Académique Documents
Professionnel Documents
Culture Documents
Milada Kozubkov
OSTRAVA 2014
Content
Content
LIST OF USED DESIGNATION ....................................................................................................... VI
1.
1.4.1.
2.1. THE TERM "GRID" AND ITS IMPORTANCE FOR THE MATHEMATICAL MODELING .................... 36
2.2. WORKBANCH, GRID ELEMENTS ............................................................................................. 36
2.3. CRITERIA FOR GRID QUALITY ASSESSMENT ............................................................................ 38
2.4. PKLAD VYTVOEN ST ...................................................................................................... 40
2.4.1. Vytvoen st ................................................................................................................. 41
2.4.2. Definition of the boundary conditions in Gambit ............................................................ 43
3.
Content
5.2. ONE - EQUATION MODEL ........................................................................................................ 67
5.3. TWO - EQUATION K- MODEL ................................................................................................. 68
5.4. RNG K- MODEL ..................................................................................................................... 69
5.5. REYNOLDS STRESS MODEL (RSM) ....................................................................................... 69
5.6. MODELING OF FLOW IN NEAR WALLS, WALL FUNCTIONS ....................................................... 70
5.6.1. Theory of wall functions by Launder and Spalding .................................................. 71
5.6.2. Modeling of flow near the wall in Fluent .................................................................... 72
5.6.3. Non-equilibrium wall function ...................................................................................... 75
5.6.4. Using of wall functions and their limitations .............................................................. 76
5.6.5. Two-layer model (near-wall modelling)...................................................................... 76
5.6.6. The influence of grid quality on the choice of wall functions for various models of
turbulence....................................................................................................................................... 77
5.6.7. The influence of roughness on the wall function ............................................................. 78
6.
DEFINOVNA.
7.1. VOD DO PROBLEMATIKY MODELOVN PENOSU TEPLA ............... CHYBA! ZLOKA NEN
DEFINOVNA.
Content
7.4.2. Okrajov podmnky na tenk dvoustrann stn .................................................... 94
7.4.3. Vpoet teploty a hustoty tepelnho toku na stn ................................................. 95
8.
10.1.
10.2.
10.2.1.
10.2.2.
10.3.
11.
11.2.
11.3.
11.4.
12.
LITERATURA......................................................................................................................... 149
13.
iv
Content
13.1.
13.2.
13.3.
13.4.
14.
14.1.
14.2.
14.3.
15.
15.1.
15.2.
A, S
plocha
m2
CD
konstanta
empirick konstanta
konstanta
C1
empirick konstanta
C 2
empirick konstanta
C3
empirick konstanta
rychlost zvuku
ms-1
cv
Jkg-1K-1
cp
Jkg-1K-1
Dn,m
m2s-1
frekvence
s-1
f i , Fi
sla
fc
Coriolisv parametr
s-1
energie
Jkg-1
empirick konstanta
Fi J
m2s-3
Gr
Grashofovo slo
thov zrychlen
ms-2
hf
Wm-2K-1
obecn vektor
vi
statick entalpie
Jkg-1
ho
celkov entalpie
Jkg-1
I
i , j, k
intenzita turbulence
bilancovan veliina
J e, J p , J w
kgm-2s-1
m2s-2
kP
m2s2
I, L
lm
smovac dlka
Machovo slo
nj
tlak
Pa
p op
operan tlak
Pa
ps
statick tlak
Pa
m2s-3
P J
Pr
Prt , h
prtok
m3s-1
J.kg-1.K-1
tepeln tok
Jm -2s-1
rezidul
normalizovan rezidul
Ra
Rayleighovo slo
Re
Reynoldsovo slo
Re y
rw
s-1
vii
S i, j
s-1
Sc
Schmidtovo slo
as
absolutn teplota
ui
ms-1
ui
ms-1
u i
ms-1
u , u*
ms-1
tec rychlost
ms-1
objem
m3
xi
y , y*
yv
y P
relaxan faktor
teplotn vodivost
m2s-1
k , ,
m2s-1
K-1
ij
Kroneckerovo delta
rychlost disipace
m2s-3
m2s-3
disipan funkce
turbulentn difusivita
m2s-1
Wm-1K-1
Wm-1K-1
yv
viii
dynamick viskozita
Pas
eff
efektivn viskozita
Pas
turbulentn viskozita
Pas
kinematick viskozita
m2s-1
turbulentn viskozita
m2s-1
ij
Pa
hustota
kgm-3
ref
referenn hustota
kgm-3
empirick konstanta
empirick konstanta
asov perioda
vazk napt
Pa
ij
Pa
Pa
turbulentn napt
Pa
druh viskozita
Pas
obecn promnn
Indexy:
stac index
j, k , l
w, e, p
W , E, P, N , S , F , B, NB
1
ix
Preface
Preface
This textbook is addressed to for students of master and doctoral programs of all
faculties who want to learn about the fundamentals of numerical modeling of fluid phenomena
in transfer, ie transfer of mass, momentum (torque), heating, etc. in laminar and turbulent flow.
In applications in teaching CFX and Fluent software will be used.
Numerical modeling of many physical phenomena are closely linked to modeling a form
of motion mathematically. Movement of fluids is associated with the solution of various
problems of the physical model:
multiphase flow, free surface flow, flow with solid particles, bubbles, respectively. drops
The mathematical model consists of defining the equations describing the above happens.
Given that this is a two-dimensional, axially symmetric or general three-dimensional and time
dependent cases, which are described by system of partial differential equations, which can
be solved by numerical methods. Their use is given by the expansion of knowledge in the field
of flow, turbulence, numerical methods, computer technology
The flow solution is possible to use commercial software systems, such as Fluent, CFX
and others. The task of the user is to build the correct calculation model, which contains some
mathematical, physical and technical principles. For such a model it is necessary to find all
input data in the existing standards, build the input data for a program that can solve
computational model, solving by the terminal, correctly interpret the results for further use in
all phases and carry out effective monitoring of all inputs and outputs. The user must safely
divide all the information on the geometric data (two-dimensional or three-dimensional
features, topology), data on the effect of external forces and physical data (information about
flowing media, its physical properties). Thus, an essential task is to know hydromechanics,
thermodynamics and other sciences to the complexity of the problem.
As regards the computational methods underlying the use of the program, the designer
should know their nature for reliable use in standard cases. For program or Fluent. CFX it is a
need to know in what shape the final volume will work, it follows the choice of network density,
Preface
approximation schemes, the nature of the time dependence of the quantities and the resulting
time step size, etc.
No less important part there is the evaluation results, particularly difficult in threedimensional problems. It is optimal to have at least the approximate value of calculated vriables
, better is to compare the results with the experiment.
Literature
Bojko, M.: Nvody do cvien Modelovn proudn Fluent
Blejcha, T.: Nvody do cvien Modelovn proudn CFX.
xi
potencial
turbulent
laminar
swirling
0
t
y (x2)
v (u2)
w (u3)
coordinates
X x, y, z
respectively
X x1 , x2 , x3 .
Velocity
z (x3)
vector
is
defined
by
components
u u, v, w
respectively
u u1 , u2 , u3 . The
u (u1)
x (x1)
Fig. 1.1 Coordinate system
1.2.
of incompressible fluid
Basic laws of physics describing the flow are the conservation laws of mass,
momentum, heat or other scalar variables. They are expressed by Navier Stokes equations
along with continuity equation and describe the laminar and turbulent flow regime. In the case
of non isothermal unsteady incompressible flow they have the following form:
Continuity equation:
u v w
O
x y z
( 1.2.1)
Navier-Stokes equation:
2u 2u 2u
1p
u uu uv uw
2
2 fx
2
x
y
t
x
y
z
x
2v 2v 2v
1p
v vu vv vw
2
2 fy
2
x
z
y
t
x
y
z
y
( 1.2.2)
2w 2w 2w
1p
w wu wv ww
fz
2
2
x
y
z
t
x
y
z
z
where using schema in Fig. 1.1 u , v a w are the velocity components, p is pressure, is
density, kinematic viscosity ane f x , y , z refers to components of volume or external forces
(gravity, centrifugal force).
The equation of heat transfer, ie. the law of conservation of energy is in the form
2T 2T 2T
T uT vT wT
t
x
y
z
y 2 z 2
x
u 2 v 2 w 2
2
x y z
( 1.2.3)
u v 2 u w 2 v w 2
y x z x z y
where
heat.
In terms of variables with three or nine components (components of speed, stress, etc.)
it is appropriate use special abbreviated designation with precisely defined rules, known as
Einstein's summation, see chap. 13, where only one member can express all three
components of velocity respectively nine stresses. The same can be expressed
mathematically for clarity using character sums. Thus, the continuity equation is written simply:
u1 u2 u3
O resp.
x1 x2 x3
uj
x
j 1
O resp.
uj
O
xj
( 1.2.4)
Navier-Stokes equations:
n
1p
ui n uiu j
2u
2i fi
t j 1 x j
xi
j 1 x j
resp.
( 1.2.5)
1p
ui uiu j
2ui
fi , i 1,..., n
t
xj
xi
x 2j
where the index i consistently expresses the component of vector and the index j (or other
alphabetically)
represents
the
summation
index
( j 1,2 resp. 3 ).
2 ,
t j n x j
j n x j
1 n n u
u
j l
2 j 1 l 1 xl x j
( 1.2.6)
resp.
2T
T u jT
xj
x 2j
t
1 u
u
j l
2 xl x j
1.3.
The equation for conservation of mass and momentum, energy equation and the
equation for the transport of chemical admixtures in the general conservative form create a
system of partial differential equations, while all the equations can be formally written
u j S
t
xj
x j x j
( 1.3.1)
Let the problem to find the temperature distribution in the body is defined. Suppose that
finding function T (ie. temperature) is a function of subsequent spatial independent variables
and time
T=T(t,x)
T
2T
2
t
x
( 1.3.2)
T=T(t,x, y)
T
2T
2T
2 2
t
x
y
T=T(t, x, y, z)
T
2T
2T
2T
2 2 2
t
x
y
z
.
If the function T depends only on the spatial coordinates and independent of time, the task
is called stationary. The problem is complicated if the function T dependends on time and
possibly the spatial coordinates, as mentioned above. For example the body is heated or
cooled with increasing time, then the case is called transient and function T is a function of
successive independent variables.
INDEPENDENT VARIABLES
stationary flow
transient flow
t
t, x
x, y
t, x, y
x, y, z
t, x, y, z
Example
For
t
example
equation
of
heat
conduction in a rod is
T=T1
T
2T
2
t
x
T=T2
T(x)=(x)
x
Fig.
T x,0 x
0 x L
( 1.3.3)
In the event that the second derivative T with respect to t occurs in the equation, the second
initial condition is defined as
T x,0
x .
t
The boundary condition defines the temperature at the beginning and end of the field
(edge rod).
T 0, t T1 t
( 1.3.4)
T L, t T2 t
The role of finding a solution of equation (1.3.1) satisfying the boundary and initial
conditions is called a mixed role. If the boundary conditions equal zero, they are called
homogeneous boundary conditions, similar to the initial conditions are zero, they are called
homogeneous initial conditions. Instead of boundary conditions (1.3.4) it can be put another
type of conditions that are also called boundary conditions. Reflection on the boundary and
initial conditions for temperature is valid for a general variable
1.4.
constants y const.
0.05
0.045
0.04
significant digits
derivative according to normal (OUTLET, heat
flux)
y x
konst1.
x
u-konst
0.03
u-polynom
0.025
u-po stech
lin. funkce
0.02
0.015
0.01
0.005
x1, y1 , x2 , y2 , x3 , y3 ,
y [m]
0.035
...
xN , y N
0.5
1.5
-1
u [m.s ]
o u tlet
ry ch lo st
tla k sta t.
ry ch lo st
tla k sta t.
boundary conditions
Conditions at the flow border - the input and output flow borders can be defined by three
types of boundary conditions
INPUT
OUTPUT
condition of stationary flow
velocity u
u
p
T
0,
0,
0
n
n
n
static pressure pstat
1 2
u
2
temperature T
Conditions on the wall - the wall can be stationary (fluid velocity is zero) or moving (eg
rotating or sliding), with friction or without friction, smooth or rough. The temperature is given
T
0 is insulated
n
wall.
Conditions of symmetry - zero normal velocity and zero normal gradients of all search
parameters, see Fig. 1.6.
Conditions of axial symmetry - define the axis of the axially symmetric two-dimensional
problems, see Fig. 1.5.
solved
region
solved
region
axis
of
symetry
plane of
symetry
Periodic (cyclic) conditions - used when the flow formations are repeated, they can be
rotational and translational type, which allows the definition of the pressure gradient in the
direction of fluid flow along the field.
All types of conditions can be time dependent, if required by their character. Other
boundary conditions do not concern of the flow as such, but other values arising from the
complexity of the mathematical model as, temperature, heat flow, radiation, mass fractions (or
molecular fraction). additions, etc.
periodical condition
periodical condition
of rotational type
of translation type
solved region
solved region
1.5.
This chapter illustrates how to enter and solve the incompressible flow in a sudden
expansion cross section using Fluent, which is one of the fundamental tasks of flow testing,
examined both experimentally and numerically. The task is: define the physical model, physical
properties of the flowing medium and walls:
define the physical model, physical properties of the flowing medium and walls
define mathematical model, boundary condition
create geometry and mesh
specify the initial and boundary conditions in FLUENT, the calculation
evaluate the computed values
Example
The physical model is given by the shape, type of flow and hydraulic parameters of
flow. The scheme of the region is shown in Fig. 1.8 and dimensions with the physical properties
in the table.
Ls
ds
u
d-ds
Oblast proudn
L
Fig. 1.8 Geometry of the problem
d ds
x
Ls
3.5
[m]
[m]
0.5
expansion
[m x m]
0.7 x 0.1
1.2
0.0000171
Boundary condition
Boundary conditions are defined at the input by the velocity and at the output there is
determined the condition of steady flow (OUTFLOW). On the walls it is assumed zero velocity.
input mean velocity u s [m.s-1]
output
OUTFLOW
Reynolds number Re
u s d
[1]
84705
Mathematical model
From the boundary and physical conditions the Reynolds number can be determined
as the dimensionless criterion for deciding whether the flow is laminar or turbulent. The value
of Reynolds number characterizes the flow in the transition region between laminar and
turbulent flow. It will therefore be solved at beginning as laminar.
Creating geometry and mesh
In a Ansys the exact geometry will be created by method similar to the CAD programs.
In addition, this program will use the possibility to form a mesh. The resulting mesh should
have the area shape taking into account the development of swirl and velocity profile near the
wall, see Fig. 1.9 Computational mesh in longitudinal section.
10
1.6.
Solve the water flow between two infinitely large plates. The physical model is determined by
the shape of the area, the type of flow and hydraulic parameters of flow. Numerical calculation
define in the programming environment ANSYS Fluent. The creation of the computational
domain (geometry) and of computer networks and uses programs DesignModeler ANSYS
Meshing.
upper wall
solved domain
outlet
inlet
s
x
lower wall
l
11
upper wall
fluid
outlet
inlet
x
lower wall
l
Fig. 1.14 Geometry of 2D rectangular computational domain
[m]
0.5
length of domain
[m]
0.033
Tab. 1.2 Physical properties of water
998
viscosity [kg.(m.s)-1]
0.001003
Boundary conditions
At the entry velocity boundary condition (VELOCITY INLET) is defined and the output is given
by the condition of static pressure (OUTLET PRESSURE). On the wall there is a boundary
condition of type WALL, where it is assumed zero velocity (is predefined). The boundary
conditions are given in Tab. 2.3
.
us
0.05
[Pa]
Mathematical model
12
Selection of a mathematical model will be resolved in later chapters, now it leaves predefined
one (laminar flow model).
Creating of geometry and networks
To create a domain and network the DesignModeler program and ANSYS Meshing is
used. Programs are available in the ANSYS Workbench program manager. After starting
Workbench is available environment shown in Fig. 1.15. On the left side of the tools Toolbox
- Analysis Systems" the "Fluid Flow"(FLUENT) is selected and draged by the mouse with the
left button and moved the item to the desktop (Project Schematic). This is a new project, which
can be renamed (eg. 2D-plate, see Fig. 1.15).
In the first phase, the program will create the geometry in DesignModeler using the
"Geometry". When the default startup command DesignModeler "Geometry" user creates 3D
geometry. If you need to create 2D geometry, it is necessary to check the right panel, under
Properties of Schematic A2: Geometry - Geometry Advanced Options - Analysis Type - 2D
and that will create 2D geometry. This setting is induced by right-click the command
"Geometry" and select "Properties". Then the program starts DesignModeler by click
"Geometry".
13
individual commands.
Commands
operations
Working area
Sketch
Plane of coordinate system
Fig. 1.17 Creating Sketch
Subsequently, the rectangle is formed by drawing individual edges, forming a closed unit. To
drawing the edges the tool from the "Draw" in the "Sketching" is used. To create a individual
entities of rectangle the "Line" is used with help of mouse to mark the two points in the
workspace and create the edge (Fig. 1.18). By the same procedure all the edges will be
14
15
Command Generate
Dimensions
16
Command Generate
17
18
19
20
After the creation computational grid, labeling and naming of all boundary conditions go back
to the Workbench environment. ANSYS Fluent starts using "Setup" (Fig. 1.29). Before running
the "Setup" it is necessary to ascertain whether the items "Geometry" and "Mesh" have a green
check mark. If not, then you must update the project with the command "Update Project".
Update of project
Start of ANSYS
Fluent
Work continues by confirmation of choosing OK. Follows checking network, showing all
21
borders and throughout the area command "Display". Select of all items in the "Surface"
appears boundary conditions and network. The network is to illustrate very rough. Practically,
it will be much smoother and compressed toward the walls.
It is necessary to check the dimensions of the network "Mesh-Scale" where you see the real
size area. If the computational domain is created in other dimensions (mm, cm, ...), you can
use the "Scaling" and "Specify Scaling Factor" convert dimensions to the units meters (m), see
Fig. 1.32.
22
Follows the checking of the existence of negative volumes in the network by command "Check
Mesh", see Fig. 1.34, which occurs in complex geometries and in this case it is necessary to
create a network again.
If all settings are correct, you proceed the menu from left to right and top to bottom. Follows
command "General", where you set the type of "Solver", and especially "Steady" or "Transient"
for the task stationary or time-dependent. It is necessary to check whether the problem solves
2D planar "Planar", axially symmetric "Axisymetric" or axially symmetric rotation "Axisymetric
Swirl". In Fig. 1.35 we can see, that there is possible to define the external force (eg. Gravity)
with acceleration "Gravity" in any direction and to change the physical units "Units" from SI unit
system to another unit or only selected variables.
23
In the "Models" (see Fig. 1.36) there is defined physical nature of the task according to the very
visual offers, whether it is a multi-phase flow "Multiphase", flow with heat "Energy", flow of
admixtures "Species", flow with discrete phases "Discrete Phase" or what type of flow
"Viscous. Here you can define laminar flow, turbulent flow using various turbulence models
and you can also solve the special case of ideal fluid flow "inviscid". Comments will be made
in the relevant chapters later. In this application, define the laminar flow model.
In Materials their physical properties can be defined, first flowing medium "Fluid" and,
secondly wall material "Solid", see Fig. 1.37. If the material does not appear in the main menu
"Materials", then it must be connected in the "Create / Edit Material" button "Fluent Database"
24
where you will select and use the "Copy" to join the list of materials. Physical properties
(Density, Specific Heat, Thermal Conductivity, Viscosity, ...) are shown at the bottom of the
menu and can be changed according to the requirements of the investigator.
The flowing material is water, so the material will be loaded into the system from the "Cell Zone
Conditions" command "Edit", see Fig. 1.38. Especially when inserting new materials, it is
necessary to check their choice and confirm here, otherwise there will remain first in the list of
material.
25
Follows the definition of boundary conditions. For each condition it is necessary to check and
correct these items. Offers are more complicated in the case of more complex models of flow.
Setting of input and output boundary conditions is shown in Fig. 1.39 and Fig. 1.40.
26
Subsequently, the initialization of the current flow field is made, ie. Definition of initial conditions
27
inside the entire area. The values are defined based on the input boundary conditions, see
Fig. 1.41.
28
29
The inlet velocity boundary condition is defined therefore the constant value of the mean velocity
us 0.05 ms-1 . Subsequently, the initialization of the flow field is done and the numerical calculation
starts. To illustrate the evaluation the velocity vectors, velocity profiles and filled isolines are displayd.
The velocity vectors are defined in each cell of the computational domain by command "Display /
Vectors / SetUp / Vectors", where it is possible to define vectors of selected quantity (the most common
is velocity and is also predefined) painted by different variable (eg. Temperature). The "Scale" will
allow to increase the size of the vector and "Skip" to skip a given number of vectors to be less dense.
Thus for Scale = 10 and Skip = 10 result is apparent from Fig. 1.44.
[ms-1])
For better overview it is possible to create the helper sections of given coordinates in which the vectors
appear. Create a section of the value of x = 0.1 m can be made eg. by typing "Surface / Iso-suface /
Surface of Constant / Mesh / X-Coocordinate" and set the Iso-Values 0.1 and call it such X-coordinate0.1, see Fig. 1.45.
30
In individual sections and boundaries the velocity profiles are defined by the command "Display /
Vectors / SetUp / Vectors". In the "Surfaces" used planes are selected, see Fig. 1.46. Then the vectors
are plotted by command "Display", see Fig. 1.47.
Fig. 1.46 Menu for the creation of the velocity vectors in the individual sections
31
[ms-1])
The evaluation shows that the length of the computational domain leads to gradual formation of the
parabolic velocity profile. To achieve the desired flow profile (from the previous solution) the
computational domain is short. The contours of velocity magnitude are shown in Fig. 1.49 and rendered
with command "Display / Contours / Contours of / Velocity". Next, specify whether the magnitude of
velocity, velocity component or other variables related to velocity will be drown. "Levels" defines the
number of isosurfaces. By checking the "Filled" in the "Options" the filled isolines appear, otherwise it
is peers, see Fig. 1.48.
32
[ms-1])
p stat
[Pa]
Further evaluation presents the velocity profiles in the various sections from inlet to outlet with a step
0.05 m along the length of the computational domain, see Fig. 1.52. This figure is very instructive, if
it is necessary to compare the profiles of quantities in input, output, or in other sections of the field. It is
set with command Display/Plot/Y Axis Function/Velocity/Velocity Magnitude. In the "Plot
Direction" we put the value 1 for X, if we plot the dependence on X or value 1 for Y if we plot the
dependence on Y. Other variables are equal zero, see Fig. 1.51.
The results indicate the formation of the velocity profile from the constant velocity value at the inlet to
the parabolic velocity profile at the outlet from the area (Fig. 1.52). The first image is a copy of the
screen and is suitable for fast orientation in the result. Far better result is obtained by checking in the
"Options / Write to File", which will offer a filename (file is in text format). This file is then read and
modify in Excel. The result is in the second part of the picture.
33
Fig. 1.52 Velocity profiles in different sections of the computational domain ( u [ms-1])
Another evaluation of the progress of static pressure along the length of the computational domain is at
the Fig. 1.53. Static pressure is evaluated in the axis of the computational domain.
34
Fig. 1.53 Progress of static pressure evaluated along the axis of the tube ( p stat [Pa])
35
modeling
The grid is a system of calculated field distribution in the partial consecutive cells in 2D
or 3D cells in two-dimensional or three-dimensional space ([23]). We can say that
computational grid coverage is the basis of mathematical modeling. Since a separate
mathematical model (a system of mathematical relations) is only a "passive" tool, which takes
effect only when it is applied to a specific problem (computational network coverage).
If you talk about mathematical models that are based on the numerical solution of
partial differential equations and thus require input and boundary conditions, we can say that
the possibility of realizing the tasks are very limited power of computer technology. This applies
several principles here:
calculation is even more difficult (slower), the more equations in the mathematical model
included in the calculation (according to the demands and complexity of the model);
calculation is even more difficult, the more the cell is calculated;
calculation is even more difficult, the less good network is prepared for calculation.
2.2.
Workbanch is much broader than just the physical drawing and networking. Fig. 2.1
schematically illustrates the basic structure of the engagement of other programs into the
network.
36
Workbanch- FLUENT
Basic structure of program
FLUENT
Meshing
-Definition of geometry
-2D/3D grid
grid on the border
of region
2D/3D grid
FLUENT
1. Geometry, generation of grid
2.
3.
5.
6.
7.
8.
Physical model
Boundary conditions
Physical properties
Parameters of calculation
Calculation
Evaluation,
result interpretation
grid
TGRID
-2D/3D grid
-hybrid grid
37
block
prizmatic
element
tetrahedron
pyramid
2.3.
cell size (with respect to the modeled process and the requirement for accuracy of
calculation)
suitability of structure in the cell compartment (eg grid refinement of interesting places
in terms of fluid flow) with respect to the particular type of task
quality of the cells (asymmetry - Skewness, the ratio of edges (faces) elements - Acpect
Ratio, etc.)
38
The most important criterion for cells quality assessment is the asymmetry when
assessing how much the cell shape closes to the ideal regular geometric shape in accordance
with the appropriate grid schema. If the cell is deformed in any way, its quality is worse.
Generally, the quality of each cell is expressed by dimensionless number in the range 0 - 1,
where 0 is the best result and 1 is worst result on the contrary, that is problematic cell for
calculations. This value is called the "degree of cell skew " (or "skewness measure"), or the
degree of deformation.
Optimal area
(equilatteral)
Optimal cell
(equilateral)
Circumscrib
ed circle
Aktual cell
Teoretical
circumscribed sphere
area
Aktual area
To determine the quality of 2D cells, respectively. its rate of deformation is the following
relationship:
Soptimal Sreal
( 2.3.1 )
Soptimal
where Soptimal presents optimal cell area, Sreal real cell area (which may or may not be
optimal) and Skewness measure (TRI ) is term for "peace cell deformation" relating to the
2D triangular diagram of the grid. For other schemes we use similar logic. The resulting value
should not exceed 0.85. Should this happen, you need a cell or a grid structure to edit, order
not to endanger feasibility and accuracy of the calculation.
To determine the quality of the 3D cells corresponding to grid diagram consisting of
tetrahedrons similar relationship is applied:
Voptimal
presents
optimal
Voptimal Vreal
Voptimal
cell
volume,
( 2.3.2 )
Vreal
real
cell
volume
and
Skewness measure (TET ) is a term for "peace cell deformation" relating to the 3D
tetrahedron diagram of the grid. The resulting value should not exceed 0.9. Should this happen,
39
Fig. 2.6 Approximate color range to assess the quality of grid cells
(0 - highest quality cells, 1 - the lowest quality cell)
The quality of the network can be tested and count in preprocessor Gambit 2.2.30, using
"Examine Mesh". After entering this command window it is available to provide several options
to control local or global mesh quality across the computing field. The test result is a grid cells
in color hue, which corresponds to the approximate range from color quality level of the cell.
2.4.
Pklad vytvoen st
geometrickho
a
proudn
st
ve
pro
ventilu
een
pi
dnou
nvaznost.
exportovat
(uloit) ve formtu, kter je pro Gambit srozumiteln a doke nast potebn data o geometrii
modelu. Proto byl proveden export modelu ze Solid Edge do souboru o formtu IGES a STEP,
co jsou dva nejpouvanj soubory pro obecn exportovn dat, tedy pedevm geometrie
modelu. Po importu do Gambitu peklada tohoto programu bohuel nebyl schopen nast
celou geometrii. Proto se geometrie penesla jako mezioperace do programu Ideas, ze kterho
se dle exportovala do IGES formtovho souboru a nsledn opt do programu Gambit.
40
Model
je
konstruktrsk
een
jako
soust,
co
je
program
Fluent.
nutno
provst
vech
sousti,
kter
objem
nejsou
oblast proudn
v bezprostednm kontaktu
s kapalinou (pruiny, tsnn, plastov obal atd.) se provede opt pomoc booleanovskch
funkc (prnik) a zstane jen objem vnitn, tedy objem fluidn (), ve kterm bude eeno
proudn.
2.4.1. Vytvoen st
Vzhledem ke sloitosti 3D geometrie byla zvolena s s tetra/hybridnmi prvky (3D
tystny), kter je nestrukturovan s velikostmi hran jednotlivch element 0.17 a bylo
vytvoeno zahutnm kolem ostrch hran uitm size function v mst, jak je naznaeno na .
S takto vytvoenou st se vygenerovalo 1 500 000 bunk.
41
42
the conditions for the continuity areas (continuum types), the specification of the area
of flow or solids.
Creation of boundary conditions is realized by specifying the properties of the selected model areas for
further work, thus creating pressure input, output, creating two symmetrical walls (rotating part) and
other walls that have no other function than to just enclose the "fluid" volume, see .
Tlakov vstup
Pressure inlet
Tlakov vpus
Pressure outlet
symetrick podmnka
Symmetry 1,2.
zbyl stny
Wall
Fig. 2.13 Boundary conditions
This kind of gid is exported to a file with suffix .msh and thus all work in the pack are ended.
43
equations
The differential method is the oldest well-known method for solving differential equations,
which is used in the illustrative example of heat conduction in a rod. It consists in replacing the
derivatives by using differential Taylor rates, deriving differential equations and their solutions
The finite volume method is briefly told in three basic points
dividing the area into discrete volumes using general curvilinear network
Fluent defines discrete final volumes using non-staggered scheme, where all variables are
stored in the centers of finite volumes.
Currently begins to form the finite element method in fow solving, which consists of those
points
the dividing the area into triangular or quadrangular elements in the two-dimensional (2D)
area or tetrahedrons resp. hexahedrons in the three-dimensional (3D) area
minimization of residual
Special method is a spectral method which is suitable for periodic flow in a single regions
(Taylor vortices arising in the gap between concentric cylinders, one of which rotates).
Other chapters are devoted only finite volumes.
3.2.
and for the simplicity the equations of one spatial independent variables which can be thought
as the flow in the three-dimensional space, where all variables in the derivative direction
and
are
zero.
Flow
is
also
44
stationary
(time
independent).
u
0
x
( 3.2.1)
momentum equation
uu 1 p u S
x
x x x
( 3.2.2)
u S
x
x x
( 3.2.3)
Fig. 3.1 Coordinate scheme using special notation of cells for 1D and 3D model instead of
subscripts, where N North, S South, E East, W West, F Front, B Back
By integrating these equations over the final volumes the initial differential equations
are converted into volume integral (dV=dx.dy.dz, dA=dy.dz), using divergence theorem (
a x
x
V
a y a z
y
z
dxdydz
a dydz a
x
geometry the capital letters indicate the center of finite volume and small letters indicate
boundaries, i.e. walls between the finite volumes ) and using discretization on the resulting
algebraic form as follows:
u
u
x dV x dxdydz u dA uA uA
e
( 3.2.4)
uAe uAw 0
Physical expressions on the left side indicate the flow rate difference
Qe Qw 0
( 3.2.5)
45
uu 1 p u S
x
x x x
u uP
u u
1
A w P W
Qeue Qwuw pe pw A e E
xe
xw
( 3.2.6)
A SV
and the equation for the scalar quantity is adjusted using the same procedure to form
P
W
w P
Qe e Qw w e E
xe
xw
A S V
( 3.2.7)
In the previous equations there are used both coefficients and variables defined in the centers
of finite volumes and on the walls of these volumes (e.g. speed in equation (3.2.6)). This is a
disadvantage and it is absolutely necessary in the next to unify storage quantities only in the
centers of finite volumes. If this value will be determined on the wall, the interpolation scheme
for the interpolation of this magnitude in the center of the cell must be used. For illustration the
simplest scheme is used, i.e. the arithmetic mean and differential scheme is simplified. For
example equation (3.2.6) can be modified as follows:
uE uP
u uW
Qw P
2
2
u uE
u uW
1
pe pw A e P
w P
xe
x w
Qe
( 3.2.8)
A SV
Qe
P E
2
Qw
P W
2
E
W
e P
w P
xe
xw
A S V
Then it is possible for this general equation in one-dimensional case to express P using the
values in neighboring finite volumes following modifications
Qe Qw
A
A
A Qe
A Qw
P e
W S V
e
w
E w
2
2
2
2
x
x
x
x
e
w
e
w
AP P AE E AW W SC
( 3.2.9)
P Ai S P Ai i SC
i
( 3.2.10)
where the sum is performed over the neighboring cells (in 1D case is i=E, W; in 3D case i=N,
S, E, W, F, B,). Ai are coefficients, which contain contributions from convection, diffusion and
source members and SC and SP are the linearized source and S = SC + SP. P. The sign is
visible from z .
Equations solved in Fluent are an extension of the previous three-dimensional
46
curvilinear coordinate system. Each iteration consists of the steps that are shown in diagram
on .
END
START
equations of motion for the unknown velocity components are solved using the pressure
values in order to update the velocity field
velocity specified in the previous point can not satisfy the continuity equation, thus the
pressure correction and subsequent correction of velocity field shall be determined
using new values of velocity the equation for the turbulent energy k and dissipation is
solved
another equation to determine the temperature and other scalar variables are solved
3.3.
FLUENT stores the components of velocity and scalar quantities in geometric centers of
finite volumes defined by grid. Because the calculation process, the required values of these
47
variables on the border of finite volume are used. These values are obtained by interpolation,
while you can choose between the following three variants differing order of accuracy
(ascending)
power interpolation
QUICK
During large changes in pressure and flow rozpotat should be available to compute with the
lowest order of accuracy (which is predefined) and after a few iterations to use higher order of
accuracy (for vortex flow with heat transfer, dissipation, etc.)
3.4.
Convergence.
3.4.1. Residuals
During flow simulation using Fluent it is very important to obtain convergent solutions.
The measure of convergence are residuals, which represent the maximum difference between
two corresponding values at the same grid point in two consecutive iterations. Residuals are
evaluated for all values computed in each iteration step and displayed for the selected
variables.
i+1-th iteration
Pi+1
i-th iteration
Pi
Strictly speaking the measure is the sum of the changes of computed variables in the
equation for all cells in the area. After discretization the equation a conservative onedimensional in shape for the general variable is derived as
48
AP P AE E AW W SC
Residual R is than defined by sum for all P
R AE E AW W SC AP P
( 3.4.1)
A
E
AW W SC AP P
A
P
( 3.4.2)
P
Residuals can be evaluated graphically in each iteration step, see Figure 3.4, and the
decreasing value of the residuals indicates good convergent task. Numerically exact values
can be observed in tab.3.1..
Fig. 3.4 Dependence of pressure and enthalpy residual on the number of iterations
Normalized reziduals
49
number
of pressure
velocity u
velocity v
entalpy
iterations
5.00000E+00
2.647l0E-01
3.62767E-01
4.01625E-01
3.49503E-01
1.00000E+01
4.21854E-02
7.31087E-02
6.09502E-02
1.15607E-01
1.50000E+01
1.61787E-02
5.57186E-02
6.77023E-02
6.28091E-02
2.00000E+01
9.91924E-03
4.11899E-02
5.52667E-02
4.24032E-02
2.50000E+01
7.78245E-03
3.71804E-02
5.02612E-02
3.19044E-02
3.00000E+01
6.71127E-03
3.33559E-02
4.61688E-02
2.55360E-02
3.50000E+01
4.96045E-03
3.13033E-02
4.34564E-02
2.11992E-02
4.00000E+0l
6.07668E-03
3.01096E-02
4.09786E-02
1.80783E-02
4.50000E+01
5.21358E-03
2.85215E-02
3.89507E-02
1.56768E-02
5.00000E+01
6.70681E-03
2.67667E-02
3.67708E-02
1.38577E-02
5.50000E+02
5.67326E-04
1.93322E-03
1.06027E-03
1.25243E-04
5.55000E+02
6.90138E-04
1.85336E-03
9.84238E-04
1.18761E-04
5.60000E+02
5.32427E-04
1.78220E-03
9.68897E-04
1.12885E-04
5.65000E+02
4.20846E-04
1.68717E-03
1.02076E-03
1.07433E-04
5.70000E+02
3.76113E-04
1.62817E-03
1.07420E-03
1.02499E-04
5.75000E+02
3.26542E-04
1.52597E-03
1.07393E-03
9.78981E-05
5.80000E+02
3.00249E-04
1.47025E-03
1.08095E-03
9.36884E-05
5.85000E+02
2.94075E-04
1.31286E-03
1.07175E-03
8.96829E-05
5.90000E+02
2.54086E-04
1.16047E-03
l.05932E-03
8.59168E-05
5.95000E+02
2.33645E-04
1.02260E-03
1.05061E-03
8.23961E-05
6.00000E+02
2.12518E-04
9.10742E-04
1.04899E-03
7.90365E-05
atd.
tab. 3.1
It is also possible to assess at what point is the highest residual value. Residuals are used to
evaluate convergence. Generally, the solution converges very well when the normalized
residuals are reduced to the value of the order of 1.10-3 and enthalpy residuals to the value of
the order 1.10-6.
50
variables defined by initialization, are often considered to be zero at the beginning of the
calculation. The most important examples of setting the initial conditions are:
temperature for solving of heat transfer problems when using the equation of state
flow with chemical reaction, when it is available to set the temperature and the mass
fraction of species
An important technique to accelerate convergence is the step by step technique (step by step
from simple to more complex tasks). To solve the problem of heat transfer it is good to start
the calculation from izothermal flow, in case of reacting flow to start the solution without the
inclusion of additives. The problem is defined at first complex and then would be necessary to
select the variables for which initial state will be resolved .
3.4.3. Relaxation
Due to nonlinearity of differential equations is not generally possible to obtain values of
all variables by solution of originally derived approximations of difference schemes.
Convergence can be achieved using relaxation, which reduces the changes of each variable
i
,
P
depends on the old value from the previous iteration P,i , the new value from the current
iteration P , i 1, vyp (or calculated changes P P , i 1, vyp P ,i and relaxation parameter
0,1 follows
P ,i 1 P .i 1,vyp 1 P .i
( 3.4.3)
resp.
P ,i 1 P ,i P
51
52
edge where the laminar boundary layer is transferred to turbulent. It can be seen that the
determination of flow is not simple and straightforward and depends on the experience of the
investigator.
53
vs
In the case of one-dimensional flow in the pipe the transition to turbulence is given
is the mean velocity in the pipe, d its diameter and kinematic viscosity. The critical value
Rekrit for a pipe of circular cross-section is the 2320. By Re Rekrit the ordered laminar flow
in the pipeline develops, the movement happens in layers and the fluid particles do not move
across section. If Re Rekrit , the flow is turbulent. At higher Reynolds numbers, the fluid
particles held disordered movement in all possible directions. This motion is irregular, random
related to motion of molecules of gas, but unlike molecules of the fluid particles can disintegrate
and thus lose their identity. The movement of particles perpendicular to the wall increases the
momentum flux to the wall and therefore the pressure drop in the flow direction is much greater
than in laminar flow. As a result of fluid mixing the velocity differences in different places are
much smaller in cross section than in laminar flow at larger distances from the wall.
The flow can be visualized by different methods to observe the differences in laminar
and turbulent flow see fig 4.2. In turbulent boundary layer the turbulent (coherent) vortex
structures characteristic just for of turbulent flow can be defined.
54
between the fluid with many different areas of velocity, which is reflected in increasing
resistance to flow as an internal fluid friction.
shear stress, resulting in turbulent flow is not only determined by internal friction in the fluid
and velocity gradient, as is the case of laminar flow (Newton's law
dv
), but by the change
dy
turbulent viscosity, which is not a physical constant of fluid, as the molecular viscosity
of laminar flow is, but it is defined as a complex functional dependence on fluid flowing
through the state and position of the point being considered, ie the sharing of
momentum fluctuations and distance from the wall. Therefore, the velocity profile for
turbulent flow in comparison with laminar is more flat (no parabolic character).
4.2.
Turbulent flow
Flow is generally called turbulent if its variables exhibit chaotic fluctuations in both space
and time, see Fig. 4.3 .
55
with regard to physics is still not resolved. Although currently the significant progress was done,
particularly in the area of nonlinear dynamical systems and chaos theory, a complete solution
of turbulence is not expected in the near future. However, interest in turbulence is not only
inspired by the desire to understand its essence, but a necessity to predict turbulent flow in
many industrial applications. Despite the randomness of turbulence the detailed studies
indicate that turbulent flow consists of the spatial structures that are usually called "eddies"
see Fig. 4.4.
56
Re l
u.l
( 4.2.1)
l2
Rel
( 4.2.2)
u.l l T
l l Tt
u
where Tt denotes the time scale of turbulent eddies transfer of macroscale l and T denotes
the time scale of molecular diffusion. The flow can be characterized based on the value of
Reynolds number as follows [6], where
T Tt tj. Rel 1
eddies vanish.
T Tt tj. Rel 1
relatively small values of the flow parameters so you can make a conclusion that most of
the flow is turbulent.
T Tt tj. Rel 1 fully developed turbulence, this means that viscous matter, which
affect the time scale T can be neglected due to the dynamics of vortices, which appear
above the value Tt . Turbulent vortices in a fully developed turbulent flow are almost
inviscid.
T Tt , tj. Rel 1 transition state, laminar steady flow, which turns into turbulent,
unsteady, if exceeded the critical Reynolds number Rel. Flow initially becomes periodic.
This qualitative change in the behavior of the flow is called a bifurcation. When increasing
the Reynolds number further instability becomes to the flow and flow is creating fully
turbulent.
An important consequence of Reynolds similarity is dissipation, respectively dissipation
rate at which the turbulent eddies lose their kinetic energy and change it into heat. Dissipation
(to unit mass) [m2s-3] is defined as
57
u3
l
( 4.2.3)
The result of dimensional analysis is length microscale of these dissipation areas, which is
called the Kolmogorov scale
1
3 4
( 4.2.4)
Re l
3
4
Rel 4 .l
( 4.2.5)
12
10
1
10
generation
l , [m]
0.1
makromtko
0.01
mikromtko
transition
0.001
l , [m]
0.0001
disipation
0.00001
0.000001
1.E+02
1.E+03
1.E+04
1.E+05
1.E+06
1.E+07
Re [1]
0
1.00E+03
1.00E+04
1.00E+05
Re [1]
1.00E+06
1.00E+07
Fig. 4.5 Microscale and macroscale of turbulent vortices in a decimal logarithmic coordinates
Turbulent flow consists of the continuous vortex structures whose length scale (size) is situated
between l and , see Figure 4.5. There are two types of vortex structures near the wall structure
(hairpin, bursts, streaks) and the structure in the middle of the stream.
58
Fig. 4.6 Turbulent eddies if the fluid flow, their splitting, conversion into heat
MODELS OF TURBULENCE
Space Filtering
LES
DES
Time Averaging
Models of
turbulent
viscosity
Reynolds
stress
model - RSM
Direct Simulation
DNS
Turbulence modeling is still in research stage, which is constantly changing with the
progress in the mathematical, physical and technical sectors. In the numerical simulation of
turbulent flow, there are three different theoretical approaches that result from modifying the
default simplification of equations describing the flow ([4]).
59
u
u
DNS
LES
RANS
60
LES - Large Eddy Simulation is based on modeling of large eddies, such as timedependent spatial formations that can be capture by grid. Eddies about the turbulent small
scales little involve in the transport phenomena, but their means is in the turbulent kinetic
energy dissipation due to viscosity to heat. These small vortices are parameterized by so called
subgrid models and removed by filtration turbulent field. Select the filter bandwidth, usually
corresponding dimension of grid cells, it is possible to achieve the number of grid cells, which
can be with current computer technology to solve.
For most engineering problems of turbulent flow the most widely used tool of statistical
turbulence models is remained. Models are based on the method of time averaging (RANS
Reynolds-Averaged Navier-Stokes equations) of turbulent flow quantities and the following
time-averaging procedure of balance equations remains.
where
T
1
d
T0
resp.
1
i
N i
, , 0, ,
. ,
where
( 4.2.6)
After substituting sum of time-averaging and fluctuation into the continuity equation it is given:
61
u j uj
0
xj
( 4.2.7)
u j uj
0
xj xj
After time averaging it is
u j uj
0
xj xj
( 4.2.8)
uj
0
xj
( 4.2.9)
Continuity equation for the fluctuation component is given by subtracting equation from the
equation :
uj
0
xj
( 4.2.10)
Similarly, you can put into the Navier - Stokes equations provided that fi f i , :
1 p p
2 ui ui
ui ui ui ui u j uj
fi
t
xj
xi
x 2j
( 4.2.11)
The result is the Reynolds equation formally similar to Navier-Stokes equation for averaging
the values that it contains also a member on the left:
1p
ui uiu j
2u
uiuj
. 2i fi
t
xj xj
xi
xj
( 4.2.12)
1 p
ui uiuj uiu j uiuj
2 ui
fi
t
x j
x j
xj
xi
x 2j
( 4.2.13)
uiuj are so called Reynolds (turbulent) stresses, which exist only in case of turbulent flow.
Turbulent vortex can mix the fluid of different velocities in cubic volume, see a).
62
a)
b)
u1u3
c)
u 3 u1
e)
d)
u1u3 . Even if considering only one cube surface, the fluid moving in one of the three
coordinate directions can be mixed and the the result are deformation u1u1, u1u2 , u1u3 . Next,
suppose that the velocity u1 is by turbulence mixed on the upper area of the original cube, Fig.
4.10 c), with an estimated speed of u3. The resulting deformation is related to Reynolds stress,
Fig. 4.10 d). Deformation is identical, just has a different orientation. In Fig. 4.10 e) there are
the two deformation regardless of the rotation. In conclusion we can say that u u u u .
3 1
1 3
The same number of combinations will be generated for the areas perpendicular to the other
two coordinate directions. Total number is nine components of the Reynolds stress, similar to
the viscous stress, and true symmetry.
63
u1u1 u1u2
u2u1 u2u2
u3u1 u3u2
u1u3
u2u3
u3u3
Turbulent flow of momentum acts as the stress and thus is called the Reynolds stress, for
formally is replaced to j (j occurs in the variable
uj
ui
which the transport equation can be derived. Comes from where the summation index k
).
t
xk
xk
xk
xi
xk2
( 4.2.14)
This equation is multiplied by u j , then the equation is writing for component j and multiplying
by u i . Equations thus obtained are added together, time averaged to get the transport
equations for Reynols stresses:
uiuj
uu
uk i j
t
xk
xk
p
uiuj uk kj ui ik uj x uiuj
k
difzn transport
u u
u p u u
uiuk j uj uk i i j 2 i j
xk
xk x j xi
xk xk
produkce
( 4.2.15)
disipace
redistribuce
Reynolds stress tensor consists of nine members, while six are independent, so it is the six
equations, which form an extensive system of differential equations solvable difficult.
Therefore, attention will be paid to theories dealing with the simpler expression of the Reynolds
stress in equation (the turbulence models).
t l.u
( 4.2.16)
The task of the various turbulence models is to express the turbulent stress, heat flows or other
scalar variables using selected scale to determine the distribution of this parameter in the flow
field. Most of the models makes use of Boussinesq's hypothesis of eddy (turbulent) viscosity.
This hypothesis assumes that, as in laminar flow when Newtonian shear stress relationship is
applied in simplified two-dimensional flow, the turbulent stress and turbulent flows are
proportional to the mean velocity gradient, temperature, concentration, etc., ie.
64
laminar flow
Boussinesq's
turbulent flow
molecule viscosity
hypothesis (analogy)
du
dy
t uv t
u
y
Generally
u uj 2
uj 2
k ij resp. uiuj t ui
k ij
uiuj t i
x
3
x
3
x
x
i
i
j
j
( 4.2.17)
1
k uj uj
2
( 4.2.18)
In great detail k
1
2
u u 2 u
1
j 1..3
viscosity is not the physical property of fluid, but the flow. It is strongly dependent on the level
of turbulence and may vary significantly within the flow field. The equation of motion shall
be adjusted using Boussinesq's hypothesis as follows:
1p
ui uiu j
2ui
2ui
fi
t
t
xj
xi
x 2j
x 2j
( 4.2.19)
The equation for heat transfer or other scalar is adjusted in a manner consistent with the
derivation of the equation for momentum transfer for averaging values. After adjusting there
is appeared in these equations member ui representing the turbulent flow of heat or other
scalar variables and like is expressed
ui t
xj
( 4.2.20)
where t is turbulent diffusivity. Most models assume that the turbulent diffusivity is
proportional to the turbulent kinematic viscosity of the relation t
t , where is turbulent
t
t
65
Direct method
DNS
Matematical
models of
flow
Direct method
Turbulent
flow
Method of large
eddies - LES
Method of time
averaging RANS
Metod of Reynolds
stresses
Boussinesqs
hypothesis
Zero - equation
model
k-
Boussinesqs
One - equation
model
hypothesis
RNG k-
Two - equation
model
k- available
k-
66
5.1.
The first model describing the distribution of turbulent viscosity t , Prandtl suggested.
Turbulent viscosity is expressed according to the mean velocity again in a simplified twodimensional form of the relationship
t t lm 2
u
y
( 5.1.1)
where lm is the mixing length defined by empirical relation l m y , where is von Krmn
constatnt (0.41). This model is suitable for modeling flow in the shear layer. The drawback is
that it assumes a local equilibrium, ie the production of turbulent kinetic energy (TKE) is equal
to the rate of dissipation of TKE. It does not affect the transport of turbulence.
5.2.
In order to affect turbulent transport parameters the differential transport equation for these
parameters must be solved. The simplest models use a transport equation for the velocity
scale of turbulent motion
1 2
1 2
2
2
k , kde k 2 u1 u2 u3 2 uj is kinetic (averaging) energy
of turbulent motion per unit weight. For k an exact equation can be derived from the Navier Stokes equations in the form:
k ujk
ulul
p
2k
u
u ul
jl 2 uluj l l
u j
t xj
xj 2
x
xj
xj xj
j
I
II
III
IV
( 5.2.1)
VI
On the right side, there are members representing diffusion due to turbulent fluctuations of
due to interaction of Reynolds stress and mean
velocity gradient and dissipation due to the conversion of energy to heat. When operating
in the area also Archimedes force, on the right side the member corresponding to production
(destruction) of kinetic energy due to buoyant forces is also placed. In the equation thus derived
the unknown correlations in diffuse and dissipation member are found. To obtain a closed
system of equations, those members must be modeled by:
67
uu p k
uu
k3/ 2
, l 2 l CD
uj l l t
k xj
x j
l
2
( 5.2.2)
where k a CD are empirical constants. After substituting for these members to the equation
for k and using the relations and for t and t has the equation for k form:
u j ul ul
k u jk
t k
k3/ 2
.
CD
t x j x j k x j t xl x j x j
l
( 5.2.3)
t C k l
where C is empirical constant. Length scale l characterizing the turbulent motion is defined
by means of empirical relations similar to the mixing length model. One- equation models affect
transport of turbulence and are suitable mainly in cases where it is realistic to describe the
distribution of length scale l , but are not suitable for modeling of complex flow cases, when
you can not define with sufficient accuracy its distribution by the empirical relationship. Here it
is necessary to define additional transport equation and switch to two - equation turbulence
model.
5.3.
Two - equation k- model determined turbulent viscosity by two transport equations for k
and . The model uses the Boussinesq 's hypothesis of eddy viscosity and relates t to k ,
and C
t C
k2
( 5.3.1)
The distribution k is given by the transport equation. The exact form of transport equation for
can be derived again from the Navier - Stokes equations, this equation contains complex
correlations, which again must be approximated. The resulting form of dissipation rate equation
used in the model is presented in this form:
u u u
k u jk
t k
k 3/ 2
t j l l CD
.
x x x
t x j x j k x j
l
l
j
j
( 5.3.2)
u j ul ul
u j
t
2
.
C
C2
k
t x j x j x j 1 t xl x j x j
( 5.3.3)
68
5.4.
RNG k- model
This model is derived from the classical k model when using a mathematical procedure
called renormalization group method (RNG). Renomalization procedure applied to the
turbulence lies in the gradual elimination of small vortices, while the equations of motion
(Navier - Stokes equations) are transform by modifying turbulent viscosity, forces and
nonlinear members. Where it is provided that these eddies are connected with dissipation ,
the turbulent viscosity t or t
t
is dependent on the scale of turbulent eddies and the
RNG viscosity method constructs this viscosity using iterative elimination of narrow band wave
numbers. In detail, this method is given in [13] with that for the iterative process uses the
session
deff
A1l 3
,
dl
(l )2
where
e
l
u
c
e
l
o
m
t
n
e
l
u
b
r
u
t
e
v
i
t
k
e
f
f
e
( 5.4.1)
By integrating this equation over the length scale l for the initial condition mol and for the
scale l l L / Re 3 / 4 , which is the Kolmogorov dissipation scale corresponding to small
d
turbulent eddies, it gets
eff l mol 3 1
3 A1 4 4
l ld
3
4mol
l ld
( 5.4.2)
This equation is the interpolation formula to calculate eff l between viscosity and the
molecular viscosity of dissipation vortices with the limits l l d corresponding to high Reynolds
numbers. For high Reynolds number can be proved that the equation has the form
eff t 0.094l 2 u
( 5.4.3)
RNG method is about one-tenth slower than two - equation classic model, but in areas of
vortices (where the liquid is slow and there is a lower Re number) it is more accurate.
5.5.
Reynolds stress model involves the calculation of the individual Reynolds stress
through a differential transport equations in the form (see )
69
uiuj
uu
p
uiuj
uk i j
uiuj uk kj ui ik uj
x
t
xk
x
k
k
difzn transport
u u
u p u u
uiuk j uj uk i i j 2 i j
xk xk
xk
xk x j xi
produkce
( 5.5.1)
disipace
redistribuce
In order closure equations Fluent approximates some members of equation , see [13]. The
calculated Reynolds stresses are then substituting into the equation for momentum transfer .
Fluent then solves:
three transport equations for averaging velocity components and the continuity
equation
transport equation for dissipation
transport equation for turbulent energy near the wall in the case of explicit boundary
conditions for the voltage as the sum of equations
5.6.
Modeling of the flow near the wall affects the accuracy of numerical solutions in the
whole area. Near the wall the solved variables are rapidly changing, significantly the transfer
of momentum and scalar quantities is here applied. Turbulence is suppressed close to the wall,
in the outer part of the boundary layer, however, a significant production of turbulent kinetic
energy due to Reynolds stresses and mean velocity gradient. Numerous experiments have
shown that the area near the wall, so called boundary layer can be divided into multiple parts.
Immediately near the wall the viscous (laminar) sublayer is placed, the flow is here nearly
laminar and molecular viscosity has a dominant influence on the transfer of momentum, heat
and mass. The outer part of the boundary layer becomes fully turbulent layer and turbulence
plays a dominant role here. between laminar and fully turbulent layer the transition layer
occurs, in which apply equally the effects of molecular viscosity and turbulence. Distribution of
the boundary layer is shown in Fig 5.2
70
8
8
u-exp
u-lam
u-turb
7
6
experiment
5
4
u [m/s]
vnj vrstva
1
0
0.0
0.5
y [m]
u=(u*/). ln((y+yo)/y
)
1.0
u=u*(y+yo)/y
)
2
1
viskzn podvrstva
pechodov vrstva
i t
0.010
0
0.001
vnj vrstva
0.100
1.000
y [m]
Fig. 5.2 Distribution of area near the wall - in linear and logarithmic coordinates 0
ln E. y
( 5.6.1)
u
u
u y
( 5.6.2)
where
= friction velocity
Friction velocity u is determined by shear stress defined as Reynolds stress (variables have
an index s ).
71
u2
w
2
2
uvs wvs
( 5.6.3)
where the total vertical momentum fluxes measured near the wall are determined by relations
Reynolds stresses are calculated using the Boussinesq 's approximation. For
turbulence
models in case of local equilibrium of turbulent energy production and dissipation rate are
defined by the relation
uv
u
u
w
y
y
( 5.6.4)
and can be derived in the logarithmic layer the equation for turbulent kinetic energy
u2
kp
C
( 5.6.5)
and dissipation
u3
p
y
( 5.6.6)
In the equations to the friction (shear) velocity appears as a constant quantity. The standard
wall functions are also defined in Fluent and are suitable for many applications. If the abovedefined functions are used for more general flow, shear rate may just cause problems. In
addition, to quantify the turbulent kinetic energy and dissipation the local equilibrium is
assumed, which is a very strong restriction in the flow of a complex character.
Wall function
Two-layer modeling
Standart wall
Nonequilibrium
function
wall funkction
72
use of wall functions by which the region between laminar sublayer and transition
layer is spanning, where the molecular and turbulent viscosity is appears, ie the area
between the wall and the area of fully developed turbulent flow
near-wall modeling includes the viscous sublayer in connection with the delicacy of
the grid.
full turbulence
fluid flow
wall
use logarithmic
wall function
near wall
modeling
Fig. 5.3 Access to the modeling of flow near the wall in Fluent.
Wall functions are a set of empirical relationships and functions that allow you to "link" variable
solved in a cell near the wall with the corresponding value on the wall. Wall functions include
logarithmic law for mean velocity and temperature and equations for turbulent quantities near
the wall.
FLUENT offers two basic variants of wall functions:
In the case of flow with high Reynolds number wall function implementation significantly
reduces the calculation and provides economical and sufficiently accurate solution for most
engineering problems. This approach is inappropriate in the case flow with low Reynolds
number. In these cases it is necessary to choose the second approach, which allows modeling
in near wall region affected by fluid flow viscosity and in FLUENT is defined as
73
When using finite volumes the general wall functions suitable for complex flow were
derived and are given by relationships
u*
ln E. y*
( 5.6.7)
where
u
*
uP k P1 / 2C1 / 4
k P1 / 2C1 / 4 yP
y
( 5.6.8)
*
This logarithmic formula for average velocity is valid for y 30 60 . In Fluent the logarithmic
*
law is recommended for y 11.225 . When the grid density near the wall is such that
y * 11.225 in the cells adjacent to the wall, then applies the laminar relationship, ie
1
*
ln E. y
u
y*
( 5.6.9)
*
u
where
the intersection of linear and logarithmic profile, and while some are
k 1p/ 2C1 / 4 yv
y
11.23
*
v
( 5.6.10)
Reynolds analogy between momentum and energy transfer allows you to define a
similar logarithmic law for mean temperature. The wall law for temperature includes two
different temperature dependence:
logarithmic law for the turbulent region, where turbulence effects are greater than the
conductivity
The thickness of thermally conductive sublayer is different from the viscous sublayer
thickness and depends on the type of fluid. The wall law for the mean temperature is defined
as follows:
T
T
Pr . y y yT
Tp c pC1 / 4 k 1p/ 2
1
q
Prt . ln Ey P y yT
74
( 5.6.11)
/4 A
P
sin / 4
1/ 2
Pr
Pr
1 t
Prt
Pr
1/ 4
( 5.6.12)
and where
Tp
Tw
Pr
Prt
= heat flux
logarithmic law for the averaging velocity by Launder and Spalding will be specified
depending on the effect of pressure gradient,
balance of turbulent kinetic energy and dissipation in the cell adjacent to the wall is
calculated in two layers, ie, laminar and turbulent.
Wall function for the mean temperature does not differ from a standard wall functions for
temperature. Logarithmic mean velocity profile is calculated using the pressure gradient
according to the relationship
u~ 1
lnEy p
u
( 5.6.13)
75
y y yv yv2
1 dp yv
u~ u
ln
2 dx k 1 / 2 yv k 1 / 2
( 5.6.14)
yv
y*
( 5.6.15)
C1 / 4k 1p/ 2
flow with low Reynolds number, the great influence of wall (eg, flow in narrow gap, the
flow of a very viscous fluid, flow with low velocity)
strong
pressure
gradient
leading
to
boundary
layer
separation
significant effect of volume forces (eg centrifugal forces, the flow near a rotating disk,
Archimedes force)
three-dimensional flow near the wall (Ekman spiral, strongly curved 3D boundary
layer).
If the character of the flow corresponds to one of the above cases, and if it is necessary to
include these phenomena in solving the problem, then we proceed to the detailed modeling
of flow near the wall (near-wall modeling). For this purpose Fluent offers two-layer model.
Re y
kn
( 5.6.16)
is the normal distance middle of the cell from the wall, in FLUENT it is interpreted as
where
76
(5.6.17)
where
rw w
rw
n min
r rw
is the
unification of all the walls at the border. This interpretation allows a uniform definition in a area
of complex shape, which includes many walls.
In the fully turbulent region (Rey 200) the turbulent models described in Chapter 4.1 - 4.3.
are used. In the area affected by the viscosity near the wall (Rey 200), Wolfstein oneequation model is applied, where the equation for momentum transfer and the equation for k
is defined in a known manner, turbulent viscosity is calculated from the relationship
t C k 1 / 2l
(5.6.18)
k 3/ 2
l
(5.6.19)
Re
l Cl y 1 exp y
(5.6.20)
Re y
l Cl y 1 exp
A
(5.6.21)
If in the whole area the Rey 200, then is not solved using the transport equation, but from
the algebraic equation (4.4.22 ). The constants A, A, occurring in these relationships are
defined according to the authors of Chen and Patel:
(5.6.22)
A = 70, A =2 cl
When using the RSM model, then the standard and non-equilibrium functions can be used as
wall functions, two-layer wall function model is not appropriate.
5.6.6. The influence of grid quality on the choice of wall functions for
various models of turbulence
Distance between cells adjacent to the wall from the wall will determine if the correct
approach of logarithmic wall functions is valid or if it is necessary to choose another one
*
logarithmic law is valid for y 30 60
*
two-layer prescription is valid for y 4 5 , ideally at least 10 cells should be in the
laminar sublayer
Spalart Allmaras model uses the-wall logarithmic function assuming a very fine grid (
Large Eddy Simulation model uses a logarithmic wall function for very fine grid (oder
77
*
of y 1 )
up
u y
ln E * u*B
u*
( 5.6.23)
where u p is the averaging velocity at the point P closest to the wall and y is the distance of
the point P from the wall. Function B depends on the type and size of roughness and it is
computed by B
ln 1 C Ks K s , where K s
K s u*
is a dimensionless roughness and K s
is a actual physical roughness in meters, CKs 0.5 for a regular roughness, higher values
90
80
70
z[m]
60
50
40
30
20
10
0
-4
-2
U [m/s]
Fig. 5.4 Graphs of velocity influenced by various aerodynamic roughnesses by the flow in
atmosphere
78
6.1.
The equations can be derived by the procedure described in Chap. 0 and 0, and their
shape is as follows:
continuity equation valid for time averaging variables
u j
0
t
xj
( 6.1.1)
u i u i u j
p
t u i
t
xj
xi x j
xj
( 6.1.2)
i 3 g f c ij 3 u j f i
gravity force
Coriolis force
In case of two-equation k- model the equation are supplemented by the equation for turbulent
kinetic energy transfer k and dissipation rate .
u j ul ul
k u j k t k
.
gj t
x x x
t
xj
x j k x j
h x j
j
j
l
( 6.1.3)
u j t
2
.
c1 P c3 G c2
t
xj
x j x j
k
( 6.1.4)
where P and G represent the production of turbulent kinetic energy due to tension and buoyant
forces
u j ul u j
, G g j t
P t
h x j
xl x j xl
( 6.1.5)
where C1=1,44 , C2=1,92, C3=1, k = 1, =1.3 are constants defined empirically and
t
c is Prandtl turbulent number.
t p
uiuj t
ui
xj
( 6.1.6)
79
Kolmogorov-Prandtl hypothesis:
t l.v C
6.2.
k2
( 6.1.7)
u1/ 2 u2/ 2 u3/ 2 ) the relative turbulence intensity is explained as the ratio of rms
fluctuation velocity component to the mean velocity in the same pointof the flow usually
expressed as a percentage. Usually there is measured only one directional component:
u1/ 2
I
u1
( 6.2.1)
uj uj
3uu
( 6.2.2)
/
The difference between the velocity fluctuations in the transverse direction u2 and u3/ is
windtunnel
80
1-5%
wake
2-10%
5-20%
in the tube
drowned stream
20%
100%
Turbulent scale l is limited by the size of the area, because turbulent vortices can not
be larger than the dimension of the area. Approximate value of the turbulent scale is
determined from the relation l 0.07 L , where L is the characteristic size or hydraulic diameter.
Turbulence intensity and hydraulic diameter are available variables that can be specified as
boundary conditions, the others are then converted by the following relationships
turbulent intensity
uj uj
3uu
l 0.07 L
turbulent scale
rate of turbulent viscosity
dissipation rate
3
u.I .l
2
3
u2 nebo k 3 uI 2
2
2
3
k2
k2
C
C t
l
3
4
Of course, turbulent energy and dissipation velocity can also be defined directly. Depending
on the complexity of the mathematical model other variables associated with heat transfer or
other scalar variables can be defined. The value of turbulent intensity in the case of LES is
defined by a random velocity fluctuations at the input.
ptot pstat
1
2
u
2
( 6.2.3)
81
1
1
ptot pstat 1
Ma 2
2
where ptot
pstat
( 6.2.4)
total pressure
static pressure
u
u
c rTs 0.5
Ma
Mach number Ma
R
, M is molecular weitght
M
cV
a
M
The difference in the evaluation of the total pressure in the flow of air into the Mach number
0 1 is shown in
0.14
1200
0.12
p dyn
p stat
p celk
p tot stlac
Ma
p [Pa]
800
0.1
0.08
600
0.06
400
1000
0.04
200
0.02
0
0
10
15
20
25
30
35
40
-1
u [ms ]
p ref gxi p
( 6.2.5)
82
at reference
temperature. Input values of total and static pressure are added to the hydrostatic pressure.
When you specify pressure conditions, it is necessary to define flow direction using
velocity components or by flow rate in the normal direction to the boundary.
Static pressure at the inlet must be specified in the case of supersonic flow. Turbulent
quantities are determined identically as in the case of mass flow boundary conditions.
6.2.5. Outflow
Outflow is not suitable for compressible flow, unsteady incompressible flow with varying
density and in case of , if specified pressure at the inlet.
83