divergence
i D ıxii Cfree field,
t ıx i.e. 0. The task
i ıxj ıxi @xi C ıxj ıxji
to find(4.3)
the
velocity fieldıx ati the time ıxi moment ıxi n C 1ıxwhich j n is
@xalso
i divergence
ıxj ıxj free
ıui
Let uni is the divergence free field, i.e. ıu n
ıxii
D 0. The task is to find the
Let uni isfield
velocity theatdivergence
the time moment ıunnC1
free field, i Ci.e.1 which
ıxi
Dis 0.
also The task
divergence is to
freefind the
velocity field at the time moment n C 1 Dwhich
0 is also divergence free (4.4)
ıxi
ıunC1inC1
Substituting (4.4) into (4.3) one ıu obtains: D0 (4.4)
ıx i i
D 0 (4.4)
n n ıx i n n
Substituting (4.4) into ıu
ı (4.3)i one u 1 ıp
j obtains: 39 C ı ıui D 0
Substituting (4.4) ıx into
(4.3) one obtains: (4.5)
i ıxj @xi ıxj ıxj
Expressing (4.5) with respect to the pressure 39 results in the Poisson equation:
39
" #
ı2pn ı 2 uni ujn ı 2 ıuni
D C (4.6)
@xi2 ıxj ıxi ıxi ıxj ıxj
The algorithm for time-advancing is as follows:
i) The solution at time n is known and divergence free.

h ı 2 un un i
2 ıun
ii) Calculation of the r.h.s. of (4.6) ıxj iıxji C ıxıi ıxj i
ıxj
iii) Calculation of the pressure p n from the Poisson equation (4.6)
iv) Calculation of the velocity unC1 38is divergence free.

. This
i
v) Go to the step ii).
In the following sections we consider the algorithm in details for the two
ıxi ıxj @xi ıxj ıxj
Expressing (4.5) with respect to the pressure results in the Poisson equation:
" # explicit time advance scheme for
Simple
2 n 2 n n 2 n
ı p
Lectures on computational fluid D
dynamics
ı u i uj ı ıu i solution of the Navier Stokes Equation
C (4.6)
@xi2 ıxj ıxi ıxi ıxj ıxj
The algorithm for time-advancing is as follows:
i) The solution at time n is known and divergence free.

h ı 2 un un i
2 ıun
ii) Calculation of the r.h.s. of (4.6) ıxj iıxji C ıxıi ıxj i
ıxj
iii) Calculation of the pressure p n from the Poisson equation (4.6)
iv) Calculation of the velocity unC1

i . This is divergence free.
v) Go to the step ii).
In the following sections we consider the algorithm in details for the two
dimensional case.
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.Œ'in C .'inC1 'in / Œ'in1 C .'in 'in1 / / D
2 2
1 n 1
D 'in c.Œ'in 'in1 C Œ'iC1 'in Œ'in 'in1 / (4.8)
2 2
where c D utx
is the CFL parameter.40The term cŒ'in 'in1 is the diffusive
1st order upwind contribution. The term c. 12 Œ'inC1 'in 12 Œ'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
'inC1 D 'in c.Œ'in 'in1 C Œ'inC1 'in east
n
Œ'in 'in1 west
n
/ (4.9)
2 2
where are limiters. Limiters functions for TVD schemes are given in ta-
ble 4.1.
Table 4.1: Limiters function for TVD schemes

Scheme 39
central 1
upwind 0
1 1
'inC1 D 'in c.Œ'in 'in1 C Œ'inC1 'in east
n
Œ'in 'in1 west
n
/ (4.9)
2 2
Simple explicit
where are limiters. Limiters functions for TVD schemes timein
are given advance
ta- scheme for
Lectures
ble 4.1.on computational fluid dynamics solution of the Navier Stokes Equation
Table 4.1: Limiters function for TVD schemes

Scheme
central 1
upwind 0
Roe minimod D max.0; mi n.r; 1//
Roe superbee D max.0; mi n.2r; 1/; mi n.r; 2//
rCmod.r/
Van Leer D 1Cmod.r/
Branley and Jones D max.0; mi n.2r; 1//
'n ' n
Here r D . @' /n =. @'
@x west
/n , . @'
@x east
/n D xiC1
@x east
i
iC1 xi
. The mixed upwind and cen-
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 collo-
cated 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

'n ' n
Here r D . @' /n =. @'
@x west
/n , . @'
@x east
/n D xiC1
@x east
i
iC1 xi
. The mixed upwind and cen-
tral difference scheme are used in Sec. 4.4 for approximation
Simple of the convective
explicit time advance scheme for
terms with the limiterfluid dynamics
(4.14). solution of the Navier Stokes Equation
4.3 Staggered grid

The grids are subdivided into collocated and staggered ones. On collo-
cated 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
Dx D C
41
@x 2 @y 2
@2 uy @2 uy
Dy D 2 C 2
@x @y
(4.10)
@ux ux @ux uy
Hx D
@x @y
@ux uy @uy uy
Hy D
@x @y
Application of the central difference scheme to the Poisson equation results

in
n n
. @p / . @p
n n
. @p / . @p
@x E
/
@x W @y N
/
@y S
n
Tx;E n
Tx;W n
Ty;N n
Ty;S
C D C
2x 2y 2x 2y
or
N pP pSS
n n n n
n
pEE pP
n n
pP pW
n pN pP
2x
2x
W
2y
2y
n
Tx;E n
Tx;W n
Ty;N n
Ty;S
C D C D QPH
2x 2y 2x 2y
The last equation is expressed in the matrix form:

X p
APp pPn C Al pln D QPH (4.11)
l
where
l D EE; W W; N N; S S , ApEE D ApW W D .21x /2 , ApN N D ApS S D .21y /2

P
and APp D l Apl
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 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,
l
where
l D EE; W W; N N; S S , ApEE D ApW W D .21x /2 , ASimple

p explicit
p time advance
N N D AS S D .2y /2
1 scheme for
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
Figure 4.1: Sample of the collocated grid.
Figure 4.2: Checkerboard pressure solution on the collocated grid.
42
Figure 4.3: Download

Grid points of staggered
free eBooks grid.
at bookboon.com
Figure 4.2: Checkerboard pressure solution on the collocated grid.
Figure 4.3: Grid points of staggered grid.

There is also the possibility that one may not be able to obtain a converged
steady-state solution. 43
A possible solution of the problem is the application of the staggered grids
(Fig. 4.3). The Poisson equation is satisfied at cell centres designated by
crosses. The ux velocities are stored at points staggered by x =2 in x-
direction (filled circles). At these points the first Navier- Stokes equation is
satisfied:
@ux @ux uj 1 @p @ @ux

D C (4.12)
@t @xj @x @xj @xj
The uy velocities are stored at points staggered by y =2 in y-direction (cir-

cles). At these points the second Navier-Stokes equation is satisfied:
WE WILL TURN YOUR CV
@uy @uy uj 1 @p INTO@ @uy AN OPPORTUNITY
D C (4.13)
@t @xj @y @xj @xj OF A LIFETIME
The staggered grid is utilized below.
ıuni ujn
4.4 Approximation of ıxj
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
Do you[4].
approach like cars? Would you to
Following like to
[4]bewe
a part of a successfulabrand?
implement smooth transition Send between
us your CV on
As a constructer at ŠKODA AUTO you will put great things in motion. Things that will
centered 2 entirely www.employerforlife.com
easedifferencing
everyday lives ofand upwinding
people using
all around Send a parameter
us your CV. We will giveit an Œ0; 1. It is defined
as new new dimension.
D mi n.1:2 t max.jux .i; j /j; juy .i; j /j/; 1/ (4.14)

Download free
The value of gamma is the maximum eBooksof
fraction at abookboon.com
cells which information can
travel in one time step, multiplied by 1:2, and capped by 1. The factor of 1:2
(Fig. 4.3). The Poisson equation is satisfied at cell centres designated by
crosses. The ux velocities are stored at points staggered by x =2 in x-
direction (filled circles). At these points the first Navier- Stokes
Simple equation
explicit is scheme for
time advance
satisfied: solution of the Navier Stokes Equation
@ux @ux uj 1 @p @ @ux

D C (4.12)
@t @xj @x @xj @xj
The uy velocities are stored at points staggered by y =2 in y-direction (cir-

cles). At these points the second Navier-Stokes equation is satisfied:
@uy @uy uj 1 @p @ @uy

D C (4.13)
@t @xj @y @xj @xj
The staggered grid is utilized below.
ıuni ujn
4.4 Approximation of ıxj
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 Œ0; 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).
@ux ux @ux uy
4.4.1 Approximation of 44@x
@y
D ux @u
@x
x
uy @u
@y
x
Point .i C 1=2; j / for x-component of velocity:
ux .i C 1=2; j / D .ux .i C 1; j / C ux .i; j //=2:0

If ux .i C 1=2; j / 0;
1 C1
then ux ux ji C1=2;j D ux .i C 1=2; j /. ux .i C 1; j / C ux .i; j //
2 2
C1 1
else ux ux ji C1=2;j D ux .i C 1=2; j /. ux .i C 1; j / C ux .i; j //
2 2
Point .i 1=2; j / for x-component of velocity:
ux .i 1=2; j / D .ux .i; j / C ux .i 1; j //=2:0

If ux .i 1=2; j / 0;
1 44 C1
then ux ux ji 1=2;j D ux .i 1=2; j /. ux .i; j / C ux .i 1; j //
2
Download free eBooks 2
at bookboon.com
C1 1
else ux ux ji 1=2;j D ux .i 1=2; j /. ux .i; j / C ux .i 1; j //
2 2
Point .i 1=2; j / for x-component of velocity:

ux .ioncomputational
Lectures 1=2; j / D .ufluid j / C ux .i 1; j //=2:0
x .i;dynamics solution of the Navier Stokes Equation
If ux .i 1=2; j / 0;
1 C1
then ux ux ji 1=2;j D ux .i 1=2; j /. ux .i; j / C ux .i 1; j //
2 2
C1 1
else ux ux ji 1=2;j D ux .i 1=2; j /. ux .i; j / C ux .i 1; j //
2 2
Point .i C 1=2; j / for y-component of velocity:
uy .i C 1=2; j / D .uy .i C 1; j / C uy .i; j //=2:0

If uy .i C 1=2; j / 0;
C1 1
then ux uy ji C1=2;j D uy .i C 1=2; j /. ux .i; j / C ux .i; j C 1//
2 2
1 C1
else ux uy ji C1=2;j D uy .i C 1=2; j /. ux .i; j / C ux .i; j C 1//
2 2
Point .i C 1=2; j 1/ for y-component of velocity:
uy .i C 1=2; j 1/ D .uy .i C 1; j 1/ C uy .i; j 1//=2:0

If uy .i C 1=2; j 1/ 0;
C1 1
then ux uy ji C1=2;j 1 D uy .i C 1=2; j 1/. ux .i; j 1/ C ux .i; j //
2 2
1 C1
else ux uy ji C1=2;j 1 D uy .i C 1=2; j 1/. ux .i; j 1/ C ux .i; j //
2 2
@ux ux @ux uy
Hx .i; j / D D
@x @y
ux ux ji C1=2;j ux ux ji 1=2;j ux uy ji C1=2;j ux uy ji C1=2;j 1

x y
@u u @uy uy @uy @uy

4.4.2 Approximation of 45@x
x y
@y
D ux @x
uy @y
Point .i; j C 1=2/ for y-component of velocity:
uy .i; j C 1=2/ D .uy .i; j / C uy .i; j C 1//=2:0

If uy .i; j C 1=2/ 0;
1 C1
then uy uy ji;j C1=2 D uy .i; j C 1=2/. uy .i; j C 1/ C uy .i; j //
2 2
C1 1
else uy uy ji;j C1=2 D uy .i; j C 1=2/. uy .i; j C 1/ C uy .i; j //
2 2
Point .i; j 1=2/ for y-component of velocity:
uy .i; j 1=2/ D .uy .i; j 1/ C uy .i; j //=2:0

If uy .i; j 1=2/ 0;
1 C1
then uy uy ji;j 1=2 D uy .i; j 1=2/. uy .i; j / C uy .i; j 1//
2 2
C1 1
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:
ux .i; j C 1=2/ D .ux .i; j C 1/ C ux .i; j //=2:0

45
If ux .i; j C 1=2/ 0;
C1 1
then uy ux ji;j C1=2 D ux .i; j C 1=2/. uy .i; j / C uy .i C 1; j //
2 2
If uy .i; j 1=2/ 0;
1 C1
then uy uy ji;j 1=2 D uy .i; j 1=2/. uy .i; j / C uy .i; j 1//
2 Simple
2 explicit time advance scheme for
Lectures on computational fluid dynamics C1 1 solution
of the Navier Stokes Equation
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:
ux .i; j C 1=2/ D .ux .i; j C 1/ C ux .i; j //=2:0

If ux .i; j C 1=2/ 0;
C1 1
then uy ux ji;j C1=2 D ux .i; j C 1=2/. uy .i; j / C uy .i C 1; j //
2 2
1 C1
else uy ux ji;j C1=2 D ux .i; j C 1=2/. uy .i; j / C uy .i C 1; j //
2 2
Point .i 1; j C 1=2/ for x-component of velocity:
ux .i 1; j C 1=2/ D .ux .i 1; j C 1/ C ux .i 1; j //=2:0

If ux .i 1; j C 1=2/ 0;
C1 1
then uy ux ji 1;j C1=2 D ux .i 1; j C 1=2/. uy .i 1; j / C uy .i; j //
2 2
1 C1
else uy ux ji 1;j C1=2 D ux .i 1; j C 1=2/. uy .i 1; j / C uy .i; j //
2 2
@ux uy @uy uy
Hy .i; j / D D
@x @y
uy ux ji;j C1=2 uy ux ji 1;j C1=2 uy uy ji;j C1=2 uy uy ji;j 1=2

x y
46
Develop the tools we need for Life Science

Masters Degree in Bioinformatics
Bioinformatics is the
exciting field where biology,
computer science, and
mathematics meet.
We solve problems from

biology and medicine using
methods and tools from
computer science and
mathematics.
Read more about this and our other international masters degree programmes at www.uu.se/master

ı ıunni
4.5
4.5 Approximation
Approximation of
ıof iıx
ıu n ı ıui
ıxj
4.5 Approximation of ıxj ıxj ıxjj ıxj
The second derivative is calculated using the Central Difference Scheme
The second
The secondderivative
derivative is calculated
is calculated using using the Central
the Central DifferenceDifference
Scheme Scheme
(CDS):
(CDS):
(CDS):
C 1;
ux .i 1; / j / x .i;
2u .i;
C uj / C u .i u1; j /j Cu1/ .i; j xC.i;1/
j / 2u
uxxx.i;.i; j / C u .i; j 1/
.i;jj/ /DDuxu.ixC
Dxx.i; .i Cj1; / 2ujxx/.i;
j 2u jx/.iCu1;xx j.i/ C
1; 2u
j / C uxx .i;
x .i; j C 1/ C2u .i;j j/1/C uxx .i; j 1/
Dx .i; j / D
D
2x
2 C y2 y22
2x x y
.iy C
uyu .i 1; / j 2u
Cj1; /y .i; C uj /.iCu1; j.i/ u1;
2ujy/.i; j /j Cu1/
y .i; 2u
.i; j C.i;1/ C2u
j / j j/1/
u .i;.i; C u .i; j 1/
D .i;jj/ /DD uy .i C 1; j / 2u
Dyy.i; 2 y 2
.i; jy/ C uyy .i C 1; j / C uyy .i; j yC21/ 2uyyy .i; j / C uyy .i; j 1/
Dy .i; j / D x C y 2
2xx yy2
4.6 Calculation of the r.h.s. for the Poisson
4.6 Calculation
4.6 equation
Calculation of the r.h.s. for the Poisson
(4.6) of the r.h.s. for the Poisson
equation (4.6)
The r.h.s. equation
of (4.6) is (4.6)
The r.h.s. ıof2 u(4.6)
n n is 2
The r.h.s. of (4.6)
i uj is ı ıuni @Hx @Hy @Dx @Dy
C D C C C
ıxıj 2ıx n
ui u n ıxi ıxj ıx
2 j @x
n @y @x @y
ı 2 uni ujn ı2 ıuni @Hx @Hy @Dx @Dy
at ithej point
The derivatives
ı ıu
C (i,j) of the ipressure
D x C @H
y C x C@H y @D @D
ıxj ıxi C ıx i ıxj ıxj
D @xstorage
C @y(designated
C @x C as X in
@y
ıxj ıxi usingıxCDS
Fig. 4.3) are calculated i ıxj ıxj @x @y @x @y
The derivatives at the point (i,j) of the pressure storage (designated as X in
The derivatives at the point (i,j) of the pressure storage (designated as X in
Fig.@H4.3) areHcalculated using CDS
x .i; j / Hx .i 1; j / @Dx Dx .i; j / Dx .i 1; j /
Fig. 4.3)
x
jij are
D calculated using CDS ; jij D ;
@x x @x x
@Hyx
@H HH x .i;
y .i; j /Hy .i;
j/ Hjx .i 1;@Djy/ @DxDy .i; j /D
1/ xD
.i; j /j
y .i; D1/x .i 1; j /
@HxjijjijDD H x .i; j / Hx .i ;1; j /j;ij@D
D x jij D Dx .i; j / Dx .i 1; j / ;
@y j
@x ij D y x @y ; j
@x ij D y x ;
@x x @x x
Hy .i; j / Hy .i; j 1/ @Dy Dy .i; j / Dy .i; j 1/
4.7 @H
@HyySolution j / the
jij D Hy .i; of j 1/ ; @Dequation
Poisson
Hy .i; j / Dy .i; j 1/
y jij D Dy .i; (4.6)
j
@y ij D y ; j
@y ij D y
@y y @y y
The numerical solution of the Poisson equation is discussed in [3].
4.7 Solution
4.7 Update
Solution of the
the Poisson equation
equation (4.6)
4.8 the of Poisson
velocity field (4.6)
The numerical solution of the Poisson equation is discussed in [3].
Thevelocity
The numerical
field solution
is updatedofaccording
the Poisson equation is discussed in [3].
to formula
4.8
4.8
unC1
x
Update
Update the
the velocity
velocity field
field
.i; j / D unx .i; j /C t .Hxn .i; j /CDxn .i; j /.p n .i C1; j /p n .i; j //=x /
nC1
u
The
y .i; j / D uynfield
velocity .i; j / C n
isupdated C Dyn .i; j /
t .Hy .i; j /according
n
.i; j C 1/ p n .i; j //=y /
to.pformula
The velocity field is updated according to formula
47
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 /
unC1
x
x .i; j / D ux .i; j /C t .Hx .i; j /CDx .i; j /.p .i C1; j /p .i; j //=x /
uynC1 .i; j / D un .i; j / C .H n .i; j / C D n .i; j / .p n .i; j C 1/ p n .i; j //= /
uynC1 .i; j / D uyyn .i; j / C tt .Hyyn .i; j / C Dyyn .i; j / .p n .i; j C 1/ p n .i; j //=yy /
47
47
47
4.9 Boundary conditions for the velocities

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
At symbols
this in boundary
stage, the Fig. 4.3. conditions
Enforcement(BC)offorthetheBC is easy,
velocity if the
field ”grey”
should be point
lies exactly
taken at the The
into account. boundary of which
nodes at the computational domain.
the BC are enforced are If not, by
shown two cases
should
grey be considered.
symbols If the Neumann
in Fig. 4.3. Enforcement of the condition the ”grey”@u
BC is easy,isif enforced D
point
@n
C , the
lies exactly
velocity at the boundary
component outsideof the computational
of the boundary domain. If not, two cases
u.0/ is calculated through the
should
interiorbequantity
considered. the Neumann condition is enforced @u
u.1/If from @n
D C , the
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
If domain,u the
the Dirichlet condition D Cvalue is calculated
u.0/ and
is enforced the point 0from the extrapolation
is outside of the
procedure:
computational domain, the value u.0/ is calculated from the extrapolation
procedure:
u.0/ C u.1/
u.0/ C u.1/ DC
2 DC
2
4.10 Calculation
4.10 Calculation of of
thethe vorticity
vorticity
@ux @u
The calculation
The calculationof of vorticity
vorticity !z @uDx
!z D @uy performed
@y @x is@x
y
is performed as follows:
as follows:
@y
1 1
uxx.i.i 1=2;
u 1=2;j jCC D D.ux.u
1=2/
1=2/ / Cj /uxC
.i;xj.i; C 1/
.i;ujx .i; C u1/x .iCu1;
jC j /C1;
x .i uxj.i/ C uxj.iC1//;
1; 1; j C 1//;
4 4
1
uxx.i.i 1=2;
u 1=2;j j1=2/ D D.u1x.u
.i; j / C ux .i 1; j / C ux .i; j 1/ C ux .i 1; j 1//;
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//;
uy .i 1=2; j 1=2/ D 4 .uy .i; j / C uy .i 1; j / C uy .i; j 1/ C uy .i 1; j 1//;
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 y x
!z .i; j / D
uy .i C 1=2; j 1=2/ uy .i 1=2; yj 1=2/

uy .i C 1=2; j 1=2/x uy .i 1=2; j 1=2/

x
48
48
48
Splitting schemes for solution of
Chapter 5
Lectures on computational fluid dynamics multidimensional problems
Chapter
Chapter 5
Splitting 5schemes for solution
Splitting
of schemes
schemes for
multidimensional
Splitting solution
forproblems
solution
of
of multidimensional
multidimensional problems
problems
5.1 Splitting in spatial directions. Alternat-
5.1 ing direction
Splitting implicit
in spatial (ADI) approach
directions. Alternat-
5.1 Splitting in spatial directions. Alternat-
ing direction
Let us consider implicit
the two dimensional unsteady(ADI) approach
heat conduction equation:
ing direction implicit 2 (ADI)
2
approach
@
Let us consider the two dimensional
@unsteady @ heat conduction equation:
Let us consider the two dimensional D unsteady C 2heat conduction equation: (5.1)
@t @x 2 @y
@ @2
Crank @2
We use implicit scheme proposed @ D by
2 C and 2 Nicolson and CDS for spatial (5.1)
derivatives: @t D @@x2 C @@y2 (5.1)
We use implicit scheme proposed @t @x 2 and
by Crank @y 2 Nicolson and CDS for spatial
We use implicit schemeproposed by Crank
derivatives: and Nicolson and CDS for spatial
nC1 n 2 n 2 n 2 nC1 2 nC1

derivatives: @ @ @ @
D C C C (5.2)
t 2 @x 2 @y 2
@x 2 @y 2

nC1 n @2 n @2 n @2 nC1 @2 nC1
nC1 n D 2 n C 2 n n C 2 nC1 C (5.2)
t 2 @n 2 @@y
iC1;j
2 2@i;jn 2 n
@xC i 1;j @2@y
nC1
2
D 2@ @x 2 C C C (5.2)
t 2 @x 2@x D @y 2 @x 2 @y 2 (5.3)
2 n i;j n
.x/ n
2
@2 n D inC1;j 2i;j C in1;j

n n (5.3)
@@x2 ii;j C1 .x/
C1;j 2i;j2C i;j i 1;j
1
i;j D (5.3)
(5.4)
@x
@y 2
i;j n
.x/
.y/ n
2
n
C1 2i;j C i;j 1
2 n
@ n D i;j n n n approximations: (5.4)
In what follows we use@2the designations
i;j2C i;j 1
for2derivative
@y2 i;j D i;j C1 .y/ (5.4)
@y22 i;j n 2.y/
n
C
2 n

ı
In what follows we use the ndesignationsi C1;j fori;jderivative
i 1;j approximations:
In what follows we useıxthe D
2 designations .x/ for derivative
2 approximations: (5.5)
Copenhagen
n n n
2
ı n i C1;j 2i;j C i 1;j
2 D n n n (5.5)
ı
ıx 2 n D ii;j C1 .x/
C1;j 2i;j 2C i;j i 1;j
1
(5.5)
(5.6)
ıy22
ıx n
.x/
.y/ n
2
n
i;j C1 2i;j C i;j 1
Master of Excellence ı n
ıy
2 D n
ı2 n
ıy 2
D
i;j C149 .y/
.y/2
n
2i;j2C i;j n
1
(5.6)
(5.6)
cultural studies
49
Copenhagen Master of Excellence 49 are
two-year master degrees taught in English
at one of Europe’s leading universities religious studies
Come to Copenhagen - and aspire!
Apply now at science

www.come.ku.dk

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 n @ @2 nC1
D fluid dynamics
C C C multidimensional
(5.2) problems
t 2 @x 2 @y 2 @x 2 @y 2

@2 n inC1;j 2i;j
n
C in1;j
D (5.3)
@x 2 i;j .x/2
n n n
@2 n i;j C1 2i;j C i;j 1
D (5.4)
@y 2 i;j .y/2
In what follows we use the designations for derivative approximations:
ı2 n inC1;j 2i;j
n
C in1;j
D (5.5)
ıx 2 .x/2
n n n
ı2 n i;j C1 2i;j C i;j 1
D (5.6)
ıy 2 .y/2
Using these designations we get from 49

the original heat conduction equation:

t ı 2 t ı 2 nC1 t ı 2 t ı 2
1 1 D 1C 1C nC
2 ıx 2 2 ıy 2 2 ıx 2 2 ıy 2

.t /2 ı 2 ı 2 nC1 n
C
4 ıx 2 ıy 2
(5.7)
The last term is neglected since . nC1 n /.t/2 @

@t
.t/3 .
Numerical solution is performed in two following steps:
Step 1: Solution of one dimensional problem in x-direction:

t ı 2 t ı 2
1 D 1C n (5.8)
2 ıx 2 2 ıy 2
The numerical solution of (5.8) is then substituted as the guess

solution for the next step:
Step 2: Solution of one dimensional problem in y-direction

t ı 2 t ı 2
1 nC1
D 1C (5.9)
2 ıy 2 2 ıx 2
It can be shown that the resulting method is of the second order of accuracy
and unconditionally stable.
5.2 Splitting according to physical processes.

Fractional step methods
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 50
D C u C Du C P (5.10)
@t
Here
1 nC1 D 1 C (5.9)
2 ıy 2 2 ıx 2
It can be shown that the resulting method is of the second order of accuracy
Splitting schemes for solution of
Lectures on computationalstable.
and unconditionally fluid dynamics multidimensional problems
5.2 Splitting according to physical processes.

Fractional step methods
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 (5.10)
@t
Here
C is the convection operator,

D is the diffusion operator,
50
P is the pressure source operator.
Simple explicit Euler method can be written as
unC1 D un C .C u C Du C P /t (5.11)

The overall numerical solution is considered as the consequence of numerical
solutions of the following partial problems:
u D un C .C un /t (*) Convection step

u D u C .Du /t (**) Diffusion step (5.12)
unC1 D u C .P /t (***) Pressure step
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/ (5.13)
@t
The splitting of (5.10) results in two steps procedure:
@u
D L1 .u /; u j t Dtn D un
@t (5.14)
@unC1
D L2 .unC1 /; unC1 j t Dtn D u
@t
where
u D un C tL1 .un / C O.t 2 /;

unC1 D u C tL2 .u / C O.t 2 / D 51
(5.15)
D un C tL1 .un /Download
C tL2 .ufreen eBooks at bookboon.com
C tL1 .un // C O.t 2 / D
D un C t.L1 .un / C L2 .un // C O.t 2 /:
@u
D L1 .u /; u j t Dtn D un
@t Splitting schemes
(5.14) for solution of
nC1
@ufluid dynamics nC1 multidimensional problems
D L2 .u /; unC1 j t Dtn D u
@t
where
u D un C tL1 .un / C O.t 2 /;

unC1 D u C tL2 .u / C O.t 2 / D
(5.15)
D un C tL1 .un / C tL2 .un C tL1 .un // C O.t 2 / D
D un C t.L1 .un / C L2 .un // C O.t 2 /:
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 ıunx unx ıunx uyn

D (5.16)
t ıx ıy
uy uyn ıuyn unx ıuyn uyn

D (5.17)
t ıx ıy
The solutions ux;y are used then for the diffusion process which is
treated implicitly:
2
u
ı 2 uBy
Brain power x ux
t
D
ı ux
ıx 2
C x 2020, wind could provide one-tenth of our planet’s
ıyelectricity
2
(5.18)
needs. Already today, SKF’s innovative know-
how is crucial to running a large proportion of the

world’s wind turbines.
uy uy ı 2 uy ı 2 uyUp to 25 % of the generating costs relate to mainte-
D C nance. These can be reduced (5.19)
dramatically thanks to our
t ıx 2 2
ıysystems for on-line condition monitoring and automatic
lubrication. We help make it more economical to create
The solutions u
x;y are cleaner, cheaper energy out of thin air.
used then for the next process which is treated
By sharing our experience, expertise, and creativity,
explicitly: industries can boost performance beyond expectations.

unC1
x u x ıp nC1 Therefore we need the best employees who can
D meet this challenge! (5.20)
t ıx
uynC1 uy The Power of Knowledge Engineering
ıp nC1
D (5.21)
t ıy
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:
ıunC1
x
ıuynC1
C
Plug into The Power of Knowledge Engineering. D0 (5.22)
Visit us at www.skf.com/knowledge ıx ıy
Applying the operator r to the equations (5.20) and (5.21) we get the Poisson
equation for the pressure:

ı 2 p nC1 ı 2 p nC1 1 52ıu
x
ıuy Click on(5.23)
the ad to read more
C D C
ıx 2 ıy 2 freet
Download eBooksıx ıy
at bookboon.com
52
ux unx ıun un ıux uy
D x x (5.16)
t ıx ıy
uy uyn ıuyn unx ıuyn uyn Splitting schemes for solution of
Lectures on computational fluid dynamics D multidimensional
(5.17) problems
t ıx ıy
The solutions ux;y are used then for the diffusion process which is
treated implicitly:
2
u
x ux ı ux ı 2 u
x
D C (5.18)
t ıx 2 ıy 2

uy uy ı 2 uy ı 2 uy
D C (5.19)
t ıx 2 ıy 2
The solutions u

x;y are used then for the next process which is treated
explicitly:
unC1
x u
x ıp nC1
D (5.20)
t ıx
uynC1 uy ıp nC1
D (5.21)
t ıy
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:
ıunC1
x
ıuynC1
C D0 (5.22)
ıx ıy
Applying the operator r to the equations (5.20) and (5.21) we get the Poisson
equation for the pressure:

ı 2 p nC1 ı 2 p nC1 1 ıu
x
ıuy
C D C (5.23)
ıx 2 ıy 2 t ıx ıy
5.3 Increase of the accuracy

52 of time deriva-
tives approximation using the Lax-Wendroff
scheme
Let us consider the general transport equation:
@u @F .u/
C D0 (5.24)
@t @x
We introduce the following designations:
@F
AD
@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 C / D u.t/ C u t .t/ C u t t .t/ C O. 3 /
2 53 (5.26)
2 3
D u.t/ Fx .t/ C .A.t/F
x .t//x C O. /
at bookboon.com
2
@F
AD
@u Splitting schemes for solution of
Lectures on computational fluid dynamics multidimensional problems
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 C / D u.t/ C u t .t/ C u t t .t/ C O. 3 /
2 (5.26)
2
D u.t/ Fx .t/ C .A.t/Fx .t//x C O. 3 /
2
A difficulty arising in the LW approach is the computation of the operator A.
One can easy derive that the Lax Wendroff scheme results in the solution of
the equation:

unC1 un @F n @ n @F
n
C D A
@x 2 @x @x
Trust and responsibility

NNE and Pharmaplan have joined forces to53
create – You have to be proactive and open-minded as a
NNE Pharmaplan, the world’s leading engineering newcomer and make it clear to your colleagues what
and consultancy company focused entirely on the you are able to cope. The pharmaceutical field is new
pharma and biotech industries. to me. But busy as they are, most of my colleagues
find the time to teach me, and they also trust me.
Inés Aréizaga Esteva (Spain), 25 years old Even though it was a bit hard at first, I can feel over
Education: Chemical Engineer time that I am beginning to be taken seriously and
that my contribution is appreciated.
NNE Pharmaplan is the world’s leading engineering and consultancy company

focused entirely on the pharma and biotech industries. We employ more than
1500 people worldwide and offer global reach and local knowledge along with
our all-encompassing list of services. nnepharmaplan.com

Chapter 6
Lectures on computational fluid dynamics Finite Volume Method
Chapter 6
Chapter 6
Finite Volume Method
6.1 Transformation of the Navier-Stokes Equa-
6.1 tions in the Finite
Transformation Volume
of the Method Equa-
Navier-Stokes
6.1 Transformation of the Navier-Stokes Equa-
tions
The Navier Stokesin the Finite Volume Method
equation
tions in the Finite Volume Method

The Navier Stokes @ui equation
@.ui uj / 1 @p @ @
The Navier Stokes C D Fi C ui (6.1)
@t equation@xj @xi @xj @xj
@ui @.ui uj / 1 @p @ @
@u
is fulfilled within@teach C @.u
mesh u / D Fi 1 @p C @j U)@x
@in uthe
i (6.1)
i @xi j j element (finite
@x volume
@x integral sense.
C D Fi i C j ui (6.1)
For that it is integrated
@t over
@xj the volume U
@xi : @xj @xj
is fulfilled within each mesh element (finite volume U) in the integral sense.
is fulfilled
For within each mesh element (finite
U : volume U) in the integral sense.
Zthat
it is integrated over the Z volume

For that@u it iis integrated
@.ui uj / over the volume1 U@p : @ @
C dU D Fi C ui d U (6.2)
Z @t @xj Z @xi @xj @xj
Z @ui C @.ui uj / d U D Z Fi 1 @p C @ @ ui d U
U U
@u (6.2)
@t i C @.u @xi u
jj
/
dU
1 @pi
Fi @x
D U results
@ @
C @xj @xj ui d U (6.2)
Application
U @t of the
@xj Gauss theorem in
@xi @xj @xj
U U
Application
Z of the
Z Gauss theorem Z results Zin 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 (6.3)
@t Z Z Z Z Z
@ UZ S U 1
Z ui uE nE dS D Z Fi d U Z p eEi nE dS C Z grad ui nE dS
S S
@ u i d U C 1 (6.3)
@t U C S ui uEapplied
ui d procedure nE dS D Uto F S p eEequation
E dS C gives
grad ui nE dS (6.3)
The Usame idU
the continuity in S
@t
Z
U S U S S
The same procedure applied to the continuity equation gives
The same procedure applied to theuE ncontinuity
E dS D 0 equation gives (6.4)
Z
Z uE nE dS D 0
S
(6.4)
S
uE nE dS D 0 (6.4)
55
S
55
55
Figure 6.1: Staggered arrangement of finite volumes.

Figure 6.1: Staggered arrangement of finite volumes.
6.2 Sample 55
6.2 Sample Download free eBooks at bookboon.com
Let us consider the two dimensional transport equation without the diffusion
term
Figure 6.1:fluid dynamics arrangement of finite volumes.
Staggered Finite Volume Method
6.2 Sample
term
8̂
@ui @ui uj @p
ˆ
< @t C @x D @x
j i
(6.5)
ˆ @uj
:̂ D0
@xj
In the integral form this equation reads
Z Z Z
@
ui d U C ui uE nE dS D p eEi nE dS (6.6)
@t
U S S
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 Pressure and unsteady terms

Source (pressure) term for x-equation:
Z
p
Q1 D p eE1 nE dS pe Se 56 pw Sw D pi C1j pij (6.7)
S
Unsteady term for x-equation:

Z
@ unC1 n
xij uxij
ux d U D 2 (6.8)
@t t
U
Pressure term for y-equation:
Z
Q2p D p eE2 nE dS pn Sn ps Ss D pij C1 pij (6.9)
S
Unsteady term for y-equation:

Z nC1 n
@ uyij uyij
uy d U D 2 (6.10)
@t t
U
6.2.2 Convection term of the x-equation

The integrand in convection term ui uE nE is represented in the table 6.1.
56
Table 6.1: nE uE andDownload

ui at different sides.
free eBooks at x-equation
bookboon.com
side nE uE ui
Unsteady term for y-equation:
Z
@ unC1 uyij
2 yij
n
uy d U D (6.10)
@t dynamics
Lectures on computational fluid t Finite Volume Method
U
6.2.2 Convection term of the x-equation

Table 6.1: nE uE and ui at different sides. x-equation

side nE uE ui
east uxe uxe
west uxw uxw
north uy n ux n
south uys uxs
The necessary velocities are approximated as shown in the table 6.2.

Herewith the convection term has the form presented in the table 6.3.
6.2.3 Convection term of the y-equation

The necessary velocities are approximated as shown in the table 6.5.
Herewith the convection term has the form presented in the table 6.6.
57
This e-book
is made with SETASIGN
SetaPDF
PDF components for PHP developers
www.setasign.com

Table 6.2: Velocities at different sides. x-equation

velocity approximation
uxe uxe D 12 .uxij C uxi C1j /
uxw uxw D 12 .uxij C uxi 1j /
ux n ux n D 12 .uxij C uxij C1 /
uxs uxs D 12 .uxij C uxij 1 /
uy n uy n D 12 .uyij C uyi C1j /
uys uys D 12 .uyij 1 C uyi C1j 1 /
Table 6.3: Convection flux. x-equation

side flux

east 4
.uxij C uxi C1j /2
west 4 .uxij C uxi 1j /2

north 4
.uxij C uxij C1 /.uyij C uyi C1j /
south 4 .uxij C uxij 1 /.uyi C1j 1 C uyij 1 /
Table 6.4: nE uE and ui at different sides. y-equation

side nE uE ui
east uxe uye
west uxw uyw
north uy n uy n
south uys uys
Table 6.5: Velocities at different sides. y-equation

velocity approximation
uxe uxe D 12 .uxij C uxij C1 /
uxw uxw D 21 .uxi 1j C uxi 1j C1 /
uy n uy n D 12 .uyij C uyij C1 /
uys uys D 12 .uyij C uyij 1 /
uye uye D 12 .uyij C uyi C1j /
uyw uyw D 12 .uyij C uyi 1j /
6.2.4 X-equation approximation

unC1 n
xij uxij
2 C
t
58
58
Table 6.6: Convection flux. y-equation

side flux

east 4
.uxij C uxij C1 /.uyij C uyi C1j /

west 4 .uxi 1j C uxi 1j C1 /.uyij C uyi 1j /

north 4
.uyij C uyij C1 /2
south 4 .uyij C uyij 1 /2

.uxij C uxiC1j /2 .uxij C uxi 1j /2 C
4 4

.uxij C uxij C1 /.uyij C uyiC1j / .uxij C uxij 1 /.uyij 1 C uyi C1j 1 /
4 4

C piC1j pij D 0
6.2.5 Y-equation approximation

nC1 n
2
uyij uyij
C
t

.uxij C uxij C1 /.uyij C uyiC1j / .uxi 1j C uxi 1j C1 /.uyij C uyi 1j /C
4 4

.uyij C uyij C1 /2 .uyij C uyij 1 /2 C
4 4

C pij C1 pij D 0
6.3 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
unC1 n
xij uxij
C
t
n
.u C unxiC1j /2 .unxij C unxi 1j /2 C
4 xij 4
n
.u C unxij C1 /.uyij
n n
C uyiC1j / .unxij C unxij 1 /.uyij
n n
1 C uyi C1j 1 /
4 xij 4

n n
C pi C1j 59pij D 0 (6.11)
nC1 n
2
uyij uyij
C
t
n n n
.u C uyiC1j /.unxij C unxij C1 / .uyij C uyin n n
1j /.uxi 1j C uxi 1j C1 /C
4 yij 4
n n 2 n n 2
.uyij C uyij C1 / .uyij C uyij 1 / C
4 4 59

n n
C pij C1 pij D 0 (6.12)
nC1 n
2
uyij uyij
C
t
Lectures
non computational fluid dynamics n Finite Volume Method
n n n n n n
.uyij C uyi C1j /.uxij C uxij C1 / .uyij C uyi 1j /.uxi 1j C uxi 1j C1 /C
4 4
n n 2 n n 2
.uyij C uyij C1 / .u C uyij 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 ij and unij are divergence free:
ıunC1
xij
nC1
ıuyij ıunxij n
ıuyij
C D 0; C D0
ıx ıy ıx ıy
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 ex-
plicitly 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
equations (6.11) and (6.12) are solved explicitly. Velocities unC1 nC1
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

Sharp
If the index is nMinds - Bright
C 1 we get implicit Ideas!
scheme
Employees at FOSS Analytical A/S are livingnC1
proof of the company value - First - using The Family owned FOSS group is
u
new inventions to make dedicated solutions unxij
2 xijfor our customers. With sharp minds and the world leader as supplier of
strive to developC
cross functional teamwork, we constantly new unique products - dedicated, high-tech analytical
Would you like to join our team? t solutions which measure and
nC1 nC1 control the quality and produc-

FOSS works diligently C unC1
with innovation
.uxij 2
xi C1j /
and development CforunC1
as basis
.uxij 2 It is
xi 1j / C
its growth. tion of agricultural, food, phar-
reflected in the 4 4 employees
fact that more than 200 of the 1200 in FOSS work with Re- maceutical and chemical produ-
search & Development in Scandinavia and USA. Engineers at FOSS work in production, cts. Main activities are initiated
development and marketing, within a wide range of different fields, i.e. Chemistry,
60 Chemometrics.
Electronics, Mechanics, Software, Optics, Microbiology,
from Denmark, Sweden and USA
with headquarters domiciled in
Hillerød, DK. The products are
We offer marketed globally by 23 sales
A challenging job in an international and innovative company that is leading in its field. You will get the companies and an extensive net
opportunity to work with the most advanced technology together with highly skilled colleagues. of distributors. In line with
the corevalue to be ‘First’, the
Read more about FOSS at www.foss.dk - or go directly to our student site www.foss.dk/sharpminds where
company intends to expand
you can learn more about your possibilities of working together with us on projects, your thesis etc.
its market position.
Dedicated Analytical Solutions

FOSS
Slangerupgade 69
3400 Hillerød
Tel. +45 70103370
www.foss.dk

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

If the index is n C 1 we get implicit scheme
2
unC1 n
xij uxij
C
t
nC1 nC1
.uxij C unC1 2
xiC1j / .u C unC1 2
xi 1j / C
4 4 xij
nC1 60 .unC1 C unC1 /.unC1 C unC1
.uxij C unC1 nC1 nC1
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)
nC1 n
2
uyij uyij
C
t
nC1 nC1 nC1 nC1 nC1 nC1 nC1 nC1
.uyij C uyi C1j /.uxij C uxij C1 / .u C uyi 1j /.uxi 1j C uxi 1j C1 /C
4 4 yij
nC1 nC1 2 nC1 nC1 2
.uyij C uyij C1 / .u C uyij 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 ij and unij are divergence free:
ıunC1
xij
nC1
ıuyij ıunxij n
ıuyij
C D 0; C D0
ıx ıy ıx ıy
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 Iterative procedure for implicit scheme

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:
@ui uj @u.m/
i uj
.m1/
D (6.15)
@xj @xj
The velocity uj is taken from the previous iteration .m1/. The system (6.11)
61
and (6.12) is rewritten in the form
61
iterative procedure. Let m be an iteration number. The nonlinear term is
represented in form:
@ui uj
@u.m/
i uj
.m1/
D (6.15)
@xj @xj
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
C
t
.m/ .m/
.uxij C u.m/ .m1/
xi C1j /.uxij C u.m1/
xi C1j / .u C u.m/ .m1/
xi 1j /.uxij C u.m1/
xi 1j /C
4 4 xij
.m/ .m/
.uxij C u.m/ .m1/
xij C1 /.uyij
.m1/
C uyi C1j / .u C u.m/ .m1/ .m1/
xij 1 /.uyij 1 C uyi C1j 1 /
4 4 xij

.m/ .m/
C pi C1j pij D 0
.m/ n
2
uyij uyij
C
t
.m/ .m/ .m1/ .m/
.uyij C uyi C1j /.uxij C u.m1/
xij C1 /
.m/
.uyij C uyi .m1/ .m1/
1j /.uxi 1j C uxi 1j C1 /C
4 4
.m/ .m/ .m1/ .m1/ .m/ .m/ .m1/ .m1/
.uyij C uyij C1 /.uyij C uyij C1 / .uyij C uyij 1 /.uyij C uyij 1 /C
4 4

.m/ .m/
C pij C1 pij D 0
or
qxi 1j u.m/ .m/ .m/ .m/ .m/

xi 1j Cqxij uxij Cqxi C1j uxi C1j Cqxij 1 uxij 1 Cqxij C1 uxij C1 C (6.16)

C 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)

Lectures on computational fluidC .m/
dynamics .m/ Finite Volume Method
pij C1 pij D ryij
where
.m1/
qxij D 2 =t C Œ.u C u.m1/
xi C1j / .uxij
.m1/
C u.m1/
xi 1j /C
4 xij62
.m1/ .m1/ .m1/ .m1/
C .uyij C uyi C1j / .uyij 1 C uyi C1j 1 /
.m1/
qxi C1j D .u.m1/
xij C u.m1/
xi C1j /; qxi 1j D .uxi 1j C uxij
.m1/
/;
4 4
.m1/ .m1/ .m1/ .m1/
qxij C1 D .uyij C uyi C1j /; qxij 1 D .uyij 1 C uyi C1j 1 /
4 4
2 .m1/ .m1/ .m1/ .m1/
qyij D C uyij C uyij C1 u yij C u yij 1 C
t 4
.m1/ .m1/
C uxij C u.m1/ xij C1 u .m1/
xi 1j C u xi 1j C1
.m1/
qyi 1j D uxi 1j C u.m1/
xi 1j C1 ; qyi C1j D u.m1/
xij C u.m1/
xij C1
4 4
.m1/ .m1/ .m1/ .m1/
qyij 1 D uyij C uyij 1 ; qyij C1 D uyij C uyij C1
4 4
n n
uxij uyij
rxij D 2 ; ryij D 2
t t
Dividing the equations (6.16) by qxij and (6.17) by qyij we obtain
axi 1j u.m/ .m/ .m/ .m/ .m/

xi 1j C uxij C axi C1j uxi C1j C axij 1 uxij 1 C axij C1 uxij C1 C (6.18)

C pi.m/
C1j pij.m/ =qxij D Rxij
.m/ .m/ .m/ .m/ .m/

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
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 (6.20)
63

xi1j C1 xij C1
4 xi1j 4 xij
.m1/ .m1/ .m1/ .m1/
qyij 1 D uyij C uyij 1 ; qyij C1 D uyij C uyij C1
4 4
n n
uxij uyij
Lectures on computational fluid D 2
rxijdynamics ; ryij D 2 Finite Volume Method
t t
Dividing the equations (6.16) by qxij and (6.17) by qyij we obtain
axi1j u.m/ .m/ .m/ .m/ .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 (6.20)
6.6 Pressure correction

63 method
The velocity field satisfying the equation (6.20) is the solution of the lin-
earized Navier Stokes equation. It doesn’t 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 so-
lution is calculated with pressure taken from the previous iteration:
u D Au C Bp .m1/ C C (6.21)

The velocity and pressure corrections
u.m/ D u C u0 ; 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)

Since u satisfies the equation (6.21) the equation for the velocity correction
reads
u0 D Au0 C Bp 0 (6.24)
The velocity at the iteration .m/ is
u.m/ D u C Au0 C Bp 0 (6.25)

It should satisfy the continuity equation
64
.mC1/
ru free eBooks
Download D 0 at bookboon.com (6.26)
what results in
u0 D Au0 C Bp 0 (6.24)
The velocity at the iteration .m/ is
u.m/ D u C Au0 C Bp 0 (6.25)
It should satisfy the continuity equation
ru.mC1/ D 0 (6.26)
what results in
ru D rBp 0 rAu0 (6.27)
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
Figure 6.2: SIMPLE algorithm.
The equation (6.28) is the Poisson equation for the pressure correction p 0 .
The solution algorithm is summarized in Fig. 6.2. Let the solution is known
at time slice n, the solution at the next time instant n C 1 is seeking. In the
first iteration all quantities are taken from the previous time instant
u.mD1/ n .mD1/
x;yij D ux;yij ; pij D pijn
At each time instant the inner loop iterations are performed until residuals
are getting smaller than some threshold
maxju.m/ .m1/
x;yij ux;yij j < "u ; maxjpijm.m/ pij.m1/ j < "p
As soon as the inner loop iterations are converged the solution at time in-
stant n C 1 is equaling to the solution from the last iteration and the next
time instant is computed. The structure of the inner loop is shown in Fig. 6.2.
65
6.7.1 Pressure correction equation
Let us consider the pressure correction equation (6.28) in details.This equa-
tion is solved for the control volume shown in Fig. 6.3. The divergency
maxju.m/ .m1/
x;yij ux;yij j < "u ; maxjpijm.m/ pij.m1/ j < "p
As soon as the inner loop iterations are converged the solution at time in-
stant n C 1 is equaling to the solution from the last iteration and the next
time instant fluid dynamics
is computed. Finite
The structure of the inner loop is shown in Fig. Volume Method
6.2.
6.7.1 Pressure correction equation

Let us consider the pressure correction equation (6.28) in details.This equa-
tion is solved for the control volume shown in Fig. 6.3. The divergency
operator rf D @f x
C @fy
is represented within Finite Volume Method as
Z Z Z
@x @y
follows @fx @fy
Z rf d U D Z . C /d U D Z fndS D .fxe fxw Cfy n fys / (6.29)
U U @xx
@f @f
@yy
rf d U D . C /d U D S65 fndS D .fxe fxw Cfy n fys / (6.29)
U U @x @y 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 1 / (6.30)
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
As of the
follows fromcontrol and (6.19) the operator Bp 0 has the following values
(6.18) volume
at faces of the control volume

.Bp /xij D pi C1j pij =qxij ; .Bp /xi 1j D pij pi 1j =qxi 1j
0 0 0 0 0 0
.Bp 0 /xij D pi0 C1j pij0 =qxij ; .Bp 0 /xi 1j D pij0 pi0 1j =qxi 1j

.Bp /yij D pij C1 pij =qyij ; .Bp /yij 1 D pij pij 1 =qyij 1 :
0 0 0 0 0 0
.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

The Wake
pi0 C1j pij0 =qxij pij0 pi0 1j =qxi 1j C pij0 C1 pij0 =qyij
pi0 C1j
the only
0
pijemission
=qxij wepij0 pi0 1jto =q
want xi 1j C
leave pij0 C1 pij0 =qyij
behind

pij pij 1 =qyij 1 D .uxij uxi 1j C uyij
0 0
uyij 1 / (6.32)
0 0
pij pij 1 =qyij 1 D .uxij uxi 1j C uyij uyij 1 / (6.32)
or
or
ˇi C1j pi0 C1j C ˇij pij0 C ˇi 1j pi0 1j C ˇij C1 pij0 C1 C ˇij 1 pij0 1 D cij (6.33)
where
where

.QYURGGF'PIKPGU/GFKWOURGGF'PIKPGU6WTDQEJCTIGTU2TQRGNNGTU2TQRWNUKQP2CEMCIGU2TKOG5GTX

1 1 1 1
cij D .u
xij uxi 1j Cuyij uyij 1 /; ˇij D
C C
6JGFGUKIPQHGEQHTKGPFN[OCTKPGRQYGTCPFRTQRWNUKQPUQNWVKQPUKUETWEKCNHQT/#0&KGUGN6WTDQ C ;
1
qxij qxi11j qyij 1
2QYGTEQORGVGPEKGUCTGQHHGTGFYKVJVJGYQTNFoUNCTIGUVGPIKPGRTQITCOOGsJCXKPIQWVRWVUURCPPKPI
qyij1 1
cij D .u xij uxi 1j Cuyij uyij 1 /; ˇij D
HTQOVQM9RGTGPIKPG)GVWRHTQPV
C C C ;
qxij qxi 1j qyij qyij 1
1
(KPFQWVOQTGCVYYYOCPFKGUGNVWTDQEQO 1 1 1
ˇi C1j D ; ˇi 1j D ; ˇij C1 D ; ˇij 1 D :
q 1 q 1 q 1 q 1
ˇi C1j D xij ; ˇi 1j D xi 1j ; ˇij C1 D yij ; ˇij 1 D yij 1 :
66
66

xij i C1j ij xij xi 1j ij i 1j xi 1j

0
.Bpon
Lectures /yij D pij0 C1
computational pij0dynamics
fluid =qyij ; .Bp 0 /yij 1 D pij0 pij0 1 1 : Volume Method
=qyijFinite
(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

pij0 pij0 1 =qyij 1 D .uxij uxi 1j C uyij

uyij 1 / (6.32)
or
where

1 1 1 1
cij D .uxij uxi 1j
Cuyij
uyij 1 /; ˇij D C C C ;
1 1 1 1
ˇi C1j D ; ˇi 1j D ; ˇij C1 D ; ˇij 1 D :
6.7.2 Summary of the SIMPLE

66 algorithm
We introduce one dimensional numbering instead of two dimensional one
according to the rule
˛ D .i 1/Ny C j
Let the solution at the time instant n be known. The task is to find the
solution at the time n C 1. At each time instant the guess solution is taken
from the previous time instant:
u.1/ n .1/ n
x;y˛ D ux;y˛ ; p˛ D p˛
The solution is found within the next substeps:
i) Calculation of the auxiliary velocity ux;y˛ from two independent sys-

tems of linear algebraic equations:
ax˛Ny ux˛Ny C ux˛ C ax˛CNy ux˛CNy C ax˛1 ux˛1 C ax˛C1 ux˛C1 C

.m1/ .m1/
C p˛CNy p˛ =qx˛ D Rx˛

ay˛Ny uy˛Ny
C uy˛ C ay˛CNy uy˛CN y
C ay˛1 uy˛1 C ay˛C1 uy˛C1 C

.m1/
C p˛C1 p˛.m1/ =qy˛ D Ry˛
ii) Calculation of the pressure correction p˛0 from the system of linear
algebraic equations: 67
0
ˇ˛CNy p˛CNy
C ˇ˛ p˛0 Download
C ˇ˛Nfree 0 eBooks at bookboon.com
0 0
y p˛Ny C ˇ˛C1 p˛C1 C ˇ˛1 p˛1 D c˛
0
tems of linear algebraic equations:
ax˛Ny ux˛Ny C ux˛ C ax˛CNy ux˛CNy C ax˛1 ux˛1 C ax˛C1 ux˛C1 C

.m1/ Finite Volume Method
C p˛CNy p˛.m1/ =qx˛ D Rx˛

ay˛Ny uy˛Ny
C uy˛ C ay˛CNy uy˛CN y
C ay˛1 uy˛1 C ay˛C1 uy˛C1 C

.m1/ .m1/
C p˛C1 p˛ =qy˛ D Ry˛
ii) Calculation of the pressure correction p˛0 from the system of linear
algebraic equations:
0
ˇ˛CNy p˛CNy
C ˇ˛ p˛0 C ˇ˛Ny p˛N
0
y
0
C ˇ˛C1 p˛C1 0
C ˇ˛1 p˛1 D c˛
iii) Calculation of the velocity correction u0x;y˛ :

u0x˛ D 0
p˛CNy
p˛0 =qx˛

0 0
uy˛ D p˛C1 p˛0 =qy˛
iv) Correction of the velocity and pressure:

0
u.m/
x;y˛ D ux;y˛ C ux;y˛ ; p˛.m/ D p˛.m1/ C p˛0
67
Figure 6.3: Control volume used for the pressure correction equation.
v) Check the difference between two iterations:
maxju.m/ .m1/
x;y˛ ux;y˛ j < "u ; maxjp˛.m/ p˛.m1/ j < "p
If these conditions are not fulfilled then

68
u.m1/ D u.m/
x;y˛Download
x;y˛
.m1/
; peBooks
free ˛ at p˛.m/
Dbookboon.com
and go to the step i). Otherwise the calculation at the time moment
Lectures
Figureon6.3:
Control volume used for the pressure correction equation.Finite Volume Method
v) Check the difference between two iterations:
maxju.m/ .m1/
x;y˛ ux;y˛ j < "u ; maxjp˛.m/ p˛.m1/ j < "p
If these conditions are not fulfilled then
u.m1/ .m/ .m1/

x;y˛ D ux;y˛ ; p˛ D p˛.m/
and go to the step i). Otherwise the calculation at the time moment
n C 1 is completed
unC1 .m/ nC1

x;y˛ D ux;y˛ ; p˛ D p˛.m/
and one proceeds to the next time instant n C 2.
68
Challenge the way we run
EXPERIENCE THE POWER OF

FULL ENGAGEMENT…
RUN FASTER.
RUN LONGER.. READ MORE & PRE-ORDER TODAY
RUN EASIER… WWW.GAITEYE.COM
1349906_A6_4+0.indd 1 22-08-2014 12:56:57

Chapter 7
Overview of pressure correction methods
Chapter
Chapter 7
Overview7 of pressure correction
Overview
Overview of
methods of pressure
pressure correction
correction
methods
methods
7.1 SIMPLE algorithm
7.1linearized
The SIMPLE algorithm
Navier Stokes equation written in operator form is
7.1 SIMPLE algorithm
The linearized Navier Stokesuequation
D Au Cwritten
Bp C Cin operator form is (7.1)
The linearized Navier Stokes equation written in operator form is
Within the SIMPLE algorithm u Dthe
Ausolution
C Bp C is
C seeking at each time step in
(7.1)
form of the loop: u D Au C Bp C C (7.1)
Within the SIMPLE algorithm the solution is seeking at each time step in
Within
form of the SIMPLE algorithm the solution is seeking at each time step in
the loop:

form of the loop: of the auxiliary velocity
Calculation
u D Au
Calculation of the auxiliary
C Bp .m1/ C C
velocity (7.2)
Calculation of the auxiliary velocity
u D Au C Bp .m1/ C C (7.2)
Calculation of the pressure
u D correction
Au C Bp .m1/ C C (7.2)
rBp 0 D ru
Calculation of the pressure correction (7.3)
Calculation of the pressure correction
rBp 0 D ru (7.3)
Calculation of the velocity correction
rBp 0 D ru (7.3)
u0 D Bp 0
Calculation of the velocity correction (7.4)
Calculation of the velocity correction
u0 D Bp 0 (7.4)
Correction u0 D Bp 0 (7.4)
u D u C u0 ; p .m/ D p .m1/ C p 0
.m/
(7.5)
Correction
Correction u.m/ D u C u690
; p .m/ D p .m1/ C p 0 (7.5)
7.2 PISO algorithm u D u C u ; p .m/ D p .m1/ C p 0
.m/ 0
(7.5)
In the SIMPLE algorithm we neglected 69 the term rAu0 (see 6.25). In PISO
69
algorithm this term is taken into account. Actually the term rAu0 can not
be calculated before the velocity correction is computed. Therefore, the term
is taken into account in an iterative way.
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)

The velocity correction is then
70
u0 free
Download D Bp 0
eBooks at bookboon.com (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 Overview of pressure correction methods
0
rBp D ru (7.6)
The velocity correction is then
u0 D Bp 0 (7.7)
7.2.2 Second iteration

The pressure correction within the second iteration is found from the Poisson
equation
rBp 00 D rAu0 (7.8)

The velocity correction within the second iteration is then
u00 D Au0 C Bp 00 (7.9)
7.2.3 Correction
Corrected velocities and pressure are
u.m/ D u C u0 C u00 ; p .m/ D p .m1/ C p 0 C p 00 (7.10)

Using formula derived above
rBp 0 D ru ; u0 D Bp 0 ; u00 D Au0 C Bp 00 ; rBp 00 D rAu0
it is easy to prove that the velocity u.m/ satisfies the continuity equation.
Now we prove the equation u.m/ D Au.m/ C Bp .m/ C C :
u C u0 C u00 D A.u C u0 C u00 / C B.p .m1/ C p 0 C p 00 / C C (7.11)
Since 70

u D Au C Bp .m1/
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
The PISO algorithm can be summarized as follows:
u D Au C Bp .m1/ C C (7.12)
Calculation of the pressure correction p 0 :
rBp 0 D ru (7.13)

71
Calculation of the velocity correction u0
Calculation
Calculation of
of the
the auxiliary
auxiliary velocity
velocity
.m1/
u D
u Au C
D Au Bp .m1/ C
C Bp CCC (7.12)
(7.12)
.m1/
u D
Lectures on computational fluid dynamics Au C Bp C COverview of pressure(7.12)
correction methods
0
Calculation
Calculation of
of the
the pressure
pressure correction p 0 ::
correction p
0
rBp 0 D
rBp D ru
ru (7.13)
(7.13)
rBp 0 D ru (7.13)
0
Calculation
Calculation of
of the
the velocity
velocity correction
correction u
u0
0 0
u0 D
u Bp 0
D Bp (7.14)
(7.14)
u0 D Bp 0 (7.14)
00
Calculation
Calculation of
of the
the pressure
pressure correction p 00 ::
correction p
00 0
rBp 00 D
rBp D rAu
rAu0 (7.15)
(7.15)
rBp 00 D rAu0 (7.15)
00
Calculation
Calculation of
of the
the velocity
velocity correction
correction uu00
00 0 00
u00 D
u Au0 C
D Au Bp 00
C Bp (7.16)
(7.16)
u00 D Au0 C Bp 00 (7.16)
Correction
Correction
Correction
0 00 0 00
u.m/
u D u C u C u ;p
.m/
Dp
.m1/
.m/ D u C u0 C u00 ; p .m/ D p .m1/ C p 0 C p 00
Cp Cp (7.17)
(7.17)
u.m/ D u C u0 C u00 ; p .m/ D p .m1/ C p 0 C p 00 (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
Technical 71training 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 OIL & GAS
experience in training technical and engineering staff and ENGINEERING
have trained people in organisations such as General
ELECTRONICS
Motors, Shell, Siemens, BHP and Honeywell to name a few.
Our onsite training is cost effective, convenient and completely AUTOMATION &
customisable to the technical and engineering areas you want PROCESS CONTROL
covered. Our workshops are all comprehensive hands-on learning
experiences with ample time given to practical sessions and MECHANICAL
demonstrations. We communicate well to ensure that workshop content ENGINEERING
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
For a no obligation proposal, contact us today POWER
at training@idc-online.com or visit our website
for more information: www.idc-online.com/onsite/
Phone: +61 8 9321 1702

Email: training@idc-online.com
Website: www.idc-online.com

Lectures on computational fluid dynamics Overview of pressure correction methods
7.3 SIMPLEC algorithm

Another way to hold the term rAu0 (see 6.25) is implemented in the SIM-
PLEC algorithm. The velocity correction at ˛ th control volume u0˛ can
be calculated using the interpolation over N adjacent control volumes:
X
N X
N
Aˇ u0ˇ u0˛ Aˇ (7.18)
ˇ D1 ˇ D1
where ˇ is the number of adjacent control volumes around the control volume
with the number ˛.
PN 0
ˇ D1 Aˇ uˇ
u0˛ PN (7.19)
ˇ D1 Aˇ
The equation for the velocity correction is
u0 D Au0 C Bp 0 (7.20)
or
X
N
u0˛ D Aˇ u0ˇ C Bp 0 (7.21)
ˇ D1
Substitution of (7.19) into (7.21) yields
X
N
u0˛ D u0˛ Aˇ C Bp 0 (7.22)
ˇ D1
and
Bp 0
u0˛ D PN (7.23)
1 ˇ D1 Aˇ
The pressure correction equation
rBp 0 D ru rAu0

takes the form
0 Bp 0
rBp D ru rA PN
1 ˇ D1 Aˇ
or

B
r B CA PN72 p 0 D ru (7.24)
1 ˇ D1 Aˇ
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).
73
Chapter 8
Lectures on computational fluid dynamics Computational grids
Chapter
Chapter 8
8
Computational grids
Computational
Computational grids
grids
8.1 Grid types
8.1computational
The Grid types grids are subdivided into:
8.1 Grid types
The computational
structured grids (seeare
grids Fig. 8.1a), into:
subdivided
The computational grids are subdivided into:
block structured
structured gridsFig.
grids (see (see8.1a),
Fig. 8.1b),
structured grids (see Fig. 8.1a),
unstructured
block structured grids (see(see
grids Fig.Fig.
8.1c).
8.1b),
block structured grids (see Fig. 8.1b),
Disadvantage of thegrids
unstructured structured grid
(see Fig. is shown in Fig. 8.2. Refinement of the
8.1c).
gridclose to the wallgrids
unstructured results inFig.
(see the refinement
8.1c). in areas where this refinement is
Disadvantage
not necessary. of thedisadvantage
This structured gridcanisbeshown in Fig.
overcome 8.2. ofRefinement
by use of the
block-structured
Disadvantage
grid
(Fig. close toand
8.1b) ofwall
the the results
structured
unstructured grid
in grid
the is shown
8.1c).ininareas
refinement
(Fig. Fig. where
8.2. Refinement of the
this refinement is
grid close to the wall results in the refinement in areas where this
not necessary. This disadvantage can be overcome by use of block-structured refinement is
not necessary.
(Fig. 8.1b) andThis disadvantage
unstructured gridcan be 8.1c).
(Fig. overcome by use of block-structured
(Fig. 8.1b) and unstructured grid (Fig. 8.1c).
Figure 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.
Figure 8.1: Samples of a) structured grid for an airfoil, b) block structured
Figure
The
grid for8.1:
quality Samples
of the
cylinder of a)has
grid
in channel structured
a strong
and grid for an
impact
c) unstructured airfoil,
ongrid
the for b)
an block
accuracy of structured
airfoil.numerical
grid for cylinder
prediction. in channel
The change of theand
cellc)topology
unstructured
withingrid
the for an airfoil. domain
computational
The quality of the grid has a strong impact on the accuracy of numerical
The qualityThe
prediction. of the gridofhas
change theacell
strong 75impact
topology on the
within accuracy of numerical
the computational domain
prediction. The change of the cell topology within the computational domain
75
75
Figure 8.2: Illustration of structured grid disadvantage.
should be smooth especially at the border between different grid blocks. The
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
special grid boundary layers close to walls.
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
Lectures boundaryfluid
computational layers close to walls.
dynamics Computational grids
8.2 Overset or Chimera grids

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-
metrical entity in the computational domain. After that the grids are com-
bined together in such a way that they overlap each other where they meet.
The crucial operation is an accurate transfer of quantities between the dif-
ferent 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 indepen-
dent 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,
u n
@E o 1
C uE UEg r uE D fE rp C E
u (8.1)
@t
Thomas and Lombard have shown that the function UEg can not be arbi-
trary 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 S
Where U and S are respectively volume and surface of cells. The equa-
tion (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, in-
dependently of the deformation of the grid. The equation (8.2) is satisfied
automatically if the control volumes don’t change their shape. The Geomet-
75
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
@
d U UEg nE dS D 0
Lectures on computational fluid dynamics Computational
(8.2) grids
@t
U S
Where U and S are respectively volume and surface of cells. The equa-
tion (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, in-
dependently of the deformation of the grid. The equation (8.2) is satisfied
automatically if the control volumes don’t change their shape. The Geomet-
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]
and [8].
I wanted real responsibili� www.discovermitas.com
Maersk.com/Mitas �e G
I joined MITAS
77 because for Engine
I wanted real responsibili� Ma
Month 16
I was
the North Sea super
advising and the No
Real work he
International
al opportunities
Internationa
�ree wo
work
Real work he
helping fo
International
�ree wo
work

Part II
Mathematical modelling of
turbulent flows
77
79 at bookboon.com
Chapter 9
Lectures on computational fluid dynamics Physics of turbulence
Chapter
Chapter 9
9 turbulence
Physics of
Physics
Physics of
of turbulence
turbulence
9.1 Definition of the turbulence
9.1 Definition
Flow motions are subdividedof the turbulence
into laminar flows and turbulent ones. The
9.1 Definition
word ” Laminar” in Greek ofmeans
thelayer.
turbulence
The fluid particles move orderly
Flow
in motions
layers are intense
without subdivided intomixing.
lateral laminarThe flows and turbulent
disruption betweenones. The
layers is
Flow
word motions
” are
Laminar” subdivided
in Greek into
means laminar
layer. Theflows and
fluid turbulent
particles ones.
move
absent. On the contrary the turbulent flow is very chaotic with strong eddies The
orderly
word ” Laminar”
in layers
and without
intense inacross
Greek
mixingintense themeans
lateral layer. The
mixing.
flow. The disruption
fluid particles movelayers
between orderly
is
in layers
absent. without
On the intense
contrary lateral
the mixing.
turbulent flow The
is disruption
very chaotic
Turbulent motion is the three dimensional unsteady flow motion with between
with layers
strong is
eddies
absent. On the contrary the turbulent
and intense mixing across the flow. flow is very chaotic with strong eddies
and intense
Turbulent mixing
motion
chaotical isacross the
the three
trajectories flow.
dimensional
of fluid particles,unsteady flow motion with
Turbulent motion is the three dimensional unsteady flow motion with
fluctuations of the velocity
chaotical trajectories and
of fluid particles,
chaotical trajectories of fluid particles,
strong mixing
fluctuations of the velocity and
fluctuations of the velocity and
strong
arisen mixing
at large Re numbers due to unstable vortex dynamics.
strong mixing
arisen at large Re numbers due to unstable vortex dynamics.
9.2 atVortex
arisen dynamics
large Re numbers due to unstable vortex dynamics.
9.2vortex
The Vortex
dynamics dynamics
is the key to understand what happens in the turbulent
9.2
flow. Vortex dynamics
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
flow. Vorticity transport equation
The vectorVorticity
9.2.1 calculus relation reads: equation
transport
9.2.1 Vorticity transport equation
The vector calculus 1relation reads:
r.A A/ D A .r A/ C .Ar/A (9.1)
The vector calculus 2relation reads:
1
Taking u D A we get:12 r.A A/ D A 81 .r A/ C .Ar/A (9.1)
r.A A/ D A .r A/ C .Ar/A (9.1)
2
1
r.u u/ D u81 ! C .ur/u (9.2)
2 81
where ! D r u is the vorticity.
Application of the curl operator to (9.2) results in
r ..ur/u/ D r .u !/ D u.r!/ C !.ru/ .!r/u C .ur/! (9.3)
Here we used the identity r . 21 r.u u// D 0.

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)

78
Let us apply the curl operator to thefree
Download Navier
eBooksStokes equation
at bookboon.com
@u 1
r . C .ur/u/ D r . rp C u/ (9.5)
Application of the curl operator to (9.2) results in
r ..ur/u/ D r .u !/ D u.r!/ C !.ru/ .!r/u C .ur/! (9.3)

Here we used the identity r . 12 r.u
u// D 0.
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)

Let us apply the curl operator to the Navier Stokes equation
@u 1
r . C .ur/u/ D r . rp C u/ (9.5)
@t
@!
C r ..ur/u/ D .r u/ D ! (9.6)
@t
Substituting (9.4) into (9.6) results in
@!
C .ur/! D .!r/u C ! (9.7)
@t
D!
D .!r/u C ! (9.8)
Dt
The equation (9.8) is the vorticity transport equation.
9.2.2 Vorticity and vortices

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).
82
www.job.oticon.dk

@t
D!
D .!r/u C ! (9.8)
Dt
The equation (9.8) is fluid
the dynamics
vorticity transport equation. Physics of turbulence
9.2.2 Vorticity and vortices

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)
The consequence of the condition (9.9) is:
All vortex lines, defined as the lines which are tangential to the vorticity
vector ! d l D 0, are closed in the three dimensional case (Fig. 9.3).
The velocity induced by vorticity ! occupied the volume U are calculated

from the Biot-Savart law:
Z
1 ! .x r/
u.x; t/ D dU (9.10)
4 U jx rj3
The velocities induced by two dimensional and three dimensional vortex
structures are shown in Fig. 9.4. An important fact is the appearance of
self induced velocities on curvilinear three dimensional vortex structures.
They are responsible for leapfrog vortex ring motion (http://www.lemos.uni-
rostock.de/galerie/). The self induced velocities is the reason for convective
instability of three dimensional vortex structures.
Figure 9.1: Vorticity and vortices.
9.2.3 Vortex amplification as an important mechanism

of the turbulence generation
80
As mentioned above the vortices are main players during the laminar- tur-
bulent transition. The vorticity can be essentially intensified (amplified) due
to action of neighboring vortices or even due to self induction.
Figure 9.1: Vorticity and vortices.
9.2.3 Vortex amplification as an important mechanism

of the turbulence generation
As mentioned above the vortices are main players during the laminar- tur-
bulent transition. The vorticity can be essentially intensified (amplified) due
to action of neighboring vortices or even due to self induction.
83
Figure 9.2: Tornado.
Figure 9.3: Vortices in two-dimensional and three dimensional cases.
The reason for that can be explained by analysis of the vorticity transport
equation
D!
D .!r/u C ! (9.11)
Dt
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 vortic-
ity in the space. The term .!r/u is responsible for the amplification of the
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
fluid stretching area @u
@x
x
> 0, the term !x @u@x
x
> 0 is positive. As a result,
D!x
Dt
> 0 is positive, what leads to the increase of the vorticity !x . As shown
analytically by Novikov [9] for a simple model problem, the vortex strength
81
84
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 vortic-
ity in the space. The term .!r/u is responsible for the amplification of the
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
fluid stretching area @u
@x
x
> 0, the term !x @u@x
x
> 0 is positive. As a result,
D!x
Dt
> 0 is positive, what leads to the increase of the vorticity !x . As shown
analytically by Novikov [9] for a simple model problem, the vortex strength
84
Figure 9.4: Velocities induced by vortices. Three dimensional curvilinear

vortices induce self induced velocities.
of vortons structures can increase exponentially up to the infinity without

viscosity effects. The vorticity growth caused by inviscid amplification term
is counterbalanced by the diffusion term. Two terms on r.h.s. of (9.11) com-
pete with each other.
R In the inviscid fluid the circulation of
how willvortex
the goods be core
transported? What re-
is constant D S !x dS D const . Increase of !x results insourcesthe decrease
will we use, of
and how many will
the cross section S . The vortex becomes thinner. The diffusionwe need? The against
acts passenger and freight traf-
and makes the vortex thicker. In some flow regions the amplification can be
stronger that diffusion. The thin vortex losses the stability movement.
and is folded.
We develop components and
As shown by Chorin [10] and [11] the folding is necessary that mechanism
operate morepre-
cleanly and more ef-
ficiently than
venting the exponential growth of vorticity. If the amplification is too strong,ever before. We are also
the vorticity goes to infinity and the energy is not kept constant. Chorin [10]
notes that ”as the vortices stretch, their cross-section decreases and the en-
ergy associated with them would increase unless they arranged corporate, and publicin
themselves use. The challeng-
such a way that their velocity canceled. The foldings achieves es aresuch
great. cancela-
We deliver the solutions and
tion”. This could be easily explained using a simple sample.offer If challenging
we havejobs.just
one straight infinite vortex it induces the velocity in the plane perpendicular
to its axis. If this vortex is tangled the vortex pieces with different vorticity
direction are approaching close each to other canceling their induction (see
Fig. 9.5).
Chorin [11] explicitly specifies typical scales of folding: ” the inertial range 1
properties are due to the appearance of folded vortex tubes, which behave on
large scales as self-avoiding walks, and on small scales contain a large num-
1
this term will be introduced in the next chapter
85 82 Click on the ad to read more

Figure 9.4: Velocities induced by vortices. Three dimensional curvilinear
Lectures
vorticesoninduce
self induced velocities. Physics of turbulence
of vortons structures can increase exponentially up to the infinity without

viscosity effects. The vorticity growth caused by inviscid amplification term
pete with each other.
R In the inviscid fluid the circulation of the vortex core
is constant D S !x dS D const . Increase of !x results in the decrease of
the cross section S . The vortex becomes thinner. The diffusion acts against
stronger that diffusion. The thin vortex losses the stability and is folded.
As shown by Chorin [10] and [11] the folding is necessary mechanism pre-
venting the exponential growth of vorticity. If the amplification is too strong,
notes that ”as the vortices stretch, their cross-section decreases and the en-
ergy associated with them would increase unless they arranged themselves in
such a way that their velocity canceled. The foldings achieves such cancela-
tion”. This could be easily explained using a simple sample. If we have just
to its axis. If this vortex is tangled the vortex pieces with different vorticity
direction are approaching close each to other canceling their induction (see
Figure
Fig. 9.5).9.4: Velocities induced by vortices. Three dimensional curvilinear
vortices
Chorin [11]induce self induced
explicitly velocities.
specifies typical scales of folding: ” the inertial range 1
properties are due to the appearance of folded vortex tubes, which behave on
of vortons
large scalesstructures can increase
as self-avoiding exponentially
walks, and up tocontain
on small scales the infinity without
a large num-
viscosity
1
effects. The vorticity growth caused by inviscid amplification term
pete with each other.R In the inviscid fluid the circulation of the vortex core
is constant D S !x dS D const . Increase 85 of !x results in the decrease of
the cross section S . The vortex becomes thinner. The diffusion acts against
stronger that diffusion. The thin vortex losses the stability and is folded.
As shown by Chorin [10] and [11] the folding is necessary mechanism pre-
venting the exponential growth of vorticity. If the amplification is too strong,
notes that ”as theFigure 9.5: stretch,
vortices Illustration
theirofcross-section
the vortex folding.
decreases and the en-
ergy associated with them would increase unless they arranged themselves in
such
ber ofa folds
way that their velocity
(=hairpins) that arecanceled.
neededThe foldings
to satisfy theachieves suchofcancela-
constraint energy
tion”. This could be easily explained using a simple sample. If we have just
conservation”.
to its axis. If this
Summarizing vortexmentioned
all effects is tangled above
the vortex pieces
one can with different
imagine vorticity
the following sce-
direction are approaching close each to other canceling their induction
nario (see Fig. 9.6). Let the vorticity at a certain point of the flow grows. (see
Fig. 9.5).
In the real physical process folding prevents the growth of the kinetic energy
Chorin [11] explicitly
and increases specifieseffect
the canceling typical scales of folding:
of viscosity. ” the
Then the inertial
vortex range 1
structures
properties
breaks down areinto
duesmall
to thestructures
appearance of to
due folded vortex tubes,
reconnection which behave
mechanism on
described
large
in thescales as self-avoiding walks, and on small scales contain a large num-
next subsection.
1
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
Lectures
breaks on computational
down into smallfluid dynamics due to reconnection mechanism described
structures Physics of turbulence
in the next subsection.
Figure 9.6: Scenario of vortex amplification.
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
Figure 9.7: Scenario of vortex reconnection.
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
84
can also lead to enlargement of small vortices if two rings approach each to
other as shown in Fig. 9.7 (see red circle).
Download Inatthis
free eBooks case the energy of small
bookboon.com
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
Figurefluid dynamics
9.7: Scenario of vortex reconnection. 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
surpasses the backward one.
9.2.5 Richardson poem (1922)

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,
In the past 5 years we have drilled around 87
95,000 km
Who are we?

gas industry.

n Operations

careers.slb.com

other as shown in Fig. 9.7 (see red circle). In this case the energy of small
surpasses the backward fluidone.
dynamics Physics of turbulence
9.2.5 Richardson poem (1922)

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 air-
Figure
plane. 9.8: Sample of the vortex reconnection of tip vortices behind an air-
plane.
Figure 9.8: Sample of the vortex reconnection of tip vortices behind an air-
plane.
Which feed on their velocity;
AndWhich feed onhave
little whorls theirlessor
velocity;
whorls,
And Which
little feed
whorlson their
have velocity;
lessor whorls,
And so on to viscosity (in the molecular sense).
And
And so on little whorls (in
to viscosity havethe
lessor whorls,sense).
molecular
And so on to viscosity (in the molecular sense).
9.2.6 Summary
9.2.6 Summary
9.2.6 Summary
Vortex arise in the fluid due to viscosity effects. They experience instability
Vortex arise in the fluid
and amplification. due to
Diffusion viscosity effects.
counteracts They experience
the amplification. If theinstability
Reynolds
and amplification.
Vortex arise in the Diffusion
fluid due counteracts
to viscosity the amplification.
effects. They If the
experience
number is large, the vortex structures are strong and concentrated. Reynolds
instability
The
number
and is large,
amplification. the vortex
Diffusion structures
counteracts are
the strong and concentrated.
amplification.
amplification can dominate at some fluid region over the diffusion. If the The
Reynolds
The
amplification
number can the
is large,
vortex instability isdominate
vortex
not dampedat some
structuresfluid
are region
by viscosity. Theover
strong flow the
and diffusion.
concentrated.
becomes The
stochastic
vortex instability
amplification
due to mutual can isdominate
not damped
interaction by viscosity.
at some
of unstable The
fluid region
vortices. flow
Theover becomes
the stochastic
diffusion.
big vortices break The
down
due to
vortex mutual
instabilityinteraction
is not of
dampedunstable
by vortices.
viscosity. The
The big
flow vortices
becomes
into small ones by means of vortex reconnection. The vortex instability break down
stochastic
into
due small
to
process mutualonesinteraction
by as
is identified means
the ofofunstable
vortex reconnection.
turbulence. vortices. The bigThe vortexbreak
vortices instability
down
process is identified as the turbulence.
into small ones by means of vortex reconnection. The vortex instability
process is identified as the turbulence.
9.3 Experimental observations
9.3 Experimental observations
9.3
DifferentExperimental observations
regimes of the fluid motion were revealed very early, perhaps, in
Different regimes
antique times. of the
Much fluid
later, motion da
Leonardo were revealed
Vinci very early,
recognized perhaps,
two states in
of the
antique
Differenttimes.
fluid motion Much
regimes
and of later,
the fluid
introduced Leonardo
termda
motion
the were Vinci
”la recognized
revealed two states
very early,
turbolenza”. Arkady of the
perhaps,
Tsinoberin
fluid
antiquemotion
times.and introduced
Much later, the term
Leonardo da ”la turbolenza”.
Vinci recognized Arkady
two
in his book ”An informal introduction to turbulence” [1] presented most Tsinober
states of the
in hismotion
fluid book results
outstanding ”An
and informal
introduced introduction
in turbulencethe research
term ”latointurbulence”
turbolenza”. [1]Arkady
chronological presented
order most
Tsinober
(Fig. 9.9).
outstanding results in turbulence research in chronological order
in his book ”An informal introduction to turbulence” [1] presented most (Fig. 9.9).
outstanding results in turbulence research 88 in chronological order (Fig. 9.9).
88 86
88 eBooks at bookboon.com
Download free
Figure 9.9: Most outstanding results in turbulence research according to [1].
9.3.1 Laminar - turbulent transition in pipe. Experi-

ment of Reynolds
Quantitative study of turbulence was started by Osborne Reynolds (1842-
1912) 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 pre-
decessors 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 experimen-
tal setup of Reynolds has been preserved at the University of Manchester in
89
87
Figure 9.10: Sketch of the Reynolds experiment.
Figure 9.11: Development of instability during the laminar- turbulent tran-

sition in the circular pipe (taken from [1]).
UK. The experiments done nowadays shown that the laminar- turbulent tran-
sition 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 Laminar - turbulent transition and turbulence in
jets
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 noz-
zle the flow is laminar. The instability is developed
88 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 concen-
transition fluid dynamics
can be delayed up to Re 40000:::50000. Physics of turbulence
9.3.2 Laminar - turbulent transition and turbulence in

jets
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 noz-
zle 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 concen-
trated 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).
branches? Kick-start your career with an
English-taught master’s degree.
liu.se/master
Figure 9.12: Development of instability in the jet (taken from [1]).
Click here!
91

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).
Figure 9.12: Development of instability in the jet (taken from [1]).
91
Figure 9.13: Development of instability in the free jet.
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 vor-
tices 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. 90 9.16 shows the effect of the
acoustic impact on jet. The original vortices (lower picture) are split into
small ones (upper picture). Download free eBooks at bookboon.com
High resolution laser diagnostics methods LIF and PIV allow to get a deep
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 vor-
tices 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 Fluores-
cence (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 9.15: Vortex structures in a free jet in a far field.

Figure 9.16: Vortex structures in a free jet with acoustic impact.[46].

Figure 9.16: Vortex structures in a free jet with acoustic impact.
Figure 9.16: Vortex structures in a free jet with acoustic impact.
91
was selected for high resolution PLIF measurements. The vortex microstruc-
was selected for high resolution PLIF measurements. The vortex microstruc-
was
turesselected for highinresolution
are presented the lowerPLIF
Figure.measurements. The vortex
A very important microstruc-
observation is the
tures are presented in the lower Figure. A very important observation is the
Figure 9.17: Vortex structures in a confined jet mixer flow. [47].
9.3.3 Laminar - turbulent transition in wall bounded

flows
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 sec-
ondary 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 ve-
locity fluid regions (see Fig. 9.20). They arise due to induction of lambda
94
92
Figure 9.18: Fine vortex structures in a confined jet mixer flow. PLIF mea-
surements by Valery Zhdanov (LTT Rostock). Spatial resolution is 31m.
structures schematically shown in Fig. 9.21.
9.3.4 Distribution of the averaged velocity in the tur-

bulent boundary layer
A remarkable feature of the turbulent boundary flow is the presence of three
typical velocity distribution regions. The instantaneous velocity distributions

678'<)25<2850$67(5©6'(*5((
can be quite different (see Fig. 9.24). However, the averaged velocity has
typical distribution close to the wall regardless of the flow type.
Let us introduce the following designations:
yLQJDQGQDWXUDOVFLHQFHVDUFKLWHFWXUHWHFKQRORJ\UHODWHGPDWKHPDWLFDOVFLHQFHV
is the distance from the wall,
w is the stress at the wall,
q
w
D
u9LVLWXVRQ&KDOPHUVVHRU1H[W6WRS&KDOPHUVRQIDFHERRN

is the friction velocity,
yC D u y

is the dimensionless wall distance.
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

Figure 9.18: Fine vortex structures in a confined jet mixer flow. PLIF mea-
surements by Valery Zhdanov (LTT Rostock). Spatial resolution is 31m.
structures fluidshown
schematically dynamics
in Fig. 9.21. Physics of turbulence
9.3.4 Distribution of the averaged velocity in the tur-

bulent boundary layer
A remarkable feature of the turbulent boundary flow is the presence of three
typical velocity distribution regions. The instantaneous velocity distributions
can be quite different (see Fig. 9.24). However, the averaged velocity has
typical distribution close to the wall regardless of the flow type.
Let us introduce the following designations:
y is the distance from the wall,
w is the stress at the wall,

q
u D w is the friction velocity,
yC D u y

is the dimensionless wall distance.
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
Figure 9.19: Scenario of laminar turbulent transition in the boundary layer

on a flat plate.
Figure 9.20: Streaks visualized by hydrogen bubbles in the boundary layer

on a flat plate.
Close to the wall the turbulent fluctuations are weak. The laminar stress
l dominates over the turbulent one t , i.e. l . We consider the thin
boundary layer, i.e. the stress is approximately equal to the wall stress w :
94
Download eBooks
free w at bookboon.com (9.13)
Applying the Newton hypothesis (1.18) to the two dimensional wall bounded
Figure 9.20: Streaks visualized by hydrogen bubbles in the boundary layer
Lectures
on a flaton plate.
computational fluid dynamics Physics of turbulence
Close to the wall the turbulent fluctuations are weak. The laminar stress
l dominates over the turbulent one t , i.e. l . We consider the thin
boundary layer, i.e. the stress is approximately equal to the wall stress w :
w (9.13)
Applying the Newton hypothesis (1.18) to the two dimensional wall bounded
flow, one gets from (9.13)
dux
w D (9.14)
dy
or
ux
D yC C C (9.15)
u
From the condition at the wall ux D 0 the unknown constant C is zero, i.e.
96
Figure 9.21: Conceptual model of the organization of the turbulence close to

Figure
the wall9.21: Conceptual
proposed model
by Adrian et of
al.the organization of the turbulence close to
(2000).
the wall proposed by Adrian et al. (2000).
Figure 9.22: Vertical distribution of the velocity ux at three different time

Figure
instants9.22: Vertical layer.
in boundary distribution of the velocity ux at three different time
instants in boundary layer. 95
ux C
ux D y C (9.16)
u D y (9.16)
instants in boundary layer.
instants in boundary layer.
ux
D yC (9.16)
u
ux C
Close to the wall the velocity increases Dy linearly ux y. This law confirmed(9.16)
u C
in measurements is valid in the range 0 < y < 5. This region is reffered to
Close
as the to the wall
viscous the velocity increases linearly ux y. This law confirmed
sublayer.
C
in
Farmeasurements
from the wallisthe valid in the stresses
laminar range 0 are < 5. This
< ysmaller than region is reffered
the turbulent to
ones
as the viscous sublayer.
t . The turbulent stress t can be found from the Prandtl mixing length
Far fromThe
model. theinstantaneous
wall the laminar stresses
velocities areare smaller as
presented than
the the
sumturbulent
of averagedonesu
fluctuation
and 0
t . The turbulent stress t can be97found from the Prandtl mixing length
u parts:
ux D ux C ux ;97
0 0
uy D uy ; uy D 0 (9.17)
The averaged velocities are defined as
ZT
ux D lim ux .t/dt (9.18)
T !1
0
The turbulent stress 12 D xy according to the definition is
12 D u0x uy0 (9.19)
Welcome to
our world
of teaching!
and open-minded professors
Study in Sweden -
close collaboration
Ericsson
TAKE THE
Figure 9.23: Illustration of the Prandtl derivation. debajyoti nag
Prandtl proposed in 1925 a very simple algebraic relation for u0x uy0sive reputation in the field
ofusing
RIGHT TRACK
Embedded Systems Re-
ideas from the kinetic gas theory developed by Boltzmann. Let us consider
the fluid particle at the distance y from the wall. Let the particle velocity
be equal to the averaged velocity at y: ux . Due to some perturbations the answer your questions at
mdustudent.com
give jumps
particle your career
from theaposition
headstart
y to at
themälardalen
position y Clxuniversity
and attains the fluid
www.mdh.se d ux
layer with the averaged velocity ux C dy lx . Since the particle has velocity
ux , the velocity at the point y C dy is changed. Obviously, this change is
ddy
ux
lx , or:
q
d ux96 Click on(9.20)
the ad to read more
ux0 2 D lx
dy
98
The turbulent stress 12 D xy according to the definition is
D u0x uy0 (9.19)

Lectures on computational fluid dynamics12 Physics of turbulence
Figure 9.23: Illustration of the Prandtl derivation.
Prandtl proposed in 1925 a very simple algebraic relation for u0x uy0 using
ideas from the kinetic gas theory developed by Boltzmann. Let us consider
the fluid particle at the distance y from the wall. Let the particle velocity
be equal to the averaged velocity at y: ux . Due to some perturbations the
particle jumps from the position y to the position y Clx and attains the fluid
layer with the averaged velocity ux C ddyux
lx . Since the particle has velocity
ux , the velocity at the point y C dy is changed. Obviously, this change is
ddy
ux
lx , or:
q
d ux
ux0 2 D lx (9.20)
dy
98
Root square of the averaged squared pulsation in vertical direction is written
in a similar form: q
d ux
uy0 2 D ly (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:
2 2
d ux 2 d ux
j12 j D j t j D Rxy lx ly D l (9.22)
dy dy
where l 2 D Rxy lx ly is the mixing length of Prandtl. The mixing length
is determined from empirical data. For instance, the length for the wall
bounded flow is
l D y (9.23)
is the first constant of the turbulence, or the constant of Karman. It is
equal to 0:41. Van Driest proposed the modification of (9.23) to take the
wall damping effect into account:
yu

l D y 1 e A (9.24)
where A is the Van Driest constant, which is equal to 26 or 27. In shear flows
97
l D C onst ı.x/, where ı is the shear layer thickness.
We consider again the thin Download
boundary layer,
free i.e.
eBooks the stress is approximately
at bookboon.com
equal to the wall stress w D 12 . Using (9.22) and (9.23) we get
is the first constant of the turbulence, or the constant of Karman. It is
equal to 0:41. Van Driest proposed the modification of (9.23) to take the
wall damping effect into account:
Lectures on computational fluid dynamics yu
Physics of turbulence
l D y 1 e A (9.24)
where A is the Van Driest constant, which is equal to 26 or 27. In shear flows
l D C onst ı.x/, where ı 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
2
2 d ux
w D l (9.25)
dy
r
d ux 1 w u
D D (9.26)
dy l y
The differential equation (9.26) reads
ux 1
D ln y C C C (9.27)
u
The constant C is approximately equal to 5:2. The region (9.27) is referred
to as the logarithmic region which takes place within 30 < y C < 300. The
region 5 < y C < 30 between the viscous and the logarithmic regions is called
the buffer layer. The region at y C > 300 is the wake region. The results of
the analysis are summarized in Fig. 9.24
99
Figure 9.24: Structure of the velocity distribution in the turbulent boundary

layer. U C D ux =u .
98
Basic denitions of the statistical theory of
Chapter 10
Lectures on computational fluid dynamics turbulence
Chapter
Chapter 10
10
Basic definitions of the
Basic
Basic definitions
statistical theory of
definitions of the
of turbulence
the
statistical
statistical theory
theory of
of turbulence
turbulence
10.1 Reynolds averaging
10.1 Reynolds
Reynolds proposed to represent averaging
any stochastic quantity in turbulent flow as
10.1
the sum of its Reynolds
averaged part and averaging
fluctuation. For instance, this representation
Reynoldsforproposed
applied velocity to represent any
components readsstochastic quantity in turbulent flow as
Reynolds
the sum ofproposed to represent
its averaged any stochastic
part and fluctuation. Forquantity
instance,inthis
turbulent flow as
representation
the sum of its averaged
applied for velocity part
ux D ucomponents 0 and fluctuation.
N x C ux I uyreads 0 For instance,
D uN y C uy I uz D uN z C uz I this
0 representation
(10.1)
applied for velocity components reads
The Reynolds uaveraged N x Cvelocities
u0x I uy areD uN y C uy0 I uz D uN z C u0z I
x D u (10.1)
ux D uN x C ux I uy D uN y C uy0 I uz D uN z C u0z9I
0
(10.1)
The Reynolds averaged velocities ZT are ZT >
>
>
uy dtI>
1 1
The Reynolds uNaveraged
x D lim
velocities are
ux dt I uN y D lim >
9
>
T !1 T ZT T !1 T ZT >
=
>
9
>
1 ZT 1 ZT >
uy dtI>
0 0
uN x D lim 1 ZT ux dt I uN y D lim 1 >
>
(10.2)
T !1 T
uN x D lim 1 0 ux dt I uN y D lim T !1 T >
=
>
uy dtI>
uN z D Tlim
!1 T T !1 T 0 >
>
=
uz dt >
; (10.2)
T !1 T Z 0T 0 >
>
>
>
(10.2)
uN z D lim 1 Z uz dt
1 T0
>
>
>
The averaged fluctuation !1 T
uN z D Tlim is zero ;
>
>
T !1 T 0
uz dt >
;

0
The averaged fluctuation is zero u0 D0
The averaged fluctuation is zero x;y;z
– YOUR CHOICE!
The root of the averaged square uof0 fluctuations
or r.m.s.. The quantity averagedu0x;y;z
x;y;z D 0
is called root mean square,
twiceDis0 equal to quantity averaged once
u D u.
The root of the averaged square of fluctuations is called root mean square,
• 32of000 studentssquare• world class research • top
The
orthe
If root
r.m.s.. thequantity
The
turbulence averaged
process averaged
is oftwice
statisticallyfluctuations istocalled
is equal(for
unsteady rootclass
quantity
instance mean teachers
averaged
r.m.s is square,
once
changed
or • modern
r.m.s.. campus • ranked nr 1 in Sweden by international students
uD
in u. the definition of the Reynolds averaging (10.2) is not applicableonce
time), The quantity averaged twice is equal to quantity averaged and
• study in English
If D
u theu.turbulence process is statistically unsteady (for instance r.m.s is changed
If
in the turbulence
time), processofisthe
the definition statistically
Reynolds101unsteady
averaging(for instance
(10.2) is notr.m.s is changed
applicable and
in time), the definition of the Reynolds averaging (10.2) is not applicable and
– Bachelor’s programme in101 Life Science
– Master’s programme in Chemistry 101
Download
brochure
here!

uN x D lim ux dt I uN y D lim uy dtI>
>
T !1 T T !1 T >
=
0 0
(10.2)
ZT >
>
1 >
>
N D lim >
>
u
Lectures on computationalTfluid
z uz dt
!1 dynamics
T >
; turbulence
0
The averaged fluctuation is zero
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 en-
semble 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:
1 X
N

u.t / D lim ui .t /
N !1 N
i D1
10.2 Isotropic and homogeneous turbulence

The turbulence is isotropic if r.m.s of all three velocity fluctuations are equal
u02 02 02
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
x;y;z .x/ E C rE/. This equality can be written for all statistical
x;y;z .x
moments. The turbulence parameters are invariant with respect to the trans-
lation of the reference system.
10.3 Correlation function. Integral length

The product of two fluctuations is the correlation. The product of two fluc-
tuations at two point separated by the distance rE is the correlation function:
E rE/ D u0i .x/u

Rij .x; E j0 .xE C rE/ (10.4)
If i D j the correlation function Ri i is reffered as to the autocorrelation
function. In homogeneous turbulence Ri i depends only on the separation:
Rij .Er / D u0i .x/u

E j0 .xE C rE/ (10.5)
The coefficient of the autocorrelation function is changed between zero and
one:
u0i .x/u
E 0i .xE C rE/
E rE/ D
i i .x; (10.6)
u02 E
i .x/
100
A sample of the autocorrealtion function coefficient for scalar fluctuation f 0
102
E rE/ D ui .x/u
Rij .x; E j .xE C rE/ (10.4)
If i D j the correlation function Ri i is reffered as to the autocorrelation
function. In homogeneous turbulence Ri i depends only on the separation:
E j0 .xE C rE/ (10.5)
The coefficient of the autocorrelation function is changed between zero and
one:
u0i .x/u
E 0i .xE C rE/
E rE/ D
i i .x; (10.6)
u02 E
i .x/
A sample of the autocorrealtion function coefficient for scalar fluctuation f 0
102
Figure 10.1: Autocorrelation function coefficient for scalar fluctuation at

three different points A; B and C across the jet mixer.[47].
f 0 .x/f
E 0 .xE C rE/
E rE/ D
f .x; (10.7)
f 02 .x/
E
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 0 .rC /f 0 .rC C r/
f .rC ; r/ D (10.8)
f 02 .rC /
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
101
103
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 .rABasic denitions
/ < 0) . Thisofisthe statistical
true in theory of
statistical fluidmost
sense, i.e. the dynamics
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
u0i103
E t/u0i .x;
.x; E t C /
E / D
i i .x; (10.9)
u02 E t/
i .x;
The integral of the spatial autocorrelation functions
Z1
E D
Lij .x/ E xj /dxj
i i .x; (10.10)
0
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)
ZD=2
Lf .x/ D f .r/dr (10.11)
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
Scholarships
Z1
E D
Ti .x/ E /d
i i .x; (10.12)
0
is the integral time length.

Coefficients of the autocorrelation function of the axial velocity fluctuations
for the free jet are presented in Fig. 10.3. The line 1 corresponds to the
Open your mind to
autocorrelation function calculated along the jet boundary line (shown by
new opportunities
the blue line in Fig. 10.3, right). The line 2 corresponds to the autocor-
relation function calculated along the jet axis. Oscillating character of the
dependency
With 31,000 Rstudents,
uu .x=d / indicates
Linnaeus theis presence of vortex structures arising
University
one jet
at the of the larger universities
boundary in Sweden. We
and attaining the jet axis. The distance between zero
are aismodern
points roughlyuniversity, known for
the vortex our strong
size.
international profile. Every year more than Bachelor programmes in
1,600 international students from all over the Business & Economics | Computer Science/IT |
world choose to enjoy the friendly atmosphere Design | Mathematics
10.3.1
and activeSome relations
student life in isotropic
at Linnaeus University. turbulence
Welcome to join us! Business & Economics | Behavioural Sciences | Computer
In the isotropic turbulence u D 02
u0l .x/u0l .x/
D u0t .x/u 0
t .x/ D
Science/IT
::: the autocor-
| Cultural Studies & Social Sciences | Design |
relation function can be represented in the form [13]Mathematics | Natural Sciences | Technology & Engineering
104

along the jet mixer centerline (Fig. 10.1, right)
ZD=2
Lf .x/ D
Lectures on computational fluid dynamics f .r/dr (10.11) turbulence
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
Z1
E D
Ti .x/ E /d
i i .x; (10.12)
0
is the integral time length.

Coefficients of the autocorrelation function of the axial velocity fluctuations
for the free jet are presented in Fig. 10.3. The line 1 corresponds to the
autocorrelation function calculated along the jet boundary line (shown by
the blue line in Fig. 10.3, right). The line 2 corresponds to the autocor-
relation function calculated along the jet axis. Oscillating character of the
dependency Ruu .x=d / indicates the presence of vortex structures arising
at the jet boundary and attaining the jet axis. The distance between zero
points is roughly the vortex size.
10.3.1 Some relations in isotropic turbulence

In the isotropic turbulence u02 D u0l .x/u0l .x/ D u0t .x/u0t .x/ D ::: the autocor-
relation 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
Figure 10.3: Autocorrelation functions in free jet flow. [46].

2 f g
Rij D u ri rj C gıij (10.13)
r2
where
u0l 105
.x/u0l .x C r/
f .r/ D ; (10.14)
u0l .x/u0l .x/
is the autocorrelation of the longitudinal velocity calculated in longitudinal
direction. For instance, the autocorrelation function of the ux fluctuation
calculated in x direction,
u0x .x/u0x .x C r/
f .r/ D ; (10.15)
u0x .x/u0x .x/
or the autocorrelation function of the uy fluctuation calculated in y direction:
uy0 .y/uy0 .y C r/
f .r/ D ; (10.16)
uy0 .y/uy0 .y/
The function f .r/ is the same in both cases (10.15) and (10.16). The au-
tocorrelaton 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/
g.r/ D x ; (10.18)
u0x .y/u0x .y/
or the autocorrelation function of the uy fluctuation calculated in x direction:
uy0 .x/uy0 .x C r/
g.r/ D ; (10.19)
uy0 .x/uy0 .x/
The function g.r/ is the same in both cases (10.18) and (10.19). The ve-
locities 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
104
connecting two points (i.e. longitudinal direction), ui t is its projection on
the transversal direction. Products like u u are the correlations between
i t i tat bookboon.com
points 1 and 2.
or the autocorrelation function of the uy fluctuation calculated in x direction:
uy0 .x/uy0 .x C r/ Basic denitions of the statistical theory of

g.r/ D
Lectures on computational fluid dynamics ; (10.19) turbulence
uy0 .x/uy0 .x/
The function g.r/ is the same in both cases (10.18) and (10.19). The ve-
locities 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
Figure 10.4: Illustrations of velocities used in calculations of the longitudinal

f and transversal g autocorrelations.
Figure 10.5: Illustration of the autocorrelation functions f and g and Taylor

microscales.
The following relations are valid between g and f in the isotropic homoge-
neous 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 autocorre-
lation 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
107
Figure 10.4: Illustrations of velocities used in calculations of the longitudinal
f and transversal g autocorrelations.
Figure 10.5: Illustration of the autocorrelation functions f and g and Taylor

microscales.
The following relations are valid between g and f in the isotropic homoge-
neous 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 autocorre-
lation 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
f .r/ D 1 C 2
.0/r 2 C O.r 4 / (10.21)
2 @r
The parabola fitted to the curve f .r/ at r D 0 intersects the horizontal axis
at the point:
s
2
f D @2 f (10.22)
@r 2
.0/
Similar relations can be derived for the transversal autocorrelation
s
2
g D @2 g (10.23)
@r 2
.0/
Today the Taylor microscale is still in use in turbulent research although
it has no physical meaning. Very popular 106is the Reynolds number based on
the Taylor microscale Download free eBooks at bookboon.com
Re D u0 = (10.24)
@r 2
Similar relations can be derived for the transversal autocorrelation

s Basic denitions of the statistical theory of
2
g D @2 g (10.23) turbulence
@r 2
.0/
Today the Taylor microscale is still in use in turbulent research although
it has no physical meaning. Very popular is the Reynolds number based on
the Taylor microscale
Re D u0 = (10.24)
which characterizes the state of the turbulence in the flow.
10.3.3 Correlation functions in the Fourier space

Any continuous function can be represented in the Fourier space:
Z1
f .Er ; t/ D E t/e i kErE d kE
fO.k; (10.25)
1
The new function fO.k;E t/ is then known as the Fourier transform and/or
the frequency spectrum of the function f . The Fourier transform is also a
reversible operation:
Z1
E t/ D 1
fO.k;
E
f .Er ; t/e i k rE d rE (10.26)
8 3
1
The Fourier transformation can be also written for the correlation function:
Z1 Z1
E i kErE E 108
E D 1 E
Rij .Er / D ˚ij .k/e d k; ˚ij .k/ Rij .Er /e i k rE d rE (10.27)
8 3
1 1
Very often one uses one dimensional correlation functions defined as
Z1 Z1 Z1
1
ij .k1 / D Rij .r1 ; 0; 0/e i k1 r1 dr1 D ˚ij .k1 ; k2 ; k3 /d k2 d k3
2
1 1 1
(10.28)
Proof of the formula (10.28) The inverse Fourier transform of the function
Rij .r1 ; 0; 0/:
Z1
1
ij .k1 ; 0; 0/ D Rij .r1 ; 0; 0/e i k1 r1 dr1 (10.29)
2
1
The inverse transformation reads:

Z1
Rij .r1 ; 0; 0/ D ij .k1 ; 0; 0/e i k1 r1 d k1 (10.30)
1
The general definition is
Z1 Z1 Z1 107 E
Rij .r1 ; r2 ; r3 / D ˚ij .k1 ; k2 ; k3 /e i k rE d k1 d k2 d k3 (10.31)
1 1 1
From the last formula we have:

1
The inverse transformation reads:

Z1 Basic denitions of the statistical theory of
Lectures on computational fluid
Rij .r 0/ D
dynamics
1 ; 0; ij .k1 ; 0; 0/e i k1 r1 d k1 (10.30) turbulence
1
The general definition is
Z1 Z1 Z1
E
Rij .r1 ; r2 ; r3 / D ˚ij .k1 ; k2 ; k3 /e i k rE d k1 d k2 d k3 (10.31)
1 1 1
From the last formula we have:
Z1 Z1 Z1
Rij .r1 ; 0; 0/ D ˚ij .k1 ; k2 ; k3 /d k2 d k3 e i k1 r1 d k1 (10.32)
1 1 1
Comparison of (10.32) with (10.30) results in the desired formula

Z1 Z1
ij .k1 ; 0; 0/ D ˚ij .k1 ; k2 ; k3 /d k2 d k3 (10.33)
1 1
10.3.4 Spectral density of the turbulent kinetic energy

109
According to the definition of the correlation function

E j0 .xE C rE/ (10.34)
the total turbulent kinetic energy is
YOUR 1 WORK
TKE D R .0/ D
1
Z
E E
AT
Z TOMTOM
E kE
˚ .k/d k D ˚ .k/d
1
WILL 1
ii (10.35)
ii ii
2 2
1 0
AROUND THEZ WORLD. EVERYDAY.

The quantity
E.k/ D E kE
˚i i .k/d (10.36)
E
jkj
follow us on
p turbulent kinetic energy. Physically it is the
is the spectral density of the
2 2 2
energy on the sphere k D k1 C k2 C k3 in the Fourier space. The total
turbulent kinetic energy is then:
Z1
TKE D E.k/d k (10.37)
0
10.4 Structure functions

#ACHIEVEMORE
10.4.1 Probability density function
We take the definitions from Wikipedia: In probability theory, a probability
density function (pdf), or density of a continuous random variable, is a func-
tion that describes the relative likelihood for this random variable to take
108
on a given value. The probability for the random variable to Click on the a
fall within ad to read more
particular region is given byDownload free eBooks
the integral at bookboon.com
of this variables density over the
region. The probability density function is non negative everywhere, and its
integral over the entire space is equal to one.
10.3.4 Spectral density of the turbulent kinetic energy
According to the definition of the correlation function
E j0 .xE C rE/ (10.34)
the total turbulent kinetic energy is
Z1 Z1
1 1 E kE D E kE
TKE D Ri i .0/ D ˚i i .k/d ˚i i .k/d (10.35)
2 2
1 0
The quantity
Z
E.k/ D E kE
˚i i .k/d (10.36)
E
jkj
p turbulent kinetic energy. Physically it is the

is the spectral density of the
energy on the sphere k D k12 C k22 C k32 in the Fourier space. The total
turbulent kinetic energy is then:
Z1
TKE D E.k/d k (10.37)
0
10.4 Structure functions

10.4.1 Probability density function
We take the definitions from Wikipedia: In probability theory, a probability
density function (pdf), or density of a continuous random variable, is a func-
tion that describes the relative likelihood for this random variable to take
on a given value. The probability for the random variable to fall within a
particular region is given by the integral of this variables density over the
region. The probability density function is non negative everywhere, and its
integral over the entire space is equal to one.
10.4.2 Structure function

Kolmogorov introduced the structure function of q order for any stochastic
function. For instance, for the longitudinal velocity along the longitudinal
direction (see Fig. 10.5) it reads:
110
109
Figure 10.6: Kurtosis of the structure function for the concentration of the
scalar field obtained in the jet mixer. [47].
Sq .l/ D h.u2l u1l /q i (10.38)

The standard deviation squared is then 2 D S2 . If the p.d.f. of the structure
function is described by the Gaussian function
1 .x/2
p:d:f:.x/ D p e 2 2 (10.39)
2
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
h.u2l u1l /4 i
Kurt D (10.40)
.h.u2l u1l /2 i/2
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
Sk D 111 (10.41)
.h.u2l u1l /2 i/3=2
is zero for the Gaussian process. For the isotropic turbulence the skewness
of the derivative
l 3
h @u
@l
i
Sk D h i (10.42)
l 2 3=2
h @u 110
i
@l
is equal to 0:5. Physically it means that negative values of the derivative @u
@l
l
are more probable than positive ones. Please prove that the skewness (10.42)
h.u2l u1l / i
S k D h.u2l u1l2/33 i3=2 (10.41)
S k D .h.u 2l2l
h.u u1l/2/i/i3=2
u1l (10.41)
S k D .h.u2l u1l /2 i/3=2 (10.41)
is zero for the Gaussian process..h.u For u1lisotropic
2l the / i/ turbulence the skewness
is zero for the Gaussian process. For the isotropic turbulence theofskewness
Basic denitions the statistical theory of
of
is the
zero
Lectures
derivative
for the Gaussian process.
on computational fluid dynamics For the isotropic turbulence the skewness turbulence
of the derivative
of the derivative l 3
h @u 3 i
S k D h h @u @ll i
@u
@ll 3i3=2 (10.42)
S k D h h@u@l 2 ii3=2 (10.42)
h
S k D hh @u l
i
2 i
@l i 3=2
@u l (10.42)
@ll 2
h @l i
is equal to 0:5. Physically it means that negative values of the derivative @u l
is equal to 0:5. Physically it means that negative values of the derivative @u @ll
@ll
areequal
is more probable than positive
0:5. Physically ones.that
Please provevalues
that the skewness (10.42)
@u
are moretoprobable it means
than positive ones. Please negative
prove that the of the derivative
skewness (10.42)
@l
is
arethe skewness
more probableof the structure
than positive functions
ones. of the
Please firstthat
prove order
the .l/ D hu2l(10.42)
S1skewness u1l i
is the skewness of the structure functions of the first order S1 .l/ D hu2l u1l i
calculated
is the skewnessat l ! 0. structure functions of the first order S1 .l/ D hu2l u1l i
of the
calculated at l ! 0.
calculated at l ! 0.
Exercise 1.
Exercise 1.
Calculate
Exercise the
1. Reynolds averaged values of the time dependent signals
Calculate the Reynolds averaged values of the time dependent signals
Calculate the Reynolds averaged values of the time dependent signals
u.t/ D cos2 t;
u.t/ D cos22 t;
u.t/ D sin
cos t;t;
t;
u.t/ D sin
1t; t;
u.t/ D sin t 1;
u.t/ D 1 t; t 1;
u.t/ D 1 0; t; ttt > 1;
1:
u.t/ D 0; > 1:
0; t > 1:
Solution:
Solution:
Solution:
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:
Exercise 2.
Exercise 2.
Find the Reynolds
Exercise 2. stresses for the isotropic turbulence.
Find the Reynolds stresses for the isotropic turbulence.
Find the Reynolds stresses for0
0the
1
isotropic turbulence.
2 r12 r13 1
0 2 r12 r13 1
@
@rr221 rr22 r13 A
12 r23 A
21 r22 r23
@rr31
21
rr32
22
rr33
23
A
r31 r32 r33
r31 r32 r33
Solution: 0 1
Solution: 02 0 01
@202 020 0001
0
Solution:
A
@0 2 0A
@00 02 20A
0 0 2
0 0 2
112
112
112

1
u.t/ D ;
2
u.t/ D 0; Basic denitions of the statistical theory of
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
Solution: 0 1
2 0 0
@0 2 0A
0 0 2
Exercise 3.
Calculate
Exercise the 112of the isotropic turbulence, if r33 D 1
3. turbulent kinetic energy
Calculate the turbulent kinetic energy of the isotropic turbulence, if r33 D 1
Solution:
Solution: k D 3=2
k D 3=2
Exercise 4.
Relation
Exercisebetween
4. longitudinal autocorrelation function
Relation between longitudinal autocorrelation function
ul .x/ul .x C r/
f .r/ D
.x C r/
u2ll.x/
ul .x/u
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
2 sin kr
f .r/ D 2 Z1 E.k/k 2 r 2 . cos kr/d k (10.43)
u2l kr
sin kr
f .r/ D 2 0 E.k/k 2 r 2 . cos kr/d k (10.43)
ul kr
Energy density of the isotropic
0 decaying turbulence is described as E.k/ D
4 2
k ExpŒk . Calculate
Energy density of the isotropic decaying turbulence is described as E.k/ D
k 4 ExpŒk 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.
Solution:
Solution:
Z1
2 sin kr
f .r/ D 2 Z1 E.k/k 2 r 2 . cos kr/d k D (10.44)
u2l 2 2 sin kr
kr
f .r/ D 2 0 E.k/k r . cos kr/d k D (10.44)
ul kr
0
Z1
2 2 2 2 sin kr 1p
2 Z
D 1 ExpŒk k r . cos kr/d k D ExpŒr 2 =4
u2l kr
sin kr 4
1 p
D 2 0 ExpŒk 2 k 2 r 2 . cos112
kr/d k D ExpŒr 2 =4
ul kr 4
0
113 eBooks at bookboon.com
Download free
113
Solution:
Z1
2 sin kr
f .r/ D 2 E.k/k 2 r 2 . cos Basic kD
kr/ddenitions (10.44)
of the statistical theory of
u kr
l
0 turbulence
Z1
2 sin kr 1p
D 2 ExpŒk 2 k 2 r 2 . cos kr/d k D ExpŒr 2 =4
ul kr 4
0
Z1
113
LD f .r/dr D
4
0
4
D
1=4
Exercise 5.
Calculate the probability density function of the time dependent signal
time 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
signal 0 0.1 0.9 0.5 0.7 0.5 0.05 0.65 0.2 0.78 0.43 0.98 0.67 0.92 0.55
using the increment 0.2.
Solution:
signal 0-0.2 0.2-0.4 0.4-0.6 0.6-0.8 0.8-1.0

p.d.f. 3/15 2/15 3/15 4/15 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 au-
tocorrelation function of u with respect to point N=2.
Solution:
uk;0 k
i D ui ui
1 X k
K
ui D ui
K
kD1
P
K
uk;0 k;0
i uN=2
kD1
R.i; N=2/ D
P
K
uk;02
N=2
kD1
114
113
Chapter 11
Lectures on computational fluid dynamics Kolmogorov theory K41
Chapter 11
Chapter 11 theory K41
Kolmogorov
Kolmogorov theory K41
Kolmogorov theory K41
11.1 Physical background
11.1
One of thePhysical background
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
Onebased on the
of the mosthypothesis
outstanding of local isotropy
results of the turbulent
in turbulence theory was motion at small
obtained by
One
scales.of
Kolmogorovthe most
The (s. outstanding
physical model
Fig. 11.1) results
in behind
1941. The in turbulence
the Kolmogorov theory
Kolmogorov theory was
theoryknown obtained
is the as by
vortex
K41
Kolmogorov
cascado
is on the(s. hypothesis
based illustrated Fig.in11.1)
Fig. ofin 1941.
11.2.
localBig The Kolmogorov
vortices
isotropy of with theory
scales
the turbulent knownatassmall
(corresponds
L motion K41
to
is based
the
scales. on
waveThe the
numbers hypothesis
physical of local
in the
=Lmodel isotropy
Fourier
behind space)
the of break
the turbulent
Kolmogorov up theory motion
to small theatvortex
is ones, small
which
scales.
in
cascado The
turn split physical
into
illustrated even model
smaller
in Fig. behind
11.2. andBig the Kolmogorov
so vortices
on up towith
the smallestheory
scales is thewith
Lvortices
(corresponds vortex
the
to
cascado
scale
the wave illustrated
. One of the =L
numbers in Fig.
most 11.2.Fourier
important
in the Big vortices
vortex break
space) with scales
up mechanisms
break (corresponds
up to Lsmall isones, to
the vortex
which
theturn
in wavesplit
reconnection numbers
described
into in the and
above.
=Lsmaller
even Fourier
The space)
energy
so on upis tobreak up tofrom
transferred
the smalles small
bigones,
vortices which
vortices
with to
the
in turn
small
scale split
.ones into
Onealmost
of theeven smaller
without
most theand
important so The
on up
loss.vortex to the
massive
break smalles
up dissipation
mechanisms vortices with
"istakes
the the
place
vortex
scale . One
at small
reconnection of
vorticesthereferred
mostabove.
described important
to asThethevortex break
dissipative
energy upthe
or mechanisms
is transferred from bigis vortices
Kolmogorov the vortex
vortices.
to
reconnection
The real
small described
onesturbulent
almost above.
vortices
without are
the The energy
similar
loss. is these
transferred
Thetomassive from by
calculated
dissipation big vortices
Isazawa
" takes to
et
place
small
al.
at smallones
(Fig. almost
11.3).
vortices without
Vortices
referred tothe
are as loss.
displayed The massive
at three
the dissipative dissipation
different
or the " takes
time instants.
Kolmogorov place
The
vortices.
at small
upper
The real vortices
pictures arereferred
turbulent obtained
vorticesto are
as the
from dissipative
the
similar lower ones
to these or by
thefiltering
Kolmogorov
calculated vortices.
outIsazawa
by the hight
et
The(Fig.
al. real 11.3).
frequencies. turbulentseenvortices
As Vortices bigare are similar
vortices at to
are revealed
displayed these
three in calculated
low
different time by
frequency IsazawaThe
simulation.
instants. et
al. the
If
upper(Fig. 11.3). are
resolution
pictures Vortices
is obtainedare displayed
increased, more
from theandat three
more
lower onesdifferent
small scaletime
by filtering instants.
vortex
out the The
filaments
hight
upper
appears pictures
on the are obtained
place of big from
smooththe lower
vortices. ones by
Thus, filtering
frequencies. As seen big vortices are revealed in low frequency simulation. the out
most the hight
important
frequencies.
physical
If As seen
processes
the resolution big vortices
isduring the vortex
increased, morearebreak
revealed
and in
up are:
more low scale
small frequency
vortexsimulation.
filaments
If the resolution
appears is increased,
on the place more and
of big smooth more small
vortices. Thus,scalethe vortex filaments
most important
appears processes
physical on the place of big
during smoothbreak
the vortex vortices.
up are: Thus, the most important
Transfer
physical energy
processes fromthe
during large scales
vortex to small
break up are: ones and
Dissipation
Transfer energy from
of the largeinscales
energy smallto small ones and
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
Figure 11.1: Andrey Kolmogorov was a mathematician, preeminent in the

20th century, who advanced various scientific fields (among them probabil-
ity theory, topology, intuitionistic logic, turbulence, classical mechanics and
computational complexity).
Figure 11.2: Illustration of the vortex cascado.
116

Figure 11.1: Andrey Kolmogorov was a mathematician, preeminent in the
20th century, who advanced various scientific fields (among them probabil-
ity theory, topology, intuitionistic logic, turbulence, classical mechanics and
computational
Lectures complexity).
on computational fluid dynamics Kolmogorov theory K41
Figure 11.2: Illustration of the vortex cascado.
116
Figure 11.3: Turbulent vortices revealed in DNS calculations performed by

Isazawa et al. (2007).
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
X @u0i @uj0 2
"D C D 2sij sij (11.1)
2 i;j @xj @xi
@u0 @uj0
where sij D 12 . @xji C @xi
/ is the fluctuating rate of strain:
1 @ui @uj 1 @ui @uj

sij D Sij S ij D . C / . C / (11.2)
2 @xj @xi 2 @xj @xi
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 116". The energy dissipated by the
small vortices is generated by large scale vortices. The energy production is
defined as
2 i;j
@xj @xi
@u0 @uj0
where sij D 12 . @xji C @xi
/ is the fluctuating rate of strain:

1 @ui
@uj 1 @ui @uj Kolmogorov theory K41
sij D Sij S ij D . C / . C / (11.2)
2 @xj @xi 2 @xj @xi
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
@ui
P D u0i uj0 (11.3)
@xj
Based on the dimension analysis, Prandtl and Kolmogorov proposed the
estimation of the integral length of the turbulent flow
k 3=2
117
L (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 informa-
tion 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 ":
D ˛ "ˇ ; u D ˛u "ˇ117
u
; D ˛ "ˇ (11.5)
The analysis of dimension allows one to derive the following dependences:
The Kolmogorov first similarity hypothesis reads:
In every turbulent flow at sufficiently high Reynolds number Re t , the statistics

of the on
Lectures small-scale motions
computational (l < lE I ) have a universal form that is Kolmogorov
fluid dynamics uniquely theory K41
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 ":
D ˛ "ˇ ; u D ˛u "ˇu ; D ˛ "ˇ (11.5)

The analysis of dimension allows one to derive the following dependences:
3 1=4
118

D ;
"
(11.6)
u D ."/1=4 ;
D .="/1=2
Here is the scale of the smallest dissipative vortices ( Kolmogorov scale),

u is the characteristic velocity of turning of Kolmogorov vortices,
is characteristic turn over time of Kolmogorov vortices.
Using expressions
k3=2
" u k1=2 (11.7)
L
some useful estimations can be derived from (11.6):
=L .Ret /3=4 ;

u =u D .Ret /1=4 ; (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 fluctua-
tion 1m=s, the turbulent Reynolds number WE WILL
in water is Re tTURN
D 106 . YOUR
The CV
INTO
Kolmogorov scale is in this case 32000 as less as the flowAN OPPORTUNITY
macroscale L. Es-
timations of the Kolmogorov scale in the jet mixer with nozzleOF A LIFETIME
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.
Do you like cars? Would you like to be a part of a successful brand?
The energy density depends Send us your CV on
As a constructer at ŠKODA AUTOon you the wave
will put great number k and
things in motion. thethat
Things dissipation
will rate
www.employerforlife.com
" ease everyday lives of people all around Send us your CV. We will give it an entirely
new new dimension.
E.k/ D "˛ k ˇ (11.9)
119 118 Click on the ad to read more

u is the characteristic velocity of turning of Kolmogorov vortices,
is characteristic turn over time of Kolmogorov vortices.
Using expressions

k3=2 Kolmogorov theory K41
" u k1=2 (11.7)
L
some useful estimations can be derived from (11.6):
=L .Ret /3=4 ;

u =u D .Ret /1=4 ; (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 fluctua-
tion 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. Es-
timations 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 ˇ (11.9)
119
Figure
Figure11.4:
11.4:Distribution of the
Distribution ofKolmogrov scale along
the Kolmogorov the centerline
scale along theofcenterline
the jet mixer
of
and
the free jet [47].and
jet mixer Thefree
dissipation
jet. The ε is calculated
ratedissipation ratefrom k – ε model
" isthecalculated and the
from the
k " model estimation
experimental of the Miller and
and the experimental Dimotakis
estimatin (1991) εand
of Miller = 48(U 3
/d )((x – x(1991)
Dimotakis
d 0
)/d )–4.
" D 48.Ud3 =d /..x x0 /=d /4 .
The analysis of dimension leads to the Kolmogorov law
k 5=3
E.k/ D ˛"2=3119 (11.10)
where ˛ 1:5 is the constant.
Figure 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 the experimental estimatin of Miller and Dimotakis (1991)
3 4
fluid dynamics Kolmogorov theory K41
" D 48.U d =d /..x x0 /=d / .
The analysis of dimension leads to the Kolmogorov law
E.k/ D ˛"2=3 k 5=3 (11.10)

where ˛ 1:5 is the constant.
11.4 Three different scale ranges of turbulent

flow
Three different ranges can be distinguished in the spectrum of scales in the
full developed turbulence at high Reynolds numbers Re t (Fig.11.5):
Energy containing range at l > lE I (according to Pope [14], lE I 16 L).

Within this range the kinetic energy of turbulence is generated and big
turbulent eddies are created.
Inertial subrange at lDI < l < lE I (according to Pope [14], lDI 60).
Within this subrange the energy is transferred along the scales towards
dissipative vortices without any significant loss, i.e. " 0. The energy
density obeys the Kolmogorov law (11.10).
Dissipation range l < lDI . The dissipation of the energy of big vortices
occurs within the dissipation range.
The inertial subrange and dissipation range belong to the universal equilib-
120
rium range. Three corresponding ranges can be distinguished in the distribu-
tion of the energy density over the wave numbers k Fig. 11.6. The presence
of the inertial and dissipation subranges was confirmed in numerous exper-
imental measurements performed after development of the K41 theory (see
Fig. 11.7, 11.8).
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.
Figure 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, 3- dissipation range.
121
Develop the tools we need for Life Science

exciting field where biology,
mathematics meet.

biology and medicine using
mathematics.

Figure 11.7: Experimental confirmation of the Kolmogorov law. The com-

pensated energy spectrum for different flows. [14].
122
Figure 11.8: Experimental confirmation of the Kolmogorov law for the con-
centration fluctuations in the jet mixer. Measurements of the LTT Rostock.[47].
11.5 Classification of methods for calculation

of turbulent flows.
The energy spectrum Fig. 11.6 is used to classify three main methods of tur-
bulent flows modelling (Fig. 11.9). The most 122 general strategy is the Direct
Numerical Simulation (DNS). Within the DNS the whole spectrum of tur-
bulent 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 con-
centration fluctuations in the jet mixer. Measurements of the LTT Rostock.
11.5 Classification of methods for calculation

of turbulent flows.
The energy spectrum Fig. 11.6 is used to classify three main methods of tur-
bulent flows modelling (Fig. 11.9). The most general strategy is the Direct
Numerical Simulation (DNS). Within the DNS the whole spectrum of tur-
bulent 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 remain-
ing 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 Limitation of K-41. Kolmogorov theory

K-62
The strongest and simultaneously the most questionable assumption of the
Kolmogorov-41 is: Dissipation rate is an universal constant for each turbulent
flow. Already in 1942, during a scientific seminar the Nobel price laureate
123
Figure 11.9: Three main methods of turbulent flows modelling.
123
Figure 11.9: Three main methods of turbulent flows modelling.
Figure 11.10: Vortex structures resolved by different models.
Landau noted, that the dissipation rate is a stochastic function, it is not

constant. We consider the consequences of the neglect of this fact.
According to the Kolmogorov - Obukhov law the structure function of the
q-th order
Sq .l/ D h.u2l u1l /q i (11.11)

has the following asymptotic behaviour at small l
Sq .l/ ."l/q .l/ q (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 phe-
124
Copenhagen
Master of Excellence cultural studies
Copenhagen Master of Excellence are

www.come.ku.dk

Sq .l/ D h.u2l u1l /q i (11.11)
has the following asymptotic behaviour at small l
Sq .l/ ."l/q .l/ q (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 phe-
nomenon called the intermittency. The intermittency is caused by the pres-
ence of laminar spots in every turbulent
124 flows even at very high Reynolds
numbers.
Figure 11.11: Power of the structure function. Experiments versus prediction

of Kolmogorov and Obukhov. [48].
After the deviation between the K-41 and measurement was documented,
Kolmogorov tried to improve his theory. New Kolmogorov theory called as
K-62 was published in 1962.
New theory is based on two following assumptions:
Assumption 1:
Sq .l/ D< ılq >< "q=3
l > l q=3 ;
q (11.13)
q D C q=3 < "ql > l q
3
Assumption 2:
.ln"a/2
P ."l / D ce 2 2
l a D lnN" l2 D A C ln.L= l/
q (11.14)
q D q.1 q/ &q D C q.3 q/ < "2l > l
2 3 18
Unfortunately, various experiments showed later that the second assumption
is proved to be wrong.
11.6.1 Exercises
Exercise 1. Calculate the Reynolds averaged values of the time dependent
signals
u.t/ D cos2 t;
D sin t;
u.t/ 125

1 t; t 1;
u.t/ D
0; t > 1:
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; Kolmogorov theory K41
u.t/ D
0; t > 1:
Solution:
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
Solution: 0 1
2 0 0
@0 2 0A
0 0 2
Exercise 3. Calculate the turbulent kinetic energy of the isotropic turbu-

lence, if r33 D 1
Solution:
k D 3=2
Exercise 4. Relation between longitudinal autocorrelation function
ul .x/ul .x C r/
f .r/ D
u2l .x/
and energy density E.k/is given by formula [13]

Z1
2 sin kr
f .r/ D 2 E.k/k 2 r 2 . cos kr/d k (11.15)
ul kr
0
Energy density of the isotropic decaying turbulence is described as E.k/ D

k 4 ExpŒk 2 . Calculate
longitudinal autocorrelation function

126 f,
integral length and
Taylor microscale

Solution:
Z1
2 sin kr
f .r/ D 2 E.k/k 2 r 2 . cos kr/d k D (11.16)
ul kr
0 126
Z1 Download free eBooks at bookboon.com
2 2 2 2 sin kr 1p
D 2 ExpŒk k r . cos kr/d k D ExpŒr 2 =4
ul kr 4
integral length and
Taylor microscale
Solution:
Z1
2 sin kr
f .r/ D 2 E.k/k 2 r 2 . cos kr/d k D (11.16)
ul kr
0
Z1
2 sin kr 1p
D 2 ExpŒk 2 k 2 r 2 . cos kr/d k D ExpŒr 2 =4
ul kr 4
0
Z1

LD f .r/dr D
4
0
4
D
1=4
Exercise 5. Calculate the probability density function of the time depen-

dent signal
time 0 1 2 3 4 5 6 7
signal 0 0.1 0.9 0.5 0.7 0.5 0.05 0.65
time 8 9 10 11 12 13 14
signal 0.2 0.78 0.43 0.98 0.67 0.92 0.55
using the increment 0.2.
Solution:
signal 0-0.2 0.2-0.4 0.4-0.6 0.6-0.8 0.8-1.0

p.d.f. 3/15 2/15 3/15 4/15 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
1 X k
K
ui D ui
K
kD1
P
K
uk;0 k;0
i uN=2
kD1
R.i; N=2/ D
P
K
uk;02
N=2
kD1
127
Reynolds Averaged Navier Stokes Equation (RANS)
Chapter 12
Chapter 12
Chapter 12
Reynolds Averaged Navier
Reynolds
Stokes Averaged
Equation Navier
(RANS)
Reynolds Averaged Navier
Stokes Equation (RANS)
Stokes Equation (RANS)
According to the Reynolds averaging each fluctuating quantity is represented
as the sum of the averaged value and its fluctuation:
According to the Reynolds averaging each fluctuating quantity is represented
According
as the sumtoofthe
the Reynolds
averaged averaging eachfluctuation:
fluctuating quantity is represented
ux D uN x C u0xvalue
I u and D uNitsy C uy I
0
uz D uN z C u0z (12.1)
as the sum of the averaged valueyand its fluctuation:
where the averaged
ux D u part
N x Cisu0xdefined
I uy D as:uN y C u0 I uz D uN z C u0 (12.1)
y z
0 0 0
ux D uN x C ux I uy D uN y C uy I uz D uN z C uz (12.1)
where the averaged part is defined as:
where the averaged Z part is defined as:
T Z T Z T
1 1 1
uN x D ux dtI uN y D uy dtI uN z D uz dt (12.2)
T ZT T ZT T ZT
1 Z0T 1 Z0T 1 Z0T
uN x D 1 ux dtI uN y D 1 uy dtI uN z D 1 uz dt (12.2)
uN x D Taveraging
The Reynolds u x dtI has uN the
y D T
following u y dtI N
u D
properties:
z
T u z dt (12.2)
T 0 T 0 T 0
0 0 0
The Reynolds
averaged averaging
fluctuation is the
has zero:following properties:
The Reynolds averaging has the following properties:
averaged fluctuation is zero: fN 0 D 0 (12.3)
averaged fluctuation is zero:
double averaged quantity is equal fN 0 to
D once
0 averaged one: (12.3)
fN 0 D 0 (12.3)
double averaged quantity is equal N
f D N
to fonce averaged one: (12.4)
double averaged quantity is equal to once averaged one:
fN D N
Brain power
averaged sum is equal to the sum
averaged sum is equal to the Cg D
f sum
N of f
f D fN
averaged:
of f N
averaged:
By 2020, wind could provide(12.4)
one-tenth of our planet’s
(12.4)
electricity needs. Already today, SKF’s innovative know-
how is crucial to running a large proportion of the
C gN world’s wind turbines. (12.5)
averaged sum is equal to the sum of averaged: Up to 25 % of the generating costs relate to mainte-
C g D fN C gNcommutate:
operators of averaging and fdifferentiation nance. These can be reduced(12.5)
dramatically thanks to our
f C g D fN C gN systems for on-line condition(12.5) monitoring and automatic
operators of averaging and @fN
@f differentiation @fNlubrication.
@f commutate: We help make it more economical to create
operators of averaging and D ; D cleaner, cheaper energy out of(12.6) thin air.
@t differentiation
@tN @x commutate:
@xN By sharing our experience, expertise, and creativity,
@f @f @f @f industries can boost performance beyond expectations.
@f D @fN ; @f D @fN Therefore we need the best (12.6)
employees who can
@t D @t ; @x D @xmeet this challenge! (12.6)
129
@t @t @x @x
129 The Power of Knowledge Engineering
129
Plug into The Power of Knowledge Engineering.

Visit us at www.skf.com/knowledge

0 0 0
The Reynolds averaging has the following properties:

averaged fluctuation is zero:
Lectures on computational fluid dynamics Reynolds Averaged Navier Stokes Equation (RANS)
fN 0 D 0 (12.3)
double averaged quantity is equal to once averaged one:
fN D fN (12.4)
averaged sum is equal to the sum of averaged:

f C g D fN C gN (12.5)
operators of averaging and differentiation commutate:

@f @fN @f @fN
D ; D (12.6)
@t @t @x @x
averaged product of two fluctuating quantities is not zero:
129
f 0 g 0 ¤ 0; fNg D fNgN (12.7)
averaged product of any averaged quantity and fluctuation is zero:
fNg 0 D fNgN0 D 0 (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 (12.9)
@t @xj @xj
Here we use the summation convention of Einstein:
@ui @ui @ui @ui

uj D u1 C u2 C u3 (12.10)
@xj @x1 @x2 @x3
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 (12.11)
@xi @x1 @x2 @x3
Using the continuity equation the convective term is written in the conser-
vative form:
@ui @ui @uj @.ui uj /

uj D uj C ui D (12.12)
@xj @xj @xj @xj
With (12.12) the NS equation reads
@ui @.ui uj / 1 @j i

C D Fi C (12.13)
@t @xj @xj
This equation is valid for fluctuating quantities represented in Reynolds
form (12.1)
@.uN i C u0i / @.uN i C u0i /.uNj C uj0 / 129

N 0 1 @.Nj i C j0 i /
C D Fi C F C (12.14)
@t @xDownload
j
free eBooks at bookboon.com@xj
Both r.h.s. and l.h.s. are averaged:

uj D uj C ui D (12.12)
@xj @xj @xj @xj
With (12.12) the NS equation reads
Lectures on computational fluid i @.ui uj /

@udynamics 1 @j i Averaged Navier Stokes Equation (RANS)
Reynolds
C D Fi C (12.13)
@t @xj @xj
This equation is valid for fluctuating quantities represented in Reynolds
form (12.1)
@.uN i C u0i / @.uN i C u0i /.uNj C uj0 / 1 @.Nj i C j0 i /

C D FNi C F 0 C (12.14)
@t @xj @xj
Both r.h.s. and l.h.s. are averaged:
@.uN C u0 / @.uN i C u0 /.uNj C u0 130 / 1 @.Nj i C 0 /

@.uN ii C u0ii / C @.uN i C u0ii /.uNj C ujj0 / D FNNi C F 00 C 1 @.Nj i C jj0 ii / (12.15)
@t C @x D Fi C F C @x (12.15)
@t @xjj @xjj
Utilization of Reynolds averaging properties results in:
Utilization of Reynolds averaging properties results in:
@uN @.uN i uNj C u0 u0 / 1 @N
@uN ii C @.uN i uNj C u0ii ujj0 / D FNNi C 1 @Njj ii (12.16)
@t C @x D Fi C @xj (12.16)
@t @xjj @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:
ˇ ˇ
ˇu00 u00 u00 u00 u00 u00 ˇ
ˇuxx uxx uxx uyy uxx uzz ˇ
RR =ˇ 0 0
ij ij ˇ u0x uy 0 uy00 uy00 uy00 u00z ˇˇ (12.18)
R ij ˇux 0
uy0
uy 0
uy0 uy0 uz0 ˇ (12.18)
ˇu
u0x u0z uy0 u0z u0z u0z ˇ
x uz uy uz uz uz
Due to symmetry condtitions the number of unknown stresses is six. The
term %u00i uj00 is called the stress since it has the same appearance in NS
term %ui uj is called the stress since it has the same appearance in NS
equation as the laminar stress:

@ui @uj
D @ui C @uj pı (12.19)
jj ii D @xj C @xi pıijij (12.19)
@xj @xi
Laminar stress appears due to viscosity effects whereas the Reynolds stress
is caused by flow fluctuations. Now the system of four fluid equations (three
URANS+ continuity) has ten unknowns: three averaged velocity compo-
nents uN , averaged pressure pN and six Reynolds stresses. The system of
nents uN ii , averaged pressure pN and six Reynolds stresses. The system of
fluid dynamics is not closed. Additional relations are necessary to express
the Reynolds stresses through the velocities and pressure. This problem of
determination of Reynolds stresses is called as the closure problem of the
130
131
Download 131
free eBooks at bookboon.com
term %u0i uj0 is called the stress since it has the same appearance in NS

Lectures on computational fluid dynamics @ui @uj Reynolds Averaged Navier Stokes Equation (RANS)
j i D C pıij (12.19)
@xj @xi
nents uN i , averaged pressure pN and six Reynolds stresses. The system of
turbulence. A huge amounts of closure models was developed within the
framework of URANS methodology.131As said above the URANS methods
models the biggest vortices of the flow. The resting scales, which are filtered
by the Reynolds avergaing out, are big enough and not universal. This is
the reason why the URANS models are not universal. Most of them are of
semi empirical character. They are based on submodels with few constants
selected for simple canonical flows. Non universality of closure models is the
biggest weakness of URANS turbulence modelling.
The majority of URANS models used in engineering are based on the Boussi-
nesq hypothesis which is the formal extension of the Newton hypothesis to

turbulent flows. Boussinesq proposed
to express the Reynolds stress through
1 @uN j @uN i
the strain rate tensor Sij D 2 @xi
C @xj
in the form of the Newton hy-
pothesis with the only difference that instead of the kinematic viscosity the
turbulent viscosity t is used

0 0
@uNj @uN i 2
ui uj D t C ıij k (12.20)
@xi @xj 3
In the Trust
simplest and
form for flow along the plate with ux .y/ and uy D uz D 0
responsibility
the formula (12.20) reads
NNE and Pharmaplan have joined forces to create – You have to be proactive and open-minded as a
NNE Pharmaplan, the world’s leading engineering
0 u0 D
d uN x
newcomer and make it clear to your colleagues what
u
and consultancy company focused entirely
x y on thet (12.21) field is new
you are able to cope. The pharmaceutical
pharma and biotech industries. dy
to me. But busy as they are, most of my colleagues
find the time to teach me, and they also trust me.
The last
Inésterm inEsteva
Aréizaga (12.20) is 25introduced
(Spain), years old to keep the
Evenconsistency.
though it was a bitIndeed
hard at fithe
rst, I can feel over
sum of Education:
three diagonal terms of the Reynolds matrix
Chemical Engineer timeis equal
that to the turbulent
I am beginning to be taken seriously and
that my contribution is appreciated.
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
NNE Pharmaplan is the world’s leading engineering and consultancy company
focused entirely on the pharma and biotech industries. We employ more than
= Re=Re
1500 people worldwide and offer global reach and local knowledge along with
our all-encompassing list of services. t cri t
nnepharmaplan.com
(12.22)
The URANS closure models are subdivided into algebraic and differential
ones. The most prominent model amount the algebraic models is the Prandtl
132

turbulent flows. Boussinesq proposed
to express the Reynolds stress through
1 @uN j @uN i
the strain rate tensor Sij D 2 @xi
C @xj
in the form of the Newton hy-
pothesis
Lectures onwith the onlyfluid
computational difference
dynamicsthat instead of the kinematic
Reynolds Averaged viscosity the
Navier Stokes Equation (RANS)
turbulent viscosity t is used

@uNj @uN i 2
u0i uj0 D t C ıij k (12.20)
@xi @xj 3
In the simplest form for flow along the plate with ux .y/ and uy D uz D 0
the formula (12.20) reads
d uN x
u0x uy0 D t (12.21)
dy
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 equa-
tion 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)
where
132
Cb1 D 0; 1355; Cb2 D 0; 622; C1 D 7; 1; D 2=3;
Cb1 1 C Cb2 (12.24)
viscosity . The equation for reads
2
@ @ 1 @ @ Cb2 d d
C uNon
Lectures D Cb1 fluid
j computational Cdynamics
w1 fw C . C /
Reynolds AveragedCNavier Stokes Equation (RANS)
@t @xj d @xk @xk dxk dxk
(12.23)
where
Cb1 D 0; 1355; Cb2 D 0; 622; C1 D 7; 1; D 2=3;

Cb1 1 C Cb2 (12.24)
Cw1 D 2 C ; Cw2 D 0; 3; Cw3 D 2; 0; k D 0; 41
k
6 1=6
3 1 C Cw3
f1 D 3 3
; f2 D ; fw D g 6 6
;
C C1 1 C f1 g C Cw3 (12.25)
6
D ; g D r C Cw2 .r r/; rD 2 2
k d

p 1 @uN i @uNj
D S C 2 2 f2 ; S D 2˝ij ˝ij ; ˝ij D (12.26)
k d 2 @xj @xi
Within the more advanced k " model the turbulent kinetic energy and the
dissipation rate are calculated from the transport equations:
9
@k @k @ t @k @uN i >
>
C uNj D C 133 C ij " >
=
@t @xj @xj k @xj @xj
(12.27)
@" @" @ t @" C"1 " @uN i C"2 "2 >
>
C uNj D C C ij >
;
@t @xj @xj " @xj k @xj k
If k and " are known the turbulent viscosity t can be found from the di-
mension analysis, applied to the dissipation rate
" k 3=2 =L (12.28)

turbulent kinematic viscosity
p
t D C kL (12.29)
Here C D 0:09 is the empirical constant. Substitution of (12.28) into (12.29)
leads to the sought relation:
p k2
t D C kL D C (12.30)
"
If t is known, the Reynolds stresses can be determined from the Boussinesq
approach (12.20).
133
Chapter 13
Reynolds Stress Model (RSM)
Chapter 13
Chapter
Reynolds13
Stress Model (RSM)
13.1 Derivation of the RSM Equations
13.1
13.1.1 Derivation
Step 1 of the RSM Equations
13.1 Derivation of the RSM Equations
The k t h Step
13.1.1 Navier-Stokes
1 equation
13.1.1 Step 1
@uk
The k t h Navier-Stokes @
equation 1 @p 1 @
C .uj uk / D C j k (13.1)
The k t h Navier-Stokes
@t equation
@x j @xk @xj
@uk @ 1 @p 1 @
is multiplied with @u C @ .ucomponent
thek velocity j uk / D 1 u@p i C 1 @ j k (13.1)
@t C @xj .u u / D @xk C @xj (13.1)
@t @xj
j k
@xk @xj
jk

is multiplied with @uthe
k @
velocity component u i 1 @p 1 @
ui C .uj uk / D ui C j k (13.2)
is multiplied with @tthe velocity
@xj component
ui @xk @xj
@uk @ 1 @p 1 @
ui @u C
The i t h equation is .uj uk /
multiplied
@ D uki th1velocity
with @p @ j k
C 1component: (13.2)
ui @tk @x
C j .uj uk / D ui @x k C @x j (13.2)
@t @xj @xk @xj
jk
The i t h equationi is multiplied with k th velocity component:
@u @ 1 @p 1 @
uk C .uj ui / D uk C j k (13.3)
The i t h equation
@t is multiplied
@xj with k thvelocity
@xi component:
@xj
@ui @ uk 1 @p C 1 @ j k
uk @u (13.2)
Resulting equations C @ and i / D are
.uj u(13.3) then1 sumed:
@p 1 @ (13.3)
uk @t C @xj .uj ui / D uk @xi C @xj j k
i
(13.3)
@t @xj @xi @xj
Resulting equations (13.2) and (13.3) are then sumed:
@.ui uequations
Resulting k/ @.u(13.2)
i uk uj /and (13.3) @p are then @p sumed:@j k @j i
C D ui uk C ui C uk (13.4)
This e-book
@t @xj @xk @xi @xj @xj
@.ui uk / @.ui uk uj / @p @p @j k @j i
@.u C @.u u u / D ui @p uk @p C ui @ C uk @ (13.4)
@ti uk / of
Substitution C Reynolds
i jk jdecomposition
@x D ui @xk uk @xi C ui @xj C uk @xj
jk ji
(13.4)
is made with SETASIGN

Substitution of u Reynolds 0
decomposition 0 0
i D u N i C ui ; p D pN C p ; j i D Nj i C j i (13.5)
Substitution of Reynolds decomposition
SetaPDF
into the equation ui (13.4)
D uN i Cresults
u0i ; p inD pN C p 0 ; D N C 0 (13.5)
ji ji ji
ui D uN i C ui ; p D pN C p 0 ; j i D Nj i C j0 i
0
(13.5)
into the equation (13.4) results in 135
into the equation (13.4) results in
135
135
PDF components for PHP developers
www.setasign.com

@t @xj @xk @xj
The i t h equation is multiplied with k th velocity component:

@u
Lectures on computational @
fluid dynamics 1 @p 1 @
j k Reynolds(13.3)
Stress Model (RSM)
i
uk C .uj ui / D uk C
@t @xj @xi @xj
Resulting equations (13.2) and (13.3) are then sumed:
@.ui uk / @.ui uk uj / @p @p @j k @j i

C D ui uk C ui C uk (13.4)
Substitution of Reynolds decomposition
ui D uN i C u0i ; p D pN C p 0 ; j i D Nj i C j0 i (13.5)

into the equation (13.4) results in
@ @ @
.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 (13.6)
@xk @xk
@pN @p 0 @j k @j0 k @j i @j0 i
uN k u0k C uN i C u0i C uN k C u0k
@xi @xi @xj @xj @xj @xj
13.1.2 Step 2
The k t h Reynolds averaged Navier Stokes equation
@uN k @ @pN @
C .uNj uN k / D C .Nj k uj0 u0k / (13.7)
@t @xj @xk @xj
is multiplied with the i th component of averaged velocity

@uN k @ @pN @ 0 0
uN i C .uNj uN k / D uN i C .Nj k uj uk / (13.8)
@t @xj @xk @xj
Again the i th Reynolds averaged Navier Stokes equation is multiplied with
the k t h component of averaged velocity

@uN i @ @pN @ 0 0
uN k C .uNj uN i / D uN k C .Nj k uj ui / (13.9)
@t @xj @xi @xj
The sum of two last equations reads
@.uN i uN k / @.uN i uN k uNj / @pN @pN @.Nj k uj0 u0k / @.Nj i uj0 u0i /
C D uN i uN k CuN i CuN k
(13.10)
13.1.3 Step 3
Subtracting the last equation from (13.6) results in
@ @ @ @p 0 @p 0
.u0i u0k / C .uNj u0i u0k / C .u0i uj0 u0k / D u0i u0k C
@t @xj @xj @xk @xi
(13.11)
@j0 k @j0 i @uN135
i @uN k
C u0i C u0k uj0 u0k uj0 u0i
@xj @x
Download
j @xj at bookboon.com
free eBooks @xj
136
@.uN i uN k / @.uN i uN k uNj / @pN @pN @.Nj k uj0 u0k / @.Nj i uj0 u0i /
C D uN i uN k CuN i CuN k
Lectures on computational fluid dynamics Reynolds Stress Model (RSM)
(13.10)
13.1.3 Step 3
Subtracting the last equation from (13.6) results in
@ @ @ @p 0 @p 0
.u0i u0k / C .uNj u0i u0k / C .u0i uj0 u0k / D u0i u0k C
@t @xj @xj @xk @xi
(13.11)
@j0 k @j0 i @uN i @uN k
C u0i C u0k uj0 u0k uj0 u0i
@xj @xj @xj @xj
Using identities (please prove them) 136
0
@p 0 @p 0 @ui @u0k @ @
u0i 0
uk Dp 0 C ıj k 0 0
.u p / C ıij 0 0
.u p /
@xk @xi @xk @xi @xj i @xj k
(13.12)

@j0 k @j0 i @2 u0k @2 u0i @2 0 0 @u0i @u0k
u0i Cu0k 0
D ui 0
Cuk 2 D 2 ui uk 2 (13.13)
@xj @xj @xj2 @xj @xj @xj @xj
we get the Reynolds stress model equation
@ @ @
.u0i u0k / C .uNj u0i u0k / C .u0i uj0 u0k / D
@t @xj @xj
0
0
@ui @u0k @ 0 0
@ 0 0
p C ıj k .u p / C ıij .u p / C
@xk @xi @xj i @xj k
@2 0 0 @u0i @u0k
C u u 2
@xj2 i k @xj @xj
@uN i @uN k
uj0 u0k
uj0 u0i
@xj @xj
which can be written in a compact form
@ 0 0 @ @
.ui uk / C uNj .u0i u0k / D Di k C Ri k C Pi k "i k (13.14)
@t @xj @xj
The physical meaning of terms on the r.h.s is as follows
@.u0i u0k / 1
Di k D u0i uj0 u0k C .ıj k u0i C ıij u0k /p 0 ! Diffusion (13.15)
@xj

1 @u0i @u0k
Ri k D C p0 ! Re distribution (energy exchange) (13.16)
@xk @xi
@uN i @uN k
Pi k D uj0 u0k uj0 u0i ! Generation (13.17)
@xj @xj
@u0i @u0k
"i k D 2 !136
Dissipation (13.18)
@xj @xj
137

1 @u0i @u0k
Ri k D C p0 ! Re distribution (energy exchange) (13.16)
@xk @xi
@uN i
Lectures on computational fluid0 dynamics
0
@uN k Reynolds Stress Model (RSM)
Pi k D uj uk uj0 u0i ! Generation (13.17)
@xj @xj
@u0i @u0k
"i k D 2 ! Dissipation (13.18)
@xj @xj
13.1.4 Analysis of terms 137

The diffusion of energy
@.u0i u0k / 1
Di k D u0i uj0 u0k C .ıj k u0i C ıij u0k /p 0 (13.19)
@xj
is due to
molecular diffusion, described by the term:
@.u0i u0k /
(13.20)
@xj
turbulent diffusion, described by the term:
u0i uj0 u0k (13.21)
turbulent diffusion caused by correlation between pressure and velocity

fluctuations
1
.ıj k u0i C ıij u0k /p 0 (13.22)

The two last terms are unclosed. Here we face with the famous problem noted
first by Friedman and Keller (1924): Effort to derive the equations for the
second Sharp
order moments Minds u0i u 0
- kBright
results inIdeas!
the necessity of determination of new
unclosed terms including third order moments u0i uj0 u- First
Employees at FOSS Analytical A/S are living proof of the company value
0
k . -Using
using
the method
The Family owned FOSS group is
proposednewby Friedman
inventions and Keller
to make dedicated solutionsin
for1924 it is With
our customers. possible to and
sharp minds derive equations
the world leader as supplier of
cross functional teamwork, we constantly strive to develop new unique products -
for moments of arbitrary order.
Would you like to join our team?
However, the equation for the m thdedicated, high-tech analytical
order
solutions which measure and
will contain unclosed moments of the m C 1 th order. Impossibility ofquality and produc-
control the
FOSS works diligently with innovation and development as basis for its growth. It is tion of agricultural, food, phar-
obtaining of a closed system of equations for a finite number of moments,
reflected in the fact that more than 200 of the 1200 employees in FOSS work with Re- maceutical and chemical produ-
known search
as the Friedman-Keller
& Development in Scandinaviaproblem is a direct
and USA. Engineers consequence
at FOSS work in production, of the cts.
nonlin-
Main activities are initiated
development and marketing, within a wide range of different fields, i.e. Chemistry,
earity of the Navier Stokes equations.
Electronics, Mechanics, Software, Optics, Microbiology, Chemometrics.
from Denmark, Sweden and USA
with headquarters domiciled in
Hillerød, DK. The products are
The also unclosed term
We offer marketed globally by 23 sales
A challenging job in an international and innovative company that is leading in its field. You will get the companies and an extensive net

opportunity to work with the most advanced technology together with highly skilled colleagues. of distributors. In line with
1 @u0 @u0k the corevalue to be ‘First’, the
Ri k- or
Read more about FOSS at www.foss.dk Dgo directly toiourCstudent site p0
www.foss.dk/sharpminds where (13.23)
company intends to expand
you can learn more about your possibilities of @xtogether
working k @x
with i projects, your thesis etc.
us on
its market position.
describes the redistribution of the energy between different tensor compo-

Dedicated Analytical Solutions
nents u0i u0k caused by correlation between the stresses and pressure fluctua-
FOSS
tions. Slangerupgade 69
3400 Hillerød
Tel. +45 70103370
www.foss.dk 138

(13.20)
@xj
turbulent diffusion, described by the term:

Lectures on computational fluid dynamics Reynolds Stress Model (RSM)
u0i uj0 u0k (13.21)
turbulent diffusion caused by correlation between pressure and velocity

fluctuations
1
.ıj k u0i C ıij u0k /p 0 (13.22)

The two last terms are unclosed. Here we face with the famous problem noted
first by Friedman and Keller (1924): Effort to derive the equations for the
second order moments u0i u0k results in the necessity of determination of new
unclosed terms including third order moments u0i uj0 u0k . Using the method
proposed by Friedman and Keller in 1924 it is possible to derive equations
for moments of arbitrary order. However, the equation for the m th order
will contain unclosed moments of the m C 1 th order. Impossibility of
obtaining of a closed system of equations for a finite number of moments,
known as the Friedman-Keller problem is a direct consequence of the nonlin-
earity of the Navier Stokes equations.
The also unclosed term

1 @u0i @u0k
Ri k D C p0 (13.23)
@xk @xi
describes the redistribution of the energy between different tensor compo-
nents u0i u0k caused by correlation between the stresses and pressure fluctua-
tions.
The term
The term
138
@uN i @uN k
Pi k D uj00 u0k0 @uN i
uj00 u0i0 @uN k (13.24)
Pi k D uj uk @xj uj ui @xj (13.24)
@xj @xj
is responsible for the energy generation, i.e. the transport of the energy
is responsible
transfer for the energy
from averaged (mean)generation, i.e. theflow
flow to oscillating transport of the energy
(fluctuations). And,
transfer
finally, from averaged (mean) flow to oscillating flow (fluctuations). And,
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
of the
the dissipation. This energy
turbulent kinetic unclosed intoterm
theisinner
responsible
energy of forthe
theflow.
transformation
of the turbulent kinetic energy into the inner energy
RSM model based on equations (13.14) is used to determine the Reynoldsof the flow.
RSM model
stresses frombased on equations
the transport (13.14) It
equations. is used
is nottobased determine
on thetheBoussinesq
Reynolds
stresses from the transport equations. It is not based
approach and takes the anisotropy of stresses into account. This model is on the Boussinesq
approach
the best oneandamong
takes RANS
the anisotropy
models. of stresses into account. This model is
the best one among RANS models.
138
Chapter 14
Equations of the k – ε Model
Chapter
Chapter 14
Equations14of the k - " Model
Equations
Equations of
of the
the k
k -- "" Model
Model
14.1 Derivation of the k-Equation
14.1 Derivation
According to definitionof the k-Equation
14.1 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
uNj
@
Reynolds @
0 0 stress model equations (13.14)
.ui uk / D Di k C Ri k C Pi k "i k (14.2)
Assuming @t i Di kk in the@x Reynolds
j stress
@xj model equations (13.14)
@ 0 0 @ 0 0
@
ans summing
@ k / C uNj @for.uk=1,2
.ui uequations i uk / Dand ik C
@ 3Dwe Ri k Cthe
obtain "i k
Pi ktransport (14.2)
equations
@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
@t kinetic
i k energy
@xj k: i k
@xj
ans summing equations for k=1,2 and 3 we obtain the transport equations
ans the
for summing equations
total kinetic @kfor k:
energy @k and @3 we obtain the transport equations
k=1,2
C uNj D Ds C P "S (14.3)
for the total kinetic energy @t k: @xj @xj
@k @k @
where @k C uNj @k D @ Ds C P "S (14.3)
@t C uNj @xj D @xj Ds C P "S (14.3)
@t @xj @xj
where
whereD D @k C 1 ı u0 p 0 u0 k 0 ; k 0 D u0 u0 =2 Diffusion (14.4)
S jk k j k k
@xj
@k 1
DS D @k C 1 ıj k u0k p 0 uj0 k 0 ; k 0 D u0k u0k =2 Diffusion (14.4)
DS D @xj C ıj k u0k p 0 0 uj00 k@0u;N k k 0 D u0k u0k =2 Diffusion (14.4)
@xj P D uj uk Generation (14.5)
@xj
0 0
@uN k
0P D 0 uj uk @u Generation (14.5)
@ukP@uDk u0 u0 @xN jk Generation (14.5)
"S D j Dissipation
k
@xj ( Pseudodissipation) (14.6)
@xj @xj
@u0 @u0k
D @uk0the
The relation"Sbetween @u0true and Dissipation ( Pseudodissipation)
pseudodissipation is (14.6)
"S D @xjk @xjk Dissipation ( Pseudodissipation) (14.6)
@xj @xj0 0 2
The relation between the@u true @u and
j pseudodissipation
141 @ @ is
" D the ktrue
The relation between C and pseudodissipation
"S C isuj0 u0k (14.7)
2 @xj @xk @xj @xk
141
For large Reynolds numbers the true141 dissipation and the pseudodissipation
are equal.
" "S (14.8)

More precise analysis shows that
" "S Re t1 (14.9)

where
p
Re t D kL= (14.10)
139
Two unknown terms in the diffusion
@k 1 0 0 0 0
" "S (14.8)
More precise analysis shows that
" "S Re t1 (14.9)
where
p
Re t D kL= (14.10)
Two unknown terms in the diffusion
@k 1
DS D C ıj k u0k p 0 uj0 k 0 (14.11)
@xj
are determined by the gradient assumption
1 t @k
ıj k u0k p 0 uj0 k 0 D (14.12)
k @xj
where k is an empirical constant. Dissipation is determined by the energy
containing motion using the formula of Prandtl- Kolmogorov
"s D CD k 3=2 =L (14.13)

The Reynolds stresses are seeking in form proposed by Boussinesq:

@uNj @uN i 2
u0i uj0 D t C ıij k (14.14)
@xi @xj 3
Substitution of all these approximations into the equation (14.3) results in
the k-Equation

@k @k @ t @k @uNj @uN i @uNj k 3=2
C uNj D C C t C CD (14.15)
@t @xj @xj k @xj @xi @xj @xi L
142

"s D CD k =L (14.13)
The Reynolds stresses are seeking in form proposed by Boussinesq:

Lectures on computational fluid
0 dynamics
0
@uNj @uN i 2
u i uj D t C ıij k (14.14)
@xi @xj 3 Equations of the k – ε Model
Substitution of all these approximations into the equation (14.3) results in

the k-Equation

@k @k @ t @k @uNj @uN i @uNj k 3=2
C uNj D C C t C CD (14.15)
@t @xj @xj k @xj @xi @xj @xi L
142
14.2 Derivation of the "-Equation
The Navier Stokes equation
@uk @uk 1 @p 1 @j k

C uj D C (14.16)
@t @xj @xk @xj
@u0
is differentiated and multiplied with the derivative @xki

@ @ui @ui 1 @p 1 @j i @u0i
C uj D C (14.17)
@xk @t @xj @xi @xj @xk
and then it is averaged. The following identities are used in transformations:

@ @ui @u0i @ @.uN i C u0i / @u0i @ @u0i @u0i
D D D
@xk @t @xk @t @xk @xk @t @xk @xk
(14.18)
1 @ @u0i @u0i 1 @"s
D D
2 @t @xk @xk 2 @t

@ @ui @u0i @ @uN i @u0i @ 0
@uN i @u0i
uj D uNj C u C
@xk @xj @xk @xk @xj @xk @xk j @xj @xk
(14.19)
@ @u0i @u0i @ 0
@u0i @u0i
C uNj C u
@xk @xj @xk @xk j @xj @xk

@ @u0i @u0i @uNj @u0i @u0i @2 u0i @u0i
uNj D C uNj D
@xk @xj @xk @xk @xj @xk @xj @xk @xk
@uN i @u0i @u0i 1 @ @u0i @u0i
D C uNj D (14.20)
@xk @xj @xk 2 @xj @xk @xk
@uN i @u0i @u0i 1 @
D C uNj "s
@xk @xj @xk 2 @xj

@ 0
@u0i @u0i @uj0 @u0i @u0i @2 u0i @u0i
uj D C uj0 D
@uj0 @u0i @u0i 1 @ 0 @u0i @u0i
D C u D (14.21)
@xk @xj @xk 2 @xj j @xk @xk
@uj0 @u0i @u0i 1 @ 0
D C u "s
@xk @xj @xk 2 @xj j
141
143
@uN i @u0i @u0i 1 @ @u0i @u0i
D C uNj D (14.20)
@xk @xj @xk 2 @xj @xk @xk
@uN i @u0i @u0i 1 @
D C uNj "s
@xk @xj @xk 2 @xj

@ 0
@u0i @u0i @uj0 @u0i @u0i @2 u0i @u0i
uj D C uj0 D
@uj0 @u0i @u0i 1 @ 0 @u0i @u0i
D C u D (14.21)
@xk @xj @xk 2 @xj j @xk @xk
@uj0 @u0i @u0i 1 @ 0
D C u "s
@xk @xj @xk 2 @xj j
143
@ 0
@uN i @u0i @uN i @uj0 @u0i @2 uN i 0 @u0i
uj D C u
@xk @xj @xk @xj @xk @xk @xk @xj j @xk
(14.22)
1 @p @u0i 1 @p 0 @u0i
D
% @xi @xk % @xi @xk
0 0
1 @ @ @u 1 @ @ @ui
jN i C j0 i i
D j0 i (14.23)
% @xk @xj @xk % @xk @xj @xk
0
@ 0 @ @ui @uj0 @2 u0i
j i D C D (14.24)
@xj @xj @xj @xi @xj @xj
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

@ 1 @ @u0i @u0i @2 u0i @2 u0i
D
@xj 2 @xj @xk @xk @xj @xk @xj @xk
(14.25)
This gives:
@"s @"s @
C uNj D D" C P" "" (14.26)
@t @xj @xj
where
@"s @u0i @p 0
D" D uj0 "0s 2 (14.27)
@xj @xk @xi
0 0
@u0i @2 uN i @ui @uj @uN i @u0i @u0i @uN i
P" D 2uj0 2 C
@xk @xj @xk @xk @xk @xj @xj @xk @xk
(14.28)
@u0i @uj0 @u0i
2
@xj @xk @xk
@2 u0i @2 u0i @u0i @u0i

"" D 2 2 ; "0s D (14.29)
@xj @xk @xj @xk @xk @xk
The terms on the r.h.s. were approximated according to the following for-
mula:
142
144

@ @ T @" C"1 " @uN i C"2 "2
D" D C I P" D ij I "" D (14.30)
@xj @xj " @xj k @xj k
Constants are taken from planar jet and mixing layer:
C"1 D 1:44; C"2 D 1:92; k D 1; " D 1:3 (14.31)

Hereby the full closed system of the k " model reads:
9
@k @k @ t @k @uN i >
>
C uNj D C C ij " >
=
@t @xj @xj k @xj @xj
(14.32)
@" @" @ t @" C"1 " @uN i C"2 "2 >
>
C uNj D C C ij >
;
@t @xj @xj " @xj k @xj k
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
" (14.33)
L
From the dimension analysis
p
The Wake
follows
t D C kL (14.34)
the only emission we want to leave behind

p k2
t D C kL D C (14.35)
"
As soon as k and " are known the turbulent kinematic viscosity t is com-
puted from (14.35) and Reynolds stresses can be calculated from the Boussi-
nesq 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 approxima-
tions used in derivation of k " model equations are not valid. To overcome
.QYURGGF'PIKPGU/GFKWOURGGF'PIKPGU6WTDQEJCTIGTU2TQRGNNGTU2TQRWNUKQP2CEMCIGU2TKOG5GTX
this disadvantage various low Reynolds k " models were proposed.

6JGFGUKIPQHGEQHTKGPFN[OCTKPGRQYGTCPFRTQRWNUKQPUQNWVKQPUKUETWEKCNHQT/#0&KGUGN6WTDQ
2QYGTEQORGVGPEKGUCTGQHHGTGFYKVJVJGYQTNFoUNCTIGUVGPIKPGRTQITCOOGsJCXKPIQWVRWVUURCPPKPI
HTQOVQM9RGTGPIKPG)GVWRHTQPV 145
(KPFQWVOQTGCVYYYOCPFKGUGNVWTDQEQO

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

" (14.33)
L Equations of the k – ε Model
From the dimension analysis

p
t D C kL (14.34)
follows
p k2
t D C kL D C (14.35)
"
As soon as k and " are known the turbulent kinematic viscosity t is com-
puted from (14.35) and Reynolds stresses can be calculated from the Boussi-
nesq 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 approxima-
tions used in derivation of k " model equations are not valid. To overcome
this disadvantage various low Reynolds k " models were proposed.
14.3 Method of wall 145

functions
Basic relations of k " model werep derived under assumption that the local
turbulent Reynolds number Re t D kL= is very high. This is not the case
in the wall flow. Due to damping effect of the wall, the velocity pulsations
and the turbulent kinetic energy are small. Application of the k " model
is not valid. An efficient solution of this problem was proposed by Spalding
who developed the method of wall functions. The wall function is an ana-
lytical representation of the solution close to the wall which is matched with
the numerical solution far from the wall. Below we derive basic relations of
the wall function method following to the textbook [15].
It is assumed that the turbulence is in equilibrium, i.e. P D ":
@uN k
P D uj0 u0k D"
@xj
Let us consider two dimensional flow along a plate:

@uN x @uN x 2
P D u0x uy0 D t D" (14.36)
@y @y

@uN x
The derivative @y
can be expressed from the mixing length model of
Prandtl:

2 @uN x @ux
t D l ) D t = l 2 (14.37)
@y @y
From analysis of dimension we obtain
t D c k 1=2 L (14.38)
144
Using the Kolmogorov-Prandtl estimation " k 3=2 =L we get
c k 2
t D (14.39)
@uN x
The derivative @y
can be expressed from the mixing length model of
Prandtl:

2 @uNx
Lectures on computational fluid dynamics @ux
t D l ) D t = l 2 (14.37)
@y @y
From analysis of dimension we obtain
t D c k 1=2 L (14.38)
Using the Kolmogorov-Prandtl estimation " k 3=2 =L we get
c k 2
t D (14.39)
"
From (14.36) and (14.37) it follows
t D l 4=3 "1=3
and
" D c3=4 k 3=2 = l
From (14.36) and (14.39) we have: 146

1=2 1=2
@uN x " "2 "
D D D
@y t c k 2 c1=2 k
for wall stress:
@uN x "
w D t D % t 1=2 D
@y c k
c k 2 "
D% D %c1=2 k
" c1=2 k
and kinetic energy:

w
kD
%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)
and turbulent kinetic energy
w
kp D
%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 (14.41)
w w x uN xp
D 145 k 1=2 c 1=4
ln EypC
p
% %
%x uN xp kp1=2 c1=4
w D
ln EypC

"p D c3=4 kp3=2 = yp (14.40)
and turbulent kinetic energy
Lectures on computational fluid dynamics w

kp D
%c1=2 Equations of the k – ε Model
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 (14.41)
w w x uN xp
D C
kp1=2 c1=4
% % ln Eyp
%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 Z Z
147
@u @u
P dxdy D ij dxdy D w dxdy D w xuxp (14.42)
@y @y
Challenge the way we run
EXPERIENCE THE POWER OF

FULL ENGAGEMENT…
RUN FASTER.
RUN LONGER.. READ MORE & PRE-ORDER TODAY
RUN EASIER… WWW.GAITEYE.COM
1349906_A6_4+0.indd 1 22-08-2014 12:56:57

Chapter 15
Large Eddy Simulation (LES)
Chapter
Chapter 15
15 Simulation (LES)
Large Eddy
Large
Large Eddy
Eddy Simulation
Simulation (LES)
(LES)
15.1 LES filtering
Within the LES
15.1 LES all filtering
vortices are subdivided into large resolved vortices and
15.1
fine modelled LES
subgrid filtering
vortices. The border between vortices should lie within
Within
the the range.
inertial LES allThe vortices are subdivided
separation of fine scale intomotions
large resolved vortices
(small fine and
vortices)
Within
fine
from the ones
modelled
large LES isalldone
subgrid vortices
usingare
vortices. The
the subdivided into large
border filtering.
spatial between Letresolved
vortices
' be vortices
should
any and
lie within
stochastic
fine
the modelled
inertial
function which subgrid
range. vortices.
The separation
is represented The border
as theofsum fine of between
scale vortices
motions
filtered part(small should lie within
fine vortices)
and fluctuation:
the inertial
from range.
large ones The separation
is done of fine scale
using the spatial motions
filtering. Let ' (small finestochastic
be any vortices)
from large ones is done using
function which is represented as 'the the spatial
D sum
'Q C ' filtering.
0 Let ' be any
of filtered part and fluctuation: stochastic
function which is represented as the sum of filtered part and fluctuation:
where the filtered part is defined as
' D 'Q C ' 0
Z 1 Z' 1 DZ'Q 1 C '0
where the filtered'. Qpart
E t/isDdefined as
x; '.xE sE; t/F .Es /d sE
where the filtered part is defined Z1 Z as Z1
1
1 1 1
Q x;
Here F(Es ) is the filtering
'. E t/ Dfunction, Z 1 '.xE the
Z 1 Z 1satisfying sE;condition
t/F .Es /d sE
Q x;
'. E t/ D 1 1 1 '.xE sE; t/F .Es /d sE
Here F(Es ) is the filtering function, Z1
1Z11 Zsatisfying
1 1
the condition
Here F(Es ) is the filtering function, satisfying F .Es /d sE the
D 1 condition
Z
1
1 Z1 Z1
1 1
Z1 Z1 Z1 F .Es /d sE D 1
Three different filtering functions shown insEFig.
F .Es /d D 1 1: ideal filter, Gauss filter
1 1 1
and top hat filter. Ideal filter is applied in Fourier space. High frequencies
1 1 1
Three
are cutdifferent
off. Low filtering
frequencies functions shown in
are simulated Fig. 1:The
directly. idealtopfilter, Gauss
hat filter filter
is some
Three
and
kind ofdifferent
top hat filter.
smoothing filtering
Ideal functions
applied filter shown
is applied
in physical in Fourier
in
space. Fig.
The 1: ideal filter,
space.
simplest HighisGauss
case filter
frequencies
smoothing
and top
are cut
over hat filter.
off. neighboring
three Ideal
Low frequencies filter
points is applied in Fourier space. High
are simulated directly. The top hat filter is some frequencies
are cut
kind of off. Low frequencies
smoothing applied inare simulated
physical directly.
space. The topcase
The simplest hat is
filter is some
smoothing
kind of smoothing applied
over three neighboring points in 1
physical space. The simplest case is smoothing
'Qi D .'i 1 C a'i C 'i C1 /
over three neighboring points b
where b D 2 C a. 1
'Qi D 1 .'i 1 C a'i C 'i C1 /
'Qi D b .'i 1 C a'i C 'i C1 /
where b D Properties
2 C a. b
15.1.1 of filtering 149
where b D 2 C a.
The spatial filtering and Reynolds averaging 149 are both filtering operations.
LES spatial filtering has properties which 149 differ from these of Reynolds av-
eraging. First, the spatialy averaged quantity is not zero. Double filtering is
not equal once filtering:
e0 ¤ 0;
' 'QQ ¤ ';
Q
Both conditions are compatible because
A
e0 D ' 'Q D 'Q 'QQ ¤ 0
'
147
Both conditions are compatible because
' A
e0 D ' 'Q D 'Q 'QQ ¤ 0
Lectures on computational fluid dynamics Large Eddy Simulation (LES)
Figure 15.1: Different filtering functions used in LES.
Other important properties are similar to these of RANS averaging:
Averaged sum of two quantities is equal to the sum of averaged quan-

150
tities:
A
' C g D 'Q C gQ
Filtering operator commutes with the differentiation operator, if filter-

ing is homogeneous:
f
@'
D
@'Q e
@'
D
@'Q
@t @t @xj @xj
A very important relation which is the consequence of these properties is
'f D 'f e e e
Q Q C ' Q 0 C ' 0 0
Q 0 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 Small scale motion

Generated by mean flow Generated by large scale structures
Depends on the flow geometry Universal
Regular 148
Stochastic
Deterministic description Stochastic description
Heterogeneous Homogeneous
Anisotrop Isotrop
A very important relation
'f D 'f
Q C ' e e e
whichQ is the0 consequence
Q C ' Q 0 C ' 0 0of these properties is
'f
f
D 'f
Q Q C '
f
In the case of Reynolds'averagingQ Q C
D 'Q Conly
' e e
e e e
Q 0 C ' 0 0
Q 00 C 'e
Q first
the
' 0 0 0 the last terms remain.
C 'and

In
Thethe case of Reynolds
properties averaging
of large and only motions
small scale the first are
andshown
the last terms
in the remain.
table 15.1.
In
Thethe
Lectures case
on of Reynolds
computational
properties of largefluidaveraging
dynamics
and only motions
small scale the first are
andshown
the last terms
Large
in the remain.
Eddy
table Simulation (LES)
15.1.
TheLarge
properties of motion
scale large and small scale motions
Small scale aremotion
shown in the table 15.1.
Large
Generated scale motion
by motion
mean flow Small scale
Generated motion
bymotion
large scale structures
Large scale Small scale
Generated
Depends onbythe mean
flowflow
geometry Generated
Universal by large scale structures
Generated
Depends by mean flow Generated by large scale structures
Regular on the flow geometry UniversalStochastic
Depends
Regular on the flow geometry Universal
Stochastic description
Deterministic description
Regular
Deterministic Stochastic
Heterogeneousdescription Stochastic
Homogeneous description
Deterministic
Heterogeneous description Stochastic
Homogeneous description
Anisotrop Isotrop
Heterogeneous
Anisotrop Homogeneous
Isotrop
Exists long time Exists short time
Anisotrop
Exists long time Isotrop
Exists short time
Diffusive Dissipative
Exists
Diffusivelong time Exists short
Dissipative time
Modelling is complicated Easy to model
Diffusive
Modelling is complicated Dissipative
Easy to model
Modelling
Tableis 15.1:
complicated Easy and
Properties of large to model
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
motion is universal.conclusion
Thereforefrom thisexpect
one can table isthat
thethe
factmodels
that the small scale
describing the
A very
motion important
is universal.conclusion
Thereforefrom
one this
can table
expect is the
that fact
the that
models
small scale motion in contrast to RANS models are also universal. the small scale
describing the
motion is universal. Therefore one can expect that the models
small scale motion in contrast to RANS models are also universal. describing the
small scale motion in contrast to RANS models are also universal.
15.2 LES equations
15.2 LES equations
15.2 LES
The governing equations
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 derived
from the
Navier Stokes equation
@ @ @ @uj @ui @p
.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
@ i @ i j @ @uj @ui @p i (15.1)
@t .ui / C @xj .ui uj / D @xj @xi C @xj @xi C gi (15.1)
@t @xj @xj 151@xi @xj @xi
Technical151 training on
151
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 OIL & GAS
experience in training technical and engineering staff and ENGINEERING
have trained people in organisations such as General
ELECTRONICS
Motors, Shell, Siemens, BHP and Honeywell to name a few.
Our onsite training is cost effective, convenient and completely AUTOMATION &
customisable to the technical and engineering areas you want PROCESS CONTROL
covered. Our workshops are all comprehensive hands-on learning
experiences with ample time given to practical sessions and MECHANICAL
demonstrations. We communicate well to ensure that workshop content ENGINEERING
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
For a no obligation proposal, contact us today POWER
at training@idc-online.com or visit our website
for more information: www.idc-online.com/onsite/
Phone: +61 8 9321 1702

Email: training@idc-online.com
Website: www.idc-online.com

Application of the filter operation to (15.1) results in
C
@
@t
.ui / C
@
@xj
.ui uj / D
@
@xj
e

@uj
@xi
C
Averaged sum is equal to the sum of averaged terms:

@ui
@xj

@p
@xi
C gi
B E F e e
@ @ @ @ui @p
.ui / C .ui uj / D C gi (15.2)
@t @xj @xj @xj @xi
Introducing the term
e
ijS GS D ui uj uQ i uQj
the equation (15.2) is rewritten in the final form

@ @ @ @uQ i S GS @pQ
.uQ i / C .uQ i uQj / D ij C gi
@t @xj @xj @xj @xi
e
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 phenomenolog-
ical model was proposed by Smagorinsky in 1963. The Smagorinsky model
is just the extension of the Boussinesq approach
1
ij kk ıij 2 t SNij ;
3
Smagorinsky introduced the subgrid viscosity S GS instead of the turbulent
kinematic viscosity
1 S GS
ijS GS kk ıij 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
d uN x
t D l 2j j
dy
According to Smagorinsky, the subgrid viscosity is proportional to the mag-
nitude of the strain rate tensor Sij and to a certain length lS squared
150 q
2 Q Q
S GS D Download
lS jSij j; free eBooks 2SQij SQij
jSij j Dat bookboon.com
Where
Prandtl, the turbulent kinematic viscosity is proportional to the mixing
length squared and the velocity gradient close to the wall
Lectures on computational fluid dynamics d uN x Large Eddy Simulation (LES)

t D l 2j j
dy
According to Smagorinsky, the subgrid viscosity is proportional to the mag-
nitude of the strain rate tensor Sij and to a certain length lS squared
q
2 Q
S GS D lS jSij j; jSij j D 2SQij SQij
Q
Where

1 @ Q
u @ Q
u
SQij D
i j
C
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
Assuming additionally that SQ 2 SQ 3 , the length lS and the Smagorinsky
constant are expressed through the Kolmogorov constant C D 1:5:
lS 1
CS D D 0:17
.7C /3=4
The Smagorinsky constant 0:17 is 153

derived analytically with a few strong
assumptions. The experience shows that numerical results agree with mea-
surements much better if a reduced value of the Smagorinsky constant is
used. Common values are 0:065 and 0:1.
Advantages and disdvantages of the Smagorinsky model are summarized in

the table 19.1.
Advantages Disadvantages
Simple Laminar flow is not modelled
Low computational costs Constant of Smagorinsky is constant in time and
space
151
Stable Actually, the constant is chosen arbitrarily
depending on theatproblem
bookboon.com
under consideration
Good accuracy in ideal Sensible to grid
surements much better if a reduced value of the Smagorinsky constant is
used. Common values are 0:065 and 0:1.
Advantages
Lectures and disdvantages
on computational of the Smagorinsky model are summarized
fluid dynamics in
Large Eddy Simulation (LES)
the table 19.1.
Advantages Disadvantages
Simple Laminar flow is not modelled
Low computational costs Constant of Smagorinsky is constant in time and
space
Stable Actually, the constant is chosen arbitrarily
depending on the problem under consideration
Good accuracy in ideal Sensible to grid
conditions
Purely dissipativ
Damping of pulsation is too strong
Table 15.2: Advantages and disadvantages of the Smagorinsky model
We finish this section with a very important comment:
the LES models are consistent when the resolution increases, i.e. ! 0.
Indeed, if the resolution is increased, ! 0, the SGS stresses disappear.

The LES equation is passed to the original Navier Stokes equations. The
LES simulation becomes the DNS simulation if ! 0. On the contrary, the
URANS simulation is not consistent when ! 0. The Reynolds stresses
don’t disappear if the resolution is increased ! 0. The turbulence is then
twice resolved.
I wanted
15.4 real responsibili�
Model of IGermano ( Dynamic www.discovermitas.com
Smagorin- Maersk.com/Mitas �e G
joined MITAS because for Engine
sky Model) I wanted real responsibili� Ma
The classical model of Smagorinsky with the parameter CS being constant

for the whole computational domain is proved to be very diffusive. Germano
proposed to calculate the Smagorinsky constant CS being variable both in
space and in time, i.e. CS D CS .x; t/. The constant is determined using the
154 Month 16
I was
the North Sea super
advising and the No
Real work he
International
al opportunities
Internationa
�ree wo
work
Real work he
helping fo
International
�ree wo
work

LES simulation becomes the DNS simulation if ! 0. On the contrary, the
URANS simulation is not consistent when ! 0. The Reynolds stresses
don’t disappear if the resolution is increased ! 0. The turbulence is then
twice resolved.
15.4 Model of Germano ( Dynamic Smagorin-

sky Model)
The classical model of Smagorinsky with the parameter CS being constant
for the whole computational domain is proved to be very diffusive. Germano
proposed to calculate the Smagorinsky constant CS being variable both in
space and in time, i.e. CS D CS .x; t/. The constant is determined using the
dynamic procedure which is then referred to as the Dynamic Smagorinsky
Model (DSM). 154
According to the definition the subgrid stress is
e
ijS GS D ui uj uQ i uQj (15.4)
Germano introduces the double filtering or the test filtering designated as

OQ Here the tilde symbol means the first filtering with filter width
u D u.
whereas the hat symbol stands for the second filtering with filter width ∼ 2.
The symbol means the resulting double filtering. Using the definition (15.4)
we can write
_
b
e
Tijt est D ui uj ui uj D ui uj uOQ i uOQj (15.5)
Filtration of (15.4) results in
b
e b
OijS GS D ui uj uQ i uQj (15.6)
Subtracting (15.6) from (15.5) yields
b
Tijt est OijS GS D uQ i uQj uOQ i uOQj (15.7)
We suppose that the double filter width is small. Therefore the Smagorinsky
model is valid for both stresses ijS GS and Tijt est :
1 S GS 2
ijS GS kk ıij D 2 Cs jSQij j SQij D 2C mSijGS ;
3 (15.8)
1 t est 2
Tijt est Tkk ıij D 2 Cs O jSOQij j SOQij D 2C mtijest ;
3
The application of the double filter to ijS GS gives:
1 S GS
OijS GS Okk O SijGS
ıij D 2C m (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
b
Cm Cm
O
We introduce the tensor Lij which is equal to the difference between test
filter and once filtered original SGS stress:
153
b
Lij D Tijt est OijS GS D uQ i uQj uOQ i uOQj (15.10)
Download free
155eBooks at bookboon.com
Oij
Okk ıij 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
b
Cm Cm
O
We introduce the tensor Lij which is equal to the difference between test
filter and once filtered original SGS stress:
b
Lij D Tijt est OijS GS D uQ i uQj uOQ i uOQj (15.10)
Using this designation we get from (15.8)

155 and (15.9)
1
2CMij D Lij Lkk ıij (15.11)
3
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
Lij SQij
C D (15.12)
2Mij SQij
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 ıij 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 ıij 2CMij D 0 (15.14)
@C 3
It follows directly from (15.14):
Mij Lij 31 Lkk ıij Mij Mij Lij

C D D (15.15)
2Mij Mij 2Mij Mij
since ıij Mij D 0. The solution (15.15) corresponds to the minimum of Q(C)
since the second derivative @2 Q=@C 2 is positive at this point
@2 Q
D 8Mij Mij > 0 (15.16)
@C 2
Theoretically the constant C can become negative. The case C < 0 and
S GS can be considered as the energy backscattering. However, this leads
to strong numerical instability. That is why the dynamic constant is limited
from below:

Mij Lij
C D max ;0 0 (15.17)
2Mij Mij
156
154
The subgrid kinematic viscosity is always positive
S GS D C2 jSQij j 0
15.5 Scale similarity models
Figure 15.2: Illustrations for derivation of the scale similarity model.
Despite the fact that diffusion of the classic Smagorinsky model was sub-
stantially 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
e
ijS GS D uQ i uQj uQQ i uQQj (15.18)
The formula (15.18) is the scale similarity model proposed by Bardina et
al. [16]. As seen the SGS stress can be calculated directly from the resolved
field uQ i .
157
www.job.oticon.dk

Figure 15.2: Illustrations for derivation of the scale similarity model.
Despite the fact that diffusion of the classic Smagorinsky model was sub-
stantially 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
e
ijS GS D uQ i uQj uQQ i uQQj (15.18)
The formula (15.18) is the scale similarity model proposed by Bardina et
al. [16]. As seen the SGS stress can be calculated directly from the resolved
field uQ i .
157
15.6 Mixed similarity models
The experience shows that diffusion produced by the scale similarity mo-
del (15.18) is too low. The numerical calculations are often unstable using
this model. Taking the fact into account, that the diffusion of the Smagorin-
sky model is too high, it was decided to combine the Samgorinsky and scale
similarity models to get the proper diffusion. The advantages and disadvan-
tages of both models are summarized as follows
Dynamic Smagorinsky Model (DSM): energy dissipation is overesti-

mated (drawback), energy backscattering is not reproduced (draw-
back),
Scale similarity model: energy backscattering is reproduced (advan-

tage), energy dissipation is underestimated (drawback).
The idea combine models to strengthen the advantages and to overcome dis-
advantages of both models. The hybrid model called as the mixed similarity
model is written as
2
e
ijr D uQ i uQj uQQ i uQQj 2 CS jSj
Q SQij (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 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)
is substituted into the SGS stress expression:
e
156 mixed model:
Finally we have Leonard’s formulation of the
ijS GS D Lij C Cij C Rij (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 e e
ui uj D .uQ i C ui /.uQj C uj / D uQ i uQj C u0i uQj C uj0 uQ i C u0i uj0
0 0 Large Eddy Simulation (LES)
(15.20)
is substituted into the SGS stress expression:
e
Finally we have Leonard’s formulation of the mixed model:
ijS GS D Lij C Cij C Rij (15.22)

where
e
Lij D uQ i uQj uQ i uQj is the Leonard stress
e e
Cij D uQ i uj0 C uQj u0i is the Cross stress
e
Rij D u0i uj0 is the Reynolds stress
The sum of the cross and Reynolds stresses is calculated via the Smagorinsky
model with the dynamically determined constant CS
2
Cij C Rij D 158 Q SQij
2 CS jSj (15.23)
A substantial disadvantage of this formulation is the fact that the Leonard
stress does not satisfy the Galilean invariance condition. The Galilean in-
variance is the independence of basic formula of mechanics on the speed of
the reference system.
The classical definition of the SGS stresses possesses the Galilean invariance.
Indeed, let V be the speed of the reference system. The velocity relative to
the reference system is
WE D uE C VE (15.24)
The SGS stress does not depend on the reference system speed:
A
Wi Wj WQ i WQj D
G GG
D .uQ i C Vi C u0i /.uQj C Vj C uj0 / .uQ i C Vi C u0i /.uQj C Vj C uj0 / D
e CC e
D .uQ i C u0i /.uQj C uj0 / .uQ i C u0i /.uQj C uj0 / D ui uj uQ i uQj
(15.25)
On the contrary, the Leonard stress is not Galilean invariant:
A e
WQ i WQj WQ i WQj D uQ i uQj uQ i uQj Vi uQ0j Vj uQ0 i (15.26)
Germano proposed an alternative formulation
ijS GS D L0ij C Cij0 C Rij0 (15.27)
where all stresses are Galilean invariant:
e
L0ij D uQ i uQj uQQ i uQQj is the Leonard stress
0 e e
Cij D uQ i uj C ui uQj uQQ i uQ0j uQ0 i uQQj is the Cross stress
0 0
e
Rij0 D u0i uj0 uQ0 i uQ0j is157
the Reynolds stress
2
Q SQij
Cij0 C Rij0 D 2 CS jSj (15.28)
A e
WQ i WQj WQ i WQj D uQ i uQj uQ i uQj Vi uQ0j Vj uQ0 i (15.26)
Germano proposed an alternative formulation
ijS GS D L0ij C Cij0 C Rij0 (15.27)
where all stresses are Galilean invariant:
e
L0ij D uQ i uQj uQQ i uQQj is the Leonard stress
0 e e
Cij D uQ i uj C ui uQj uQQ i uQ0j uQ0 i uQQj is the Cross stress
0 0
e
Rij0 D u0i uj0 uQ0 i uQ0j is the Reynolds stress
2
Q SQij
Cij0 C Rij0 D 2 CS jSj (15.28)
Exercise: Prove the following facts:
equivalence of subgrid stresses159computed from formulations of Ger-

mano (15.27) and the original one proposed by Leonard (15.22),
Galilean invariance of stresses in Germano‘s formulation (15.27).
15.7 A-posteriori and a-priori tests

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 round-
ing 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
e
ijS GSdef D ui uj uQi uQj (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 ıij D 2.CS /2 jSQij jSQij (15.30)
3
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
is a big challenge to measure 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
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
is a big challenge to measure three components of velocity in a volume with
high spatial and temporal resolutions. The 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,
are filtered and utilized for the test. However, it should be noted that DNS
simulation is restricted by relatively low Reynolds numbers, whereas main
laws of LES are valid for high Re numbers.
160

how will goods be transported? What re-
sources will we use, and how many will
we need? The passenger and freight traf-
movement. We develop components and
that operate more cleanly and more ef-
ficiently than ever before. We are also
corporate, and public use. The challeng-
es are great. We deliver the solutions and
offer challenging jobs.

161
Chapter
Chapter 16 16
Lectures on computational fluid dynamics Hybrid URANS-LES methods
Chapter 16
Chapter
Hybrid 16
Hybrid URANS-LES
URANS-LES methods
methods
Hybrid URANS-LES methods
Hybrid URANS-LES
16.1 Introduction
methods
16.1 Introduction
As
16.1 discussed above, the most promising approach to resolve the flow un-
Introduction
As discussed above, the most promising approach to resolve the flow un-
steadiness 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.
As discussed above, theTypical Reynolds approach
most promising numbers in to engineering
resolve the flow are very
un-
large even at model scales. The grid resolution necessary for a pure LES
large even
steadiness at
is themodel
Large scales.
Eddy The grid
Simulation resolution
(LES),
As discussed above, the most promising approach to resolve the flow un- necessary
which is for
already a pure
widely LES
used
is so huge that it makes the direct application of LES impossible (see Sec.
is
for soresearch
huge is
steadiness that itLarge
purposes.
the makes the Simulation
Typical
Eddy direct application
Reynolds numbers
(LES), of LES
which impossible
inisengineering
already widely (see Sec.
are used
very
16.4). A practical solution of this problem is the use of hybrid URANS-LES
16.4).
large A practical
even
for research atpurposes.
model solution
scales. of The
Typical thisReynolds
problem is the use
grid resolution
numbers of engineering
hybrid
necessary
in forURANS-LES
a pure LES
are very
methods, where the near body flow region is treated using URANS and far
methods,
is
large evenwhere
so huge that itthe
at model near
makes body
the
scales. flow
direct
The region
application
grid is treated
resolution using
ofnecessary
LES URANS
impossible
for a pure andLES
(see far
Sec.
flow regions are treated with LES.
flow
is so regions
16.4). huge are it
A practical
that treated
makeswith
solution theof LES.
this problem
direct is theofuse
application LESof hybrid
impossibleURANS-LES
(see Sec.
methods, where the near body flow region is treated
16.4). A practical solution of this problem is the use of hybrid URANS-LES using URANS and far
According
flow regions to Peng [17] the hybrid techniques can be subdivided into flow
methods,
According toare
where Pengtreatednearwith
the [17] body
the LES.flow techniques
hybrid region is treated
can beusing URANSinto
subdivided andflowfar
matching and turbulence matching methods. Within the flow matching
flow regions
matching andareturbulence
treated with LES. methods. Within the flow matching
matching
methods the interface between URANS and LES is explicitly defined. LES
methods
Accordingthe to interface
Peng [17]between the hybrid URANS and LES
techniques canis be
explicitly
subdivided defined.
into LES
flow
filtered equations are solved in the LES region, whereas URANS equations
filtered
matching equations
and are
turbulence solved in the
matching LES region,
methods.
According to Peng [17] the hybrid techniques can be subdivided into flow whereas
Within URANS
the flow equations
matching
are solved in the URANS domain. The flow parameters (velocities, kinetic
are solvedthe
methods
matching in interface
and the URANS
turbulence domain.
between URANS
matching The flow LES
and
methods. parameters (velocities,
is explicitly
Within the defined.
flow kinetic
matchingLES
energy) are matched at the interface between the URANS and the LES re-
energy)
filtered are
thematched
methodsequations are at
interface the interface
solved
between in URANS
the LES between
region,
and thewhereas
LES isURANS anddefined.
URANS
explicitly theequations
LESLES re-
gions. Among the most important contributions to the development of flow
gions.
are Among
solved in the
the most
URANS important
domain. contributions
The flow to the
parameters
filtered equations are solved in the LES region, whereas URANS equations development
(velocities, of flow
kinetic
matching methods we mention the works of (Davidson, Dalstroem [18]; Ter-
matching
energy)
are solved methods
areinmatched
the URANSweatmention
thedomain.the works
interface The of (Davidson,
between
flow the URANS
parameters Dalstroem
and the[18];
(velocities, LES Ter-
re-
kinetic
racol [19]; Jakirlic et. al. [20]; Temmerman et. al. [21]) and others. A
racol
gions. [19];
Among Jakirlic
the et.
most al. [20];
important Temmerman
contributions
energy) are matched at the interface between the URANS and the LES re-et.to al.
the [21]) and
development others.
of flow A
serious weakness of this approach is the development of robust procedures to
serious
matching weakness
gions. Among methodstheofmost
thismention
we approach
important theisworks
the development
of (Davidson,
contributions to theof robust
Dalstroem procedures
development [18];
of Ter-
flowto
set the URANS-LES interface for complicated flow geometries. Within the
set
racol the[19];
matching URANS-LES
Jakirlic we
methods et.interface
al. [20];
mention for complicated
theTemmerman flowal.
et.
works of (Davidson, geometries.
[21]) and Within
Dalstroem others. the
[18]; Ter- A
framework of the turbulence matching method an universal transport equa-
framework
serious of
weakness the turbulence
of this matching
approach is the method
developmentan
racol [19]; Jakirlic et. al. [20]; Temmerman et. al. [21]) and others. A universal
of robust transport
procedures equa-to
tion is solved in the whole computational domain. The stress terms in this
tion
set
seriousis solved
the in the
URANS-LES
weakness whole
of this computational
interface
approach foriscomplicateddomain.
the development flow The stress procedures
geometries.
of robust terms
Within in this
the
to
equation are treated in different ways in LES and URANS domains. There
equation
framework are treated in
of the turbulence
set the URANS-LES different
interfacematchingways in LES and
method flow
for complicated URANS
an universal
geometries.domains.
transport
Within There
equa-
the
are various procedures to distinguish between LES and URANS cells. The
are
tion various
is solved
framework procedures
of in the
the wholeto distinguish
turbulence matchingbetween
computational domain.
method LES and stress
The
an universal URANS cells.inequa-
terms
transport The
this
most popular hybrid method is Detached Eddy Simulation (DES) proposed
most ispopular
equation
tion(Spalart hybrid
are treated
solved method
in is Detached
different ways in LES Eddyand Simulation
URANS (DES)
domains. proposed
There
by et in the whole
al.[22]). Thecomputational
original versiondomain. The stress
of this method terms
is based in
on this
the
are various
equation are procedures
treated in to distinguish
different ways between
in LES LES
and and
URANS URANS
domains.cells. The
There
classic Smagorinsky LES model and 163 the Spalart-Allmaras (SA) URANS ap-
most popular hybrid method is Detached
163 Eddy Simulation (DES)cells.
proposed
proach. SA is used close to the wall, between
are various procedures to distinguish whereas LES LES andin theURANS
rest part of The the
most popular
flow. hybrid method
The switching between isthe Detached Eddy Simulation
two techniques is smooth(DES) proposed
and occurs in
163
a ”gray” subdomain. There are two major improvements of DES, devel-
oped recently. The first one, DDES163 (Delayed DES), has been proposed to
detect the boundary layers and to prolong the RANS mode, even if the wall-
parallel grid spacing would normally activate the DES limiter (Spalart [23]).
The second one, IDDES (Improved DDES), allows one to solve the prob-
lems 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 blend-
ing functions to switch the solution between LES and URANS modes were
proposed by (Peng [17]; Davidson, Billson [25]; Abe, Miyata [26]) and others.
160
A very critical point of the turbulence matching methods is the transition
from the time (or ensemble)Download
averaged
free smooth
eBooks atURANS flow to the oscillat-
bookboon.com
ing 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-
see the principle
serious review (Spalart [23]).thoroughly
limitations In spite ofanalyzed
a wide application
by (Menter,area DES[24]).
Egorov has
serious versions
Other principleoflimitations thoroughly
the turbulence analyzed
matching by (Menter,
methods Egorovblend-
using different [24]).
Other
ing versionstoofswitch
functions the turbulence
the solutionmatching
betweenmethods
LES andusing
URANSdifferent
modesblend-
were
ing functions
proposed to switch
by (Peng [17]; the solution
Davidson, between
Billson [25];LES
Abe,and URANS
Miyata [26])modes were
and others.
Lectures
proposedon computational fluid
by (Peng [17]; dynamics Billson [25]; Abe, Miyata [26])Hybrid
Davidson, URANS-LES methods
and others.
A very critical point of the turbulence matching methods is the transition
A very
from thecritical
time (orpoint of the turbulence
ensemble) matching
averaged smooth methods
URANS flowistothe
thetransition
oscillat-
fromLES
ing the flow,
time see
(or (Menter,
ensemble)Egorov,
averaged smooth
2005). TheURANS flow have
oscillations to thetooscillat-
appear
ing LESa short
within flow, see
flow(Menter,
domain in Egorov,
a ”gray2005).
zone” The oscillations
between LES and have to appear
URANS. Ex-
within a short
perience showsflow thatdomain in a ”gray
it is extremely zone” between
difficult to provideLES and URANS.
a smooth Ex-
transition
perience
of shows that
the turbulent it isenergy
kinetic extremely difficult
passing from theto provide
URANSatosmooth transition
LES domain. To
of the turbulent
overcome kinetic (Schlueter
this problem energy passing
et al.from
[27])the
andURANS
(Benerafato LES
et al.domain. To
[28]) used
overcome
an this forcing
additional problemterm (Schlueter et al. [27])
in the Navier Stokesandequation
(Benerafa et al. [28])
artificially used
enhanc-
an additional
ing fluctuationsforcing
in theterm
grayinzone.
the Navier
However, Stokes
the equation
problem of artificially enhanc-
smooth solution
ing fluctuations
transition in the gray
from URANS zone.still
to LES However,
remainsthe asproblem
the mainofchallenge
smooth solution
for the
transition from
turbulence URANS
matching to LES still remains as the main challenge for the
methods.
turbulence matching methods.
16.2 Detached Eddy Simulation (DES)

16.2 Detached Eddy Simulation (DES)
The most popular hybrid method -detached eddy simulation- was proposed
The
in most
1997 by popular hybrid
Spalart et method
al. [22]. -detachedofeddy
The principle DES simulation-
is illustratedwas proposed
in Fig. 16.1.
in 1997tobythe
Close Spalart
body etthe
al.solution
[22]. The is principle of using
calculated DES istheillustrated
URANSinmode.
Fig. 16.1.
Far
Closethe
from to wall
the body the equations
the LES solution isare calculated using
solved. The greythezone
URANS mode.
between URANS Far
fromLES
and the wall
is thethe LES solution.
mixed equations are solved. The grey zone between URANS
and
The LES is the
classical mixedofsolution.
version the DES approach is based on the Spalart Almaras
The classical
(SA) version of with
model formulated the DES approach
respect is based on
to the modified the Spalart
turbulent Q D
Almaras
viscosity
(SA)
t =f1model
. The formulated for Q to
with respect
transport equation the modified turbulent viscosity Q D
reads
t =f1 . The transport equation for Q reads
164
164
Figure 16.1: Zones of the Detached Eddy Simulation.
2
@Q @Q Q
C vNj D Cb1 SQ Q Cw1 fw C
@t @xj „ ƒ‚ … d
Generation „ ƒ‚ …
Destruction
(16.1)
1 @ @Q Cb2 @Q @Q
C . C /
Q C
@x @xk @xk @xk
„ k ƒ‚ …
Diffusion
where
Cb1 D 0:1355; 161

Cb2 D 0:622; C1 D 7:1;
DownloadC
free 1 CatCbookboon.com
b1 eBooks b2
D 2=3; Cw1 D 2 C ;

C D 0:3; C D 2:0; D 0:41;
@t @xj d
„ ƒ‚ …Generation
Destruction
(16.1)
1 @ @Q Cb2 @Q @Q
C . C /
Q C
@xk @xk @xk @xk
„ dynamics
Lectures on computational fluid ƒ‚ … Hybrid URANS-LES methods
Diffusion
where
Cb1 D 0:1355;Cb2 D 0:622; C1 D 7:1;

Cb1 1 C Cb2
D 2=3; Cw1 D 2 C ;

Cw2 D 0:3; Cw3 D 2:0; D 0:41;
3
f1 D 3
; f2 D 1 ;
C C1
3 1 C f1
6 1=6
1 C Cw3
fw D g 6 6
; D =;
Q
g C Cw3
Q
g D r C Cw2 .r 6 r/; r D ;
Q 2d 2
S
Q p
SQ D S C 2 2 f2 ; S D 2˝ij ˝ij ;
d
1 @N i @Nj
˝ij D
2 @xj @xi
165
In the past 5 years we have drilled around
95,000 km
Who are we?

gas industry.

n Operations

careers.slb.com

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 equilib-
rium state:
2
Q
Cb1 SQ Q Cw1 fw 0
„ ƒ‚ … d
Generation „ ƒ‚ …
Destruction
The kinematic viscosity is then calculated from the formula
Cb1 Q 2
Q Sd
Cw1
which is similar to the Smagorinsky one:
q
t D lS2 jSQij j; jSQij j D 2SQij SQij
lS D CS
DES inventors proposed to use the following expression for d :
d D minfd; CDE S g; D maxfx; y; zg
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 re-
finement. 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 acti-
vated. 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:
Figure 16.2: Squires K.D., Detached-eddy simulation: current status and
current status and perspectives)
perspectives.
Figure
Figure16.3:
16.3:Flow around
Squires combat
K.D., aircraft. (Squires
Detached-eddy K.D., Detached-eddy
simulation: simulation:
current status and
current status and perspectives)
perspectives.
16.3 Hybrid model based on integral length

as parameter switching between LES and
URANS
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
@ui @.ui uj / @p @.ijl C ijt /

C D C ; (16.2)
@t @xj @xi @xj
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;

(16.3)
if L < then the cell is in the URANS region:
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
LDC ; (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=ı < 0:2, C 0:35 at 0:2 < y=ı <
0:7 and C 1:0 in the outer area of the bounder layer y=ı > 0:7, where ı 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
branches?regions. The
Kick-start yourextended LES
career with an filter is computed as
English-taught master’s degree. q
D dmax 2 C ı2; (16.5)
liu.se/master
where dmax is the maximal length of the cell edges dmax D max.dx ; dy ; dz /
1=3
and ı D .the cell volume/ is the common filter width used in LES. This
Click here!
choice ensures that very flat cells in the boundary layer (for which ı 0
but dmax > 0) are treated correctly. depends only on the mesh and it is
168

length scale L and the extended LES filter according to the following rule:
if L > then the cell is in the LES region;

Lectures on computational fluid dynamics Hybrid(16.3)
URANS-LES methods
if L < then the cell is in the URANS region:
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
LDC ; (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=ı < 0:2, C 0:35 at 0:2 < y=ı <
0:7 and C 1:0 in the outer area of the bounder layer y=ı > 0:7, where ı 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
D dmax 2 C ı2; (16.5)
where dmax is the maximal length of the cell edges dmax D max.dx ; dy ; dz /
and ı 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 ı 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) recog-
nizes 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 dis-
tribution 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:

2 1 @uj
166 @ui
SGS D cD ı jSij j; Sij D C ; (16.6)
2 at
Download free eBooks @xbookboon.com
i @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
Lectures as the subgrid
on computational viscosity and calculated according to the
fluid dynamics dynamic
Hybrid URANS-LES methods
model of Smagorinsky:

2 1 @uj @ui
SGS D cD ı 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 Estimations of the resolution necessary

for a pure LES on the example of ship
flow
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 separa-
tion 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. Sec-
ond, 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 devel-
opment 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 applica-
tion 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
turbulent kinetic energy should be directly resolved and the rest is modeled
167
in a properly resolved LES simulation. Implementation of this idea implies
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
opment 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 applica-
Lectures
tion ofonhybrid
computational
methods.fluid dynamics Hybrid URANS-LES
The impossibility of pure LES is illustrated below methods
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
turbulent kinetic energy should be directly resolved and the rest is modeled
in a properly resolved LES simulation. Implementation of this idea implies
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
wave number k separating the resolved and modeled turbulence is found
from the condition 170
R1
E.k/d k
Rk1 0:2: (16.8)
0 E.k/d k
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 simula-
tions 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 lat-
ter is dictated by the wish to resolve the most intensive vortex flow structures
which have the strongest influence on the propeller operation. Since LES ap-
plication 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

678'<)25<2850$67(5©6'(*5((
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 ıBL . The grid for a pure LES is generated using the
following algorithm. The minimum Kolmogorov length min is determined
in the LQJDQGQDWXUDOVFLHQFHVDUFKLWHFWXUHWHFKQRORJ\UHODWHGPDWKHPDWLFDOVFLHQFHV
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
9LVLWXVRQ&KDOPHUVVHRU1H[W6WRS&KDOPHUVRQIDFHERRN
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 ı , where ı is the Kolmogorov scale at y D ıBL . 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 simi-
lar estimations for the nonlinear k-" model these results show that the LES
171
wave number k separating the resolved and modeled turbulence is found
from the condition
R1
E.k/d k
Rk

Lectures on computational fluid dynamics 0:2: Hybrid (16.8)
URANS-LES methods
1
0 E.k/d k
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 simula-
tions 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 lat-
ter is dictated by the wish to resolve the most intensive vortex flow structures
which have the strongest influence on the propeller operation. Since LES ap-
plication 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 ıBL . 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 ı , where ı is the Kolmogorov scale at y D ıBL . 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 simi-
lar estimations for the nonlinear k-" model these results show that the LES
171
169
·min
··
min
·min
=1 min
+
y
··
=11 ·min
++
yy = min
··

z
x z y
z
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 com-
putations with
grid hundred millions andmillions
even with nodes.
a few billions of nodes are
grid should
should have
have the
the order
order ofof tens
tens of
of millions of
of nodes. Nowadays,
Nowadays, the the com-
com-
becoming available
putations in the researchand community. However, a numerical study
putations with
with hundred
hundred millions
millions and even
even with
with aa few
few billions
billions of
of nodes
nodes areare
of engineering
becoming problems
available in implies
the usually
research many computations
community. However, a which havestudy
numerical to be
becoming available in the research community. However, a numerical study
performed within
of a reasonable time with moderate computational have resources.
of engineering
engineering problems
problems implies
implies usually
usually many
many computations
computations which which have to to be
be
In this
performedsense, the
within results
a of
reasonable the present
time withsubsection
moderate clearly demonstrate
computational that
resources.
performed within a reasonable time with moderate computational resources.
the puresense,
In LES is impossible for ship applications so far. To verify that the
In this
this sense, thethe results
results ofof the
the present
present subsection
subsection clearly
clearly demonstrate
demonstrate that that
resolution
the 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
applied forestimation
resolution turbulent boundary
procedure layer
we (TBL)
used benchmark.
gives meaningful We foundit from
results, has me-
been
resolution estimation procedure we used gives meaningful results, it has been
thodical calculations
applied that the pure LES with 1M cells isWe quite accurate for
applied for
for turbulent
turbulent boundary
boundary layer
layer (TBL)
(TBL) benchmark.
benchmark. We found found from
from me-me-
prediction of the velocity
thodical distribution, TBL thickness, TBL displacement
thodical calculations
calculations thatthat the
the pure
pure LES
LES with
with 1M1M cells
cells is
is quite
quite accurate
accurate for for
thickness
prediction and
of the
the wall
velocityshear stress.
distribution, The
TBLestimation
thickness, procedure
TBL presented
displacement
prediction of the velocity distribution, TBL thickness, TBL displacement
above predictedthe
thickness the necessary resolution around 0.5M. Therefore,presented
the esti-
thickness and
and the wall wall shear
shear stress.
stress. The
The estimation
estimation procedure
procedure presented
mations presented
above for a ship model are rather lower bound for the resolution
above predicted
predicted the the necessary
necessary resolution
resolution around
around 0.5M.
0.5M. Therefore,
Therefore, the the esti-
esti-
required
mations for a pure
presented LES.
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 at bookboon.com
Mathematical model of the ice protection
of a human body at high temperatures
Chapter 17
Lectures on computational fluid dynamics of surrounding medium
Chapter 17
Chapter 17
Mathematical model of the ice
Mathematical
protection of amodel
humanof the ice
Mathematical model ofbody at
the ice
protection
high of a human
temperatures of body at
protection of a human body at
high temperatures
surrounding mediumof
high temperatures of
surrounding medium
surrounding medium
17.1 Designations
17.1 Designations
17.1.1 List of symbols
17.1 Designations
A area (m2 )
17.1.1 List of symbols
c specific
17.1.1 List
p heatofcapacity
symbols (J/(kgK))
2
F area
A (m of) heat release(W=m3 )
density
2
hp area
cA sector(mheight
specific ) capacity
heat (m) (J/(kgK))
cF specific heat capacity 3
(J/(kgK))
Kp density of
correlation heat release(W=m
factor of the rectal) temperature
3
F
h
k density
sector
the of heat release(W=m
heightconductivity
thermal (m) )
(W/(mK))
h
K
M sectordone
work height
correlation by (m)
factor of the
person (W)rectal temperature
k
m correlation
K body
the thermal factor of the rectal
conductivity
mass (kg) temperature
(W/(mK))
k
M the
work thermal
done
P body height (cm)byconductivity
person (W) (W/(mK))
m
Q work done
M internal
body mass by
(kg)
energy person
(J) (W)
QmP body
P heat flux mass
height (kg)
(W)(cm)
R sector radius(cm)
P
Q body
internalheight
energy (m)(J)
P internal
heat fluxenergy
Q temperature
T (W)(deg)(J)
Q P heat flux (W)
sector coordinate
R radial
x radius (m) along the sector (m)
R
T
˛ sector
centralradius
temperature (m)
sector(deg)
angle (deg)
T
x
temperature
radial
skin thickness (deg)
coordinate(m)along the sector (m)
x
˛
radial
thermal coordinate
central sector
diffusivity along
angle 2 the sector (m)
(deg)
(m =s)
˛

central
skin sector (kg=m
massthickness
density angle
(m) (deg)
3
)

skin
initialthickness
thermal (m) (m
diffusivity
temperature 2
(deg) =s)

thermal
mass density (kg=m(m)2 =s)
diffusivity 3
178
3

mass
initial density (kg=m
temperature
17.1.2 Subscripts )
(deg)

initial temperature (deg) 178
core human body core 178
hum human body
i number of human body layer
j number of point on the human body
med surrounding medium
melt melting ice
outer outer boundary of clothes
skin human body skin
17.1.3 Superscripts
0 degree of Celsius
172
17.2 Introduction
j number of point on the human body
med surrounding medium
melt melting ice Mathematical model of the ice protection
outer outer boundary of clothes of a human body at high temperatures
Lectures
skin onhuman
computational
bodyfluid
skindynamics of surrounding medium
17.1.3 Superscripts
0 degree of Celsius
17.2 Introduction
The heat transfer inside the human body is an extremely complicated prob-
lem 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. How-
ever, the thermal diffusion being complicated is not the main difficulty of
modeling. A large fraction of the heat generated by internal organs is trans-
ported 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 char-
acter and must relay on measurements data to diminish the modelling errors.
Especially interesting is the case of human thermodynamics at high tem-

peratures 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 don’t work [34], [35]. The most effective principle of the protec-
tion 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, B-

liver, C - kidney.
Figure 17.1: Sketch of the human body used in simulations. A- heart, B-
liver, C - kidney. 173
technologically easy and renewable
Downloadcan be used
free eBooks for protection clothes of res-
at bookboon.com
technologically easy and renewable can be
cue teams working under high temperatures up toused for 50
protection
0 clothes of res-
and more.
cue teams working under high temperatures up to 500 and more.
Figure 17.1: Sketch of the human body used in simulations. A-body
of a human heart, B- temperatures
at high
Lectures
liver, Con-computational
kidney. fluid dynamics of surrounding medium
technologically easy and renewable can be used for protection clothes of res-
cue teams working under high temperatures up to 500 and more.
17.3 Human body and ice protection models

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 ac-

cording 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 dis-
placed 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
Welcome to
our world
of teaching!
and open-minded professors
Study in Sweden -
close collaboration
Ericsson
TAKE THE
debajyoti nag
sive reputation in the field
RIGHT TRACK
of Embedded Systems Re-
answer your questions at
mdustudent.com
give your career a headstart at mälardalen university
www.mdh.se

Table 17.1: Sizes of the body used in simulations
Element Size in meter

1 0.127
2 0.127
3 0.234
4 0.090
5 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 mod-
elled 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

It is assumed, that the heat exchange between the sectors is negligible at
every cross section of the human body. Therefore both the vertical and cir-
cumferential heat transfer processes are not taken into account. The heat
transfer along the radius x within each sector is described by the one di-
mensional heat conduction equation derived from the energy conservation
consideration. The change of the internal energy in the sector element with
coordinates x and x C dx due to heat transport within the time dt is

@T @T @ @T
dQ1 D k.x C dx/S.x C dx/ k.x/S.x/ dt k.x/S.x/ dt
@x @x @x @x
(17.1)
where T .x; t/ is the temperature, k is the thermal conductivity, S.x/ D ˛xh
is the sector arc length multiplied with its height h and ˛ is the central angle
181
175
Figure 17.2: Horizontal cross section of the human body represented as an

ellipse with five layers: 1- inner core, 2- outer core, 3- muscles, 4- fat, 5- skin.
Figure 17.3: Ice protection construction. 1- polyurethane foam, 2- ice

briquette, 3- human body, 4-special overheating protection clothes, 5-
polyurethane net (air layer).
182
176
Table 17.2: Radii of the elliptical cross sections used in simulations
radial upper elliptical lower elliptical

line cylinder, m cylinder, m
OJ 0.118 0.109
OB, OB1 0.122 0.112
OC, OC1 0.137 0.121
OD, OD1 0.159 0.135
OE, OE1 0.181 0.156
OF, OF1 0.195 0.176
OG, OG1 0.200 0.190
OH, OH1 0.187 0.173
OK, OK1 0.163 0.145
OL, OL1 0.138 0.121
OM, OM1 0.120 0.106
ON, ON1 0.110 0.980
OP 0.106 0.940
Table 17.3: Radii of layers used in simulations in fraction of the skin thickness

Elliptical Element Area Layer Radius

cylinder length, m m2
Core 36:15
upper 0.476 0.489 Muscle 53:70
Fat layer 59:25
– YOUR CHOICE! Skin
Core
60:25
36:15
• 32 000 students0.317
lower • world 0.293
class research
Muscle • 53:70
top class teachers
• modern campus • ranked nr 1 inFat Sweden
layer by international students
59:25
• study in English Skin 60:25
– Bachelor’s programme in Life Science

of the sector. Within the material we have additionally the heat release with
– Master’s programme in Chemistry
the heat density F (heat per unit volume):
dQ2 D SF dxdt (17.2)

Download
brochure 183
here!

OP 0.106 0.940

Table 17.3:
Lectures Radii of layers
on computational used in simulations in fraction of the skin thickness
fluid dynamics of surrounding medium

Elliptical Element Area Layer Radius

cylinder length, m m2
Core 36:15
upper 0.476 0.489 Muscle 53:70
Fat layer 59:25
Skin 60:25
Core 36:15
lower 0.317 0.293 Muscle 53:70
Fat layer 59:25
Skin 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 the internal energy of the material

Due to the heat transfer and heat release
is changed by dQ3
@T
dQ3 D cp S dxdt (17.3)
@t
where cp is the specific heat capacity and is the mass density of the material.
By conservation of energy
dQ1 C dQ2 D dQ3
or

@T @ @T
S.x/ D S.x/ C S.x/f .x/ (17.4)
@t @x @x
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)

and the boundary conditions formulated at the sector center
T .0; t/ D Tcore (17.6)

and at the outer boundary of the clothes
T .xout er ; t/ D Tmed (17.7)

The temperature of the surrounding medium was Tmed D 500 . It was as-
sumed that the core center (at x D 0) temperature was constant in time
Tcore D 36:70 .
178
The heat transfer process isDownload
subdivided into at
free eBooks three steps. Within the first
bookboon.com
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 (17.6)
and at the outer boundary of the clothes
T .xout er ; t/ D Tmed (17.7)
The temperature of the surrounding medium was Tmed D 500 . It was as-
sumed 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 (17.8)
@x
and the human body Qhum caused by metabolism
@T
QP hum D ski n Aski n (17.9)
@x
where med is the thermal diffusivity184of the clothing layer adjacent to the
ice layer, Amed is the surface between the ice layer and adjacent clothing
layer in a sector under consideration, ski n is the thermal diffusivity of the
human skin and Aski n is the skin surface in the sector. The second step is the
ice layer melting. The phase transition in matter is described by the Stefan
problem [40]. In this paper we used a simplified approach based on the heat
balance conditions. It is assumed that the water and ice mixture is uniform
with the constant temperature of zero degree of Celsius. The dynamics of
the boundary between the water and ice is not considered. The heat fluxes
(17.8) and (17.9) are constant in time. The second step is finished in time t
once the ice is melted:
Qmelt D .QP med C QP ski n /t (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 Design of the protection clothes
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
179 with thickness of 0:001m, iso-
work we used three layers: outer clothing layer
lating layer of 0:009m and the air layer
Download free of 0:003m.
eBooks All layers have the initial
at bookboon.com
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
derivatives and Crank Nicolson implicit representation of of athe unsteady
human body atterm.
high temperatures
An inhouse code was developed for
Lectures on computational fluid dynamics this purposes. of surrounding medium
17.5 Results
17.5.1 Design of the protection clothes
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, iso-
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
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
Table 17.4: Thermodynamic coefficients of the body layers used in simula-

tions
layer heat thermal density initial heat

capacity, conductivity, temperature, production,
KJ/K W/(mK) kg m3 Deg. Celsius W
core 39.97 Scholarships
0.4186 1050 36.89 76.24-80.00
muscles 74.68 0.3959 1050 36.28 8.30-9.00
fat 17.73 0.3348 850 34.53 3.57-4.00
skin 5.07 0.0807 1000 33.62 0.66-0.7
Open your mind to

The numerical simulations are used to find the distribution of the ice thick-
ness necessary to keep the body core (both inner an outer) temperature at less
new opportunities
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
With 31,000 students, Linnaeus University is
for rescue team
one of the largerworking
universitiesunder emergency
in Sweden. We conditions in the mining industry.
Sincearethe
a modern university,distribution
continuous known for our strong
of the ice protection is difficult from the
international
technology pointprofile. Everyand
of view year more than
undesirable from ergonomics
Bachelorconsiderations
programmes in the
1,600 international students from all over the Business & Economics |the
Computer Science/IT |
jacket protection was designed using
world choose to enjoy the friendly atmosphere the discrete distribution of
Design | Mathematics
ice in
formandofactive
briquettes embedded
student life into the jacket (see Fig. 17.4). The further
at Linnaeus University.
Welcome to join us!
simulations have been performed to prove the ability of &the
Business new| Behavioural
Economics designed Sciences | Computer
jacket to satisfy the protection requirements. Fig. 17.5 Science/IT | Cultural Studies
illustrates the&tem-
Social Sciences | Design |
Mathematics | Natural Sciences | Technology & Engineering
perature 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 180 tested in further Click experimental
on the ad to read more
investigations. Download free eBooks at bookboon.com
17.5.2 Experimental proof of numerical prediction

(see Eq. (17.4)). of a human body at high temperatures
Table 17.4: Thermodynamic coefficients of the body layers used in simula-

tions
layer heat thermal density initial heat

capacity, conductivity, temperature, production,
KJ/K W/(mK) kg m3 Deg. Celsius W
core 39.97 0.4186 1050 36.89 76.24-80.00
muscles 74.68 0.3959 1050 36.28 8.30-9.00
fat 17.73 0.3348 850 34.53 3.57-4.00
skin 5.07 0.0807 1000 33.62 0.66-0.7
The numerical simulations are used to find the distribution of the ice thick-
ness 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 tem-
perature 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 Experimental proof of numerical prediction

The numerical prediction and the jacket designed on basis of numerical sim-
ulations 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 17.4: Overheating protection jacket designed on the base of simula-

tions.
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 be-
tween the measurement (solid line) and the numerical simulations (dotted
Figure 17.6: Development of the averaged temperature in the air gap between
line).
tween 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
watts which corresponds to a hard work0:425
typical for the mining industry. The
from the empirical formula A D 0:24m P 0:4 [41]. 0The candidates weared
temperature of the surrounding medium was T 50 . The temperature has
been measured in the air gap between the underwear and the ice at seven
watts which corresponds to a hard work typical for the mining industry. The
points around the chest (section 2 in Fig. 17.1). The temperature averaged
over these points is presented in Fig. 17.6. The discrepancy between the
measurement and the numerical simulations does not exceed 14 percent af-
ter 60 minutes of the real time. This agreement can be considered as quite
satisfactory taking the simplicity of the used model and complexity of the
problem into account.
188
188

tween the measurement (solid line) and the numerical simulations
Mathematical (dotted
model of the ice protection
line). of a human body at high temperatures

from the empirical formula A D 0:24m0:425 P 0:4 [41]. The candidates weared
watts which corresponds to a hard work typical for the mining industry. The
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 tempera-
ture 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)

where Tj sensors readings according to the numeration shown in Fig. 17.7.
The prediction accuracy can be sufficiently184 increased when the core temper-
ature is calculated with account for the test person feeling and the rectal
temperature Trect al according to the formula [41]
ture 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)
where Tj sensors readings according to the numeration shown in Fig. 17.7.
The prediction accuracy can be sufficiently increased when the core temper-
ature is calculated with account for the test person feeling and the rectal
temperature Trect al according to the formula [41]
Ti nner D K Trect al C .1 K/T (17.12)

the time history of the inner temperature Ti nner . The most important con-
The correlation
the time historyfactor is taken
of theKinner from the table
temperature Ti nner17.5 [39].
. The Figure
most 17.8 shows
important con-
clusion drawn from this figure is that the inner temperature Ti nner doesn’t
the timedrawn
clusion history of the
from this0inner
figuretemperature
is that the T
inner . The most important
i nner temperature con-
Ti nner doesn’t
exceed the threshold 360 within 60 minutes. Thus, the aim of the design has
clusion drawn
exceed the from this
threshold 36 figure
withinis60that
189the inner
minutes. Thus,temperature
the aim of Tthe doesn’t
design
i nner has
been achieved. 0
exceed the threshold
been achieved. 36 within 60 minutes. Thus, the aim of the design has
been achieved.
Table 17.5: Coefficient K depending on the test person feelings and energy
Table 17.5: Coefficient K depending on the test person feelings and energy
expenditure E D M=A .W =m22 /. A is the body surface (m22 ) and M is the
Table 17.5: Coefficient
expenditure E D M=A K .Wdepending
=m /. A isonthethebody
test surface
person feelings
(m ) and and
M energy
is the
work (W ) 2 2
expenditure
work (W ) E D M=A .W =m /. A is the body surface (m ) and M is the
work (W )
feeling j E 70 88 113 145 178
feeling j E 70 88 113 145 178
coldj E
feeling 0.55
70 0.5788 0.59113 0.62
145 0.64
178
cold 0.55 0.57 0.59 0.62 0.64
chilly 0.57 0.58 0.60 0.64 0.65
cold
chilly 0.55
0.57 0.57
0.58 0.59 0.62 0.65
0.60 0.64 0.64
slightly chilly 0.59 0.60 0.63 0.65 0.67
chilly
slightly chilly 0.57
0.59 0.58
0.60 0.600.63 0.64
0.65 0.65
0.67
comfort 0.62 0.64 0.65 0.67 0.69
slightly chilly 0.59
comfort 0.62 0.60
0.64 0.630.65 0.65
0.67 0.67
0.69
slightly warm 0.70 0.70 0.70 0.70 0.70
comfort
slightly warm 0.62
0.70 0.64
0.70 0.650.70 0.67
0.70 0.69
0.70
warm 0.79 0.77 0.74 0.72 0.72
slightly
warm warm 0.70 0.70 0.70 0.70
0.79 0.77 0.74 0.72 0.72 0.70
hot 0.86 0.82 0.79 0.76 0.73
warm
hot 0.79
0.86 0.77
0.82 0.740.79 0.72
0.76 0.72
0.73
hot 0.86 0.82 0.79 0.76 0.73
The second series of measurements was carried out directly during the work
The second series of measurements was carried out directly during the work
in a cole mine at the temperature not higher than 6000 without fire action.
The
in a second series
cole mine at of
themeasurements was higher
temperature not carriedthan
out directly during
60 without firetheaction.
work
During one year of observations no equipment fault has 0 been documented.
in a cole mine at the temperature not higher than 60 without
During one year of observations no equipment fault has been documented.fire action.
The temperature of workers was kept at a prescribed level during at least
During
The one year ofof observations
temperature workers was no equipment
kept fault has
at a prescribed been
level documented.
during at least
55 60 minutes as predicted both in numerical simulations and thermal
The temperature
55 60 minutes asof workers
predictedwas keptin atnumerical
both a prescribed level during
simulations at least
and thermal
chamber tests.
55 60 minutes
chamber tests. as predicted both in numerical simulations and thermal
chamber tests.
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 trans-
fer within the human body at high temperatures of the surrounding medium.
This chapter
fer within thepresents relatively
human body simple
at high and efficient
temperatures model
of the of the heat
surrounding trans-
medium.
The human body cross section is represented as an ellipse with a few lay-
fer within
The human thebody
human body
cross at high
section is temperatures
represented asofan theellipse
surrounding
with a medium.
few lay-
ers, modelling the internal organs. The heat generation by internal organs
The
ers, modelling the internal organs. The heat generation by internalfew
human body cross section is represented as an ellipse with a lay-
organs
is taken into account. The ice protection and the clothes are modelled as
ers,taken
is modelling the internal
into account. Theorgans. The heat
ice protection andgeneration
the clothes byare
internal organs
modelled as
additional layers covering the human body. Neglecting the heat transfer in
is taken into
additional account.
layers coveringThetheicehuman
protection
body.and the clothes
Neglecting the are
heatmodelled
transfer asin
vertical and circumferential directions the problem is reduced to the solution
additional layers covering the human body. Neglecting the heat
vertical and circumferential directions the problem is reduced to the solution transfer in
of the one dimensional heat conduction equation with variable heat diffu-
vertical
of anddimensional
the one circumferential
heatdirections
conductionthe equation
problem is reduced
with to the
variable heatsolution
diffu-
sivity. The simulations were performed using the finite differential method.
of the one
sivity. The dimensional heat performed
simulations were conductionusing
equation with differential
the finite variable heat diffu-
method.
Results of simulation were used for calculation of the ice layer thickness nec-
sivity. The simulations were performed using the finite differential
Results of simulation were used for calculation of the ice layer thickness nec- method.
essary to prevent the body overheating and to keep the temperature of the
Resultstoofprevent
essary simulation
the were
bodyused for calculation
overheating and toofkeep
the the
ice layer thicknessofnec-
temperature the
essary to prevent the body overheating and to keep the temperature of the
190 185
190
190eBooks at bookboon.com
Download free
Figure 17.8: Inner human body temperature versus time.
body core at 36:70 ˙10 within one hour. The results of simulations were used
for designFigure 17.8:overheating
of special Inner human body temperature
protection versus
jacket. The time.
labour tests of the
jacket and trials at real conditions confirmed the prediction of the numerical
simulations.
body core at 36:70 ˙10 within one hour. The results of simulations were used
for design of special overheating protection jacket. The labour tests of the
jacket and trials at real conditions confirmed the prediction of the numerical
simulations.
YOUR WORK AT TOMTOM WILL

AROUND THE WORLD. EVERYDAY.
follow us on
191
191
#ACHIEVEMORE

CFD Design of cloth for protection of divers at
Chapter 18
Lectures on computational fluid dynamics low temperatures under current conditions
Chapter 18
Chapter 18
CFD Design of cloth for
CFD Design
protection of of clothatfor
divers low
CFD Design of cloth for
protection of divers
temperatures under at low
current
protection of divers at low
temperatures
conditions under current
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 hu-
Study
man body was was
performed
assumed fortothe
be model
constantTinand Man presented
equal to 330 at in the
Fig.temperature
18.1. Inner
Study was
thermodynamics performed
in thefor the
body of model
was10not Tin Man
modelled. presented in
The temperatureFig. 18.1.theInner
of the surrounding medium degrees of Celsius. 0
Thus, theofDirichlet hu-
thermodynamics
man body was in the body
assumed to be was not modelled.
constant and equal The
to temperature
33 at the of the hu-
temperature
boundary conditions were enforced for the temperature whereas the heat
man body
of was assumed to beofconstant and of equal to 330 Thus,
at the thetemperature
fluxthe
fromsurrounding
the body medium
was calculated.10 degrees
Heat exchange Celsius.
by shivering, Dirichlet
breathing
of the
boundary surrounding
conditions medium of
were enforced10 degrees of
forexchange Celsius.
the temperature Thus, the
whereas Dirichlet
and radiation was neglected. The heat between the body the
and heat
sur-
boundary
flux from conditions
the body were
was enforced for
calculated. Heat the temperature
exchange by whereas breathing
shivering, the heat
rounding medium is mostly determined by the convection due to wind. The
flux radiation
and from the bodyneglected.
was calculated. Heat exchange by shivering,
the bodybreathing
incompressiblewas flow was calculatedThe heat
using exchange between
steady RANS (Reynolds and sur-
Averaged
and radiation
rounding mediumwas neglected.
is mostly usingThe heat
determined exchange
by!the between
convection the body and The
sur-
Navier- Stokes Equations) the k SST model. due to wind.
Temperature was
rounding medium
incompressible is was
mostly determined
usingbysteady
the convection due to wind. The
considered as aflow passive calculated
scalar determined from RANS (Reynolds
the temperature Averaged
transport
incompressible
Navier- flow was calculated using
k steady
SSTRANS (Reynolds Averaged
equation.Stokes Equations)
The framework using the
OpenFOAM was! utilizedmodel.
for the Temperature
numerical solu- was
Navier- Stokes
considered as a Equations)
passive using
scalar the k !from
determined SSTthemodel. Temperature
temperature was
transport
tion of this problem.
considered The
equation. as a passive scalar determined wasfrom the for
temperature transport
First, CFD was framework
validated for OpenFOAM
the case of the utilized
air the numerical
flow. Distribution of the solu-
heat
equation.
tion of thisThe framework
problem. OpenFOAM was utilized for the numerical solu-
transfer coefficient along the Tin Man body is shown in Fig. 18.1, left. The
tion ofCFD
First, this was
problem.
validated for the caseseparation
of the air flow. Distribution
minimal heat flux takes place in the area with reduced of flowtheveloc-
heat
First,
transferCFD was validated for the case of the air flow. Distribution 18.1, left. heat
of the
ities, i.e.coefficient
behind the along
bodythe Tinand
arms Manhead,
bodyasiswellshown in the
as in Fig.stagnation The
area
transfer coefficient
minimal heat flux alongplace
takes the Tin
in Man
the body is area
separation shown in Fig.
with 18.1,flow
reduced left.veloc-
The
in the front part of the body. Figure 18.1 illustrates the integral coefficient
minimal
ities, heat fluxthe
takes place in the separation areaaswith reduced flow veloc-
of thei.e. behind
heat transfer body arms
hc obtained and
by head, as well
integration overinthethe stagnation
whole body area
sur-
ities,
in thei.e. behind
front part the
of body
the arms
body. and
Figure head,
18.1 as well as
illustrates in
the the stagnation
integral area
coefficient
face. Results of authors marked by crosses are compared with experimental
in the front
of part of the body. Figure 18.1 illustratesoverthe
theintegral coefficient
data ofheat transfer
de Dear obtained
et al.hc(triangles) by integration
and different calculations. whole body
Shadowed sur-
grey
of theResults
face. heat transfer obtainedbybycrosses
hc marked
of scattering
authors integration over thewith whole body sur-
area shows the of data otbainedare compared
using experimental
various approaches. Big
face. Results of authors marked by crosses are compared
data of de Dear et al. (triangles) and different calculations. Shadowed grey with experimental
data shows
area of de Dear et al. (triangles)
the scattering of dataand different
otbained calculations.
using Shadowed grey
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
187 show results of the present
works. The figure is taken from [2]. Blue crosses
work.
scattering of data illustrates the complexity of the problem of the human

CFD Design of cloth for protection of divers at
Figure 18.1: Left: Heat transfer coefficient at air speed of 1m=s. Right:
Whole body convective heat transfer coefficient hcc from various published
works. The figure is taken from [2]. Blue crosses show results of the present
work.
scattering of data illustrates the complexity of the problem of the human

body thermodynamics interaction with a surrounding medium. Our results
agree satisfactory with measurements. Note that the calculations at small
air speeds were performed assuming the laminar character of the flow.
Within the next step the Tin Man body was weared in a neoprene cloth
with the thickness of 1:0 cm to reproduce the case of a diver working at
0
low temperatures of 100 in a sea current with the speed of 1 m=s. The Tin
Man body temperature was uniformly distributed along the body and was 33
degrees of Celsius. The cloth was assumed to be waterproof and uniformly
contaminated by oil products. This case is typical for ship repair works and
rescue and technical operations in oil spill zones. Contamination in terms
of gram per centimeter squared as well as the heat diffusivity coefficient of
contaminated cloth are given in the table 18.1. The size of the computational
domain was 12 12 6 meters, whereas the size of the body was 1:65 meters.
C
For discretization a mesh with 1:3 million of cells was used with y C values
below 0:5 on the whole body. The results for heat flux depending on the
contamination are presented in the last column of table 18.1 and in Fig. 18.2.
As seen the heat flux is almost linear function of the contamination.
The Tin Man body is a primitive simplest geometry. From one side, it makes
the solution of the inner thermodynamics problem much easier. From the
other side, such simplified geometry is not suitable for aerodynamic calcula-
tions since it consists of sharp edges with numerous strong separations. Nu-
merical solution becomes unstable leading to unsatisfactory convergence. Al-
though the real human body is geometrically more complex, the flow around
it can be calculated much easier than 194in the Tin Man case. Therefore, fur-
ther investigations done by our group were performed for real human body
geometry (Fig. 19.1) designed by the Hohenstein Institute [43] based on the
detailed antropological study of different people categories.
Contamination Heat diffusivity Heat flux

g=cm2 m2 =s W=m2 s
0:0 2:07 106 1:89
6
0:5 2:72 10 2:41
2:5 5:27 106 4:46
6
4:5 7:74 10 6:33
5
6:5 1:01 10 8:08
Table 18.1: Heat flux from the diver depending on the
cloth contamination
188
2:5 5:27 106 4:46
4:5 7:74 106 6:33
6:5 1:01 105 8:08Design of cloth for protection of divers at
CFD
Table 18.1: Heat flux from the diver depending on the
cloth contamination
Figure 18.2: Heat flux from the

diver depending on the cloth con-
tamination
195

CFD application for design of cloth for
protection from low temperatures under wind
Chapter 19
Lectures on computational fluid dynamics conditions. Infl uence of the wind on the cloth
deformation and heat transfer from the body.
Chapter 19
Chapter 19
CFD application for design of
CFD
cloth application
for for from
protection design
lowof
CFD application for design of
cloth for protection
temperatures under from
wind low
cloth for protection from low
temperatures
conditions. under wind
Influence of the
temperatures under wind
conditions.
wind on theInfluence of the
cloth deformation
conditions. Influence of the
wind
and on the
heat cloth from
transfer deformation
the
wind on the cloth deformation
and
body.heat transfer from the
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
study was and Brink [44]
performed havetype
in open shown
windthetunnel
influence
withof windtest
closed on heat transfer
section up to
Lebedeva
due to anddeformation
cloth Brink [44] have
caused
5 shown
by the
wind influence
induced of wind on
pressures. heat transfer
Experimental
Reynolds number of 3:6 10 based on the air speed and diameter of cloth
due towas
study cloth deformation caused by wind induced
withpressures. Experimental
packages inperformed in open
form of cylinders. type
Thewind
aim tunnel
of the present closed
worktest section
is to up to
get similar
5 type wind tunnel with closed test section up to
study was performed
Reynolds number
estimations 3:6 10body
in open
for realofhuman based on the air speed and diameter of cloth
form.
Reynolds in
packages number 3:6 105 based
form ofofcylinders. The aimon the
of theairpresent
speed and
workdiameter
is to getofsimilar
cloth
packages in for
estimations form of human
real cylinders. The
body aim of the present work is to get similar
form.
estimations for real human body form.
19.1 Wind tunnel measurements of pressure
19.1 distribution
Wind tunnel measurements of pressure
19.1 Wind tunnel measurements of pressure
distribution
The experiment was conducted in the wind tunnel of the Chair of Ocean
distribution
Engineering at the Rostock University. The wind tunnel of the Göttingen
The
type experiment
has the testwas conducted
section in the
of lengths wind
of 2:8 m tunnel of square
with the the Chair of section
cross Ocean
The experiment
Engineering at was
the conducted
Rostock in the
University. wind
The tunnel
wind of the
tunnel Chair
of the of Ocean
Göttingen
of 1:4 1:4 m2 . The model of the height of 700 mm was manufactured by
Engineering
type has theat
testthesection
Rostock
of University. The
lengths of 2:8 m wind
with tunnel of the
the square Göttingen
cross section
2
type
of 1:4
1:4has themtest section
. The of lengths
model of 2:8ofm700
of the height withmm
thewas
square cross section
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).
downsizing the real body with the scale factor190 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. SinceDownload free eBooks at bookboon.com
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
Figure 19.1: Human body model in wind tunnelprotection of the Rostock university under wind
from low temperatures
Lectures
(left). on computational
Positions fluid dynamicspoints (right). conditions. Infl uence of the wind on the cloth
of measurement
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 of pressure dis-

tribution and comparison with measure-
ments
Numerical simulations were performed using the commercial software pack-
age 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 sep-
aration 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 sec-
tion of 2:1 2:1 m2 . A length of the domain is 2:1 m in front of the human
body (0:7 m height) and 4:2 m behind 198 it. The grid consists of 1:2 106
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 of CFD results with mea-

surements
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 measure-
ment 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
191
degrees in the area of the separation (point 5 in Fig. 19.1). The experiment
Download
doesn’t predict the strong under free eBooks
pressure at bookboon.com
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
flow character, simulations were performed both for pureCFD application
turbulent andforpure
design of cloth for
laminar flows with and without roughness. protection from low temperatures under wind
Lectures on computational fluid dynamics conditions. Infl uence of the wind on the cloth
19.3 Comparison of CFD results with mea-

surements
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 measure-
ment 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
doesn’t 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 differ-
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 is fixed at the widest
axis and doesn’t 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
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

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
application widest
design of cloth for
axis and doesn’t depend on the flow character.protection
Second,from low temperatures
strong unsteady under wind
flow oscillation makesfluid
thedynamics conditions.
difference between laminar Infl uence of
and turbulent the wind on the cloth
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
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
Figure 19.2: Contrours of torso (left) and pressure coefficient Cp distribution

around the body at three different altitudes z D 0:329; 0:418 and 0:476m.
between Cp determined for 10 and 15 m/s seems to be very questionable, at

least within the zone of weak Reynolds number influence, i.e. at angles less
than 90 degrees. The flow at angles larger than 90 degrees is unsteady. The
scattering of the data due to unsteady effects is shown both for simulations
and measurement. Obviously, the unsteady pressure oscillations can not
be reason for the disagreement between simulation and experiments. More
detailed experimental study should be carried out in the future to clear this
problem.
19.4 Change of thermal conductivity caused

by wind induced pressures
Estimations of pressure influence on cloth thermal conductivity was per-
formed 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.
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
193
but not stretched. With the other words, the cloth has no deformation if
the pressure difference p DDownload
p patfree eBooks at bookboon.com
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 per-
formed 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
protection from lowand Taffeta under wind
temperatures
as a lining. The cloth fluid dynamics
is used conditions.
in oil industry for work at very Infl
lowuence of the wind on the cloth
temperatures
under oil contamination conditions. deformation and heat transfer from the body.
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 us-

ing 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 k is the thermal conductivity coefficient in W=mK, C1 D 9:322
104 mKP
W
a
and C2 D 9:775 106 mKPW
a2
. Fig. 19.4 shows distribution of the
thermal conductivity around the body at three sections along the human
body. As seen the conductivity can201 increase in the stagnation area at the
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
wind was integrated over the body at five sections:
I
f .z/ D k.wi nd ¤ 0/=k.wi nd D 0/r d# (19.2)
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
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
wind was integrated over the body at five sections: CFD application for design of cloth for
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
Figure 19.4: Change of thermal conductivity due to pressure induced by wind

of 10 m/s (left) and 20 m/s (right). Thermal conductivity without wind is
0:22 W/mK.
203
195
Simulation of human comfort conditions in car cabins
Chapter 20
Chapter
Chapter 20
20
Simulation of human comfort
Simulation
conditions of
Simulation in human
of car cabins
human comfort
comfort
conditions
conditions in
in car
car cabins
cabins
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
The conducted
comfort in this
conditions field at
of drivers the chair of modelling and simulation of
research topics in car industry. Inandthispassengers
chapter we is one of the
present most
a fragmentimportant
of the
the University of
research Rostock. The calculations presented
presentbelow were carried
research topics
conducted in carin industry.
this field In at this
the chapter
chair of we modelling a fragment
and simulationof the
of
out by
research Stefan
conductedKnochenhauer.
in this field A compact car was chosen as the object of
the University of Rostock. Theatcalculations
the chair ofpresented
modellingbelow and simulation
were carried of
investigations.
the The cabin geometry is represented in Fig. 20.1. The inlet
out University of Rostock.
by Stefan Knochenhauer. TheA calculations
compact carpresentedwas chosen below
as thewere carried
object of
of
outthe
by air stream
Stefan coming from A
Knochenhauer. thecompact
ventilationcar channel
was is marked
chosen as the by green
object of
investigations. The cabin geometry is represented in Fig. 20.1. The inlet
color whereas
investigations. the
Theair outlet
cabin from by
geometry red one. The
is representedgrid with
in Fig.6.5 million
20.1. The of cells
inlet
of the air stream coming the ventilation channel is marked by green
is
of shown
the air instream
Fig. 20.2.
coming The
from mathematical
the ventilationmodel is based
channel is on the by
marked URANS
green
color whereas the air outlet by red one. The grid with 6.5 million of cells
equation
color (12.17)
whereas theand
air temperature
outlet redtransport equation (1.30). The turbulent
is shown in Fig. 20.2. The by mathematicalone. The gridiswith
model based6.5onmillion of cells
the URANS
character
is shown of the
in Fig.temperature
20.2. The transport
mathematical is takenmodelintoisaccount
based by introduction
on the URANS
equation (12.17) and temperature transport equation (1.30). The turbulent
of the additionaland
equation turbulent heat conduction. It means thatThe instead of
character (12.17) temperature
of the temperature transporttransport
is taken equation
into account(1.30). turbulent
by introduction
in the
character equation (1.30)
of the temperature we use the
transport sum C
is taken tinto, where
account is the turbulent
t by introduction
of the additional turbulent heat conduction. It means that instead of
heat
of conduction coefficient. t is expressed through the turbulent kinematic
in the equation (1.30) we use the sum C t , where tthat
the additional turbulent heat conduction. It means is theinstead of
turbulent
viscosity t
in theconduction and the turbulent
equation coefficient.
(1.30) we use Prandtl
the number
sum C P r t
t , where t is the kinematic
turbulent
heat t is expressed through the turbulent
heat conduction coefficient.
viscosity t and the turbulentt Prandtl is expressed
t number
t through
P rt the turbulent kinematic
viscosity t and the turbulent Prandtl D number P r t (20.1)
c
tp Pr t
t
The turbulent kinematic viscosityc D t
ist computed from the URANS closure(20.1) mod-
p D P r t (20.1)
els (see Chapters 12, 13 and 14)cwhereas P r the turbulent Prandtl number is
The turbulent kinematic viscosity ispcomputed t
from the URANS closure mod-
assumed
The to bekinematic
turbulent constant P r t 0:7. The URANS closure modelclosure
used inmod-this
els (see Chapters 12, 13 viscosity
and 14) iswhereas
computed thefrom the URANS
turbulent Prandtl number is
work is the k ! SST model [45]. Numerical simulations were performed
els (see Chapters
assumed 12, 13Pand
to be constant
using thetofinite volume method
r t 14) 0:7.whereas
The URANS
(seeTheChapter
theWE WILL
turbulent
closure
6). closure
Boundary TURN
Prandtl
model used
conditions
inYOUR
number is
this
are CV

assumed
given is
work
be
in the ktable
using thethefinite
constant
work is the k ! SST model
! 20.1.
volume
P
SST model
r t
method[45].
0:7.
[45]. NumericalURANS
(see Numerical
Chapter 6).
INTOsimulations
simulations
AN OPPORTUNITY
model
Boundarywere
were used in
performed
performed
conditions
this
are
A few
using in
given
selected
thethefinite results
tablevolume of calculations are illustrated
20.1. method (see Chapter 6). Boundary conditions
in Fig. OF
20.3. TheA distri-
LIFETIME
are
butions
given in of
the airtable
velocity
20.1. and temperature inside of the car cabin are strongly
A few selected results of calculations are illustrated in Fig. 20.3. The distri-
A few selected
butions results of
of air velocity calculations
and temperature are inside
illustrated
of theincar Fig.cabin
20.3.areThe distri-
strongly
butions of air velocity and temperature 205 inside of the car cabin are strongly
205
205
Do you like cars? Would you like to be a part of a successful brand?

Send us your CV on
As a constructer at ŠKODA AUTO you will put great things in motion. Things that will
www.employerforlife.com
ease everyday lives of people all around Send us your CV. We will give it an entirely
new new dimension.

character of the temperature transport is taken into account by introduction
of the additional turbulent heat conduction. It means that instead of
in the equation (1.30) we use the sum C t , where t is the turbulent
heat conduction
Lectures coefficient.
on computational t is expressed Simulation
fluid dynamics through the turbulent
of human kinematic
comfort conditions in car cabins
viscosity t and the turbulent Prandtl number P r t
t t
D (20.1)
cp P rt
The turbulent kinematic viscosity is computed from the URANS closure mod-
els (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 distri-
butions 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 20.1: Sketch of the car cabin studied numerically.
Figure 20.2: Grids with 6.5 million of cells generated with snappyHexMesh.
197
Lectures on computational fluid dynamics Simulation of human comfort conditions in car cabins
Surface physical quantity Condition

Velocity ux uy uz
Pressure grad p D 0
Inlet Temperature T D konst:
turbulent kinetic energy k D konst:
specific turbulent dissipation rate ! D konst:
Velocity grad U D 0
Pressure pD0
Outlet Temperature grad T D 0
turbulent kinetic energy grad k D 0
specific turbulent dissipation rate grad ! D 0
Velocity U D0
Pressure grad p D 0
T D konst: or
Other surfaces Temperature grad T D 0 or
coupled boundary condition
turbulent kinetic energy modelled by wall function
specific turbulent dissipation rate modelled by wall function
Table 20.1: Boundary conditions
(a) Velocity field (b) Temperature field
Figure 20.3: Results of numerical simulations.
207
198
Lectures on computational fluid dynamics 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 trans-
fer simulation with cfd- the advances and challenges.building simulation
[2] Novieto D. Zhang
2009. pages Y. andEleventh
2162–2168. Yingchun J. Human envoronmental
International heat trans-
IBPSA Conference, Glas-
fer
gow,simulation
Scotland,with
Julycfd- the 2009
27–30, advances
2009.and challenges.building simulation
2009. pages 2162–2168. Eleventh International IBPSA Conference, Glas-
[3] gow, Scotland,
Ferziger J. and July
Peric27–30, 2009 2009.
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 incompress-
ible navier-stokes equations on rectangular domains. Applied Mathemat-
[4] ics,
Seibold B. A compact
Massachusetts and offast
Institute matlab code
Technology, solving the incompress-
www-math.mit.edu/ seibold,
ible
2008.navier-stokes equations on rectangular domains. Applied Mathemat-
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.
Mathematics, and Neunhoeffer
Philadelphia, PA, USA,T. 1998.
Numerical simulation in
fluid dynamics: A practical introduction. Society for Industrial and
Applied Mathematics,
[6] Phillips A.B. Turnock Philadelphia,
S.R. and FurlongPA, M.
USA, 1998.simulations of static
Urans
drift and dynamic manouveres of the KVLCC2 tanker. pages F63–F68.
[6] Proceedings
Phillips A.B.ofTurnock S.R. and
the SIMMAN Furlong 2008.
workshop, M. Urans simulations of static
[7] Liseikin V.Develop
Proceedings of the
Grid SIMMAN the
Generation toolsSpringer,
workshop,
Methods. we need
2008. for Life Science
2010. 390p.
[7] Liseikin
[8] V. Grid
Warsi Z.U.A Generation
Thompson I.F. Methods.
and MastinSpringer, 2010. 390p.grid genera-
C.W. Numerical
tion. Foundations and applications, 1997.
[8] Warsi Z.U.A Thompson I.F. and Mastin C.W. Numerical grid genera-
tion. Foundations
[9] Novikov and applications,
E. Generalized dynamics1997.
of three-dimensional singularities
(vortons). Sov. JETP, 84(3):975 – 981, 1984.
[9] Novikov E. Generalized dynamics of three-dimensional singularities
(vortons).
[10] Chorin Sov. evolution
A. The JETP, 84(3):975 – 981, vortex.
of a turbulent 1984. Comm. in Mathematical
Physics, 83:517 – 535, 1982. exciting field where biology,
[10] Chorin A. The evolution of a turbulent vortex. Comm. in Mathematical
Physics, 83:517 – 535, 1982. mathematics meet.
209
209 biology and medicine using
mathematics.

ics, Massachusetts Institute of Technology, www-math.mit.edu/ seibold,
2008.
[5] Dornseifer T. Griebel M. and Neunhoeffer T. Numerical simulation in

fluid dynamics: fluid dynamics introduction. Society for Industrial and Bibliography
A practical
Applied Mathematics, Philadelphia, PA, USA, 1998.
[6] Phillips A.B. Turnock S.R. and Furlong M. Urans simulations of static
Proceedings of the SIMMAN workshop, 2008.
[7] Liseikin V. Grid Generation Methods. Springer, 2010. 390p.
[8] Warsi Z.U.A Thompson I.F. and Mastin C.W. Numerical grid genera-
tion. Foundations and applications, 1997.
[9] Novikov E. Generalized dynamics of three-dimensional singularities

(vortons). Sov. JETP, 84(3):975 – 981, 1984.
[10] Chorin A. The evolution of a turbulent vortex. Comm. in Mathematical

Physics, 83:517 – 535, 1982.
[11] Chorin A. Constrained random walks and vortex filaments in turbulence
209
theory. Comm. in Mathematical Physics, 132:519 – 536, 1990.
[12] Schlichting H. Boundary layer theory. Springer, 2000.
[13] Batchelor G. The Theory of Homogeneous Turbulence. Cambridge Uni-

versity Press, Cambridge, 1956.
[14] Pope S. Turbulent flows. Cambridge University Press, Cambridge, 2000.
[15] Below I.A. and Isaev S.A. Modelling of turbulent flows. Saint Petersburg,
(2001).
[16] Ferziger H. Bardina J. and Reynolds W. Improved subgrid models for

large eddy simulation. AIAA Paper, pages 80 – 1357, 1980.
[17] Peng S. Hybrid rans-les modeling based on zero- and one- equation
models for turbulent flow simulation. pages 1159 – 1164. Proc. of 4th
Int. Symp. Turb. and Shear Flow Phenomena, 2005.
[18] Davidson L. and Dalstroem S. Hybrid rans-les: An approach to make les

applicable at high reynolds number. Int. J. of Comp. Fluid Dynamics,
19:415 – 427, 2005.
[19] Terracol M. Airframe noise prediction by mean of a zonal rans/les ap-

proach. pages 1165 – 1169. In: Proc. of 4th Int. Symp. Turb. and Shear
Flow Phenomena, 2005.
[20] Kniesner B. Kadavelil G. Basara B. Jakirlic S., Saric S. and Chaouat B.

Sgs modelling in les of wall-bounded flows using transport rans models:
from a zonal to a seamless hybrid les/rans method. pages 1057 – 1062.
Proc. of 6th Int. Symp. Turb. and Shear Flow Phenomena, 2007.
[21] Leschziner M. Temmerman L., Hadziabdec M. and Hanjalic K. A hybrid

two-layer urans-les approach for large-eddy simulation at high reynolds
numbers. Int. J. of Heat and Fluid, 26:173 – 190, 2005.
[22] Strelets M. Spalart P. R., Jou W. H. 200 and Allmaras S. R. Comments on

the feasibility of les for wings and on a hybrid rans/les approach. pages
137 – 148. Advances in LES/DNS: Proc. of the first AFOSR Interna-
tional Conf. on DNS/LES, 1997.
Flow Phenomena, 2005.
[20] Kniesner B. Kadavelil G. Basara B. Jakirlic S., Saric S. and Chaouat B.

Sgs modelling in les of wall-bounded flows using transport rans models:
Lecturesfrom a zonal to fluid
on computational a seamless
dynamicshybrid les/rans method. pages 1057 – 1062. Bibliography
Proc. of 6th Int. Symp. Turb. and Shear Flow Phenomena, 2007.
[21] Leschziner M. Temmerman L., Hadziabdec M. and Hanjalic K. A hybrid

two-layer urans-les approach for large-eddy simulation at high reynolds
numbers. Int. J. of Heat and Fluid, 26:173 – 190, 2005.
[22] Strelets M. Spalart P. R., Jou W. H. and Allmaras S. R. Comments on

the feasibility of les for wings and on a hybrid rans/les approach. pages
137 – 148. Advances in LES/DNS: Proc. of the first AFOSR Interna-
tional Conf. on DNS/LES, 1997.
[23] Spalart P.R. Detached-eddy simulation. Annu. Rev. Fluid Mech., 41:181
– 202, 2009.
[24]
[24] Menter
Menter F.
F. and
and Egorov
Egorov J.
J. Turbulence
210
Turbulence models
models based
based on
on the
the length-scale
length-scale
equation.
equation. pages 941 – 946. Proc. of
pages 941 – 946. Proc. of 4th
4th Int. Symp. Turb. and
Int. Symp. Turb. and Shear
Shear
Flow
Flow Phenomena,
Phenomena, 2005.
2005.
[25] Davidson
[25] Davidson L.L. and
and Billson
Billson M.
M. Hybrid
Hybrid les-rans
les-rans using
using synthesized
synthesized turbulent
turbulent
fluctuations
fluctuations for forcing in the interface region. Int. J. of Heat and Fluid,
for forcing in the interface region. Int. J. of Heat and 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
[27] H. Schlueter
Pitsch H. Schlueter J.
J. U.
U. and
and Moin
Moin P.
P. onsistent
onsistent boundary
boundary conditions
conditions for
for
integrated les/rans simulations: Les outflow conditions. volume
integrated les/rans simulations: Les outflow conditions. volume 3121,3121,
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.
term approach. pages
pages 1141
1141 –– 1146.
1146. 4th
4th Int.
Int. Symp.
Symp. Turb.
Turb. and
and Shear
Shear Flow
Flow
Phenomena, 2005.
Phenomena, 2005.
[29] Shchukin
[29]
[29] Kornev N.,E.
Shchukin E.Taranov
KornevA.,
Kornev N.,Shchukin
N., TaranovE.,
Taranov A. and
A. and Kleinsorge
and KleinsorgeL.L.
Kleinsorge L.Development
Development
Development of
hybrid
of
of urans-les
hybrid
hybrid methods
urans-les
urans-les for flow
methods
methods forsimulations
for flow in the ship
flow simulations
simulations in area.
in Oceanstern
the ship
the ship Engineering,
stern area.
area.
Ocean Engineering,
38(16):1831–1838,
Ocean Engineering, 38(16):1831 –– 1838,
20111.
38(16):1831 1838, 2011.
2011.
[30]
[30] Durbin
Durbin P.
P. Near-wall
Near-wall turbulence
turbulence closure
closure modeling
modeling without
without damping
damping func-
func-
tions.
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., Bensow
Bensow R.
R. and
and Svennberg
Svennberg U.
U. Current
Current
capabilities
capabilities of des and les for
of des and les for submarines
submarines at straight course.
at straight course. J.
J. Ship
Ship
Research,
Research, 54:184
54:184 –196,
–196, 2010.
2010.
[33]
[33] Guminer
Guminer P.K.
P.K. Thermoregulation
Thermoregulation studying
studying in
in hygiene
hygiene and
and work
work physiol-
physiol-
ogy. Medgiz, Moscow, page 147, 1962.
[34]
[34] Novieto
Novieto D.
D. Zhang
Zhang Y.
Y. and
and Yingchun
Yingchun J.
J. Human
Human environmental
environmental heat
heat trans-
trans-
fer simulation with cfd the advances and
fer simulation with cfd the advances and challenges. pages 2162 – 2168.
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.
[32] Fureby C. Huuva T. Alin N., Bensow R. and Svennberg U. Current
capabilities of des and les for submarines at straight course. J. Ship
Research, 54:184 –196, 2010.
[33] Guminer
Lectures P.K. Thermoregulation
on computational fluid dynamics studying in hygiene and work physiol- Bibliography
[34] Novieto D. Zhang Y. and Yingchun J. Human environmental heat trans-

fer simulation with cfd the advances and challenges. pages 2162 – 2168.
Eleventh International IBPSA Conference, Glasgow, Scotland, 2009.
[35] Cherunova I.V. New technologies of heat protection wear design, (in
russian). Technologies of textile industry, pages 51 – 54, 2009.
[36] Palella B.I. Alfano1 F. R. and G.

211Riccio. Thermode 193: an enhanced
stolwijk thermoregulation model of the human body. page 8. 7th Interna-
tional Thermal Manikin and Modelling Meeting, University of Coimbra,
2008.
[37] Smith T. Anatomie-atlas: Aufbau und Funktionsweise des menschlichen

Körpers. Dorling Kindersley, Deutschland, page 448, 2011.
[38] Herman I.P. Physics of a human body. Springer, 860p., 2007.
[39] Zhavoronkov A. Grivina I. and Postnikov N. Pecularities of develop-

ments of imitation model ”human-clothes-medium” (in russian) textile
industry. 3:36 – 37, 1988.
[40] Alexiades V. and Solomon A.D. Mathematical modeling of melting and

freezing processes: Basic concepts & applications. Taylor & Francis,
1992.
[41] Cherunova I. Theoretical foundations of complex design of special clothes

for protection at high temperatures conditions. PhD thesis, South Rus-
sian University of Economics and Service, Shakhty, 2008.
[42] Vitte N.K. Thermal exchange and its hygienic value. State medical
publishing, Kiev, page 145, 1956.
[43] www.hohenstein.de.
[44] Brink I. and Lebedeva E. Investigation of wind influence on packages of

heat protecting cloth. Textile Industry, 3:34–36 (in Russian), 2005.
[45] Menter F.R. Zonal two equation k-! turbulence models for aerody-
namic flows. AIAA 93-2906, 24th Fluid Dynamics Conference, Orlando,
Florida, 1993.
[46] Ginevsky A.S., Vlasov Ye.V., Karavosov R.K., 2004, Acoustic control of turbulent
jets, Springer, 232 p.
[47]
Kornev N., Zhdanov V. and Hassel E. (2008) Study of scalar macro- and
microstructures in a confined jet, Int. Journal Heat and Fluid Flow, vol. 29/3,
665–674.
[48] Frick P., Turbulence, Models and Methods, Lecture course, Perm State University,
Vol.~2, pp.~138, 1999.
202
212
Lectures on computational fluid dynamics Index
Index
"-Equation, 141 continuity equation, 17
a-posteriori test, 158 convective acceleration, 24
"-Equation, 143 energy containing range, 120
a-priori test, 158
energycorrelation function, 101
production, 117
a-posteriori
alternating test,implicit,
direction 160 49 Courant Friedrich
ensemble averaged quantity,Levy criterion,
10237
a-priori test,
amplification 160
of the vorticity, 81 Crank-Nicolson
experiment scheme, 89
of Reynolds, 31
alternating
artificial direction
viscosity, 33 implicit, 49 cross
explicit stress, 157,
scheme, 31 158
amplification of the vorticity, 84
autocorrelation function, 101 explicit time advance, 39
artificial viscosity, 33
density, 25
autocorrelation function, 102 Finite Volume Method, 55
backward difference scheme, 27 Detached Eddy Simulation, 161, 163
folding, 85
backward
Biot-Savart law,difference
80 scheme, 28 Direct Numericalscheme,
Simulation,
forward difference 28123
Biot-Savart
block structured law,
grids, 83
74 Dirichlet condition, 26, 48
Fourier law, 25
block structured grids, 75
body forces, 17 dissipation,
fractional step 139methods, 50
body forces, 18
boundary conditions, 24, 26 dissipation
friction velocity, range,
95 120
boundary conditions, 24, 26
Boussinesq hypothesis, 144 Friedman-Keller
dissipation rate,problem,
117 138
Boussinesq hypothesis, 132
buffer
bufferlayer, 98 99
layer, Dynamic Smagorinsky Model, 152
Galilean invariance, 159
energy containing range, 120
central
central difference
difference scheme,
scheme, 28 28 energy production,
heat conduction 117
coefficient, 26
collocated grid, 41 heat conduction
ensemble averaged quantity, 10026
equation, 25,
collocated grid, 41
constant of Karman, 99 heat flux, 25 of Reynolds, 87
constant of Karman, 97 experiment
continuity equation, 17 heat sources, 25
convective acceleration, 23 homogeneous turbulence, 102
correlation function, 102 hybrid URANS-LES methods, 163
Courant Friedrich Levy criterion, 37
Crank-Nicolson scheme, 31 implicit scheme, 31
cross stress, 158, 159 inertial subrange, 119, 120
inner energy balance equation, 25
density, 25 integral length, 102
Copenhagen
Detached Eddy Simulation, 163, 164
Direct Numerical Simulation, 123
Dirichlet condition, 26, 48
isotropic turbulence, 102
k - " Model, 141
Master of Excellence
dissipation, 139
dissipation range, 120
dissipation rate, 117
k-Equation, 141
Kelvin Helmholtz instability, 91
Kolmogorov first similarity hypothe- cultural studies
Dynamic Smagorinsky
Copenhagen Model,
Master 154
of Excellence aresis, 118
213

www.come.ku.dk

explicit scheme, 31 Large Eddy Simulation, 123, 147

explicit time advance, 38 Lax-Wendroff scheme, 53
Leonard stress, 157, 158
Finite Volume Method, 55 LES equations, 149
folding, 83 LES filtering, 147
forward difference scheme, 28 limiters function, 40
Fourier law, 25 local acceleration, 24
fractional step methods, 51 logarithmic region, 98
friction velocity, 94
Friedman-Keller problem, 138 material substantial derivative, 24
mixed schemes, 39
Galilean invariance, 157 mixed similarity models, 156
morphing grids, 75
heat conduction coefficient, 26
heat conduction equation, 25, 26
Navier Stokes Equation, 21, 79
heat flux, 25
Neumann condition, 26, 48
heat sources, 25
Newton hypothesis, 22
homogeneous turbulence, 101
Newtonian liquids, 22
hybrid URANS-LES methods, 160
normal stress, 21
implicit scheme, 31
overset or chimera grids, 75
inertial subrange, 119, 120
inner energy balance equation, 25
pairing, 89
integral length, 102
PISO algorithm, 70
isotropic turbulence, 103
Poisson equation, 39
Prandtl mixing length model, 96
k - " Model, 139
pressure, 23
k-Equation, 139
pressure correction method, 64
Kelvin Helmholtz instability, 89
Kolmogorov first similarity hypothesis, 117 probability density function, 109
Kolmogorov hypothesis of local isotropy, 118 properties of Reynolds averaging, 128
Kolmogorov law, 120 properties of surface forces, 19

Kolmogorov scale, 119 properties of surface forces, 19
Kolmogorov second similarity hypothesis, 119 pseudodissipation, 139
Kolmogorov theory K-62, 123
Kolmogorov theory K41, 114 Reynolds averaged Navier Stokes, 123
kurtosis, 110 Reynolds averaging, 100
Reynolds stress, 131, 157, 158
lambda structures, 92 Reynolds Stress Model, 134
laminar, 81 Richardson poem, 86
204
scale similarity models, 156 time-advancing, 40

shear stress, 21 turbulence, 81
SIMPLE method, 65 turbulent kinematic viscosity, 132
SIMPLEC algorithm, 73 turn over time, 118
skewness, 110 TVD schemes, 41
Smagorinsky model, 152
Spalart Allmares (SA model), 132
unstructured grids, 75
specific heat capacity, 25
upwind difference scheme, 28
splitting according to physical processes, 51
staggered grid, 41
Van Driest constant, 98
streaks, 92, 94
viscous sublayer, 96
structure function, 109
structured grids, 74 vortex cascado, 116
subgrid stress (SGS), 150 vortex reconnection, 86

subgrid viscosity, 151 vorticity, 82
surface forces, 19 vorticity transport equation, 81
Taylor microscale, 107 wake region, 98
205

Lectures On Computational Fluid Dynamics

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Lectures On Computational Fluid Dynamics

Transféré par

Droits d'auteur :

Formats disponibles

Prof. Dr.-Ing. habil. Nikolai Kornev & Prof. Dr.-Ing. habil.

Lectures on computational fluid dynamics

I Introduction into computational methods for solution of transport equations 16

1 Main equations of the Computational Heat and Mass Transfer 17

2 Finite difference method 27

4 Click on the ad to read more

3 Stability and artificial viscosity of numerical methods 33

4 Simple explicit time advance scheme for solution of the Navier

5 Click on the ad to read more

5 Splitting schemes for solution of multidimensional problems 49

6 Finite Volume Method 55

7 Overview of pressure correction methods 70

How will people travel in the future, and

6 Click on the ad to read more

II Mathematical modelling of turbulent flows 77

10 Basic definitions of the statistical theory of turbulence 99

In the past 5 years we have drilled around

Who are we?

Who are we looking for?

We’re looking for high-energy, self-motivated graduates

7 Click on the ad to read more

11 Kolmogorov theory K41 114

12 Reynolds Averaged Navier Stokes Equation (RANS) 128

13 Reynolds Stress Model (RSM) 134

14 Equations of the k - " Model 139

15 Large Eddy Simulation (LES) 147

16 Hybrid URANS-LES methods 160

III CFD applications to human thermodynamics 171

17 Mathematical model of the ice protection of a human body at high

18 CFD Design of cloth for protection of divers at low temperatures under

20 Simulation of human comfort conditions in car cabins 196

11.11 Power of the structure function. Experiments versus prediction of

19.3 Left: Pressure distribution p on the body obtained using StarCCM+

15 Click on the ad to read more

1.1.2 Classification of forces acting in a fluid

1.1.2.1 Body forces

fE D gU kE (1.5)

1.1.2.2 Surface forces

with the unit kgs 2m m12 D ms

dependence of pEn on the orientation of the surface S.

with the unit kgs 2m m12 D ms

dependence of pEn on the orientation of the surface S.

1.1.2.3 Properties of surface forces

Figure 1.2: Forces acting on the liquid element.

pEn D pEx cos.nx/ C pEy cos.ny/ C pEz cos.nz/ (1.13)

20 Click on the ad to read more

Figure 1.3: Stresses acting on the liquid cube with sizes a.

The stress matrix is symmetric and contains 6 unknown elements:

1.1.3 Navier Stokes Equations

The second law (1.16) takes the form: 21

The second law (1.16) takes the form:

where p is the pressure,

pxx C pyy C pzz

Innovation, flat hierarchies

23 Click on the ad to read more

The local acceleration is due to the23change of the velocity in time. The

1.2 Heat conduction equation

q.x; t/  n.x/ (1.24)

and using the Gauss theorem

Here f is the heat sources within the volume U .

q.x; t/ D rT .x; t/ (1.28)

Since the volume U is arbitrary, (1.29) is reduced to

Z   Z Main equations of the Computational

11.11 Power of the structure function. Experiments versus prediction of

19.3 Left: Pressure distribution p on the body obtained using StarCCM+

fE D gU kE (1.5)

dependence of pEn on the orientation of the surface S.

dependence of pEn on the orientation of the surface S.

q.x; t/ n.x/ (1.24)

q.x; t/ D rT .x; t/ (1.28)

Z Z Main equations of the Computational

rT .x; t/ n.x/ D F1 .x; t/; x2S (1.32)

'i D ax 2 C bx C 'i 1

g nC1 D g n C G.g; t/t (2.16)

g nC1 D g n C G.g n ; t/t (2.17)

g nC1 D g n C G.g nC1 ; t/t (2.18)

The derivatives at i 1 th point can be expressed through these at i th

The derivatives at i 1 th point can be expressed through these at i th

inC1 in n in1

inC1 D in .1 c/ C cin1 (3.13)

inC1 D in .1 c/ C c in1 D ".1 c/

inC2 D inC1 .1 c/ C c inC1 2

inC2 nC1 nC1

inC3 D inC2 .1 c/ C c inC2

inC3 nC2 nC2

l D EE; W W; N N; S S , ApEE D ApW W D .21x /2 , ApN N D ApS S D .21y /2

l D EE; W W; N N; S S , ApEE D ApW W D .21x /2 , ASimple