Documentos de Académico
Documentos de Profesional
Documentos de Cultura
User Manual
contains:
NUMECA International
5, Avenue Franklin Roosevelt
1050 Brussels
Belgium
FINE™ iii
Contents
iv FINE™
Contents
Introduction 5-15
Default Mixing Plane Approach 5-16
Full Non-matching Technique for Mixing Planes 5-19
Domain Scaling Method 5-21
FINE™ v
Contents
vi FINE™
Contents
FINE™ vii
Contents
viii FINE™
Contents
FINE™ ix
Contents
x FINE™
CHAPTER 1: Getting Started
1-1 Overview
Welcome to the FINE™ User’s Guide, a presentation of NUMECA’s Flow INtegrated Environ-
ment for computations on structured meshes. This chapter presents the basic concepts of FINE™
and shows how to get started with the program by describing:
• what is CFD,
• what FINE™ does and how it operates,
• how to use this guide,
• how to start the FINE™ interface.
1-2 Introduction
1-2.1 What is CFD?
All the mathematical sciences are founded on relations between physical laws and laws of numbers,
so that the aim of exact science is to reduce the problems of nature to the determination of quanti-
ties by operations with numbers.
James Clerk Maxwell, 1856
In the late 1970’s, the use of supercomputers to solve aerodynamic problems began to pay off. One
early success was the experimental NASA aircraft called HiMAT, designed to test concepts of high
manoeuvrability for the next generation of fighter planes. Wind tunnel tests of a preliminary design
for HiMAT showed that it would have unacceptable drag at speeds near the speed of sound; if built
that way the plane would be unable to provide any useful data. The cost of redesigning it in further
wind tunnel tests would have been around $150,000 and would have unacceptably delayed the
project. Instead, the wing was redesigned by a computer at a cost of $60,000.
Paul E. Ceruzzi, 1989
Computational Fluid Dynamics (CFD) is commonly accepted as referring to the broad topic encom-
passing the numerical solution, by computational methods, of the governing equations that describe
FINE™ 1-1
Getting Started Introduction
fluid flow related conservative equations (for instance involving energy or species concentration). It
has grown from a mathematical curiosity few decades ago, to become an essential tool in almost
every branch of fluid dynamics, from aerospace propulsion to weather prediction.
Key elements of CFD are briefly described below. Please notice these informations have mostly
been extracted from the web at http://www.crankfield.ac.uk/sme/cfd/whatiscfd.htm and http://
ltp.larc.nasa.gov/aero/login.htm. The reader is invited to consult the web at www.cfd-online.com
for more insight in CFD theory.
By 1900, theoretical and mathematical developments in fluid dynamics had not achieved signifi-
cant progress as quickly as experimental efforts. Although the Navier-Stokes equations describing
fluid behaviour had existed for over 75 years before the Wright brothers' first flight, they were sim-
ply too complex to solve in their complete form for anything but simplified problems under specific
conditions. Even today, theoretical solutions to the Navier-Stokes equations are rare and only suited
to specific classes of problems.
As a developing science, CFD has received extensive attention throughout the international com-
munity since the advent of the digital computer. CFD interests are mainly driven by the desire to
model complex physical fluid phenomena, that couldn't be easily or cost effectively simulated with
a physical experiment.
Computational techniques differ from analytical or theoretical solutions in the sense that they only
solve equations at a finite number of points rather than for the entire flow field. Choosing these
points may become quite difficult - especially for a complex geometry and may require hundreds of
thousands or even million of points. In general, a dense grid with many points will give a solution
of great detail, but require more computer memory and time to reach a solution. Since this trade-off
in computer resources and solution quality is required, the current trend is often to use a dense grid
in areas where the solution may change rapidly such as in the boundary layer or near a shock wave,
but use of a coarser grid with fewer computational points in areas where the solution is expected to
change more gradually.
The selection of grid nodes is the science of grid generation. It is a complex field on its own and has
many applications outside CFD, such as in constructing solid models for stress, vibration and heat
transfer analysis. There are several different techniques that are commonly used to develop compu-
tational grids for CFD.
Grid nodes can be arranged either in a regular or irregular pattern. Depending on the pattern
selected, the method of generation is referred to as structured or unstructured. Both techniques have
their own advantages and disadvantages.
Basically, structured methods enable the generation of regular grids, with high quality standards
and good control in the distribution of grid nodes in shear and boundary layers. However, they may
sometimes be hard to generate for very complex geometries. On the contrary, unstructured methods
allow a limited control in the grid quality (and thus in the grid solution) but the grid generation
process does usually require limited user resources, even for complex geometries.
It is thus important that the user recognizes these so that the best grid can be created for a particular
geometry. The quality of the grid is definitely important, since it strongly influences the solution -
including whether or not a solution can be found at all.
FINE™/Turbo and FINE™/Design3D environments developed by NUMECA rely on structured
grid methods, based on AutoGrid and IGG™ software tools. NUMECA offer also includes unstruc-
tured capabilities, through HEXPRESS™ hexahedral grid generator and FINE™/HEXA environ-
ment. More details can be provided upon request.
1-2 FINE™
Introduction Getting Started
Usually, about 4 to 5 unknowns are computed at each grid node. This makes methods of presenting
CFD results a difficult matter, mainly due to the large number of data to manage. Adequate post-
processing is then required to easily and quickly outline the major characteristics of the flow. This
is usually done through surface color techniques and particle paths/ribbons. The latter method is
mostly used to depict vortices and complex flow-wall interactions. The user is invited to refer to
CFView™ manual for a detailed information on NUMECA offer in that area.
1-2.2 Components
The resolution of Computational Fluid Dynamics (CFD) problems involves three main steps:
• spatial discretization of the flow domain,
• flow computation,
• visualization of the results.
To perform these steps NUMECA has developed three software systems. The first one, IGG™, is
an Interactive Geometry modeller and Grid generation system for multiblock structured grids. The
second software system, the flow solver EURANUS, is a state of the art 3D multiblock flow solver
able to simulate Euler or Navier-Stokes (laminar or turbulent) flows. The third one, CFView™, is a
highly interactive Computational Field Visualization system.
These three software systems have been integrated in a unique and user friendly Graphical User
Interface (GUI), called FINE™, allowing the achievement of complete simulations of 3D internal
and external flows from the grid generation to the visualization, without any file manipulation,
through the concept of project. Moreover, multi-tasking capabilities are incorporated, allowing the
simultaneous treatment of multiple projects.
1-2.3 Multi-tasking
FINE™ has the particularity of integrating the concept of multi-tasking. This means that the user
can manage a complete project in FINE™ interface; making the grid using IGG™, running the
computation with EURANUS and visualizing the results with CFView™. Furthermore, the user has
the possibility to start, stop and control multiple computations. Please note that the flow simulation
can be time consuming, therefore the possibility of running computations in background has been
implemented. See chapter 13 for more detail on how to manage multiple tasks through the interface
or in background.
FINE™ 1-3
Getting Started Introduction
— a .cgns file that contains the solution and is used for restarting the solver,
— a .res file used by the Monitor to visualize the residual history (see Chapter 15),
— two files used to visualize the convergence history in the Steering with extensions .steering
and .steering.binary (see Chapter 15).
— two files with extensions .mf and .wall that contain global solution parameters.
— two files with extensions .std and .log that contain information on the flow computation
process.
— a .batch file used to launch the computation in batch (see Chapter 15).
• The CFView™ visualization files: In addition to the .run file, the flow solver creates a series
of files, which can be read by CFView™. These files have different extensions. For example in
case of turbomachinery flow problem, the solver will create a file for the azimuthal averaged
results with extension .me.cfv.
Through the interface, the user can modify all the information stored in the files associated to the
project.
When creating a new project a new directory is made, e.g.; \project. In this directory the project file
is stored: \project\project.iec and a directory is created called \project\_mesh. In this directory the
grid files used for the computations can be stored. It is however also possible to select a grid that is
located in another directory.
Only one mesh file should be used for all computations in a project. If computations
need to be done on another mesh file it is advised to duplicate the project (see section 2-
3.1.4) or to create a new project (see section 2-3.1.1) for those computations.
batch
1-4 FINE™
How To Use This Manual Getting Started
At first time use of FINE™ it is recommended to read this first chapter carefully and certainly sec-
tion 1-4 to section 1-6. Chapter 2 gives a general overview of the FINE™ interface and the way to
manage a project. For every computation the input parameters can be defined as described in the
chapters 3 to 13. Chapter 14 gives an overview of how to run computations using the Task Manager
or using a script. Chapter 15 finally describes the available tools to monitor the progress on a com-
putation.
The expert user finds in chapter 3 to 13 a section describing advanced options that are available in
expert user mode. Additionally Annex C provides a list with all supported expert parameters on the
page Computation Steering/Control Variables in expert user mode. For each parameter a refer-
ence is given to the section in the manual where it is described.
The use of non-supported parameters is at own risk and will not guarantee correct
results.
1-3.2 Conventions
Some conventions are used to ease information access throughout this guide:
• Commands to type in are in italics.
• Keys to press are in italics and surrounded by <> (e.g.: press <Ctrl>).
• Names of menu or sub-menu items are in bold.
• Names of buttons that appear in dialog boxes are in italics.
• Numbered sentences are steps to follow to complete a task. Sentences that follow a step and are
preceded with a dot (•) are substeps; they describe in detail how to accomplish the step.
A light bulb in the margin indicates a section with a description of expert parameters.
FINE™ 1-5
Getting Started First Time Use
1-6 FINE™
How to start FINE™ Interface Getting Started
b) Background Color
The background color of the Graphics area can be changed by setting the environment variable
NI_IGG_REVERSEVIDEO on UNIX/LINUX platforms or IGG_REVERSEVIDEO on Windows
platforms. Set the variable to ’ON’ to have a black background and set it to ’OFF’ to have a white
background. The variable can be manually specified through the following commands:
On UNIX:
in csh, tcsh or bash shell:
setenv NI_IGG_REVERSEVIDEO ON
in korn shell:
NI_IGG_REVERSEVIDEO=ON
export NI_IGG_REVERSEVIDEO
The selection will take effect at the next session.
On Windows:
• Log in as Administrator.
• Launch System Properties from the Start/Settings/Control Panel/System menu.
• Go in the Environment Variables.
• Modify or add the IGG_REVERSEVIDEO entry to either ON or OFF.
The selection will take effect at the next session.
When multiple versions of FINE™ are installed the installation note should be con-
sulted for advice on how to start FINE™ in a multi-version environment.
On Windows click on the FINE™ icon in Start/Programs/NUMECA software/fine#. Alterna-
tively FINE™ can be launched from a dos shell by typing:
<NUMECA_INSTALLATION_DIRECTORY>\fine#\bin\fine <Enter>
where NUMECA_INSTALLATION_DIRECTORY is the directory indicated in section 1-4.1 and #
is the number corresponding to the version to be used.
FINE™ 1-7
Getting Started Required Licenses
1-8 FINE™
CHAPTER 2: User Interface
2-1 Overview
When launching FINE™ as described in Chapter 1 the interface appears in its default layout as
shown in Figure 2.1.0-1. An overview of the complete layout of the FINE™ interface is shown on
the next page in Figure 2.1.0-2. In the next sections the items in this interface are described in more
detail.
Together with the FINE™ interface a Project Selection window is opened, which allows to create a
new project or to open an existing project. See section 2-2 for a description of this window.
A File Chooser window is available for browsing through the file system and to select a file. More
detail on the File Chooser window is given in section 2-11.
To define a profile through the FINE™ interface a Profile Manager is included. The Profile Man-
ager is described in more detail in section 2-12.
FINE™ 2-1
User Interface Project Selection
Parameters
button
(Section 2-6)
View button
(Section 2-7)
2-2 FINE™
Project Selection User Interface
5. A new directory is automatically created with the chosen name as illustrated in Figure 2.3.1-5
indicated by point (1). All the files related to the project are stored in this new directory. The
most important of them is the project file with the extension .iec, which contains all the project
settings (Figure 2.3.1-5, no. 2). Inside the project directory FINE™ creates automatically a sub-
directory "_mesh" (Figure 2.3.1-5, no. 3).
6. A Grid File Selection window (Figure 2.2.1-3) appears to assign a grid to the new project. There
are three possibilities:
• to open an existing grid click on Open Grid File. A File Chooser allows to browse to the grid
file with extension .igg. Select the grid file and press OK to accept the selected mesh. A win-
dow appears to define the Grid Units and Project Configuration. Set the parameters in this
window as described in section 2-2.3 and click OK to accept.
• if the grid to use in the project was not yet created use Create Grid File to start IGG™. Create
a mesh in IGG™ or AutoGrid (Modules/AutoGrid) and save the mesh in the directory
"_mesh" of the new project. Click on Modules/Fine Turbo to return to the FINE™ project.
Select the created mesh with the pull down menu in the icon bar. A window will appear to
define the Grid Units and Project Configuration. Set the parameters in this window as
described in section 2-2.3 and click OK to accept.
• When using the AutoBlade™ or Design 3D Module it is not necessary to select a grid. In that
case close the Grid File Selection window and select Modules/AutoBlade or Modules/
Design 3D menu to use these modules.
2.2 3
1.1
2.1
1.2
FINE™ 2-3
User Interface Main Menu Bar
The mesh properties are always accessible after linking the mesh to the
project through the menu Mesh/Properties...
2-4 FINE™
Main Menu Bar User Interface
Each time the user creates a new computation, a subdirectory is added (4). The output files gener-
ated by the flow solver will be written in the subdirectory of the running computation.
(1) project directory
(2)
After the new project is created a Grid File Selection window (Figure 2.2.1-3) appears to assign a
grid to the new project. There are two possibilities:
• to open an existing grid click on Open Grid File. A File Chooser allows to browse to the grid
file with extension .igg. Select the grid file and press OK to accept the selected mesh. A win-
dow appears to define the Grid Units and Project Configuration. Set the parameters in this
window as described in section 2-2.3 and click OK to accept.
• if the grid to use in the project was not yet created use Create Grid File to start IGG™. Create
a mesh in IGG™ or AutoGrid (Modules/AutoGrid) and save the mesh in the directory
"_mesh" of the new project. Click on Modules/Fine Turbo to return to the FINE™ project.
Select the created mesh with the pull down menu in the icon bar. A window will appear to
define Grid Units and Project Configuration. Set the parameters in this window as described
in section 2-2.3 and click OK to accept.
b) Selecting the File From the List of Files in the File Menu.
The File menu lists the names of the five most recent projects that have already been opened. Click
on a project name from this menu to open this project.
FINE™ 2-5
User Interface Main Menu Bar
FINE™ will check the write permission of the project directory and the project file and will issue a
warning if the project or directory is read only. In both cases the project can be modified, but the
changes will not be saved.
Only one project can be active. Opening a second project will save and close the first
one.
An attempt to open a no longer existing project will remove its name from the list in the
File menu.
If more than one computation is selected in the Computations list, then all parameters in
the Parameters area will be assigned to all selected (active) computations. This may be
dangerous if the parameters for the currently opened information page are not the same
for all the selected computations. To avoid this it is recommended to select only one
computation at a time.
The run file is automatically updated (or created if not existing) when starting a computation
through the main menu (Solver/Start...).
2-3.1.6 Preferences
When clicking on File/Preferences... the Preferences window appears that gives access to some
project specific and global settings:
a) Project Units
In this section the user can change the units system of the project. The units can be changed inde-
pendently for a selected physical quantity or for all of them at the same time. To change the units
for one specific quantity select the quantity in the list (see (1) in Figure 2.3.1-6). Then select the
unit under Select New Units (2). To change the units for all physical quantities at the same time use
the box Reset Units System To (3).
The default system proposed by NUMECA is the standard SI system except for the unit for the
rotational speed, which is [RPM].
When the user changes the units, all the numerical values corresponding to the selected physical
quantity are multiplied automatically by the appropriate conversion factor. The numerical values
2-6 FINE™
Main Menu Bar User Interface
are stored into the project file in the same units as they appear in the interface, but in the .run file,
which is used by the flow solver, the numerical values are always stored in SI units. The results can
be visualized with the flow visualization system CFView™ in the units specified in FINE™.The
change of the units can be done at any stage of the creation of the project.
User-defined units can be defined through a text file euranus.uni. Please contact NUMECA support
team at support@numeca.be for more information.
b) Global Layout
Under the Global Layout thumbnail the user may specify project independent preferences for the
layout of the interface. More specifically the user may activate or deactivate the balloon help, which
gives short explanation of the buttons when the mouse pointer is positioned over them. By default
balloon help is activated. Deactivation of the balloon help will only be effective for the current ses-
sion of FINE™. When FINE™ is closed and launched again the balloon help will always be active
by default.
2-3.1.7 Quit
Select the File/Quit menu item to quit the FINE™ integrated environment. The active project will
be saved automatically. Note that this action will not stop the running calculations. A TaskManager
window will remain open for each running computation. Closing the TaskManager window(s) will
kill the running calculation(s) after confirmation from the user.
FINE™ 2-7
User Interface Main Menu Bar
2-3.2.3 Unload
Mesh/Unload clears the mesh from the memory of the computer.
2-3.2.4 Properties
Clicking on Mesh/Properties... opens a dialog box that displays global information about the
mesh:
• Grid Units: the user specifies the grid units when importing a mesh. When importing a mesh
the Grid Units and Project Configuration window appears, which allows the user to check the
mesh units (see section 2-2.3). The chosen units can be verified and, if necessary, changed.
• Space Configuration: allows the user to verify whether the mesh is cylindrical or Cartesian and
the dimensionality of the mesh. When importing a mesh the Grid Units and Project Configu-
ration window allow the user to check these properties. Those chosen properties can be veri-
fied and, if necessary, changed.
• The total number of blocks and the total number of points are automatically detected from the
mesh.
• For each block: the current (depending on the current grid level) and maximum number of
points in each I,J and K directions, and the total number of points are automatically detected
from the mesh.
2-8 FINE™
Main Menu Bar User Interface
2-3.3.2 Save
To save an intermediate solution while the flow solver is running click on Solver/Save....
FINE™ 2-9
User Interface Icon Bar
• Solver/Kill... allows to stop immediately the computation for the active project.In this case no
output is created and the only solution left is the last one that was output during the run. This
may become a dangerous choice if asked during the output of the solution. The output is then
ruined and all the computation time is lost. To avoid this, it is better to use Solver/Suspend...
or to kill the computation some iterations after the writing operations.
The icon bar contains several buttons that provide a short cut for menu items in the main menu bar.
Also the icon bar contains a File Chooser to indicate the mesh linked to the project and a pull down
menu to choose between Standard Mode and Expert Mode.
2-10 FINE™
Icon Bar User Interface
The Open Project icon is a short cut for the menu item File/Open.
The Save Project icon is a short cut for the menu item File/Save.
The Save RUN Files icon is a short cut for the menu item File/Save Run Files.
To select a grid not only the grid file with extension .igg needs to be present but all four
files (with extensions .igg, .geom, .bcs and .cgns) need to be available in the same direc-
tory with the same name for the chosen mesh.
If a mesh has already been linked to the current project and another mesh is opened, FINE™ will
check the topological data into the project against the data into the mesh file and will open a new
dialog box (Figure 2.4.2-8) The dialog box will inform the user that a merge of the topologies has
been performed and will also show at which item the first difference was detected.
FINE™ 2-11
User Interface Computations Button
The merge may also occur when opening an existing project, or when switching between IGG™
and FINE™, because the check of the topology is performed each time when a project is read.
In case the mesh file is no longer existing, or not accessible, an appropriate warning will pop-up
when opening the project. In this case it will not be possible to start the flow solver, so the user
needs to locate the mesh by means of the File Chooser button on the right of the Grid Selection bar.
Once the grid file selected and its topology checked the Grid Units and Project Configuration win-
dow appears. This window allows to check the properties of the mesh and to modify them if neces-
sary. See section 2-2.3 for more detail on this window.
The Suspend Flow Solver icon is a short cut for the menu item Solver/Suspend....
The Kill Flow Solver icon is a short cut for the menu item Solver/Kill....
The Start CFView icon is a short cut for the menu item Modules/CFView.
2-12 FINE™
Parameters Button User Interface
When selecting multiple computations, the parameters of the active page in the Parame-
ters area are copied from the first selected computation to all other computations.
With the buttons New and Remove computations can be created or removed. When clicking on New
the active computation is copied and the new computation has the same name as the original with
the prefix new_ as shown in the example below. In this example computation_2 was the active
computation at the moment of pressing the button New. A new computation is created called
new_computation_2 that has the same project parameters as computation_2.
To rename the active computation click on Rename, type the new name and press <Enter>.
FINE™ 2-13
User Interface View Area
The third page provides visualization commands on the grid. It consists of two rows: a row of but-
tons and a row of icons.
The row of buttons is used to determine the viewing scope, that is the grid scope on which the view-
ing commands provided by the icons of the second row will apply. There are five modes determin-
ing the scope, each one being represented by a button: Segment, Edge, Face, Block, Grid. Only
one mode is active at a time and the current mode is highlighted. Simply left-click on a button to
select the desired mode.
• : in Grid mode, a viewing operation applies to all the blocks of the grid.
The icons of the second row and their related commands are listed in the following table:
Icon Command
Toggles vertices.
Toggles edges.
Toggles shading.
2-14 FINE™
Mesh Information User Interface
The display of vertices should be avoided since it allows to modify interactively in the
Graphics area the location of the vertices and therefore alter the mesh. The Graphics
area is for display purposes only and any modification on the mesh in the Graphics area
will not be saved in the mesh files and therefore not used by the flow solver.
FINE™ 2-15
User Interface Graphics Area
button to resize
Graphics area
For systems that only accept a mouse with two buttons, the middle mouse button can be
emulated for viewing options by holding the <Ctrl> key with the left mouse button.
2-16 FINE™
Graphics Area User Interface
2-10.1.3 Scrolling
This button is used to translate the contents of active view within the plane of graphics window in
the direction specified by the user. Following functions can be performed with the mouse buttons:
• Left: press and drag the left mouse button to indicate the translation direction. The translation
is proportional to the mouse displacement. Release the button when finished. The translation
magnitude is automatically calculated by measuring the distance between the initial clicked
point and the current position of the cursor.
• Middle: press and drag the middle mouse button to indicate the translation direction. The
translation is continuous in the indicated direction. Release the button when finished. The
translation speed is automatically calculated by measuring the distance between the initial
clicked point and the current position of the cursor.
FINE™ 2-17
User Interface Graphics Area
2-10.1.10Cutting Plane
This option displays a movable plane that cuts the geometry and the blocks of the mesh. The plane
is symbolically represented by four boundaries and its normal, and is by default semi-transparent.
After having selected the button:
• Press and drag the left mouse button to rotate the plane
• Press and drag the middle mouse button to translate the plane
• Press < x >, < y > or < z > to align the plane normal along the X, Y or Z axis
• Press <n> to revert the plane normal
• Press < t > to toggle the transparency of the plane (to make it semi-transparent or fully trans-
parent). It is highly advised to deactivate the plane transparency when using X11 driver to
increase the execution speed.
2-18 FINE™
File Chooser User Interface
FINE™ 2-19
User Interface Profile Manager
The Profile Manager is used to interactively define and edit profiles for both fluids and boundary
conditions parameters. The user simply enters the corresponding coordinates in the two columns on
the left. The graph is updated after each coordinate (after each pressing of <Enter> key).
The button Import may be used if the profile exists already as a file on the disk. The Export button
is used to store the current data in the profile manager as a file (for example to share profiles
between different projects and users). The formats of the profile files are explained in Appendix B
where all file formats used by FINE™ and EURANUS are detailed.
If the mouse cursor is placed over a point in the graph window, this point is highlighted and the cor-
responding coordinates on the left will be also highlighted, giving the user the possibility to verify
the profile.
The button OK will store the profile values and return back to the main FINE™ window.
2-20 FINE™
CHAPTER 3: Fluid Model
3-1 Overview
Every FINE™ project contains a fluid with the corresponding properties. In FINE™ the fluid to use
for a computation can be defined by:
1. Using the fluid defined in the project file of which the properties are shown on the fluid selec-
tion page (see section 3-2.1), or
2. selecting a fluid from the fluid database included in the release that contains a set of pre-defined
fluids (see section 3-2.2), or
3. creating a new fluid (see section 3-2.3).
In the next section the interface is described in detail including advice for use of the fluid defini-
tion. The theoretical background for the different fluid models is described in section 3-3.
FINE™ 3-1
Fluid Model The Fluid Model in the FINE™ Interface
filters
list of properties
3-2 FINE™
The Fluid Model in the FINE™ Interface Fluid Model
Every fluid is associated with the user who created it and can be modified or removed only by its
owner (the owner has Read Write Delete permissions). All other users will have Read Only permis-
sions only.
The user may select a fluid from the list. The selection will be highlighted and the properties of the
selected fluid are displayed in the information box below the fluid list. Those fluid properties will
be saved in the project file ’.iec’ as soon as the project is saved.
Information linked to fluid creation is stored in the fluid database only when closing the
FINE™ interface. As a consequence, this information will not be available to other users
as long as this operation is not completed.
FINE™ 3-3
Fluid Model The Fluid Model in the FINE™ Interface
• The viscosity law which may be constant, temperature dependent or specified by the Sutherland
law. In case the viscosity depends on the temperature the user can specify the law through a pol-
ynome or as a profile through the Profile Manager. Furthermore, when the viscosity law is
defined by the Sutherland law, the Sutherland parameters are accessible by clicking on the but-
ton ( ) right next to the pull down menu.
The user can select a law for the dependence of the laminar viscosity in function of the static tem-
perature. The default is the Sutherland’s law that imposes a temperature dependence of the dynamic
viscosity. The viscosity may also be maintained constant.
–3 –2 –1 1 2 3
Cp ( T ) = A0 T + A1 T + A2T + A 3 + A 4 T + A5 T + A 6 T , (3-1)
–3 –2 –1 1 2 3
γ ( T )= B0T + B1 T + B2 T + B 3 + B4 T + B 5 T + B6 T , (3-2)
where the coefficients A0, A1, A 2, A 3, A 4, A5, A6, B 0, B 1, B 2, B 3, B4, B5 and B 6 are chosen by the user
and set in the dialog box for Cp and γ respectively. In the definition of real gas properties, lower and
upper bounds of temperature variations [Tmin-Tmax] allowed in the domain must be provided.
Depending on the local temperature during the iterative process, the physical properties will then be
able to vary according to the laws defined in the fluid database. Note that it is necessary to cover the
whole temperature range occurring in the system otherwise some inaccuracy in the results might be
introduced. In addition, the local temperature may not fit at some occasion in the prescribed range;
this is especially true during the transient. If so, as indicated by the formula editor (section 3-2.3.5),
the physical properties are kept constant, at a value that corresponds to the lower/upper bound
allowed by the range.
For the default definition of a real gas, involving a compressibility factor, the user has to specify:
• The specific heat at constant pressure characteristic and the specific heat ratio (γ=Cp/Cv) of the
fluid, that can be dependent on the temperature through a polynome or a profile. In both cases,
the user has to specify the temperature range over which the specific heat is defined. For alterna-
tive modelling methods for a real gas see section 3-2.9 and section 3-3.2.2.
• The heat conductivity which is constant, defined by a polynome or profile in temperature or
specified through a constant Prandtl number.
• The viscosity law which may be constant, dependent on temperature or specified by the Suther-
land law. In case the viscosity depends on the temperature the user can specify the law through a
polynome or as profile. Furthermore, when the viscosity law is defined by the Sutherland law,
the Sutherland parameters are accessible by clicking on the button ( ) right next to the pull
down menu.
Compared to a perfect gas computation (Cp and γ constant), the real gas option results in an
increase of about 25% of the CPU time. Real gas calculations are usually robust enough to avoid
any use of preliminary run with average physical properties. However, convergence difficulties
may arise in the transient (say the first 100-200 iterations) if either the temperature range allowed is
too large or if the properties are expected to vary very significantly in the temperature range
allowed. At these occasions, it is often useful to run an equivalent perfect gas, with average proper-
3-4 FINE™
The Fluid Model in the FINE™ Interface Fluid Model
ties, in order to remove any temperature dependency and ease the convergence process in the tran-
sient.
For liquids, the reference temperature and pressure must lie in the expected range of the
static temperature and pressure of the flow field. The reference pressure and temperature
are defined on the Flow Model page (Eq. 4-85).
FINE™ 3-5
Fluid Model The Fluid Model in the FINE™ Interface
steam_tables/ on Windows before launching FINE™. When the subdirectory is created, the con-
densable gas is recognized by FINE™ and the fluid will appear in the list with respectively Fluid
Name and Fluid Type as "_name" and Condensable Fluid.
FIGURE 3.2.3-2 The Formula Editor for polynomials in the fluid properties
The Formula Editor is used to define polynomial values for the fluid properties. The user has to
define the seven coefficients of the polynomial according to the displayed formula on the top, and
also the lower and upper limits of the range, in which the polynomial function is defined. Press
<Enter> after each entry. The constant values equal to the minimal and maximal values will be con-
sidered outside of this range (as shown in the graph).
If a fluid with "Read Only" permissions contains formulas they may be visualized by
means of the Show Fluid Properties... button that opens the Formula Editor window, but
they can not be modified - a message will warn the user that the selected fluid can not be
modified.
3-6 FINE™
The Fluid Model in the FINE™ Interface Fluid Model
If the mouse cursor is placed over a point in the graph window, this point is highlighted and the cor-
responding coordinates on the left will be also highlighted, giving the user the possibility to verify
the profile.
The button OK will store the profile values into the fluid definition.
If a fluid with "Read Only" permissions contains profiles they may be visualized by
means of the Show Fluid Properties... button that opens the Profile Manager window,
but they can not be modified - a message will warn the user that the selected fluid can not
be modified.
Information linked to fluid removal is stored in the fluid database only when closing the
FINE™ interface. As a consequence, this information will not be available to other users
as far as this operation is not completed. A message will however warn users that the
fluid properties have been modified (up to removal in the present situation) by the owner.
Information linked to fluid edition is stored in the fluid database only when closing the
FINE™ interface. As a consequence, this information will not be available to other users
as far as this operation is not completed. A message will however warn users that the
fluid properties have been modified by the owner.
FINE™ 3-7
Fluid Model The Fluid Model in the FINE™ Interface
3-2.7 Filters
On the Fluid Model page two filters are available to display only a limited amount of fluids in the
list of fluids. Select the owner(s) and fluid type(s) to display from the pull down menus to limit the
list of fluids.
Please notice that under Windows™ the fluid database is not owner-oriented. The data-
base is stored locally and user defined fluids is accessible with full permissions to all
Windows™ users.
When loading an existing project and opening the Fluid Model page, please notice that
FINE™ is automatically checking if the fluid is existing or not in the fluid database. If
the fluid name is existing but the properties are different, a comparison window will
invite the user to select the fluid to use (the default fluid in the database or the fluid
defined in the project).
3-8 FINE™
Theory Fluid Model
3-3 Theory
3-3.1 Transport Properties
3-3.1.1 The Laminar Viscosity
In general the kinematic viscosity (in m2/s) is defined in FINE™. The laminar kinematic viscosity
may be
• constant,
• given in terms of a polynomial function of the temperature,
• given by a profile for a certain temperature range,
• varying according to Sutherland’s law.
The dynamic viscosity is then computed within the code by multiplying the kinematic viscosity by
the reference density.
T 0
µ ( T ) = µ ( 120 ) --------- T ≤ 120 K
120
(3-3)
T 1.5 T + TSUTHE
∞
µ ∞ ⎛ ------⎞ ⎛ -----------------------------------
-⎞ 0
µ( T) = T ≥ 120 K
⎝T ⎠ ⎝ T + TSUTHE ⎠
∞
For T ∞ ≤ 120K
T 0
µ ( T ) = µ ∞ ------ T ≤ 120 K
T∞
(3-4)
T 1.5 120 + TSUTHE
µ ( T ) = µ ( 120 ) ⎛ ---------⎞ ⎛ ---------------------------------------⎞
0
T ≥ 120 K
⎝ 120⎠ ⎝ T + TSUTHE ⎠
The viscosity µ ∞ in Eq. 3-3 and Eq. 3-4 is the dynamic viscosity specified in the Fluid Model page
or is obtained from the kinematic viscosity and the density both specified in the Fluid Model page,
while T ∞ and TSUTHE are respectively the reference temperature and the Sutherland temperature
specified also in the Fluid Model page.
FINE™ 3-9
Fluid Model Theory
µC
κ = ---------p- (3-5)
Pr
The turbulent viscosity is calculated iteratively using one of the turbulence models discussed in sec-
tion 4-3.3. The turbulent conductivity is obtained from the turbulent viscosity and a turbulent
Prandtl number whose value can be controlled through the expert parameter PRT (default: 1.0):
µt Cp
κ t = ----------
-. (3-6)
Pr t
p = ρrT (3-7)
with r the gas constant for the perfect gas under consideration:
R
r = ----- = c p – c v , (3-8)
M
with R the universal gas constant and M the molecular weight of the perfect gas.
γ is the specific heat ratio, C p the specific heat at constant pressure, and C v the specific heat at
constant volume with:
Cp
γ = ------ . (3-9)
Cv
When the gas temperature is so low that the vibrational and electronic modes are frozen, the inter-
nal energy of the gas will be proportional to the temperature and the specific heats as well as γ are
constant. The gas is "calorically perfect". This is the usual assumption of moderate speed aerody-
namics.
3-10 FINE™
Theory Fluid Model
γ–1
r = ----------- C p . (3-10)
γ
The values of γ and C p are user input. The static pressure is obtained from the conservative varia-
bles through the following relation:
2
( ρw ) .
p = ( γ – 1 ) ρE – -------------- (3-11)
2ρ
R
p = ρrT = ρ ----- T , (3-12)
M
• the enthalpy equation:
⎛ T ⎞ p ⎧
T
⎫
h = ⎜
∫
⎝ T0
C p dT⎟
⎠
= e + --- =
ρ ⎨ ∫
⎩ T0
C v dT ⎬ +
⎭
rT , (3-13)
where T0 is a reference temperature. In practice T0 is set equal to the minimum temperature of the
temperature range specified in the Fluid Model page.
These two equations lead to:
⎛ T ⎞
rT = ⎜
∫
⎝ T0
( Cp – C v ) dT⎟ .
⎠
(3-14)
In the case of a real gas with varying specific heats one can notice that it is not possible to respect
the 2 above equations with a constant value of r, unless the difference (Cp-Cv) does not vary with
temperature.
Two solutions are proposed in EURANUS in order to model real gases:
• Perfect gas with compressibility factor (expert parameter IRGCON=0)
— the equation of state is modified to include a compressibility factor Z:
R
p = ZρrT = Zρ ----- T (3-15)
M
⎛ T ⎞
⎜
∫
⎝ T0
( Cp – C v ) dT⎟
⎠
Z = -------------------------------------------- (3-16)
rT
• Constant r model
FINE™ 3-11
Fluid Model Theory
— the user specifies the gas constant r (expert parameter RGCST). Then Cp is provided and γ
Cp
calculated as γ = -------------- (expert parameter IRGCON=1).
Cp – r
a) Liquid
The density is user input:
ρ = Cst . (3-17)
This implies a decoupling of the mass and momentum conservation equations from the energy con-
servation equation, since no term in the mass and momentum equations depends on the tempera-
ture.
b) Barotropic Liquid
The density is a user defined function of the pressure:
ρ = f(p) . (3-18)
The corresponding values can be entered either through a polynomial approximation or through a
profile (see section 3-2.3.5 and section 3-2.3.6).
The particularity of the barotropic liquid formulation is that although the density is not constant, a
decoupling of the energy equation is still adopted, exactly as in the full incompressible formulation.
Modification of the treatment of the energy equation
A modification of the treatment of the energy is required in order to account for the density varia-
tions. The temperature equation for any type of fluid can be written as:
dT ∂p
ρc v ------ = – T ⎛ ------⎞ ⎛ ∇V⎞ + ∇ ⎛ k∇T⎞ + ε v , (3-19)
dt ⎝ ∂T⎠ ⎝ ⎠ ⎝ ⎠
V
where εv is the viscous dissipation term and V is the specific volume (inverse of density). Since for
a barotropic fluid there is a unique relation between the pressure and the density the first term on
the right hand side of the above equation vanishes and the temperature for a barotropic fluid
becomes:
dT
ρc v ------ = ∇ ⎛ k∇T⎞ + ε v . (3-20)
dt ⎝ ⎠
∂c ⎛ ∂ 2 p⎞
--------v = T ⎜ --------2⎟ , (3-21)
∂V ⎝ ∂T ⎠
V
the specific heat at constant volume depends only on the temperature cv=cv(T). Note that we also
have as for an incompressible fluid that the specific heats at constant pressure and at constant vol-
ume are equal (cv=cp).
The modification of the formulation of the energy for a barotropic fluid can be derived from the
above relations. Since we have for a barotropic fluid:
3-12 FINE™
Theory Fluid Model
∂p
de = c v dT + ⎛ T ⎛ ------⎞ – p⎞ dV = c v dT – pdV . (3-22)
⎝ ⎝ ∂T⎠ ⎠
V
The barotropic relation can be integrated from a reference state to obtain finally
T p
⎛ ⎞
p ∂ρ
e ( p, T ) = e 0 +
⎜∫
⎜ c dT⎟
⎝ T0
v ⎟
⎠
+ ∫ ----ρ- -----
2 ∂p
- dp . (3-23)
p0
Compared to an incompressible formulation the energy depends on both pressure and temperature.
a) Introduction
The aim of the Condensable Fluid module is the modelling of the real thermodynamic properties of
the fluid by means of interpolation of the variables from dedicated tables.
The module can be used for a single-phase fluid whose properties are too complex to be treated
with a perfect or real gas model. It can also be used in order to treat thermodynamic conditions that
are close to the saturation line. Note that the model can be used on the liquid or on the vapour side
of the saturation curve. In case the thermodynamic state lies inside the two-phase region a homoge-
neous equilibrium two-phase mixture of vapour and liquid is considered. The hypothesis of an equi-
librium mixture is however not valid if the dryness (wetness) fraction exceeds 20%. The module
can not be used above these fractions, as it completely ignores evaporation-condensation phenom-
ena.
For a real fluid the equation of state may be a complicated and usually implicit expression, generat-
ing unacceptable computational overhead if the corresponding equations are explicitly introduced
in the solver. Similarly when an equilibrium mixture of vapour and droplets (wet steam for
instance) is considered as a single fluid, the calculation from the saturation properties of thermody-
namic variables must be done iteratively.
The approach that has been adopted in EURANUS consists of using a series of thermodynamic
tables, one table being required each time a thermodynamic variable must be deduced from two
other ones. This implies the creation of many tables as input, but presents the advantage that no iter-
ative inversion of the tables is done in the solver, with as a consequence a very small additional
CPU time. This additional time corresponds to the one that is needed by the bilinear (or bicubic)
interpolation procedures through the tables. In order to optimize the efficiency of these interpola-
tions the input tables are always built on basis of a Cartesian mesh in the plane of the input varia-
bles (V1,V2) (N1 and N2 values for the variables V1 and V2 respectively).
b) Thermodynamic Tables
Tables are generated based on NUMECA internal tools, based on various sets of well-known equa-
tions as described below:
• Benedict-Webb-Rubin equation (usually used with refrigerants),
• Modified Benedict-Webb-Rubin model (selected equation for hydrogen),
• Vander Waals equation.
Single tables are generally used, which means that the tables cover both single-phase and two-phase
regions, as for instance depicted in the figure 3.3.2-4, showing one of the tables used to model
hydrogen. A bilinear interpolation approach is well adapted to these tables, providing an adequate
smoothing of the saturation region. Bicubic interpolation techniques have also been implemented
and tested. They provide a higher accuracy for a given number of data points but tend to create spu-
FINE™ 3-13
Fluid Model Theory
rious oscillations in the saturation region. These oscillations can be avoided by adopting separated
tables covering respectively the single and the two-phase regions, these two tables intersecting
along the saturation line.
Five categories of tables are used by the solver
1. Basic tables: p(e,ρ) and T(e,ρ)
The numerical scheme being based on a formulation of the equations on basis of the energy and the
density, the basic tables are used in order to update the pressure and the temperature after each
update of the formulation variables.
2. Entropy tables: p(h,s), ρ(h,s), s(h,p), h(s,p)
The entropy tables are used at the inlet/outlet boundary conditions, and are required in order to cal-
culate the total (static) thermodynamic conditions from the static (total) ones.
3. The (p,T) tables: e(p,T) and ρ(p,T)
These tables are required in order to allow the use of the inlet boundary conditions based on total
pressure and temperature. They are also required by the turbomachinery initial solution procedure.
ρ [kg/m3]
3-14 FINE™
Theory Fluid Model
The viscosity and conductivity can also be interpolated from tables. If these tables are not present
the other laws available within the user interface can be used.
5. The saturation table
This table contains the liquid and vapour values of all thermodynamic variables along the saturation
line. It is only required if the user activates the dryness fraction output and/or if the inlet boundary
condition based on total enthalpy and dryness fraction is selected.
∂p p ∂p
c = ⎛ ------⎞ + ----2- ⎛ ------⎞
2
(3-24)
⎝ ∂ρ⎠
e ρ ⎝ ∂e⎠ ρ
The Baldwin-Lomax, Spalart-Allmaras and (k,e) turbulence models can be used. The turbulent con-
ductivity calculation is based on the usual relation:
µt
κ t = -------- c p (3-25)
Prt
The specific heat at constant pressure can either be constant (the value being provided in the user
interface) or deduced from the specific table.
Condensable fluid option is not compatible with the use of cooling/bleed module and/or
upwind schemes for space discretization.
By default, a set of thermodynamic tables for water (steam) is proposed starting FINE™/
Turbo v6.2-9. More details can be found in Appendix D. The installation of the thermo-
dynamic tables must be performed as described in section 3-2.3.4
d) Specific Output
Specific outputs can be created by the condensable fluid module:
• enthalpy: static, total absolute and total relative
• dryness fraction (0<x<1), and generalised dryness fraction
FINE™ 3-15
Fluid Model Theory
Below is a list of the buttons appearing in the Outputs/Computed Variables page under the thumb-
nail Thermodynamics when the condensable gas is selected as fluid model.
3-16 FINE™
CHAPTER 4: Flow Model
4-1 Overview
The Flow Model page, as displayed in Figure 4.1.0-1, can be used in order to specify several char-
acteristics of the flow:
• the time configuration to define time dependence of the equations to solve,
• the mathematical model to:
— choose between viscous and non-viscous flow,
— choose between laminar and turbulent flow,
— activate gravity,
— activate pre-conditioning for low speed flow,
• characteristic scales defining the Reynolds number of the flow,
• reference values of the temperature and pressure in the flow.
This chapter is organised in the following way:
• section 4-2 describes the interface, best practice and theoretical background for unsteady com-
putations,
• section 4-3 describes the interface, best practice and theoretical background for the definition
of the mathematical model and especially the available turbulence models,
• section 4-4 provides details on the Reynolds number related information and reference values.
FINE™ 4-1
Flow Model Time Configuration
a) Rotating Machinery
On the Rotating Machinery page, under the Rotor-Stator thumbnail, the domain scaling rotor/stator
interaction is activated. Furthermore, the Phase Lagged capability can be activated on the top of
the page. When activated, the phase-lagged rotor/stator interaction will be applied.
Finally in both cases, the Number of Angular Positions (in one periodicity) is requested in the Com-
putation Steering/Control Variables page instead of the time step.
4-2 FINE™
Time Configuration Flow Model
• fct(time) to define a constant value in space but varying in time. Click on the profile data but-
ton ( ) to activate the Profile Manager.
• fct(space-time) to define a space and time dependent variation of the boundary condition
value. In this case a one-dimensional space profile has to be defined. The time profile to
impose is a scaling factor that is applied to the space profile. This means that the profile in
space is only varying in time with a certain scaling factor.
Another way to define an inlet/outlet unsteady boundary condition is to set an absolute rotational
speed to the boundary condition by activating Rotating Boundary Condition and specifying the
rotational speed in Rotational Speed Unsteady respectively under the INLET and OUTLET thumb-
nails, and to set the periodicity of the signal entered in FINE™ through the expert parameter
NPERBC.
It is important to notice that EURANUS allows only one rotation speed (combined
with a zero rotation speed).
0 π
θ
θ1 θ2 θ3 θ4 2π
v2.2
t0 t0+T
The blade channel is initially (t0) located at θ1,θ2 and will after one period of the signal (T) be
located at θ3,θ4. Τhe circumferential signal has thus to be imposed from at least θ1 to θ4.
When Rotating Boundary Condition is activated, the time step is no longer specified by the physical
time step in the FINE™ interface in the Computation Steering/Control Variables page. Instead the
time step is imposed by defining the Number Of Angular Positions.
c) Phase Lagged
When activating Rotating Boundary Condition in the Boundary Conditions page to rotate the refer-
ence system of the boundary condition at the inlet and/or outlet, the period of the boundary condi-
tion should be compared to the blade passing period. In the case of a single blade row calculation
with the period of the varying inlet or outlet boundary conditions different from the blade passing
period, the Phase Lagged option should be activated in the Configuration/Rotating Machinery page
FINE™ 4-3
Flow Model Time Configuration
and the periodicity of the signal entered in FINE™ has to be specified through the expert parameter
NPERBC. See section 4-2.3.2 and section 4-2.4.5 for more detailed information on the phase
lagged boundary condition.
In the case of phase lagged approach, the Number Of Angular Positions is the number
of time iterations per full machine rotation.
It is important to notice that EURANUS allows only one rotation speed (combined
with a zero rotation speed).
d) Control Variables
On the Computation Steering/Control Variables page the following parameters appear when select-
ing an unsteady time configuration non-turbomachine:
• Physical Time Step: in general cases the user specifies the magnitude of the physical time step
in seconds. This time step is constant through the whole calculation. In cases of turbomachin-
ery applications where Rotor/Stator interfaces are detected or Rotating Boundary Condition
activated, the Number Of Angular Positions is requested instead of the time step as described
in section 4-2.2.
• Number Of Physical Time Steps: to define the number of time steps to perform. In cases of tur-
bomachinery applications where Rotor/Stator interfaces are detected or Rotating Boundary
Condition activated, the Number of Periods can also be specified.
• Save Solution Every: allows to save the solution after a constant number of time steps. By
default the solution is overwritten in the same output files. In order to keep the successive
unsteady solutions, the multiple output (Multiple Files) option has to be activated.
• Outputs For Visualization:
— Output For Visualization/At end only and Multiple Files: multiple set of output files saved
for restart of the unsteady computation and visualization in CFView™ only available at the
end of the computation.
— Output For Visualization/At end only and One Output File: one set of output files saved for
restart of the unsteady computation and visualization in CFView™ only available at the end
of the computation.
— Output For Visualization/Intermediate and Multiple Files: multiple set of output files saved
for restart of the unsteady computation and visualization in CFView™ during the computa-
tion.
— Output For Visualization/Intermediate and One Output File: one set of output files saved for
restart of the unsteady computation and visualization in CFView™ during the computation.
• Number Of Steady Iterations: the number of iterations to initialize the unsteady computation.
Such initialization is performed with the steady state algorithm, using fixed geometry and
mesh as well as constant inlet/outlet boundary conditions. In case this initialization procedure
is used the steady state iterations can even be preceeded by a full multigrid process allowing a
rapid initialization of the flow (see section 9-4.3).
• Save Steady Solution Every: the number of iterations at which the solution of steady initializa-
tion is saved
By default only one output file is created. For the generation of multiple output files,
4-4 FINE™
Time Configuration Flow Model
Note that only limited output is saved when selecting Multiple Files with Output For
Visualization/At end only and a solution is requested every x time steps, only solution
files with extension ’.cgns’ are written every x time steps instead of the full set of solu-
tion files. This limited output still allows to restart from this solution (see section 15-2).
FINE™ 4-5
Flow Model Time Configuration
4-6 FINE™
Time Configuration Flow Model
Depending on the periodicity of the boundary condition the phase lagged approach needs to be
used:
1. The circumferential distribution of the imposed inlet/outlet conditions has the same period as
the blade row. The calculation of such time-dependent flow does not require to activate phase
lagged approach.
2. The period of the time varying inlet/outlet boundary condition is different from the blade pass-
ing period. The calculation of such time-dependent flow requires to activate the Phase Lagged
approach at the top of the Rotating Machinery page.
A circumferential profile must be entered in the FINE™ interface for the full range
[0,2π] and NPERBC = 1 or for part of the range [0,2π/N] covering one or multiple peri-
ods of the signal and NPERBC = N. The profile should be based either on θ, z-θ or r-θ.
The profile is automatically rotating at the speed of rotation that is specified through the
rotation speed specified in the Boundary Conditions page.
A profile as a function of r-θ must be structured. The profiles must be given at stations
of constant θ for increasing r. Also θ must be increasing going from one station to the
next in the file.
FINE™ 4-7
Flow Model Time Configuration
’.cgns’ files will be generated. The files are automatically renamed accordingly to the
time step computed using the suffix ’_t#.’ that is thus determining the #th physical time
step. The corresponding ’.cfv’ files can be read in CFView™ only if in addition of the
Multiple Files the Intermediate has been selected in the Output For Visualization option.
4-8 FINE™
Time Configuration Flow Model
Note that the ’.run’ and ’.cgns’ files (without ’_t#.’ suffix) are linked to the initial state of calcula-
tion.
∂ ∂U
----
∂t ∫ ∫ ∫ U dV + ∫ ∫ ∫ ------
∂τ ∫∫
- dV + F ⋅ dS =
∫ ∫ ∫ ST dV , (4-1)
V V S V
where t is the physical time and τ is pseudo time, U is the solution vector of the conservative variables, V is
the volume, F are the conservative fluxes and ST are the source terms. The time discretization of the deriva-
tive of the conservative variables with respect to the physical time t is made to meet the desired accuracy in
time. Two formulations are available:
First order upwind in time:
⎛∂ ⎞n +1 U
n+1
V
n+1
–U V
n n
⎜ ----
⎝ ∂t
∫∫∫ U dV⎟
⎠
= --------------------------------------------- .
∆t
(4-2)
V
This first order scheme is used only for the first time step of the unsteady computation except for a restart
when the two previous solutions are available (see section 4-2.2).
Second order upwind in time:
⎛∂ ⎞n+1 1.5U
n+1
V
n+1 n
– 2U V + 0.5U V
n n–1 n–1
⎜ ----
⎝ ∂t
∫∫∫ U dV⎟
⎠
= ------------------------------------------------------------------------------------------------ .
∆t
(4-3)
V
All other terms in Eq. 4-1 are computed at time n+1. The equation is then treated as a modified steady state
problem in the pseudo time τ:
∂U
------- V n + 1 + R TA ( U ) = 0 . (4-4)
∂τ
Denoting by R the residual corresponding to the steady state problem, the new residual used for time accurate
computations RTA is given by:
FINE™ 4-9
Flow Model Time Configuration
For the second order upwind scheme we have β1=1.5, β0=-2, β-1=0.5 while for the first order
scheme we have β1=1, β0=-1, β-1=0. The new residual contains terms that depend only on the solu-
tion of previous time steps, all other terms depend on the current solution vector U.
When steady state is reached at each physical time step, the left hand side of Eq. 4-5 tends to zero
and the time accurate solution is obtained.
l+1 l ∂U
U = U + ------- ∆τ . (4-6)
∂τ
Putting Eq. 4-6 into Eq. 4-5, a modified pseudo time step is introduced:
∆τ
∆τ mod = ---------------------- . (4-7)
∆τ
1 + β 1 ------
∆t
The pseudo time step is modified by an additional term containing the physical time step. This term
becomes dominant if the physical time step is significantly smaller than the pseudo time step. This
situation does not often happen since the physical time step is taken as large as possible. However,
in the far field where mesh dimensions are increased the pseudo time step might reach high values
and the contribution of the physical time step will have a stabilizing effect there.
The physical time step ∆t is limited by the level of desired accuracy. For periodic flows the time
step is usually taken a fraction lying between 1/20-1/40 of the expected period.
n+1 n+1 n n
U initial V = U V (4-8)
Vn+1 being the cell volume evaluated at the next time step. The volume evaluation at each physical
time step is only performed for moving meshes.
In Jameson (1991) it is suggested to take as initial solution an extrapolated solution from the previ-
ous time steps. This is obtained by a Taylor expansion:
n
n+1 n ∂(U ⋅ V ) 2
(U ⋅ V) = ( U ⋅ V ) + ----------------------- ∆t + O ( ∆t ) (4-9)
∂t
This expression leads to the following initial solution:
In practice however, using this expression does not show any convergence enhancement for the
next time step compared to the convergence obtained with Eq. 4-8. Eq. 4-8 is therefore imple-
mented.
4-10 FINE™
Time Configuration Flow Model
In the particular case of turbomachinery applications a discretization in time of the blade passing
period is performed by the unsteady algorithm, and once one full period has been accomplished by
the algorithm, the solution may be initialised using the solution obtained at the same position at the
previous rotation. This requires the storage of all the intermediate solutions and the experience has
shown that it did not bring significant convergence improvements. The approach has however been
implemented in EURANUS, and can be activated by setting the expert parameter PREROT to 1.
∂ – 1 ∂Q
----
∂t ∫ ∫ ∫ U dV + ∫ ∫ ∫ Γ ------- dV +
∂τ ∫ ∫ F ⋅ dS = ∫ ∫ ∫ ST dV , (4-11)
V V S V
where t is the physical time and τ is pseudo time, U is the solution vector of the conservative varia-
bles, Q is the solution vector of chosen dependent variables, Γ is the preconditioning matrix, V is
the volume, F are the conservative fluxes and ST are the source terms. For all fluids we have:
Q = ( p g, u, v, w, E g, ρ k, ρε, Y i ) . (4-12)
The time discretization of the derivative of the conservative variables with respect to the physical
time t is made to meet the desired accuracy in time. The equation is then treated as a modified
steady state problem in the pseudo time τ:
∂Q n + 1
------- V + R TA ( Q ) = 0 , (4-13)
∂τ
with
l+1 l ∂U ∂Q
U = U + ------- ------- ∆τ (4-15)
∂Q ∂τ
Putting this relation into equations Eq. 4-13 we obtain the following updating procedure:
∆Q n + 1 ∂U ∆τ – 1
-------- V = – ⎛ I + Γβ 1 ------- ------⎞ R ( U ) (4-16)
∆τ ⎝ ∂Q ∆t ⎠ TA
where I is the identity matrix. The updating procedure has been modified and includes now the time
steps ratios, leading to a stabilizing effect.
FINE™ 4-11
Flow Model Time Configuration
n+1 n+1 n n
U initial V = U V
(4-17)
n+1 n+1
Q initial = Q ( U initial )
Since the dependent variables are Q, these variables have to be computed at the beginning of each
physical time step from the conservative variables U.
where qo is a space distribution (uniform, 1D or 2D) and f(t) is a scaling function read from a file:
(tn, f(tn)).
2) rotation of the reference system
In turbomachinery the relative motion between stator and rotor blades induces an unsteadiness, due
to the fact that for instance in a turbine stage the flow conditions that a rotor sees at the inlet are
time dependent, whereas these conditions would be fixed for a fixed reference system.
If one of the imposed variable along the inlet or the outlet is not uniformly distributed in the tangen-
tial direction in the fixed space the inlet/outlet section sees a boundary condition that changes in
time. In this case, the space distribution of the imposed quantity being defined by qo (1D or 2D),
each cell face centre coordinate is rotated by an angle ∆θ = Ω relative • ∆t at each time step, before
performing a new space interpolation.
The activation of this procedure along the inlet and/or the outlet is obtained by activating Rotating
Boundary Condition in the Boundary Conditions page and specifying the rotational speed in Rota-
tional Speed Unsteady respectively under the INLET and OUTLET thumbnails.
The use of a Rotating Boundary Condition imposes a limitation for the specified pro-
files at the inlet and outlet. Only θ profiles (f(θ), f(r-θ) or f(θ-z)) are allowed and the 2D
space profile f(r-θ) must be structured (profile f(r) with the same number of points for
different θ)
4-12 FINE™
Time Configuration Flow Model
in magnitude as the gap between successive blade rows is decreased, and affecting consequently the
performance of the machine.
Ω
D
C
n∆t
G
F
Y H
E
B A
Z X
NumberOfAngularPositions NOFROT
n= --------------------------------------------------------------------------- = ---------------------------
NumberOfBlades NBLADES
FINE™ 4-13
Flow Model Time Configuration
Let us consider that the computational domain is the one defined by ABCDHGFE as shown in
Figure 4.2.4-2, where the solid blade walls correspond to the segments BC and FG. Under the con-
text of phase-lagged boundary conditions, the scalar variables (density, pressure, turbulent viscos-
ity,...)on AB, EF and CD, GH are not equal and the velocity vectors are not related by a simple
rotation operator, since the flow is not periodic.
U GH ( t ) = U CD ( t – ( n ⋅ ∆t ) ) , (4-18)
Therefore, the flow variables of segments AB and CD need to be stored for the n preceding time
steps.
Boundary conditions for segments AB and CD:
In the absolute system of reference, the actual positions of segments AB and CD correspond respec-
2π – ϕ
tively to the positions of segments EF and GH, -------------------------------------- time steps earlier, with ϕ the angle
( NOFROT – n )
between D and H. The variables at time t on segments AB, CD are obtained from those on segments
EF, GH using:
2π – ϕ
U CD ( t ) = U GH ⎛ t – -------------------------------------- ∆t⎞ .. (4-19)
⎝ ( NOFROT – n ) ⎠
2π – ϕ
Therefore, the flow variables of segments EF and GH need to be stored for the -------------------------------------- pre-
( NOFROT – n )
ceding time steps.
Since at the beginning of the computation no solution is available at the previous time steps, stand-
ard periodic boundary conditions are applied at the first iteration and then frozen during n time
steps for segments EF, GH and (NOFROT-n) time steps for segments AB and CD. Although this
procedure is the simplest, it might contribute to slow down the convergence to a periodic state.
2π
∆t = --------------------------------------------------------------- , (4-20)
Ω ⋅ Nperiods ⋅ NOFROT
4-14 FINE™
Mathematical Model Flow Model
2π 1
∆t = ------ ------------------------- . (4-21)
Ω NOFROT
4-2.4.7 References
Arnone, A., Liou, M. and Povinelli, L., 1995, " Integration of Navier-Stokes Equations using dual
time stepping and multigrid method:, AIAA-Journal, Vol. 33, No. 6, 1995, pp. 985-990.
Eliasson, P., and Nordstrom, J., 1995, " The development of an unsteady solver for moving meshes
", FFA/TN 1995-39.
Fatsis, A., Pierret, S., Van den Braembussche, R., 1995, " 3D unsteady flow and forces in centrifu-
gal impellers with circumferential distortion of the outlet static pressure ", ASME, 95-GT-33.
Hakimi N. (1997) ’Preconditioning methods for time dependent Navier-Stokes equations’, PhD
Thesis, Dept of Fluid Mechanics, Vrije Universiteit Brussel.
Hirsch Ch. (1990) 'Numerical Computation of Internal and External Flows. Volume 2', John Wiley
& Sons.
Jameson, A., 1991, "Time dependent calculations using multigrid, with applications to unsteady
flows past airfoils and wings", AIAA-Paper 91-1596.
Melson, N.D., Sanetrik, M.D., Atkins, H.L., 1993, "Time accurate Navier-Stokes calculations with
multigrid acceleration", AIAA paper, pp. 1041-1042.
Pierce, N.A., Alonso, J.J., 1997,"A preconditioned implicit multigrid algorithm for parallel compu-
tation of unsteady aeroelastic compressible flows:, AIAA-Paper 97-0444.
FINE™ 4-15
Flow Model Mathematical Model
a) Boundary Condition
Inlet
• Spalart-Allmaras: the value of the turbulent viscosity must be defined. It might be constant or
varying in space and time. The user can specify the time or spatial variation through the profile
manager.
• k-ε models: the values of the turbulent quantities (k and ε) must be defined. They might be con-
stant or vary in space and time. The user can specify the time or spatial variations through the
profile manager.
b) Initial Condition
• Spalart-Allmaras: a constant initial value of the turbulent viscosity is defined through the expert
parameter NUTFRE on the Control Variables page under the Expert Mode.
• k-ε models: the initial values of the turbulent quantities (k and ε) are specified by the user if con-
stant or turbomachinery initial solutions are chosen.
4-16 FINE™
Mathematical Model Flow Model
When problems occur during the wall distance calculation (wall distance not equal to
0 on the wall), we recommend to rise the value of NTUPTC to 40. Furthermore, RTOL
has to be decreased to 0.9 if the wall distance gradient is not smooth.
Baldwin-Lomax
IATFRZ: multigrid parameter
= 0: apply the turbulence model on all grid levels,
= 1: freeze turbulence on coarser grids (default),
= 2: freeze turbulence on all grids (see ITFRZ and RESFRZ).
ITFRZ: freeze the turbulent viscosity field if the fine grid iteration exceeds ITFRZ (default value =
1000000).
RESFRZ: freeze the turbulent viscosity field if the fine grid residual is lower than RESFRZ (default
value = -12).
Spalart-Allmaras
MUCLIP: controls the maximum allowable value for the ratio MUT/MU (default value = 5000).
NUTFRE: constant to initialize the turbulent viscosity field (default value = 0.0003).
k-ε models
CMU: constant for the k-ε model in Eq. 4-59 (default value = 0.09).
CE1: constant for the k-ε model in Eq. 4-57 (default value = 1.44 automatically set to 1.35 for the
Chien k-ε model).
CE2: constant for the k-ε model in Eq. 4-57 (default value = 1.92 automatically set to 1.8 for the
Chien k-ε model).
SIGK: constant for the k-ε model in Eq. 4-56 (default value = 1.).
SIGE: constant for the k-ε model in Eq. 4-57 (default value = 1.3).
PRT: constant for the k-ε model - the turbulent Prandtl Number (default value = 1.).
INEWKE: defines the high Reynolds k-ε model when k-epsilon (Extended Wall Function) is
selected:
= 0: standard formulation of the wall functions,
= 10 (default): mesh independent formulation of the wall functions (extended wall functions).
ICOPKE: (= 1) applies the pressure gradient-velocity model to take into account a compressibility
correction. It is useful only for high Mach number flows (M>3). The default value is 0.
FINE™ 4-17
Flow Model Mathematical Model
SIGRO: constant in the k-ε model when ICOPKE =1 in Eq. 4-70 and Eq. 4-72 (default value = 0.5).
C3: constant for the k-ε model when ICOPKE = 1 in Eq. 4-72 (default value = 2.).
ICODKE: parameter to active the compressible dissipation,
= 0 (default): no compressible dissipation,
= 1: Sarkar model for compressible dissipation (Eq. 4-75),
= 2: Nichols model for compressible dissipation (Eq. 4-77).
ALF: constant of Sarkar model for compressible dissipation when ICODKE=1 (default = 0.5).
CP1: constant of Nichols model for compressible dissipation when ICODKE=2 (default value = 4.).
MAVRES: allows to control the update of k and ε (default = 0.95),
> 0: according to the restrictions defined by Eq. 4-78 and Eq. 4-79,
= 0: according to the restriction defined by Eq. 4-78,
< 0: according to the restriction defined by Eq. 4-80.
MAVREM: allows to control the multigrid corrections for k and ε (default = 0.95),
> 0: according to the restrictions defined by Eq. 4-78 and Eq. 4-79,
= 0: according to the restriction defined by Eq. 4-78,
< 0: according to the restriction defined by Eq. 4-80.
MUCLIP: controls the maximum allowable value for the ratio MUT/MU (default value = 5000).
LTMAX: maximum turbulent length scale (default = 1.E+6),
= -1: control of turbulent length scale with automatic clipping.
IYAP: (= 1) applies the Yap’s modification to control the turbulent length scale (default value = 0).
The value of the expert parameter LTMAX must be the default value.
TEDAMP: parameter to improve the robustness of the k-ε models (default value = -1),
> 0: a minimum of TEDAMP multiplication factor of the clipping value (EKCLIP) is used
for k in the factor 1 ⁄ T in Eq. 4-57,
<= 0 : the minimum values used for k and ε correspond to the expert parameter EPS,
GAMMAT: introduces the turbulence time scale in the computation of the local time step to control
stability (default value = 10).
PRCLIP: sets an upper bound for the ratio between production and dissipation (default value = 50).
LIPROD: (= 1) corresponds to a linearization of the production term (default value = 0) when strain
rate is large (i.e. impinging flow) and PRCLIP has no more effect on the flow field.
EKCLIP: clipping value for the turbulent kinetic energy k (default value = 1.E-5).
EPCLIP: clipping value for the turbulent dissipation rate ε (default value = 1E-5).
KEGRID: Full multigrid parameter corresponding to the finest grid level on which the Baldwin-
Lomax model is used (default = 2). On the lower grid levels, the k-ε model is used.
IUPWTE: (= 1) uses an upwind scheme for the convection of k-ε instead of a centre scheme
(default = 1).
4-18 FINE™
Mathematical Model Flow Model
IKENC: (= 1) solves the k-ε equations with a non-conservative approach (default = 1).
IKELED: (= 1) actives the LED scalar scheme for k-ε instead of the standard central scheme
(default = 0).
IOPTKE: (= 1) optimized implementation for k-ε (Yang-Shi, Wall function, Lauder Sharma)
(default = 1).
FINE™ 4-19
Flow Model Mathematical Model
Another model often used in design is "standard" k-ε. This model employs an empirically based
logarithmic function to represent the near-wall physics and requires a lower grid resolution in this
region as a result. One drawback of this treatment is that the logarithmic function does not apply for
separated flows, although the NUMECA extended wall function will still work. If it is expected that
a significant amount of separation will have to be predicted, one of the low-Reynolds number tur-
bulence models would be more appropriate. All of the listed turbulence models employ constant
turbulent Prandtl numbers which is somewhat of a restriction when performing heat transfer calcu-
lations. However, experience has shown quite successful prediction of heat transfer coefficients
when using Baldwin-Lomax and Launder-Sharma k-ε.
a) Cell Size
Separating flow
logarithmic law
Note that the sublayer extends up to y+=5 but 10 is an acceptable approximation for
design calculations.
Note: The variable v* displayed in the figure is uτ.
Picture from White, F.M., Viscous Fluid Flow, McGraw Hill, 1991.
When calculating turbulence quantities it is important to place the first grid node off the wall within
a certain range (ywall). This can be done for the blade and the endwalls (hub and shroud) independ-
ently. When doing computations including viscosity (Navier-Stokes equations) the boundary layer
near a solid wall presents high gradients. To properly capture those high gradients in a numerical
simulation it is important to have a sufficient amount of grid points inside the boundary layer. When
Euler computations are performed no boundary layer exists and therefore the cell size near solid
walls is of less importance.
4-20 FINE™
Mathematical Model Flow Model
To estimate an appropriate cell size ywall for Navier-Stokes simulations including turbulence, the
local Reynolds number based on the wall variable y+is computed. The value of y+ associated with
the first node off the wall will be referred to here as y1+:
ρu τ y wall
y1+ = --------------------- (4-22)
µ
where uτ is the friction velocity:
τ wall 1---
uτ = ----------- = ( V ) 2 Cf (4-23)
ρ 2 ref
In Figure 4.3.5-3 is represented the evolution of u+ against y+ from the measurements of Lind-
gren(1965) with:
u+= u / u τ
Low-Re models resolve the viscous sublayer and are well suited for y1+ values between 1 and 10
whereas high-Re models apply analytical functions to the log-layer and are appropriate to y1+ val-
ues ranging from 20 to 50 (it depends on the extension of the buffer layer for the considered flow).
Moreover one can notice that the logarithmic function does not apply for separated flow. So
whether it is expected that a significant amount of separation will have to be predicted, one of the
low-Reynolds number turbulence models would be more appropriate.
Recommendations are given in the table below for ranges of y1+ specific to the different types of
models.
One way to estimate ywall as a function of a desired y+ value is to use a truncated series solution of
the Blasius equation:
Vref – 7 / 8 ⎛ L ref⎞ 1 / 8 +
y wall = 6 ⎛ ---------⎞ --------- y ... (4-24)
⎝ ν ⎠ ⎝ 2 ⎠
Note that the reference velocity, Vref, can be taken from an average at the inlet. For instance, if the
mass flow is known the value can be calculated using density and the cross-sectional area of the
inlet. If the mass flow is not known the reference velocity may be calculated from the inlet total
pressure and an estimated static pressure using isentropic relations. The reference length, Lref,
should be based on streamwise distance since an estimation of boundary layer thickness is implied
FINE™ 4-21
Flow Model Mathematical Model
in this calculation. For instance, in the case of a turbomachinery simulation one could use the dis-
tance of hub and shroud curves that exist upstream of the first row of blades. This is approximate,
of course, as the thickness of boundary layers will vary widely within the computational domain.
Fortunately it is only necessary to place y+ within a range and not at a specific value.
Another method of estimating ywall is to apply the 1/7th velocity profile. In this case the skin fric-
tion coefficient Cf is related to the Reynolds number:
0.027-
Cf = ------------- (4-25)
Re x1 / 7
where Rex should be based on average streamwise values of Vref and Lref as discussed above. Since
uτ is based on Cf it may be calculated based on Eq. 4-23, and ywall may then be calculated from
Eq. 4-22. Note that either method is not exact but they will yield results that are quite close to each
other. In fact, it can be instructive to calculate ywall using both methods as a check. Since only one
wall distance is being calculated, the particular flow being studied should be kept in mind. For
instance if it is a diffusing flow Cf, and hence y+, can be expected to drop. Since a certain range is
desired (e.g., 20< y+<50 for high-Re Standard k- ε) the user may choose to base the calculation of
wall distance on an average of that range (e.g., 40).
c) General Advice
• What grid resolution is adequate?
The resolution method employed in the EURANUS flow solver requires approximately 9 nodes
across a free shear-layer and approximately 15 across a boundary layer to provide grid-independent
results for turbulent flows. If wall functions are used the boundary layer only requires approxi-
mately 9 nodes.
Of course the flow field under study will realistically consist of shear layers of which the width var-
ies substantially throughout the flow field. The user must therefore decide what is important to cap-
ture and what is not. For instance, in the design-cycle analysis of a compressor with a volute it
would probably be acceptable to choose a fully-developed boundary layer. The number of nodes
across the diameter would therefore be approximately 29. However, it would be wise to select a
number like 33 to maintain a a large number of multi-grid levels. The selection of nodes in the
streamwise direction should be governed by what resolution adequately represents the studied
geometry. Regions of concentrated high gradients, such as airfoil leading or trailing edges or any
geometrical corners should contain a relatively high clustering of nodes.
4-22 FINE™
Mathematical Model Flow Model
d) Verification of y1+
By following this instructions it should be possible to generate a grid of reasonable quality for tur-
bulent flows. It is recommended however, that the user checks values of y1+ after approximately
one hundred iterations on the fine grid to ensure the proper range has been specified. At the same
time, it can be useful to plot contours of residuals (continuity, momentum, energy and turbulence)
over selective planes. If the level of skewness is too high, this will be indicated by local peaks in
residuals that are orders of magnitude greater than the rest of the flow field. If a multi-block grid is
used, the residual levels in each block can be compared in the monitor.
a) Spalart-Allmaras Model
When the Spalart-Allmaras model is selected the user should specify in the inlet boundary condi-
tion the kinematic turbulent viscosity νΤ (m2/s). If no information is available on the turbulence
properties of the flow, estimates can be made based on the following assumptions that:
ν
• For internal flows (e.g. turbomachinery): -----T = 1 to 5.
ν
ν
• For external flows (e.g. vehicle aerodynamics): -----T = 1 .
ν
b) K-ε Models
• Estimation of k
The value of the turbulent kinetic energy can be derived from the turbulence intensity Tu or from
the wall shear stress.
FINE™ 4-23
Flow Model Mathematical Model
— From the turbulence intensity. The turbulence intensity Tu can be expressed against the
streamwise fluctuating velocity u" and the streamwise velocity Uref:
″2
u
T u = ------------ . (4-26)
U ref
For internal flows the value of Tu is about 5% and for external flows it is reduced to 1%. With
these considerations k can be calculated in considering an isotropic turbulence:
3 ″2 2
k = --- ( u ) (4-27)
2
— From the wall shear stress. If the wall shear stress is known, the user can use the wall func-
tions defined for the fully turbulent flow:
τ
wall
k = ----------- (4-28)
ρ Cµ
This value of k could be used as an initial value and also for the inlet boundary condition.
• Estimation of ε
The value of the turbulent dissipation can be specified through one of the following rules:
— Specify the ratio of the turbulent viscosity to the laminar viscosity
2
µ ρref k
ε = C µ ------ -------------- (4-29)
µ it µ
3 3
--- ---
4 2
cµ k
ε = ---------- . (4-30)
l
— Derive ε from the asymptotic turbulent kinetic equation:
In a free uniform flow the turbulent kinetic energy equation reduces to
∂k
u ----- = – ε . (4-31)
∂x
This relation can be used to specify the value of the turbulent dissipation in the following way:
∆k
ε = – u ------ , (4-32)
L
where u is the inlet velocity, ∆k the decay of the turbulent kinetic energy over a length L. For
example, in a turbomachine, L is the maximum geometric length and ∆k could be set to 10% of
the inlet value of k.
4-24 FINE™
Mathematical Model Flow Model
Using this method, the user must make sure that the value of the turbulent viscosity obtained
µ
from these values of k and ε is not too big or too small. i.e. 1 < ----t < 1000 . If this condition is not
µ
satisfied, it is advised to scale down or up the value of k inlet or the ∆k and compute again the
turbulent dissipation
— Specify the wall shear stress.
If the wall shear stress is known, the user can use the wall functions defined for the fully turbu-
lent flow.
3
---
2
( τ wall ⁄ ρ )
ε = ------------------------
- (4-33)
l
If there is an initial solution file containing k and ε , the k and ε values of this file are used to
initialize the fields.
The initial values mentioned above can also be used to set the inlet boundary conditions for the k
and ε fields. However, to reduce the possibility of oscillations in skin friction due to non-physical
relaminarisation during convergence, it is recommended to insure ε ≈ 0.1 ε inlet .
Either of the above methods can be applied for setting the boundary conditions in tur-
bomachine applications. However, if the given values of k and ε lead to the killing of the
turbulence shortly after the inlet section, we suggest to apply Eq. 4-32 and select an inlet
values of k such that the ratio of the turbulent viscosity to the laminar viscosity equals
50.
In some cases a cross-check between Eq. 4-29 and Eq. 4-31 may result in very differ-
ent values for ε. In such a case it is recommended to re-evaluate k and ε in the following
manner:
1. Use relation Eq. 4-32:
∆k 0.1k
ε = – u ------- = u ---------- . (4-34)
∆L ∆L
This relation expresses that k0 (k at the inlet) is expected to be decreased by about 10% over a
length ∆L that is characteristic to the size of the domain.
2. Use relation Eq. 4-29. In this relation Cµ=0.09 and ν represents the laminar viscosity of the
fluid.
3. Combine the relations of the two previous steps to remove ε0, leading to:
0.1U µ T ν
k 0 = ------------ ⎛⎝ ------ ------⎞⎠ . (4-35)
∆L µ C µ
4. Using either relation Eq. 4-29 or Eq. 4-32 then easily leads to an estimation of ε0.
FINE™ 4-25
Flow Model Mathematical Model
4-26 FINE™
Mathematical Model Flow Model
are then reset using the classical procedure while the other variable fields (density, velocity and
energy) are transferred adequately. To overcome this difficulty, KEGRID must again be set to a
value higher than the number of grids. This procedure results in a much better initialization of k and
ε, thus preventing some relaminarisation while enhancing convergence.
FINE™ 4-27
Flow Model Mathematical Model
4-3.6.1 Baldwin-Lomax
The Baldwin-Lomax algebraic turbulence model, Baldwin & Lomax (1978) is a two layer model
where the turbulent viscosity in the inner layer is determined using Prandtl’s mixing length model,
and the turbulent viscosity in the outer layer is determined from the mean flow and a length scale.
The strain-rate parameter in Prandtl’s mixing length model is taken to be the magnitude of the vor-
ticity.
The influence on the mean flow equations through the turbulent kinetic energy is neglected.
The turbulent viscosity is given by
⎧ ( µ t ) i, n ≤ n c
µt = ⎨ (4-36)
⎩ ( µ t ) 0, n > n c
where n is the normal distance to the wall, and n c is the smallest value of n at which the inner and
outer viscosity is equal.
The inner viscosity is
2
( µ t ) i = ρl ω (4-37)
where
+ +
–y ⁄ A
l = kn ( 1 – e ) (4-38)
+ ⎛ ρ w τ w⎞
y = ⎜ ---------------
-⎟ n (4-39)
⎝ µw ⎠
and
∂u j
ω i = ε ijk (4-40)
∂ xk
n max F max
2 (4-42)
and C wk n max 2 2 2 2 2 2 /F max
( u + v + w )max – ( u + v + w ) min
The term n max is the value of n corresponding to the maximum value of F , F max , where
+ +
– n /A
F(n) = n ω (1 – e ) (4-43)
and
4-28 FINE™
Mathematical Model Flow Model
–1
F Kleb = 1 + 5.5 ( nC 6
Kleb /n max )
(4-44)
+
The constants used are hard coded and equal: A = 26 , C wk = 1. , C cp = 1.6 , k = 0.41 ,
C kleb = 0.3 , K = 0.0168 .The turbulent Prandtl number, needed to calculate the turbulent con-
ductivity, is accessible through the expert parameter PRT (see section 3-3.1.2).
The expert parameter IATFRZ allows to control the interaction of the model with the multigrid sys-
tem. When IATFRZ is set to 0, the model is applied separately on all grid levels. When it is set to 1,
the model is only applied on the finest grid and the turbulent viscosity is restricted (through the
restriction operator) to the coarser grids where the turbulent viscosity is frozen. It is possible to
freeze the turbulent viscosity field on all grid level by setting IATFRZ to 2. The value can be auto-
matically changed by using a criteria based on a maximum iteration number or on a minimum con-
vergence. The maximum iteration number is specified by the expert parameter ITFRZ. The
minimum reduction of order of magnitude is set in the expert parameter RESFRZ.
4-3.6.2 Spalart-Allmaras
The Spalart-Allmaras turbulence model is a one equation turbulence model which can be consid-
ered as a bridge between the algebraic model of Baldwin-Lomax and the two equation models. The
Spalart-Allmaras model has become quite popular in the last years because of its robustness and its
ability to treat complex flows. The main advantage of the Spalart-Allmaras model when compared
to the one of Baldwin-Lomax is that the turbulent eddy viscosity field is always continuous. Its
advantage over the k-ε model is mainly its robustness and the lower additional CPU and Memory
usage.
The principle of this turbulence model is based on the resolution of an additional transport equation
for the eddy viscosity. The equation contains an advective, a diffusive and a source term and is
implemented in a non conservative manner. The implementation is based on the papers of Spalart
and Allmaras (1992) with the improvements described in Ashford and Powell (1996) in order to
avoid negative values for the production term ( S˜ in Eq. 4-50).
The turbulent viscosity is given by
ν t = ν̃f v1 (4-45)
3
χ
f v1 = ------------------
3
- (4-46)
χ + c v1
with χ is the ratio between the working variable ν̃ and the molecular viscosity ν ,
ν̃
χ = --- (4-47)
ν
The turbulent working variable obeys the transport equation
∂ν̃ 1
------ + V ⋅ ∇ν̃ = --- { ∇ ⋅ [ ( ν + ( 1 + c b2 )ν̃ ) ∇ν̃ ] – c b2 ν̃ ∆ν̃ } + Q (4-48)
∂t σ
FINE™ 4-29
Flow Model Mathematical Model
where
˜
ν̃P ( ν̃ ) = c b1 Sν˜ (4-50)
2
ν̃
ν̃D ( ν̃ ) = c w1 f w ⎛ ---⎞ (4-51)
⎝ d⎠
ν̃
S˜ = Sf v3 + ----------
-f ;
2 2 v2
(4-52)
κ d
1 ( 1 + χf v1 ) ( 1 – f v2 )
f v2 = ------------------------------3- ; f v3 = -------------------------------------------- (4-53)
( 1 + χ ⁄ c v2 ) χ
where d is the distance to the closest wall and S the magnitude of vorticity.
1-
6 --
⎛ 1 + c w3 ⎞ 6
fw = g ⎜ ------------------
6 6
-⎟ (4-54)
⎝ g + c w3⎠
with
6 ν̃
g = r + c w2 ( r – r ) ; r = -------------
- (4-55)
˜S κ 2 d 2
The constants arising in the model are
2
c w1 = c b1 ⁄ κ + ( 1 + c b2 ) ⁄ σ , c w2 = 0.3 , c w3 = 2. , c v1 = 7.1 , c v2 = 5.
The equation Eq. 4-48 is solved with the appropriate boundary conditions:
on solid wall ν̃ = 0 , along the inflow boundaries the value of ν t is specified ( ν̃ is obtained by
using a Newton-Raphson procedure to solve Eq. 4-45) and along the outflow boundaries it is
extrapolated from the interior values.
4-30 FINE™
Mathematical Model Flow Model
a) General Formulation
In the k – ε turbulence model two additional transport equations for the turbulent kinetic energy, k ,
and the turbulent dissipation rate, ε , are solved. In EURANUS, 4 linear and 2 non-linear models
are currently used.
In the k – ε turbulence model two additional differential equations need to be solved respectively
for k and ε . These additional equations can be put in the following general form:
∂ρk µ ⎧ ⎫
--------- + ∇•⎛ ρwk – µ + -----t ∇k⎞ = – ⎨ ρw″ ⊗ w″ S ⎬ – ρε (4-56)
∂t ⎝ σk ⎠
⎩ ⎭
∂--------
ρε̃- µ 1⎛ ⎧ ⎫ ⎞
+ ∇•⎛ ρwε̃ – µ + -----t ∇ε̃⎞ = – --- ⎜ C ε1 f 1 ⎨ ρw″ ⊗ w″ S ⎬ + C ε2 f 2 ρε̃⎟ + E (4-57)
∂t ⎝ σε ⎠ T⎝ ⎩ ⎭ ⎠
where S is the mean strain tensor and – ρw″ ⊗ w″ the turbulent Reynolds stress tensor.
ε̃ = ε – D (4-58)
µ t = ρC µ f µ kT (4-59)
b) Linear Models
In the linear models, the turbulent Reynolds stress tensor is related to the mean strain tensor in a lin-
ear way.
2
( – ρ w″ ⊗ w″ )ij = 2µ t S ij – 2
--- ( ∇w )δ ij – --- ρkδ ij
3 3
(4-60)
1 ⎛ ∂w̃ ∂w̃ ⎞
S ij = --- ⎜ i + j⎟
2 ⎝ ∂ xj ∂ x i ⎠
FINE™ 4-31
Flow Model Mathematical Model
The constants or functions Cµ, Cε1, Cε2, σk, σε, fµ,f1, f2, D, E and T are model dependent and they
2 0.5
k k y
are defined in the Table 4-3 where Re t = ------ and Re y = -----------
νε ν
T k ⁄ ε̃ k⁄ε k ⁄ ε̃ k ⁄ ε + (ν ⁄ ε)
0.5
2 2
D 2νk ⁄ y 0. 2ν ( ∇ k ) 0.
+
– 0.5y
2µε˜ e
–---------------------------- 2 2
E - 0. 2νµ t ( ∇•S ) νµ t ( ∇•S )
2
y
2 0.5
kw uτ ⁄ Cµ /
0. 0. 0.
DNS
3
εw u τ ⁄ ( κy ) / 2
0. 0. 2ν ( ∇ k )
DNS
The constants Cµ, Cε1, Cε2, σk, σε can however be changed by the user (respectively the expert param-
eters CMU, CE1, CE2, SIGK, SIGE in the Computation Steering/Control variables page. The tur-
bulent Prandtl number, needed to calculate the turbulent conductivity, is also an expert parameter
(PRT).
4-32 FINE™
Mathematical Model Flow Model
Two variants of the high Reynolds k-ε with wall functions (section d) Solid Boundary
and Wall Functions) exist:
• If the expert parameter INEWKE=10 (Default), the wall functions are derived from DNS
(Direct Numerical Simulation) curve fitting. The k-ε model is in that case derived from the
Yang-Shih model. This new model, named extended wall functions, allows to obtain accurate
results on fine mesh contrary to the standard high Reynolds k-ε model.
• if the expert parameter INEWKE=0, standard wall function are applied and the y+ value of the
near wall cell should be greater than 20.
The Chien and the Launder-Sharma models belong to the same family (zero value of ε
at the wall) while the standard model and Yang-Shih model belong to a second family
(non zero value of ε at the wall). Launching the restart is allowed as long as the user
sticks to the same family.
i
4 fµ ⎛ γRi t ⎞ 2
– ρw″ ⊗ w″ = --- ----------------------
- ⎜ 1 – ---------------------
i
-⎟ kTS – --- ρkI
3A + f s⎝ 1 + γRi ⎠ 3
1 µ t
C τ2 2 2
kT' ⎛⎝ 2S – SΩ + ΩS – --- { S }⎞⎠
2 2
– -----------------------------
3 3
(4-61)
A2 + s + ω 3
C τ3 2 2
kT' ⎛ 2S + SΩ – Ω S – --- { S }⎞
2 2
– -----------------------------
3 3 ⎝ 3 ⎠
A2 + s + ω
1⁄3
A2
T' = 1 ⁄ ( f t1 ⁄ T w + f t2 ⁄ T ) ; T w = ------------------
-
2
2{S }
1 f t1
f t1 = 1 – ---------------------------------------------2- ; f t2 = 1 – --------------------------
-
2 1 + Re t ⁄ 70
( 1 + 2 { S }ν ⁄ ε )
2 2 2
s = T' 2 { S } ; ω = T' – 2 { Ω } ; Ri t = – ωs – ω
FINE™ 4-33
Flow Model Mathematical Model
• T’ is replaced by T
The production term becomes:
i
4 1 ⎛ γRi t ⎞ 2
– ρw″ ⊗ w″ = --- -------------- ⎜ 1 – ---------------------
i
-⎟ kTS – --- ρkI
3 A1 + s ⎝ 1 + γRi ⎠ 3
t
C τ2 2 2
kT ⎛ 2S – SΩ + ΩS – --- { S }⎞
2 2
– ----------------------------- (4-62)
3 3 ⎝ 3 ⎠
A2 + s + ω
C τ3 2 2
kT ⎛ 2S + SΩ – Ω S – --- { S }⎞
2 2
– -----------------------------
3 3 ⎝ 3 ⎠
A2 + s + ω
2
s = T 2{ S }
(4-63)
2
ω = T –2 { Ω }
Before launching the non-linear model, it is recommended to apply first the linear
model and restart from this solution. This procedure will provide a good initial solution
for the non-linear model and prevent the non-linear terms to cause divergence of the
code at the early stage.
Since the non-linear model is from Yang-Shih, the restart can be done only from the
Yang-Shih model (linear or not) and the Standard model with wall functions.
For the models that do not use wall functions (Chien, Launder Sharma, Yang & Shih
and Low Reynolds non linear model), the profile of the boundary layer is directly
inferred from the input values at the wall. Consequently, particular attention must be paid
to the grid refinement in this region in order to capture the viscous sublayer. A typical
+
value is given by 1 < y < 10 at the first node from the wall.
The wall functions are used to mimic the presence of the walls by reflecting the effect of the steep,
non-linear variations of the flow properties through the turbulent boundary layer. They define the
shear stress and the heat flux on the cell faces lying on solid boundaries as well as the values of k
and ε in the vicinity of the wall.
The wall functions are applied in the Standard (see kw and εw in Table 4-3 ) and in the High Rey-
nolds non linear k-ε models. For the extended wall functions model, the wall functions for k and ε
are fitted with polynomials to the DNS data.
4-34 FINE™
Mathematical Model Flow Model
d.1) Velocity
For smooth walls:
The wall function in the viscous sublayer is given by:
u +
----- = y (4-64)
uτ
yu
with y = -------τ- , the dimensionless normal distance from the walls and u τ =
+
τ wall ⁄ ρ , the wall
ν
friction velocity directly related to the wall shear stress.
While in the turbulent layer, the wall function becomes:
u 1 +
----- = --- ln y + B (4-65)
uτ κ
where κ is the von Karman constant (default value 0.41) and B another constant (default value
5.36). They can be modified by the user in the Boundary Condition page.
For rough walls:
The wall function is given by (no viscous sublayer considered):
u 1 y–d
----- = --- ln -------------0- + B (4-66)
uτ κ k0
where κ and B are defined just above and k0 and d0 are constants asked for each rough wall. The
constant k0 is the equivalent roughness height while d0 is known as the height of the zero displace-
ment plane, defined by u ( k 0 + d 0 ) = 0. (if B=0). For small roughness elements, d 0 = 0 . For high
roughness elements, the flow behaves as if the wall was located at a distance d0 from the real wall
position.
ks
k 0 = ------ where ks is the average roughness height
30
+
+ ks y
For rough walls, the constant B is usually zero (when ks+ > 70 with k s = ------------ ). Otherwise the
y
value of B is evolving as presented in below graph.
+
s
FINE™ 4-35
Flow Model Mathematical Model
When a rough wall is specified, the distance from the wall to the cell-centre of the first
inner cell should be bigger than k0 + d0. In general, the first inner cell should be located
within the fully turbulent layer.
d.2) Temperature
Similarly, for isothermal walls, the following laws are considered in the code:
+ +
T = Pr y in the viscous sublayer (4-67)
+ Tw – T
where T = ---------------- .
Tτ
q w = ρCpuτ T τ (4-69)
A complete description of the available thermal boundary conditions at walls is provided in section
8-5.3.
The "law of the wall" distributions of velocity, temperature and other variables are
assumed to prevail across the boundary layer. They are imposed at the node of a single
+
grid cell for which the user is advised to check that 20 < y < 100 . As a result, if the first
cell node from the wall is placed too close, calculation is conducted in the viscous sub-
layer and the law is no longer valid. Similarly, if the first node is placed too far away,
then a discrepancy may exist between the profiles and their assumed shape.
e) Numerical Concerns
This paragraph provides a description of the numerical method used to deal with the k and ε equa-
tions.
e.1) Choice of the initial and boundary conditions
Initial values for k and ε must be specified by the user, as well as the inlet boundary condition.
The specified values at the inlet boundary can have a strong effect on the flow-field prediction. Sev-
eral methods are proposed in section 4-3.5.4 to help the user in the choice of suitable values.
e.2) Compressibility correction
The compressibility correction is present only in the k-ε model of Chien and is useful only for high
Mach number flows (M>3). Two kinds of compressibility effects are implemented:
1) the contributions from the modelled pressure gradient-velocity term and
2) the compressible dissipation.
They can be activated by using respectively the expert parameters ICOPKE or ICODKE.
The first compressibility correction is associated with the modelling of pressure gradient-velocity
term in the equation of the turbulent kinetic energy, and, according to Zhang et al. (1991), leads to
the source term Rk:
4-36 FINE™
Mathematical Model Flow Model
µ t ∂ρ ∂p
R k = – ----------- , (4-70)
ρ σp ∂ xi ∂ xi
2
µ t ∂ρ ∂p
D' = D + ----------- . (4-71)
ρ σp ∂ xi ∂ xi
3
Similarly a term R ε can be added to the ε equation. After Vandromme (1991) one has:
µ t ε ∂ρ ∂p
R ε = – C 3 ----------- -- , (4-72)
ρ σ k ∂ xi ∂ xi
2
p
µ t ε ∂ρ ∂p
E' = E – C 3 ----------- -- . (4-73)
ρ σ k ∂ xi ∂ x i
2
p
The above terms are included in the k – ε equations if the expert parameter ICOPKE is set to 1; if
ICOPKE=0 (Default) they are set to zero.
The empirical constants in these equations are two expert parameters (SIGRO and C3), with default
values:
The second correction concerns the modelling of compressible dissipation. The physical foundation
for this correction is the observation of decreased turbulent viscosity with increased Mach number
(Nichols, 1990). The type of compressible dissipation is controlled by the expert parameter
ICODKE.
For ICODKE=0 no correction for compressible dissipation is done.
In the correction of Sarkar (1989) (ICODKE=1), the dissipation of k is expressed as the sum of
solenoidal ( ε ) and dilatational ( ε d ) components. The dissipation in the source term of the k -equa-
tion (Eq. 4-56) is then corrected according to:
2
ε corr = ε ( 1 + αM t ) (4-75)
2 k
M t = ----2- (4-76)
a
where a is the speed of sound.The constant α is an expert parameter (ALF), with default value
=0.5.
Nichols (1990) developed a model that corrects the production term in the turbulent kinetic energy
equation (ICODKE=2):
⎛ – ρw″ k
⊗ w″ S⎞ = – ρw″ ⊗ w″ S ⎛ 1 – C p1 ( γ – 1 )M ----2-⎞ (4-77)
⎝ ⎠ ( corr ) ⎝
a⎠
FINE™ 4-37
Flow Model Mathematical Model
where a is the local speed of sound, γ is the specific heat ratio and C p1 an empirical constant (CP1
in the list of expert parameters available on the Control Variables page, with default value =4.).
This correction improves the prediction of the turbulent production level by directly taking the
effect of the Mach number (M) into account.
e.3) Limiters for the turbulence variables
Successive limitations are set for the turbulence variables to ensure physical values and the stability
of the calculation. Indeed, the term sources in the k-ε equations can vary strongly especially at the
beginning of the computation and lead to unphysical values and divergence.
1. To control the stability of the calculation, the expert parameter MAVRES, denoted here ζ , is
used This parameter limits the variation of the turbulence variables for each Runge-Kutta stage.
n+1
Denoting the new value of k or ε as q , the following restrictions are imposed when MAVRES
> 0:
n+1 n+1 n
q = max ( q , EPS,ζq ) (4-78)
n+1 n+1 n
q = min ( q , ( 2 – ζ )q ) (4-79)
where EPS is an expert parameter with a small value, default 1E-28, also used to avoid possible
divisions by zero.
The restriction defined in Eq. 4-79 is not applied to the turbulent dissipation rate (ε) when this tends
to increase, in order to avoid excessive values of the turbulent viscosity.
If MAVRES = 0, only the restriction defined in Eq. 4-78 is applied (avoiding negative values)
If MAVRES < 0, the restriction is:
n+1 n
n+1 n (1 + ζ)(q –q )
q – q = --------------------------------------------- (4-80)
⎛ n + 1 n
q –q ⎞
⎜ 1 + -------------------------
n
-⎟
⎝ q ⎠
2. A similar system is applied to the multigrid corrections with the expert parameter MAVREM,
denoted below as ζ :
n+1 n n
If MAVREM > 0: q – q < ( 1 – ζ )q
If MAVREM = 0: restriction to positive values
n+1 n
n+1 n (1 + ζ )( q –q )
If MAVREM < 0: q – q = ---------------------------------------------
⎛ n + 1 n
q –q ⎞
⎜ 1 + -------------------------
n
-⎟
⎝ q ⎠
3⁄2
k
3. The expert parameter LTMAX is used to limit the turbulence length scale L turb = --------- to a
ε
value compatible with the domain scale.
If Lturb>LTMAX and MAVRES > 0, the value of ε is set to respect the maximum length scale.
If Lturb>LTMAX and MAVRES <= 0, the values of k and ε are set to a small value EPS.
The default value for LTMAX is 1E+6.
4-38 FINE™
Mathematical Model Flow Model
4. An automatic control of the turbulent length scale is possible by setting the expert parameter
LTMAX to a negative value (LTMAX = -1). In this case, the following length scale control is pro-
posed:
with le max the maximum equilibrium length scale in the computation, defined by
–3 ⁄ 4
le max = κd max C µ (4-82)
where d max is the maximum distance to the wall among all grid points of all blocks. The factor 5 in
Eq. 4-81 is set to avoid to clip the length scale too early.
For the update of turbulent variables (previous paragraph), LTMAX is replaced by 5le max
5. Another way to control the turbulent length scale is the Yap’s modification. It is applied by set-
ting the expert parameter IYAP to 1. In this case, the expert parameter LTMAX should be set to its
default value (1.E+6) to avoid conflicts.
Yap’s modification consists of adding to the epsilon equation a source term that control the turbu-
lent length scale. It can be applied to all k-ε models. The source term to be added writes:
l l 2ε
S YAP = 0.83max ⎛ --- – 1, 0⎞ ⎛ ---⎞ --- (4-83)
⎝l ⎠ ⎝l ⎠ T
e e
where l and l e are respectively the turbulent length scale ( l = kT ) and the equilibrium
–3 ⁄ 4
length scale ( l e = κdC µ ).
A weakness of Yap’s correction is that it might conflict with Inlet boundary conditions if these are
set without respecting a turbulent length scale criteria.
6. The expert parameter TEDAMP is another parameter to control stability. This parameter prevents
the factor 1 ⁄ T in the ε equation from being singular as k tends to zero (see Eq. 4-57 and the rela-
tionship between k and T in Table 4-3 ). A minimum of TEDAMP multiplication factor of the clip-
ping value (expert parameter EKCLIP) is used for k .
The same parameter is used to damp 1/ε coming from T in the calculation of the turbulent viscos-
ity, Eq. 4-59: a minimum of TEDAMP multiplication factor of the clipping value (expert parameter
EPCLIP) is used for ε .
If TEDAMP<=0, the minimum values used for k and ε correspond to the expert parameter EPS.
The default value of TEDAMP is -1.
Irrespective of the value of TEDAMP, this strategy always avoids the use of possible negative val-
ues of k in the source term and of ε in the turbulent viscosity.
FINE™ 4-39
Flow Model Mathematical Model
7. Another control on the stability is available through the parameter GAMMAT which introduces
the turbulence time scale Tturb in the computation of the local time step:
1 1 GAMMAT
--------- = ----- + -------------------------- (4-84)
∆t kε ∆t T turb
The default value of GAMMAT is set to 10.
8. Another important consideration is the amount of turbulence production compared to its dissipa-
·
tion, Prod ⁄ ( ρε ) . This quantity should stay within physically admissible ranges so that the turbu-
lence production remains limited. The parameter PRCLIP allows the control of this ratio. As an
example, Prod ⁄ ( ρε ) ≈ 1 for homogeneous core flows and thin shear layers, whereas
Prod ⁄ ( ρε ) < 1 for separated shear layers and wakes. Flows of high shear, such as jets in cross
·
flow, can exhibit much higher values of Prod ⁄ ( ρε ) . An upper bound of PRCLIP=50 is recom-
mended to the user. However, this control is removed by the use of a linearised turbulent production
model introduced in all linear k-ε models. This simple modification is relevant for impinging flows,
for which the turbulence models are well known to overestimate the turbulent production of kinetic
energy. It corresponds to a linearisation of the production term when the strain rate is large when the
expert parameter LIPROD is set at 1.
9. The expert parameters EKCLIP and EPCLIP are respectively the minimum allowable values of
the turbulent kinetic energy (k) and the turbulent dissipation rate (ε). They avoid non-physical neg-
ative values. Moreover the minimum value of k (EKCLIP) allows to maintain a minimum residual
turbulence in the domain. The default value is 1.E-5.
The minimum value of ε is chosen so that the computed value of the turbulent viscosity is reasona-
ble. With a default value set to 1.E-5, the turbulent viscosity has a value of the order 1.E-6.
10. The input parameter KEGRID allows the use of a particular full multigrid strategy for k-ε com-
putations. The Baldwin-Lomax model is used on the coarse grids during the full multigrid stage and
the k and ε initial solution is obtained from the Baldwin-Lomax model solution.
KEGRID corresponds to the last grid level on which the algebraic model is used. For example, if
the computation is done with 4 levels of multigrid and KEGRID is equal to 2, the Baldwin-Lomax
model is used on the 333 (or level 4), 222 (or level 3) and 111 (or level 2) and the k-ε computation
is initiated when transferring to the finest grid. If KEGRID is set to 3, the k-ε computation will be
initiated when transferring to 222. If KEGRID is greater than the number of multigrid levels, the
computation will be started on the coarsest level with k-ε.
11. The expert parameter IKENC allows solving k-ε equations with a non-conservative approach.
This settings avoids the appearance of an artificial source term in the k-ε equations during the con-
vergence process. This artificial term comes from the residuals of the equation of the mass conser-
vation that can be non-negligible at the beginning of the convergence process. This approach is new
and not sufficiently tested to ensure that the loss of conservation is not excessive in all cases.
12. The expert parameter IKELED allows to avoid an excessive artificial dissipation of the standard
central scheme into the boundary layer. Indeed this scheme can lead to a non-physical relaminarisa-
tion in some specific test cases. An alternative computation of the artificial dissipation, based on the
Local Extrema Diminishing (LED) version of the Jameson-Schmidt-Turkel treatment has been
implemented. It is active when the expert parameter IKELED is set to 1.
All the above mentioned limiters are very important for the robustness of k-ε compu-
tations. With the exception of the length scale LTMAX all limiters are given appropriate
defaults and we do not recommend to modify them. For LTMAX we recommend to set it
to ten times the geometrical length scale.
4-40 FINE™
Mathematical Model Flow Model
4-3.6.4 References
Ashford G.A. , and Powell K.G. 1996, "An unstructured grid generation and adaptative solution
technique for high-Reynolds number compressible flow", VKI (Von Karman Institute) Lecture
Series 1996-06.
Baldwin B., Lomax H. (1978) 'Thin Layer Approximation and Algebraic Model for Separated Tur-
bulent Flows', AIAA-78-257.
Chien K.Y. (1982) 'Predictions of Channel and Boundary-Layer Flows with a Low-Reynolds-
Number Turbulence Model', AIAA J., Vol. 20, No. 1.
Hakimi N. (1997) ’Preconditioning methods for time dependent Navier-Stokes equations’, PhD
Thesis, Dept of Fluid Mechanics, Vrije Universiteit Brussel.
Jameson A., Schmidt W., Turkel E. (1981) 'Numerical Solutions of the Euler Equations by Finite
Volume Methods using Runge-Kutta Time-Stepping Schemes', AIAA-81-1259.
Jameson A., 1995, "Positive schemes and shocks modelling for compressible flows", International
Journal for Numerical Methods in Fluids, Vol 20, pp 773-776.
Launder B.E. and Sharma B.I., 1974,"Application of the Energy-Dissipation Model of turbulence to
the Calculation of Flow Near a Spinning Disc", Letters in Heat and Mass Transfer, Vol. 1, pp. 131-
138.
Launder B.E. and Spalding D.B., 1974, "The numerical computation of turbulent flow", Comput.
Methods App. Mech. Eng., vol. 3, pp. 269-289.
Nichols R.H. (1990) 'A Two-Equation Model for Compressible Flows', AIAA paper 90-0494.
Sarkar S., Erlebacher G., Hussaini M.Y., Kreiss H.O. (1991) 'The analysis and modelling of dilata-
tional terms in compressible turbulence', J Fluid Mech, Vol.227, pp.473-493.
Spalart P.R. and Allmaras S.R. (1992),"A one equation turbulence model for aerodynamic flows",,
AIAA 92-0439
Vandromme D. (1991) 'Turbulence Modelling for Compressible Flows and Implementation in
Navier-Stokes Solvers', VKI Lecture Series 1991-02.
Yang Z. and Shih T.H., 1993, "A k-e model for turbulence and transitional boundary layer", Near-
Wall Turbulent Flows, R.M.C. So., C.G. Speziale and B.E. Launder(Editors), Elsevier-Science Pub-
lishers B. V., pp. 165-175.
Zhang H.S., So R.M.C., Speziale C.G., Lai Y.G. (1991) 'A Near-Wall Model for Compressible Tur-
bulent Flow', ICASE Report 91-82.
When the gravity is taken into account in the Navier-Stokes equations, the source terms ρg
and ρ ( g ⋅ V ) are respectively introduced in the momentum and energy conservation equations
where ρ is the density, g the gravity vector and V the velocity vector.
FINE™ 4-41
Flow Model Mathematical Model
If the fluid is a liquid, the density is constant and you do not have any influence of the pressure or
temperature on the flow by interaction with the gravity. To simulate these interactions, one can use
the Boussinesq approximation in the gravity source terms. With this approximation, the density is
developed in the first order of the MacLaurin series. It becomes:
with α the compressibility and β the dilatation coefficients specified in the fluid properties.
This variation of the density is only applied to the gravity source term (Boussinesq approximation).
Otherwise, the density is kept constant in the conservation equations. The value of ρ ref is the
characteristic density specified in the Flow Model page. p ref and T ref are the reference values in
the Flow Model page.
When the gravity is taken into account, a reference pressure must be specified at a ref-
erence altitude to add the hydrostatic pressure in the initial pressure field. The reference
pressure is the reference value in the Flow Model page. The reference altitude is speci-
fied in the expert parameter IREFPT.
This option is only implemented for central schemes. Consequently, the upwind spa-
tial scheme (Numerical Model page) is not available if the preconditioning option is cho-
sen.
4-42 FINE™
Mathematical Model Flow Model
For steady state applications solved by time marching algorithms the time derivatives of the
unknowns arising in the flow equations are of no physical meaning and can thus be modified with-
out altering the final steady state solution. The idea of preconditioning precisely uses this property
and consists of multiplying the time derivatives of the dependent variables with a matrix called pre-
conditioning matrix. The main property of this matrix is to remove the stiffness of the eigenvalues.
In addition, reduced flow variables such as the dynamic pressure and the dynamic enthalpy are
introduced, reducing drastically the round off errors at low Mach numbers. The acoustic wave
speed c is replaced by a pseudo-wave speed c’ of the same order of magnitude as the fluid speed. To
be efficient the selected preconditioning matrix should be valid for inviscid computations as well as
for viscous computations with heat transfer.
The preconditioning methodology developed (Hakimi, 1997) is of sufficient generality and can
treat any type of fluids including perfect gases and incompressible Newtonian and non Newtonian
fluids. On the numerical level, the solution procedure including space discretization, time integra-
tion and boundary conditions have been adapted to the new transient behaviour of the conservation
equations.
The low speed preconditioner has been validated for inviscid flows, viscous flows, turbulent flows
and unsteady flows. Efficient convergence rates and accurate solution have been obtained for Mach
numbers from M=0.1 to M=10-6, Reynolds numbers from Re=106 to 10-6 and aspect ratios from 1
to 2000.
∂p- ---------
1- -----
---- ∂ρu ∂ρv ∂ρw
2 ∂t
+ + --------- + ---------- = 0
β ∂x ∂y ∂z
αu- ∂p ∂τ xx ∂τ yx ∂τ zx
------ ------ + ∂p
------ + ρ du ------ = ---------
- + ---------- + ----------
2 ∂t dt ∂x ∂x ∂y ∂z
β
αv ∂p dv ∂p ∂τ xy ∂τ yy ∂τ zy (4-86)
- + ---------- + ---------- – ρg
------ ------ + ρ ------ + ------ = ---------
2 ∂t dt ∂y ∂x ∂y ∂z
β
αw ∂p dw ∂p ∂τ xz ∂τ yz ∂τ zz
-------- ------ + ρ ------- + ------ = ---------
- + ---------- + ---------
2 ∂t dt ∂z ∂x ∂y ∂z
β
de
ρ ------ + p ⎛ ∇ ⋅ V⎞ = ∇ ⋅ ⎛ k∇T⎞ + ε v
dt ⎝ ⎠ ⎝ ⎠
V = ( u, v, w ) is the velocity vector. At this stage the eigenvalues as well as the eigenvectors associ-
ated with the preconditioned method defined by Eq. 4-86 are already completely defined.
FINE™ 4-43
Flow Model Mathematical Model
– 1 ∂Q
∫∫∫Γ ------- dΩ +
∂t ∫ ∫ F ⋅ dS = ∫ ∫ ∫ ST dΩ (4-87)
Ω S Ω
with
T
Q = ( p g, u, v, w, E g, ρk, ρε̃ ) . (4-88)
The variable p g = p – p ref is the gauge pressure and the variables E g is the total gauge energy.
2
V
E g = C v ( T – T ref ) + ----- (4-89)
2
⎛ ⎞
⎜ ⎟
⎜
1 0 0 0 0 0 0 0 . . 0 ⎟
-----
⎜ 2 ⎟
β
⎜ ⎟
⎜ ( 1 + α )u ⎟
⎜ ---------------------
2
ρ 0 0 0 0 0 0 . . 0 ⎟
⎜ β ⎟
⎜ ⎟
⎜ ( 1 + α )v ⎟
–1 ⎜
--------------------
- 0 ρ 0 0 0 0 0 . . 0 ⎟
Γ = β
2 (4-91)
⎜ ⎟
⎜ ( 1 + α )w ⎟
⎜ ---------------------
2
- 0 0 ρ 0 0 0 0 . . 0 ⎟
⎜ β ⎟
⎜ ⎟
2
⎜ αv + Eg ⎟
⎜ 2
- ( – 1 )∗ 0 0 0 ρ 0 0 0 . . 0
--------------------- ⎟
⎜ β ⎟
⎜ ⎟
⎜ 0 0 0 0 0 1 0 0 . . 0 ⎟
⎝ 0 0 0 0 0 0 1 0 . . 0 ⎠
4-44 FINE™
Mathematical Model Flow Model
λ 1 ,2, 3 = V ⋅ n
2 2 2 2 (4-92)
1 ⎛ β⎞ ⎛ ⎛ β ⎞⎞ ⎛ 2 ( V ⋅ n) ⎞ 2
λ 4 ,5 = --- V ⋅ n ⎜ 1 – α + ----2-⎟ ± ⎜V ⋅ n ⎜ 1 – α + ----2-⎟ ⎟ + 4 ⎜ n – -----------------
2
-⎟ β
2 ⎝ c ⎠ ⎝ ⎝ c ⎠⎠ ⎝ c ⎠
The two parameters α and β are chosen so that the stiffness of the eigenvalues is minimised at low
speed.
The corresponding eigenvalues for incompressible fluids are:
λ 1 ,2, 3 = V ⋅ n
(4-93)
1 2 2 2
λ 4 ,5 = --- V ⋅ n ( 1 – α ) ± ( V ⋅ n ( 1 – α ) ) + 4n β
2
Compared to preconditioned system of compressible fluids, the first three eigenvalues are unchanged
and the eigenvalues of the acoustic waves can be obtained by setting in Eq. 4-92 the speed of sound to
infinity. We note in particular that the acoustic eigenvalues are always of opposite sign showing that
the flow remains subsonic with respect to the pseudo-sonic speed. As for the compressible precondi-
tioner the two parameters α and β are chosen so that the stiffness of the eigenvalues is minimised at
low speed.
Note that if a k-ε model is considered, the corresponding eigenvalues will be:
Pref ,T ref are accessible through the Flow Model page and are referred to as reference pressure and
reference temperature.
A value of Tref of the order of magnitude of the expected temperature field will reduce
the machine round-off errors influence on the temperature field.
The value of Pref fixes the absolute level of the pressure in the field. This value is used in
the equation of state for a compressible fluid and in the second order artificial dissipation
term. A value of Pref close to the real pressure level is therefore recommended.
The parameter α is accessible in the ‘Expert parameters’ list of the Control variables page as ALP-
HAP. According to several computations accomplished with the central scheme, good convergence
rates were obtained for values of α lying in the range [-1,1]. So far the best value is found to be α=-1.
This optimal value is not surprising since it increases the spectral radius and therefore adds an addi-
tional amount of artificial dissipation.
Τhe preconditioning parameter β is imposed by the user in the Numerical Model page through a coef-
ficient β∗ and a characteristic velocity U ref such that
FINE™ 4-45
Flow Model Mathematical Model
2 2
β = β∗ U ref (4-95)
U ref is representative of the maximum velocity in the flow field. For example, in external flows,
Uref could be taken as the free stream velocity whereas in internal flows it could be taken as the
maximum inlet velocity. For the user’s convenience, the constant Uref is however always set equal
to the reference velocity specified in the Flow Model page and used in the calculation of the Rey-
nolds Number.
The default value of β∗ is 3. Based on our experience with the presented low speed preconditioner
the parameter β∗ can be taken of order unity for inviscid flow computations. For viscous computa-
tions associated with Reynolds numbers greater than about Re=1000, a constant value β∗ of order
unity is also adequate. If convergence difficulties are encountered at the very beginning of a compu-
tation, it is recommended to increase the value of the parameter β∗. Remember however that a too
large value of β∗ will introduce excessive artificial dissipation into the solution.
For lower Reynolds number the parameter β∗ has to vary in order to preserve numerical stability and
a good convergence rate. The parameter β∗ should increase as the Reynolds number decreases and
may vary over several orders of magnitude (see Figure 4.3.8-4).
If the local velocity scaling option is activated in the Numerical Model page, another definition of β
is used:
2 2 2
β = β∗ max ( Uref, Uloc ) (4-96)
10000
1000
100
10
1
1 10 100 1000 10000 Re
For high Reynolds flows it is suggested to choose β among the following values:
• β=3
• β=30
• β=300
Too small or too big values of β may lead to divergence and to a too dissipative solution (if β is too
high)
4-46 FINE™
Characteristic and Reference Values Flow Model
4-3.8.6 References
Chorin, A. J. (1967) ’A numerical method for solving incompressible viscous flow problems’, J. of
Comput. Phys., Vol 2, pp. 12-26.
Hakimi N. (1997) ’Preconditioning methods for time dependent Navier-Stokes equations’, PhD
Thesis, Dept of Fluid Mechanics, Vrije Universiteit Brussel.
FINE™ 4-47
Flow Model Characteristic and Reference Values
4-48 FINE™
CHAPTER 5: Rotating Machinery
5-1 Overview
The Rotating Machinery page contains the following thumbnails:
• Rotating Blocks defining for each block typical information concerning cylindrical cases. This
thumbnail is only available when the mesh is cylindrical (see Mesh/Properties... to see
whether the mesh is cylindrical or Cartesian). See section 5-2 for more detail on this thumb-
nail.
• Rotor/Stator defining the properties of each rotor/stator interface. This thumbnail is only avail-
able when the mesh contains Rotor/Stator interfaces. Rotor/Stator interfaces are defined in
IGG™ as ROT boundary condition. For more detail on this thumbnail see section 5-3.
• Throughflow Blocks giving access to two-dimensional throughflow Euler simulation and this
is only accessible for non-viscous flow (Euler selected as mathematical model on the Flow
Model page). See Chapter 6 for more detail on the options available under this thumbnail.
FINE™ 5-1
Rotating Machinery Rotating Blocks
When selecting the Rotating Blocks thumbnail the page appears as shown in Figure 5.2.0-1. On the
left a list of all blocks is displayed. It is possible to group the blocks that have the same properties
on this page:
1. Select several blocks in the list of blocks by simply clicking on them. Clicking on a block dese-
lects the currently selected block(s). It is possible to select several blocks situated one after
another in the list by clicking on the first one and holding the left mouse button while selecting
the next. To select a group of blocks which are not situated one after another in the list, click on
each of them while keeping the <Ctrl> key pressed.
2. Click on the Group button. A text box will appear just below the Group button. Place the cursor
in the text box and type the name of the group and press <Enter>. The name of the group will
appear on the list of blocks in red with a plus sign on the left of the name. To see the blocks that
are included in a certain group double click on the name of the group or click on the + symbol in
front of the name.
3. To remove a group of blocks select the group and click on the Ungroup button. The group will
be removed and the blocks that were in this group will be listed again in the list as separate
blocks.
On the right of the list with blocks the information is given for the selected block or group of
blocks. If multiple blocks are selected the information is only displayed for the first selected block.
For each block the current and maximum number of nodes in the three grid directions is given. Four
items need to be defined by the user: the streamwise direction, the spanwise direction, the azimuthal
direction and the rotational speed:
— The streamwise, spanwise and azimuthal directions allow to determine the block orientation.
Remark that if the AutoGrid module is used for the grid generation, the streamwise direction
is automatically taken as being the k-direction, the spanwise direction as the j-direction and
the azimuthal direction as the i-direction.
— EURANUS flow solver solves the equations in a relative frame of reference. This implies
that the mesh blocks surrounding rotating blades rotate with the blades and should therefore
been set as "rotating" in this page. The rotational speed of the block has to be given in RPM
(Revolutions Per Minute). A positive value for the rotation speed indicates a rotation in pos-
5-2 FINE™
Rotor/Stator Interaction in the FINE™ Interface Rotating Machinery
The different computations in a project can not have different configuration settings. If
any of the settings in this area is changed, the modifications will be taken for all (active
and not active) computations.
The rotation axis should be the z-axis in order to be compatible with the EURANUS
flow solver.
On the left the list of patches is displayed which are defined as part of a rotor/stator interface in
IGG™ (boundary condition type: "ROT"). It is possible to display only a limited amount of patches
by using the Filter. Type in the text box the part of the name that is common to all patches to display
and press <Enter>. All patches containing in the name the entered text will be displayed. Note that
the filter is case insensitive.
It is possible to group patches that have the same definition by selecting them in the list and click-
ing on the Group button. A text box will appear below this button which allows to enter a name for
the group. The group will be displayed in red in the list of patches. Double click on the group name
to see which patches are in the group. To select multiple patches select them while keeping the
<Ctrl> key pressed. To remove the group simply select the group from the list and click on
Ungroup. The patches in the selected group will be put back into the list of patches and the group
name will be removed from the list.
FINE™ 5-3
Rotating Machinery Rotor/Stator Interaction in the FINE™ Interface
Multiple rotor/stator interfaces can be treated in one single project. Each rotor/stator interface is
indicated by an ID number and each of them may contain an arbitrary number of upstream and
downstream patches. For each patch in the list the ID number of the corresponding rotor/stator
interface has to be provided by the user. Also it should be specified whether the patch is located in
the upstream or downstream side of the interface. It should be mentioned that no hypothesis is made
on the direction of the flow through the interface. The purpose of the upstream and downstream
denominations is only to establish the two groups of patches belonging to respectively the rotor and
the stator side.
.The ID number of the rotor/stator interface cannot be higher than the total number of
rotor/stator interfaces.
In addition to the ID number and the upstream/downstream side specifications the user has to select
also:
• the order of extrapolation (zero or first order). Zero order extrapolation (default) is used in
most cases and usually provides satisfactory results. First order extrapolation may be useful in
case of important flow gradients in the direction normal to the interface together with a rela-
tively coarse mesh.
• one of the four available steady state techniques (for steady state applications only), i.e. the
Local Conservative Coupling, Conservative Coupling by Pitchwise Rows, Full Non Matching
Mixing plane and Full Non Matching Frozen Rotor approaches.
1. The first approach is only recommended for the rotor/stator interface between an impeller and a
volute. Experience has shown that in cases where significant flow variations are observed in the cir-
cumferential direction (as for instance in the case of rotor-volute interactions) it is more stable to
base the flux decomposition on the local flow direction as it is done in the Local Conservative Cou-
pling. Small conservation errors can be observed with this technique, which is therefore only used
for special configurations.
5-4 FINE™
How to Set-up a Simulation with Rotor/Stator Interfaces? Rotating Machinery
2. The second approach is recommended due to its capability to provide an exact conservation of
mass flow, momentum and energy through the interface. This approach adopts the same coupling
procedure for all the nodes along the circumferential direction, even if the local flow direction is
different from the averaged one. This technique offers the advantage of being able to guarantee a
strict conservation of mass, momentum and energy through the interface. Moreover it shows to be
very robust and is therefore used by default.
3. The third approach is respecting an exact conservation of mass flow, momentum and energy
through the interface as the second approach but has less constraint on the interface geometry.
4. The fourth approach is considering the rotor/stator interface as a perfect connection and is
neglecting the rotor movement in the connecting algorithm. In this approach, the periodicities must
be equal between the rotor and the stator.
FINE™ 5-5
Rotating Machinery How to Set-up a Simulation with Rotor/Stator Interfaces?
• The patches on both sides of the rotor/stator interface must cover the same range in spanwise
direction,
• The meshes need not to match in the spanwise direction, but the azimuthal mesh lines on the
boundary must be circular arcs.
An illustration of these constraints is given in Figure 5.4.1-4. Consisting of two stripes, the depicted
configuration provides a complete description of the capabilities. The upstream side of the quasi-
steady interface consists of three blocks, represented by the bold frames and the surface meshes,
while a single block is assumed for the downstream side. A gap may exist between two blocks, e.g.,
if the mixing plane coincides with the exit plane of a blade row having a blunt trailing edge. The
grid points distribution are different for each of the boundary patches involved, but the pitchwise
grid lines have constant radius and the pitchwise line along which two patches of the same side join
(e.g., BC 1 and BC 3) exists as an inner grid line on all other boundaries. In the example, the inter-
face is thus decomposed into two stripes, stripe 1, consisting of BC 1 and BC 2 on the upstream side
and BC 5 on the downstream side, and stripe 2, consisting of BC 3 and BC 4 on the upstream side
and BC 6 on the downstream side.
.
A new interpolation technique (Full Non Matching Mixing plane) has been implemented to avoid
those geometrical constraints. This new technique is based on the full non-matching algorithm and
has clear advantages:
• The geometrical constraints have been removed, the only remaining constraint being that the
rotor/stator patches belonging to a given interface should lie on the same common axisymmet-
ric surface.
5-6 FINE™
How to Set-up a Simulation with Rotor/Stator Interfaces? Rotating Machinery
When setting the expert parameter IRSNEW=1 only Full Non Matching Mixing plane
approach will be applied on all rotor/stator interfaces even if another mixing plane
approach has been selected in the Rotating Machinery page under Rotor-Stator thumb-
nail.
ID = 2
Upstream Downstream
Upstream Downstream
ID = 1
Flow Direction
FINE™ 5-7
Rotating Machinery How to Set-up a Simulation with Rotor/Stator Interfaces?
a) Selection of Technique
IRSNEW = 0: default mixing plane interpolation technique imposing geometrical constraints as
listed in section 5-4.1.1.
= 1: new full non-matching algorithm for rotor/stator interfaces removing the geometrical con-
straints of the default mixing plane technique. More detail on this interpolation technique can be
found in section 5-5.3.
5-8 FINE™
How to Set-up a Simulation with Rotor/Stator Interfaces? Rotating Machinery
2. To create a mesh covering the same pitchwise distance on both sides of the interface and to
impose Full Non Matching Frozen Rotor in the Rotating Machinery page under the Rotor-Stator
thumbnail. Furthermore the user has to specify for each of these patches the ID and the
FINE™ 5-9
Rotating Machinery How to Set-up a Simulation with Rotor/Stator Interfaces?
Note that for example for an impeller-volute computation the complete impeller
would need to be meshed in all three cases since the periodicity of the volute is 1.
As an illustration the frozen rotor method is compared to the sliding mesh approach for the Aachen
Turbine in Figure 5.4.2-7 and Figure 5.4.2-8.
5-10 FINE™
How to Set-up a Simulation with Rotor/Stator Interfaces? Rotating Machinery
FINE™ 5-11
Rotating Machinery How to Set-up a Simulation with Rotor/Stator Interfaces?
• Output For Visualization/At end only and Multiple Files: multiple set of output files saved for
restart of the unsteady computation and visualization in CFView™ only available at the end of
the computation.
• Output For Visualization/At end only and One Output File: one set of output files saved for
restart of the unsteady computation and visualization in CFView™ only available at the end of
the computation.
• Output For Visualization/Intermediate and Multiple Files: multiple set of output files saved for
restart of the unsteady computation and visualization in CFView™ during the computation.
• Output For Visualization/Intermediate and One Output File: one set of output files saved for
restart of the unsteady computation and visualization in CFView™ during the computation.
Note: When a sliding mesh simulation with IRSNEW=1 is started from a mixing
plane simulation with IRSNEW=0 it is not possible to use the *.fnmb file of the mixing
plane simulation to read in the initialization of the sliding mesh simulation. It is neces-
sary to compute the full non-matching connections again.
5-12 FINE™
How to Set-up a Simulation with Rotor/Stator Interfaces? Rotating Machinery
FINE™ 5-13
Rotating Machinery How to Set-up a Simulation with Rotor/Stator Interfaces?
• A steady state initialization is performed in a separate computation with steady Time Configu-
ration. It is suggested to model all rotor/stator interfaces in this case using frozen rotor
approach. The unsteady domain scaling computation is using the result of this steady computa-
tion as initial solution. See Chapter 10 for more information on how to start a computation
from an initial solution file.
• The steady state initialization is automatically performed before starting the time accurate
computation (that means within the same computation), switching to unsteady Time Configu-
ration in the Flow Model page and selecting in the interface an appropriate Number of Steady
Iterations in the Steady Initialization parameters on the Control Variables page.
4. The Number Of Angular Positions defined in the Control Variables page is the number of time
step per full revolution (2π) of the rotor. In theory, the number of time step per full rotation has
to be a multiple of the number of blades in the rotor (Nbrotor) and number of blades in the stator
(Nbstator).
Number Of Angular Positions = i x Nbrotor x Nbstator
The two periods of blade passing frequencies are respectively (ixNbrotorx∆t) and
(ixNbstatorx∆t). Since the number of time step per passage is usually imposed around 40,
the value of i should be adjusted so that (ixNbrotor) and (ixNbstator) are both close to this
number. By experience, if the number of time step per passage is lower than 30 the solu-
tion starts to be deteriorated.
5. The Number Of Time Steps in the Control Variable page is defined as:
Number Of Time Steps = j x (Number Of Angular Positions)/min(Nbrotor,Nbstator)
By default, the Phase Lagged method used to compute the viscous fluxes at the rotor/
stator interface from dummy cells values set from other side inner cells instead from
same side inner cells by extrapolation (more detail in section 5-5.4.3).
5-14 FINE™
Theoretical Background on Rotor/Stator Interfaces Rotating Machinery
IRSVFL = 0 IRSVFL = 1
FINE™ 5-15
Rotating Machinery Theoretical Background on Rotor/Stator Interfaces
Solving for such flows requires an unsteady and viscous flow solver with the capability to manage
enormous data storage. One way to optimize the computer memory requirement is to resolve the
steady flowfield on a truncated computational domain. This requires a pitchwise averaging process
to be performed at the so-called rotor/stator interfaces. This particular boundary condition is availa-
ble under the FINE™/Turbo environment as the mixing plane approach. The rotor-stator interaction
is done by exchanging circumferentially averaged flow quantities. Physically, this means that the
blade wake or separation occurring in the blade passage are mixed circumferentially before entering
the downstream component. As a result, the velocity components and pressure is uniform in the cir-
cumferential direction at the rotor/stator interface. This physical approximation tends to become
more acceptable as rotational speed is increased. This mixing plane technique is by far the most
used rotor/stator modeling in industry approach. See section 5-5.2 and section 5-5.3 for more detail
on the mixing plane approach.
The second step allowing for a higher level of accuracy consists of considering an unsteady rotor/
stator interaction. Various techniques allowing to treat configurations with an arbitrary number of
blades in the rotor and the stator have been proposed by many authors (Giles, 1990, Lemeur, 1992,
Erdos et al, 1977, He, 1997), each of them having their own advantages and disadvantages, each of
them affecting differently the memory requirement, solution time and the most important parame-
ter, the accuracy of the computed flow solution. But none of these techniques allows to resolve on a
truncated computational domain all time frequencies present in the turbomachinery configuration,
except if the pitch distance is the same in the successive rotors and stators. In fact, each of them are
affecting the computed time frequencies and Fourier modes that are propagating the flow informa-
tion inside the gap region.
The Domain Scaling Method is another type of unsteady rotor/stator interface treatment (Rai,
1989), that has been selected and implemented in the EURANUS flow solver. This technique is
based on the constraint that the pitch distance must be identical on both sides of the interface. If the
numbers of blades in the rotor and the stator are different, this implies either to increase the number
of passages modelled or to scale the geometry. The Domain Scaling Method presents the advantage
that it allows to resolve all time frequencies with an error that is proportional to the scaling coeffi-
cient. The accuracy of the computed unsteady flow solution may then be improved at the expense
of more data storage and larger solution time if more blade passages are considered.
Finally, the Phase Lagged Method is the last type of unsteady rotor/stator interface treatment that
has been selected and implemented in the EURANUS flow solver. This technique is removing the
Domain Scaling constraint that the pitch distance must be identical on both sides of the interface.
The Phase Lagged Method presents the advantage that it allows to resolve all time frequencies with
an error that is proportional to the scaling coefficient. The accuracy of the computed unsteady flow
solution may then be improved at the expense of more data storage and larger solution time if more
blade passages are considered.
5-16 FINE™
Theoretical Background on Rotor/Stator Interfaces Rotating Machinery
For compressible flows, the density, velocity vector and pressure are derived from Q ρ, Q v, Q E ,
respectively the flux of mass, the flux of momentum plus the pressure force and the flux of energy.
The other variables such as k or ε are also derived from their fluxes. They are referred to, in the fol-
lowing as X an QX. The fluxes are constructed from the primitive variables using:
FINE™ 5-17
Rotating Machinery Theoretical Background on Rotor/Stator Interfaces
Qρ = ρ ( V ⋅ S )
Q v = ρV ( V ⋅ S )
. (5-1)
2
Q E = ρ ( CpT + V ⁄ 2 ) ( V ⋅ S )
Q X = ρX ( V ⋅ S )
where ρ is the density, V the absolute velocity vector, T the temperature, Cp the heat capacity and
S the surface vector of the cell face. After application of the mixing process the primitive variables
are calculated from the fluxes using:
2 · 2
Qv ⋅ S ⎛ Qv ⋅ S ⎞ ( γ – 1 ) ( Q v – ( 2Q ρ Q E ) )
-±
p s = --------------------- ⎜ ---------------------
-⎟ + ----------------------------------------------------------
2 2 2
S (γ + 1) ⎝ S ( γ + 1 )⎠ S (γ + 1)
Qv – ps S
v = -------------------
-
Qρ (5-2)
Qρ
ρ = ---------
-
v⋅S
Q
X = ------X-
Qρ
For incompressible flows, the velocity vector and pressure are derived from Q ρ, Q v , respectively
the flux of mass and the flux of momentum plus the pressure force (See Eq. 5-1). The other varia-
bles like k or ε are also derived from their fluxes. They are referred to, in the following, as X and
QX. As far as the energy equation is concerned, the temperature is also derived from its flux like the
other X variables.
2
( Qv ⋅ S – Qρ ⁄ ρ )
p s = --------------------------------------
2
S
( Qv – ps S ) (5-3)
V = ------------------------
-
Qρ
QX
X = -------
Qρ
5-18 FINE™
Theoretical Background on Rotor/Stator Interfaces Rotating Machinery
Therefore a new boundary conditions has been developed, in which the decision on the direction of
the flow is taken only once per pitchwise row, and which avoids the fluxes decoding. This new
boundary condition has shown to be more robust in most cases and guarantees a strict global con-
servation. As mentioned in section 5-3 the only type of application for which a local approach
would better suit is the impeller-volute interaction.
One image mesh is constructed on both sides of the interface. It respects the following constraints:
FINE™ 5-19
Rotating Machinery Theoretical Background on Rotor/Stator Interfaces
• The image is a "concatenation" of all the patches constituting the left or right side of the inter-
face.
• The pitchwise mesh lines lie on circular arcs.
• The left and right images have the same spanwise distribution of the nodes.
• The mesh density of the image is similar to the one of the initial patch.
The creation of the image mesh is illustrated by Figure 5.5.3-11, showing the creation of the image
of a butterfly configuration. With the previous version of the mixing plane module such a configu-
ration would have required the creation of an intermediate block with one face connected to the but-
terfly and the opposite face respecting the constraints of rotor/stator patches.
F 1 = ρV n ∆S
F 2 = ρV r V n ∆S
F 3 = ρV θ V n ∆S
F 4 = ρV z V n ∆S
F 5 = ρHV n ∆S
2. For all spanwise positions perform a pitchwise averaging of the flux and of the primitive varia-
bles.
3. For all spanwise positions determine the upstream and downstream side according to the flow
direction, and build the mixing plane fluxes.
5-20 FINE™
Theoretical Background on Rotor/Stator Interfaces Rotating Machinery
The local fluxes on the upstream side are taken from the upstream, with a contribution of the aver-
aged downstream static pressure.:
ups ups
F1 = F1
ups ups down
F2 = F2 + p n r ∆S
ups ups
F3 = F3
ups ups down
F4 = F4 + p n z ∆S
ups ups
F5 = F5
On the downstream side the local fluxes are taken as the averaged fluxes on the upstream side with
a contribution of the local static pressure:
down ups
F1 = F1
down ups down
F2 = F2 + p n r ∆S
down ups
F3 = F3
down ups down
F4 = F4 + p n z ∆S
down ups
F5 = F5
K1 P1 = K2 P 2 . (5-4)
where the indices 1 & 2 refer to the upstream and downstream sides of the interface, and P1 & P2
are the pitches of the two connected blade rows.
If the numbers of blades in respectively the rotor and the stator are not equal, the user should make
the choice between:
• modelling a higher number of blade passages (or even the entire machine if a common multi-
ple can not be found). If for instance the numbers of blades are 16 & 24, the numbers of pas-
sages to model are respectively 3 & 2. This permits to avoid the introduction of an error due to
the geometry scaling, at the cost of an increased memory requirement.
• scaling the geometry. The easiest way to proceed in order to obtain the required scaling is to
generate a mesh with a blade number different from the real one. The ratio between the modi-
fied and the real blade numbers should be as close as possible to 1.0.
In some cases the most appropriate treatment may be a combination of the 2 above approaches. If
for instance the numbers of blades are 18 & 24, one could scale the stator by changing the number
of blades from 18 to 16, and then choose to model 2 & 3 passages. The scaling factor is then 1.125,
instead of 1.33 in case only one blade passage is modelled.
The major advantage of having the same pitch distance on both sides of the interface is that it per-
mits to avoid considering any time periodicity in the boundary condition treatment. The boundary
conditions to be imposed along the periodic boundaries inside the gap region are given by
FINE™ 5-21
Rotating Machinery Theoretical Background on Rotor/Stator Interfaces
R R R R
U i ( r, θ i , z, t ) = U i ( r, θ i ± p, z, t ) i = 1, 2 , (5-5)
where p = K 1 P 1 is the pitch distance on both sides of the interface. The “+” sign is to be used
when imposing boundary conditions on the lower periodic segment (minimum azimuthal coordi-
nates) and the “-” sign is to be used for the higher periodic segment (maximum azimuthal coordi-
nates). Along the rotor/stator interface, the boundary condition is defined as
Finally, the previous implementation lets the rotor do the full rotation. The problem in this method
is that the rotor and the stator are in front of each other for a little fraction of time. During this time,
no time shift is necessary at the R/S interface which acts then as a pure connection. The modifica-
tion proposed aims at increasing the time during which the R/S interface acts as a connection. To
perform this, only the periods during which the rotor and the stator are in front of each other, are
repeated. But, since the length of the periods are different in the 2 blades rows, the solution has to
be stored and reset in the row with the smallest period.
Example: if row 1 has the highest number of blades and therefore is submitted to the larger period,
this last period T1 is repeated over and over, while a fractional number of periods will be simulated
in row 2. The solution is thus stored at the end of the last completed period in row 2, to be reused at
the beginning of each T1. This method increases a lot the speed of exchange between the two blade
rows. It increases also the robustness of the computation.
5-22 FINE™
Theoretical Background on Rotor/Stator Interfaces Rotating Machinery
FINE™ 5-23
Rotating Machinery Theoretical Background on Rotor/Stator Interfaces
X’ and X” are respectively extrapolated on R/S interface from X3 and X4. Then X2’ is interpolated
from X’ and X”. Finally, X2” is defined in the dummy cell by extrapolating X2 (result coming from
the stator) and X2’ (result coming from the rotor). This last method, automatically activated when
dealing with phase-lagged simulation, can be activated for non phase-lagged simulation through the
expert parameter IRSVFL (by default set at 0) that has to be set at 1.
5-5.4.5 References
Erdos J.I., Alzner E. & McNally W. (1977), Numerical Solution of Periodic Transonic Flow
Through a Fan Stage, AIAA Journal, Vol. 15, No. 11, pp. 1559-1568
Giles M.B. (1990), Stator/Rotor Interaction in a Transonic Turbine, Journal of Propulsion, pp. 621-
627
Van Leer B. (1985), Upwind-Difference Methods for Aerodynamic Flows Governed by the Euler
Equation, Lectures in Applied Mathematics, Vol. 23, Part 2, AMS, Providence, pp. 327-336
5-24 FINE™
CHAPTER 6: Throughflow Model
6-1 Overview
The Throughflow module is intended for the design and the analysis of turbomachinery flows. This
kind of simulation is based on the two-dimensional steady state axisymmetric Euler equations with
a specific model for blade rows. It is fully integrated in EURANUS, the three-dimensional Euler-
Navier-Stokes flow solver. As a consequence:
• Options that apply to comparable 3D calculations also apply to throughflow calculations: the
physical fluid model, the boundary conditions, the spatial and time discretization, the conver-
gence acceleration techniques and the output, including data for visualization.
• The code accepts any number of blade rows.
• Throughflow blocks can be combined with 3D blocks in a single computation.
• Any improvement introduced in the 3D solver also applies to the througflow blocks.
The throughflow simulation should be used only with the perfect gas model currently
available through the FINE™ interface.
The throughflow simulation can be activated directly from the FINE™ interface by clicking on the
Throughflow Blocks thumbnail in the Configuration/Rotating Machinery page. This thumbnail is
only accessible when the Mathematical Model is set to Euler on the Flow Model page. In the next
section the Throughflow Blocks option in the FINE™ interface is described including all related
parameters. Finally, section 6-5 provides its theoretical background.
FINE™ 6-1
Throughflow Model Throughflow Blocks in the FINE™ Interface
6-2 FINE™
Throughflow Blocks in the FINE™ Interface Throughflow Model
Two different types of throughflow blocks are available. The throughflow module is fully active by
selecting either the Compressor or Turbine options. The difference between the two lies in the defi-
nition of the loss coefficient: for a turbine the pressure used as reference is taken at the trailing edge
and for a compressor at the leading edge.
FINE™ 6-3
Throughflow Model Throughflow Blocks in the FINE™ Interface
In all modes except the analysis mode with Throat Control the geometry can be defined by:
• 2D input file: the file defining the blade geometry should contain the (Z,R,θcl,θss,θps) coordi-
nates of the blade camber line, suction and pressure sides throughout the meridional domain
covered by the blade. For each (Z,R) meridional position the circumferential coordinate θ is
given at the camber line, the suction and pressure sides. For the format of a 2D input file see
section 6-3.
• 1D input parameters where the user specifies:
— the streamwise location of the maximum thickness at hub and shroud (a linear evolution is
assumed from hub to shroud),
— a file containing the spanwise distribution of maximum tangential thickness,
— the normalised streamwise distribution of blade thickness: This distribution can be com-
puted using a power function whose exponent has also to be specified or loaded from a 1D
input file with the format as specified in section 6-3. For most configurations a power func-
tion with an exponent of 2 leads to a correct modelling of the blade generating a quadratic
thickness evolution. The thickness function is set to zero at the leading and trailing edges
and presents a maximum at the prescribed position of maximum thickness.
An additional blockage can be introduced to take the endwall boundary layers (Endwall boundary
layer and wake blockage factors) into account. The two real parameters refer to the inlet and outlet
values of the endwalls blockage. A linear variation is assumed between the inlet and the outlet.
6-4 FINE™
Throughflow Blocks in the FINE™ Interface Throughflow Model
In case the analysis mode with Throat Control is selected the user specifies the streamwise evolu-
tion of the effective blockage, which results from the tangential blockage and the flow angle beff= b
cosβ, with b the tangential blockage and β the relative flow angle at the trailing edge.
The user specifies:
— the streamwise location of the throat at hub and shroud. If a negative value is entered, it is
assumed that there is no throat and a linear evolution of the effective blockage from the lead-
ing to the trailing edge will be adopted.
— a file containing the spanwise distribution of the throat section. The throat is by definition
the section in which the effective blockage is minimum. In transonic turbines it determines
the mass flow.
— the normalised streamwised distribution of effective blockage: This distribution can be com-
puted using a power function whose exponent has also to be specified, or loaded from an
input file. The streamwise distribution of effective blockage results from the values specified
by the user at the leading and trailing edges and at the throat (if a throat is given).
FINE™ 6-5
Throughflow Model Throughflow Blocks in the FINE™ Interface
tion, the streamwise distribution of the flow angle or of the tangential velocity is iteratively com-
puted according to the calculated value at the leading edge and the value defined at the trailing
edge. The default value of the exponent is 1, which generates linear evolutions from leading to trail-
ing edge.
Two definitions of the flow angle are available, respectively based on the axial and
meridional directions. The choice between these definitions is made through the expert
parameter ITHVZM:
Wθ=Wz.tan β(ITHVZM=0, default)
Wθ=Vm.tan β(ITHVZM=1).
6-6 FINE™
Throughflow Blocks in the FINE™ Interface Throughflow Model
FINE™ 6-7
Throughflow Model Throughflow Blocks in the FINE™ Interface
Throughflow blocks can also be connected to 3D blocks. This implies the use of a quasi-steady
interface (rotor-stator interface).
∇b
• Grad(B)/B-1st component: radial component of the ------
- vector,
b
6-8 FINE™
File Formats for Throughflow Blocks Throughflow Model
∇b
• Grad(B)/B-3rd component: axial component of the ------
- vector,
b
• Lean vector (tangential vector along the blade in the spanwise direction),
• Loss coefficient ψ ,
• Blade force vector fB ,
FINE™ 6-9
Throughflow Model File Formats for Throughflow Blocks
— if for a streamwise distribution, the leading and/or trailing edge values should be extended to
the upstream/downstream ends of the block. Values of 1 and 3 extend the upstream region,
whereas values of 2 and 4 extend the downstream regions. This switch is used to achieve a
blockage factor below 1. upstream and/or downstream of the blade passages.
The angles are in degrees, the length in meters, except for the blade thickness, which is in millime-
tres.
Some examples of the geometrical files are given hereafter:
• Spanwise distribution of the maximum blade thickness:
14 4 1 0
.1181262 8.20958
.1282913 8.22191
.138422 8.16546
.1484767 8.03777
.1583352 7.85892
.1680261 7.58743
.1778827 7.26814
.1877939 6.86779
.197697 6.45259
.2076666 6.11613
.217709 5.78311
.2277728 5.65696
.2378241 5.84635
.2477884 6.27821
6-10 FINE™
File Formats for Throughflow Blocks Throughflow Model
ALPHA = arctg(VT/VZ)
BETA = arctg(WT/VZ)
GAMMA = arctg(VR/VZ)
Mach Number
Meridional Mach Number
Static Pressure, Static Temperature,
Density
Absolute Total Pressure
Rothalpy Total Pressure
Relative Total Pressure
Absolute Total Temperature
Rothalpy Total Temperature
Relative Total Temperature
Isentropic Efficiency
Polytropic Efficiency
Mass Flow
Total Area
FINE™ 6-11
Throughflow Model Expert Parameters Related to Throughflow Blocks
In addition of files which can be read by CFView™, a summary file is created when the through-
flow module is activated. It has the extension ".gtf" and gives the average inlet and outlet results for
each block in SI units.
6-4.2 Others.
• ITHVZM = 0 (default): Wθ=Wz.tan β
= 1: Wθ=Vm.tan β
Definition of the flow angle. The selection of ITHVZM to 1 is crucial for a correct treatment of
radial machines.
• PRCON to define the pressure for the initial solution at rotor/stator interfaces. If this parameter
is not modified a linear pressure distribution from inlet to outlet is assumed.
6-12 FINE™
Theoretical Background on Throughflow Method Throughflow Model
0
ρ bρW
∂ 1
---- + --- ∇ ⋅ b ( ρW ⊗ W + pI ) = ∇b (6-1)
∂t ρW b ρQ + p ------- + ρ ( f B + f F )
b
ρE bρWH ρQ
This is a system of 5 PDE for 5 unknowns (the dependent variables, e.g., the primitive variables in
cylindrical coordinates, ρ, Wr, Wθ, Wz, p) in two spatial dimensions (r and z). Apart from the
absence in the differential operators of derivatives with respect to the tangential direction, the sys-
tem differs from the three-dimensional Euler equations in three ways:
• The presence of the blockage factor b in the fluxes.
∇b
• Source terms p ------
- and fB representing the inviscid action of the blades.
b
• A source term f F grouping all the effects of viscosity and heat conduction.
The system is solved in Cartesian coordinates using explicit Runge-Kutta integration in time and
either central or upwind spatial discretization.
FINE™ 6-13
Throughflow Model Theoretical Background on Throughflow Method
d 2πr
b = 1 – --- where s = --------- (6-2)
s z
Thus, b = 1 outside blade rows and 0 < b ≤ 1 inside blade rows (including the hypothetical case of
blades with zero thickness). The blockage factor is an external input to the computation.
k+1 k t
( ρF Bθ ) = ( ρF Bθ ) + κ ( W θ – W θ ) (6-3)
where the constant κ is user input. The target tangential velocity Wθt is either direct external input
(design mode) or locally calculated from the prescribed relative flow angle β and the current merid-
ional velocity (analysis mode: Wθ=Vmtanβ). The vector of the blade force, fB , is perpendicular to
the mean camber surface, which is internally defined by the flow angle and the lean angle. Cur-
rently, a zero lean angle is assumed.
Robust analysis mode
In the case of the analysis mode another formulation of the above equation is adopted, based on the
magnitude of the blade force instead of its tangential component:
··
k+1 k F̃
( ρFB ) = ( ρF B ) – κ ⎛⎝ W n⎞⎠ ----- (6-4)
W̃
where FB is the magnitude of the blade force and n is the normal vector whose direction results
from the imposed relative flow angle distribution. In case the angle is measured with respect to the
axial direction (ITHVZM=0):
6-14 FINE™
Theoretical Background on Throughflow Method Throughflow Model
p W
ρf F = – ------------ p ref ∂ m ψ --------- (6-6)
pt rot W
This formulation ensures that a loss coefficient of zero will give exactly a zero friction force. The
imposed loss coefficient is defined as:
LE
ptrot – ptrot
ψ = ------------------------------- (6-7)
p ref
where ptrot is the total pressure associated with rothalpy defined as:
γ -
----------
Tt rot γ – 1 2
W – ( ωr )
2
ptrot = p ⎛ -------------⎞ where
Tt rot = T + -----------------------------
⎝ T ⎠ 2c p (6-8)
where W is the relative velocity magnitude and pref is a reference dynamic pressure taken at the
leading edge (for compressors) or trailing edge (for turbines) on each streamwise mesh line.
FINE™ 6-15
Throughflow Model Theoretical Background on Throughflow Method
6-16 FINE™
CHAPTER 7: Optional Models
7-1 Overview
FINE™ contains the following three optional models:
• Fluid-Particle Interaction to model solid particles in the flow.
• Conjugate Heat Transfer to model the heat transfer between the fluid and solid walls.
• Cooling/Bleed to model cooling holes or bleed.
• Transition Model to model the flow transition.
Each of these models requires a special license feature. If the model is not accessible due to a miss-
ing license it is displayed in gray under Optional Models. Contact NUMECA sales or support team
for more information about the possibility to obtain a license.
FINE™ 7-1
Optional Models Fluid-Particle Interaction
tion and of the particle traces tracking can be repeated several times until no changes are observed
in the results (usually 3 to 10 times).
In the present Lagrangian model, the motion of the particles is only influenced by their own inertia
and by the flow, i.e. the local aerodynamic or hydrodynamic forces. Interactions between the parti-
cles are currently not considered. This implies the hypothesis of a dilute flow. On the contrary in a
dense flow, particle motion is governed by their direct interaction, which implies more complex
models, such as the Eulerian ones, in which both phases are modelled as continuous phases.
Particle-laden flows are found in many industrial applications such as cyclone separators, pneu-
matic transport of powder, droplet and coal combustion system, spray drying and cooling or sand-
blasting. In all cases, the flow consists of a continuous phase - also named carrier flow - and a
dispersed phase in the form of solid particles, liquid droplets or gas bubbles. In general, the motions
of the carrier flow and of the dispersed phase are interdependent.
In the lagrangian approach, the physical particle cloud is represented by a finite number of compu-
tational particles. The continuous phase (fluid) is treated as a continuum using a boundary-control-
led method whereas the dispersed phase (particles) is treated as a number of separate particles. The
particles are successively tracked as a single probe particle through the flow field using a Lagrang-
ian formulation.
In the case of laminar flows, each particle will follow a unique deterministic trajectory. In the case
of turbulent flows however, particles will have their own random path due to interactions with the
fluctuating turbulent velocity field. This feature is taken into account in EURANUS through a sto-
chastic treatment. This approach consists of employing a sampling technique based on a Gaussian
probability density function to introduce the random nature of the turbulent velocity field. The par-
ticle diffusion due to turbulence can therefore be modelled without introducing artificial diffusion
properties of the cloud.
According to the distribution and contents of the particles, an averaging procedure provides phase
coupling source terms to be added into the Eulerian carrier phase conservation equations. The two-
way coupling can be achieved through global iterations.
7-2.1.3 Overview
In this chapter an overview is given of the fluid particle interaction module incorporated in
NUMECA’s FINE™ environment:
7-2 FINE™
Fluid-Particle Interaction Optional Models
• section 7-2.2 describes the main functionalities of the module and the way to proceed in the
FINE™ interface.
• section 7-2.3 gives the format of the files associated with the module.
• section 7-2.6 summarizes the main theoretical features of the Lagrangian model.
FINE™ 7-3
Optional Models Fluid-Particle Interaction
3. in this new computation activate the Lagrangian Fluid-particle calculation and perform a com-
putation with Lagrangian coupling.The particles encounter several forces: the friction force, i.e.
the resultant of the pressure and viscous stress in the carrier phase over the particle surface, as
well as potential inertia and gravity forces. The user is free to introduce the gravity force
through the vectorial components of the acceleration of gravity on the Flow Model page. An
inertial force due to the rotation of the system of reference can also be included. If some of the
blocks are rotating this button should systematically be activated. In this case the reference rota-
tion velocity is the one specified on the Rotating Machinery page.
a) Inlet Boundary
The user can enter a number of classes of particles, each class representing a different type of parti-
cles. For each class the inlet boundary conditions have to be specified. The displayed boundary con-
ditions are the one for the class specified in the bar as indicated (a) in Figure 7.2.2-2. To select
another class use the scroll buttons on both sides of this bar. The purpose of this is to be able to sim-
ulate the presence of different types of particles in the same patch.
(a)
FIGURE 7.2.2-2 Inlet boundary conditions for the fluid-particle interaction model
7-4 FINE™
Fluid-Particle Interaction Optional Models
Note however that multiple classes are of interest for fully couples calculations only, by allowing
the flow calculation to account for the presence of different particle types. In the case of a weak
coupling strategy it is recommended to create one computation per particle class, because only one
particle class can be visualized in CFView™. The Expert parameter (in Expert Mode on the Control
Variables page) CLASID permits to specify which particle class must be included in the CFView™
output.
The inlet boundary conditions are the following:
• spatial distribution of the injection points: uniform over the chosen patch or located at a chosen
point of coordinates (x,y,z).
• inlet velocity of the particles (absolute or relative frame of reference). The velocity is always
considered in the same relative frame of reference as the one of the block in which the inlet
particle is located. The relative velocities here defined as the difference between the particle
and the flow velocity (a zero relative velocity is therefore very commonly used) whereas the
absolute velocity boundary conditions is used when the velocity of the particle is well known
and different from the velocity of the flow. Note that only the Cartesian coordinate system is
available.
• density and mean diameter of the particles.
• particle concentration (number of particles/m3).
• number of particles per cell should be seen as the number of trajectories per cell.
The use of multiple trajectories per cell permits to ensure a more uniform covering of the inlet (and
hence of the computational domain) by the particles. This is only of interest in the case of fully cou-
pled calculations where it is important to ensure that the source terms are correctly evaluated in all
mesh cells. The use of multiple trajectories does not modify the number of particles entering the
domain. Each trajectory must be seen as an ensemble of particles. It is also very often referred to as
a computational particle.
In the case of full coupling, the number of particles per cell should be statistically large enough to
provide accurate averaging of the dispersed phase parameters. Therefore it is recommended in this
case to have a high number of particles per cell. This increases the computational time required for
the particle tracking process, but gives rise to a smoother distribution of the particles in the domain.
In the case of a weak (1-way) coupling, only one particle per cell is usually enough to provide a suf-
ficiently detailed description of the particles distribution in the field.
FINE™ 7-5
Optional Models Fluid-Particle Interaction
FIGURE 7.2.2-3 Solid boundary conditions for the fluid-particle interaction model
7-2.3 Outputs
The following quantities are calculated by the solver (see Figure 7.2.3-4):
⎛ ⎞
• the number concentration ⎜
⎝
∑ Ni⎟⎠ ⁄ Ω where Ni is the number of particles of class i in the cell
i
of volume Ω,
⎛ ⎞
• the mass concentration ⎜
⎝
∑ ρipart Ni Ωipart⎟⎠ ⁄ Ω (mass of particles per cell volume) where
i
• the weighted mass and weighted absolute or relative velocity vectors number,
⎛ ⎞
• the components of the interphase force F = ⎜
⎝
∑ Fi Ni⎟⎠ ⁄ Ω .
i
The solver automatically writes an output file containing the particle traces that is read by the
CFView™ post processing software.
7-6 FINE™
Fluid-Particle Interaction Optional Models
• Scalars
Number concentration (51)
Mass of particles per cell volume (52)
Volume of particles per cell volume (53)
Presence (1) or absence (0) of particles in cells (54)
Vx: x-component of mass-weighted velocity vector (55)
Vy: y-component of mass-weighted velocity vector (56)
Vz: z-component of mass-weighted velocity vector (57)
Fx: x-component of interphase force (58)
Fy: y-component of interphase force (59)
Fz: z-component of interphase force (60)
FINE™ 7-7
Optional Models Fluid-Particle Interaction
• The maximum residence time (TMAX in corresponding ’.run’ file), which is the time limita-
tion for particle residence in the flow. It may be used either to prevent particle stopping or to
avoid an infinite loop, having for instance a particle captured in a separation bubble. After the
time has expired, the particle disappears.
• The governing equations for particles in a carrier flow consist of a set of non-linear ordinary
equations. Two methods are provided to integrate this equations with respect to time (that can
be selected in expert mode):
— The Runge-Kutta scheme, in which the transfer to the next time level is fulfilled through a
finite number of intermediate steps. A two-stage and a four-stage option (NRKS in corre-
sponding ’.run’ file) are available. This scheme is limited to big particles.
— The exponential scheme that lies on a semi-analytical method. This scheme is particularly
suitable for small particles.
• The Drag force is in general the only considered part of the interphase force. It corresponds to
the force on a particle moving in a still fluid at constant velocity (or using the particle refer-
ence system, a fixed particle in a uniform laminar flow). At extremely low Reynolds numbers
(Rep << 1 i.e. at low relative velocities), it is expressed through the Stokes formula. For higher
Reynolds numbers, an optional standard curve fit may be chosen (section 7-2.6.1). The choice
between the two types of forces can be made in expert mode.
Other parameters have a standard value, which may be modified only on rare occasions. They are
accessible in the Expert Mode of the Computation Steering/Control Variables page:
CLASID: selection of class for CFView™ output
LMAX: the maximum number of segments along a trajectory may be used to avoid an infinite
loop in the case of an oscillatory movement for instance.
DDIMAX: tolerance factor to establish that a particle has left a cell (default=1.E-4).
I2DLAG: simplification for 2D problems (0 if 3D, 1 if 2D).
DIAMR: diameter ratio under which the particle tracking process stops (default value: 0.1).
CELT: coefficient for eddy life time (for turbulent flows only): 0.28.
KOUTPT: enables to write the ASCII ’.tr.out’ file containing the output of the trajectories
(default value: 0). Contact support at support@numeca.be to have more details on this file.
7-8 FINE™
Fluid-Particle Interaction Optional Models
7-2.6 Theory
In a Lagrangian approach the particle cloud is modelled by a finite number of computational parti-
cles, representing a group of real physical particles with similar properties. The particles are tracked
separately, each as a single probe particle, through the flow field.
∆V
F = ma = m ------- (7-1)
∆t
where a is the acceleration vector of the particle of velocity V and F is the sum of the friction force
between the fluid and the particle and the eventual inertia and gravity forces:
The gravity force is defined by the user on the Flow Model page:
Fgravity = mg
It is assumed that the mesh blocks for which a rotation of the reference system is introduced in the
carrier flow equations (when rotation of the reference system is selected under the Parameters
thumbnail of the Fluid-Particle interaction page) are also considered as rotating ones for the parti-
cles calculation. The inertia force due to the rotation of the reference system contains the two terms
related to the inertia and Coriolis effects:
F inertia = – m [ ω × ( ω × r ) + 2ω × V ] (7-3)
Fd is the drag force between the flow and the particles, calculated according to the Stokes law:
D
Fd = – ----- ( V – U ) (7-4)
Fs
where D is the diameter of the particles, U is the velocity of the fluid and Fs depends on the viscos-
ity of the fluid and on the Reynolds number Re p based on D:
F s = 1 ⁄ ( 3πµC ) (7-5)
where:
ρ ( V – U )D
C = 1 (simple Stokes law for Re p ≡ --------------------------- « 1 )
µ
0.5
or C = 1 + 0.179Re p + 0.013Re p (standard curve approximation).
Up to now, no model is incorporated in which the mass and the volume of the particle would change
(for instance by combustion or evaporation).
FINE™ 7-9
Optional Models Fluid-Particle Interaction
2
τ drag = ( ρD ) ⁄ ( 18µC ) (7-6)
For a stationary reference system (no inertia force) and in the case of no gravity, the relaxation time
is the time required by the particle to reach the flow velocity U, whereas in the presence of an iner-
tia and/or a gravity force, the velocity reached by the particle after the relaxation time is not U, but
U + g . τdrag.
In some cases (for instance with very small particles) the relaxation time is very small, which leads
to a large number of time steps per mesh cell, and hence to a larger CPU time for the particle track-
ing algorithm. A faster convergence may then be reached by using the so-called exponential scheme
for the time integration step.
∆V
F interphase = m ------- – Finertia – F gravity (7-7)
∆t
U = U + U' (7-8)
The fluctuation U' is randomly sampled from the normal Gaussian distribution with the probabil-
ity-density-function f:
u'
⎛ u' ⎞2
1
f ( u' ) = ----------------- exp ⎜ – ----------⎟ and
2 ⎝ 2u' 2⎠
U' = ∫ f ( u' ) du' (7-9)
2πu' –∞
The individual components of the velocity fluctuation are assumed to be non-correlated. The mean
square deviation is evaluated from the turbulent kinetic energy:
7-10 FINE™
Conjugate Heat Transfer Optional Models
2 2
u' = --- k (7-10)
3
It is accepted that this sample velocity fluctuation remains unchanged within the turbulent eddy
during its life time. The eddy life time Te and size Le are calculated according to:
k 1⁄2
Te = C lt -- and Le = Te ⎛⎝ 2--- k⎞⎠ (7-11)
ε 3
The calculated fluctuation gives the instantaneous flow velocity, which is used while tracking the
particle. As soon as the life time has expired or the particle has left the eddy, a new fluctuation is
chosen as described before, assuming no correlation with the previous sample.
Note that turbulence is therefore taken into account only for the different linear and
non linear k-ε models. In the particular case of the Baldwin-Lomax model, turbulence
does not have a direct influence on the particles trajectories.
7-2.7 References
Chang, K. C., Wang, M. R., Wu, W. J., Liu, Y. C., Theoretical and Experimental Study on Two-
Phase Structure of Planar Mixing Layer, AIAA Journal, Vol. 31, No 1, pp 68-74, 1993.
The current version of the model has been validated for steady state applications. It assumes that
some of the blocks of the multiblock mesh are "solid", whereas the other ones are "fluid". The
connections between the solid and fluid blocks become "thermal connections" along which the fluid-
solid coupling is performed.
First section 7-3.2 explains how to set up a project involving the Conjugate Heat Transfer model
under the FINE™ user environment. In section 7-3.3 a theoretical description is provided, present-
ing the approach used to solve the conduction problem and describing the adopted solid-fluid cou-
pling strategy.
FINE™ 7-11
Optional Models Conjugate Heat Transfer
The only action required from the user to set up a computation involving the Conjugate Heat Trans-
fer model can be decomposed in three steps:
1. Determine the Solid and Fluid Blocks.
2. Set the Global Parameters.
3. Define the thermal connections.
Those steps are described in the following paragraphs in more detail.
7-12 FINE™
Conjugate Heat Transfer Optional Models
• full non-matching connections (one patch or one group of patches on each side). The connec-
tion algorithm is implemented in a general way, allowing on both sides a grouping of solid and
fluid patches. This is useful for instance in order to facilitate the coupling of a turbine block
with holes with the external flow. One single connection can be built, with on the inner side a
group of patches including the blade surface and the outer face of the holes.
7-3.3 Theory
This section describes the Conjugated Heat Transfer model implemented within EURANUS:
• the technique used to solve the conduction in the solid bodies,
• the thermal coupling strategy.
∆T
ρC p ------- Vol = Σ ⎛ k∇T⎞ n∆S (7-12)
∆τ ⎝ ⎠
where Vol is the volume of the control volume, n∆S is the normal vector to the faces of the control
volume and k,ρ and Cp are respectively the conductivity, density and specific heat of the solid body.
This equation is applied to calculate the change of temperature T at a given time step resulting from
the equilibrium of the conduction fluxes through the faces of the control volume.
One can notice that the absence of any advective contribution in the equilibrium permits to avoid the
use of artificial dissipation (case of a centred scheme). It also imposes to adopt a pure diffusive
approach for the calculation of the time step. For a 2D problem the time step is calculated according
to:
∆τ ρC Vol
--------- = ---------p ----------------------------------------- (7-13)
Vol k
Si + Sj + 2Si·Sj
2 2
One can notice that once the 2 above equations are put together the heat capacity Cp is eliminated
from the equation, and is hence not required as input data.
Exactly as for the flow solver the adopted time integration approach is a 4- or 5-step Runge-Kutta
scheme whose efficiency is largely enhanced by the use of an implicit residual technique and of a
multigrid strategy.
This finite volume approach can be used for both steady and unsteady conduction problems. Its
efficiency has been compared for steady state applications to the classical Gauss-Seidel and PSOR
schemes often employed for the resolution of pure elliptic equations. The results have shown that the
use of a time marching approach together with multigrid is much more efficient.
FINE™ 7-13
Optional Models Conjugate Heat Transfer
connections", along which the heatflux and temperature are imposed to be equal.
The calculation is performed simultaneously on the fluid and solid blocks. The thermal coupling is
applied each time the boundary conditions are applied. Note that no particular procedure is required
in order to initiate the calculation. The full multigrid strategy is very efficient in order to rapidly
establish an overall temperature distribution in the domain.
On both solid and fluid sides, the thermal boundary condition imposes a temperature profile. This
profile is iteratively recalculated in order to ensure equality of the heatfluxes calculated on both
sides:
hf(Tw-Tf)=-hs(Tw-Ts), (7-14)
where hf and hs are the heat transfer coefficients and Tf and Ts are the inner temperatures. The heat
transfer coefficients are calculated in a simple way, assuming a sufficient orthogonality of the mesh:
k
h = ------ , (7-15)
∆y
where k is the local conductivity and ∆y is the distance between the inner cell centre (where the
temperature of Tf and Ts is taken) and the wall.
i i i
Tw = ∑ T n w n + T inner W i with ∑ wn + Wi = 1 . (7-16)
elmts elmts
In the generic case of Figure 7.3.2-5:
• the cell 1 is not connected and the above equation is applied with Wi=1,
• the cell 2 is partially connected and 0<Wi<1,
• the cell 3 is fully connected and Wi=0.
The heatflux density is calculated as the summation of the contribution of the elements:
i ⎛ i ⎞ i
ϕw = ⎜
⎝
∑ ϕ n S n⎟ ⁄ S
⎠
(7-17)
elmts
Note that the heatflux density is saved instead of the heatflux so that the thermal connection bound-
ary condition is compatible with the boundary condition imposing the heatflux density. In the rou-
tine imposing the heatflux into the residual the heatflux density is re-multiplied by the cell surface.
It should be mentioned that hybrid connections are allowed, this means that each side of the inter-
face can be composed of fluid and solid patches.
7-14 FINE™
Cooling/Bleed Optional Models
7-4 Cooling/Bleed
7-4.1 Introduction
The Cooling/Bleed model allows both the simulation of cooling flows injected through solid walls
into the flow or bleed flows where mass flow leaves the main flow through solid wall.
The adopted technique makes use of additional sources of mass, momentum and energy located at
given points or along given lines of the solid walls and does not require the meshing of the cooling
flow injection channels. The objective of this model is not to describe the details of the cooling flow
itself but rather to consider its effect on the main flow.
In the next sections the following information is given on the Cooling/Bleed model:
• section 7-4.4 provides a theoretical description of the approach used to solve the cooling flow
problem.
• section 7-4.3 gives a summary of the available expert parameters related to cooling or bleed
flow.
• section 7-4.2 describes how to set up a project involving the Cooling/Bleed model under the
FINE™ user environment.
FINE™ 7-15
Optional Models Cooling/Bleed
a)
b)
c)
The list box on the top of the page (a) displays all the injectors (Cooling or Bleed) imported or
added by the user. The buttons below the list enable to add, edit, remove or import (more details on
data file in section 7-4.5) injectors (b). Clicking with the left mouse button over a selected injector
will display its corresponding parameters in the lower part of the page (c). Clicking with the right
mouse button over a selected injector will display a popup menu enabling to add/edit/remove the
injector.
On the first page of this wizard the user can enter the name, the role, the solid patches, the geometry
definition and the type of positioning of the injector.
7-16 FINE™
Cooling/Bleed Optional Models
It is strongly recommended to use only standard english letters, arabic numerals and
underscore "_" symbol when naming an injection sector. Non-standard characters may
not be recognized or may be misinterpreted when using the project on other computers.
In case a slot (or a line) should traverse several patches (as for instance in a HOH
mesh configuration) all the patches should be selected and one single slot should be cre-
ated.
FINE™ 7-17
Optional Models Cooling/Bleed
• Relative mode: this mode is used for blade walls and is valid for AutoGrid meshes only. In
this case, the spanwise and streamwise positions of the hole (or of the 2 extremities in the case
of a line of holes) are specified.
• Space Cartesian mode: the user specifies the (X,Y,Z) Cartesian coordinates of the hole or of
the 2 extremities for a line. For a slot the user enters the radius and the Z-coordinate of one
point of the slot.
• Space Cylindrical mode: the user specifies the (R,θ,Z) cylindrical coordinates of the hole or
of the 2 extremities for a line. For a slot the user enters the radius and the Z-coordinate of one
point of the slot.
• Grid mode: the user specifies the (I,J,K) indices of the hole or of the 2 extremities of a line of
holes.
The injection sectors of type "Slot" can not be presented in relative coordinates.
To cancel the modifications to the injector definition and close the wizard click on the Cancel but-
ton.
Once the injector sector role, type and mode are correctly set, click on Next>> to go to the next
page of the wizard. On this page, depending of the role, type and mode of injector, the flow param-
eters and geometry/positioning parameters have to be defined in the second page of the wizard.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
7-18 FINE™
Cooling/Bleed Optional Models
As for the old algorithm, the holes are located by means of a spanwise and a streamwise relative
position. The algorithm can be applied to H, I or O-meshes generated by AutoGrid. For O-mesh the
blade surface is divided in two different sub-patches (one for the pressure side and one for the suc-
tion side). Therefore, the user has to select on which side of the blade (Side1/Side2) the injection
will be located. A button has been added in the interface in order to select the side. This button is
active only for O-meshes.
Only the version of AutoGrid provided with FINE™ 6.1-1 (and latter versions) can be
used to generate the mesh and this relative positioning is only available for blade sur-
faces.
The location is defined as a fraction of the arc length from hub to shroud and from the leading edge
to the trailing edge. Spanwise and streamwise coordinates are computed for each cell vertex. The
closest point found in this relative coordinate frame is the one that is concerned by the injection.
An option has been set-up in order to use the axial chord (Axial Coordinates button) instead of the
arc length (only available for axial machines).
The relative positioning does not accept meshes with a blunt at the leading edge
FIGURE 7.4.2-10 . Relative positioning of a hole at 70% spanwise and 50% streamwise.
Finally, when a cooling injection sector has been selected, three ways to impose the injection direc-
tion are proposed through the interface: Cartesian, cylindrical or grid indices.
FINE™ 7-19
Optional Models Cooling/Bleed
When Grid Indices is selected, the injection direction is provided by the user through 2 angles
expressed in degrees (depending of the units selected in the menu File/Preferences). Each angle is
measured with respect to the local normal and one of the 2 grid line directions tangent to the wall.
I ( J, K )
In case of an = ct surface, the angles are measured with respect to the grid lines
J ( K, I )
K ( I, J )
respectively. Figure 7.4.2-11 illustrates the definition of angle βk for an I = cst surface.
I=constant surface
V
J
βk
n
βk is the angle of the velocity projection in the (K,n) plane with respect to the normal
n to the surface.
FIGURE 7.4.2-11 Angle definition in the cooling flow page of the FINE™ interface
Vi, j, k
β i, j, k = atan ⎛ -------------⎞ (7-18)
⎝ V ⎠
n
For example, in order to make the velocity direction oriented along the normal, the two angles
should be set equal to 0. A 90 degrees angle is not allowed as it will induce a velocity vector tan-
gent to the corresponding grid line direction, with null mass flow.
For holes located on the blade walls and when the mesh is generated with Numeca’s AutoGrid soft-
ware, the K- and J- lines are the two tangent directions along the blade wall, respectively oriented in
the streamwise and spanwise directions.
e.2) Relative positioning for a line of holes
The user specifies the diameter of the hole in case of a line of holes. Furthermore, the number of
holes has to be specified and will be uniformly distributed on the line.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
As for the old algorithm, the line of holes is defined by two points (holes) that are located by means
of a spanwise and a streamwise relative position. The algorithm can be applied to H, I or O-meshes
generated by AutoGrid. For O-mesh the blade surface is divided in two different sub-patches (one
for the pressure side and one for the suction side). Therefore, the user has to select on which side of
the blade (Side1/Side2) the injection will be located. A button has been added in the interface in
order to select the side. This button is active only for O-meshes.
7-20 FINE™
Cooling/Bleed Optional Models
For each point of the line the algorithm works as for the single hole positioning.
Only the version of AutoGrid provided with FINE™ 6.1-1 (and latter versions) can be
used to generate the mesh and this relative positioning is only available for blade sur-
faces.
The location is defined as a fraction of the arc length from hub to shroud and from the leading edge
to the trailing edge. Spanwise and streamwise coordinates are computed for each cell vertex. The
closest point found in this relative coordinate frame is the one that is concerned by the injection.
An option has been set-up in order to use the axial chord (Axial Coordinates button) instead of the
arc length (only available for axial machines).
The relative positioning does not accept meshes with a blunt at the leading edge
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
e.3) Cartesian positioning for a single hole
The user specifies the diameter of the hole in case of a single hole.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
As for the old algorithm, the holes are located by means of the (X,Y,Z) Cartesian coordinates.
For a single hole, the cell that is concerned by the injection is the closest cell face cen-
tre to the defined point. In case of repetition, the point is automatically transferred to the
meshed patch
FINE™ 7-21
Optional Models Cooling/Bleed
Furthermore, in order to ease the capture of the right coordinates, the button Get Coordinates ena-
bles the user to interactively select the point in the mesh view after opening the mesh through the
menu Mesh/View On/Off or Mesh/Tearoff graphics (only available on UNIX).
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
e.4) Cartesian positioning for a line of holes
The user specifies the diameter of the hole in case of a line of holes. Furthermore, the number of
holes has to be specified.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
As for the old algorithm, the line of holes is defined by two points (holes) that are located by means
of the (X,Y,Z) Cartesian coordinates.
For the two points defining the line, the cells that are concerned by the injection are
the closest cell face centre to the defined point. In case of repetition, the point is automat-
ically transferred to the meshed patch
FIGURE 7.4.2-14 Geometry Parameters for space Cartesian positioning of a line of holes
7-22 FINE™
Cooling/Bleed Optional Models
Furthermore, in order to ease the capture of the right coordinates, the button Get Coordinates ena-
bles the user to interactively select the point in the mesh view after opening the mesh through the
menu Mesh/View On/Off or Mesh/Tearoff graphics (only available on UNIX).
A line is drawn from those two points (this line do not necessarily follow the surface of the patch,
see Figure 7.4.2-15). The segment is then uniformly divided into N-1 sub-segments (where N is the
number of holes on this line). Each N-2 limits of the sub-segment is then projected on the patch (the
closest cell face centre is found). Those are the cells where the injection will be applied. They are
not necessarily uniformly distributed on the patch.
For a mesh with an important clustering, or for an important number of points in the
line, several projected points could end on the same cell. This is taken into account in the
algorithm. The mass flow injected is then multiplied by the relevant factor. Therefore,
one should be aware that the number of cells that are highlighted in the interface should
not strictly match the number of hole in the line.
within
Parametric Space
without
Parametric Space
FIGURE 7.4.2-16 Definition of a line of holes without or within the parametric space
FINE™ 7-23
Optional Models Cooling/Bleed
On the contrary for a mesh with a relatively poor spatial discretisation, the location of the cell face
centres where the injection will be applied will not be strictly on a line. A limitation of this algo-
rithm is that the line does not necessarily follow the patch surface. Therefore, defining a line of
holes at the hub of a gap (upper surface of a rotor blade) could produce the problem presented on
the left side of Figure 7.4.2-16. The same problem can arise generally on a convex surface. In order
to solve this problem, it has been decided to allow the user to define a line of holes following a par-
ametric surface defined by the grid lines by activating Parametric Space. Therefore, a line of hole
on a convex surface could be more easily defined (right side of Figure 7.4.2-16).
One limitation of this option is that a line of hole cannot cross a connection. For
instance defining a line of holes (in the parametric space) close to the trailing edge of a O
mesh will produce the result sketched bellow.
POSSIBLE
For turbomachinery configuration with periodic conditions, points could be located either on the
meshed passage or on one of its repetitions. Therefore, the injections should automatically be trans-
ferred to the meshed patch. Two methods are available.
Meshed Meshed
Repetition Repetition
Patches Patches
7-24 FINE™
Cooling/Bleed Optional Models
• In FINE™ 6.1, the method was transferring only the extremities of the line of holes to the
meshed patch.
• In FINE™ 6.2, the method is rotating only the final position of the injections in order to be
located on the meshed patch. As illustrated in Figure 7.4.2-18, those two options could lead to
rather different results.
In case of multi-patch selection the algorithm described above is applied independently for each
patch. Therefore, a direct definition of a line of hole crossing several patches is not possible. In this
case it is recommended to define as many line as necessary (see figure bellow). In order to over-
come this problem the multi-patch selection has been restricted to the slots defined by Cartesian or
cylindrical coordinates.
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
e.5) Cartesian positioning for a slot
The user specifies the width of the slot. has to be specified.
The multi-patch selection has been restricted to the slots defined by Cartesian or cylin-
drical coordinates.
FINE™ 7-25
Optional Models Cooling/Bleed
For the positioning of the slot, the radius and z values are asked to the user. The closest cell face of
the patch to this point is found. The radius and axial position of this cell face centre become the slot
position in the (r,z) plane. It should be noted that this slot position is not identical to the initial user
input. For most axial machines the radius can be easily identify by calculating the closest point on
the patch. In this case only the axial position of the slot is required in the interface. For radial
machines, or for axial machines with complex vein geometry, the two components of the original
point are required (Figure 7.4.2-21).
Axial Machine
Radial Machine
The algorithm that defines the cell faces that are concerned by this slot is the following one. For
each cell face of the patch, the two extremities in the (r,z) plane of this polygon (A,B and A’,B’) are
identified (i.e., the two vertices with the maximum distance in this plane). They define the limits of
the cell face in the (r,z) plane. If those two points are called A,B for cell 1 and A’,B’ for cell 2, and
S is the slot location specified by the user in the (r,z) plane through FINE™, as presented on
Figure 7.4.2-22, S is only included in the domain of cell 1 limited by A,B in the (r,z) plane, there-
fore, the cell 1 is the only cell concerned by the slot S.
FIGURE 7.4.2-22 Identification of the cells concerned by the slot S in the (r,z) plane. δ is the
distance between A and B.
7-26 FINE™
Cooling/Bleed Optional Models
On the contrary, S is not inside the domain defined by A',B'. Thus, the cell 2 will not be concerned
by the slot.
For O-meshes with small cells close to leading or trailing edges, this algorithm could
leads to missing cells. This problem is due to a (r,z) position specified by the user in
FINE™ for the slot that is not precise enough.
Furthermore, in order to ease the capture of the right coordinates, the button Get Coordinates ena-
bles the user to interactively select the point in the mesh view after opening the mesh through the
menu Mesh/View On/Off or Mesh/Tearoff graphics (only available on UNIX).
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
e.6) Grid positioning for a single hole
The user specifies the diameter of the hole in case of a single hole.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
The grid positioning used in EURANUS depends on the grid indices of the centre of the cell faces.
It is different from the grid indices used in IGG™ and in CFView™. The grid indices are directly
used by the flow solver in order to introduce the flow of mass, momentum and energy at this cell
face.
FINE™ 7-27
Optional Models Cooling/Bleed
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
e.7) Grid positioning for a line of holes
The user specifies the diameter of the hole in case of a line of holes. Furthermore, the number of
holes has to be specified.
Each hole is circular. The diameter of the hole is specified by the user but a hole is
included in one cell only even when the diameter exceeds the cell width.
From the two cell face centres defined in the interface, two geometrical points are defined by their
(x,y,z) coordinates. Then, a line is constructed in the same way as for the Cartesian positioning
(section e.4). In this case there is no problem of repetition because the two extremities of the line
are located on the meshed patch. Furthermore, concerning the Parametric Space option, refer to
section e.4) on page 7-24 for more details.
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
7-28 FINE™
Cooling/Bleed Optional Models
Finally, three ways to impose the injection direction are proposed through the interface: Cartesian,
cylindrical or grid indices. When Grid Indices is selected, refer to the end of section e.1) for more
details.
f) Flow Parameters
The user has to specify the Mass flow: the specified mass flow is actually the total mass flow
through all the blades and through all the holes in the case of a line of holes. Thus the mass flow
through one hole is the specified mass flow divided by the number of blades and the number of
holes. The mass flow is often responsible for initialization troubles as it might be too high. The
FINE™ interface allows a tuning of the specified mass flow through an expert parameter. This is
described in section 7-4.3.
For cooling flow the mass flow specified in FINE™ will be considered as positive and
as negative for bleed flow.
FINE™ 7-29
Optional Models Cooling/Bleed
Furthermore, the Arrows scaling factor parameter enables the user to control the shape of the
arrows representing the injection direction after opening the mesh through the menu Mesh/View
On/Off or Mesh/Tearoff graphics (only available on UNIX).
For cooling additional parameters have to be defined:
• static or total temperature.
• turbulent kinetic energy and dissipation or turbulent intensity when the k-ε model is used.
The parameters area will then be overlapped by the graphic window, and the small control button
on the upper - left corner of the graphic window can be used to resize the graphics area in order to
visualize simultaneously the parameters and the mesh (see section 2-10).
Mesh/Tearoff graphics is only available on UNIX and enables to open the mesh in a
new window.
7-30 FINE™
Cooling/Bleed Optional Models
Furthermore, new outputs have been added into FINE™ on the Outputs/Computed Variables page
under the thumbnail Solid data in order to enable the visualisation in CFView™ of the injected flow
and the injection direction.
b) ZMNMX
In the old algorithm (COOLFL=0) for the relative mode, the injector position in streamwise direc-
tion can be provided with 0 and 1 defined in two different ways:
ZMNMX = 0 (default): the leading and trailing edges are the grid edges, i.e the points where the
periodic lines of the mesh meet the blade walls
= 1: the points located at respectively the minimum and maximum axial positions.
c) COOLFL
There is a possibility to specify the cooling/bleed flow data through an external input file with the
extension ’.cooling-holes’. This file can be useful in case the flow calculation is integrated into an
automatic procedure. For more details on the format of the file, refer to section 7-4.5.
COOLFL = 0: the old algorithm of the cooling/bleed model will be used (model used in FINE™
6.1 and previous releases). The backward is ensured through this parameter set at 0.
= 1: the old algorithm of the cooling/bleed model will be used (model used in FINE™
6.1 and previous releases) based on an external file with the extension ’.cooling-holes’ located in
the corresponding computation subfolder will be read and the data specified through the FINE™
interface will be ignored.
= 2 (default): the new algorithm of the cooling/bleed model will be used.
FINE™ 7-31
Optional Models Cooling/Bleed
d) MAXNBS
The maximum allowed number of injection sectors is controlled through the expert parameter
MAXNBS (the default value is 50). If the input file contains a higher number of sectors the param-
eter should be modified.
7-4.4 Theory
If COOLFL=0, the model consists in the addition of local source terms to all the flow equations
(mass, momentum, energy and k-e equations when a 2-equations turbulence model is used). These
terms are added as fluxes through the solid boundary cells where a cooling hole has been placed by
the user. Fluxes are computed as:
mass
• Mass: F = ρV n S
momentum
• Momentum F x, y, z = ρVx, y, z V n S + p∗ n x, y, z S
energy
• Energy F = V n ( ρE + p∗ )S
k, ε
• k,ε-equations: F = ρV n ( k, ε )S
where ρ is the density, Vn is the velocity component normal to the solid wall and S is the cooling
hole surface.The increased static pressure is computed as:
2
p∗ = p + --- ρk (7-19)
3
and the total energy E* is:
V2
E∗ = e + --- + k (7-20)
2
The density required to calculate the velocity components is obtained from the computed pressure
field along the walls and from the imposed static or total temperature.
Note that for the momentum equations only the velocity term is added as a source
term. The pressure term already exists even in the absence of holes.
If COOLFL=2, the cooling injection/bleed are treated as additionnal inlet or outlet boundary condi-
tion.
Finally, the turbulent kinetic energy and dissipation can either be directly imposed by the user or be
calculated from the turbulent intensity. In that case, the turbulent kinetic energy is obtained as:
2
k = 1.5 ( Tu ⋅ V ) (7-21)
where V is the velocity magnitude and Tu is the turbulent intensity. The turbulent dissipation is
obtained from the relation:
0.75 1.5
Cµ k
ε = -------------------
- (7-22)
0.1D
where D is the diameter of the hole and Cµ = 0.09.
It should be noted that the velocity remains equal to zero at the walls even in the cool-
ing flow region. Thus the boundary condition on the velocity is left unchanged and the
7-32 FINE™
Cooling/Bleed Optional Models
effect of the cooling flow will be observed in the surrounding cells. However, the tem-
perature field will be affected on the wall itself.
Each injection sector should contain all the appropriate data. If one data is missing, the
flow solver will stop and a message will appear on the screen.
b) Boundary patch
The user should specify the block and face number, as well as the boundary patch on which the
holes are located. The face number can be 1 (I=1), 2 (J=1), 3 (K=1), 4 (K=KM), 5 (J=JM), 6 (I=IM)
The patch number is only required if several solid patches are present on the face (for instance on a
hub or shroud wall).
Note that this "blade wall" positioning option is only valid for H- or I-type meshes, the
FINE™ 7-33
Optional Models Cooling/Bleed
pressure and suction surfaces being assumed to be located on two opposite faces of the
mesh.
c.2) space Cartesian (X,Y,Z) positioning data
• in case of a hole or of a line of holes the user specifies the Cartesian coordinates of the hole or
of the 2 extremities.
• in case of slot the user specifies the radius and the Z-coordinate of one point of the slot.
A searching procedure is applied to find the closest solid wall cell to the specified point(s). The
point is periodically repeated around the machine axis so that the specified point does not need to
be located in the blade passage used for the computation.
c.3) Grid positioning data
The user specifies the grid indices of the hole, of one point of the slot or of the 2 extremities of the
line of holes. Depending on the face number, only 2 indices have to be provided (for instance, along
the I=1 face, only the J and K indices are required).
e) Flow properties
• the mass flow through the hole(s): the specified mass flow is the total mass flow through all
the blades. In case of a line of holes, the specified mass flow is the overall mass flow through
all the holes. Therefore the mass flow through one hole is the specified mass flow divided by
the number of blades and by the number of holes. Mass flow has to be negative for Bleed flow.
• the static or total temperature (if both are present only the static temperature will be read).
• he turbulent kinetic energy and dissipation, or the turbulent intensity (if the k-ε model is used).
The kinematic turbulent viscosity (if Spalart-Allmaras model is used) cannot be speci-
fied through an external file ’cooling-holes’.
f) Velocity direction
Finally, when a cooling injection sector has been selected, three ways to impose the injection direc-
tion are proposed through the interface: Cartesian, cylindrical or grid indices. When the injection
direction is provided by the user through grid indices, 2 angles have to be expressed in degrees.
Each angle is measured with respect to the local normal and one of the 2 grid line directions tangent
to the wall. The user specifies 2 of the following 3 angles (in degrees), depending on the face on
which the hole is located:
• Angle with respect to I-grid line Vi , j , k
• Angle with respect to J-grid line β i , j , k = arctg ( )
Vn
• Angle with respect to K-grid line
Note that the two angles should be equal to 0 degree to make the velocity direction oriented in the
normal direction to the blade wall, whereas a 90 degrees angle induces the velocity vector to be tan-
gent to the corresponding grid line direction.
In the case of holes located on the blade walls, and if the mesh has been generated with the AutoG-
rid software, the K- and J-lines are the two tangent directions along the blade wall, respectively ori-
ented in the streamwise and spanwise directions. The K-angle is the streamwise angle, whereas the
J-angle is the spanwise angle.
7-34 FINE™
Cooling/Bleed Optional Models
a) Global Data
Keyword
Single hole (1), line(2) or slot(3) type
Number of holes (if line) number of holes
Diameter (single hole or line) diameter
Width (slot) width
Block Number block
Face Number face
Patch Number (only if several patches) patch
Input Mode (1-relative, 2-Cartesian(X,Y,Z), 3-Grid(I,J,K)) mode
b) Positioning
FINE™ 7-35
Optional Models Cooling/Bleed
c) Flow Properties
Keyword
Mass Flow (negative if bleed flow) mass flow
Static Temperature (only for cooling flow) static temperature
Total Temperature (only for cooling flow) total temperature
Turbulent Kinetic Energy (only for cooling flow and k-ε model) width
Turbulent Dissipation (only for cooling flow and k-ε model) turbulent dissipation
Turbulent Intensity (only for cooling flow and k-ε model) turbulent intensity
7-4.5.3 Examples
Example 1 - Bleed Flow through a Slot define in Grid (I,J,K) mode
7-36 FINE™
Transition Model Optional Models
Furthermore, predicting the onset of turbulence is a critical component of many engineering flows.
It can have a tremendous impact on the overall drag, heat transfer, and performances especially for
low-Reynolds number applications. However, most of the turbulent models fail to predict the
transition location.
FINE™ 7-37
Optional Models Transition Model
user input or via a separate guess. It should be pointed out that these transition terms should only be
active when the transition module is used. When it is not used, the fully turbulent version of the
Spalart-Allmaras turbulence model remains unchanged.
In the next sections the following information is given on the Transition Model:
• section 7-4.4 provides a theoretical description of the approach used to solve the transition
region.
• section 7-4.3 gives a summary of the available expert parameters related to the Transition
model.
• section 7-4.2 describes how to set up a project involving the Transition Model under the
FINE™ user environment.
A list of all the blades appears in the left box and the remaining area is divided into two boxes for
suction and pressure side, respectively. Four choices are possible:
— Fully Turbulent
— Fully Laminar
— Forced Transition
— Abu-Ghannam and Shaw (AGS) Model
Furthermore, for Forced Transition, the user has to specify the position of the transition lines on each
side of each blade. It is based on the relative stream-wise position and it is defined as a line on the
blade pressure and/or suction sides from hub to shroud. The user specifies two anchor points through
which a straight line will be built by the solver. It should be noted that only blades are transitional
while end walls are fully turbulent.
The transition model is only available for the one-equation Spalart-Allmaras model
using AutoGrid mesh and is applied to blade boundaries and work in both steady and
7-38 FINE™
Transition Model Optional Models
unsteady modes.
R/S interfaces should be correctly oriented (from upstream to downstream) and solid
wall of the blades should be adiabatic.
For the Abu-Ghannam and Shaw model: only one "downstream" R/S upstream of
each blade is allowed. Furthermore, there are no restrictions on the number of inlets.
Finally, a new output has been added into FINE™ on the Outputs/Computed Variables page under
the thumbnail Solid data in order to enable the visualisation in CFView™ of Intermittency (more
details on this quantity in section 7-5.4).
Fully
Turbulent
MultiGrid
Initialization
Full INITRA
MultiGrid Transition
Model
Coarse Grid
Initialization MultiGrid
FINE™ 7-39
Optional Models Transition Model
FTRAST (real): prevents the transition to start in an area defined from the leading edge of the mesh.
The value set is a percentage of the chord upstream of which the transition is not allowed (by
default 0.05: 5%).
ITRWKI (integer): switches on the wake induced transition model. When activated (ITRWKI set
to 1), the free stream turbulence intensity (defined at the inlet or at the rotor/stator interface) and the
turbulence intensity from the boundary layer edge are used to define the transition region. When
this parameter is not checked, only the free stream turbulence intensity is used.
7-5.4 Theory
The Transition Model consists of introducing a so-called intermittency, Γ, defined as the fraction of
time during which the flow over any point on a surface is turbulent. It should be zero in the laminar
boundary layer and one in fully developed turbulent boundary layer.
The intermittency function is defined on each point of the blade surface and at each iteration. It is
extended to the computational domain by a simple orthogonal extension (i.e., each point of the
computational domain has the same intermittency than the closest point on the blade surface).
The intermittency is used to multiply the turbulence production term. For the Spalart-Allmaras tur-
bulence model this gives:
~ ~
S = ΓS (7-23)
where S˜ is the production term in the Spalart-Allmaras in Eq. 4-52. Free-stream wakes (with only
a weak production term) will not be affected by the intermittency. A laminar boundary layer will be
preserved upstream of the transition location and turbulence could freely develop thereafter.
The transition location imposed by the user through the Forced Transition option and
by the AGS Model is only appearing if it has been defined in a turbulent boundary layer.
7-40 FINE™
Transition Model Optional Models
In Forced Transition mode, each transition line is positioned through a point on the hub and a point
on the shroud. By default, these points are defined by a percentage of the arc length
(Appendix 7.5.4-32). However, a special treatment is also available where the transition points are
defined by a percentage of the axial chord (only available for axial machines). This option can be
activated via the expert parameter INTERL that should be set to 1 (default value = 0).
The intermittency, Γ, is then computed on each point of the blade surface (it is set to 0 upstream of
the transition line and tends towards 1 downstream). Each point of the computational domain has
the same intermittency as the one of the closest point on the blade surface. However a special treat-
ment is required close to the trailing edge and in tip gaps.
FIGURE 7.5.4-32 Example of the calculation of a transition point on a blade suction side as a
percentage of the chord length.
A last aspect of the implementation is the post-processing. Spalart & Allmaras propose in their arti-
cle a turbulent index to be able to distinguish turbulent zone from laminar zone. This index is
implemented in EURANUS as a solid data. It is defined as:
1 ∂ν~
it = (7-24)
κ ν ω ∂n
where n is the direction normal to the solid wall. It is near 0 in the laminar zone and near 1 in the
turbulent zone.
FINE™ 7-41
Optional Models Transition Model
and Shaw [1980] and derived from experimental data from transition on a flat plate with pressure
gradients. According to these authors transition starts at a momentum thickness Reynolds number:
⎛ F (λθ ) ⎞
Rθs = 163 + exp⎜ F (λθ ) − τ⎟
⎝ 6.91 ⎠ . (7-25)
θ 2 dU e
λθ =
ν ds (7-26)
where Ue is the velocity at the edge of the boundary layer, s is the streamwise distance from the
leading edge, and θ denotes the momentum thickness in the laminar region. τ is the free stream tur-
bulence level (in %). The function F(λθ ) depends on the sign of the pressure gradient:
F (λθ ) = 6.91 + 12.75λθ + 63.64(λθ ) 2 for adverse pressure gradient (λθ < 0), (7-27)
F (λθ ) = 6.91 + 2.48λθ − 12.27(λθ ) 2 for a favourable pressure gradient (λθ > 0). (7-28)
Therefore, according to these relations, transition is promoted in adverse pressure gradient whereas
it is retarded in favourable pressure gradient.
The range of application of the AGS correlation is 0.1 > λθ > -0.1, and for a free stream turbulent
level ranging from 0.3 to 10%.
Γ = 1 − exp(−0.412ξ 2 ) (7-29)
1
ξ= Max( s − st ,0) (7-30)
λ
where st is the position of the transition onset and s is the current position on the arc from leading
edge to trailing edge. λ is the characteristic extent of the transition region and is determined from
the correlation:
Re λ = 9 Re st
0.75
(7-31)
Close to the trailing edge the intermittency is set to 1. This is necessary in order to allow a turbulent
wake to be generated downstream of the blade. This special treatment is controlled by the expert
parameter FTURBT that is set to 0.95 by default (i.e., only the last 5% of a blade side will have an
7-42 FINE™
Transition Model Optional Models
intermittency of 1). In order to get a tri-dimensional distribution of the intermittency, the intermit-
tency of a cell is the one of the closest cell on a solid surface (hub, shroud, and tip gap are fully tur-
bulent and have an intermittency equal to unity).
FINE™ 7-43
Optional Models Transition Model
7-44 FINE™
CHAPTER 8: Boundary Conditions
8-1 Overview
During the IGG™ grid generation process, the user has to define the type of boundary condition to
be imposed along all boundaries. The parameters associated with these boundary condition types
can be fully defined using the Boundary Conditions page.
This chapter gives a description of the boundary conditions available on the Boundary Conditions
page. In section 8-2 the Boundary Conditions page is described. For expert use section 8-3 gives an
overview of the available expert parameters. In section 8-4 some advice is provided on the combi-
nations of boundary conditions to use. For more detailed information section 8-5 provides a theoret-
ical description of each available boundary condition
Note that the connecting (matching (CON) and non-matching (NMB)) boundary con-
ditions (without periodicity) do not appear in this menu, because they do not require any
input from the user. Only the "periodic" conditions appear under the Periodic menu.
FINE™ 8-1
Boundary Conditions Boundary Conditions in the FINE™ Interface
A common particularity of the five Boundary Conditions pages is their subdivision into two areas.
The left area contains always a list of the different patches that are of the boundary condition type
of the selected thumbnail. A patch can be identified either by the name of the block face to which it
belongs and by its number, or by the numbers of the block and face and by its local index on the
face (if no name is provided in IGG™). The right area is the area where the boundary condition
parameters are specified for the selected patch(es).
One or several patches may be selected in the left area by simply clicking on them. Clicking on a
patch unselects the currently selected patch(es). It is possible to select several patches situated one
after another in the list by clicking on the first one and holding the left mouse button while selecting
the next. To select a group of patches that are not situated one after another in the list, the user
should click on each of them while simultaneously holding the <Ctrl> key pressed.
Several patches can be grouped by selecting them and clicking on the Group button. A dialog box
will appear asking for a group name. The name of the group will appear in red color in the list of
patches to indicate that this is a group of patches. Every change in the parameters when a group is
selected applies to all patches in the group. The Ungroup button removes the group and its patches
are displayed individually in the list.
The Ungroup button is active only when at least one group is selected.
Clicking with the left mouse button on the plus sign + left of a group name in the list will display
the patches included in this group.
If some of the patches have been given names in IGG™, these names will appear in
8-2 FINE™
Boundary Conditions in the FINE™ Interface Boundary Conditions
FINE™. If the patches have been grouped in IGG™, by giving the same name to many
patches, they will appear ungrouped in FINE™ (with block, face and patch number
shown after the name) to avoid contradiction between grouping in IGG™ and FINE™.
Ungroup button will toggle the IGG™ name of the selected patch(es) with the default
FINE™ name (block, face and patch number).
If the Graphics Area window with the mesh topology of the project is opened (Mesh/View On/Off
menu), all the selected patches and/or groups will be highlighted. A click with the right mouse but-
ton over a selected patch will select all the patches and groups in the list that have the same param-
eters.
If the user selects several patches that have different parameters, a warning dialog box will appear.
When clicking on OK all parameters will be set equal to the ones of the first patch. Selecting Cancel
instead will cancel the selection.
(1)
(2)
(3)
FIGURE 8.2.0-2 Two patches have been grouped into a group called HUB.
The right area of all the notebook boundary condition pages is created in a generic way by means of
a resource file, where all the boundary condition parameters are described altogether with their
default values.
FINE™ 8-3
Boundary Conditions Boundary Conditions in the FINE™ Interface
FIGURE 8.2.1-3 Subsonic inlet boundary conditions page for cylindrical problems.
FIGURE 8.2.1-4 Subsonic inlet boundary conditions page for Cartesian problems.
8-4 FINE™
Boundary Conditions in the FINE™ Interface Boundary Conditions
The upper part of the inlet boundary condition page allows the user to select the type of inlet condi-
tion to apply to the selected patches through a series of toggle buttons. There are two categories of
inlet conditions: conditions applicable to subsonic inlets and conditions applicable to supersonic
inlets.
Note that to determine whether an inlet (or an outlet) boundary is subsonic or super-
sonic, the velocity component normal to the boundary should be considered.
For each inlet condition type, the lower part of the page is adapted in order to provide input boxes
only for the physical variables that are required to fully determine the boundary condition. These
variables are for instance: the pressure and temperature (static or total), the velocity components,
the velocity magnitude, the velocity angles (specified in radians), or the mass flow. In case of a sim-
ulation involving the k-ε turbulence model, the inlet values of the parameters k and ε are also
required. A turbulent viscosity entry appears in case the Spalart-Allmaras turbulence model is used.
For the inlet boundary conditions:
• The velocity components are defined in the Cartesian or cylindrical coordinate system,
depending on the selected type.
• The pressure and the temperature values to specify can be the static or the total values.
• For rotating configurations, the specified values are always the absolute quantities.
Taverage + ∆T
Inlet
FIGURE 8.2.1-5 Meridional view of the "Coupling with Outlet ID" functionality
When mass flow is imposed at an inlet patch, the temperature may be chosen to be coupled to an
outlet patch. The coupling is performed through the static temperature. Only in that case, the aver-
age temperature Taverage is computed at the specified outlet and imposed at the inlet as Tinlet = Taver-
age + ∆T (see Figure 8.2.1-5). The quantity ∆T is a constant temperature difference (>=0 or <0)
FINE™ 8-5
Boundary Conditions Boundary Conditions in the FINE™ Interface
specified by the user. This is done in FINE™ by choosing Mass Flow Imposed as boundary condi-
tion type for the inlet and clicking on the Coupling Temperature With Outlet button. A pull-down
menu allows to select which patch of the outlet must be involved in the coupling. The value to be
entered in the static temperature box is ∆T.
To visualize and/or edit the profile data press the small button ( ) on the right side of the label
"profile data". This button opens the Profile Manager with the existing profile. See section 2-12 for
a detailed description of the Profile Manager. It offers the possibility to modify or to create a data
profile interactively. Click on the OK button to set the new profile to the selected patch(es).
The button Surface data toggles 1D and 2D editing modes. It can be used to change the dimension
of an existing profile. 2D profiles are displayed as a "cloud of points". Selecting a point will display
the f(x,y) value in blue in the corresponding column of the Profile Manager.
8-6 FINE™
Boundary Conditions in the FINE™ Interface Boundary Conditions
Three types of subsonic outlet boundary condition are available, as shown in Figure 8.2.2-7. These
three different boundary conditions are described in the following paragraphs.
An option for treatment of Backflow Control can be activated in the case of radial dif-
fusers outlet. The purpose of this option is to control the total temperature distribution
along the exit section. In case the flow partially re-enters the domain through the bound-
ary, the total temperature of the entering flow is controlled so that the entering and outgo-
ing flows globally have the same total temperature.
.
FINE™ 8-7
Boundary Conditions Boundary Conditions in the FINE™ Interface
The pressure imposed at the outlet can be constant or as a function of space and/or time. To define a
profile as a function of space and/or time click on the profile button ( ) right next to the input text
box. The Profile Manager will appear as described in section 8-2.1.3.
(static pressure)
(mass flow)
8-8 FINE™
Boundary Conditions in the FINE™ Interface Boundary Conditions
In a design process, with a variation of the geometry, it is not always known in advance where in the
performance curve the computations are performed. Therefore it is not always possible to choose
the appropriate boundary condition at the outlet for the complete design process.
To overcome this problem this boundary condition allows to impose a relation between the mass
flow and the pressure at the outlet. It is no longer needed to choose between imposing pressure
when working around the chocking part and imposing mass flow when working close to the stall
limit.
The user has to impose a very simple characteristic line defined through 3 parameters: a target out-
let mass flow and a target outlet pressure (at point 2 in Figure 8.2.2-8) and the pressure at zero mass
flow (point 1 in Figure 8.2.2-8). For more detail on this boundary condition see section 8-5.2.
In the present version, periodic connection is not allowed for blocks with different
rotation speed on each side of an interface. Thus no frozen rotor calculation should be
performed with periodic connections.
Normally no user input is required on this page. The only periodic boundary condition cases for
which a user input is required are those in which some of the boundary conditions have to be
applied with a periodicity angle that differs from the global periodicity angle of the block.
In case some input is required this section explains how to define the periodic conditions using the
PERIODIC thumbnail.
The type of periodicity connection between the patches (Matching or Non Matching) can be
entered. A connection is Matching if the numbers of mesh points along the connected patches are
FINE™ 8-9
Boundary Conditions Boundary Conditions in the FINE™ Interface
identical, and if all the corresponding points along these patches coincide. The Non Matching con-
nection requires the use of an interpolation process to establish the connection, whereas a matching
one consists of a single communication of the flow variables.
Some additional characteristics need to be given to fully determine the periodic boundary condi-
tion:
• For rectangular (Cartesian) problems, the user has to specify the translation vector defining the
periodicity. The positive translation vector goes from the current patch to the periodicity
patches.
• For cylindrical problems, the user has to specify the rotation angle in degrees. The rotation
angle is calculated from the connected patch to the current patch, according to the rule of the
right handed system.
FIGURE 8.2.4-10 Solid boundary conditions page in case of a cylindrical boundary condition
8-10 FINE™
Boundary Conditions in the FINE™ Interface Boundary Conditions
When the area defined option is selected, a small picture defining the parameters is displayed (the z
axis is the rotation axis, the r axis is the radial axis). The rotation velocity is set to ω 1 outside the
domain and to ω 2 inside the domain.
The specified range must be a valid range for the used geometry. For example, for an
axial machine the limits for rotation of the hub can be defined by setting the lower and
higher axial limit to appropriate values. In such a case it is important to set the radial lim-
its such that they include the full solid (hub) patch.
b) Thermal Condition
Three options are available for the thermal condition: constant heat flux (in W/m2), adiabatic or iso-
thermal. The imposed heat flux or temperature can be constant on the patch or defined as a profile.
Use the pull down menu to change Constant Value to for example Fct(space). Click on the profile
button to launch the Profile Manager as described insection 8-2.1.3.
∑ F ⋅ nz . (8-1)
S
The projection of the torque along a given direction z , i.e. the couple exerted by the global force
along the rotation axis:
⎛ ⎞
⎜
⎝
∑ r × F⎟⎠ ⋅ z (8-2)
S
In all other cases the force and torque are computed as:
— the lift,
— the drag,
— the moment calculated at (0,0,0) by default.
The direction of the forces and torque as well as the location of the point for the moment can be
determined with the expert parameters IDCLP, IDCDP, IDCMP and IXMP (see section 8-3).
Within EURANUS v5.1-6 it is possible to calculate and to store the partial torque in the corre-
sponding ".wall" file of the computation through the use of the expert parameter IFRCTO within
FINE™ GUI.
FINE™ 8-11
Boundary Conditions Boundary Conditions in the FINE™ Interface
The partial torque is the torque computed for each layer on the whole blade. When IFRCTO is set to
2 (1 or 3), the torque is defined on each J-direction (I-direction or K-direction) layer (i.e. when
using AutoGrid meshes, J-direction corresponds to the spanwise direction).
IFRCTO = 2
IFRCTO = 0
FIGURE 8.2.4-11 File ".wall" when computing partial torque along the blade
Finally, the resulting ".wall" will contain, in addition of the global torque and force on the selected
blade, the partial torque for each layer along the J-direction of the blade as presented in
Figure 8.2.4-11.
8-12 FINE™
Expert Parameters for Boundary Conditions Boundary Conditions
Use an external condition rather than an inlet condition for cases for which it is not
known if the flow enters or leaves the domain.
FINE™ 8-13
Boundary Conditions Expert Parameters for Boundary Conditions
imposed. In addition to that the extrapolation of Vz must be selected. Note also that both flow
angles and total conditions are then treated in the relative mode.
The following parameters have to be defined:
INLREL: allows to specify relative angles for the cylindrical inlet boundary conditions with
total quantities imposed (default value = 0)
= 1: inlet for a stator (extrapolation of the axial velocity only). The expert parameter
OMGINL has also to be specified,
= 2: inlet for a rotor.
ANGREL (for stator only): data use with the expert parameter INLREL (default = 1 0.05 0.05):
1st real: relaxation angle (in degrees),
2nd real: distance (%) from hub where the absolute flow angle is extrapolated,
3rd real: distance (%) from shroud where the absolute flow angle is extrapolated.
8-14 FINE™
Best Practice for Imposing Boundary Conditions Boundary Conditions
FINE™ 8-15
Boundary Conditions Theory on Boundary Conditions
Radial turbine
FIGURE 8.4.3-13 Example of case with strongly varying radius in the inlet1
1. Picture from D. Japikse, N.C. Balines, Introduction to Turbomachinery, Concept ETI Inc. and Oxford Uni-
versity Press, 1994.
8-16 FINE™
Theory on Boundary Conditions Boundary Conditions
V
α = atan -----θ-
Vz
(8-3)
V
γ = atan -----r
Vz
• the magnitude of the velocity and other flow angles δ and ε are specified (OPSEL = 5):
V
δ = acos ------r
Vm
(8-4)
V
ε = atan ------t
Vm
2 2
with V m the meridional velocity ( V m = V r + V z ). This option has been designed to allow
radial inlet with zero axial velocities,
• the cylindrical velocity components are specified (OPSEL = 7).
The static pressure is extrapolated from the interior by Eq. 8-13, which allows to calculate all the
primitive variables on the boundary.
For a turbulent calculation with the Spalart-Allmaras model, the turbulent viscosity has also to be
specified on the inlet boundary. When the two-equation turbulent model is selected, the boundary
values of k and ε have to be specified.
These conditions can also be used for incompressible flows.
FINE™ 8-17
Boundary Conditions Theory on Boundary Conditions
• the module of the absolute velocity vector is obtained by extrapolation from the interior field,
cf. Eq. 8-15. (OPSEL=2)
• the axial velocity component is extrapolated (OPSEL = 3):
( Vz )0 = ( Vz )1 (8-5)
Note that it is assumed that the angular velocity vector is in the z-direction,
ω = ω ⋅ 1z , (8-6)
which makes this direction the axial one. Also note that, as a result of Eq. 8-6 the absolute and
relative axial velocity component are equal.
• the mass flow is extrapolated, assuming that the meridional component of the velocity is the
one contributing to the mass flow.
( Qm )0 = ( Qm )1 , (8-7)
with Q m = ρV m S . (8-8)
This boundary condition is specially dedicated to radial turbomachinery for which problem of
mass flow conservation can arise. It is not interfaced but it can be used by selecting the previous
boundary condition (absolute flow angles with extrapolation of velocity) and setting the expert
parameter IMASFL to 1.
the inlet patches have to be grouped in one group to use this last boundary condition
(extrapolation of mass flow).
b.2) Relative flow angles from axial direction
For a stator calculation it may be convenient to impose the velocity angles of the relative flow at the
exit of the upstream rotor. In case of rotors the user may also prefer to impose relative boundary
conditions. This procedure is only available in case total conditions and flow angles are imposed. In
addition to that the extrapolation of Vz must be selected. Note also that both flow angles and total
conditions are then treated in the relative mode. This boundary condition is not directly available in
the Boundary Conditions page of the interface. It can be used by selecting the previous boundary
condition (absolute flow angles with extrapolation of the axial velocity component) and using the
expert parameter INLREL. In this case, relative flow angles, relative total pressure and relative total
temperature are specified in the interface instead of the absolute values. Set INLREL to 1 in the
case of stators. In the case of rotors (INLREL=2) the approach is exactly the same as the one used
when absolute conditions are imposed.
The adopted approach consists of imposing boundary conditions, which are still defined in the
absolute frame of reference, but which vary from one iteration to the other in order to respect the
imposed relative value. This technique provides a robust algorithm with an under-relaxation of the
evolution of absolute conditions and a special treatment of the hub and shroud boundary layers.
Indeed as the hub and shroud walls are usually not rotating, the relative flow angle tends to 90
degrees in the boundary layer, which complicates the boundary condition treatment if the relative
conditions are imposed. The above approach (switch to the absolute frame of reference) allows
computing the absolute flow angles distribution through an extrapolation in the boundary layers.
The under-relaxation factor as well as the percentage of the distance from the hub and the shroud
where the angle is extrapolated are specified in the expert parameter ANGREL.
The relation between the relative (β) and absolute (α) angles is the following:
8-18 FINE™
Theory on Boundary Conditions Boundary Conditions
ωr
tan α = tan β + ------ (8-9)
Vz
where ω is the speed of rotation of the upstream rotor specified in the expert parameter OMGINL
and r the radial position.
The update of the flow angle is followed by the update of the absolute total conditions according to:
( ωr ) ωrV
2
T 0 = T 0 – ------------- + ------------θ-
a r
2c p cp
a γ ⁄ (γ – 1)
. (8-10)
a r ⎛T ⎞
p0 = p 0 ⎜ ----0r-⎟
⎝ T 0⎠
The initial values of the absolute flow angle and total conditions are set to the value of
the relative ones except with the initial solution for turbomachinery. In the initial solution
for turbomachinery, Eq. 8-9 and Eq. 8-10 are used to compute the initial absolute flow
angle and total conditions from the specified relative values.
b.3) Velocity direction
The user first specifies a direction (given under the form of a vector) that corresponds to the direc-
tion of the absolute velocity vector. Note that this vector does not have to be a unitary vector. The
r,θ,z components of the direction vector are given. E.g. if the user enters the direction (0,1,1) the
user imposes that the θ- and z-components of the absolute velocity vector at inlet are equal and that
the radial velocity is zero.
The absolute total pressure and the absolute total temperature on the inlet boundary are also speci-
fied. These values will be imposed.
The module of the absolute velocity vector is extrapolated from the interior field. Using subscript 0
for the boundary, and subscript 1 for the first internal cell:
V0 = V 1 (8-11)
FINE™ 8-19
Boundary Conditions Theory on Boundary Conditions
V
α = atan -----θ-
Vr
, (8-12)
V
β = acos -----z-
V
with the angles given in radians.
For a turbulent calculation with two-equation model the boundary values of k and ε also have to be
specified. With the Spalart-Allmaras model, the turbulent viscosity has to be specified along the
boundary.
d.2) Static quantities
The flow at inlet is completely specified, by giving the static pressure, static temperature and the
three absolute cylindrical velocity components (ITYPE = 23, OPSEL = 2).
For turbulent models with additional equations, the corresponding quantities are also specified and
imposed on the inlet boundary.
It should be mentioned that the use of this boundary condition requires the presence of
the saturation table (’PSA.atab’) of the fluid in the corresponding subfolder.
The boundary condition is based on the dryness fraction (X), whose value is bounded
between 0 and 1. The inlet thermodynamic must be either saturated or be located inside
8-20 FINE™
Theory on Boundary Conditions Boundary Conditions
The validity of the condensable fluid model is limited to gases with small fraction of
liquid (X > 0.9) or reversely to liquids with small fraction of gas (X < 0.1).
The implementation of the boundary condition is based on extrapolation of the static pressure,
which permits to evaluate the saturated liquid and gas values of the density and enthalpy. The dry-
ness fraction permits to deduce the values of the local static enthalpy and density. The other bound-
ary conditions are then used to calculate the velocity vector.
p0 = p 1 , (8-13)
M0 = M1 , (8-14)
• the module of the absolute velocity vector is extrapolated from the interior field.
V0 = V1 (8-15)
FINE™ 8-21
Boundary Conditions Theory on Boundary Conditions
For a turbulent calculation with the Spalart-Allmaras model, the turbulent viscosity has also to be
specified on the inlet boundary. For a turbulent calculation with a two-equation model, the bound-
ary values of k and ε have to be specified.
8-22 FINE™
Theory on Boundary Conditions Boundary Conditions
The purpose of the treatment of backflow is to control the total temperature distribution along the
exit boundary. In case the flow partially re-enters the domain through the exit boundary, the total
temperature of the entering flow is controlled so that the entering and outgoing flow globally have
the same total temperature.
The outlet patches have to be grouped in one group to use this boundary condition.
This boundary condition is only valid on surfaces with mesh lines at constant radius.
The static pressure in the outlet section is assumed to be uniform in the tangential direction and to
vary in radial direction according to:
2
∂p v θ
= ρ ⋅ ------- . (8-16)
∂r r
The user specifies the static pressure at a specific radius, followed by that radius and the radial
direction. The radial direction is specified by using the block orientation (I-direction, J-direction or
K-direction).
The pressure distribution in the outlet section is then found by integrating Eq. 8-16 with pitchwise
averaged values of Vθ and r.
The other variables are extrapolated following the system of options exposed above. See “Static
Pressure Imposed” on page 22. The treatment of backflow is not implemented for this boundary
condition.
The outlet patches have to be grouped in one group to use this boundary condition.
FINE™ 8-23
Boundary Conditions Theory on Boundary Conditions
The scaling of the velocity vector is not totally free. The value of the scaling factor is constrained
between VELSCA and 1/VELSCA, where VELSCA is an expert parameter available in the Control
variables page in the expert mode.
Given that the pressure is also extrapolated with this condition, it is imperative that the inlet bound-
ary condition fixes the pressure through the total pressure.
Fixing the mass flow is not as robust as to impose the pressure and this is particularly sensitive with
full-multigrid. A pressure is thus requested as input and is imposed during the full-multigrid proc-
ess. Consequently, the mass flow computed at outlet is not exactly the target mass flow during the
computation on the coarse grids.
This option is not recommended in case significant backflow is detected along the exit. The next
option (exit pressure adaptation) is then better recommended.
d.2) Pressure Adaptation
OPSEL = 30
This boundary conditions is an adaptation of the "uniform static pressure imposed" and "static pres-
sure imposed with radial equilibrium" boundary conditions.
The only difference is that an automatic procedure is included that introduces a variation of the
imposed exit pressure at each iteration of the calculation. The pressure is iteratively updatet in order
to reach at convergence the imposed mass flow. The successive pressure modifications are calcu-
lated according to:
·
new old r gas T ref actual imposed
p = p 2
-(Q
+ RELAXP --------------------- –Q ) , (8-17)
L ref
where Lref and Tref are the characteristic length and reference temperature specified in the Flow
Model page. rgas (RGAS) and RELAXP are expert parameters defined in the Control Variables
page. The expert parameter RELAXP introduces an under-relaxation of the successive modifica-
tions of the exit pressure. The default value is 1.
The value of the exit pressure provided by the user is the initial one. In case the radial equilibrium is
used, the initial value of the exit pressure is imposed at midspan.
Contrary to the previous option fixing the mass flow through velocity scaling, it is only after con-
vergence of the procedure that the mass flow reaches the imposed value.
e) Characteristic Imposed
ITYPE = 29
This boundary condition has been implemented to increase the robustness in the frame of a design
process. Indeed it is not always suitable to impose the pressure at the exit because it can drive the
machine to stall when the geometry change affects its characteristic line.
The proposed solution is to impose a relation between the mass flow and the pressure at the outlet.
It is no longer needed to choose between imposing pressure when working around the chocking part
and imposing mass flow when working close to the stall limit.
The user has to impose a line defined through 3 parameters: a target outlet mass flow ( Qt ), a target
outlet pressure ( p Qt ) and the pressure at zero mass flow ( p Q0 ). See section 8-2.2 for a schematic
representation of this boundary condition.
The implementation follows three principles:
8-24 FINE™
Theory on Boundary Conditions Boundary Conditions
• The relation between mass flow and pressure is imposed from a function of the type:
F ( Q, p ) = 0 , (8-18)
with
( p Qt – p Q0 ) 2
F = p – p Q0 – --------------------------
2
-Q . (8-19)
Qt
(8-20)
∂F ∂F
F + ------- δQ + ------ δp = 0 , (8-21)
∂Q ∂p
where the δ are the variations of the mass flow and averaged exit pressure and where F and its
derivatives are computed from the old values of p and Q.
• The local outgoing characteristic variables are extrapolated.
• the relation between p 1 and p 0 is imposed as follows:
p 0 = p 1 + ∆p , (8-22)
with ∆p is constant over the whole outlet and with the subscripts 0 and 1 corresponding respec-
tively to the value in the first outer cell and in the first inner cell.
Combining these three principles and expressing Q and p in function of ∆p at the outlet, Eq. 8-21
can be written as a polynome of second order in ∆p. The suitable root is chosen so that the imposed
pressure always respects the relation defined between the mass flow and the pressure during the
computation.
FINE™ 8-25
Boundary Conditions Theory on Boundary Conditions
• the absolute total temperature is imposed at a value equal to the averaged total temperature
along the outlet (the averaging is performed throughout the region where the flow leaves the
domain.
This treatment permits to avoid the introduction of an uncontrolled total temperature level in the
regions where the flow re-enters the domain, and in this way to better respect the physics of the
flow.
Limitations:
• only for compressible fluid,
• it is applied separately on each outlet patch when an uniform static pressure is imposed (it is
applied to a group of patches if ITYPE=27, i.e. averaged static pressure and mass flow
imposed).
• not available for the outlet boundary conditions where static pressure is imposed with radial
equilibrium.
w∗ w = w 1
(8-23)
3 1
w∗ w = --- w 1 – --- w 2
2 2
with subscripts w,1,2 denoting respectively the wall, the first and the second inner cell.
The tangential part of the extrapolated velocity vector is:
**
w w = w∗ w – ( w∗ w ⋅ n )n , (8-24)
with n the normal to the wall. The velocity vector on the wall is finally found by a scaling process
such that its module equals:
ww = w1
3 1 . (8-25)
w w = --- ⋅ w 1 – --- ⋅ w 2
2 2
The wall density and pressure are obtained through extrapolation from the interior with zero order:
ρw = ρ1
pw = p1
8-26 FINE™
Theory on Boundary Conditions Boundary Conditions
or first order:
3 1
ρ w = --- ρ 1 – --- ρ 2
2 2
. (8-26)
3 1
pw = --- p 1 – --- p 2
2 2
Alternatively the pressure can be solved from the normal momentum equation (ITYPE=15):
n ⋅ ∇p = ρv ⋅ ⎛ v ⋅ ∇⎞ n , (8-27)
⎝ ⎠
which is written in a local curvilinear system as:
∂n ∂n ∂p ∂p ∂p
ρv wall ( v wall ⋅ S i ) ------ + ( v wall ⋅ S k ) ------ = S i ⋅ n ------ + S j ⋅ n ------ + S k ⋅ n ------ , (8-28)
∂ξ ∂ζ ∂ξ ∂η ∂ζ
where ξ, η, ζ represent the coordinate directions, and where it is assumed that the j-direction and η
is the direction away from the wall (which is not necessarily perpendicular to the wall).
∂p
Eq. 8-28 is then solved for ------ , whereas zero-order extrapolation is used for the density.
∂η
a) Adiabatic Walls
The velocity vector on the wall vanishes. ITYPE = 11 or 13 are used for Cartesian boundary condi-
tion, 111 or 113 for cylindrical.
If condition 111 or 113 is used, the angular velocity of the wall (in the absolute frame of reference)
has to be specified.
The velocity relative to the wall should be zero, leading to:
w = – ( u s – uw ) , (8-29)
n ⋅ ∇p = – ρ n ⎛ V ⋅ ∇⎞ V + n ⋅ ⎛ ∇ ⋅ τ⎞ . (8-30)
⎝ ⎠ ⎝ ⎠
The normal pressure gradient can be written as a function of pressure derivatives along the coordi-
nate lines:
1 ∂p ∂p ∂p
n ⋅ ∇p = ------------ S j ⋅ S i + S j ⋅ S j + S j ⋅ S k , (8-31)
Sj Ω ∂ ζ ∂ η ∂ζ
where ξ, η, ζ represent the coordinate in the i, j and k directions, and where it is assumed that the j-
direction is directed away from the wall (not necessarily perpendicular to the wall). S i, j, k are the
surface vectors of the corresponding cell faces.
Combining Eq. 8-30 and Eq. 8-31 and considering that the velocity vanishes on the wall, yields:
FINE™ 8-27
Boundary Conditions Theory on Boundary Conditions
∂p 1 ∂p ∂p Ωρ ⎛ Ω
= – --------2- S j ⋅ S i + S j ⋅ S k – -------- ⋅ n V ⋅ ∇⎞ V + ------- n ⎛ ∇ ⋅ τ⎞ (8-32)
∂η ∂ξ ∂ξ S ⎝ ⎠
Sj
⎝ ⎠
Sj j
Ωρ 2
COR2 = --------2- ⋅ S j r ⋅ ω w (8-33)
Sj
Note that COR2 vanishes for a stator wall.
In the more complicated implementation (ITYPE=11 or 111) the complete normal momentum
equation is considered in the evaluation of the pressure derivative, i.e. the term COR3 is also
accounted for. Applying Gauss' theorem this term can be rewritten as:
n n n Sj
COR3 = ----- Ω ⎛ ∇ ⋅ τ⎞ = -----
S ⎝ ⎠ S ∫ ⎛⎝ ∇ ⋅ τ⎞⎠ dΩ = -----S °∫ τdS ≈ -------
S
2
-
∑τ ⋅S i i . (8-34)
faces
Once ∂p/∂η is determined, the pressure on the wall is obtained as (assuming the direction points
inside the interior field, and with indices w,1 representing the wall and the first inner cell):
1 ∂p
p w = p 1 – --- ------ . (8-35)
2 ∂η
The temperature on the wall is obtained by expressing that the normal temperature gradient van-
ishes, or in terms of the derivatives along the coordinate lines, cf. Eq. 8-31:
∂T 1 ∂T ∂T
= – --------2- S j ⋅ S i + S j ⋅ S k . (8-36)
∂η ∂ξ ∂ξ
Sj
The wall temperature is then found using a similar equation as Eq. 8-35.
The density follows from pressure and temperature through the appropriate relation, depending on
the type of gas.
b) Isothermal Walls
The implementation of this boundary condition is very similar to the adiabatic Navier-Stokes wall.
The only difference is in the determination of the wall temperature, which is fixed to the specified
temperature in the current boundary condition. Again two versions are available, depending on
whether all the terms in the normal momentum equation are taken into account when determining
the wall pressure (ITYPE 12 ≡ 11; 14 ≡ 13; 112 ≡ 111; 114 ≡ 113).
For the cylindrical isothermal walls, #112 and #114, the angular velocity of the wall has to be spec-
ified. For all versions (#12,#112,#14,#114) the wall temperature is specified.
The pressure is determined as in the adiabatic condition and the density follows from pressure and
temperature through the appropriate relation, depending on the type of gas.
8-28 FINE™
Theory on Boundary Conditions Boundary Conditions
± 2c
R n = Vn ± ----------- , (8-37)
γ–1
+
where R n are the Riemann variables associated with the direction n and c is the local speed of
sound.
+ +
For subsonic inflow, R n corresponds to a positive, incoming characteristic. R n is then obtained
from the free stream values:
+ 2c ∞
R n = V ∞ ⋅ n + ----------
- , (8-38)
γ–1
where V ∞ is the free stream velocity, c ∞ the free stream speed of sound.
-
R n , on the other hand, corresponds to an outgoing characteristic and has to be estimated from
inside the computational domain by an appropriate extrapolation:
- 2c
R n = V i ⋅ n – ----------i- , (8-39)
γ–1
where subscript i refers to values from the interior. Linear extrapolation of the variables to the
external boundary will be used to obtain a sufficiently accurate solution:
3 1
q i = --- q 1 – --- q 2 , (8-40)
2 2
where subscript 1 and 2 refer to the first two inner points.
The normal velocity and the local speed of sound can now be obtained on the boundary by adding
+ -
and subtracting both Riemann variables R n and R n :
+ -
Rn + Rn γ–1
-, c wall = ( R +n – R -n ) ----------- .
V w ⋅ n = ----------------- (8-41)
2 4
The entropy on the wall is set to the free stream entropy. All the flow variables can then be
obtained.
+
Corresponding formulas hold for a subsonic outflow with the difference that R n and the wall
entropy are obtained from interior values.
FINE™ 8-29
Boundary Conditions Theory on Boundary Conditions
For supersonic flows, both Riemann variables and the entropy are obtained from free stream values
for inflow, interior values for outflow.
For preconditioning computation, the same principle is used with modified invariants defined as:
2 2 2 2
± pg ⎛ β ⎞ 2⎛ β ⎞ ⎛ vn ⎞ 2
R n = v n – -----------2- v n ⎜ 1 – ----2- – α⎟ −
+ n
v ⎜ 1 – ----
2
- β ,
- – α⎟ + 4 ⎜ 1 – ------
2⎟
(8-42)
2ρβ ⎝ c ⎠ ⎝ c ⎠ ⎝ c ⎠
where
p g = p – p ref
. (8-43)
β = β∗ U ref
2 2
The free stream variables of static pressure, static temperature and the three velocity components
are specified by the user in the boundary condition file.
For a turbulent calculation with a two-equation model, the k and ε values on the boundary are set to
their free stream values (inflow) or extrapolated from the interior (outflow).
8-30 FINE™
CHAPTER 9: Numerical Model
9-1 Overview
To define numerical parameters of the computation, the Numerical Model page allows to define
several aspects of the computation:
• the CFL number,
• the multigrid parameters,
• preconditioning parameters (if applicable).
These parameters are described in the next section.
In expert user mode (Expert Mode) additional parameters are available as shown in Figure 9.1.0-2:
additional multigrid parameters and unsteady parameters. These interfaced expert parameters are
described in section 9-3. Also the non-interfaced expert parameters are described in this section.
More theoretical information on the available parameters related to the numerical model (spatial
and temporal discretizations) is provided in section 9-4.
FINE™ 9-1
Numerical Model Numerical Model in FINE™
9-2 FINE™
Expert Parameters for the Numerical Model Numerical Model
The coarsest grid level available is then 4 along i and 5 along j and k.
• The current grid level is the finest grid level (for each of the i, j and k directions) on which the
computation will take place. The selected levels should be in the range between 0 and the coars-
est grid level available for each of the i, j and k directions. Referring to the example given just
above, the user can do a first run on level 3 3 3 to validate the computational parameters and
then switch to level 1 1 1 or 0 0 0 for a finer solution. It gives a high flexibility to the system
since with one grid, the user can run simulations on several sub-meshes.
All combinations between the i, j and k grid levels, in the specified range, are possible
such as 2 3 1 or 0 3 2...etc.
Press <Enter> after each modification to validate the new specified levels.
The use of multigrid is highly recommended in order to ensure fast convergence of the flow solver.
The mesh used to discretise the space can have multiple grid levels in each direction of the compu-
tational domain i, j and k. These levels are numbered from 0 (finest grid) to N (coarsest grid).
The grid level (N) available in one direction (I, J, or K) is the smallest grid among all
patches on that particular direction (I, J, or K).
The third box, named ’Number of Grid(s)’, permits to significantly accelerate the convergence to
steady state if several grid levels are available: the flow calculation is performed simultaneously on
all the grid levels. This technique is referred to as the multigrid strategy. This number should be
chosen as high as possible, and deduced from the information displayed on the available grid level
and the grid level on which the computation will be performed (’Current Grid Level’).
Finally, the toggle button ’Coarse Grid Initialization’ enables, before calculating the flow on the
mesh contained in the IGG™ files, to perform a preliminary flow calculations on a coarser mesh
automatically created by the flow solver by coarsening the initial one. This provides a rapid estima-
tion of the flow. This technique is referred to Full Multigrid.
FINE™ 9-3
Numerical Model Expert Parameters for the Numerical Model
• The maximum number of cycles to be performed on each grid level is reached. This parameter
is specified through the input box in the Full MultiGrid Parameters area as the Maximum
number of cycles per grid level.
• The residual on the current grid level has dropped a certain order of magnitude as indicated by
the Convergence criteria on each grid level in the Full MultiGrid Parameters area.
• An additional input box permits to select the number of sweeps to be performed on the coarse
grid levels through the input box ’Number of sweeps on successive grid levels’. The amount of
sweeps is the amount of times the Runge-Kutta operator respectively the chosen relaxation
operator is applied (see section 9-4.4 for more detail). In the default configuration of the solver
called ’Linear progression’, the number of sweeps on each level equals the level number (2
sweeps on second level,...). It has been observed in many cases that the convergence rate may be
significantly improved by performing a higher number of sweeps on the coarsest levels. Recom-
mended sets of values are proposed in the Sweeps area through the parameter Scheme definition:
for instance (1,2,16) in case of 3 grid levels. However, it is should be noted that more sweeps
have as a consequence more time required for an iteration.
Condensable fluid option is not compatible with the use of upwind schemes for space
discretization.
If the computation is an unsteady computation, unsteady parameters box appears. Please refer to
chapter section 4-2.1 for a description of those unsteady parameters.
9-4 FINE™
Expert Parameters for the Numerical Model Numerical Model
9-3.2.1 Multigrid
IPROLO: Indicates the order of the prolongation within the multigrid approach:
= 0: is piecewise constant prolongation,
= 1 (default): is linear prolongation (no damping if -1),
= 2 linear prolongation except at boundaries(-2).
When used with a preconditioned formulation, this damping option is somewhat use-
less and can be damaging. FINETM sets it therefore to off when the preconditioning is
involved.
IRESTR: Gives the order of the restriction operator of the multigrid.
= 0 (default): corresponds to linear restriction,
= 1: to quadratic restriction.
MGRSTR: Applies only for multigrid and fixes the type of multigrid cycle. MGRSTR=1: V-cycle;
MGRSTR=2: W-cycle; MGRSTR=3: F-cycle; MGRSTR=4: V-cycle in sawtooth;
MGRSTR=5: W-cycle in sawtooth; MGRSTR=6: F-cycle in sawtooth. Default
MGRSTR=1.
MGSIMP: Applies only for multigrid. If MGSIMP>1 a simplification will be used on coarser lev-
els. If MGSIMP=1 a more dissipative scheme is used on coarser meshes, i.e. first-order
upwind (if IUPWIN=1) or central scheme with increased 4th order dissipation on solid
walls (if IUPWIN=0). If MGSIMP=2, in addition to the simplification for MGSIMP=1,
viscous and source terms (of turbulence) are neglected on coarser levels. No simplifica-
tion for MGSIMP=0. Default MGSIMP=1
RSMPAR: Residual smoothing parameter σ*/σ. A value less or equal to zero means no residual
smoothing.
SMCOR: Only used in case of multigrid. SMCOR has the same meaning as RSMPAR but applies
to the interpolated multigrid corrections instead of the residuals. SMCOR indicates
whether interpolated corrections should be smoothed or not. The default value for
SMCOR is 0.
a) Central Discretization
CDIDTE: Constant used in the exponential damping factor that is used in the numerical dissi-
pation for k and ε, both if central or upwind schemes are used. The default value for
CDIDTE is 100
FINE™ 9-5
Numerical Model Expert Parameters for the Numerical Model
VIS2: Coefficient κ(2) for 2nd order dissipation in the central scheme (default VIS2=1.0).
VIS2KE: Used only for k-ε turbulence model. Coefficient κ(2) for 2nd order dissipation in the
central scheme, as applied to the k and ε equations. Default VIS2KE=1.0
VIS2SW: This parameter switches off all second order dissipation for incompressible fluids
(when set to the default: VIS2SW=1). Second order dissipation is intended for stabi-
lization of the shocks but there are no shocks in incompressible flows. Therefore the
default is to switch off all second order dissipation for incompressible fluids. Some-
times it may be stabilizing to use some second order dissipation. In that case set this
parameter to zero.
VIS4: Coefficient κ(4) for 4th order dissipation in the central scheme. Default VIS4=0.1
VIS4KE: Used only for k-ε turbulence model. Coefficient κ(4) for 4th order dissipation in the
central scheme, as applied to the k and ε equations. Default VIS4KE=0.1
VISNUL: Eliminates 2nd order dissipation fluxes along the physical boundaries (central
scheme). Default VISNUL=1. If VISNUL=0, the procedure is not applied.
b) Upwind Discretization
ENTRFX: Entropy coefficients, see Eq. 9-2, on page 9-8 and Eq. 9-3, on page 9-8. The first
value is the entropy coefficient applied to the linear field, the second value is applied
to the non-linear field. ENTRFX<1. means a constant value, ENTRFX>1. means
scaled with the spectral radius through a factor ENTRFX-1.
IRATIO: Vector of two values, that indicate which effective ratio is to be used, see Eq. 9-
27, on page 9-13. The first value applies to the linear field, the second to the non-lin-
ear field.
IROEAV:1 indicates that Roe averages are to be used to calculate cell face data. See Eq. 9-14, on
page 9-10. If IROEAV=0, arithmetic averaging will be used.
9-6 FINE™
Theory Numerical Model
NSTAGE: number of stages for the explicit Runge-Kutta scheme. In practice, 4 or 5 stage schemes
are mostly used.
IRSMCH: specifies the type of residual/correction smoothing.
= 1: standard version, Eq. 9-58,
= 2 (default in incompressible cases): Radespiel & Rossow, Eq. 9-59,
= 3 (default in compressible cases): Zhu, time step dependent coefficients, see
Eq. 9-61,
= 4: Swanson & Turkel viscous,
= 5: Vatsa, Eq. 9-63. On meshes without high aspect ratio.
ISWV: governs the recalculation of the dissipative residual in the different Runge-Kutta stages.
A value α between 0 and 1 is allowed. Zero means no recalculation of dissipative resid-
ual (latest available value will be used); 1 means recalculation. For intermediate values,
a weighted averaging of the latest dissipation residual and of the preceding one is
applied with the weight α and (1-α) respectively. E.g. for a 5-stage scheme: ISWV = 1 0
1 0 1, means that the dissipative residual is calculated in 1st,3rd and 5th stage only.
ISWV is used in combination with IBOTH.
RSMPAR: residual smoothing parameter σ*/σ. A value less or equal to zero means no residual
smoothing. Default value 2.
9-4 Theory
9-4.1 Spatial Discretization
The discretization in space is based on a cell entered control volume approach. The general Navier
Stokes equation Eq. 9-1 is discretised as:
∂U
∫ Ω∂ t
dΩ +
∑ F ∆S + ∑ F ∆S = ∫ Q dΩ ,
I V (9-1)
faces faces Ω
FINE™ 9-7
Numerical Model Theory
1 1
Ω ∫
∇Φ = ---- ∇Φ dΩ = ---- ΦdS .
Ω °∫ (9-2)
FIGURE 9.4.1-3 : Different control volumes used to calculate gradients in cell corners or cell
faces.
If IFACE=1 (Default), gradients are calculated directly on cell faces. Figure 9.4.1-3 (b) and (c)
illustrate the control volumes used to find the gradient on respectively the cell face i+1/2 and j+1/2.
The option IFACE=1 is more robust and therefore preferred (default), although it is more expensive
as the total number of cell faces is approximately 3 times the number of cell corners.
* 1
( Fn ) i+1⁄2 = --- { ( Fn ) i + ( Fn ) i + 1 } – d i + 1 ⁄ 2 . (9-3)
2
The first term in the right-hand-side of Eq. 9-3 corresponds to a purely central evaluation of the
flux. The term d i + 1 ⁄ 2 represents a numerical dissipation term, that may be an artificial dissipation,
used in combination with central schemes, or the dissipation associated with upwind schemes.
In Eq. 9-3 an arithmetic averaging technique is used. Alternatively, instead of averaging the fluxes
in Eq. 9-3, a flux can be used, based on the averaged unknowns, i.e.
9-8 FINE™
Theory Numerical Model
* Ui + Ui + 1
( Fn ) i+1/2 = Fn ⎛⎝ ------------------------⎞⎠ – d i + 1 ⁄ 2 . (9-4)
2
Theoretically, the formulation Eq. 9-3 is to be preferred, but, in combination with the central
scheme, the formulation of Eq. 9-4 (default) is sometimes more robust, especially for high speed
flows.
Both central schemes and upwind schemes are available within EURANUS.
Currently, the arithmetic averaging given by Eq. 9-4 is the only available option.
(2 ) ( 4) 3
di + 1 ⁄ 2 = ε i + 1 ⁄ 2 δU i + 1 ⁄ 2 + εi δ Ui + 1 . (9-5)
(2 ) 1 (2) *
ε i+1⁄2 = --- κ λ max ( ν i – 1, ν i, ν i + 1, ν i + 2 )
2
. (9-6)
(4) 1 (4 ) * (2 )
ε i+1⁄2 = max ⎛ 0, --- κ λ -ε i + 1 ⁄ 2⎞
⎝ 2 ⎠
(2) (4 )
The coefficients κ , κ are user input, respectively VIS2,VIS4 in the non-interfaced parame-
ters.
(4)
The cell entered values of ε in Eq. 9-5 are obtained by arithmetic averaging of the cell face val-
ues of Eq. 9-6. The variables v i are sensors to activate the second-difference dissipation in regions
of strong gradients, such as shocks, and to deactivate it elsewhere. They are based on pressure and
temperature variations (depending on the value for the expert parameter IARTVZ) and are defined
as:
⎧ p ⎫
i + 1 – 2p i + p i – 1 , T i + 1 – 2T i + T i – 1
v i = max ⎨ ----------------------------------------- (9-7)
- ------------------------------------------- ⎬
⎩ i + 1 + 2p i + p i – 1 T i + 1 + 2T i + T i – 1
p ⎭
for all equations except the k and ε equations. For liquids, only a pressure switch is used. For the k
and ε equations, v i is based on the pressure, on the turbulent kinetic energy and on the dissipation
rate:
i + 1 – 2p i + p i – 1 , k i + 1 – 2k i + k i – 1 , ε i + 1 – 2ε i + ε i – 1
⎧ p ⎫
v i = max ⎨ -----------------------------------------
- ----------------------------------------- ----------------------------------------- ⎬ . (9-8)
⎩ p i + 1 + 2p i + p i – 1 k i + 1 + 2k i + k i – 1 ε i + 1 + 2ε i + ε i – 1 ⎭
*
λ in Eq. 9-6 is a measure of the inviscid fluxes and is commonly chosen as the spectral radius
multiplied with the cell face area:
* *
λ = λ i+1⁄2 = ( v∆S + c∆S ) i + 1 ⁄ 2 . (9-9)
On high aspect ratio meshes, the resulting dissipation in the streamwise direction may become very
low. Increasing it may improve the robustness and the convergence rate. According to Martinelli
(1987), the spectral radii in the other directions are also accounted for. Instead of Eq. 9-9 one uses:
FINE™ 9-9
Numerical Model Theory
* * * 1–σ * σ * 1–σ * σ
λ = max { λ i + 1 ⁄ 2, (λ i + 1 ⁄ 2) (λ j + 1 ⁄ 2) ,(λ i + 1 ⁄ 2) (λ k + 1 ⁄ 2) } (9-10)
The spectral radii in the other directions are defined similar as in Eq. 9-9, e.g. for the j-direction:
*
λ j+1⁄2 = ( v i + 1 ⁄ 2 ∆S j + 1 ⁄ 2 + c i + 1 ⁄ 2 ∆S j + 1 ⁄ 2 ) (9-11)
The cell face normal ∆S j + 1 ⁄ 2 is obtained by averaging the normals on the 4 cell faces in the j-
direction, belonging to the 2 cells that share the i+1/2 cell face.
The parameter σ is user input (EXPMAR). The default value is 0.5, i.e. Eq. 9-9 is used. In the cur-
rent version of EURANUS, the Martinelli approach is not yet implemented for the turbulence equa-
tions.
Note that if the k – ε turbulence model is chosen, the resulting equations for k and ε are solved
(2) (4 )
with the central scheme. Separate coefficients κ , κ for the artificial dissipation can be cho-
sen, respectively VIS2KE and VIS4KE in the non-interfaced expert parameters. Both the 2nd and
4th order dissipation are damped by multiplying them with the following exponential function:
+ 2 ⎛ V 2⎞
y
1 – exp ⎛ – ⎛ -----⎞ ⎞ exp – -----
⎜ 3⎟
. (9-12)
⎝ ⎝ C⎠ ⎠
⎝ S ⎠
The constant C corresponds to the input variable CDIDTE. The above expression means that the
damping is applied in the region where y+<CDIDTE, the default is CDIDTE=100; V is the cell vol-
ume and S the cell face area.
1 l
d i + 1 ⁄ 2 = --- R i + 1 ⁄ 2 diag ( α i + 1 ⁄ 2 )δ i + 1 ⁄ 2 W , (9-13)
2
where δW represents the variation of the characteristic variables which is related to the conserva-
tive variables U or the primitive variables V through
–1 –1
δW = R δU = L δV . (9-14)
–1
The matrices R and R contain respectively the right and left eigenvectors of the Jacobian matrix.
l l
diag ( α i + 1 ⁄ 2 ) in Eq. 9-13 represents a diagonal matrix with α i + 1 ⁄ 2 the element in row and col-
umn l.
They are to be evaluated on the cell faces, and therefore require the knowledge of the primitive var-
iables on the cell face. These can be obtained either by arithmetic averaging of the primitive varia-
bles in the cell centres left and right of the cell face (input variable IROEAV=0) or by a Roe-type of
averaging (IROEAV=1). In the latter case the following formulas are used:
9-10 FINE™
Theory Numerical Model
ρi + 1 ⁄ 2 = ρi ρi + 1
vi ρ i + vi + 1 ρi + 1
v i + 1 ⁄ 2 = ---------------------------------------------
-
ρi + ρi + 1
(9-15)
Hi ρi + Hi + 1 ρi + 1
H i + 1 ⁄ 2 = ------------------------------------------------
-
ρi + ρi + 1
2
c i+1⁄2 = (γ – 1) H 2
i+1⁄2 – v i+1⁄2
Theoretically the Roe averages have to be used. However, in practice, arithmetic averages can be
used in most applications without any repercussion on accuracy or convergence. Arithmetic averag-
ing has the advantage that it is computationally less expensive.
Different upwind schemes result, depending on the expression for α. In EURANUS a TVD version
of the Flux Difference Splitting scheme (FDS TVD), Roe (1981), and Symmetric TVD (STVD),
Yee (1987), are available.
For both schemes α can be expressed as:
α i + 1 ⁄ 2 = λi + 1 ⁄ 2 [ 1 – Qi + 1 ⁄ 2 ] (9-16)
where λ i + 1 ⁄ 2 represents the eigenvalues of the Jacobian matrix and Q i + 1 ⁄ 2 is a limiter function,
ensuring monotonicity.
Both the FDS TVD and STVD scheme do not have a built-in entropy criterion, which would guar-
antee that the 2nd law of thermodynamics is satisfied. These schemes therefore may allow unphys-
ical solutions, such as expansion shocks.
In order to avoid unphysical solutions an entropy fix is used. The eigenvalue λ i + 1 ⁄ 2 in Eq. 9-16 is
)
replaced by λ i + 1 ⁄ 2 :
2 2
λ i+1⁄2 + δ
)
λ i + 1 ⁄ 2 = -----------------------------
- if λ i + 1 ⁄ 2 ≤ δ
2δ . (9-17)
= λ i + 1 ⁄ 2 if λ i + 1 ⁄ 2 > δ
Eq. 9-17 limits the minimum value of λ i + 1 ⁄ 2 to δ . In subsonic and transonic cases a small con-
stant value (<1.) for the entropy fix δ suffices (choose the input variable ENTRFX < 1.). For flow
simulations in high speed regime, it is advised to scale the entropy fix with the spectral radius, by
choosing the input variable ENTRFX > 1, e.g. ENTRFX=1.05 or 1.1. The scaling factor is then
ENTRFX-1., i.e.:
*
δ = ( ENTRFX – 1. ) λ i+1⁄2 . (9-18)
The latter choice also increases the robustness, at the expense of a slightly increased dissipation.
Two values for ENTRFX are to be specified. The first one applies to the linear field (the character-
istic variables that propagate with speed un ), the second value to the non-linear field (the charac-
teristic variables that propagate with speed un ± c ).
For first-order upwind schemes, the limiter vanishes and both FDS TVD and STVD reduce to the
same first-order accurate upwind scheme.
FINE™ 9-11
Numerical Model Theory
- δi – 1 ⁄ 2W + δi + 3 ⁄ 2W
-, r i + 1 ⁄ 2 = --------------------
r i + 1 ⁄ 2 = -------------------- - (9-19)
δi + 1 ⁄ 2W δi + 1 ⁄ 2W
For the FDS TVD scheme, Q is function of only one of the ratios, according to the sign of the
eigenvalue:
-
Q = Q ( ri + 1 ⁄ 2 ) λi + 1 ⁄ 2 > 0
(9-20)
+
Q = Q ( ri + 1 ⁄ 2 ) λi + 1 ⁄ 2 < 0
The following limiters are available:
Q ( r ) = minmod ( 1, r ) , (9-21)
2
r+r
Q ( r ) = -------------2- r>0
1+r (9-22)
= 0 r<0
r+ r
Q ( r ) = -------------- , (9-23)
1+r
which represent respectively the Min Mod limiter (LIMITE=1), the Van Albada limiter (LIM-
ITE=2), the Van Leer limiter (LIMITE=3) and the Super Bee limiter (LIMITE=4). The function
minmod chooses the value with the minimum module.
In the STVD scheme, the check on the eigenvalue sign is avoided by choosing Q as a function of
- +
the two ratios r and r
- +
Q = Q ( r i + 1 ⁄ 2, r i + 1 ⁄ 2 ) . (9-25)
Since the 'classical' limiters such as Van Leer, Van Albada, Min Mod, Super Bee are defined as a
function of one ratio, Eq. 9-25 would exclude these limiters in a STVD context.
A new family of non-separable limiters, containing also smooth limiters, is proposed within EURA-
NUS, Lacor et al. (1993).
*
The classical limiter functions are applied to a so-called effective ratio r , which is an average of
- +
r and r :
- +
Q ( r , r ) = Q ( r* ) . (9-26)
The same functions Q as given in Eq. 9-21 to Eq. 9-24 can be chosen, where the ratio r is replaced
by the effective ratio r* (the same limiters as for the FDS scheme are used).
The following possible choices for the effective ratio are proposed:
9-12 FINE™
Theory Numerical Model
+ - 2 - + 2 + -
+ - r (r ) + r (r ) 2r r -
min (r ,r ) ≤ --------------------------------------
2 2 +
-- ≤ r + r ≤ .5 ( r + + r - ) ≤ max (r +,r -) .
- ≤ -------------- (9-27)
+ -
(r ) + (r ) r +r
It can be shown that these new limiters guarantee monotonicity provided that:
r* ≥ 0 , (9-28)
1 1 –1
d i + 1 ⁄ 2 = --- R i + 1 ⁄ 2 diag ( λ i + 1 ⁄ 2 )L i + 1 ⁄ 2 [ 1 – Q i + 1 ⁄ 2 ]δ i + 1 ⁄ 2 V . (9-29)
2
The limiter function is the same as before and acts on ratios as in Eq. 9-19 or Eq. 9-27 where the
characteristic variable variations are replaced by primitive variable variations.
Eq. 9-29 can be simplified to
1
d i + 1 ⁄ 2 = --- An i + 1 ⁄ 2 ( 1 – Q i + 1 ⁄ 2 )δ i + 1 ⁄ 2 V . (9-30)
2
∂ ( Fn )
The Jacobian matrix in Eq. 9-30 is the Jacobian with respect to the primitive variables --------------- :
∂V
–1
An = RΛL . (9-31)
FINE™ 9-13
Numerical Model Theory
a certain direction, the grid can not be coarsened any more (because of an odd number of cells)
before reaching the desired number of grid levels, EURANUS will create the missing grids by
coarsening only in the remaining directions.
Coarse level : 2 2 2
Coarsest level : 3 3 3
MG Cycle
The multigrid method is based on the Full Approximation Storage (FAS) approach. V (input param-
eter MGRSTR=1), W (MGRSTR=2) and F-cycle (MGRSTR=3) can be chosen by the user, or if
preferred their sawtooth variant (MGRSTR =4,5,6 for respectively V,W,F sawtooth). In the saw-
tooth variant no smoothing (i.e. Runge-Kutta solver or implicit solver) will be applied in the coarse-
to-fine part of the cycle. As a result, the solution is only prolonged when passing from coarse to
fine.
Consider a set of meshes denoted with an index l = 1,...,L with L being the finest level. The Navier-
Stokes problem on the finest level can be written as:
L
∂U
---------- + N L ( U L ) = 0 , (9-32)
∂t
L
where N L ( U ) is the spatial discretization of the Navier-Stokes operator on the finest mesh L . The
problem is then approximated on coarser levels l as:
l
∂U
--------- + N l ( U l ) = F l , (9-33)
∂t
with F l the forcing function, defined recursively as:
l
) + ˆI l + 1 [ F l + 1 – N l + 1 ( U ) ] ,
l l+1 l+1
F l = N l ( Il + 1 U (9-34)
9-14 FINE™
Theory Numerical Model
l
where I l + 1 and ˆI l + 1 represent restriction operators of respectively the unknowns and the residu-
l
∑R ,
ˆI l R l + 1 = l+1
(9-35)
l+1
∑Ω U
l+1 l+1
l l+1
Il + 1 U = --------------------------------- , (9-36)
∑Ω
l+1
l+1
where R is defined as:
l+1 l+1
R = Fl + 1 – Nl + 1 ( U ) (9-37)
and Ω represents the cell volume. The summation in Eq. 9-35 and Eq. 9-36 is over the 8 fine cells
contained within a coarse cell.
If IRESTR=1 a more complicated quadratic restriction is used both for the residuals and the
unknowns.
After temporal discretization, Eq. 9-33 becomes:
l l(0)
S∆U + N l ( U ) = Fl . (9-38)
l (0 )
U is the current solution on mesh l, around which the equations have been linearized (in an
implicit method) and which has to be smoothed. One has:
l (0 ) l l+1
U = Il + 1 U (9-39)
l l (0 )
∆U is an update of U and is to be calculated.
S is the smoother. It is the operator that corresponds to the chosen time integration method.
In EURANUS, an explicit multi-stage Runge-Kutta time-marching scheme has been implemented,
see below.
l l
The linear problem Eq. 9-38 can be solved for ∆U . The updated solution U will be smoothed
(provided S is a good smoother) and can be restricted to the next coarser level, according to Eq. 9-
39 with l replaced by l-1.
Note that the number of smoothing sweeps (i.e. the number of times the Runge-Kutta operator
respectively the chosen relaxation operator is applied) can be chosen by the user for each grid level
(Numerical Model page in Expert Mode: Scheme definition and Number of sweep on successive
grid levels) as described in section 9-3.1.1.
Once the solution on the coarsest mesh is smoothed, the coarse-to-fine sweep of the multigrid cycle
is initiated. The current solutions on finer grids are updated with the solution on the next coarser
level:
l
The operator Il-1 is a prolongation operator, which may be of zero- (i.e. piecewise constant, input
parameter IPROLO=0) or first-order (IPROLO=1).
FINE™ 9-15
Numerical Model Theory
In the basic V,W,F cycle (MGRSTR = 1,2,3 respectively) the new solution on the finer mesh is
l ( 0) l
smoothed before proceeding to the next finer level, by solving Eq. 9-35, with U = U . The
number of sweeps of the smoothing operator is user input (Numerical Model page in Expert Mode:
Scheme definition and Number of sweep on successive grid levels).
In the sawtooth type cycles (MGRSTR=4,5,6) the system of equations is not solved in the coarse-
to-fine part of the multigrid cycle.
The user also has the possibility to smooth the corrections after prolongation (2nd term in the right-
hand-side of Eq. 9-40) before adding them to the current finer grid solution, by applying the resid-
ual smoothing operator to the corrections. This is governed by the input parameter SMCOR which
has the same definition as RSMPAR, see below, and allows to calculate the smoothing parameter,
according to the chosen type of residual smoothing (parameter IRSMCH), see below.
In EURANUS, the computing cost of a multigrid cycle is significantly reduced by using simplify-
ing assumptions on coarser (i. e. all levels but the finest) grids (input parameter MGSIMP). If
MGSIMP=1, a first-order accurate upwind scheme is used on coarser levels (if Upwind spatial dis-
cretization scheme is selected in the Numerical Model page in Expert Mode), else a more diffusive
central scheme is used on coarser levels (if Central spatial discretization scheme is selected in the
Numerical Model page in Expert Mode). If MGSIMP=2 in addition to the changes for MGSIMP=1
also viscous terms as well as source terms (of turbulence and chemistry) are omitted on coarser lev-
els.
Coarse level : 2 2 2
Coarsest level : 3 3 3
It should be mentioned that there are two different kinds of prolongation in Full Multigrid. One is
the same as that used in the standard multigrid cycle, the other is the prolongation of solutions,
which replaces corrections in a standard multigrid prolongation. The latter could be different from
the former. Normally a zero-order prolongation is often used for the solution interpolation in a full
multigrid cycle.
Based upon numerical experiments, it has been found out that it is very important to check the con-
vergence level of solutions on the coarser grid before switching to the next finer grid. The reason
9-16 FINE™
Theory Numerical Model
for this is that if solutions do not converge on the coarser grid, it may cause divergence on the next
finer grid.
In addition, an initial solution on coarse grids is less sensitive to multigrid convergence than on the
finest grid. A large cell size makes solutions converge easily and fast. Hence it takes much less
computational time and risk of blowing up to get reasonable solutions on the finest grid, when com-
pared to the start directly on the finest grid. A full multigrid method can therefore improve the
robustness and efficiency of numerical iterations methods.
dU
= F(U) (9-41)
dT
can be written:
1 n n
u = u + α 1 ∆tF ( u )
2 n 1
u = u + α 2 ∆tF ( u )
… (9-42)
q n q–1
u = u + ∆tF ( u )
n+1 q
u = u
The coefficients α i determine the stability area and the order of accuracy of the Runge-Kutta
scheme. They can be chosen in such a way that they suit the problem to be solved.
For inviscid calculations the local (i.e. for each cell) inviscid time step is calculated as:
⎛∆ t
-----⎞
CFL
= -------------------------------------------------------------------------------------------------- . (9-43)
⎝ Ω⎠
I wS i + wS j + wS k + c [ S i + S j + S k ]
CFL is the user specified CFL number (variable CFL on the Numerical Model page). The vectors S
are cell normals at the cell centre (obtained by averaging the normals on the cell faces) in respec-
tively i,j and k-direction. The module corresponds to the cell face area.
For viscous calculations one also calculates a local viscous time step in each cell.
If a viscous CFL number (expert parameter CFLVIS by default set at -1) is specified by the user in
the Control Variables page, i.e. the input variable CFLVIS > 0, the following formula is used for the
viscous time step:
⎛∆ t
-----⎞
CFLVISΩρ
= ------------------------------------------------------------------------------------------------------------------------
- , (9-44)
⎝ Ω⎠ 2 2 2
ν 8µ { S i + S j + S k + 2 [ S i S j + S i S k + S k S j ] }
with µ the local laminar viscosity (laminar simulation) or the sum of local laminar and turbulent
viscosity (turbulent simulation).
The actual local time step is then the minimum of both time steps:
FINE™ 9-17
Numerical Model Theory
⎛∆ t
-----⎞
⎧ ∆t ∆t ⎫
= min ⎨ ⎛ -----⎞ ,⎛ -----⎞ ⎬ . (9-45)
⎝ Ω⎠ ⎝ Ω⎠ ⎝ Ω⎠
⎩ I ν⎭
If the input variable CFLVIS < 0 (Default) the viscous CFL number, CFLVIS, is replaced by the
inviscid CFL number, CFL in Eq. 9-44. The actual local time step is then obtained by weighting the
inviscid and viscous time step according to a harmonic mean formula:
⎛∆ t ∆t
-----⎞ ⎛ -----⎞
∆ t ⎝ Ω⎠ ⎝ Ω⎠
ν
⎛ -----⎞ = -------------------------------
I
- (9-46)
⎝ Ω⎠
⎛∆ t⎞ ⎛ ∆t⎞
----- + -----
⎝ Ω⎠ ⎝ Ω⎠
I ν
If local time stepping is used i.e. each cell in the computation domain has its own time step given by
Eq. 9-43, for inviscid calculations and Eq. 9-45 or Eq. 9-46 (depending on CFLVIS) for viscous
calculations.
If global time stepping is used, i.e. in all cells the same time step is used, which is, for stability rea-
sons, the minimum of the local time steps. This option is only to be used for unsteady simulations
when the dual time stepping approach (default) is not selected. For steady simulations, it is recom-
mended always to choose local stepping, as it will increase the convergence rate.
In practice, 4 or 5 stage Runge-Kutta schemes are mostly used. In EURANUS the following coeffi-
cients are default for 4 and 5 stage schemes:
For central schemes
The user has the possibility to provide the Runge-Kutta coefficients through the expert parameter
IRKCO. The first element of IRKCO corresponds to the first coefficient, the 2nd element to the sec-
ond, and so forth. A maximum of 10 coefficients (i.e. a 10-stage scheme) is foreseen. If all elements
of IRKCO are zero the default values are used as listed in section 9-3.2.3.
In Jameson (1985) the dissipative terms are calculated only once or a few times in a q-stage Runge-
Kutta scheme. This approach reduces the computational cost substantially.
The disadvantage is that the dissipative terms need to be stored separately. For the laminar Navier-
Stokes equations with 5 equations, 5 extra arrays are required to store the dissipative residuals from
the artificial dissipation.
F(U)=Q(U)-D(U)
(9-50)
The artificial (central or upwind) dissipation terms, cf. Eq. 9-2, are accounted for either in the con-
vective part (parameter IBOTH = 0) or in the dissipative part (parameter IBOTH = 1).
9-18 FINE™
Theory Numerical Model
where m is the current stage and r is the latest stage at which the dissipative residuals have been
evaluated.
The recalculation of the dissipative residuals is governed by the (vector) variable ISWV. For a m-
stage Runge-Kutta scheme the m first elements of this vector are used, the first element correspond-
ing to the 1st stage, the 2nd element to the 2nd stage and so on. If the element of ISWV is 0 it means
that the dissipative residual should not be calculated at the corresponding stage, and the latest avail-
able dissipative residuals will be used; if it is 1, D will be calculated.
Note that for consistency reasons, the first value of ISWV should always be one.
For upwind schemes, in order to run at high CFL numbers, it is advised to recalculate the upwind
dissipation at every stage. The best strategy is to chose IBOTH=0 (upwind dissipation put in con-
vective residuals, and hence calculated at each stage). ISWV monitors then only the recalculation of
the physical dissipation (i.e. Navier-Stokes terms) and a calculation on the first stage only is usually
enough (ISWV=1 0...0).
Central schemes are usually less sensitive to the recalculation of the artificial dissipation, so that
IBOTH=1 can be used. The recalculation of artificial dissipation is then also governed by ISWV.
However only one calculation (ISWV=1 0...0) may not be enough; a better choice is for instance
recalculation once every 2 stages, e.g. ISWV= 1 0 1 0 1 for 5-stage Runge-Kutta.
For IBOTH=2 (Default), the same strategy as for IBOTH=1 is used but in addition the physical vis-
cous terms are evaluated only once per iteration.
For non-equilibrium real gas calculations (ITYGAS=4 in corresponding ’.run’ file) the input varia-
ble ISWS governs the recalculation of the source terms of the species equations. Apart from the
source terms, ISWS also governs the recalculation of the diagonal terms of the Jacobians of the
source terms (which are used in the point implicit procedure).
From Eq. 9-42 it can be seen that, after the update in the last stage, the residual is normally not re-
calculated any more. Strictly spoken however, within a multigrid approach, one should use in the
calculation of the forcing function for the coarser level, Eq. 9-34, the residual corresponding to the
latest solution. For single grid computations, re-calculation is never done.
Within multigrid, the Runge-Kutta operator can be applied more than once on each grid level, to
ensure a better smoothing. The parameter "Number of sweeps on successive grid levels" give the
number of times the Runge-Kutta operator has to be applied to each grid level during the fine-to-
coarse - respectively the coarse-to-fine - part of the multigrid cycle. They can take different values
on each grid level.
Within the explicit Runge-Kutta approach, the source terms of the two-equation turbulence models
and of the chemistry equations (non-equilibrium real gas) are treated in a point implicit way.
For the k-ε equations the Jacobian of the negative part of the source term is diagonalized by assum-
ing that the turbulent viscosity is approximately constant. This leads to the following relations
between ρk and ρε :
µ t ρε
ρk = ----------
-
Cµfµ
. (9-52)
2
( ρk )
ρε = C µ f µ -------------
µt
FINE™ 9-19
Numerical Model Theory
For instance, for the Chien model, this results in the following diagonal terms of the Jacobian,
including the compressibility corrections of Sarkar and of Nichols, cf. Eq. 4-75, on page 4-37,
Eq. 4-77, on page 4-37:
-
∂S k Cµ 1 µ C p1 ( γ – 1 )Mk
= – 2 ------ ρk ⎛ 1 + --- αM t ⎞ – 2 -------2- – 2 --------------------------------
2
- Pk , (9-53)
∂ ( ρk ) µt ⎝ 2 ⎠
ρy ρka
2
-
∂S ε 3 C µ ρε µ – 0.5y+
= – --- C ε 2 f 2 ------------
- – 2 ----2 e . (9-54)
∂ ( ρε ) 2 µt y
Note that in the equations above the damping function f µ is not taken into account.
m+1 n m n m
u = u + α m ∆tF ( u ) = u + α m R ( u ) (9-55)
The residual R may then be smoothed first, before applying the update of Eq. 9-55. The smoothing
may be obtained by applying a central type operator leading to a smoothed residual R̃ :
2 2 2
( I – ε i ∆ i ) ( I – ε j ∆ j ) ( I – ε k ∆ k )R̃ = R (9-56)
2
with the operator ∆ i
∆ i R˜ = R˜ i – 1 – 2R˜ i + R˜ i + 1
2
(9-57)
* 2
1 σ
ε > --- ----- –1 , (9-58)
4 σ
*
where σ , σ are the CFL numbers of the smoothed and unsmoothed Runge-Kutta scheme respec-
tively. The result of the residual smoothing is thus that one can run at higher CFL numbers: choose
*
σ /σ > 1 and by an adapted choice of ε, according to Eq. 9-58, the scheme remains stable. Though
*
theoretically Eq. 9-58 guarantees stability for any value of σ /σ , in practice this is not the case. A
* *
good practical value is σ /σ =2. The parameter σ /σ is user input (RSMPAR).
Different types of residual smoothing, corresponding to different definitions of ε are available.
Each one corresponds to a different value of the input parameter IRSMCH.
For IRSMCH = 1 the definition of Eq. 9-58 is used, with the restriction that ε should be positive.
For higher values of IRSMCH, the definition of ε also takes the mesh aspect ratio into account.
For IRSMCH = 2 one uses, after Radespiel & Rossow:
9-20 FINE™
Theory Numerical Model
2
1 σ* 1 + max ( ( λ j * ⁄ λ i * ), ( λ k * ⁄ λ i * ) )
ε i = --- ------ --------------------------------------------------------------------------------------- –1 , (9-59)
4 σ 1 + max ( λj * ⁄ λ i *, λ k * ⁄ λ i * )
with the spectral radius (scaled with the cell face area), e.g.:
λ i * = [ ( un + c ) ⋅ S ] i . (9-60)
1 σ* 1 2
ε i = --- ------ -------------------------------------------------------------------------------------------- –1 (9-61)
4 σ 1 + 0.0625 ⋅ ( ( λ j * ⁄ λ i * ) + ( λ k * ⁄ λi * ) )
IRSMCH=4 is also after Swanson & Turkel but for viscous flows. First Eq. 9-61 is applied. Denot-
ing the obtained value as ε*, the final value is:
1 10 ⋅ λ d
ε i = max [ε*,--- ------------------------------------
-] . (9-62)
4 λ i * + λj * + λ k *
2
1 σ* 1 + ( λ j * ⁄ λ i * ) + ( λ k * ⁄ λ i * )
ε i = --- ------ -------------------------------------------------------------------------- –1 . (9-63)
4 σ 1 + λ j * ⁄ λi * + λ k * ⁄ λ i *
FINE™ 9-21
Numerical Model Theory
9-22 FINE™
CHAPTER 10:Initial Solution
10-1 Overview
FINE™ provides the possibility to start a computation from an initial solution. The following initial
solution types are available:
• constant values,
• from file,
• for turbomachinery,
• from throughflow.
Each of these types is described in the next sections.
FINE™ allows to define a different initial solution type for each block or group of blocks. Such a
block dependent initial solution provides more flexibility in the restart of previously computed
solutions. The next section describes how to set up a block dependent solution.
FINE™ 10-1
Initial Solution Block Dependent Initial Solution
10-2 FINE™
Initial Solution Defined by Constant Values Initial Solution
2. A turbine includes 10 stages. Even with the turbomachinery-oriented initial solution, the code
does not succeed to converge properly. The designer could split the turbine in several parts and
calculate the solution for each of these parts. Then a global project is created. In the Initial Solu-
tion page, a from file initial solution is assigned to each part of the project. The computation will
have a far better chance to succeed.
FINE™ 10-3
Initial Solution Initial Solution from File
To continue on the result of a previous computation the user must simply select the solution. Click-
ing on the Open browser button will open a File Chooser window to select the ’.run’ (or ’.cgns’)
file of a computation.
To restart from a solution the computation definition file with extension ’.run’ is not
sufficient. It is required that the solution file with extension ’.cgns’ is present, which con-
tains all the saved and requested quantities.
The toggle button Reset convergence history can be selected. When selecting this button the ’.res’
file, which contains all the residuals from the previous computation, will be erased. The new resid-
uals will be calculated with respect to the first iteration of the new computation. If the from file
solution is already very close to the exact solution, the normalized residuals will decrease slowly. It
is just because the starting state is already a good approximation of the solution.
When all the blocks are grouped in only one group, the user is able to unselect the button Reset con-
vergence history. In this case, the new residuals are appended to the ’.res’ file and are calculated
with respect to the first iteration of the old computation.
If only one group of blocks exists the button to Reset convergence history can be set
on or off. If several groups of blocks exist, the button is disabled and automatically the
convergence history is reset.
There is a limitation in the current version of the FINE™ interface: if the user wants to keep the
convergence history, the ’.run’ file MUST belong to the active computation. The following steps
1. Compute a solution.
10-4 FINE™
Initial Solution from File Initial Solution
The initialization of the flow computation from a solution file is currently limited to
CGNS files.
As the initial pressure and temperature fields are block dependent, discontinuities in
these quantities can be present at block connections and the flow solver could have diffi-
culties to converge in that case. Therefore boundary conditions should be carefully
checked in order to avoid any important discontinuity at block connections.
The flow solver can start from files resulting from calculations performed on different grid levels.
Only the grid levels that are used in the multigrid cycle are considered here. For instance, a compu-
tation of a three-stage turbine on 0/0/0 with 3 grid levels can start from three solutions for each
stage, each one on a different grid level (0/0/0, 1/1/1, and 2/2/2). Furthermore, the full multigrid
cycle is automatically removed by the flow solver when at least one file is provided on the finest
grid level.
FINE™ 10-5
Initial Solution Initial Solution for Turbomachinery
When using the turbomachinery-oriented initial solution, make sure that the INLET
boundary condition is set in cylindrical mode and not in Cartesian mode. This last mode
is not compatible with the turbomachinery initial solution in the present version.
The turbomachinery-oriented page displays two major panels: one for the inlet patches and one for
the Rotor/Stator interfaces (see Figure 10.5.0-1). On the left of each panel, a tree shows the names
of the patches. If no patch name has been given in IGG™, the default name will be a concatenation
of the name of the block, the index of the face and the index of the patch.
The Inlet Patches grouping displayed in this initial solution page is identical to the Inlet Patches
grouping created under the INLET thumbnail of the Boundary Conditions page. Similarly, the
Rotor/stator interface grouping displayed in the initial solution page is based on the ID given in the
Rotor-Stator thumbnail of the Configuration/Rotating Machinery page. There is no way to modify
the groups of patches in the Initial Solution page. If the user wants to edit the groups of patches, the
user must go back to the Boundary Conditions and the Rotating Machinery pages.
An input dialog box is provided per group of patches in order to select the type of static pressure
specification: constant distribution or radial equilibrium. One or two input boxes are provided per
group of patches in order to define the static pressure value to be used and, if radial equilibrium is
chosen, the radius on which this radial equilibrium will be based.
Provided that a k-ε turbulence model is selected, a third panel is displayed that requires the initial
values for the turbulent kinetic energy (k) and the turbulent dissipation rate (epsilon). For advice on
the initial values to choose for k and epsilon see section 4-3.5.4.
Make sure that there is an inlet patch among the blocks for which a Turbomachinery
initial solution has been chosen. If there is no inlet patch for the selected group of blocks,
a warning message is displayed in the FINE™ interface asking the user either to edit the
10-6 FINE™
Throughflow-oriented Initial Solution Initial Solution
Note that there is no direct link between the throughflow module integrated in the
FINE™ interface and the format currently available. Work is in progress to make the
bridge between both tools.
The solver will create a virtual mesh based on the hub and shroud definitions, forcing points on the
data points. Two additional stations will be created, at the inlet and outlet, respectively. The inlet
data are extrapolated from the first spanwise station while the outlet data come from the last station.
It is therefore recommended that the inlet and outlet boundaries of the mesh lie outside the meridi-
onal domain covered by the throughflow data.
FINE™ 10-7
Initial Solution Throughflow-oriented Initial Solution
There are 4 spanwise stations in this example (Z=-0.025, Z=0.0, Z=0.025 and Z=0.05), each with 3
points in spanwise direction (R direction). The point order in a streamwise and spanwise station
goes respectively from hub to shroud and from inlet to outlet.
Each data line is made of the two (R,Z) coordinates (in m), the static pressure (in Pa), the static tem-
perature (in K), the meridional velocity and the tangential velocity (in m/s).
The hub is described in increasing Z coordinate from -0.02 to 0.051. The first column contains the
R coordinate for each of the two points on the hub.
10-8 FINE™
Throughflow-oriented Initial Solution Initial Solution
BLADE
3D View
INLET OUTLET
HUB
FINE™ 10-9
Initial Solution Throughflow-oriented Initial Solution
10-10 FINE™
CHAPTER 11:Output
11-1 Overview
With the EURANUS flow solver in the FINE™ environment the user has the possibility to specify
the content of the output generated by the solver. The output does not only contain the flow varia-
bles for which the equations are solved (density, pressure, velocity components), but also other var-
iables such as the temperature, the total conditions, the Mach number, customized velocity
projections, etc...
Furthermore CFView™ allows to calculate derived quantities, which permits to limit the memory
size of the output and to avoid re-launching the flow solver in case one quantity of interest is not
available. With this functionality the user can find the compromise between an acceptable memory
size for the output and the ease of having all flow quantities of interest immediately available for
visualization.
In the next sections the way to control the output from the EURANUS flow solver is described in
more detail:
• field quantities: calculated at all the mesh nodes (see section 11-2.1),
• solid data: calculated along the solid wall boundaries (see section 11-2.1.5),
• customized averaged data:
— the evolution of some averaged quantities can be provided along the streamwise direction as
described in section 11-2.2,
— for turbomachinery applications, a pitchwise averaged output can be created for visualiza-
tion as described in section 11-2.3,
• global performance output (see section 11-2.5),
• additional output for the ANSYS code, a widely used finite element code for structural analy-
sis (see section 11-2.4).
The global performance output is written in a ASCII file with the extension ’.mf’. All other output
generated by the solver is written in one single ’.cgns’ file, compatible with the CFView™ visuali-
sation software. It is also possible to request for the output to be written in Plot3D format as
described in section 11-2.6.
FINE™ 11-1
Output Output in FINE™
11-2.1.1 Thermodynamics
Table 11-1 and Table 11-2 give the formulae used to calculate the thermodynamic quantities.
Total pressure 2 γ
V +k ----------
-
Ptot = P stat + ρ --------------- T tot γ – 1
2 Ptot = Pstat ⎛⎝ -----------⎞⎠
T stat
Total temperature 2 2
V +k V +k
T tot = T stat + --------------- T tot = T stat + ---------------
2C p 2C p
Density ρ ρ
Entropy ( P stat ⁄ P statinlet )
T stat
s = C p ln --------------------
- s = C v ln -----------------------------------------
-
T statinlet γ
( ρ ⁄ ρ inlet )
Dynamic viscosity µ µ
11-2 FINE™
Output in FINE™ Output
Furthermore five definitions of the pressure coefficient are available: CP1, CP*1, CP2, CP*2, CP3:
2 ( P – Ptinlet )
— CP1 = -------------------------------
2
-,
ρinlet U ref
Pt inlet – Pt
— CP∗ 1 = -------------------------
-,
Pt inlet
2( P – Pinlet )
— CP2 = -----------------------------2- ,
ρ inlet W inlet
2 ( Pt – Pt )
— CP∗ 2 = ---------------------------------
inlet
2
-,
ρ inlet W inlet
2( P – Pinlet )
— CP3 = -----------------------------
2
-.
ρU
where Xinlet corresponds to the mean value of the variable X on the inlet, and where Uref is defined
as:
• ω RLEN in cylindrical cases, with RLEN, the reference length given in the Flow Model page,
and ω, the rotational speed (defined block per block),
• UENTR (Cartesian case) to be specified in the expert parameters list of the Computational
Steering/Control Variables page.
When applying FINE™ to external flow situations (expert parameter IINT set to 0), only the quan-
tity CP2 is well-defined. In external flows this pressure coefficient is defined in terms of the Refer-
ence values and Reynolds Number Related Info specified in FINE™ on the Flow Model page:
2 ( P – P ref )
CP2 = --------------------------
2
-
ρ ref U ref .
The user must then ensure that the far field or free stream quantities set on the Boundary Conditions
page as well as the initial solution values (set on the Initial Solution page) are fully consistent with
the reference values specified on the Flow Model page.
FINE™ 11-3
Output Output in FINE™
Finally, when a condensable fluid is selected in the Fluid Model page, new quantities at the bottom
of the page will appear. Those variables are defined using the thermodynamic tables defining the
condensable gas (see section 3-2.3.4 for more details).
When the condensable fluid option is used, 6 flow variables are stored at all mesh cell centers
instead of 5 for a perfect gas or liquid. These variables are the following: density, velocity compo-
2
V
nents, static pressure and total energy ( ρE = ρ ⎛⎝ e + ------⎞⎠ ). The static and total enthalpy can easily be
2
derived using:
2
V ⎞
• absolute total enthalpy H a = ⎛⎝ h + -----
- .
⎠ 2
2
• relative total enthalpy H r = ⎛⎝ h + W
-------⎞ .
⎠ 2
The dryness fraction permits to evaluate the location of the local thermodynamic state with respect
to the saturation curve. For each value of the pressure an interpolation from the saturation table
(’PSA.atab’) permits to determine the saturated liquid (ρl) and gas (ρv) values of the density. The
dryness fraction is then computed by:
( ρ – ρl )ρv
• dryness fraction X = -------------------------
-.
( ρv – ρl )ρ
The generalised dryness fraction directly results from the above expression whereas the standard
dryness fraction is bounded in the [0,1] interval.
11-2.1.2 Velocities
11-4 FINE™
Output in FINE™ Output
·
Vis = V – ⎛⎝ V ⋅ Si⎞⎠ ⋅ Si .
Vθ = Vy cos θ – V x sin θ ,
2 2
Vm = Vr + Vz .
W= V – ω × r ,
Wθ = Wy cos θ – W x sin θ .
• Relative velocity projections: Wi, Wj, Wk, |Wi|, |Wj|, |Wk|, Wis, Wjs, Wks
·
Wi = ⎛⎝ W ⋅ Si⎞⎠ ⋅ Si with Si , surface vector and
·
Wis = W – ⎛ W ⋅ Si⎞ ⋅ Si .
⎝ ⎠
11-2.1.3 Vorticities
FINE™ 11-5
Output Output in FINE™
ζ= ∇ × V ,
where ∇ is the gradient operator. From the above expression, ζx, ζy, ζz, ζxyz, and ζ are available.
11-2.1.4 Residuals
The residuals are computed in FINE™/Turbo by a flux balance (the sum of the fluxes on all the faces
of each cell). The absolute value |RES| resulting from the flux balance is written as output for each
cell. The residuals for the following quantities can be selected:
• density ρ,
• energy E,
• velocity component Wx,
• velocity component Wy,
• velocity component Wz,
• turbulent kinetic energy k for two equations turbulence models,
• turbulent energy dissipation rate ε for two equations turbulence models.
The solid wall data refer to quantities that are estimated at the solid boundaries:
• static pressure at walls (see section 11-2.1.1).
• static temperature at walls (see section 11-2.1.1).
τ
• normalized tangential component C f = ----------------
2
of the viscous stress at walls, where τ is the vis-
ρV ⁄ 2
cous stress, V is the reference velocity and ρ is the reference density defined in Flow Model page.
DT
• heat transfer at walls in general q w = K ⋅ -------- and for k-ε extended wall function
Dn
11-6 FINE™
Output in FINE™ Output
11-2.1.6 Turbulence
ρy wall u w
y + = ---------------------- ,
µ
11-2.1.7 Throughflow
For the output specific for Throughflow computations see Chapter 6.
FINE™ 11-7
Output Output in FINE™
Since the integration is performed on mesh surfaces, some differences could appear.
For example, in a case with splitter, two curves will be generated and a sum on this curve
should be performed to see the evolution of the quantity in streamwise direction of one
channel.
Since the surface averaging is performed on grid surfaces this option is mainly suita-
ble for H-topology meshes. Especially computing the evolution of the mass flow in
streamwise direction with HOH mesh topology is not correct.
For Cartesian projects as defined in the Mesh/Properties... menu the Surface Averaged Variables
page is displayed as shown in Figure 11.2.2-8. In this case the streamwise direction needs to be
selected directly in this page by selecting to average on surfaces of constant index (in the example
of Figure 11.2.2-8 surfaces of constant k). The blocks to include in the averaging need to be
selected in the same way as for a cylindrical case as described before. Contrary to a cylindrical
project, in this case no relative quantities are available.
Like for a cylindrical project the surface averaged evolution of the quantities is shown
for each block separately.
11-8 FINE™
Output in FINE™ Output
Automatic setting of the meridional averaging patches only works if no manual action
is involved (like renaming of some patches).
FINE™ 11-9
Output Output in FINE™
11-2.4 ANSYS
The FINE™/Turbo environment provides the user with a specific ANSYS output. The objective is
to export results of the CFD computation on solid surfaces (nodes and elements) into ANSYS as a
boundary condition for the finite element model. This approach is called FEM loading method for
loading the CFD results on a Finite Element Model. The transferred results on the interface fluid-
structure can be the pressure field for a structural analysis and the temperature field or the heat
fluxes for a thermal analysis in the structure.
The FEM approach needs to have the finite element mesh of ANSYS. If this mesh
does not exist before the CFD computation, it is impossible to create directly a file that
can be imported in ANSYS to apply the suitable boundary condition. Consequently the
best way to manage the interface between FINE/TURBO and ANSYS can be a new post-
processing tool. The post-processing tool is needed with the specification of the ANSYS
mesh, the common surfaces where the data must be interpolated and the type of bound-
ary condition on each surface. The transfer will be carried out through the ANSYS file
saved with the command CDWRITE ("project_ANSYS.cdb") in ANSYS. This file will
be modified and read again in ANSYS to impose the CFD results as boundary conditions
(see Figure 11.2.4-10).
11-10 FINE™
Output in FINE™ Output
The coded value can be modified in order to avoid confusion with physical values.
Note that the surface must be relatively flat or smooth (without sharp angle) so that the
six surfaces define a cube. Moreover the surface closed on itself as a cylinder or a blade
must be divided into two parts.
Validation tests are performed with ANSYS release v8.0. Please refer to ANSYS user
manual for more details on the ’.cdb’ file format
FINE™ 11-11
Output Output in FINE™
(2)
(3)
(4)
(5)
Firstly, the user needs to specify under the thumbnail Global Parameters in Boundary Conditions
(Figure 11.2.4-11 - (1)) the coded constant value used in the file ’project_ANSYS.cdb’. The default
coded values are given in the table 11-3 for the different types of boundary condition. Furthermore,
the user has to specify the required units of the data for each type of boundary condition. The units
will be specified next to the coded value.
Then, the user needs to load the file ’project_ANSYS.cdb’ (Figure 11.2.4-11 - (2)) created in
ANSYS (refer to section 11-2.4.1 for more details) in Input File.
The data transfer between the CFD mesh and the structural mesh is possible if the common
geometries are matching. It is possible that this condition is not fulfilled and the causes can be
diverse.
• the units: a first difference can come from the units of the meshes. Consequently the units used
in the ANSYS mesh must be specified in ANSYS Mesh Units. The conversion Factor will be
deduced automatically (Figure 11.2.4-11 - (3)).
• the coordinate system: the second cause of mismatching is the use of different coordinate sys-
tems (Figure 11.2.4-11 - (3)). For example, the original geometry can be only rotated for the
CFD model so that the axial direction corresponds to the Z axis. The necessary change of coor-
dinate will be done through the specification of the position of the coordinate center of the
ANSYS mesh in the coordinate system of the CFD mesh and the specification of the location
of three common points in both meshes in Reference System.
Note that these points cannot be located on a line or a plane including the zero point
(not linearly dependent). The coordinates are specified in their respective units.
11-12 FINE™
Output in FINE™ Output
Finally in Output File, the user specifies the new output file ’project_CFD.cdb’ (Figure 11.2.4-11 -
(4)) containing the transferred CFD results on the solid boundaries coded in ANSYS (refer to sec-
tion 11-2.4.1 for more details).
In order to create the new output file, the user has to complete the page under the Surface Selection
thumbnail before clicking on the Create Output File button (Figure 11.2.4-11 - (5)).
(1) (3)
(4)
(2)
FINE™ 11-13
Output Output in FINE™
Two different interpolation types (Interp Type) are proposed: a conservative and a non-conserva-
tive. The suitable interpolation scheme is dependent on the type of boundary conditions. The spatial
field of pressure or temperature can be interpolated from cell faces to nodes whereas a conservative
transfer of the heat fluxes can be preferred.
An allowed maximal distance (Max Dist) can be used to control the interpolation process by ensur-
ing that it is performed in the nearest element of the selected solid CFD patches.
Then, the user specifies the solid CFD patches in the list of Solid Patches by <Ctrl> or <Shift> and
left-click. To add or remove solid patches linked to the highlighted ANSYS surface, the buttons
Add Selection and Remove Selected are respectively used. When the solid patches are well selected,
the window has to be closed (Close) and the patches will appear in the list Selected Solid Patches
and Groups (Figure 11.2.4-12 - (3)).
Note that the ANSYS surface must be completely overlapped by the selected IGG
solid patches in order to ensure a complete data transfer.
The area of the selected IGG solid patches can be larger than the area of the ANSYS
surface.
Since the IGG™ solid patches defining a periodic structure can be divided (Figure 11.2.4-14), a
periodic shift can be required to match the common ANSYS surface. This shift must be specified
for each solid patch, highlighted in Selected Solid Patches and Groups list, in Periodicity by acti-
vating Rotation and Translation (Figure 11.2.4-12-(4)). The rotation and translation shifts are
respectively specified degrees and IGG™ units mesh.
11-14 FINE™
Output in FINE™ Output
Moreover one ANSYS surface can be covered by a group of solid patches and by the same group
rotated (or translated). Consequently, a duplication of the periodic solid patches is possible by acti-
vating Duplicate patch(es) after applying a Rotation and Translation to the highlighted solid patch
(Figure 11.2.4-12-(4)).
Finally, the new output file can be created by clicking on the Create Output File button (Figure
11.2.4-11 - (5)).
ries. Those quantities are inferred from the global force F = Fpressure + Fviscous exerted by the flow
at a given point. This force can be estimated from the pressure and velocity fields at each point. The
calculated quantities for an internal configuration are:
• Axial thrust, i.e. the projection of the global force on the rotation axis: ∑ F ⋅ nz .
S
• The projection of the torque along a given direction z , i.e. the couple exerted by the global
⎛ ⎞
force ⎜
⎝
∑ r × F⎟⎠ ⋅ z along direction z . This direction is given through the expert parameter
S
IDCMP, the default value of which is (0,0,1).
For external flow configurations the expert parameter IINT must be set to 0 and the output file will
contain the quantities as described in section 11-3.2 instead of the axial thrust and the torque
FINE™ 11-15
Output Output in FINE™
2 γ–1 T2
-----------
p γ p 1 – ------- T
⎛ ------
2⎞
–1 ln ⎛ -----2-⎞ T t1 ln ⎛ ------2-⎞
⎝p ⎠ ⎝T ⎠
- η is
⎝p ⎠ R t1 = ----------------------------- cp t1
η is t1
= ----------------------------- η pol = ---- ⋅ ----------------- γ----------
– 1- η pol = ---- ⋅ -----------------
-
T cp T p2 ⎞ γ R p
------2- – 1 ln ⎛ -------⎞2
⎛
1 – ------ ln ⎛ 2⎞
-----
-
⎝T ⎠ ⎝p ⎠ ⎝p ⎠
T t1 t1 t1 t1
3 γ–1 T
-----------
p γ p 1 – ----2- T
⎛ ----2-⎞ –1 ln ⎛ ----2-⎞ T1 ln ⎛ ----2-⎞
⎝p ⎠ ⎝T ⎠
- η is =
⎝p ⎠ R 1 --------------------------- - cp 1
η is 1
= ---------------------------- η pol = ---- ⋅ ---------------- γ----------
– 1- η pol = ---- ⋅ ----------------
-
T cp T p γ R p
----2- – 1 ln ⎛ -----2⎞ 1 – ⎛ ----2-⎞
⎛
ln ----- 2⎞
⎝T ⎠ ⎝p ⎠ ⎝p ⎠
T1 1 1 1
TABLE 11-4 Isentropic and polytropic efficiency definitions for perfect gases
Global performance data, available in the ’.mf’ file, are calculated using the results at the inlet and
outlet boundaries of the computational domain. The ’.mf’ file also provides global averages at each
rotor/stator interface. If the project does not involve any of these boundaries or interfaces, the ’.mf’
file is empty.
For turbomachinery cases, the output includes an efficiency. Several definitions are available,
selected through the expert parameter EFFDEF (default: 1).
For projects with a rotor, the expressions proposed in table 11-4 are used (index 1 and 2 refer to the
inlet and exit sections respectively).
The same definitions can be applied to the real gas in the case IRGCON=0 but with γ instead of γ.
is h 02 – h 01
h 02 – h 01
η is = -------------------- η is = --------------------
is
h 02 – h 01 h 02 – h 01
TABLE 11-5 Isentropic efficiency definition for real gases with ct r (IRGCON=1)
is
The definitions in Table 11-5 correspond to real gases with IGRCON=1. The value of h 02 is
is 2is C 2is r
obtained from the temperature T02 by using the isentropic relation: ∫1 -----p- dT 0 =
T0 ∫1 ------ dP 0 .
P0
11-16 FINE™
Output in FINE™ Output
3 ∆p t m· ρCp∆T t
η = -------------------
- η = -------------------
-
ρCp∆T t ∆p t m·
p t2 – p 2
η = -----------------
- . (11-1)
p t1 – p 2
Finally, relative and absolute flow angles are derived from the averaged velocity components:
vθ wθ
α = arc tan ⎛ ------⎞ β = arc tan ⎛ ------⎞ cylindrical , (11-2)
⎝V ⎠ ⎝V ⎠
m m
v w
α = arc tan ⎛ ----y⎞ β = arc tan ⎛ ------y⎞ cartesian . (11-3)
⎝v ⎠ ⎝w ⎠
x x
When selecting the Unformatted file (binary) option use the corresponding option in
CFView™ to open the file: click on the File Format... button in the menu File/Open
Plot3D Project... and select Unformatted. Do not select Binary format. In CFView™ the
user needs to select also little endian or big endian formats. On PC platforms (Windows
and LINUX) select binary low endian and on all other platforms select binary big
endian.
FINE™ 11-17
Output Expert Parameters for Output Selection
MERMAR: Safety margin for azimuthal averaged CFView™ solution. The mesh is shrunk
with a factor of 1.- MERMAR.
1st entry = axial safety margin
2nd entry = radial safety margin
The purpose of these two margins is to avoid points near the boundaries for which the azimuthal
averaging can not be performed. In general the azimuthal averaging is performed by taking a point
on the defined patch and to average the values for points in azimuthal direction starting from this
point. If there is not a sufficient amount of points in azimuthal direction the average is not com-
puted for that point and therefore the value for this point is not represented in CFView™. To avoid
such ’blank’ points near the edges of the patches it may be necessary to take a certain margin from
the edges of the patches.
The default values for the safety margins are appropriate for meshes on surfaces of revolution. If
the mesh is not a mesh on surfaces of revolution higher margins may be required. See section 11-
4.3 for more detailed information on the methods available for azimuthal averaging and the use of
these margins.
11-18 FINE™
Expert Parameters for Output Selection Output
• Lift coefficient, i.e. normalized projection of the global force in a chosen direction, generally
2
perpendicular to that used for the Drag coefficient: ∑ F ⋅ n2 ⁄ ( 0.5ρref Uref S ref ) .
S
• Momentum, i.e. normalized projection of the torque in a chosen direction (default 0, 0, 1):
⎛⎛ ⎞ ⎞ 2
⎜⎜
⎝⎝
∑ r × F⎟⎠ ⋅ n3⎟⎠ ⁄ ( 0.5ρref Uref Sref Cref ) .
S
In the above equations, S is the surface of the selected patch(es), ρ ref and U ref are user defined ref-
erence density and velocity (see the Flow Model menu of the FINE™ interface). S ref is a surface of
reference for which the default value SREF is set to 1 m2 and C ref is a length of reference for which
the default value CREF is set to 1. n 1, n 2, n 3 are normalized vectors used for the Drag, Lift and
Momentum coefficient and defined through the expert parameters IDCDP, IDCLP and IDCMP.
Their default value is (0., 0., 1.). r is the distance from the reference point IXMP (default value: 0.,
0., 0.). n z is the normalized vector directed along the rotation axis. The parameters mentioned here
can be changed in the Computational steering/Control variables page in expert mode.
IWRIT: Add some additional output on all block faces such as mass flow through these faces:
= 0: deactivated,
= 1: mass flow through rotor/stator interfaces,
= 3: summarizes the mass flow through different boundary types (SOL, ROT, INL,...),
= 4: mass flow through all block faces.
SREF: Reference area to non-dimensionalize the lift, drag and momentum, to get respectively the
lift coefficient, drag coefficient, moment coefficient. The default SREF=1.
CREF: Reference chord to non-dimensionalize the calculated moment, to get the moment coeffi-
cient.
IDCDP: This parameter gives the direction (i.e. 3 values) on which the body force has to be pro-
jected to give the axial thrust or the drag component. E.g. 1,0,0 means that the drag is the x-
component of the body force.
IDCLP: For flows around bodies this parameter gives the direction (i.e. 3 values) on which the body
force has to be projected to give the lift component. E.g. 1,0,0 means that the lift is the x-
component of the body force. For internal (IINT=1) cylindrical cases this parameter is not
used.
FINE™ 11-19
Output Theory
IDCMP: This parameter gives the direction (i.e. 3 values) on which the moment vector on the body
has to be projected to give the moment. E.g. 1,0,0 means that the moment is the x-compo-
nent of the moment vector.
IXMP: Specifies the x,y,z coordinate of the point around which the moment has to be calculated.
11-4 Theory
11-4.1 Computed Variables
The flow solution is stored in the .cgns file and will be used for the 3D CFView™ output.
CFView™ requires a cell-vertex representation of the flow solution, i.e. the flow variables need to
be provided at the mesh nodes. Since EURANUS considers the flow variables at cell centres, an
interpolation process is performed by the flow solver in order to compute the flow variables at the
mesh nodes.
The required interpolation of the flow solution is done by arithmetic averaging. If u represents one
of the primitive variables ρ, w x, w y, w z, p and u the interpolated value,
8
1
u = ---
8 ∑u , i (11-4)
i=1
where the summation is over the eight surrounding original locations of the respective variable. The
basic flow variables are calculated at the mesh nodes using the above procedure. Derived quantities
( v r, p t, H, s, ... ) are computed from this consistent representation of solution and geometry. By
contrast, if a quantity involves the surface normals, these are first calculated on the cell faces of the
original mesh and then averaged to the required location. Other exceptions are quantities involving
gradients or which are provided by the code:
∇ × V = ∇ × W + 2ω system . (11-5)
If a two equation turbulence model is used, k and ε are additional variables treated in the same
way as the primitive variables. The residuals too, are provided by the code in the cell centres and
therefore treated in the same way as the primitive variables, with the difference that they are only
defined in the interior and thus copied to the dummy cells as described above.
The above described arithmetic averaging process used to construct the CFView™ solution is first
order accurate, which has an important consequence if the user wants to extract quantitative global
quantities:
11-20 FINE™
Theory Output
In case of high flow gradients the calculation within CFView™ of the average (sur-
face integral) of a given quantity throughout an I, J or K=constant surface of the mesh or
a cutting plane will be performed with an error with respect to the flow solution calcu-
lated by the flow solver.
FIGURE 11.4.2-16 Stencil for 1D averaged flow. The numbers are the relative weight of each
cell in the contribution of the central cell to the surface average.
The weight coefficients are computed on the cell face centres (index F ), since this is the natural
location of the required surface normals, yielding:
1
m· F = ---
4 ∑ q corner . (11-6)
four corners
The surface average is then calculated as
1
∑
q = --------------------------------------- wFqF , (11-7)
FINE™ 11-21
Output Theory
the volumetric mesh. The domains and indices defining these sub-surfaces are provided in the input
file. Choosing sub-surfaces of the actual 3D mesh ensures that the resolution of the meridional
mesh is commensurate with the accuracy of the numerical solution.
Starting from each point ( r, z ) of the meridional mesh, the solution is defined by the primitive var-
iables ρ, w r, w θ, w z, p averaged along a circular arc passing through that point,
1
∫
u = -------------- wu dθ (11-8)
∫
w dθ
2 2
Vm = v r + v z is the meridional velocity. Either mass or area averaging can be requested, but the
same type of averaging will be used for all the primitive variables.
The integrals in Eq. 11-8 are approximated by trapezoidal sums. Practically, the r, z -constant line
along which the integration proceeds may be composed of k fragments of n ( j ) data points,
j = 1, ...,k , so that Eq. 11-8 is evaluated as:
k n(j)
u = ∑ ∑a ij u ij , (11-10)
j = 1i = 1
with
n ( j)
Σ i = 1 b ij
a ij = b ij ------------------------------
k n(j)
- (11-11)
Σ j = 1 Σ i = 1 b ij
and
⎧ 2w i θ 2 – θ 1 i = 1
⎪
bi = ⎨ w i θi + 1 – θi – 1 i = 2, ..., n-1 (11-12)
⎪
⎩ 2w i θ n – θ n – 1 i = n
All requested variables are derived, if possible, from the azimuthal-averaged primitive variables.
Those variables that can not be computed from the azimuthal-averaged primitive variables are com-
puted first. The fact that they are computed by the same routines that are used for generating 3D
CFView™ output ensures consistency. The suction and pressure side of the blades can be superim-
posed on the azimuthal averaged view.
For meshes that do not lie on surfaces of revolution the accuracy of the averaged solution is
decreased close to curved boundaries (see Figure 11.4.3-17). Although all boundary grid points lie
on the same surface of revolution, the straight line segments forming the cell edges do not. The
r, z -constant ray associated with point 1 will therefore miss some cell faces or pick values from the
inside instead of boundary values. In the worst case it will not encounter any cell faces at all. Strong
curvature and cells with high aspect ratio increase the inaccuracy.
This problem has been addressed in two ways. First, meridional grid points encountering fewer than
two cell faces are set to zero (a message printed to the screen indicates for how many points this
11-22 FINE™
Theory Output
was necessary). Secondly, the possibility is provided to shrink the meridional mesh by a small
amount to avoid the outermost regions of the volumetric mesh as shown in Figure 11.4.3-18. Inde-
pendent shrink factors k can be specified in the streamwise and spanwise directions using the
expert parameter MERMAR. Figure 11.4.3-18 illustrates the global effect. Each mesh line is shrank
individually by uniformly contracting towards its centre. The grid points thereby slide along the
segments of the original line (dashed line in Figure 11.4.3-18).
2 2
1 1
a) b)
FIGURE 11.4.3-17Azimuthal-averaging along circular arcs at curved boundary (light lines =
meridional mesh supporting the azimuthal-averaged solution, heavy lines = an arbitrary
cell face of the computational mesh), situation (a) before shrinking and (b) after shrinking.
h
kh/2
FIGURE 11.4.3-18: Shrinking of the meridional mesh, schematic illustration of (a) the global
effect and (b) the modification applied to each mesh line (dashed line and circles = original
line, solid line and squares = shrank line, C = centre point of the original line)
FINE™ 11-23
Output Theory
∫ Q dS = ∑ QB SB ∫ qd S = ∑ qB SB (11-13)
surface1/2 cell faces surface1/ 2 cell faces
The index B indicates a boundary cell face centre value, obtained by averaging the coordinates
( r = x, y, z ) of the four corners:
4
1
r B = ---
4 ∑r i (11-14)
i=1
and the solution ( U = ρ, w x, w y, w z, p ) in the first inner (index 1) and first outer cell (index 0):
1
U B = --- [ U 0 + U 1 ] . (11-15)
2
All variables except flow angles are first computed locally from this consistent representation of
solution and geometry and then summed, multiplied by the appropriate weight factor
m· B = ( ρWS ) B or A B = S B . Averages are calculated for the following variables:
VS
w x, w y, w z, w r, w t, v x, v y, v z, v r, v t, W , V , -------, p, ρ, p t, rel, p t, abs, p t, rot, T t, rel, T t, abs, T t, rot .
S
Static pressure is area averaged, all other variables are mass averaged. Explicitly, q representing
any of the enumerated scalars (other than p ) and q its average:
1 1
q = ---·-
m ∑ m· B q B p = ---
A ∑ AB pB . (11-16)
cell faces cell faces
The mass flow m· is obtained from Eq. 11-13 with Q = ρW , the area A with q = 1 . The values
of p t, rel, p t, abs, T t, rel, T t, abs are also computed from p, ρ and the averaged velocity components
to be compared with the directly averaged values.
11-24 FINE™
CHAPTER 12:Blade to Blade Module
12-1 Overview
This chapter describes the quasi-three-dimensional Blade-to-Blade Module for turbomachinery cas-
cade analysis. The module is fully automatic and can be used (and acquired) independently from
the other NUMECA tools.
Furthermore, an additional module is associated with the present one, called FINE™/Design2D,
which permits to redesign the blades for an improved pressure distribution on the blade surfaces
and is presented in Chapter 13.
The module assumes that the flow in the turbomachinery cascade remains on an axisymmetric
streamsurface, whose shape and thickness can either be provided by the user, or automatically con-
structed by the module (using the given hub and shroud walls).
The geometrical input data required from the user can be:
1. The streamsurface and the blade section on this last surface or,
2. The whole three-dimensional blade shape and the hub and shroud walls.
The module comprises a fully automatic mesh generator (which is able to treat any type of meridi-
onal configuration) and a customised version of NUMECA’s turbulent Navier-Stokes flow solver.
In the next section the interface is described in detail including advice for use of the Blade-to-Blade
Module. The theoretical background for the mesh generator and the flow solver is described in sec-
tion 12-4. Finally, the Blade-to-Blade Module requires geometrical data files and generates output,
which are detailed in section 12-5.
FINE™ 12-1
Blade to Blade Module Blade-to-Blade in the FINE™ Interface
When an existing Blade-to-Blade project should be opened click on the button Open an Existing
Project in the Project Selection window and select the project in the File Chooser window. The
most recently used projects can also be selected from the list of recent projects. If the selected
project was saved in the Design 2D module, the FINE™ interface is automatically switched to this
module showing the interface as in Figure 12.2.1-1.
12-2 FINE™
Blade-to-Blade in the FINE™ Interface Blade to Blade Module
FINE™ 12-3
Blade to Blade Module Blade-to-Blade in the FINE™ Interface
- 3D type (from .geomTurbo file): The whole 3D blade geometry is given under the form of a
geometry turbo file containing the hub, shroud and blade definition (more details in
AutoGrid User Manual). The stream surface can either be specified by the user (the
lower and the upper axisymmetric surfaces have to be specified), or can be automati-
cally constructed by the solver (geometrical division). In the latter case, the user only
has to specify the spanwise position of the stream surface and its relative thickness.
a) Streamsurface Data
Figure 12.2.2-2 shows the Parameter area for entering Stream surface data. The following stream
surface data files have to be specified, depending on the type of input data:
• 2D or Q3D types: stream surface definition file name,
• 3D case: hub and shroud file names or .geomTurbo file name.
When the 3D type of input data is selected additional parameters have to be defined:
• - Geometrical division or streamtube provided by the user.
• - If geometrical division: spanwise position between 0 (hub) and 1 (shroud).
• - If streamtube provided: lower and upper surfaces file names (2 lines).
2D&Q3D:
3D:
In both cases the module automatically calculates the intersection of the 3D blade with the consid-
ered stream surface.
12-4 FINE™
Blade-to-Blade in the FINE™ Interface Blade to Blade Module
In a 3D case with geometrical division, the streamtube height defines the thickness of the stream-
tube (the recommended value is 1%).
In a 2D, Q3D or 3D case with streamtube provided, the blockage ratio permits to scale the stream-
tube thickness distribution.
For all types of input data the number of points along meridional stream surfaces has to be speci-
fied. This input data is the number of points that the module will generate along the streamsurface
in order to obtain an accurate calculation of the meridional coordinate (default: 300).
3D
FIGURE 12.2.2-3 Blade geometry information page for 3D cases (from hub to shroud)
FINE™ 12-5
Blade to Blade Module Blade-to-Blade in the FINE™ Interface
12-6 FINE™
Blade-to-Blade in the FINE™ Interface Blade to Blade Module
• Number of points in the pitchwise direction: The number of cells should better be a multiple of
8, so that four multigrid levels are available.
• Periodic boundaries type: the user defines if the periodic lines should be straight or curved. In
the default configuration the periodic lines are curved.
• Five thumbnails give access to parameters to influence the mesh quality and points distribution
as described in the next paragraphs.
a) Inlet
The inlet location of the mesh is imposed on the hub and the shroud. The position is specified in
meridional coordinates (Z, R) or in spanwise coordinate (S).
b) Outlet
The outlet location of the mesh is imposed on the hub and the shroud. The position is specified in
meridional coordinates (Z, R) or in spanwise coordinate (S).
c) Clustering
The clustering coefficient in streamwise direction can be imposed from 0.0 to 1.0. A value of 1.0
gives a uniform distribution, whereas a 0.0 value concentrates all the points at the edges.
The second input defines if an Euler (no clustering in pitchwise direction) or a Navier-Stokes mesh
(with clustering to capture the boundary layer) should be generated. For a Navier-Stokes mesh the
following parameters should be specified:
• Size of the first mesh cell in the pitchwise direction (in meters).
• Number of constantly spaced cells in the pitchwise direction.
FINE™ 12-7
Blade to Blade Module Blade-to-Blade in the FINE™ Interface
• Constant or decreasing clustering from the blade edges to the inflow/outflow boundaries.
d) Smoothing
An elliptic smoother can be used in order to improve the quality of the mesh. In case the smoothing
is selected, the user has to specify if the smoothing should be performed with or without the control
of the clustering along the walls, and the number of smoothing steps (sweeps) to be performed (see
Figure 12.2.2-6).
e) Local Pre-smoothing
A local pre-smoothing can be performed in the leading and trailing edge regions, which can be
appropriate in cases of thick rounded leading and/or trailing edges (Figure 12.2.2-7).
FIGURE 12.2.2-7 Local pre-smoothing page of the Mesh generation parameters page
12-8 FINE™
Blade-to-Blade in the FINE™ Interface Blade to Blade Module
• Mass flow (kg/s) + direction of absolute velocity (radians) + static temperature (K).
• Mass flow (kg/s) + direction of relative velocity (radians) + static temperature (K).
• Mass flow (kg/s) + absolute Vθ velocity component (m/s) + static temperature (K).
• Mass flow (kg/s) + relative Wθ velocity component (m/s) + static temperature (K).
Selecting one of these options activates a submenu in which the user can set the inlet values. An
example corresponding to a ’Direction of absolute velocity + absolute total conditions’ is shown in
figure 12.2.3-8.
Both zero and first order extrapolation techniques are available, and can be selected in the Expert
mode (the default technique is the zero-order one).
FINE™ 12-9
Blade to Blade Module Blade-to-Blade in the FINE™ Interface
It should be mentioned that the inverse design method is more stable when the mass flow is explic-
itly imposed as a boundary condition. Therefore it is recommended to impose the mass flow during
the analysis also. The recommended strategies are therefore the one imposing the mass flow at the
inlet and the pressure at the outlet, or the one imposing the total conditions at the inlet and the mass
flow at the outlet.
12-10 FINE™
Expert Parameters Blade to Blade Module
In the case the initial solution has been obtained with a frozen turbulent viscosity field
(using the ITFRZ expert parameter), the parameter IATFRZ should be set to 2 in order to
read the turbulent viscosity field from a file and to keep this field unchanged.
12-4 Theory
The main characteristics of the mesh generation tool and of the blade-to-blade flow solver are pro-
vided in this section.
FINE™ 12-11
Blade to Blade Module Theory
2 2
∂p V Vm
------ = ρ -----θ- cos δ – ρ ------
-, (12-1)
∂n r Rc
where n is the spanwise direction, Vm and Vθ are respectively the meridional and tangential compo-
nents of the absolute velocity, and δ and Rc are respectively the slope and the curvature radius of
the meridional trace of the streamsurface.
A procedure is applied to the numerical scheme in order to force the flow to be aligned to the
streamsurface. This procedure eliminates the momentum residual in the direction normal to the
streamsurface. In case of multigrid the same procedure is applied to the flow solution after restric-
tion and prolongation of the solution.
12-12 FINE™
File Formats used by Blade-to-Blade Module Blade to Blade Module
a) Streamsurface Data:
For the 2D and Q3D types, one ASCII file defining the surface and its thickness is required. For the
3D type (form hub to shroud), 2 files are required to define the hub and shroud walls separately and
for the 3D type (from .geomTurbo), 1 file is required to define the hub, shroud and the blade. In
case the user wants to specify the streamsurface, 2 additional files are required to define the lower
and upper axisymmetric surfaces of the streamtube.
The structure of the streamsurface data file is the following:
FINE™ 12-13
Blade to Blade Module File Formats used by Blade-to-Blade Module
12-14 FINE™
File Formats used by Blade-to-Blade Module Blade to Blade Module
θ
W
β>0
α<0
V RPM (<0)
z
FIGURE 12.5.1-12 Coordinate system and reference for flow angles and speed of rotation
In a 2D case, if the ZTR coordinate system is used, the radius does not have to be specified in the
blade geometry input file. The number of columns can be 2 or 3, the third column being not read by
the solver (the radius is automatically set to the value given in the streamsurface data input file).
The suction side is by convention located above the pressure side in the θ-direction
Figure 12.5.1-12. It does not necessarily have to be the physical suction side.
FINE™ 12-15
Blade to Blade Module File Formats used by Blade-to-Blade Module
a) ’project_computationname.cgns’
This file contains the results of the calculation, i.e. the five flow variables (density, x-y-z components
of the relative velocity and the static pressure) at all the mesh cell centers (including the dummy cells
generated by the boundary conditions). It also contains all output variables to be displayed by
CFView™.
b) ’project_computationname.mf’
This file is a summary file containing the averaged flow quantities at the inlet and the outlet bounda-
ries.
c) ’project_computationname.mfedge’
This file is a summary file containing the averaged flow quantities at the leading and trailing edge
planes.
γ–1
⎛ -----------
γ ⎞
2 p
Mis = ----------- ⎜ ⎛ ----0-⎞ – 1⎟ , (12-2)
γ – 1 p⎠ ⎜ ⎝ ⎟
⎝ ⎠
where p is the local static pressure, and p0 is the local relative total pressure calculated assuming a con-
stant rothalpy and a constant entropy in the field.
The pressure coefficient is defined by:
p – p exit
Cp = ------------------- , (12-3)
p exit
where pexit is the exit static pressure, and p is the local static pressure.
12-16 FINE™
File Formats used by Blade-to-Blade Module Blade to Blade Module
These files can be read and plotted by the NUMECA MonitorTurbo (Loading diagram menu). The
’.velini’ file contains the initial distributions, whereas the ’.vel’ contains the actual ones, and is iter-
atively updated during the resolution process.
h) ’project_computationname.split.loadini’ &
’project_computationname.split.load’ (if splitter blades)
These files contain the same results as the ’.load’ and ’.loadini’ files along the splitter blades. These
files can be read and plotted by the NUMECA MonitorTurbo provided with the FINE™ user inter-
face (Loading diagram menu).
b) ’project_computationname.res’
This file contains the evolution of the mean and maximum residuals and of the inlet and outlet mass
flows. This file can be read and plotted by the NUMECA MonitorTurbo (Convergence history
menu).
FINE™ 12-17
Blade to Blade Module File Formats used by Blade-to-Blade Module
12-18 FINE™
CHAPTER 13:Design 2D Module
13-1 Overview
This chapter describes the new quasi-three-dimensional FINE™/Design 2D Module for turboma-
chinery cascade inverse design. This module is associated with the FINE™/Design 2D Blade-to-
Blade module presented in Chapter 12. In addition to the mesh generator and the flow solver, the
FINE™/Design 2D module includes one of the latest inverse methods for the redesign of the blades
with improved performance. Such a method permits to redesign the blade shape for a target pres-
sure distribution along the blade surfaces. The use of the design method requires a license for blade-
to-blade simulation.
The FINE™/Design 2D module can be used (and acquired) independently from the other
NUMECA tools.
In the near future NUMECA intends to extend the FINE™/Design 2D environment to other types
of applications and to other design techniques. This first version is limited to the inverse design of
turbomachinery cascades, and has therefore been integrated with the Blade-to-Blade method in the
FINE™/Design 2D user environment.
Both analysis and design modules are based on the assumption that the flow in the turbomachinery
cascade remains on an axisymmetric streamsurface. The geometrical inputs required from the user
to perform an inverse design are not different from the ones required to perform an analysis (as
described in Chapter 12).
For more information about the Blade-to-Blade method, see the description provided in the
Chapter 12. In the next section the interface is described in detail including advice for use of the
inverse design method. The theoretical background for the inverse method is described in section
13-3. Finally, the Design 2D Module requires geometrical data files and generates output as
detailed in section 13-4.
FINE™ 13-1
Design 2D Module Inverse Design in the FINE™ Interface
When an existing Design 2D project should be opened click on the button Open an Existing Project
in the Project Selection window and select the project in the File Chooser window. The most
recently used projects can also be selected from the list of recent projects. If the selected project
was saved in the Design 2D module, the FINE™ interface is automatically switched to this module
showing the interface as in Figure 12.2.1-1.
13-2 FINE™
Inverse Design in the FINE™ Interface Design 2D Module
13-2.2.1 Recommendations
It is often useful to save the results obtained after the analysis computation. Therefore it is recom-
mended to perform the inverse design in a new computation. With the analysis computation
selected in the Computations area on the top left of the FINE™ interface click on the New button. A
new computation is created with the same parameters as the initial computation. The Rename but-
ton may be used to rename the new computation.
analysis
inverse design
The expert integer parameter IATFRZ is usually set to 2 during the inverse design procedure in
order to freeze the computation of the turbulent viscosity. The consequence of this is to drastically
reduce the computational time per inverse design iteration.
FINE™ 13-3
Design 2D Module Inverse Design in the FINE™ Interface
At any moment, a zoom of the curves can be obtained by defining a new rectangular
window, using the left button of the mouse. The previous unzoomed view can be
retrieved by pressing the right button of the mouse.
13-4 FINE™
Inverse Design in the FINE™ Interface Design 2D Module
Finally, the number of iterations performed for the inverse design calculation.
FINE™ 13-5
Design 2D Module Inverse Design in the FINE™ Interface
Exactly as in the case of the restart of a flow analysis the user has to specify the name
of the initial solution file. It is recommended to save the results of the flow analysis
under a different name, so that several inverse design calculations can be performed suc-
cessively, starting from the same initial solution.
An interrupted inverse design calculation can be "restarted". For this purpose the user
should set the expert parameter INVMOD to 1 instead of 0. This implies that in addition
to the correct flow solution file, the actual blade geometry will be read by the solver and
considered as the initial one.
13-6 FINE™
Theory Design 2D Module
13-3 Theory
The purpose of an inverse method is to redesign the blade shape in order to obtain a prescribed
pressure distribution along the blade surfaces. This provides a detailed control of the boundary
layer behaviour, by limiting the amount of diffusion and by eliminating the eventual spurious accel-
eration and deceleration detected along the initial blade profile. Such a method also offers an indi-
rect control of the secondary losses, as a control of the blade loading distribution can be obtained.
The inverse design method adopted in FINE™/Design 2D consists of modifying iteratively the
blade geometry, until a target pressure distribution is reached on the blade surfaces. The geometry
modification algorithm is based on the permeable wall concept. Each iteration of the process is
composed of 2 separate steps, respectively related to the geometry and to the flow field updates.
Both viscous and inviscid problems can be treated with this approach.
Three formulations of the inverse problem are possible:
FINE™ 13-7
Design 2D Module File Formats used for 2D Inverse Design
• the classical formulation, in which both suction and pressure sides are redesigned in order to
obtain a pressure distribution prescribed along the whole blade surface.
• the mixed formulation, in which the suction side is redesigned in order to obtain a pressure dis-
tribution specified along the suction side only. This formulation is for instance interesting for
rotor blades on which strict mechanical constraints are imposed. In those cases the classical
inverse formulation may be less efficient because the blade thickness is a result of the calcula-
tion, and it is often difficult and time consuming to find an appropriate target leading to a blade
with an acceptable thickness distribution.
• the loading formulation, in which the distribution of blade loading (pressure-to-suction side
pressure difference) from the leading to the trailing edge is prescribed, and in which the cam-
ber line of the blade is redesigned (the blade thickness distribution being kept unchanged).
This approach is interesting especially for radial machines. This is due to the fact that in radial
machines, the average velocity level in the blade channel is mainly controlled by the shape of
the hub and shroud endwalls, and therefore a modification of the blade shape mainly influ-
ences the loading distribution. The "loading" formulation is therefore more appropriate and
efficient for radial machines. However no general rule can be set, and improvements can be
reached in many cases using the second formulation.
Several geometrical constraints and options have been implemented, to increase the flexibility of
the method:
• Control of the leading and trailing edges: several points can be kept unchanged in the leading/
trailing edge regions.
• Constant blade chord.
• Specification of the fixed reference point (leading or trailing edge).
• Inverse problem formulated in terms of isentropic Mach number or of pressure coefficient.
• Constant meridional position of leading and trailing edges (recommended for radial
machines): with this option, the points are displaced along the circumferential direction, which
guarantees that their meridional position is unchanged.
An important advantage of the present method is the possibility to control the outlet flow angle or
swirl. A major drawback of most inverse methods is the lack of control of the turning angle or work
exchange in the cascade. The user does not have the guarantee that the specified target pressure dis-
tribution will lead to an unchanged turning angle. This implies that several iterations are sometimes
required before finding the target leading to an acceptable design.
FINE™/Design 2D offers a unique solution to that problem, allowing the control of the turning
angle or of the change of swirl from inlet to outlet. A degree of freedom is introduced in the target
pressure distribution, that is automatically modified in order to respect the outlet flow angle con-
straint. A part of the target pressure distribution is defined using a fourth order polynomial curve.
The parameter defines the vertical position of a point of this polynomial curve. A smooth transition
is ensured between the fixed part of the target and variable one. The parameter is iteratively and
automatically adjusted in order to respect the outlet flow angle or the outlet swirl RVθ.
13-8 FINE™
File Formats used for 2D Inverse Design Design 2D Module
FINE™ 13-9
Design 2D Module File Formats used for 2D Inverse Design
The coordinates are given in the same unit length chosen by the user for the input data.
γ----------
– 1-
⎛ γ ⎞
2 p
Mis = ----------- ⎜ ⎛ ----0-⎞ – 1⎟ (13-1)
γ – 1⎜⎝ p ⎠ ⎟
⎝ ⎠
where p is the local static pressure, and p0 is the local relative total pressure calculated assuming a
constant rothalpy and a constant entropy in the field.
The pressure coefficient is defined by:
p – p exit
Cp = ------------------
- (13-2)
p exit
where pexit is the exit static pressure and p is the local static pressure.
These files can be read and plotted by the NUMECA MonitorTurbo (Loading diagram
menu). In the case of an inverse design the visualization of the two files permits to com-
pare the initial and the new pressure distributions.
13-10 FINE™
File Formats used for 2D Inverse Design Design 2D Module
formulation of the inverse design method is selected. The ’.loadini’ file contains the initial distribution,
whereas the ’.load’ file contains the actual one.
These files can be read and plotted by the NUMECA MonitorTurbo (Loading diagram
menu). In the case of an inverse design the visualization of the two files permits to compare
the initial and the new distributions.
These files can be read and plotted by the NUMECA MonitorTurbo (Loading diagram
menu). The comparison of the .tarini and ’.velini’ (or ’.loadini’) permits to observe the
inverse problem to be solved, whereas the comparison of the ’.tar’ and ’.vel’ (or ’.load’)
files permits to verify the correct convergence of the inverse design.
FINE™ 13-11
Design 2D Module File Formats used for 2D Inverse Design
13-12 FINE™
CHAPTER 14:The Task Manager
14-1 Overview
Using the Solver menu in FINE™/Turbo computations can be started, suspended or killed. For
basic task management these menu items are sufficient.
The Task Manager provides more advanced features for the management of (multiple) tasks. It
allows to manage tasks on different machines on a network, to define parallel computations or to
delay tasks to a given date and time.
Before using the Task Manager for the first time it is important to read the next section first. This
section provides important information for getting started with the Task Manager. Read this section
carefully to fully benefit of the capabilities of the Task Manager.
From FINE™/Turbo the Task Manager can be accessed by the Modules/Task Manager menu
item. The interface as shown in Figure 14.3.1-1 will appear.
In section 14-3 the Task Manager interface is described in detail including a description of all capa-
bilities. The current limitations of the Task Manager are listed in section section 14-6.
To manage tasks through the use of scripts is also possible and has the benefit that it is not neces-
sary to stay logged in on the machine on which the tasks are launched. See section 14-5 for more
detail on the scripts to use to launch NUMECA software.
FINE 14-1
The Task Manager Getting Started
14-2 FINE
Getting Started The Task Manager
login1 and login2 are optional. They must be set if the login on remote host is different from
the login on the local host.
The .rhosts file needs to be ended by a blank line.
The .rhosts file permission must be set to ’rw’ (chmod 600 ~/.rhosts).
2. Test the rsh command on the desired machine: an external check that the .rhosts file is set cor-
rectly is to enter the following command line:
% rsh remote_host ls <Enter>
where remote_host is the name of the machine to connect to. If the login on the remote host is not
the same, ensure that the .rhosts file contains the line:
remote_host local_login
and in this case the command line is:
% rsh remote_host -l local_login ls
If the .rhosts is set up correctly, a listing of the files is shown on the remote host.
Due to a limitation of the PVM library, FINE™ can not be connected to a machine
where FINE™ has already been started. When PVM tries to establish the connection, it
detects that a server FINE™ is already running and/or pvm daemons are still running and
finally refuses the connection with a warning. To solve this problem, the PVM daemons
must be stopped on the machine on which FINE™ must be connected:
1. Log on the machine where the connection must be established.
2. Start FINE™ and open the Task Manager through Modules/Task Manager and use the button
Shutdown of the page Hosts definition. The button Shutdown can be used to switch off all the
PVM daemons of all the machines connected to FINE™. This action will stop all the daemons
on all the machines connected, kill all the tasks and finally exit FINE™.
3. On UNIX: remove all the /tmp/pvmd.<userID> and /tmp/pvml.<userID> files. Under Windows
these files should be removed from the directory defined by PVM_TMP (by default this direc-
tory is C:\tmp).
4. Repeat this operation for all the machines on which the problems appear.
All these operations apply only to the user’s daemons and user’s files. Multiple users
can use the Task Manager simultaneously. No interaction appears between PVM dae-
mons and PVM log files of different users.
The Task Manager does not allow to launch processes from a UNIX platform to a
Windows platform.
FINE 14-3
The Task Manager Getting Started
From time to time, when a user tries to connect on a remote host, an error message can
appear very briefly: "Can not connect to RSH Port!!!" This may happen on Win-
dows connecting to a host, disconnecting, and trying to reconnect again to the same host.
It has the inconvenience to stop the interface FINE™. There is a socket release time
parameter which keeps the connection for some time.
14-4 FINE
The Task Manager Interface The Task Manager
When the user accepts (Accept), the host name appears in the host list, with type of operating sys-
tem and its connection status. At this point, the host is in the virtual machine, and is ready to receive
a task.
if the connection can not be established, check for more detail in section 14-2.1, sec-
FINE 14-5
The Task Manager The Task Manager Interface
14-3.1.3 Shutdown
When the user clicks on Shutdown button, all hosts in the virtual machine are removed. FINE™
exits and the virtual machine halts. All running tasks are killed. This option is useful when the user
wants to change the machine on which FINE™ is running. See section 14-2.1.1 for more detail on
virtual machines and when to change this.
When the user validates the data, a little watch in the task list indicates that the task is scheduled:
FIGURE 14.3.2-2Task List with Delay window to start a task at a later moment
When exiting FINE™, all the delayed tasks are automatically started and stay in sleep
mode until their starting dates are reached. When starting FINE™ again, all the task
14-6 FINE
The Task Manager Interface The Task Manager
manager related to the delayed tasks still in sleep mode are killed and FINE™ Task Man-
ger gets back the control of these tasks.
— The mesh generation system AutoGrid requires three files as arguments: the trb file, the
geomTurbo file and as output the grid file name (see the AutoGrid manual for more details)
FINE 14-7
The Task Manager The Task Manager Interface
— The flow visualization system CFView™ requires two files: the run file and a macro file
(see CFView™ manual for more details)
The run file is automatically opened by CFView™ before starting the macro. The macro file
may open other ’.run’ files to perform flow comparisons (see CFView™ manual for more
information about the macro ’.py’ file).
c) Remote Copy
The Remote copy to button allows the user to specify that the files are not visible from the remote
machine and must be transferred. This is only available for AutoGrid, Design3D and EURANUS.
d) Parallel Computations.
d.1) MPI system
The flow solver euranusTurbo has integrated multiprocessor machine concept. The button Parallel
Computation can be activated to specify that the parallel version of the flow solver can be used for
this computation.
The button Flow solver parallel settings gives access to the dialog box used to set up a parallel
computation.
When the dialog box is opened, the block list displayed all the block of the mesh. The user can
defined new processes on all the available machines and defined manually or automatically the load
balancing (using the Automatic Load Balancing button).
The parallel processing can only be used on multiprocessor machine (shared memory
mode) or on machines with the same operating systems (homogenous distributed mem-
ory mode)
See section 14-4 for more detailed information on parallel computations.
14-8 FINE
The Task Manager Interface The Task Manager
FINE 14-9
The Task Manager The Task Manager Interface
Process Management
Once the process is starting, the ".run" file is modified automatically depending of the number of
processor. Then a script file ".sge" is created at the same time as the ".p4pg" file when using MPI
system.
The name of the SGE job is limited to 8 characters and the following format is thus
chosen "comp#000" where 000 stands for a any number depending of the number of
SGE jobs already running under the same environment.
To test if a computation is running, because there are no communication with the task manager
since EURANUS has not been launched using PVM daemon, the "qstat" command with the job-ID
number of the SGE job is used.
To suspend the SGE job, the user could use the Suspend function of the Task Manager. It will create
automatically a ".stop" file in the running computation directory that will enable EURANUS to sus-
pend after saving the solution at the next iteration or after the full multigrid initialization.
To kill the SGE job, the user could use the Stop function. The Task Manager will then use the SGE’s
function "qdel" to kill the computation without saving the solution.
To kill EURANUS properly, the SGE’s "execd parameters" variable should be set to
"NOTIFY_KILL=TERM" and the "notify time" variable of nodes should be set to a suf-
ficient value to allow EURANUS to exit properly. We invite the user to ask the SGE
administrator to check these variables.
e) Message Window
When a task is started, a Task Manager process is launched to take care about the task management:
process communication, automatic launching of the sequential and parallel subtasks. A message
window linked to the Task Manager is opened. In this window the user can follow the evolution of
the status of all the subtasks of the current task. In case of flow solver computations the iterations
are also displayed in this window as shown in Figure 14.3.2-4.
At the bottom of the Task Manager window a Refresh Rate can be chosen. This rate determines how
frequently the Task Manager window is updated. A higher refresh rate will update the information
in the window more frequently but will use more of the available CPU.
14-10 FINE
Parallel Computations The Task Manager
If the user quits FINE™, all the message windows will stay open and the tasks will
continue. Closing a message window will immediately kill all the subtasks of the corre-
sponding task.
FINE 14-11
The Task Manager Parallel Computations
The present MPI implementation is based on a host/node approach. One host process is dedicated to
input/output management. The host reads the mesh and input files, sends the data to the computa-
tion processes (the nodes), receives the computation results from the nodes and writes the output
files. Figure 14.4.1-5 shows the task distribution on a 4 processors configuration. Note that the host/
node implementation will be replaced progressively as parallel MPI input/output functionalities are
now available.
14-12 FINE
Parallel Computations The Task Manager
that type. This is the case for block edges data that are exchanged several times during each itera-
tion. Significant latency times are induced by these communication processes and message group-
ing highly reduces the number of such messages.
Large message limitation
Very large messages are usually exchanged during initialization and at the end of the computation
between the nodes and the host. These also need to be avoided as they can cause important memory
needs and eventually lead to system failure. Messages are split into buffers of maximum size
MXBFSZ (see section 14-4.4) in order to avoid excessive MPI memory requests.
The total memory need Mtotal for a parallel computation involving n ( n ≥ 2 ) processors can be esti-
mated as:
n
M total = M real ⋅ 1, 2 ⋅ ------------ + M integer ⋅ n
n–1
where Mreal and Minteger are the real and integer memory needed for the same computation in
sequential mode.
From a practical point of view, the user does not specify the total memory needed but the integer
real M
and real memory required per processor. This can be estimated as ---------------
- ⋅ 1, 2 for the real mem-
n
ory and Minteger for the integer memory.
FINE 14-13
The Task Manager Task Management Using Scripts
The following table states the total memory need for a parallel computation using respectively 4
and 6 and 8 processors. The values of Mreal and Minteger are set to 10 Mb and 1 Mb respectively.
The equivalent memory request on a single processor is thus 11Mb.
14-4.5 Troubleshooting
If the computation is interrupted for an unknown reason, more detailed information is provided in
the ’.std’ and ’.log’ computation files. The file contains all relevant informations including MPI
error messages and should be sent to the NUMECA support team.
14-4.6 Limitations
• The distributed memory mode is restricted to homogeneous (i.e. with the same operating sys-
tem) clusters.
• The number of processors should not exceed the number of blocks.
Please notice that the capability to launch in batch is currently not supported for
EURANUS in Parallel Mode on PC.
14-14 FINE
Task Management Using Scripts The Task Manager
FINE 14-15
The Task Manager Task Management Using Scripts
#! /bin/csh
euranusTurbo /users/project/project_computation_1/project_computation_1.run -niversion 62_7
euranusTurbo /users/project/project_computation_2/project_computation_2.run -niversion 62_7
The script is started in a C-shell that is obtained by having as its first line: " #! /bin/csh". The com-
mand line, to start the flow solver EURANUS, contains its name, the full path name of the ’.run’ file
to launch and the FINE™ release that will be used:
"euranusTurbo COMPUTATION_PATH/computation.run -niversion <version>"
2. Set Permission for execution for the script file ’batch.sh’ by typing the command: "chmod 755
batch.sh".
3. Create an input file (’.run’) for each computation. To create those files through the FINE™ interface,
click on the File/Save Run Files menu in order to save the ’.run’ file of the activated computation,
after opening the corresponding project.
4. Launch the script ’batch.sh’ by typing the script file name: ’./batch.sh’.
When saving the ’.run’ file of the computation (File/Save Run Files) and launching the
computation (Solver/Start...) killed just afterwards (Solver/Kill...), a new script ’.bat’ on
WINDOWS (’.batch’ on UNIX) is created in the corresponding computation subfolder that
enables the user to launch the same computation in batch.
C:\NUMECA_SOFTWARE\fine62_7\bin\euranus.exe \\
C:\users\project\project_computation_1\project_computation_1.run -seq
C:\NUMECA_SOFTWARE\fine62_7\bin\euranus.exe \\
C:\users\project\project_computation_3\project_computation_3.run -seq
14-16 FINE
Task Management Using Scripts The Task Manager
C:\NUMECA_SOFTWARE\fine62_7\bin\euranus.exe \\
C:\users\project\project_computation_2\project_computation_2.run -seq
The command line, to start the flow solver EURANUS, contains its full path name, the full path
name of the ’.run’ file to launch and the sequential mode selection:
INSTALLATION_DIRECTORY\fine62_7\bin\euranus.exe \\
COMPUTATION_PATH\project_computation_1.run -seq
When saving the ’.run’ file of the computation (File/Save Run Files) and launching
the computation (Solver/Start...) killed just afterwards (Solver/Kill...), a new script
’.bat’ on WINDOWS (’.batch’ on UNIX) is created in the corresponding computation
subfolder that enables the user to launch the same computation in batch.
#! /bin/csh
euranusTurbo_parallel /users/project/computation/computation.run -niversion 62_7 -p4pg /
users/project/hosts.p4pg
The script is started in a C-shell that is obtained by having as its first line: " #! /bin/csh". The
command line, to start the flow solver EURANUS in parallel, contains the script name that ena-
bles parallel computation, the full path name of the ’.run’ file to launch, the FINE™ release and
the full path name of the file ’hosts.p4pg’ containing the definition of the machines that will be
used to launch the computation in parallel:
The name of the executable is different from the one for sequential computations.
2. Create Hosts Definition File ’hosts.p4pg’. A file needs to be created to define the hosts (e.g.
’hosts.p4pg’). This file specifies the machine information regarding the various processes. An
example of a ’hosts.p4pg’ file:
Example 1:
Hostname1 4 euranusTurbo62_7(master host on Hostname1 and 4 processes on Hostname1)
Hostname2 2 euranusTurbo62_7(2 processes on Hostname2)
FINE 14-17
The Task Manager Task Management Using Scripts
Example 2:
Hostname1 0 euranusTurbo62_7 (master host on Hostname1 and no other process on Hostname1)
Hostname2 2 euranusTurbo62_7 (2 processes on Hostname2)
For each machine, a line must be added consisting of the machine name (the machine hostname), the
number of processes to run on the machine, the name of the executable and the FINE™ release:
"hostname 4 euranusTurbo<version>"
If additional machines are to be used, subsequent lines are required for each one. If the number of proc-
essors is set to 0 for the first machine (first line of the ’hosts.p4pg’ file), only the master host process
will run on that machine while the nodes will run on the next machines declared in the following lines.
The first machine (hostname) specified in the file ’hosts.p4pg’ should be the one from which
the script will be launched.
The total number of processes defined in the file ’hosts.p4pg’ should be equal to the number
of nodes running.
3. Set permission for execution ’batch.sh’ by using the command: "chmod 755 batch.sh"
Create input files ’.run’ to launch the computation. The parallel settings of the computational file (’.run’)
are managed by the Task Manager. Therefore before launching the parallel computation script, the follow-
ing steps must be performed through the FINE™ interface:
4. Create computational file (’.run’) by activating the corresponding computation and clicking on the File/
Save Run Files menu in the interface.
5. To set the final parallel settings:
— create a task in the Task Manager,
— define the parallel settings of EURANUS subtask (parallel computation),
— launch the task through the Task Manager by clicking on Start,
— kill the task when EURANUS flow solver starts by clicking on Stop. This operation is mandatory in
order to have all the parallel settings correctly imposed in the computation file (’.run’).
6. Launch the script ’batch.sh’ by typing the script file name: "./batch.sh".
When saving the ’.run’ file of the computation (File/Save Run Files) and launching the com-
putation in parallel as explained in step 5 through the Task Manager (Modules/Task Manager)
and killed it just afterwards, a new script ’.batch’ and a file ’.p4pg’ are created in the correspond-
ing computation subfolder that enable the user to launch the same computation in batch and in
parallel.
D:\NUMECA_SOFTWARE\Fine62_7\bin\mpirun.exe /
"D:\test\test_computation_1\test_computation_1.p4pg" /
"D:\test\test_computation_1\test_computation_1.run"
The command line, to start the flow solver EURANUS in parallel, contains the script name that enables
parallel computation, the full path name of the ’.run’ file to launch and the full path name of the file
’.p4pg’ containing the definition of the machines that will be used to launch the computation in parallel:
14-18 FINE
Task Management Using Scripts The Task Manager
"hostname 4 NUMECA_RELEASE_PATH\bin\euranus.exe"
If additional machines are to be used, subsequent lines are required for each one. If the number of proc-
essors is set to 1 for the first machine (first line of the ’.p4pg’ file), only the master host process will run
on that machine while the nodes will run on the next machines declared in the following lines.
The first machine (hostname) specified in the file ’.p4pg’ should be the one from which the
script will be launched.
The total number of processes defined in the file ’.p4pg’ should be equal to the number of
nodes running.
3. Set permission for execution ’batch.bat’
Create input files ’.run’ to launch the computation. The parallel settings of the computational file (’.run’) are
managed by the Task Manager. Therefore before launching the parallel computation script, the following
steps must be performed through the FINE™ interface:
4. Create computational file (’.run’) by activating the corresponding computation and clicking on the File/
Save Run Files menu in the interface.
5. To set the final parallel settings:
— create a task in the Task Manager,
— define the parallel settings of EURANUS subtask (parallel computation),
— launch the task through the Task Manager by clicking on Start,
— kill the task when EURANUS flow solver starts by clicking on Stop. This operation is mandatory in
order to have all the parallel settings correctly imposed in the computation file (’.run’).
6. Launch the script ’batch.bat’ by typing the script file name: "./batch.bat" in a DOS-shell.
When saving the ’.run’ file of the computation (File/Save Run Files) and launching the com-
putation in parallel as explained in step 5 through the Task Manager (Modules/Task Manager)
and killed it just afterwards, a new script ’.bat’ and a file ’.p4pg’ are created in the corresponding
computation subfolder that enable the user to launch the same computation in batch and in paral-
lel.
FINE 14-19
The Task Manager Task Management Using Scripts
#! /bin/csh
#$ -S /bin/sh
#$ -o /home/sgeuser/parallel_sge/parallel_sge_computation_1/ parallel_sge_computation_1.std -j y
#$ -N comput1
#$ -pe numecampi 3
#$ -notify
P4_GLOBMEMSIZE=15000000
Export P4_GLOBMEMSIZE
NI_VERSIONS_DIR=/usr/numeca
MPIR_HOME=$NI_VERSIONS_DIR/_mpi
cd /home/sgeuser/parallel_sge/parallel_sge_computation_1
$MPIR_HOME/bin/mpirun -np $NSLOTS -machinefile $TMPDIR/machines \
$NI_VERSIONS_DIR/bin/euranusTurbo62_7 \
/home/sgeuser/parallel_sge/parallel_sge_computation_1/parallel_sge_computation_1.run
where:
#$ -S /bin/sh requires Bourne shell to be used by SGE for job submission and hence, only the .profile file
of the user is executed if exists on each computation host. There is nothing specific to Numeca software
that must be written . the .profile file.
#$ -o /home/sgeuser/parallel_sge/parallel_sge_computation_1/ parallel_sge_computation_1.std -j y
tells SGE system that the standard output has to be redirected in the ".std" in the computation directory,
the option "-j y" indicates that the standard error is redirected into the same file.
#$ -N comput1 is the name of the job given to SGE and that will be seen when monitoring the job with the
graphical monitoring utility qmon (Job Control button ) or with the qstat SGE command. SGE has a limi-
tation of 8 characters for the job name.
#$ -pe numecampi 3 requests 3 slots (processors) to the SGE system for executing the job using the
numecampi parallel environment described in chapter 2.1. This must correspond to NTASK +1 where
NTASK is the number of computation processes in the ".run" file, the "+1" being the host process that
manages inputs/outputs.
#$ -notify gives a delay between the send of the SIGKILL signal and SIGUSR2 signal. The delay is
defined in the SGE interface.
P4_GLOBMEMSIZE=15000000
Export P4_GLOBMEMSIZE
NI_VERSIONS_DIR=NUMECA_RELEASE_PATH
MPIR_HOME=$NI_VERSIONS_DIR/_mpi
are the environment variables indicating respectively the Numeca software installation directory and the
mpi directory that are used in the following command:
14-20 FINE
Task Management Using Scripts The Task Manager
$NSLOTS is the number of slots (processors) that have been allocated by SGE for the job, $TMP-
DIR is generated by SGE for providing the machines file used by the mpirun script.
$NI_VERSIONS_DIR/bin/euranusTurbo is a symbolic link to the numeca_start startup script for
all Numeca softwares and 62_7 is the version of the code that will be used and requires that it is
installed under /NUMECA_RELEASE_PATH/fine62_7.
/home/sgeuser/parallel_sge/parallel_sge_computation_1/parallel_sge_computation_1.run is the
argument to the euranusTurbo executable.
For sequential runs, the line "#$ -pe numecampi 3" must be removed and the final
command must be: $NI_VERSIONS_DIR/bin/euranusTurbo62_7 \ /home/sgeuser/
parallel_sge/parallel_sge_computation_1/parallel_sge_computation_1.run -seq
2. Set permission for execution ’batch.sge’ by using the command: "chmod 755 batch.sge"
Create input files ’.run’ to launch the computation. The parallel settings of the computational file
(’.run’) are managed by the Task Manager. Therefore before launching the parallel computation
script, the following steps must be performed through the FINE™ interface:
3. Create computational file (’.run’) by activating the corresponding computation and clicking on
the File/Save Run Files menu in the interface.
4. To set the final parallel settings:
— create a task in the Task Manager,
— define the parallel settings of EURANUS subtask (parallel computation),
— launch the task through the Task Manager by clicking on Start,
— kill the task when EURANUS flow solver starts by clicking on Stop. This operation is man-
datory in order to have all the parallel settings correctly imposed in the computation file
(’.run’).
5. Launch the script ’batch.sge’ by typing the script file name: "./batch.sge".
When saving the ’.run’ file of the computation (File/Save Run Files) and launching
the computation in parallel as explained in step 4 through the Task Manager (Modules/
Task Manager) and killed it just afterwards, a new script ’.sge’ is created in the corre-
sponding computation subfolder that enables the user to launch the same computation in
batch and in parallel.
FINE 14-21
The Task Manager Limitations
14-6 Limitations
The task manager have some limitations due to the current PVM and MPI libraries:
14-22 FINE
Limitations The Task Manager
FINE 14-23
The Task Manager Limitations
14-24 FINE
CHAPTER 15:Computation Steering
and Monitoring
15-1 Overview
This chapter describes the Computation Steering pages and the additional tool MonitorTurbo.
First section 15-2 describes the Computation Steering/Control Variables page. Furthermore this
chapter is completely dedicated to the monitoring tools available in FINE™ to monitor the global
solution during and after a computation:
• Computation Steering/Convergence History and Task Manager/Convergence History, see sec-
tion 15-3 and
• MonitorTurbo, see section 15-4.
In section 15-5 advice is provided on how to use the monitoring tools in analysing the progress of a
computation.
FINE™ 15-1
Computation Steering and Monitoring Convergence History
or a k-ε turbulence model is used. Also the amount of selected output may require more mem-
ory. To allocate more memory for the computation select the button Set the requested memory.
This will allow to define the amount of memory in Mb used for reals and integers.
For unsteady computations the Control Variables page is updated to give access to
additional parameters as described in section 4-2.1 and section 4-2.3.
Furthermore this page contains in expert mode two lists of expert parameters. Only the expert
parameters that are described in the manual are supported. Use of the other parameters is not rec-
ommended. For a summary of all supported expert parameters see Appendix C.
(1)
(5)
15-2 FINE™
Convergence History Computation Steering and Monitoring
When the flow solver EURANUS is invoked, a communication is automatically established with
FINE™ allowing the user to follow the selected quantities. Two representations of the quantities are
currently available: the convergence curves displayed in the area 5 and the quantity value at the last
iteration in the area 4. All these data are stored in files with extension .steering and .steering.binary.
These files are created and managed automatically by FINE™
If the communication between the flow solver and FINE™ is interrupted (i.e.: net-
work problem), parts of the convergence curves can be lost.
(a)
(b)
(c) (d)
2. All the steering files already loaded can be selected by <Ctrl>-leftclick and <Shift>-leftclick in
the list of available computations.
3. The file can also be loaded manually using the button Add Computation (c). A file chooser
prompts the user to select the file.
4. The button Remove Computation allows the user to remove from the memory the selected com-
putation (the file is not removed and remains available for selection).
5. Additionally, when a task is started, the steering files of the flow solver computation called in
the task are automatically loaded and become available through the list (b).
The Export Curves To File... button enables the user to save in an ASCII file ’.cvh’ all the curves
plotted in the Graphics View.
FINE™ 15-3
Computation Steering and Monitoring Convergence History
.
(a)
(b)
(c)
When a quantities has been added, it appears in the selection list and becomes the selected quantity.
If the flow solver is running, the value and the convergence curve of the new quantities is displayed
after a few seconds. The convergence curve begins at the iteration corresponding to the moment
when the new quantity has been added in the Selected Variables list.
When a quantity is added for selection it appears in the selection list with a number behind the
name. This number is only added to avoid to have two times the same variable name in the list. This
is especially necessary when, for example, the static pressure is monitored at two different points in
the domain.
Care should be taken when using the steering of a parameter on a local 3d grid point. It
should be carried out in the following way:
- select the parameter (relative velocity u, static pressure...).
- add for selection. It becomes available for the selected variables.
15-4 FINE™
Convergence History Computation Steering and Monitoring
- set the block id and the I,J,K indices of the grid point on which the evolution of the
selected parameter will be visualized.
All the user modifications (I,J,K indices...) can also be done (and modified) during the
computation run.
(d) (e)
Selecting a quantity will show its history in the graphics view as described in the next paragraph. To
select multiple quantities click on them (with the left mouse button) in the list while holding the
<Ctrl> or <Shift> key. To select quantities that are next to each other in the list simply click on the
first one and keep the mouse button pressed while moving to the last variable.
RES =
∑ fluxes . (15-1)
The root mean square of the residuals is computed with the following formula:
RES
RMS RES = log ⎛ RMS ⎛ -----------------------------⎞ ⎞ , (15-2)
⎝ ⎝ cellvolume⎠ ⎠
RES
MAXRES = log MAX ⎛ -----------------------------⎞ , (15-3)
⎝ cellvolume⎠
FINE™ 15-5
Computation Steering and Monitoring Convergence History
(a)
(b)
15-6 FINE™
MonitorTurbo Computation Steering and Monitoring
The convergence history of the selected curves are displayed inside the graphics view.
In the Computation Steering/Convergence History page, the color of the curves is different (up to 8
selected quantities). The automatic chosen curves colors are also used for the displayed value (a).
Additionally, the error between the two first selected quantities is also indicated (b). This is espe-
cially important for checking the difference between the mass flow at inlet and outlet
(Figure 15.3.6-7).
In the Task Manager/Convergence History page, the color of the curves is the same (but with differ-
ent markers) if related to the same computation and is different for each curves if related to different
computations (up to 10 selected quantities).
15-4 MonitorTurbo
15-4.1 Introduction
The MonitorTurbo can be launched independently from FINE™ in order to facilitate the batch
mode control. It is a separate graphic control window in which the user can visually monitor:
• the convergence history of one or several computations (Convergence history),
• the blade loading distribution (Loading diagram available with FINE™/Design 2D only),
• the blade profile (Blade profile available with FINE™/Design 2D only).
The two last displays are only available under the FINE™/Design 2D environments, as described in
Chapter 12 and Chapter 13. This section only describes the functionalities of the convergence his-
tory display.
FINE™ 15-7
Computation Steering and Monitoring MonitorTurbo
When multiple versions of FINE™ are installed the installation note should be con-
sulted for advice on how to start FINE™ in a multi-version environment.
On Windows click on the Monitor icon in Start/Programs/NUMECA software/fine#. Alterna-
tively FINE™ can be launched from a dos shell by typing:
<NUMECA_INSTALLATION_DIRECTORY>\fine#\bin\monitor <Enter>
FINE™ allows multi-process analyses. Therefore, the convergence histories of multiple (running)
projects can be visualized at the same time. It is also possible to compare the convergence history
associated with different computational parameters for the same project.
The monitored variables are displayed as a x-y graph in the upper part of the window (a). The x-
axis represents the number of multigrid cycles or work units (see section 15-4.3.3 and Section B-
4.3) achieved by the flow solver. The y-axis is a logarithmic axis representing the power of ten of
the residual values. The user can choose the residual of any of the equations solved and some global
parameters using the buttons visible in the lower right part of the graphic control window (b) as
shown in Figure 15.4.1-9.
The lower part of the graphic control window contains four boxes. The small box (c) located on the
left side contains a Print button (which allows to save the current graph as a postscript file) and a
Quit button (which allows to close the MonitorTurbo window). The three other boxes are described
in the next sections.
.
(a)
(c) (b)
15-8 FINE™
MonitorTurbo Computation Steering and Monitoring
Since the RMS and maximum residuals are normalized differently it may occur that
the RMS value shown in the MonitorTurbo is higher than the maximum residual.
FINE™ 15-9
Computation Steering and Monitoring Best Practice for Computation Monitoring
15-10 FINE™
Best Practice for Computation Monitoring Computation Steering and Monitoring
Representation menu the velocity will be displayed on the selected patches and this will be the
rotation speed everywhere.
15-5.3 MonitorTurbo
The MonitorTurbo follows the same quantities as the Convergence History, but its main interest is
to offer the possibility to follow the convergence block by block. Thus, one can localize problems
and, if necessary, modify the mesh in the region associated to the block, or change a boundary con-
dition.
Finally, coming back to the Task Manager, it is possible to get a deeper control, locally, by adding
control points (local 3d grid point). These points allow to track in important parts of the domain,
the speed, the pressure and turbulent quantities, which could be a requested information sometimes.
In the case of steady computations, the values associated to these points should converge to a con-
stant value.
One additional advice is to use CFView™ during the computation with intermediate and non-con-
verged solutions. When having convergence problems for example CFView™ may be used to look
for the zones were the residuals have too high values compared to other regions or an incorrect tur-
bulence field. To know in which block(s) the residuals start to increase first in case of divergence
problem, the MonitorTurbo may already give an indication. Combining the information from the
FINE™ 15-11
Computation Steering and Monitoring Best Practice for Computation Monitoring
MonitorTurbo and CFView™ allows to find the cause of convergence problems in the computa-
tions.
15-12 FINE™
APPENDIX A:Governing Equations
A-1 Overview
The flow solver integrated into the FINE™ user environment is named EURANUS ("EURopean
Aerodynamic NUmerical Simulator"). EURANUS is a multipurpose code for 2D and 3D flows in
complex geometries, using the latest numerical developments in CFD. A structured mesh is
required and complex geometries can be easily handled through a flexible multiblock meshing pro-
cedure.
This appendix describes the basic governing equations solved in EURANUS. For more detailed
information on the different aspects like turbulence, fluid modeling, multi grid strategy etc., see the
related chapters in this manual.
∂
U + ∇F I + ∇F V = Q , (A-1)
∂t
ρ
U = ρv , (A-2)
ρE
FINE™ A-1
Governing Equations Reynolds-Averaged Navier-Stokes Equations
ρv i 0
ρv 1 v i + pδ 1i τ i1
F li = ρv 2 v i + pδ 2i and – F vi = τ i2 , (A-3)
ρv 3 v i + pδ 3i τ i2
( ρE + p )v i q i + v j τ ij
where the stress and the heat flux components are given by:
∂w̃ ∂w̃
τ ij = ( µ + µ t ) --------i + --------j – 2
--- ( ∇w )δ ij (A-4)
∂x j ∂x i 3
∂ ˜
and q i = ( κ + κ t ) T. (A-5)
∂xi
0
Q = ρf e , (A-6)
Wf
with f e expressing the effects of external forces and W f the work performed by those external
forces: W f = ρf e ⋅ v . Other source terms are possible, like gravity, depending on the chosen func-
tionalities.
q = q + q' (A-7)
where where q is the time averaged value and q' the fluctuating part and
q' = 0 (A-8)
The energy, velocity components and temperature are density weighted averages defined as:
ρq
q̃ = ------ (A-9)
ρ
∂w̃ i ∂w̃ j 2 2
– ρw i ″w j ″ = µ t + – --- ( ∇w̃ )δ ij – --- ρkδ ij (A-10)
∂ xj ∂ x i 3 3
A-2 FINE™
Formulation in Rotating Frame for the Absolute Velocity Governing Equations
with w i the xi component of the relative velocity. In this equation k is the turbulent kinetic energy
and is defined as:
1
k = --- ρw i ″w i ″ ⁄ ρ . (A-11)
2
In contrast to the laminar case, both the static pressure and the total energy contain contributions
from the turbulent kinetic energy k and are defined as:
2
p* = p + --- ρk , (A-12)
3
1
E˜ = ẽ + --- w̃ i w̃ i + k (A-13)
2
Note that Eq. A-13 does not contain a term in the angular velocity. This term is accounted for in the
source term, and assuming stationary flow, corresponds to the last term of Eq. A-15.
ρ ρw̃ i 0
ρw̃ 1 p*δ 1i + ρw̃ i w̃ 1 τ i1
U = ρw̃ 2 F Ii = p*δ 2i + ρw̃ i w̃ 2 – F vi = τ i2 (A-14)
where the shorthand notation , i, in the Fvi expression, is used to denote derivatives with respect to
x i . The source term vector Q contains contributions of Coriolis and centrifugal forces and is given
by:
0
Q = ( – ρ ) [ 2ω × w + ( ω × ( ω × r ) ) ] (A-15)
2 2
ρw∇ ( 0.5ω r )
with ω the angular velocity of the relative frame of reference.
FINE™ A-3
Governing Equations Formulation in Rotating Frame for the Absolute Velocity
components. This formulation is different from the one generally used to solve internal turboma-
chinery problems, where the equations are solved for the relative velocity. The two formulations
should lead to the same flow solution. However, experience shows that the solution can be differ-
ent, especially in the far field region. For propeller problems, the formulation based on relative
velocities has the disadvantage that the far field relative velocity can reach high values. This
induces an excess of artificial dissipation leading to a non physical rotational flow in the far field
region, this dissipation being based on the computed variables.
This formulation is valid only if all boundary conditions are uniform in the azimuthal
direction. If the flow field boundary conditions are not uniform in the azimuthal direc-
tion, the boundary conditions expressed in this formulation must be unsteady. In this
case, the formulation for the relative velocity is suggested.
Except for the non-linear k-ε turbulence model, first-order closure, based on Boussinesq's assump-
tion, is used for the Reynolds stress:
∂ṽ i ∂ṽ j 2 2
– ρv i ″v j ″ = µ t + – --- ( ∇ṽ )δ ij – --- ρkδ ij (A-16)
∂ xi ∂ xi 3 3
with v i the xi component of the absolute velocity. The flux vectors are decomposed into Cartesian
components:
F I = f I1 1 x + f I2 1 y + f I3 1 z
(A-17)
F v = f v1 1 x + f v2 1 y + f v3 1 z
and
ρ ρw̃ i 0
ρṽ 1 p*δ 1i + ρw̃ i ṽ 1 τ i1
U = ρṽ 2 F Ii = p*δ 2i + ρw̃ i ṽ 2 – F vi = τ i2 (A-18)
where the shorthand notation , i, in the Fvi expression, is used to denote derivatives with respect to
x i . The velocity w i is the xi component of the relative velocity. This formulation envolves thus
both the absolute and the relative velocity components.
The source term vector Q is given by:
0
Q = –ρ( ω × v ) , (A-19)
with ω the angular velocity of the relative frame of reference. Other source terms are possible, like
gravity, depending on the chosen functionalities.
A-4 FINE™
Formulation in Rotating Frame for the Absolute Velocity Governing Equations
The averaged Navier-Stokes equations are obtained by the Favre averaging as described in section
A-2.2.
In contrast to the laminar case, both the static pressure and the total energy contain contributions
from the turbulent kinetic energy k and are defined as:
2
p* = p + --- ρk , (A-20)
3
1
Ẽ = ẽ + --- w̃ i w̃ i + k . (A-21)
2
The stress and the heat flux components are given by:
∂ṽ ∂ṽ
τ ij = ( µ + µ t ) -------i + -------j – 2
--- ( ∇v )δ ij (A-22)
∂x j ∂x i 3
∂ ˜
and q i = ( κ + κ t ) T, (A-23)
∂ xi
FINE™ A-5
Governing Equations Formulation in Rotating Frame for the Absolute Velocity
A-6 FINE™
APPENDIX B:File Formats
B-1 Overview
This Appendix describes the files used by FINE™ and the flow solver EURANUS. It is divided in
two parts. The first one gives the file format information needed to use FINE™, while the second
one describes the format of the files used and produced by EURANUS. As FINE™ is intended to
handle the file treatment for the user, knowing the exact format of all the files used in a simulation
process is not required. This Appendix is therefore written for advanced users.
In the following description, it is assumed that all the files are related to a generic project called
'project'. The chapter is divided in five sections:
• files produced by IGG™,
• files produced and used by FINE™,
• files produced and used by the flow solver EURANUS,
• files used as data profile,
• resource files used to control the layout and which contain default values and reference infor-
mation.
To simplify the notations, it is assumed that the related mesh has a topology of one block.
FINE™ B-1
File Formats Files Produced by FINE™
Manual modification of these files is not supported since it may corrupt the file or pro-
vide incorrect results. Such a modification should only be done on explicit advice of
NUMECA support team (support@numeca.be).
a) File Header
The file always has the following header containing the version number and the project type:
NUMECA_PROJECT_FILE VERSION 5.0
PROJECT_TYPE STRUCTURED
After the header, 3 new lines are used to store the name of the files linked to the project:
B-2 FINE™
Files Produced by the Flow Solver EURANUS File Formats
GRID_FILE /usr/_turnkey_tutorials/_rotor37/rotor37/_mesh/rotor37.igg
TRB_FILE /usr/_turnkey_tutorials/_rotor37/rotor37/rotor37.trb
GEOMETRY_FILE /usr/_turnkey_tutorials/_rotor37/rotor37.geomTurbo
These are respectively the mesh, the template and the geometry files. The template and the geome-
try files are not used by FINE™/Turbo and are only there for backward compatibility reasons.
The menu File/Save Run Files enables to save the ’.run’ file without starting the
solver EURANUS.
FINE™ B-3
File Formats Files Produced by the Flow Solver EURANUS
2. For block i= 1 to n:
The grid point coordinates and the 3D output flow variables as selected by the user through FINE™
(see Chapter 11). This data will be read by CFView™. Flow quantities are interpolated at mesh
nodes by the flow solver as required by CFView™ (see section 11-4 for mathematical details).
If the user selected solid data output, azimuthal averaged output or surface averaged output (see
Chapter 11), the corresponding mesh coordinates and computed quantities are added to the corre-
sponding ’.cgns’ file.
B-4 FINE™
Files Produced by the Flow Solver EURANUS File Formats
Mass flow in/out = total mass flow for both inlet and outlet boundaries. When no inlet or outlet is
present in the computation, the corresponding mass flow is set to zero.
Finally, the ".res" file is given with the CPU time included. In case of parallel computations the wall
clock time will be given instead of the CPU time.
FINE™ B-5
File Formats Files Produced by the Flow Solver EURANUS
The first 3 files can be written in ASCII or binary format, the binary format being the Fortran unfor-
matted format. The .name file is always written in ASCII.
B-6 FINE™
Files Used as Data Profile File Formats
Velocity-X
Velocity-Y
Velocity-Z
The imported values will be interpreted in the current project units. These values
would change (in the database) if the user changes the corresponding units. Therefore all
values will be converted for the flow solver EURANUS in SI system units (in radians for
the angles).
The values for 1D profiles should be given in increasing order of the x-coordinate. For example,
when entering a profile of temperature as a function of R, FINE™ checks whether the two first
points are in increasing order. If this is not the case, the profile will be inverted automatically in
order to ensure compatibility with the flow solver. The next time the Profile Manager ( ) is
opened, the profile will be shown in increasing order, contrary to what was initially entered.
For 2D profiles there is no such constraint except for profiles as a function of r-θ (see section 4-
2.3.2).
FINE™ B-7
File Formats Files Used as Data Profile
The theta angle is defined as θ = arctg (y/x). The θ profile should cover the patch
geometry, so it may take negative values. For unsteady calculations with profile rotation
the θ profile must be given form 0 to 2π (see section 4-2.3.2).
• The next lines contain the coordinate(s) of each point with the associated physical value.
Example 1:
r pressure
45
0.5 101000
0.6 102000
0.6 103000
0.7 102000
0.8 101000
This example is related to the interpolation of the pressure along the radius. There are 5 point coor-
dinates in the file. This is a space profile.
FINE™ uses the same file formats to import space, time, and space and time profiles. Thus if there
is only one profile in the file, it will be interpreted as a space profile, if a second profile exists,
FINE™ will recognize it as a time profile. Following is an example of a space and time profile:
Example 2:
R Pt
46
.3 95600.13
.251765 98913.46
.248412 100909.6
.24445 101527.6
.240182 101740.4
.23622 101811.4
100 10
.216713 101811.4
.210922 101872.1
.205435 101872.1
.199644 101872.1
.195682 101872.1
.191414 101872.1
.187452 101872.1
.18349 101740.4
.179222 99946.98
.1 99946.98
B-8 FINE™
Files Used as Data Profile File Formats
The two profiles are separated by an empty line. The interpolation type in time is specified as 100.
If this file is imported with the profile viewer invoked as "fct(space)"- only the first part will be
taken into account. If "fct(time)" is specified - only the second part will be read by the flow solver.
Both profiles will be imported if "fct(space-time)" is specified for the variable type.
If there are less coordinates supplied than the number of points specified, FINE™ will put zero for
the missing coordinates.
The format for the Cp and Gamma profiles is slightly different because the two profiles are speci-
fied in the same file (with extension .heat_capacity). It is as shown in the example below:
Example 2:
T Cp
11 4
295 1004.9
298 1005.3
301 1010.8
318 1011.0
11 3
296 1.31
298 1.34
303 1.36
FINE™ B-9
File Formats Resource Files
The first set of points is for the Cp profile, the second is for the Gamma profile as functions of the
temperature.
# -----------------------------------------------------------------------
# these are the names of the systems
# used to change all the quantities at the same time
DEFAULT_SYSTEM SI
DEFAULT_SYSTEM Default
DEFAULT_SYSTEM American 1
DEFAULT_SYSTEM American 2
# -----------------------------------------------------------------------
# -----------------------------------------------------------------------
B-10 FINE™
Resource Files File Formats
NEW_PROJECT_DEFAULT_SYSTEM Default
# -----------------------------------------------------------------------
NI_BEGIN UNITS_RECORD
NUMBER_OF_SYSTEMS 4
VALUE_UNITS_NAME length
UNITS_NAME 1 [m]
CONV_FACTOR 1 2 1.
CONV_FACTOR 1 3 3.280839895
CONV_FACTOR 1 4 39.37007874
UNITS_NAME 2 [m]
CONV_FACTOR 2 1 1.
CONV_FACTOR 2 3 3.280839895
CONV_FACTOR 2 4 39.37007874
UNITS_NAME 3 [ft]
CONV_FACTOR 3 1 0.3048
CONV_FACTOR 3 2 0.3048
CONV_FACTOR 3 4 12.
UNITS_NAME 4 [in]
CONV_FACTOR 4 1 0.0254
CONV_FACTOR 4 2 0.0254
CONV_FACTOR 4 3 0.08333333333
NI_END UNITS_RECORD
NI_BEGIN UNITS_RECORD
NUMBER_OF_SYSTEMS 4
VALUE_UNITS_NAME mass
UNITS_NAME 1 [kg]
CONV_FACTOR 1 2 1.
CONV_FACTOR 1 3 0.06852176586
CONV_FACTOR 1 4 0.005710147155
UNITS_NAME 2 [kg]
CONV_FACTOR 2 1 1.
CONV_FACTOR 2 3 0.06852176586
CONV_FACTOR 2 4 0.005710147155
UNITS_NAME 3 [lbf s2/ft]
FINE™ B-11
File Formats Resource Files
CONV_FACTOR 3 1 14.59390294
CONV_FACTOR 3 2 14.59390294
CONV_FACTOR 3 4 0.08333333333
UNITS_NAME 4 [lbf s2/in]
CONV_FACTOR 4 1 175.1268352
CONV_FACTOR 4 2 175.1268352
CONV_FACTOR 4 3 12.
NI_END UNITS_RECORD
......
If, for example, a new system is added on the fifth position, the user should supply the factors to
convert the physical quantities from all the existing systems (1, 2, 3, and 4) to the fifth, and back
from the fifth to all the others as shown below:
NI_BEGIN UNITS_RECORD
NUMBER_OF_SYSTEMS 5
VALUE_UNITS_NAME length
UNITS_NAME 1 [m]
CONV_FACTOR 1 2 1.
CONV_FACTOR 1 3 3.280839895
CONV_FACTOR 1 4 39.37007874
CONV_FACTOR 1 5 new_factor_value_from_1_to_5
UNITS_NAME 2 [m]
CONV_FACTOR 2 1 1.
CONV_FACTOR 2 3 3.280839895
CONV_FACTOR 2 4 39.37007874
CONV_FACTOR 2 5 new_factor_value_from_2_to_5
UNITS_NAME 3 [ft]
CONV_FACTOR 3 1 0.3048
CONV_FACTOR 3 2 0.3048
CONV_FACTOR 3 4 12.
CONV_FACTOR 3 5 new_factor_value_from_3_to_5
UNITS_NAME 4 [in]
CONV_FACTOR 4 1 0.0254
CONV_FACTOR 4 2 0.0254
CONV_FACTOR 4 3 0.08333333333
CONV_FACTOR 4 5 new_factor_value_from_4_to_5
UNITS_NAME 5 [new_units_name]
CONV_FACTOR 5 1 new_factor_value_from_5_to_1
CONV_FACTOR 5 2 new_factor_value_from_5_to_2
CONV_FACTOR 5 3 new_factor_value_from_5_to_3
CONV_FACTOR 5 4 new_factor_value_from_5_to_4
NI_END UNITS_RECORD
B-12 FINE™
Resource Files File Formats
The first (SI) and the second (default) systems are identical except for the rotational
speed units (RPM vs. radian).
FINE™ B-13
File Formats Resource Files
B-14 FINE™
APPENDIX C:List of Expert
Parameters
C-1 Overview
On the Control Variables page under Expert Mode a list of non-interfaced expert parameters is
available. As stated in the interface these parameters should not be used unless explicitly stated in
this manual. This chapter contains a list of all non-interfaced expert parameters that are described in
this manual. For each parameter a reference is given to the corresponding section. If more informa-
tion is desired on another parameters please contact NUMECA support at support@numeca.be.
FINE™ C-1
List of Expert Parameters List of Float Expert Parameters
C-2 FINE™
List of Float Expert Parameters List of Expert Parameters
FINE™ C-3
List of Expert Parameters List of Float Expert Parameters
C-4 FINE™
APPENDIX D: Characteristics of
Water (steam) Tables
D-1 Overview
As described in section section 3-2.3.4, the Condensable Fluid module aims at the modelling of the
real thermodynamic properties of a given fluid by means of interpolation of the variables from ded-
icated tables.
The approach that has been adopted in EURANUS consists of using a series of thermodynamic
tables, one table being required each time a thermodynamic variable must be deduced from two
other ones. This implies the creation of many tables as input, but presents the advantage that no iter-
ative inversion of the tables is done in the solver, with as a consequence a very small additional
CPU time.
Information related to
• the range of variation of admissible values for thermodynamic variables
• the discretization of the grid
• the nature of the interpolation algorithm selected
• the relative mean error (mean and maximum) on thermodynamic variables
• the size of the table
FINE™ D-1
Characteristics of Water (steam) Tables Main Characteristics
RHS: interpolates static pressure over density as a function of entropy and total enthalpy
HSP: interpolates total enthalpy as a function of entropy and static pressure
MER: interpolates dynamic viscosity as a function of internal energy and density
KER: interpolates thermal conductivity as a function of internal energy and density
PSA: defines the saturation line
Table 1 on page 2 lists the range of admissible variations for the thermodynamic variables, in the
different tables proposed. Ranges have been extended from the default tables proposed up to
FINE™ /Turbo v6.2-7, especially at low pressures (< 1000 Pa). Significant improvements, both in
terms of robustness and accuracy, are expected in that area.
D-2 FINE™
Main Characteristics Characteristics of Water (steam) Tables
Both RMS (root mean square) and maximum errors should be interpreted as relative errors. RMS
error does not exceed 2%, while in most tables it is ranging from 0.0001% to 0.1%. Maximum rela-
tive errors are located nearby the saturation line, where the interpolation naturally degrades despite
the use of bi-cubic double precision algorithms.
Tables RPT and EPT have been tuned so as to guarantee maximum accuracy in the vapor phase (<
0.1% at maximum). However, run performed in the liquid phase are consequently prohibited since
they would lead to large relative errors (> 100%) in that area. A dedicated set covering RPT and
EPT tables balancing maximum relative errors both on vapor and liquid phases can however be
obtained upon request. Please take direct contact at support@numeca.be for any question on this
purpose.
FINE™ D-3
Characteristics of Water (steam) Tables Main Characteristics
D-4 FINE™
Index
INDEX
A C
Abu-Grannam-Shaw Model 7-41 CFD
Add Fluid 3-3 Governing equations A-1
ANSYS Introduction 1-1
Outputs 11-10 CFL Number 7-12, 9-2, 9-17
AutoBlade 2-3, 2-10 CFView™ 1-3–1-4, 2-10
AutoGrid 1-2, 2-3, 2-5, 2-10 CGNS B-3
Axial Thrust 8-11, 11-15 Characteristic Values, see Reference Values
Azimuthal Averaged Output 11-9 Computation Definition Area 2-12
Condensable fluid
B Boundary conditions 8-20
Background Color 1-7 Outputs 11-4
Barotropic Liquid, see Fluid, Barotropic Liquid Condensable Fluid, see Fluid, Condensable
Benedict-Webb-Rubin 3-13 Conductivity 3-3–3-5
Blade To Blade 12-1–?? Laminar 3-10
Boundary Conditions 12-8 Turbulent 3-10
File Formats 12-13–12-17 Turbulent, Condensable Fluid 3-15
Flow Solver 12-12 Conjugate Heat Transfer 7-11–7-15
Geometrical Data 12-3 Blocks Types 7-12
Initial Solution 12-10, 13-5 Theory 7-13–7-15
Input Files 12-13–12-15 Control Variables 15-1
Inverse Design 12-8 Unsteady 4-4
Mesh Generator 12-6, 12-11 Convention 1-5
New Project 12-2 Convergence History 10-4
Numerical Model 12-10 Cooling
Open Project 12-2 Theory 7-32, 7-33
Cooling flow
Output Files 12-15–12-17
Bleed flow 7-15
Stream Surface Data 12-4
Data file 7-33
Theory 12-11
Flow Parameters 7-29
Bleed flow
Data file 7-33 Outputs 7-31
Flow Parameters 7-29 Positioning 7-18
Outputs 7-31 Visualization 7-30
Positioning 7-18 Wizard 7-16
Create
Visualization 7-30
Grid File 2-3
Wizard 7-16
Project 2-2, 2-4, 12-2, 13-2
Block
Conjugate Heat Transfer 7-12 Run Files 2-6
Throughflow 6-2–6-3 D
Boundary Conditions 8-1 Delete Fluid 3-7
Blade To Blade 12-8 Design 2D 2-10, 13-1–??
Lagrangian Module 7-4, 7-4–7-6 Formulation 13-4, 13-7
Rotor/Stator 5-17–5-19 Input Files 13-3, 13-8
Throughflow 6-7–6-8 Interface 13-2
Turbulence 4-16 Output Files 13-9
Unsteady 4-2 Start 13-6
Boussinesq 4-47 Theory 13-7–13-8
Design 3D 1-2, 2-3, 2-10
FINE™ i
Index
INDEX
Discretization Blade 6-14
Central Scheme 9-9–9-10 Friction 6-15
Limiters 9-12 Forced Transition 7-38, 7-40
Spatial 9-4, 9-7–9-13 Formula Editor 3-6
Temporal 9-4 Full Non-matching 5-20
Time 9-17–9-21 Fully Laminar 7-40
Upwind Scheme 9-10–9-13 Fully Turbulent 7-40
Drag 8-11
Drag Coefficient 11-19 G
Driver 1-6 Gauss Theorem 9-8
Dryness Fraction 3-15 Global Layout 2-7
Duplicate Active Project 2-6 Global Performance 11-15, B-4
Graphics 1-6
E Graphics Area 2-16
Edit Fluid 3-7 Gravity 3-5, 4-41
Efficiency 11-16 Grid
Enthalpy 3-11, 3-15 Create Grid File 2-3
Entropy 3-14 Grid Generation 1-2–1-4
Euler 4-15 Units 2-4, 2-8
2D 6-1 Group
Equations 6-13 Block Rotation 5-2
EURANUS 1-3 Boundary Conditions 8-2
EuranusTurbo Rotor/Stator 5-3
Parallel Computation 14-8, 14-17
H
Expert Mode 2-12
Hand Symbol 1-5
Expert Parameters C-1, D-1
Host Definition 14-5
External Boundary Condition 8-13
Hybrid Analysis 6-3
F
I
File Chooser 2-19
Icon Bar 2-10
File Format 6-9, B-1
IGG™ 1-3, 2-3, 2-5, 2-10
File Management 1-3
Incompressible
File Menu 2-4 Fluid Model, see Fluid Model, Incompressible
File Names, Limitations 2-7 Local Conservative Coupling 5-18
Fluid Initial Solution 10-1–10-8
Add 3-3
Blade To Blade 12-10, 13-5
Condensable 3-13–3-16
Block Dependent 10-1–10-3
Delete 3-7
Coarse Grid 9-3
Edit 3-7
Constant Values 10-3
Incompressible 5-18
File 10-4–10-5
Interface 3-2
Full Multigrid Strategy 9-16
Liquid 3-12
Throughflow 6-8–6-9, 10-7–10-8
List 3-2
Turbomachinery 10-6–10-7
Models 3-1–3-16
Turbulence 4-16
Perfect Gas 3-10
Unsteady 10-5
Real Gas 3-11 Initialization
Fluid-Particle Interaction. See Lagrangian Module. Full Multigrid Strategy 10-1
Force 8-11
Unsteady 4-4
ii FINE™
Index
INDEX
Installation 1-6 Information 2-15
Interface 1-7 Properties 2-8
General Description of FINE™ Interface 2-1–2-19 Selection 2-11
Toggles 2-14
L
View 2-8, 2-16–2-18
Lagrangian Module
View Area 2-13
Boundary Conditions 7-4, 7-4–7-6
Modules 2-10
Global Strategy 7-3
Moment 8-11
Interaction With Turbulence 7-10
Momentum 11-19
Outputs 7-6
Monitor
Particles Traces 7-7
Convergence History 15-11
Theory 7-9–7-11
Display 15-10
Laminar 4-15
Quantities 15-3
Layout 2-7
Residual File Box 15-9
License 1-8
Solution 15-1–15-12
Lift 8-11
Steering File 15-3
Lift Coefficient 11-19
Zoom 15-10
Light Bulb 1-5
MonitorTurbo 15-7
Limitations 2-7, 2-8
MSW Driver 1-6
Limiter
Multigrid 9-2
Lacor 9-12
Full 9-16
Minmod 9-12
Prolongation 9-16
Superbee 9-12
Strategy 9-3, 9-13–9-17
Van Albada 9-12
Van Leer 9-12 N
Linearization 4-27 New Project 2-4, 12-2, 13-2
LIPROD 4-27 Numerical Model 9-1
Liquid, see Fluid, Liquid
Blade To Blade 12-10
Local Conservative Coupling 5-17–??
Loss Coefficient 6-6 O
Low Speed Flow 4-42 Open
Project 2-3, 2-5
M OPENGL Driver 1-6
Mach Number
Outputs 3-15, 11-1–11-24
Absolute 11-5
3D Quantity 11-2–11-7, 11-20
Condensable Fluid 3-15
Azimuthal Averaged 11-9, 11-18, 11-21–11-23
Isentropic 13-4, 13-10
Lagrangian Module 7-6
Relative 11-5
mf File 11-15, 11-18
Mathematical Model 4-15
Particle Traces 7-7
Menu
File 2-4 Surface Averaged 11-8, 11-21
Menu Bar 2-4 Theory 11-20–11-24
Mesh 2-8 P
Modules 2-10 Pair of Scissors 1-5
Solver 2-9 Parallel Computation 14-13
Merge
Parameter Button 2-13
Merge Mesh Topology 2-11
Parameters Area 2-2
Patch For Azimuthal View 11-9
Particles 7-1
Mesh 1-2–1-4
Perfect Gas, see Fluid, Perfect Gas
FINE™ iii
Index
INDEX
Performance 11-15, B-4 Rotor/Stator
Periodic Boundaries 6-7, 8-9 Domain Scaling 5-11–5-12, 5-21–5-24
Phase Lagged 4-3, 4-13 FNMB Mixing Plane Coupling 5-19–5-21
Plot3D 11-17, B-6 Frozen Rotor 5-9–5-10, 5-24
Post Processing 1-3–1-4 Initial Solution 10-6
Prandtl Number 3-10 Interface 5-3–5-5
Preconditioning 4-42 Local Conservative Coupling 5-17–5-18
Parameters 4-45, 9-3 Mixing Plane Coupling 5-5–5-8, 5-16–5-21
Preferences 2-6 Phase Lagged 5-13
Profile Manager 2-19, 3-3–3-6 Pitchwise Rows Coupling 5-18–5-19
File Formats B-7 Steady 5-5–5-8, 5-9–5-10, 5-16–5-21, 5-24
Project Theory 5-15–5-24
Configuration 2-4 Unsteady 5-11–5-12, 5-13, 5-21–5-24
Creation 2-2 Rough Wall 4-35
Duplicate 2-6 Run File 1-3, 2-6
Management 1-3 Runge-Kutta
New 2-4 Scheme 9-17
Open 2-3, 2-5
S
Save 2-6 Save
Units 2-6 Intermediate Solution 2-9
Properties Project 2-6
Fluid 3-1
Scripts 14-14
Mesh 2-8
Second Order Restart 4-5
PVM
SI System 2-6
Daemons 14-1–14-5
Smooth Wall 4-35
Q Solid Data 11-6
Quit 2-7 Solver Menu 2-9
Specific Heat 3-10–3-12, 3-15
R
Standard Mode 2-12
Real Gas, see Fluid, Real Gas
Reference Start Flow Solver 2-9
Density 3-9, 4-47 Stator
Rotor/Stator Interface. See Rotor/Stator.
Length 4-47 Steady
Pressure 3-5, 4-45 Rotor/Stator 5-5–5-8, 5-9–5-10, 5-16–5-21, 5-24
Reference Values 4-47 Stop Flow Solver 2-9
Temperature 3-5, 3-9, 4-45 Surface Averaged Output 11-7
Velocity 4-45, 4-47 Suspend Flow Solver 2-9
Residual 11-6, 15-5, B-4 Sutherland 3-9
Implicit Smoothing 9-20 Sutherland Law 3-9
Radespiel & Rossow 9-20
Swanson & Turkel 9-21 T
Vasta 9-21 Task Definition 14-6
Restart 2-10 Taskmanager 14-1–14-23
Design 2D 13-6 Add Host 14-5
Unsteady 4-5 Delay 14-6
Reynolds Number 4-47 Limitations 14-22
Rotation Parallel Computation 14-8
Blocks 5-2–5-3 PVM 14-1–14-5
iv FINE™
Index
INDEX
Remove Host 14-6 U
Scripts 14-14–14-22 Units
Grid 2-4, 2-8
Shutdown 14-6
Project 2-6
Subtask 14-7–14-11
Unload Mesh 2-8
Task Definition 14-6–14-11
Unsteady 4-2
Tear-off Graphics 2-8
Temperature Boundary Conditions 4-2, 4-6, 4-12
Reference 3-9 Control Variables 4-4
Sutherland 3-9 Create 4-6
Thermal Connections 7-12 Initial Solution 10-5
Thermodynamic Tables 3-13 Initialization of 4-7
Throat Control 6-3 Phase Lagged 4-3
Throughflow Rotor/Stator 5-11–5-12, 5-21–5-24
Analysis Mode 6-3 Rotor/Stator Phase Lagged 5-13
Blade Geometry 6-3–6-5 Second Order in Time 4-5
Block Type 6-2 Turbomachinery 4-12
Boundary Conditions 6-7–6-8 User Mode 2-12
Expert Parameters 6-12
File Format 6-9–6-11 V
Flow Angle/Tangential Velocity 6-5–6-6 Vander Waals 3-13
Velocity
Global Parameters 6-2
Absolute 11-5
Initial Solution 6-8–6-9, 10-7–10-8
Friction 4-21
Loss Coefficient 6-6
Local Scaling 4-46
Mesh 6-6
Projections 11-5
Model 6-1–6-15
Reference 4-47
Theory 6-13–6-15
Relative 11-5
Time Configuration 4-2
Relative Projections 11-5
Time Step
Global Time 9-4 View Area 2-13
Local Time 9-4 View Manipulation 2-16
Physical 4-14 View On/Off 2-8, 2-13
Viscosity
Torque 8-11, 11-15
Inviscid Flux 9-8
Transition Model 7-37
Laminar 3-9
Abu-Grannam-Shaw Model 7-41
Turbulent 3-10, 4-39
Forced Transition 7-38, 7-40
Viscous Flux 9-8
Fully Laminar 7-40
Vorticity 11-5
Fully Turbulent 7-40
Turbulence W
Boundary Condition 4-16 Wall
Initial Solution 4-16 Rough 4-35
Linearization 4-27 Smooth 4-35
Output 11-7 Work Unit B-4
Turbulence Models
Baldwin-Lomax 4-28 X
k-epsilon 4-31 X11 Driver 1-6
Non Linear k-epsilon 4-33
Z
Spalart-Allmaras 4-29
Zoom In/Out/Region 2-18
TVD 9-11
FINE™ v