0% found this document useful (0 votes)

22 views148 pages

Elmer Tutorials

Uploaded by

vxtniilzvhnlfsocwn

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

22 views148 pages

Elmer Tutorials

Uploaded by

vxtniilzvhnlfsocwn

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 148

Elmer Tutorials

CSC – IT Center for Science

April 30, 2009

Copyrights

This document gives practical examples of the use of Elmer, a finite element software for multiphysical
problems.
The copyright of this document belongs to CSC – IT Center for Science, Finland, 1995–2009. Note that
even though the software is licensed under GPL this document is not.
Elmer program is free software; you can redistribute it and/or modify it under the terms of the GNU
General Public License as published by the Free Software Foundation; either version 2 of the License, or (at
your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
the GNU General Public License for more details.
CSC assumes no responsibility or liability on any errors or inaccuracies in Elmer program or docu-
mentation. Any kind of material damages or other indirect consequences resulting from any Elmer part,
documentation discrepancies and errors or non-anticipated program behavior are limited to the value of
appropriate products purchased from CSC.
This document is for informational use only. All information and specifications given have been care-
fully prepared by the best efforts of CSC, and are believed to be true and accurate as of time publishing.
CSC reserves the right to modify Elmer and its documents without notice.

Additional software product copyrights included in Elmer

Copyright, and License:
UMFPACK Version 4.3, Jan. 16, 2004. Copyright (c) 2004 by Timothy A. Davis, University of Florida,
[email protected]. All Rights Reserved.
UMFPACK License:
Your use or distribution of UMFPACK or any derivative code implies that you agree to this License.
THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY EXPRESSED OR
IMPLIED. ANY USE IS AT YOUR OWN RISK.
Permission is hereby granted to use or copy this program, provided that the Copyright, this License,
and the Availability of the original version is retained on all copies. User documentation of any code that
uses UMFPACK or any modified version of UMFPACK code must cite the Copyright, this License, the
Availability note, and "Used by permission." Permission to modify the code and to distribute modified code
is granted, provided the Copyright, this License, and the Availability note are retained, and a notice that
the code was modified is included. This software was developed with support from the National Science
Foundation, and is provided to you free of charge.

1
About Elmer Tutorials

The Elmer Tutorials is part of the documentation of Elmer finite element software. Elmer Tutorials gives
examples on the use of Elmer in different field of continuum physics. Also coupled problems are included.
The present manual corresponds to Elmer software version 6.0 for Windows NT and Unix platforms. Lat-
est documentations and program versions of Elmer are available (or links are provided) at http://www.csc.fi/elmer.

2
Contents

I ElmerGUI Problems 5
1 Heat equation – Temperature field of an L-shaped domain 6

2 Heat equation – Temperature field of a solid object 11

3 Linear elasticity equation – Loaded elastic beam 16

4 Navier-Stokes equation – Laminar incompressible flow passing a step 20

5 Navier-Stokes equation – Turbulent incompressible flow passing a step 25

6 Transient flow and heat equations – Rayleigh-Benard instability 30

7 Electrostatic equation – Computation of fringe capacitance 35

II non-GUI Problems 39
8 Radiation heat transfer 40

9 Eigenvalue analysis of an elastic beam 44

10 Elastic linear plate 49

11 Compressible flow passing a step 53

12 Flow through a hole – determining the acoustic impedance 58

13 Electrostatics 65

14 Lossless acoustic waves 69

15 Induction heating of a graphite crucible 71

16 Fluid flow around an elastic beam 75

17 Thermal actuator driven with electrostatic currents 81

18 Axisymmetric coating process 86

19 Blood ejection from a ventricle into aorta 91

III Utility Problems 96

20 Operator splitting in the evolutionary heat equation 97

3
21 Temperature distribution with BEM 103

22 Adding user defined equation solver 106

23 Volume flow boundary condition 110

24 Streamlines 114

25 Electroosmotic flow and advected species 118

26 Active and passive elements 124

IV Obsolite Problems 128

27 Temperature distribution 130

28 Loaded elastic beam 133

28.1 Solution with linear model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
28.2 Solution with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

29 Incompressible flow passing a step 138

29.1 Solution with linear triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
29.2 Solution with 2nd order rectangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

30 Transient flow and heat equations - the Rayleigh-Benard instability 142

Part I

ElmerGUI Problems

5
Tutorial 1

Heat equation – Temperature field of an

L-shaped domain

Directory: TemperatureAngleGUI
Solvers: HeatSolve
Tools: ElmerGUI
Dimensions: 2D, Steady-state

Problem description
An L-shaped structure (see figure 27.1) is heated by an internal heat source, which magnitude is 1 W/m3 .
The density of the structure is 1 kg/m3 and the heat conductivity is 1 W/mK. All the boundaries Γi are
kept on constant temperature of 0 K. The problem is to solve the temperature distribution in the structure.
Mathematically the problem to be solved is

−κ∆T = ρf inΩ
(1.1)
T = 0 onΓ

where κ is the heat conductivity, T is the temperature and f is the heat source. It is assumed that density and
heat conductivity are constants.

Figure 1.1: L-shaped domain

6
1. Heat equation – Temperature field of an L-shaped domain 7

Figure 1.2: The finite element mesh in ElmerGUI

Choose Accept to close the window.

In the equation section we choose the relevant equations and parameters related to their solution. In this
case we’ll have one set only one equation – the heat equation.
When defining Equations and Materials it is possible to assign the to bodies immediately, or to use mouse
selection to assign them later. In this case we have just one body and one boundary and therefore its easier
to assign the Equation and Material to it directly.

c CSC – IT Center for Science

1. Heat equation – Temperature field of an L-shaped domain 8

For the linear system solvers we are happy to use the defaults. One may however, try out different
preconditioners (ILU1,. . . ) or direct Umfpack solver, for example.
Model
Equation
Add
Name = Heat Equation
Apply to bodies = 1
Heat Equation
Active = on
Apply
OK
The Material section includes all the material parameters. They are divided to generic parameters which
are direct properties of the material without making any assumptions on the physical model, such as the
mass. Other properties assume a physical law, such heat conductivity.
Model
Material
Add
Name = Ideal
Apply to bodies = 1
General
Density = 1.0
Heat Equation
Heat Conductivity = 1.0
Apply
OK
A Body Force represents the right-hand-side of a equation that in this case represents the heat source.

Model
Body Force
Add
Name = Heating
Heat Source = 1.0
Apply to bodies = 1
Apply
OK
No initial conditions are required in steady state case.
In this case we have only one boundary and set it to zero.

Model
BoundaryCondition
Add
Heat Equation
Temperature = 0.0
Name = Zero
Apply to boundaries = 1
Apply
OK
For the execution ElmerSolver needs the mesh files and the command file. We have know basically
defined all the information for ElmerGUI to write the command file. After writing it we may also visually
inspect the command file.

c CSC – IT Center for Science

1. Heat equation – Temperature field of an L-shaped domain 9

Figure 1.3: The output log of ElmerSolver when used under ElmerGUI

Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. In saving the project all the
necessary files for restarting the case will be saved to the destination directory.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver
A convergence view automatically pops up showing relative changes of each iteration. As the case is linear
only one iteration was required for the solution and the second one just is needed to check the convergence.
The resulting output log is shown in figure 1.3.
Note: if you face problems in the solution phase and need to edit the setting, always remember to save
the project before execution.
To view the results we may start the the internal VTK widget, or use ElmerPost for postprocessing as is
done here
Run
Start postprocessor
The default configuration shows the surface mesh colored with temperature as in picture 1.4. You may test
the different options. For example, choosing in the Color Mesh Edit window the Mesh Style to be
Line and Edge Style to be All and pressing Apply the wireframe of the computational mesh is seen.

c CSC – IT Center for Science

1. Heat equation – Temperature field of an L-shaped domain 10

Figure 1.4: The temperature distribution of the L-shaped domain as visualized ElmerPost

Results
As a reference result the maximum temperature in the structure is given. For a comparison the same problem
was solved six times with different element sizes using the -relh flag of ElmerGrid format located in the
Mesh menu under Configure. After choosing Remesh and saving the mesh the solver may be recalled
with the modified mesh.
The maximum temperature obtained by using different meshes is recorded in Table 1.1. From the results
one can see that the result converges. With a denser mesh the result is more accurate, but solving the problem
takes more calculation time. For reference, the CPU-time used in each case is also shown in the table (using
Lenovo 60 laptop).

Table 1.1: Results with different element sizes

Model mesh h [m] Elements max(T ) [K] cpu time [s]
0.2 75 0.1452 0.14
0.1 300 0.1477 0.19
0.05 1200 0.1489 0.41
0.025 5120 0.14925 1.22
0.01 30800 0.14937 7.77

c CSC – IT Center for Science

Tutorial 2

Heat equation – Temperature field of a

solid object

Directory: TemperatureGenericGUI
Solvers: HeatSolve
Tools: ElmerGUI,tetgen,OpenCascade
Dimensions: 3D, Steady-state

Problem description
This tutorial tried to demonstrate how to solve the heat equation for a generic 3D object. The solid object
(see figure 2.1) is heated internally by a heat source. At some part of the boundary the temperature is fixed.
Mathemetically the problem is described by the Poisson equation

−κ∆T = ρf in Ω
(2.1)
T = 0 on Γ

where κ is the heat conductivity, T is the temperature and f is the heat source. It is assumed that density and
heat conductivity are constants.
To determine the problem we assume that the part of the boundary is fixed at T0 = 293 K, the internal
heat generation is, h = 0.01 W/kg, and use the material properties of aluminium.

Figure 2.1: Generic object being heated

11
2. Heat equation – Temperature field of a solid object 12

Figure 2.2: The computational mesh showing the three joined boundaries

Solution procedure
Start ElmerGUI from command line or by clicking the icon in your desktop. Here we describe the essential
steps in the ElmerGUI by writing out the clicking procedure. Tabulation generally means that the selections
are done within the window chosen at the higher level.
The geometry is given in step format in file pump_carter_sup.stp in the samples/step direc-
tory of ElmerGUI, This file is kindly provided at the AIM@SHAPE Shape Repository by INRIA. The heat
equation is ideally suited for the finite element method and the solution may be found even at meshes that for
some other problems would not be feasible. Therefore you may easily experiment solving the same problem
with different meshes. If you lack either OpenCascade or tetgen you might try to solve the problem with the
grd files angle3d.grd, angles3d.grd, bench.grd, or cooler.grd, for example.
The CAD geometry defined by the step file is transformed on-the-fly by OpenCascade library into a stl
file for which tetgen (tetlib) creates tetrahedral volume discretization.
Load the input file:
File
Open -> pump_carter_sup.stp
You should obtain your mesh and may check in the number of element in the Model summary. With
tetgen the default setting generate 33902 nodes and 110010 tetrahderal elements. In order to affect the mesh
density study the command-line options in the tetgen manual. Here we continue with the default mesh.
The stl description of the mesh only provides with one body and one surface. Therefore we need to
divide the existing surface. First choose the surface by clicking on it (it should turn red) and then
Mesh
Divide Surface
Divide
You may as well use the red icon with arrows pointing to different directions. The algorithm separates all
boundaries with more than 20 angle separation between surfaces. We want to set the temperature at the
inside of the holes and in that aim again join three boundaries (see figure 2.2). For that aim choose the three
boundaries as shown in the picture by pressing the Ctrl-key down.
Mesh
Unify Surface
After we have the mesh we start to go through the Model menu from the top to bottom. In the Setup we
choose things related to the whole simulation such as file names, time stepping, constants etc. The simulation

c CSC – IT Center for Science

2. Heat equation – Temperature field of a solid object 13

is carried out in 3-dimensional cartesian coordinates and in steady-state. Only one steady-state iteration is
needed as the case is linear.
Model
Setup
Simulation Type = Steady state
Steady state max. iter = 1
Choose Accept to close the window.
In the equation section we choose the relevant equations and parameters related to their solution. In this
case we’ll have one set only one equation – the heat equation.
When defining Equations and Materials it is possible to assign the to bodies immediately, or to use
mouse selection to assign them later. In this case we have just one body and therefore its easier to assign the
Equation and Material to it directly, whereas the active boundary is chosen graphically.
For the linear system solvers we are happy to use the defaults. One may however, try out different
preconditioners (ILU1,. . . ), for example.
Model
Equation
Add
Name = Heat Equation
Apply to bodies = Body 0
Heat Equation
Active = on
Add
OK
The Material section includes all the material parameters. They are divided to generic parameters which
are direct properties of the material without making any assumptions on the physical model, such as the
mass. Other properties assume a physical law, such heat conductivity. We choose Aluminium from the
Material library which automatically sets for the needed material properties.
Model
Material
Add
Material library
Aluminium
Apply to bodies = Body 0
Add
OK
A Body Force represents the right-hand-side of a equation that in this case represents the heat source.
Model
Body Force
Add
Name = Heating
Heat Source = 0.01
Apply to bodies = Body 0
Add
OK
No initial conditions are required in steady state case.
In this case we have only one boundary and set it to room temperature. First we create the boundary
condition
Model
BoundaryCondition

c CSC – IT Center for Science

2. Heat equation – Temperature field of a solid object 14

Add
Heat Equation
Temperature = 293.0
Name = RoomTemp
Add
OK
Then we set the boundary properties
Model
Set boundary properties
Choose the defined group of three boundaries by clicking with the mouse and apply the condition for this
boundary.
Boundary condition
RoomTemp
For the execution ElmerSolver needs the mesh files and the command file. We have know basically
defined all the information for ElmerGUI to write the command file. After writing it we may also visually
inspect the command file.
Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. In saving the project all the
necessary files for restarting the case will be saved to the destination directory.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver
A convergence view automatically pops up showing relative changes of each iteration. As the case is linear
only one iteration was required for the solution and the second one just is needed to check the convergence.
The norm of the solution should be around 389.853 K.
Note: if you face problems in the solution phase and need to edit the setting, always remember to
regenerate the sif file and save the project before execution.

Postprocessing
To view the results we may use the ElmerPost postprocessor or start the the internal VTK widget as is done
here,
Run
Postprocessor (VTK)
The default configuration shows just the object. To color the surface with the temperature choose
Surfaces
Surface: Temperature
Apply
There seems to be some handing nodes generated by tetgen with zero temperature. Thus, fixing the Min to
293 and setting crossing the Keep limits box gives better color scale. You may also turn on opasity in
order to see through the object, 10-20% is a good value. This way you’ll able to see some isosurfaces that
you might want to define. Some examples of the visualizations may be seen in figure 2.3.

c CSC – IT Center for Science

2. Heat equation – Temperature field of a solid object 15

Figure 2.3: The temperature distribution of the solid object domain as visualized using the VTK-based
postprocessor

c CSC – IT Center for Science

Tutorial 3

Linear elasticity equation – Loaded

elastic beam

Directory: ElasticBeam3D
Solvers: StressSolve
Tools: ElmerGUI
Dimensions: 3D, Steady-state

Case definition
Assume a homogenous, elastic beam being rigidly supported on one end. On the other end it is subjected
with a load of 2000 N resulting from an attached object in the gravitational field. The gravity affects also
the beam itself. The length of the beam is 1 m and the thickness is 0.05 m, and the width 0.1 m. Material
properties of the beam are those of dry pine timber: Poisson ratio 0.37, Young’s modulus 10 · 109 N/m2 ,
and density 550 kg/m3 . The problem is to solve the displacement and stress field of the beam. Here the
StressSolve routine based on the linear theory of elasticity is applied.

Solution procedure
The mesh is given in ElmerGrid format in file beam3d.grd, load this file.
File
Open -> beam3d.grd
You should obtain your mesh and may check that it consists of 6073 nodes and of 1200 quadratic hexahedral
elements. The second order elements give improved accuracy compared to the first order elements as they
avoid the phenomenom known as locking.

Figure 3.1: The mesh used in the computations

16
3. Linear elasticity equation – Loaded elastic beam 17

The Material section includes all the material parameters. They are divided to generic parameters which
are direct properties of the material without making any assumptions on the physical model, such as the
mass. Other properties assume a physical law, such as Young’s modulus and Poisson ratio.
Model
Material
Name = Pine
General
Density = 550
Linear Elasticity
Youngs Modulus = 10.0e9
Poisson ratio = 0.37
Apply to Bodies = Body 1
Add
OK
In this case there is a body force i.e. the gravity acting on the beam. We assume that the gravity points
to the negative y direction.

Model
BodyForce
Name = Gravity
Linear Elasticity
Force 2 = $ -9.81 * 550
Apply to Bodies = Body 1
Add
OK

c CSC – IT Center for Science

3. Linear elasticity equation – Loaded elastic beam 18

Here we use a MATC expression for computing the volume force. This expression is constant and is computed
when the command file is interpreted.
Convergence should be obtained with the default initial condition i.e. zero for all fields, hence no initial
condition is applied.
The first boundary condition fixes the beam rigidly at the wall. The second boundary condition distributes
the load of 2000 N uniformly on the area of 5.0e-3 m2 .
Model
BoundaryCondition
Name = Wall
Linear elasticity
Displacement 1 = 0.0
Displacement 2 = 0.0
Displacement 3 = 0.0
Add
New

Name = Mass
Linear elasticity
Force 2 = -4.0e5
Add
The conditions may also be assigned to boundaries in the Boundary condition menu, or by clicking
with the mouse. Here we use the latter approach as that spares us of the need to know the indexes of each
boundary.
Model
Set boundary properties
Choose the wall end of the beam -> set boundary condition Wall
Choose the other end of the beam -> set boundary condition Mass
For the execution ElmerSolver needs the mesh files and the command file. We have know basically
defined all the information for ElmerGUI to write the command file. After writing it we may also visually
inspect the command file.
Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. The project includes all the files
needed to restart the case.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver
A convergence view automatically pops up showing relative changes of each iteration. Two iterations are
performed, the second one only to ensure convergence at the nonlinear level.

Results
When there are some results to view we may start the postprocessor, this time we use ElmerPost.
Run
Start postprocessor

c CSC – IT Center for Science

3. Linear elasticity equation – Loaded elastic beam 19

As a result the absolute value of maximum displacement is shown. The maximum displacement is 6.36 cm
To visualize the displacement in the geometry using ElmerPost can be done with the following command in
the Elmer-Post command line.
math n0=nodes
math nodes=n0+Displacement

To redraw the picture with new settings use the rightenmost icon on the top row. The resulting picture is
shown in Fig 3.2 Note that the displacement are so large that the assumption of linearity may be severely

Figure 3.2: The displaced shape of the elastic beam colored with the von Mises stresses

questioned. When further increasing the loading one should resort to a solver that is able to catch the
geometric nonlinearities.

Extra task: Gravity in x direction

The beam should be more rigid if the beam is oriented differently. For that aim, change the direction of
gravity to orient in the negitive x. Change the body force

Model
BodyForce
Linear Elasticity
Force 1 = $ -9.81*550
Update
OK

and the boundary condition

Model
BoundaryCondition
Linear elasticity
Force 1 = -4.0e5
Update
OK
The rigidity should scale as dh3 and hence the maximum displacement should be reduced roughly to one
quarter of the original.

c CSC – IT Center for Science

Tutorial 4

Navier-Stokes equation – Laminar

incompressible flow passing a step

Directory: FlowStepGUI
Solvers: FlowSolve
Tools: ElmerGUI
Dimensions: 2D, Steady-state

Case definition
This tutorial represents the canonical step flow of viscous fluid. A fluid, flowing past a step (see figure 29.1),
has the density 1 kg/m and viscosity 0.01 kg/ms. The velocity profile at the inlet is parabolic with a mean
velocity < vx >= 1.0 m/s and vy = 0.0 m/s. At the outlet only the vertical component is defined, vy =
0.0 m/s. At all other walls the no-slip boundary condition, ~v = 0, is applied. Thus the Reynolds number for
the case is around 100.

Figure 4.1: Geometry of the step flow problem

Mathematically the problem to be solved is

−∇ · (2µε) + ρ~u · ∇~u + ∇p = 0 in Ω
(4.1)
∇ · ~u = 0 in Ω

with the boundary conditions


 ux = 1 on Γinlet
ux = 0 on Γno−slip (4.2)
uy = 0 on Γinlet ∪ Γoutlet ∪ Γno−slip


where µ is the viscosity, ε is the strain tensor, ρ is the density, ~u is the velocity and p is the pressure. It is
assumed that the density and viscosity are constants.

20
4. Navier-Stokes equation – Laminar incompressible flow passing a step 21

Solution procedure
The mesh is given in ElmerGrid format in file step.grd, load this file.
File
Open -> step.grd
You should obtain your mesh and may check that it consists of 9696 nodes and of 9442 bilinear elements.

Model
Summary...
After we have the mesh we start to go through the Model menu from the top to bottom. In the Setup we
choose things related to the whole simulation. The steady-state simulation is carried out in 2-dimensional
cartesian coordinates, which are also the defaults.

Model
Setup
Simulation Type = Steady state
Coordinate system = Cartesian

In the equation section we choose the relevant equations and parameters related to their solution. In this case
the only the Navier-Stokes equation is needed.
When defining Equations and Materials it is possible to assign the to bodies immediately, or to use
mouse selection to assign them later. In this case we have just one body and therefore its easier to assign the
Equation and Material to it directly. One could also edit the solver setting in order to try different strategies
for solving the nonlinear or linear system. Initially the Navier-Stokes solver uses the more robust Picard
iteration which is changed to Newton iteration after few initial steps. For the given viscosity the default
values are ok, but may need tuning when going into higher Reynolds numbers.
Model
Equation
Name = Navier-Stokes
Apply to Bodies = 1
Navier-Stokes
Active = on
Edit Solver Setting
Nonlinear System
Max. iterations = 20
Newton after iterations = 3
Add
OK
The Material section includes all the material parameters. They are divided to generic parameters which
are direct properties of the material without making any assumptions on the physical model, such as the
density. Other properties assume a physical law, such as viscosity.
Model
Material
Name = Ideal
General
Density = 1.0
Navier-Stokes
Viscosity = 0.01
Apply to Bodies = 1
Add
OK

c CSC – IT Center for Science

4. Navier-Stokes equation – Laminar incompressible flow passing a step 22

The current case does not have any body forces. Convergence should also be obtained using the default
initial condition which sets all field values to zero. Hence no setting for initial condition are needed.
Only one boundary condition may be applied to each boundary and therefore all the different physical
BCs for a boundary should be grouped together. In this case the Temperature and Velocity. The side walls
are assumed to be adiabatic.
The parabolic inlet-profile is achieved using the MATC environment. To be able to edit the content of the
inlet profile click Enter to open an edit box for the Velocity 1. The given expression will be interpreted
at run-time so that vx = 6(y − 1)(2 − y). As y ∈ [1, 2] thereby creating a parabolic velocity profile with a
mean velocity of unity.
Model
BoundaryCondition
Name = Inlet
Navier-Stokes
Velocity 1 = Variable Coordinate 2; Real MATC ‘‘6*(tx-1)*(2-tx)’’
Velocity 2 = 0.0
Add
New

Name = Outlet
Navier-Stokes
Velocity 2 = 0.0
Add
New

Name = Walls
Navier-Stokes
Velocity 1 = 0.0
Velocity 2 = 0.0
Add
The conditions may also be assigned to boundaries in the Boundary condition menu, or by clicking
with the mouse. Here we use the latter approach as that spares us of the need to know the indexes of each
boundary.
Model
Set boundary properties
Choose Inlet -> set boundary condition Inlet
Choose Outlet -> set boundary condition Outlet
Choose Walls -> set boundary condition Walls
For the execution ElmerSolver needs the mesh files and the command file. We have now basically defined
all the information for ElmerGUI to write the command file. After writing it we may also visually inspect
the command file.
Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. The project includes all the files
needed to restart the case. Create a suitable directory for the case if needed.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver

c CSC – IT Center for Science

4. Navier-Stokes equation – Laminar incompressible flow passing a step 23

A convergence view automatically pops up showing relative changes of each iteration. The problem should
converge in nine iterations. When there are some results to view we may start the postprocessor also
Run
Start postprocessor

Results
The results may be viewed using the postprocessor as shown in Figure 4.2 and 4.3. One may also register
specific values, for example the pressure difference is 4.23 Pa, the minimum and maximum lateral velocities
are -0.164 m/s and 1.3709 m/s, respectively. One special result of interest is the point, on the x-axis, at which
the direction of the flow changes. In this case its position is about 6.6 m after the step.

Figure 4.2: Absolute value of the velocity field

Figure 4.3: Pressure field

Extra task: Decreasing the viscosity

Try what happens if the viscosity is further decaresed by a factor 10. Convergence may be difficult to obtain.
Some tricks that may be tested include
• Introducting a relaxation factor (typically in the range 0.5-0.7)
• Increasing number of nonlinear iterations

• Favoring Picard iteration over Newton

• Increasing mesh density (and length of domain)
Don’t be worried if you fail to find convergence. This task will mainly act as a motivator in using turbulence
models for higher Reynolds numbers.
Remember to re-perform the following phases in order to get the updated results

c CSC – IT Center for Science

4. Navier-Stokes equation – Laminar incompressible flow passing a step 24

Sif
Generate
File
Save Project
Run
Start solver
You may just reload the results in the postprocessor rather than closing and opening the program.

c CSC – IT Center for Science

Tutorial 5

Navier-Stokes equation – Turbulent

incompressible flow passing a step

Directory: FlowStepGUI
Solvers: FlowSolve,KESolver
Tools: ElmerGUI
Dimensions: 2D, Steady-state

Case definition
This tutorial is a natural contination of the tutorial 4 where the same case was solved with a smaller Reynolds
number. It is advicable to study that case before.
When Reynolds number increases the Navier-Stokes equations do not posses any steady-state solution.
Basically the solution can be averaged simulating the transient flow over time. However, the computational
cost of this approach is often very heavy particularly while at high Reynolds numbers the computational
mesh in direct numerical simulation needs to be very dense. Instead its customary to solve timeaveraged
equations. Unfortunately these equations include unknown correlations between quantities that need to
modeled in some way.
The workhorse of turbulence modeling is the k − ε model which is used in this tutorial. The k − ε model
is a two-equation model that introduces two additional variables – the turbulent kinetic energy k and the
turbulent dissipation ε which determines the scale of the turbulence.
The case under study is the canonical step flow of viscous fluid. A fluid, flowing past a step has the
density 1 kg/m3 and viscosity 1.0e − 4 kg/ms. The velocity profile at the inlet is defined by a parabolic
profile with mean velocity vx = 1.0 m/s and vy = 0.0 m/s. This way the Reynolds number will be 10000.
At the outlet only the vertical component is defined, vy = 0.0 m/s. At all other walls the no-slip boundary
condition, ~v = 0, is applied.
Also the new turbulent variables require boundary conditions. In the inflow the condition could reflect the
values for developed turbulent profile. Here we roughly estimate the turbulent kinetic energy and elongate the
distance before the step to have a fully developed turbulent profile. From the literature it is the the turbulent
intensity i.e. the kinetic energy of turbulence vs. the kinetic energy of mean flow scalas as 0.16Re−1/8 .
For our current configuration a estimate for the turbulent kinetic energy is 0.00457. For the walls a no-slip
condition is applied which also set the values of the turbulent parameters accordingly. Also boundary layer
model could be used but here our mesh should be able to capture even the boundary phenomena quite well.

Solution procedure
The mesh is given in ElmerGrid format in file steplong.grd, load this file.
File
Open -> steplong.grd

25
5. Navier-Stokes equation – Turbulent incompressible flow passing a step 26

You should obtain your mesh and may check that it consists of 14584 nodes and of 14175 bilinear elements.
Model
Summary...
After we have the mesh we start to go through the Model menu from the top to bottom. In the Setup we
choose things related to the whole simulation. The steady-state simulation is carried out in 2-dimensional
cartesian coordinates, which are also the defaults. The coupled system converges unfortunately quite slowly
and hence we need to increase the number of maximum iterations.
Model
Setup
Simulation Type = Steady state
Coordinate system = Cartesian
Steady state max. iter = 100
In the equation section we choose the relevant equations and parameters related to their solution. In
this case the Navier-Stokes and k − ε equations are needed. We want to solve the Navier-Stokes and k − ε
equations iteratively using only one nonlinear iteration for optimal convergence of the coupled system. Some
relaxation is needed in order to achieve convergence at all. We also relax a little bit on the steady state
convergence tolerance. Initially the Navier-Stokes solver uses the more robust Picard iteration which may
be changed to Newton iteration after the iteration progresses. However, here we want to supress the use of
Newton lineariarization since it seems to cause problems with the k − ε equation.

Model
Equation
Name = Flow equations
Apply to Bodies = Body 1
Navier-Stokes
Active = on
Edit Solver Setting
Nonlinear System
Max. iterations = 1
Relaxation factor = 0.5
Newton after tolerance = 0.0
Steady state
Convergence tol. = 1.0e-4
K-Epsilon
Active = on
Edit Solver Setting
Nonlinear System
Max. iterations = 1
Relaxation factor = 0.5
Steady state
Convergence tol. = 1.0e-4
Add
OK

c CSC – IT Center for Science

5. Navier-Stokes equation – Turbulent incompressible flow passing a step 27

Density = 1.0
Navier-Stokes
Viscosity = 1.0e-4
Viscosity Model = K-Epsilon
Apply to Bodies = Body 1
Add
OK
The current case does not have any body forces. To help in the convergence we make an rude intial
guess.
Model
Initial Condition
Name = Initial Guess
Navier-Stokes
Velocity 1 = 0.0
Velocity 2 = 0.0
K-Epsilon
Kinetic Energy = 0.00457
Kinetic Dissipation = 1.0e-4
When defining Boundary conditions it is possible to assign the to boundaries immediately, or to use
mouse selection to assign them later. In this case we have use the latter since we do not necessarily know
the numbering of boundaries by heart. There is a special boundary condition that takes care of the Boundary
conditions for the noslip walls for both the Navier-Stokes and k − ε equation. Additionally there are inlet
and outlet conditions. For the inlet click Enter to open an edit box for the Velocity 1 when typing in
the expression. which will be evaluated at run-time so that vx = 6(y − 1)(2 − y).
Model
BoundaryCondition
Add
Name = Inlet
Navier-Stokes
Velocity 1 = Variable Coordinate 2; Real MATC ‘‘6*(tx-1)*(2-tx)’’
Velocity 2 = 0.0
K-Epsilon
Kinetic Energy = 0.00457
Kinetic Dissipation = 1.0e-4
Add
New

Name = Outlet
Navier-Stokes
Velocity 2 = 0.0
Add
New

Name = Walls
Navier-Stokes
Noslip Wall BC = on
Add
The conditions may also be assigned to boundaries in the Boundary condition menu, or by clicking
with the mouse. Here we use the latter approach as that spares us of the need to know the indexes of each
boundary.
Model

c CSC – IT Center for Science

5. Navier-Stokes equation – Turbulent incompressible flow passing a step 28

Set boundary properties

Choose inlet -> set boundary condition Inlet
Choose outlet -> set boundary condition Outlet
Choose walls -> set boundary condition Walls
For the execution ElmerSolver needs the mesh files and the command file. We have now basically defined
all the information for ElmerGUI to write the command file. After writing it we may also visually inspect
the command file.
Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. The project includes all the files
needed to restart the case. Create a suitable directory for the case if needed.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver
A convergence view automatically pops up showing relative changes of each iteration. The convergence
if far from monotonic but computation should terminate when sufficient convergence is reached after 30
iterations.

Results
When there are some results to view we may start the postprocessor. This time we use the internal VTK
based postprocessor for visualization. Also ElmerPost could be used.
Run
Postprocessor (VTK)
The results may be viewed using the postprocessor as shown in Figures 5.1 and 5.2. One may also
register specific values, for example the pressure difference is 0.302 Pa, the minimum horizontal and vertical
velocities are -0.213 m/s and -0.0834 m/s, respectively. One special result of interest is the point, on the
x-axis, at which the direction of the flow changes. In this case its position is about 5.1 m after the step.

Figure 5.1: Variables of the Navier-Stokes solver: absolute velocity on top and pressure on bottom

c CSC – IT Center for Science

5. Navier-Stokes equation – Turbulent incompressible flow passing a step 29

Figure 5.2: Variables of the k − ε solver: kinetic energy on top and its dissipation on bottom

c CSC – IT Center for Science

Tutorial 6

Transient flow and heat equations –

Rayleigh-Benard instability

Directory: RayleighBenardGUI
Solvers: HeatSolve, FlowSolve
Tools: ElmerGUI
Dimensions: 2D, Transient

Case definition
This tutorial is about simulating the developing of the Rayleigh-Benard instability in a rectangular domain
(Figure 30.1) of dimensions 0.01 m height and 0.06 m length. The simulation is performed with water and
the material parameters of water required by the Elmer model are presented in Table 30.1. The temperature
difference between the upper and lower boundary is set to 0.5 so that lower one has the temperature of 293.5
K and the upper one has the temperature of 293 K.
The density of water is inversely proportional to its temperature. Thus, heated water starts to flow
upwards, and colder downwards due to gravity. In this case we assume that the Boussinesq approximation is
valid for thermal incompressible fluid flow. In other words, the density of the term ρf~ in the incompressible
Navier-Stokes equation can be redefined by the Boussinesq approximation

ρ = ρ0 (1 − β(T − T0 ))

where β is the heat expansion coefficient and the subscript 0 refers to a reference sate.

Figure 6.1: Domain.

30
6. Transient flow and heat equations – Rayleigh-Benard instability 31

Table 6.1: Material parameters for water

parameter value
density 998.3 kg/m3
viscosity 1040e-6 Ns/m2
heat capacity 4183 J/(kg·K)
heat conductivity 0.58 W/(m·K)
heat expansion coefficient 2.07e-4 K−1
reference temperature 293 K

Solution procedure
The mesh is given in ElmerGrid format in file box.grd, load this file.
File
Open -> box.grd
You should obtain your mesh and may check that it consists of 3036 bilinear elements.
There is a possibility to divide and unify edges to simplify the case definition in the future.
Choose (left wall + right wall (Ctrl down)) -> unify edge

After we have the mesh we start to go through the Model menu from the top to bottom. In the Setup we
choose things related to the whole simulation such as file names, time stepping, constants etc. The simulation
is carried out in 2-dimensional cartesian coordinates. 2nd order bdf time-stepping method is selected with
200 steps and with step size of two seconds.
Model
Setup
Simulation Type = Transient
Steady state max. iter = 20
Time Stepping Method = bdf
BDF Order = 2
Time Step Intervals = 200
Time Step Sizes = 2.0
Gravity = ...
In the equation section we choose the relevant equations and parameters related to their solution. In this case
we’ll have one set of equations (named “Natural Convection”) which consists of the heat equation and of the
Navier-Stokes equation.
When defining Equations and Materials it is possible to assign the to bodies immediately, or to use
mouse selection to assign them later. In this case we have just one body and therefore its easier to assign
the Equation and Material to it directly. It is important to select the convection to be computed since that
couples the velocity field to the heat equation.
The system may include nonlinear iterations of each equation and steady state iterations to obtain conver-
gence of the coupled system. It is often a good idea to keep the number of nonlinear iterations in a coupled
case low. Here we select just one nonlinear iteration for both equations. For the linear system solvers we are
happy to use the defaults. One may however, try out different preconditioners (ILU1,. . . ) or direct Umfpack
solver, for example.
Model
Equation
Name = Natural Convection
Apply to Bodies = 1
Heat Equation

c CSC – IT Center for Science

6. Transient flow and heat equations – Rayleigh-Benard instability 32

Active = on
Convection = Computed
Edit Solver Setting
Nonlinear System
Max. iterations = 1
Navier-Stokes
Active = on
Edit Solver Setting
Nonlinear System
Max. iterations = 1
Add
OK
The Material section includes all the material parameters. They are divided to generic parameters which are
direct properties of the material without making any assumptions on the physical model, such as the mass.
Other properties assume a physical law, such as conductivities and viscosity.
Here we choose water at room temperature from the material library. You may click trough the material
parameters of the various solvers to ensure that the properties are indeed as they should be. Any consistant
set of units may be used in Elmer. The natural choice is of course to perform the computations in SI units.
Apart from the properties from the material database, we reference temperature for the Boussinesq ap-
proximation.

Model
Material
Material library
Water (room temperature)
General
Reference Temperature = 293
Apply to Bodies = 1
Add
OK

A Body Force represents the right-hand-side of a equation. It is generally not a required field for a
body. In this case, however, we apply the buoyancy resulting from heat expansion as a body force to the
Navier-Stokes equation.

Model
Body Force
Name = Buoyancy
Apply to Bodies = 1
Navier-Stokes
Boussinesq = on
Add
OK
Initial conditions should be given to transient cases. In this case we choose a constant Temperature field
and an small initial velocity that initializes the symmetry break.

Model
Initial Condition
Name = Initial Guess
Heat Equation
Temperature = 293
Navier-Stokes
Velocity 1 = 1.0e-9
Velocity 2 = 0.0

c CSC – IT Center for Science

6. Transient flow and heat equations – Rayleigh-Benard instability 33

Only one boundary condition may be applied to each boundary and therefore all the different physical
BCs for a boundary should be grouped together. In this case the Temperature and Velocity. The side walls
are assumed to be adiabatic.
Model
BoundaryCondition
Name = Bottom
Heat Equation
Temperature = 293.5
Navier-Stokes
Velocity 1 = 0.0
Velocity 2 = 0.0
Add
New

Name = Top
Heat Equation
Temperature = 293
Navier-Stokes
Velocity 1 = 0.0
Velocity 2 = 0.0
Add
New

Name = Sides
Navier-Stokes
Velocity 1 = 0.0
Velocity 2 = 0.0
Add
The conditions may also be assigned to boundaries in the Boundary condition menu, or by clicking
with the mouse. Here we use the latter approach as that spares us of the need to know the indexes of each
boundary.
Model
Set boundary properties
Choose Bottom -> set boundary condition Bottom
Choose Top -> set boundary condition Top
Choose Sides -> set boundary condition Sides
For the execution ElmerSolver needs the mesh files and the command file. We have know basically
defined all the information for ElmerGUI to write the command file. After writing it we may also visually
inspect the command file.
Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. The project includes all the files
needed to restart the case.
File
Save Project
After we have successfully saved the files we may start the solver
Run
Start solver

c CSC – IT Center for Science

6. Transient flow and heat equations – Rayleigh-Benard instability 34

A convergence view automatically pops up showing relative changes of each iteration. When there are some
results to view we may start the postprocessor also
Run
Start postprocessor

Results
Due to the number of the time-steps the simulation may take around ten minutes. You may inspect the
results with ElmerPost as the time-steps are computed, or wait until all timesteps have been computed. When
opening the result file using ElmerGUI ElmerPost only opens the first time-step. Therefore it is important
to reopen the file and load the time-steps of interest. Pressing the button All selects all the calculated time
steps. A video of the results can be viewed by selecting the option Timestep Control and pressing the
button Loop under the Edit menu.
In Figures 30.2 and 30.3 the obtained temperature distribution and the velocity vectors are presented.
The maximum velocity in the system should be about 0.5 mm/s.

Figure 6.2: Temperature distribution at 260 s.

Figure 6.3: Velocity vectors at 260 s.

Extra task: Sensitivity to temperature difference

If you have time you may try to solve the case with different parameters. Changing the temperature difference
is one way of affecting the instability of the system. Decreasing the tempereture differences the system
eventually becomes steady state and the convection rolls vanish alltogether. Increasing the temperature
difference may increase the number of convection rolls and eventually the system becomes fully chaotic.
Note that changing the temperature difference also affects to the time scale of the wake.

c CSC – IT Center for Science

Tutorial 7

Electrostatic equation – Computation of

fringe capacitance

Directory: FringeCapacitance
Solvers: StatElecSolver
Tools: ElmerGUI
Dimensions: 2D, Steady-state

Case definition
This case presents solving the Laplace equation for electric potential.

−∇ · ε∇φ = 0 ∈ Ω (7.1)

where the eletric potential φ is given at conducting surfaces. This is a standard type of equation with many
variants in physics, electrostatics being just one of them. From the solution one may calculate derived fields,
and capacitance which is obtained from the total electric energy
Z
1
E= ε|∇φ|2 dΩ (7.2)
2

and the relation E = CU 2 /2.

Figure 7.1: The geometry of the capacitor

The geometry studied is a 2D plate capacitor having the well known approximation for Capacitance
Capp = εr ε0 A/d. With the help of the simulation one may evaluate the fringe capacitance resulting from
the end effects of the capacitor geometry. The measuments of the capacitor are 10 × 1, and the distance
to ground is also 1. Defining the permittivity of vacuum to be 0 = 1 the comparison to the analytical
approximation becomes trivial since then then Capp = 10.

35
7. Electrostatic equation – Computation of fringe capacitance 36

The derived fields in the StatElecSolver are computed by averaging the fields over elements – not using
the Galerkin method which would provide optimal accuracy. The fields may, however, be sufficient for
visualization purposes.

Figure 7.2: Computational mesh used in the simulation

Solution procedure
The definitions for the electrostatic equation are not loaded into ElmerGUI by default. Hence, one needs to
load these before starting the simulations.
File
Definitions
Append -> electrostatics.xml

The additional definitions should recide in the directory edf-extra within the distribution. Moving the
desired xml files to the edf-directory enables automatic loading of the definitions at start-up. By inspecting
the definitions in the Elmer Definitions File editor one may inspect that the new definitions
were really appended.
The mesh is given in ElmerGrid format in file disc.grd, load this file.

File
Open -> disc.grd
You should obtain your mesh and may check that it consists of 30484 nodes and of 30348 bilinear elements.
After we have the mesh we start to go through the Model menu from the top to bottom. In the Setup we
choose things related to the whole simulation such as file names, time stepping, constants etc. The steady-
state simulation is carried out in 2-dimensional cartesian coordinates. For convenience we also set ε equal
to one.
Model
Setup
Simulation Type = Steady state
Permittivity of Vacuum = 1.0
In the equation section we choose the relevant equations and parameters related to their solution. In this case
we’ll have only the electrostatics solver.
When defining Equations and Materials it is possible to assign the to bodies immediately, or to use
mouse selection to assign them later. In this case we have just one body and therefore its easier to assign the
Equation and Material to it directly.

c CSC – IT Center for Science

7. Electrostatic equation – Computation of fringe capacitance 37

In the solver specific options we want to activate some flags that are needed to invoke the computation
of derived fields. For the linear system solvers we are happy to use the defaults. One may however, try out
different preconditioners (ILU1,. . . ) or direct Umfpack solver, for example.
Model
Equation
Name = Electrostatics
Apply to Bodies = 1
Electrostatics
Active = on
Edit Solver Settings
Solver specific options
Calculate Electric Field = True
Calculate Electric Energy = True
Add
OK

The Material section includes all the material parameters. In this case we only have the relative permittivity
which we set to one.
Model
Material
Electrostatics
Relative Permittivity = 1.0
Apply to Bodies = 1
Add
OK
We have two boundary conditions for the potential at the ground and at the capacitor. For other bound-
aries the do-nothing boundary results to zero flux over the boundary.

Model
BoundaryCondition
Name = Ground
Electrostatics
Potential = 0.0
Add
New

Name = Capacitor
Electrostatics
Potential = 1.0
Add
The conditions may also be assigned to boundaries in the Boundary condition menu, or by clicking
with the mouse. Here we use the latter approach as that spares us of the need to know the indexes of each
boundary.

Model
Set boundary properties
Choose Ground -> set boundary condition Ground
Choose Capacitor -> set boundary condition Capacitor
For the execution ElmerSolver needs the mesh files and the command file. We have know basically
defined all the information for ElmerGUI to write the command file. After writing it we may also visually
inspect the command file.

c CSC – IT Center for Science

7. Electrostatic equation – Computation of fringe capacitance 38

Sif
Generate
Edit -> look how your command file came out
Before we can execute the solver we should save the files in a directory. The project includes all the files
needed to restart the case.

File
Save Project
After we have successfully saved the files we may start the solver

Run
Start solver
A convergence view automatically pops up showing relative changes of each iteration. The equation is
fully linear and hence only two iterations are needed – the second one just ensures that convergence of the
nonlinear level was really obtained. When the solution has finished we may start the postprocessor to view
some results.

Run
Start postprocessor

Results
From the output of the simulation one may see that the capacitance in this case was 13.70 compared to the
analytical estimate of 10. Hence the fringe capacitance in this case increases the capacitance by 37 %.

Figure 7.3: The electrostatic potential

Figure 7.4: The electrostatic energy density, a close-up

c CSC – IT Center for Science

Part II

non-GUI Problems

39
Tutorial 8

Radiation heat transfer

Directory: TemperatureRadiation
Solvers: HeatSolve
Tools: ElmerGrid, editor
Dimensions: 2D, Axi-Symmetric

Case definition
At high temperature the radiation heat transfer between walls is often the dominating heat transfer mecha-
nism. In this tutorial we look how radiation heat transfer between cocentric cylinders is modeled.

Solution Procedure
The problem is a pure heat transfer problem that may be solved with HeatSolve. The view and Geb-
harht factors associated with the radiation are solved as a first step in solving the equations. Thereafter the
nonlinear heat equation is solved until convergence is reached.
The computatonal mesh is done with ElmerGrid in directory radiation with the command
ElmerGrid 1 2 radiation
The directory is given in the header of the command file
Header
Mesh DB "." "radiation"
End
The only constant required is the Stefan-Boltzmann constant that gives the relationship between temperature
and radiation power
Constants
Stefan Boltzmann = 5.67e-8
End
The geometry is axisymmetric and the case is solved in steady state. As there is only one equation only 1
iteration for the system is required.
Simulation
Coordinate System = Axi Symmetric
Simulation Type = Steady State
Steady State Max Iterations = 1
Output Intervals = 1
Output File = "radiation.result"
Post File = "radiation.ep"
End

40
8. Radiation heat transfer 41

There are two bodies with the same equation but different properties.
Body 1
Equation = 1
Body Force = 1
Material = 1
Initial Condition = 1
End

Body 2
Equation = 1
Material = 2
Initial Condition = 1
End
The nonlinear equation requires realistic initial conditions. Otherwise convergence may not be obtained.
Initial Condition 1
Temperature = 250.0
End
The body force is the heating power in units W/kg.
Body Force 1
Heat Source = 10000
End
The material properties differ only in heat conductivity. Heat capacity is not actually needed since the case
is not transient.ă
Material 1
Density = 1.0
Heat Conductivity = 10.0
Heat Capacity = 1.0
End

Material 2
Density = 1.0
Heat Conductivity = 1.0
Heat Capacity = 1.0
End
The heat equation is solved with an itrative procedure that requires some relaxation for better convergence.
There are two different ways to discretize the radiation. There are two keywords defining when to switch to
the true Newtonian iteration which should give better convergence.
Solver 1
Equation = Heat Equation
Stabilize = True
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Convergence Tolerance = 1.0e-12
Linear System Max Iterations = 500
Linear System Preconditioning = ILU
Nonlinear System Newton After Iterations = 1
Nonlinear System Newton After Tolerance = 1.0e-4
Nonlinear System Max Iterations = 50
NonLinear System Convergence Tolerance = 1.0e-8

c CSC – IT Center for Science

8. Radiation heat transfer 42

Steady State Convergence Tolerance = 1.0e-8

Nonlinear System Relaxation Factor = 0.7
End
The only solver is the heat equation.

Equation 1
Active Solvers = 1
End
The radiation boundary conditions are set for two different boundaries. The first one is for the internal
object and the second one for the insulation. The normal direction of the surfaces is important since a wrong
direction may result to badly set problem for the view factor computation. Internal and external surfaces are
very different. The normal direction may be switched with the keyword Radiation Target Body. A
good sign of properly set case is that the view factors add up to about one.
Boundary Condition 1
Target Boundaries = 1
Heat Flux BC = True
Radiation = Diffuse Gray
Radiation Target Body = -1
Emissivity = 0.6
End

Boundary Condition 2
Target Boundaries = 2
Heat Flux BC = True
Radiation = Diffuse Gray
Radiation Target Body = -1
Emissivity = 0.1
End
The third boundary condition is the Dirichtlet condition for the extranal boundary. Dirichtlet conditions
boost up the convergence even though the heat equation is basically well defined also with external radiation
conditions.

Boundary Condition 3
Target Boundaries = 3
Temperature = 100.0
End

Results
With the given computational mesh the problem is solved in around 30 seconds. With 1 231 second order
9-noded rectangular elemenets the maximum temperature is 565.7 K. The corresponding results are shown
in Fig. 8.1.

c CSC – IT Center for Science

8. Radiation heat transfer 43

Figure 8.1: Temperature distribution in the radiation heat transfer problem

c CSC – IT Center for Science

Tutorial 9

Eigenvalue analysis of an elastic beam

Directory: ElasticEigenValues
Solvers: StressSolve, EigenSolve
Tools: ElmerGrid,Editor
Dimensions: 3D, Steady-state

Case definition
A homogenous, elastic silicon beam of dimensions 1 m length, 0.1 m height and 0.2 m width is supported
on its both ends (boundaries 1 and 2). A beam has the density 2330 kg/m3 , Poisson ratio 0.3 and Young’s
modulus 1011 N/m2 . The problem is to calculate the eigenvalues of the beam. Mathematically the equation
to be solved is
−ρω 2 φ = ∇ · τ (φ)
where ρ is the density, ω 2 is the eigenvalue, ω is the angular frequency, φ is the corresponding vibration
mode and τ is the stress tensor.

Figure 9.1: Beam.

44
9. Eigenvalue analysis of an elastic beam 45

Solution procedure
The mesh has been created by using Gambit software and it consists of 2500 elements. The mesh can be
converted to Elmer format with ElmerGrid with the command

ElmerGrid 7 2 mesh.FDNEUT

This command creates the directory which contains the Elmer mesh files.

Header
Mesh DB "." "mesh"
Include Path ""
Results Directory ""
End

A steady-state three-dimensional analysis is defined in the simulation section.

Simulation
Coordinate System = "Cartesian 3D"
Coordinate Mapping(3) = 1 2 3
Simulation Type = "Steady State"
Steady State Max Iterations = 1
Solver Input File = "eigen_values.sif"
Output File = "eigen_values.dat"
Post File = "eigen_values.ep"
End

The geometry of the problem is simple and it includes only one body and material.

Body 1
Equation = 1
Material = 1
End

Material 1
Youngs Modulus = 100e9
Poisson Ratio = 0.3
Density = 2330
End

The problem is solved according to linear elastic theory and due to that stress analysis is set to true.

Equation 1
Stress Analysis = True
End

In the solver section Stress Analysis is selected. In addition, the value of the keyword Eigen
Analysis have to set to true. The keyword Eigen System Values defines the number of the
computed eigenvalues. The problem also is possible to solve with iterative solver but we have used direct
solver in this time.

Solver 1
Equation = "Stress Analysis"
Eigen Analysis = Logical True
Eigen System Values = Integer 5
Linear System Solver = "direct"

c CSC – IT Center for Science

9. Eigenvalue analysis of an elastic beam 46

Variable = "Displacement"
Variable Dofs = 3
Linear System Iterative Method = "BiCGStab"
Linear System Max Iterations = 1000
Linear System Convergence Tolerance = 1.0e-08
Linear System Abort Not Converged = True
Linear System Preconditioning = "ILU0"
Linear System Residual Output = 1
Steady State Convergence Tolerance = 1.0e-05
Nonlinear System Convergence Tolerance = 1.0e-05
Nonlinear System Max Iterations = 1
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-02
Nonlinear System Relaxation Factor = 1
Linear System Precondition Recompute = 1
End
The beam is supported on its both ends and therefore displacements is set to zero in all the directions.
Boundary Condition 1
Target Boundaries(1) = 1
Displacement 1 = 0
Displacement 2 = 0
Displacement 3 = 0
End

Boundary Condition 2
Target Boundaries(1) = 2
Displacement 1 = 0
Displacement 2 = 0
Displacement 3 = 0
End
After that, the problem is ready to solve.

An anisotropic model

The same problem can also be solved as an anisotropic problem which causes a couple of changes in the
sif-file. First, it is reasonable to rename the files in the simulation section
Solver Input File = "eigen_values_aniso.sif"
Output File = "eigen_values_aniso.dat"
Post File = "eigen_values_aniso.ep"
For anisotropic material Young’s modulus have to redefine as a matrix. In this case the matrix is defined as
follows
Youngs Modulus
Size 6 6
Real 200e9 60e9 60e9 0 0 0
60e9 200e9 200e9 0 0 0
60e9 60e9 200e9 0 0 0
0 0 0 80e9 0 0
0 0 0 0 80e9 0
0 0 0 0 0 80e9
End
No more changes are needed in the sif-file.

c CSC – IT Center for Science

9. Eigenvalue analysis of an elastic beam 47

Results
Both the eigenvalues of the isotropic and the eigenvalues of the anisotropic model are shown below in Elmer
outputs. Figure 9.2 presents the computed eigenvectors of the beam with the isotropic model. The formula
ω = 2πf have been used in calculating frequencies (f ) (Table 9.1). According to the results the anisotropic
model yielded greater eigenvalues with these values of Young’s modulus.

EigenSolve: Computed Eigen Values:

EigenSolve: --------------------------------
EigenSolve: 1 (16737546.4275755,0.00000000000000D+000)
EigenSolve: 2 (48175589.4544061,0.00000000000000D+000)
EigenSolve: 3 (99674749.0526558,0.00000000000000D+000)
EigenSolve: 4 (110392974.959463,0.00000000000000D+000)
EigenSolve: 5 (253947166.278411,0.00000000000000D+000)

Isotropic model.

EigenSolve: Computed Eigen Values:

EigenSolve: --------------------------------
EigenSolve: 1 (29608629.8775828,0.00000000000000D+000)
EigenSolve: 2 (88782964.0905879,0.00000000000000D+000)
EigenSolve: 3 (198583949.415515,0.00000000000000D+000)
EigenSolve: 4 (205085884.544046,0.00000000000000D+000)
EigenSolve: 5 (480903841.387323,0.00000000000000D+000)

Anisotropic model.

Table 9.1: Computed frequencies.

step isotropic anisotropic

1 651.127 Hz 866.023 Hz
2 1104.673 Hz 1499.633 Hz
3 1588.959 Hz 2242.809 Hz
4 1672.210 Hz 2279.229 Hz
5 2536.249 Hz 3490.191 Hz

c CSC – IT Center for Science

9. Eigenvalue analysis of an elastic beam 48

Figure 9.2: Eigenvectors

c CSC – IT Center for Science

Tutorial 10

Elastic linear plate

Directory: ElasticPlateLinear
Solvers: SmitcSolver
Tools: ElmerGrid, editor
Dimensions: 2D

Case definition
This tutorial demonstrates how to use the Smitc solver to solve small deflections of plates. The Smitc solver
is for elastic linear plates and uses the theory of Reissner and Mindlin.
The case under investigation is a L-shaped steel plate under pressure. The plate is shown in figure 10.1
The longer sides have the length of 2 m and the shorter 1 m. So the area of the plate is 3 m2 . The plate has
a thickness of 1 cm. We assume that on the plate there is about 15300 kg of sand. The sand is uniformly
distributed on the plate and the sand stays uniformly distributed even if the plate undergoes small deflection.
The sand exerts to the plate a pressure of 50000 P a. The plate is clamped from all sides meaning that both
deflection and rotation are zero on all edges.

Figure 10.1: The geometry of plate and the numbering of edges.

Solution Procedure
The first thing to do is create a mesh with ElmerGrid. The definition of mesh is in the file simple_plate.grd.
The mesh is about uniform and consist of 1000 linear square elements. The mesh is created with command

49
10. Elastic linear plate 50

ElmerGrid 1 2 simple_plate
One thousand element should easily do the trick in this case but if more elements is needed you can edit the
file simple_plate.grd. More specifically the line
Surface Elements = 1000
The solver input file simple_plate.sif starts with turning on the warnings and the definition of the
proper mesh directory.
check keywords warn

Header
Mesh DB "." "simple_plate"
End
The simulation uses 2D cartesian geometry. The simulation is not time dependent i.e. Steady State. There
is no coupled solvers so only one iteration is needed. The output interval is one meaning all intervals (now
there is only one interval). Numerical results are written to file simple_plate.result and ElmerPost
file is simple_plate.ep.
Simulation
Coordinate System = Cartesian 2D
Simulation Type = Steady State
Steady State Max Iterations = 1
Output Intervals = 1
Output File = "simple_plate.result"
Post File = "simple_plate.ep"
End
There is just one body, the plate, and it uses Equation and Body Force 1 and is of Material 1.
Body 1
Equation = 1
Body Force = 1
Material = 1
End
The equation block is now more than easy. It only states that we use Solver 1 to solve the equation.
Equation 1
Active Solvers(1) = 1
End
In Body Force block we give the equations right hand side. It is the sands pressure and it is the same constant
in every point.
Body Force 1
Pressure = 5.0e4
End
In Material block we define the plates properties i.e. Poisson ratio, Young’s modulus and density. We also
give the plates thickness and possible pretension. Now there is no pretension.
Material 1
Poisson ratio = 0.3
Youngs modulus = 209e9
Density = 7800.0

Thickness = 1.0e-2
Tension = 0.0
End

c CSC – IT Center for Science

10. Elastic linear plate 51

Next the Solver block.

• First we define that we use SmitcSolver and give the name of the subroutine file Smitc and subroutine
name SmitcSolver.
• We name the variable Deflection and state that it has 3 degrees of freedom. First degree is the deflection
and the remaining two are actually the components of rotation vector.
• We don’t need eigen analysis nor is there any holes in the plate.
• We solve the matrix equation iteratively with stabilized biconjugate gradient method. We precondition
the iteration with incomplete LU-factorization.
• Tolerance for the matrix system is 1 · 10−8 and the tolerance should be achieved in less than 300
iteration.

Solver 1
Equation = "SmitcSolver"
Procedure = "Smitc" "SmitcSolver"

Variable = Deflection
Variable DOFs = 3

Eigen Analysis = False

Hole Correction = False

Linear System Solver = Iterative

Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU2
Linear System Convergence Tolerance = 1.0e-8
Linear System Max Iterations = 300
End

Finally we give the boundary conditions. The plate has 6 edges and the edge numbering is in figure 10.1.
All the edges are clamped i.e. no deflection (Deflection 1) and no rotation (Deflection 2 and 3).
Boundary Condition 1
Target Boundaries(6) = 1 2 3 4 5 6
Deflection 1 = 0.0
Deflection 2 = 0.0
Deflection 3 = 0.0
End

Results
The problem is solved in few seconds and the results are viewed with ElmerPost. It it possible to make
ElmerPost to show deflection in 3D. First we determine the number of nodes. Give commands
math tmp = size(Deflection.1)
math n = tmp(1)
to ElmerPost. Next we put the values of deflection to nodal z-values. Deflection is rather small so the values
are scaled by 50.
math nodes(2,0:n-1) = 50*Deflection.1
Result is shown in figure 10.2.
Deflection.2 and Deflection.3 are the x- and y-components of rotation vector. Values are transformed to
vector Rotation with commands

c CSC – IT Center for Science

10. Elastic linear plate 52

math Rotation = 0
math Rotation(0,0:n-1) = Deflection.2
math Rotation(1,0:n-1) = Deflection.3
math Rotation(2,0:n-1) = Deflection.2*0
The length of vector is calculated with

math AbsRotation = sqrt( vdot(Rotation,Rotation) )

Result is shown in figure 10.2.
It is rather cumbersome to write all the commands every time you solve the problem. It is possible to
write the commands to file. The file, let us name it Draw, would be

math tmp = size(Deflection.1);

math n = tmp(1);

math nodes(2,0:n-1) = 50*Deflection.1;

math Rotation=0;
math Rotation(0,0:n-1) = Deflection.2;
math Rotation(1,0:n-1) = Deflection.3;
math Rotation(2,0:n-1) = Deflection.2*0;

math AbsRotation = sqrt( vdot(Rotation,Rotation) );

display;
The file is executed in ElmerPost with command source Draw.

Figure 10.2: The deflection of the plate and the corresponding rotation.

c CSC – IT Center for Science

Tutorial 11

Compressible flow passing a step

Directory: FlowStepCompressible
Solvers: FlowSolve, HeatSolve
Tools: ElmerGrid, Editor
Dimensions: 2D, Steady-state

Case definition
This tutorial demonstrates how to simulate the compressible air flow passing a step. The whole step has
length of 1.4 m and the height of 0.2 m and the first part of it has length of 0.4 m and the height of 0.1 m
(Figure 11.1). The needed material parameters of air are shown in Table 11.1. The model has three sets of
boundary conditions. The air flows into the step from the inlet region and withdraws from the outlet region.
The other edges of the step compose the third boundary. The flowing air is considered as an ideal gas in this
case, and its density ρ depends on the pressure p and temperature T through the equation of state
p
ρ= ,
RT
where R is the gas constant.

Table 11.1: Material parameters.

parameter value
viscosity 16.7e-6 Ns/m2
heat conductivity 0.026 W/(m·K)
heat capacity 1.01e3 J/(kg·K)
specific heat ratio 1.4
reference pressure 1e5 Pa

Solution procedure
The mesh consists of 500 rectangular elements and it is constructed using ElmerGrid with the following
command

ElmerGrid 1 2 mesh.grd

This command creates the subdirectory mesh which contains the Elmer mesh files.

53
11. Compressible flow passing a step 54

Figure 11.1: Step.

Header
Mesh DB "." "mesh"
Include Path ""
Results Directory ""
End

The simulation uses 2D cartesian geometry and the problem is solved in steady state using no more than
twenty steady state iterations.

Simulation
Coordinate System = Cartesian 2D
Coordinate Mapping(3) = 1 2 3
Simulation Type = Steady
Steady State Max Iterations = 20
Solver Input File = "compress_step.sif"
Post File = "compress_step.ep"
Output File = "compress_step.dat"
End

The solvers are coupled and therefore the convection is computed.

Equation 1
Navier-Stokes = True
Heat Equation = True
Convection = "Computed"
End

Due to the simplicity of the model only one body is needed.

Body 1
Equation = 1
Material = 1
Initial Condition = 1
End

c CSC – IT Center for Science

11. Compressible flow passing a step 55

Our intention is to model compressible flow and that is why we have to set the value ”Perfect Gas” for the
keyword Compressibility Model. Furthermore, because perfect gas model has been chosen the
settings Reference Pressure and Specific Heat Ratio must also be given. The
Navier-Stokes equation also needs the value of viscosity and the heat equation needs the values of heat
capacity and heat conductivity.

Material 1
Compressibility Model = String "Perfect Gas"
Reference Pressure = 1e5
Specific Heat Ratio = 1.4
Viscosity = 16.7e-6
Heat Conductivity = 0.026
Heat Capacity = 1.01e3
End

For the initial value of temperature we have chosen 300 K.

Initial Condition 1
Temperature = 300
End

The Navier-Stokes equation is solved first. Here we give the linear system solver and convergence
criterions for linear, nonlinear and steady state solution of the Navier-stokes equation. Note that we are
solving for the compressible Navier-stokes equation and that is why a bubble function formulation is used
for stabilization of the equation.

Solver 1
Equation = "Navier-Stokes"
Linear System Solver = "Iterative"
Linear System Iterative Method = "BiCGStab"
Linear System Max Iterations = 500
Linear System Convergence Tolerance = 1.0e-08
Linear System Abort Not Converged = True
Linear System Preconditioning = "ILU2"
Linear System Residual Output = 1
Steady State Convergence Tolerance = 1.0e-05
Bubbles = Logical True
Nonlinear System Convergence Tolerance = 1.0e-05
Nonlinear System Max Iterations = 1
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-02
Nonlinear System Relaxation Factor = 1
End

The corresponding parameters for the solver of the heat equation are defined in the following solver section.

Solver 2
Equation = "Heat Equation"
Variable = "Temperature"
Linear System Solver = "Iterative"
Linear System Iterative Method = "BiCGStab"
Linear System Max Iterations = 350
Linear System Convergence Tolerance = 1.0e-08
Linear System Preconditioning = "ILU0"
Linear System Residual Output = 1
Steady State Convergence Tolerance = 1.0e-05

c CSC – IT Center for Science

11. Compressible flow passing a step 56

Bubbles = Logical True

Nonlinear System Convergence Tolerance = 1.0e-05
Nonlinear System Max Iterations = 1
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-02
Nonlinear System Relaxation Factor = 1
End

Finally, the boundary conditions are specified. There are three sets of boundary conditions, so three
Boundary Condition sections are needed. The first one is used to prescribe the boundary conditions
in the inlet region. Note that we have defined the x-velocity and temperature as a variable of y-coordinate.
This is done by setting different values for the x-velocity and temperature (the numerical values of the
second column between the words Real and End) in the different y-points (the numerical values of the
first column between words Real and End) of the inlet region. This kind of procedure prevents occuring
singularities in the corner points of the inlet region. In addition, this kind of definition is more realistic than
a condition, inwhich the values of the x-velocity and temperature remain the same in the whole inlet region.

Boundary Condition 1
Target Boundaries = 1
Velocity 1 = Variable Coordinate 2
Real
0.1 0
0.15 0.02
0.2 0
End

Velocity 2 = 0
Temperature = Variable Coordinate 2
Real
0.1 300
0.15 350
0.2 300
End
End

After the rest boundary conditions have been defined the problem is ready to solve.

Boundary Condition 2
Target Boundaries = 2
Velocity 2 = 0
End

Boundary Condition 3
Target Boundaries = 3
Velocity 1 = 0
Velocity 2 = 0
Temperature = 300
End

Results
Figure 11.2 presents the temperature distribution of the step in steady state. The maximum and minimum
values of x- and y-velocities are also given as a result and they are shown in Table 11.2.

c CSC – IT Center for Science

11. Compressible flow passing a step 57

Figure 11.2: Step.

Table 11.2: Computed velocities.

velocity value
min x-velocity -0.0014 m/s
min y-velocity -0.0016 m/s
max y-velocity 0.0008 m/s

c CSC – IT Center for Science

Tutorial 12

Flow through a hole – determining the

acoustic impedance

Directory: FlowResistance
Solvers: FlowSolve
Tools: ElmerGrid, editor
Dimensions: 3D, Steady-state

Case definition
The problem at hand consists of finding the resistance that a fluid faces as it is forced to flow through a
hole. The flow resistance is stated by the ratio of pressure drop over the hole and the input velocity. In
microsystem modeling, the hole resistance is often needed to analyse the gas damping effect in perforated
structures. Here, the contribution of the holes is homogenised over the perforated structure based on a single
hole resistance. For homogenisation in Elmer, the specific acoustic impedance is used to represent the flow
resistance. Specific acoustic impedance zh is defined as
p F
zh = = , (12.1)
v vAh
where F is the net force due to gas on the moving surface, v is the velocity of the gas on the same surface,
and Ah is the area of the moving surface. The calculation is best performed in a unit cell of the geometry.
In order to the homogenisation to be possible, the dependence of input velocity and the net force should
be linear. Further, there should not be a phase change between these two quantities. These conditions are
satisfied when the flow is incompressible. In a linear case, the fluid flow can be treated with the linear form
of Navier-Stokes equations called the Stokes equation
∂~u
ρ − ∇ · (2ηε) + ∇p = ρf~, (12.2)
∂t
where ~u is the unknown velocity field, p is the pressure, η is the viscosity of the fluid, ρf~ is a body force
and ε is the linearised strain tensor. Note, that the stationary version of the above equation can be used in
homogenisation calculations.
The condition for Stokes equation to apply is that the Reynolds number Re of the problem should be
small
ρU L
Re = , (12.3)
η
where ρ is density of the fluid and U and L are, respectively, the velocity and length scales of the problem.
The issue of compressibility is more difficult to answer. A classical condition for the compressibility is
that the Mach number M a of the problem should be small
U
Ma = < 0.3, (12.4)
a

58
12. Flow through a hole – determining the acoustic impedance 59

where a is the speed of sound in the gas in operating conditions and the value 0.3 is often stated limit
for a small Mach number (actually, the condition is that M a2 has to be small). Also the frequency and
amplitude of the oscillations of the system have an effect on the validity of the linearity and incompressibility
assumptions, since they affect the velocity scale of the problem.
However, also other factors have an effect on the compressibility of the gas. In microsystems, the viscous
effects on pressure, or even temperature changes, can affect the density of the gas. A condition for viscous
pressure changes is that M a2 /Re has to be small, and for temperature, in addition, that the Prandtl number
P r may not be too large
ηcp
Pr = , (12.5)
k
where cp is the heat capacity (ie. specific heat) in constant pressure and k is the thermal conductivity.
The conditions listed here for the flow to be approximately incompressible are only an overview and
the validity of incompressibility assumption should be considered in each case separately. In microsystems,
refer for example to the article M. Gad-el-Hak, J. Fluids Eng., 121, 5–33, 1999. Additionally, it is advisable
to perform numerical checks on the issue.
One final point on the applicability of the Stokes (or Navier-Stokes) equations is the effect of gas rarefi-
cation. If the dimensions of the problem are very small the continuity assumption may not be valid anymore.
The importance of the gas rarefication effects are given by the Knudsen number Kn
L
Kn = , (12.6)
L
where L is the mean free path of the gas molecules. The mean free path depends inversely on ambient
pressure, which has to take into account in stating the Knudsen number. For Knudsen numbers close to and
less than 1, slip boundary conditions should be used.
To summarise, the motivation of this tutorial is to perform a linear incompressible simulation of fluid
flowing through a hole. The wake for the flow is a constant velocity boundary condition for a boundary
before the hole. On the same boundary, the force caused by the fluid is recorded. These two quantities can
then be used to determine the specific acoustic impedance of a single hole. The constant velocity boundary
condition may be interpreted as describing a moving wall with small displacement. In this particular tutorial,
a symmetrical quadrant of a square-shaped hole is used.

Solution procedure
The solution for the problem is found by solving iteratively the Stokes equation. Nonlinear iterations are not
needed, since the problem is linear.
The computational mesh should include enough free space after the hole so that any artificial effects due
to the boundaries of the mesh are avoided. In this tutorial, the geometry is created and meshed using the
ElmerGrid program by the command elmergrid 1 2 hole.grd. The default mesh consists of about
12000 nodes and 10500 eight-noded hexahedrons.
The header section of solver input file includes only the location of the mesh files.
Header
Mesh DB "." "hole"
End
In the simulation section, a steady-state three-dimensional analysis is defined.
Simulation
Coordinate System = Cartesian 3D
Simulation Type = Steady State
Steady State Max Iterations = 1
Output File = "flow.result"
Post File = "flow.ep"
End
The geometry contains only one body and no body forces or initial conditions are present. The body
section reads thus as follows.

c CSC – IT Center for Science

12. Flow through a hole – determining the acoustic impedance 60

Body 1
Equation = 1
Material = 1
End
For solving the flow patterns the Navier-Stokes solver is used but the nonlinearity through convection is
switched off in the equation block. Also, solvers for the fluidic force and saving data are enabled.
Equation 1
Active Solvers(3) = Integer 1 2 3
NS Convect = False
End
Just a single iteration of the Navier-Stokes solver is needed, since the equation is linear. This can be
verified by switching the number of nonlinear iterations to a value more than one, and observing the change
in solution between iteration steps.
Solver 1
Equation = Navier-Stokes
Variable = Flow Solution
Variable DOFs = 3
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU0
Linear System Max Iterations = 200
Linear System Convergence Tolerance = 1.0e-08
Stabilize = True
Nonlinear System Convergence Tolerance = 1.0e-05
Nonlinear System Max Iterations = 1
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-08
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0e-05
End
The fluidic force solver needs to be run only once, after the flow solution is finished. With the keyword
Calculate Viscous Force it is possible to define whether the viscous forces of the fluid are included
in the force or not. If this is set to false, only the pressure integral is calculated.
Solver 2
Exec Solver = After All
Equation = Fluidic Force
Procedure ="FluidicForce" "ForceCompute"
Calculate Viscous Force = True
End
The final solver is used to save data from the analysis. With the following definitions, the input velocity
and the net force on the input boundary as well as the area of the boundary are written into a file called
flowdata.dat.
Solver 3
Exec Solver = After All
Equation = SaveScalars
Procedure = "SaveData" "SaveScalars"
Filename = "flowdata.dat"
Save Variable 1 = Velocity 3
Save Coordinates(1,2) = 0.0 0.0
End

c CSC – IT Center for Science

12. Flow through a hole – determining the acoustic impedance 61

The fluid is defined to be air. Note the Elmer MEMS units used.
Material 1
Name = Air
Density = 1.293e-12
Viscosity = 1.67e-5
End
Finally, the boundary conditions. BC 1 defines the input boundary, where also the fluidic force is calcu-
lated. BCs 2 and 4 are define the symmetry boundaries, BC 3 defines the no-slip conditions for the walls,
and BC 5 defines an open boundary.

Boundary Condition 1
Target Boundaries = 4
Velocity 1 = 0.0
Velocity 2 = 0.0
Velocity 3 = 1.0e3
Calculate Fluidic Force = True
End

Boundary Condition 2
Target Boundaries(2) = 8 10
Velocity 2 = 0.0
End

Boundary Condition 3
Target Boundaries(4) = 1 2 3 7
Velocity 1 = 0.0
Velocity 2 = 0.0
Velocity 3 = 0.0
End

Boundary Condition 4
Target Boundaries(2) = 6 9
Velocity 1 = 0.0
End

Boundary Condition 5
Target Boundaries = 5
Pressure = 0.0
End

Slip boundary conditions

The same simulation can also be performed using slip boundary conditions. These are appropriate, as stated
in introduction, when the Knudsen number is between 10−3 and 1. The slip boundary condition implemented
in Elmer is of first order
S · ~u = σ · ~n, (12.7)
where S is a vector containing the slip coefficients si for each velocity component, µ is the viscosity, and σ
is the stress tensor. For Newtonian fluids and for tangential directions of the boundary this gives

∂ui
si ui = µ , (12.8)
∂n
where si and ui refer to the same tangential component of the slip coefficient and the flow velocity.

c CSC – IT Center for Science

12. Flow through a hole – determining the acoustic impedance 62

The value of the slip coefficient is related to the mean free path of the gas molecules λ. For example,
Maxwell’s first order slip boundary condition may be used (as in e.g. A. Beskok, Num. Heat Transfer, B, 40,
451–471, 2001):
2 − σv ∂ui
ui = λ , (12.9)
σv ∂n
where σv is the tangential momentum accommodation coefficient, which models the momentum exchange
of gas molecules and the surface. The accommodation coefficient is dependent on the gas and on the surface,
and recent measurements give a result of σv ' 0.80 for various monoatomic gases such as Argon in contact
with prime Silicon crystal.
The slip coefficient of Elmer can finally be written as
µ σv
si = . (12.10)
λ 2 − σv
The mean free path is defined as r
µ πM
λ= (12.11)
ρ 2RT ,
where ρ is density, M is the molar mass, T is the temperature, and R = 8.3145 J/mol K is the molar gas
constant.
In the Elmer analysis, only a few changes in the sif-file are needed to make the slip conditions active. The
flow force boundary conditions have to be turned on and the numerical value of the slip coefficient has to be
defined on each boundary (here s =2e-4 is used for air). Further below is a list of the Boundary Condition
blocks. Note that there are more BCs than in the no-slip simulation, since a separate condition is needed for
surfaces oriented differently in space.
Generally, a normal-tangential orientation scheme for the boundary conditions are needed, since the
surfaces are not necessarily having a normal vector pointing in one of the coordinate directions. This would
be done for each such boundary by the line
Normal-Tangential Velocity = True
after which the Velocity component 1 points to the normal direction and the other components to the tangen-
tial directions.

! Definitions for slip boundary conditions:

Boundary Condition 1
Target Boundaries = 4
Flow Force BC = True
Slip Coefficient 1 = 2e-4
Slip Coefficient 2 = 2e-4
Velocity 3 = 2.0e3
Calculate Fluidic Force = True
End

Boundary Condition 2
Target Boundaries(2) = 8 10
Velocity 2 = 0.0
End

Boundary Condition 3
Target Boundaries(2) = 2 3
Flow Force BC = True
Velocity 3 = 0.0
Slip Coefficient 1 = 2e-4
Slip Coefficient 2 = 2e-4
End

c CSC – IT Center for Science

12. Flow through a hole – determining the acoustic impedance 63

Boundary Condition 4
Target Boundaries(2) = 6 9
Velocity 1 = 0.0
End

Boundary Condition 5
Target Boundaries = 5
Pressure = 0.0
End

Boundary Condition 6
Target Boundaries = 1
Flow Force BC = True
Velocity 1 = 0.0
Slip Coefficient 2 = 2e-4
Slip Coefficient 3 = 2e-4
End

Boundary Condition 7
Target Boundaries = 7
Flow Force BC = True
Velocity 2 = 0.0
Slip Coefficient 1 = 2e-4
Slip Coefficient 3 = 2e-4
End

Results
The computation takes about 200 cpu seconds on an AlphaServer with 1 GHz central processor when trilinear
elements are used. The results for two different input velocities taken from the file flowdata.dat are
summarised in Table 12.1. Also the specific acoustic impedance zh is calculated in the table. The results of
slip and no-slip simulations are also compared. Note that for the force, only the component perpendicular to
the surface should be used since the other components cancel out due to symmetry. The values in the table
are again given in Elmer MEMS units.

Table 12.1: Results of flow simulations for two input velocities

v slip model Fz zh
1.0 · 103 no-slip 36.13 1.45 · 10−3
2.0 · 103 no-slip 72.25 1.45 · 10−3
1.0 · 103 slip 29.30 1.17 · 10−3
2.0 · 103 slip 58.60 1.17 · 10−3

The identical values obtained for the spesific acoustic impedance in Table 12.1 prove by no means that
the flow in reality is linear, since this was the assumption and the simulation performed can and should not
reveal any nonlinear behavior. The results indicate, though, that allowing slip on the boundaries reduces the
resistance that the fluid faces. This example shows that in microsystems, as the dimension of the smallest
flow channel is in the range of a micrometer, it is reasonable to use slip boundary conditions for the velocity.
Finally, a picture of the results obtained with no-slip conditions is presented. The Fig. 12.1 shows a lot
of pressure isosurfaces which are coloured using the absolute value of the velocity.

c CSC – IT Center for Science

12. Flow through a hole – determining the acoustic impedance 64

Figure 12.1: The linear flow results.

c CSC – IT Center for Science

Tutorial 13

Electrostatics

Directory: Electrostatics
Solvers: StatElecSolve, ElectricForce
Tools: ElmerGrid, editor
Dimensions: 3D, Steady-state

Case definition
This case presents solving the Poisson equation for electric potential and calculating appropriate derived
quantities, such as capacitance, based on the result. The geometry studied is a symmetric quadrant of a plane
capacitor having a rectangular hole in another plate. A setting of this kind can be used to study the effects of
geometrical features on the capacitance and on the electrostatic force, which both are meaningful quantities
for coupled simulations in e.g. microsystems.

Solution procedure
The mesh is constructed using ElmerGrid with the following command
ElmerGrid 1 2 elmesh.grd
The mesh is extended above the hole to avoid undesired boundary effects. The geometry is presented in the
Figure 13.1

Figure 13.1: The geometry of problem.

65
13. Electrostatics 66

The simulation problem includes a single body, and thus one material and one equation set, as well as
three solvers. The solvers are used to compute the electric potential and related quantities, to calculate the
electric force, and to save relevant data into a file. This tutorial is defined in Elmer MEMS units. The sif-file
is presented below.

Check Keywords Warn

Header
Mesh DB "." "elmesh"
End

Only a single steady state iteration is needed, since the Poisson equation is linear.
Simulation
Coordinate System = Cartesian 3D
Simulation Type = Steady State
Steady State Max Iterations = 1
Output File = "elstatics.result"
Post File = "elstatics.ep"
End
The permittivity of vacuum has to be defined in the Constants section.
Constants
Permittivity Of Vacuum = 8.8542e-12
End

Body 1
Equation = 1
Material = 1
End
Electric energy density is added into the results in Equation section. This allows energy density to
be visualised in ElmerPost. Note also, that calculating electric flux (or the electric displacement field) is
disabled in the Solver 1 block. Further, the potential difference used in calculating the capacitance of the
system has to be defined in this section. This should be the same as the boundary conditions define for the
capacitance calculation to be sensible.
Equation 1
Active Solvers(2) = 1 2
Calculate Electric Energy = True ! (default False)
End

Solver 1
Equation = Stat Elec Solver
Variable = Potential
Variable DOFs = 1
Procedure = "StatElecSolve" "StatElecSolver"
Calculate Electric Field = True ! (default True)
Calculate Electric Flux = False ! (default True)
Potential Difference = 1.0e6
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Max Iterations = 200
Linear System Convergence Tolerance = 1.0e-07
Linear System Preconditioning = ILU1
Linear System ILUT Tolerance = 1.0e-03

c CSC – IT Center for Science

13. Electrostatics 67

Nonlinear System Max Iterations = 1

Nonlinear System Convergence Tolerance = 1.0e-4
Nonlinear System Newton After Tolerance = 1.0e-3
Nonlinear System Newton After Iterations = 10
Nonlinear System Relaxation Factor = 1
Steady State Convergence Tolerance = 1.0e-4
End
The static electric force solver does not need a lot of information:
Solver 2
Equation = Electric Force
Procedure = "ElectricForce" "StatElecForce"
End
Finally, some data is saved in file scalars.dat in working directory.
Solver 3
Exec Solver = After All
Equation = SaveScalars
Procedure = "SaveData" "SaveScalars"
Filename = "scalars.dat"
End

Only the relative permittivity of the material has to be defined.

Material 1
Relative Permittivity = 1
End
The boundary conditions include the values of electric potential (voltage) and indication on which bound-
ary the electric force should be calculated. On all the other boundaries a natural boundary condition is used,
basically stating that the electric flux through these boundaries is zero.
Boundary Condition 1
Target Boundaries = 4
Potential = 0.0
Calculate Electric Force = True
End

Boundary Condition 2
Target Boundaries = 3
Potential = 1.0e6
End

Results
The results obtained for capacitance and electric force are compared to those of a complete plane capacitor.
For a plane capacitor, the capacitance is
A
C = εr ε0 , (13.1)
d
and the electrostatic force is
1 A
Fe = εr ε0 2 Φ2 , (13.2)
2 d
where εr is the relative permittivity, ε0 is the permittivity of vacuum, A is the area of a capacitor plate, d is
the separation of the capacitor plates, and Φ is the potential difference between the plates.

c CSC – IT Center for Science

13. Electrostatics 68

Table 13.1: Comparison of numerical results to analytic values

simulation analytic ratio

Capacitance 2.1361 · 10−10 2.2136 · 10−10 0.965
Electric Force 1.0406 · 102 1.1068 · 102 0.940

The results of the simulation as well as the comparison to the complete plane capacitor values are shown
in Table 13.1 (in Elmer MEMS units). Note that the fringe fields on capacitor edges are not calculated. This
would require much larger mesh extending outside the capacitor boundaries.
Finally, a picture of the results is presented. The Figure 13.2 shows the isosurfaces of the electric po-
tential with the color marking the strength of the electric field. From the picture it is clearly seen that the
electric field is constant between the plates except for the proximity of the hole which causes weakening of
the field magnitude. There are also strong electric fields at the edges of the hole.

Figure 13.2: Isosurfaces of the potential coloured with electric field magnitude.

c CSC – IT Center for Science

Tutorial 14

Lossless acoustic waves

Directory: AcousticWaves
Solvers: HelmholtzSolve
Tools: ElmerFront
Dimensions: 2D, Harmonic

Introduction
Elmer provides two alternative ways of conducting acoustic analyses in the frequency domain. Firstly, one
may simply use the Helmholtz equation which is based on the assumption of lossless flow, i.e. the effects of
viscosity and heat conduction are assumed to be negligible. More refined analyses where these effects are
taken into account may be carried out by using the specific solver for the set of time-harmonic dissipative
acoustic equations. The aim of this tutorial is to demonstrate the usage of the solver for the basic Helmholtz
equation, which is frequently taken as the starting point in acoustic analyses.

Case description
In this problem the fluctuations of the pressure in an air-filled cavity shown in Figure 14.1 are considered.
The cavity is connected with the surrounding air by an open narrow pipe. The pressure fluctuations are
generated by a vibrating membrane on the boundary ΓS with the frequency of the motion being f = 100
Hz. The remaining parts of the boundary are assumed to be rigid walls. In addition, the effects of gravity are
assumed to be negligible.
Γ0
6
0.1
?
6
0.3 -
0.08
-

0.3

ΓS
?
0.25 - 0.15 -0.1-

Figure 14.1: The geometry of the cavity.

69
14. Lossless acoustic waves 70

Suitable boundary conditions in terms of the pressure must be given. On the rigid walls the pressure flux
is prescribed to vanish which corresponds to the assumption that there is no velocity in the direction normal
to the boundary. At the open end Γ0 the impedance boundary condition suitable for forward traveling plane
waves is given by setting Z = −c with c being the sound speed. We assume that c = 343 (m/s). Finally, the
wave source is given by defining a non-vanishing pressure flux on the corresponding part of the boundary.
We take simply ∇P · ~n = 1 where P is the (complex) amplitude of the pressure and ~n is the outward unit
normal to the boundary.

Solution procedure
• Before starting Elmer copy the geometry file (domain.egf) to the working directory and then launch
Elmer Front by giving the command

ElmerFront

• Open the geometry file by choosing Open Cad-file in the File menu. To enable browsing with the
mouse click the button on the right-hand side of the field where the file name may be written. Here
the correct Cad file type is Elmer. Give also the model name (for example helmholtz) and write
the path of the working directory in the Model directory field.
• Select the equation to be solved by selecting Equations in the Problem menu. Choose the Helmholtz
equation and press Add button.

• Define the angular frequency for the simulation by selecting Simulation parameters in the Problem
menu. Enter the value 628.3 to the field and accept the value by clicking OK button.
• Define the sound speed for the medium by selecting Materials in the Model menu. Enter the value 343
for the sound speed and press Add button.

• Prescribe the boundary conditions by selecting Boundary conditions in the Model menu. Select (with
the mouse) Boundary1 and give the value for the boundary flux:

Wave flux Re = 1
Wave flux Re = 0

Finally, press Add button. Then proceed to give the other boundary conditions in a similar manner
(the value for the pressure is prescribed).
• Create a finite element mesh by selecting Define mesh in the Mesh menu. To begin with give a name
for the mesh. Define the number of element edges on each boundary and then create the mesh by
pressing Generate Mesh button.
• The problem may now be solved by selecting Solver in the Run menu.
• After the solution is done, view the results by selecting the Postprocessor from the Run menu.

Run -> Postprocessor

• To save the created model, select Save model file from the File menu.

File -> Save model file

Results
Using a mesh consisting of 3900 (quadratic) elements with 7601 nodes the difference of the maximum and
the minimum value of the pressure is found to be ∆p ≈ 0.205

c CSC – IT Center for Science

Tutorial 15

Induction heating of a graphite crucible

Directory: InductionHeating
Solvers: StatMagSolve
Tools: ElmerGrid, editor
Dimensions: 2D, Axi-Symmetric

Case definition
At high temperatures the most practical method to heat up the crucible is by electromagnetic induction. The
induction coil generates an alternating current that flows through the crucible. The Ohmic resistance en-
countered by this current dissipates energy, thereby directly heating the crucible via internal heat generation.
The tutorial case is a simple axi-symmetric crucible that could be used, for example, to grow silicon
carbide (SiC) by the sublimation method. The crucible is made of dense graphite and isolated by porous
graphite. At the bottom of the crucible there is some SiC powder. The physical properties of the material are
given in Table 15.1. The dimensions of the induction heating crucible are given in Table 15.2. Additionally,
the powder thickness is 1.0 cm and there are 10 spirals in the coil. The frequency of induction heating f
is 50 kHz and the current I is 10 A. The permeability of the space is 4π10−7 if the other variables are in
SI-units.

Solution Procedure
At low frequencies the free charges may be neglected and the induction heating problem may be solved
in terms of an magnatic vector potential. The proper solver to do this is StatMagSolver. However,
the induction heating problem can only be modeled if the helicity of the coil is neglected and an avarage
current density is assumed. This current density may be computed easily when the area of the coil is known
j0 = nI/A, where A is the coil area.
The mesh for this problem may easily be created by ElmerGrid. The provided mesh is quite sufficient
for this case but for higher frequencies the mesh should be tuned to solve the thin boundary layers. The
computational mesh is created from file crucible.grd by the command
ElmerGrid 1 2 crucible

Table 15.1: Material parameters of the crucible

material ε κ [W/mk] σ (1/Ωm)

graphite 0.7 10.0 2.0E4
insulation 0.9 1.0 2.0E3
powder 0.5 25.0 1.0E4

71
15. Induction heating of a graphite crucible 72

Table 15.2: Dimensions of the crucible

body part rinner router hinner houter
graphite 2.0 2.5 6.0 8.0
insulation 2.5 4.0 8.0 12.0
coil 5.0 5.5 8.0

The mesh consists of 5 different bodies which need 4 different materials sets. Only on set of boundary
conditions are required for the external boundary. Thus the header information of the command file is as
follows
Header
Mesh DB "." "crucible"
Include Path ""
Results Directory ""
End
In the Simulation section the coordinate system and time dependendy is set, among other things. Also
we know that the equation is linear and therefore only one steady state iteration is requited. If the electric
properties depend on the magnitude of the field several iterations are required.
Simulation
Coordinate System = "Axi Symmetric"
Simulation Type = Steady State
Steady State Max Iterations = 1
Output File = "crucible.result"
Post File = "crucible.ep"
End
In the Constants section the permittivity of vacuum must be given.
Constants
Permittivity Of Vacuum = 8.8542e-12
End
In the differential equation for the magnetic vector potential the source the is the current density. Thus, it is
given in the Body Force section.
Body Force 1
Current Density = 2.5e5
End
In the Body section the different bodies are assigned with correct equation sets and material parameters, for
example
Body 3
Name = "Insulation"
Equation = 1
Material = 2
End
In the Equation block all the relavant solvers are set to active.
Equation
Name = "Vector Potential Equation"
Active Solvers = 1
End

c CSC – IT Center for Science

15. Induction heating of a graphite crucible 73

The only solver in this simple tutorial is the solver for the magnetic vector potential. Look for the relevant
model manual for information about the options. Here the equation is solved iteratively and the local Joule
heating and magnetic flux are computed as a postprocessing step. The Joule heating is scaled so that the total
heating power is 3.0 kW. This option may be used when the total heating efficiency is known. The nonlinear
solver parameters are not really needed as the material parameters are constant. Sometimes the parameters
may depend on the magnetic field and thus the nonlinear problem must be solved iteratively.
Solver 1
Equation = Potential Solver
Variable = Potential
Variable DOFs = 2

Angular Frequency = Real 50.0e3

Calculate Joule Heating = Logical True
Calculate Magnetic Flux = Logical True
Desired Heating = Real 3.0e3

Procedure = "StatMagSolve" "StatMagSolver"

Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Max Iterations = 300
Linear System Convergence Tolerance = 1.0e-10
Linear System Preconditioning = ILU1
Linear System ILUT Tolerance = 1.0e-03
Linear System Residual Output = 1
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-6
Nonlinear System Relaxation Factor = 1
Steady State Convergence Tolerance = 1.0e-6
End
In the Material sections all the necessary material parameters are given, for example
Material 2
Name = "Insulation"
Electric Conductivity = 2.0E3
End
The magnetic field must vanish at infinity. Unfortunately the computational domain is bounded and therefore
the infinite distance becomes very finite. A proper distance may be checked by gradually increasing it until
no change in the result occurs.

Boundary Condition 1
Target Boundaries = 1
Potential 1 = Real 0.0
Potential 2 = Real 0.0
End

Results
With the given computational mesh the problem is solved in a few seconds. With the 20 072 bilinear elements
the heating efficieny is 16.9 W. The corresponding results are shown in Fig. 15.1.

c CSC – IT Center for Science

15. Induction heating of a graphite crucible 74

Figure 15.1: Induction heating of a simple crucible. a) in-phase component of the vector potential b) out-of-
phase component of the vector potential c) Joule losses in the conductors

c CSC – IT Center for Science

Tutorial 16

Fluid flow around an elastic beam

Directory: FluiStructureBeam
Solvers: ElasticSolve, FlowSolve, MeshSolve
Tools: ElmerGrid, editor, Fortran 90 compiler
Dimensions: 2D, steady-state

Case definition
A homogenous, elastic beam (Ω2 ) is in a fluid flow (which takes place in the region Ω1 ), see figure 16.1. The
beam is 5.0 m long and it is rigidly supported on the end Γ6 . At the incoming boundary Γ1 , the fluid velocity
in the x-direction is a parabola which reaches its maximum value 1 m/s at the point y=5.0 m. At the incoming
and outcoming boundaries Γ1 and Γ2 velocities in the y-direction are 0 m/s. At the outcoming boundary the
pressure is 0 Pa. The velocities on boundaries Γ3 , Γ4 and Γ5 are 0 m/s. The fluid is incompressible and
has density 1 kg/m, dynamic viscosity 0.5 kg/ms and Poisson ratio 0.3. Material properties for the beam
are the density 1000 kg/m, Poisson ratio 0.3 and Young’s modulus 1.0e3 N/m. The problem is to solve the
maximum displacement of the beam and the pressure change in Ω1 .

Γ4

Γ3
Γ1 u1,x Ω1 Γ2

y Ω2

x
Γ6 Γ5

Figure 16.1: Geometry of the problem

The flow generates normal and shear stresses to the beam boundary Γ3 . These forces deform the beam
and this changes the flow. This coupled problem can be modelled using the Navier-Stokes and elasticity
equations.
For the incompressible fluid the Navier-Stokes equations can be written as

−∇ · (2µε) + ρ~u · ∇~u + ∇p = 0 in Ω1

∇ · ~u = 0 in Ω1
~ux = (10y − y 2 )/25 on Γ1 (16.1)
~ux = 0 on Γi , i = 3, 4, 5
~uy = 0 on Γi , i = 1, . . . , 5,

75
16. Fluid flow around an elastic beam 76

where µ is the dynamic viscosity, ε is the strain tensor, ρ is the density, ~u is the velocity vector and p is the
pressure. It is assumed that the density and viscosity are constants.
For the homogeneous, elastic beam the elasticity equations are

−div[(I + ∇u)Σ] = 0 in Ω2
Σ = λ(trE)I + 2µE in Ω2
E = 12 (∇uT + ∇u + ∇uT ∇u) in Ω2 (16.2)
u = 0 on Γ6
(I + ∇u)Σn = q on Γ3

where u is the displacement vector, Σ is the second Piola-Kirchhoff stress tensor, λ and µ are the Lamé
constants and E is the Green-St Venant strain tensor, q is the surface traction and n is the unit normal to the
beam boundary. The edge load q is calculated from the stresses obtained using the Navier-Stokes equations.
For updating the fluids mesh a linear elasticity equation is used. In mathematical form it can be written
as,
−div{λtr[ε(v)]I + 2µε(v)} = 0 in Ω1
v = u|Γ3 on Γ3 (16.3)
v=0 on Γi , i = 1, 2, 4, 5
where λ and µ are the Lamé constants, ε is linear strain tensor and v is the displacement vector. Here u is
the solution to the elasticity equations.

Solution procedure
In this problem geometry and mesh is created by using ElmerGrid by the command
ElmerGrid 1 2 fsi.grd
In this problem, external Fortran functions are used to express a parabolic velocity inflow profile and to
define a variable stiffness for the deforming mesh. The stiffness of the mesh is controlled by the Youngs
modulus. In the case of deforming mesh, the absolute value of the parameter is not important but its changes
over the domain are influential. The following function defines the mesh to be stiffer near the corner of the
beam where the mesh is deformed most. The idea is to avoid too irregularly shaped elements in that area.

FUNCTION Youngs( Model, n, x ) RESULT( s )

USE Types
IMPLICIT NONE

TYPE(Model_t) :: Model
INTEGER :: n
REAL(KIND=dp) :: x,s,xx,yy

xx = Model % Nodes % x(n)

yy = Model % Nodes % y(n)

s = 1.0d0 / SQRT( (xx-11.0)2 + (yy-4.9)2 )

END FUNCTION Youngs

The following function is used to give the inflow boundary condition.

FUNCTION InFlow( Model, n, x ) RESULT( vin )

USE Types
IMPLICIT NONE

TYPE(Model_t) :: Model
INTEGER :: n

c CSC – IT Center for Science

16. Fluid flow around an elastic beam 77

REAL(KIND=dp) :: yy,xx,x,vin,v0,vt

xx = Model % Nodes % x(n)

yy = Model % Nodes % y(n)

v0 = (-(yy**2)+10*yy)/25

IF(x < 8.0) THEN

vt = x/8.0
ELSE
vt = 1.0
END IF

vin = v0*vt

END FUNCTION InFlow

In general, the user may supply his/hers own functions for defining general dependencies of variables or
parameters. The function header should read as follows

FUNCTION FunctionName( Model, n, x ) RESULT( res )

where Model is a structure holding various informationăof the problem and n is the number of the node
being considered. These are automatically given by the program. The paramter x is the value of the variable,
which user defines to be given for the function as input. The variable is defined in the sif file. Finally, a the
function returns the result res for the parameter in consideration.
If the Elmer environment is succesfully setup the compilation command should look like

elmerf90 -o FsiStuff FsiStuff.f90

The functions may be compiled also locally on a PC either for Linux or WindowsNT, provided that a
Fortran90 compiler and the Elmer software are installed.
The next step is to edit the solver input file (sif file) with a text editor, e.g. Emacs.

• The Header section, where the mesh is defined.

Header
Mesh DB "." "fsi"
End

• The Constants section need not contain any constant this time.

Constants
End

• The Simulation section gives the control data for the case.

Simulation
Coordinate System = Cartesian 2D
Simulation Type = Steady State
Steady State Max Iterations = 50
Steady State Min Iterations = 2
Output Intervals = 1
Output File = "fsi.result"
Post File = "fsi.ep"
End

c CSC – IT Center for Science

16. Fluid flow around an elastic beam 78

• The Body section is used to define which equations and materials are used for the simulation bodies.
(Ω2 ).

Body 1
Equation = 1
Material = 1
End

Body 2
Equation = 2
Material = 2
End

• The Material section defines the material properties of each body. Also the procedure Youngs is
used.

Material 1
Density = 1.0
Viscosity = 0.5
Poisson Ratio = 0.3
Youngs Modulus = Variable time
Real Procedure "FsiStuff" "Youngs"
End

Material 2
Density = 1000
Youngs Modulus = 1.0e3
Poisson Ratio = 0.3
End

• The Solver section defines various control parameters for each solver.

Solver 1
Equation = Navier-Stokes
Stabilize = True
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU1
Linear System Max Iterations = 500
Linear System Convergence Tolerance = 1.0e-8
Nonlinear System Max Iterations = 10
Nonlinear System Convergence Tolerance = 1.0e-5
Nonlinear System Newton After Tolerance = 1.0e-5
Nonlinear System Newton After Iterations = 20
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0e-4
End

Solver 2
Equation = Elasticity Solver
Variable = Displacement
Variable DOFs = 2
Procedure = "ElasticSolve" "ElasticSolver"
Linear System Solver = Direct
Linear System Direct Method = Banded

c CSC – IT Center for Science

16. Fluid flow around an elastic beam 79

Nonlinear System Newton After Tolerance = 1.0e-3

Nonlinear System Newton After Iterations = 20
Nonlinear System Max Iterations = 100
Nonlinear System Convergence Tolerance = 1.0e-5
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0e-4
End

Solver 3
Equation = Mesh Update
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU1
Linear System Max Iterations = 500
Linear System Convergence Tolerance = 1.0e-8
Steady State Convergence Tolerance = 1.0e-4
End

• The Equation section. This section defines the equations for a body. Plane stress model for the
elastic beam is applied. Setting this keyword to False enables the plane strain assumption.

Equation 1
Active Solvers(2) = 1 3
End

Equation 2
Active Solvers = 2
Plane Stress = True
End

• In the Boundary Condition section different boundary conditions are defined. In Boundary
Condition 1 the shape of inflow is defined with the Fortran procedure InFlow. Note also the bc 3,
where keyword FSI BC is supplied to define the boundary on which the fluid affects a force on the
elastic beam. Also, on the same boundary, the mesh is constraint to follow the deformation of the
beam by the two Equals lines.

Boundary Condition 1
Target Boundaries = 1
Velocity 1 = Variable Time
Real Procedure "FsiStuff" "InFlow"

Velocity 2 = 0.0
Mesh Update 1 = 0.0
End

Boundary Condition 2
Target Boundaries = 2
Velocity 2 = 0.0
Pressure = 0.0
Mesh Update 1 = 0.0
End

Boundary Condition 3
Target Boundaries = 3
Velocity 1 = 0.0

c CSC – IT Center for Science

16. Fluid flow around an elastic beam 80

Velocity 2 = 0.0
FSI BC = True
Mesh Update 1 = Equals Displacement 1
Mesh Update 2 = Equals Displacement 2
End

Boundary Condition 4
Target Boundaries(2) = 4 5
Velocity 1 = 0.0
Velocity 2 = 0.0
Mesh Update 2 = 0.0
End

Boundary Condition 5
Target Boundaries = 6
Displacement 1 = 0.0
Displacement 2 = 0.0
End

Results
After the solution is done, view the results with ElmerPost. Read the result file fsi.ep in ElmerPost by
choosing Read Model File. The fsi.ep file is located in the same directory as the mesh files.

Table 16.1: Max. displacement and pressure difference

Elements max |u| [m] max ∆p [Pa]

540 1.34 6.36

As a result the maximum absolute value of displacement and the pressure difference is given (see ta-
ble 16.1).

c CSC – IT Center for Science

Tutorial 17

Thermal actuator driven with

electrostatic currents

Directory: ThermalActuator
Solvers: StatCurrentSolve, HeatSolve, StressSolve
Tools: ElmerGrid, editor
Dimensions: 3D, Steady-state

Case definition
The tutorial introduces a micro mechanical thermal actuator as shown in Fig. 17.1. A static electric current
is driven through the actuator. The power loss due to the resistance of the actuator is transformed into heat
which in turn causes thermal stresses into the structure. The electric current thus results in deformation
of the actuator. In industry, such an actuator might be used to control the position of a micromechanical
component.

Figure 17.1: The geometry of the actuator.

Solution procedure
The problem is solved by first iterating the electrostatic current solver and heat equation until both are
converged. The temperature distribution is then used as a load for stress analysis solver which calculates the
actual deformation of the structure. The electric conductivity of the actuator depends on the temperature and
thus the electrostatic - thermal problem is coupled in both directions.
The computational mesh for this particular tutorial is created by using Ansys software. The details of the
mesh are written into files called ExportMesh by a certain Ansys macro and converted to Elmer format by
the ElmerGrid program. The command to use is
ElmerGrid 4 2 ExportMesh -order 1.0 0.1 0.001 -o thermal

81
17. Thermal actuator driven with electrostatic currents 82

The above command reads in the Ansys mesh files, arranges the mesh nodes in a reasonable way and saves
the mesh in Elmer format in a directory called thermal.
The geometry of the problem includes only one body and material. Boundary conditions are defined on
the actuator legs, which are kept at constant electric potential, temperature and position. Thus, only Dirichlet
boundary conditions are used.
The header and simulation blocks of the solver input file are

Header
Mesh DB "." "thermal"
End

Simulation
Coordinate System = Cartesian 3D
Simulation Type = Steady State
Steady State Max Iterations = 30
Output Intervals = 1
Output File = "actuator.result"
Post File = "actuator.ep"
End

An initial condition for temperature is defined in order to ease the convergence of the iterative solvers.
Also, a body force for the heat equation solver defining the Joule heating is needed. These both have to be
declared in the body section as follows:

Body 1
Equation = 1
Material = 1
Initial Condition = 1
Body Force = 1
End

The solution procedure requires the use of three solvers: Static current solver, heat equation solver and
the stress analysis solver. The equation block below defines that these solvers are used.

Equation 1
Active Solvers(3) = Integer 1 2 3
Calculate Joule Heating = True
End

The solver blocks define the parameters of the respecting solvers. The static current conduction problem
is tackled by an iterative conjugate gradient method (CG). For heat equation, a stabilized biconjugate gradient
method is used. The coupled problem of these two solvers is difficult since the static current calculated heats
the structure on each step, and the rise of temperature makes the current conduction more and more difficult.
To overcome this problem, a relaxation factor of 0.5 is defined for the heat equation solver.

Solver 1
Equation = Stat Current Solver
Procedure = "StatCurrentSolve" "StatCurrentSolver"
Variable = Potential
Variable DOFs = 1
Calculate Volume Current = True
Calculate Electric Conductivity = True
Linear System Solver = Iterative
Linear System Iterative Method = CG
Linear System Preconditioning = ILU3
Linear System Max Iterations = 300

c CSC – IT Center for Science

17. Thermal actuator driven with electrostatic currents 83

Linear System Convergence Tolerance = 1.0e-8

Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0-6
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-12
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0e-6
End

Solver 2
Equation = Heat Equation
Variable = Temperature
Variable DOFs = 1
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU1
Linear System Max Iterations = 350
Linear System Convergence Tolerance = 1.0e-9
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-07
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-12
Nonlinear System Relaxation Factor = 0.5
Steady State Convergence Tolerance = 1.0e-07
End

For stress analysis, a direct solver is used instead of an iterative solver. It is often difficult for the
iterative solver to find a solution for a structure that contains parts with varying stiffness properties, which
is obviously the case here (try the iterative solver and see!). The stress analysis solver is called first only
after the coupled iteration of two previous solvers is complete. This is possible since the deformation of the
structure is so small that it does not change the current density distribution. Defining stress analysis this way
saves computational time. It is possible to iterate all the three solvers until convergence by commenting the
Exec Solver line.

Solver 3
Exec Solver = After All
Equation = Stress Analysis
Variable = Displacement
Variable DOFs = 3
Linear System Solver = Direct
Linear System Direct Method = Banded
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-6
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-12
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0e-6
End

The material of the structure has a temperature dependent electric conductivity. This, as well as other
material parameters, is defined in the material block. Note that a MEMS unit system is used.

Material 1
Electric Conductivity = Variable Temperature
Real

c CSC – IT Center for Science

17. Thermal actuator driven with electrostatic currents 84

298.0 4.3478e10
498.0 1.2043e10
698.0 5.1781e9
898.0 2.7582e9
1098.0 1.6684e9
1298.0 1.0981e9
1683.0 1.0
2000.0 1.0
End

Density = 2.3e-15
Heat Conductivity = 32.0e6
Youngs Modulus = 169.0e3
Poisson Ratio = 0.22
Heat Expansion Coefficient = 2.9e-6
Reference Temperature = 298.0
End

Finally, the initial condition, thermal heat load for stress analysis, and the boundary conditions are de-
fined.

Initial Condition 1
Temperature = 298.0
End

Body Force 1
Heat Source = Equals Joule Heating
End

Boundary Condition 1
Target Boundaries = 1
Potential = 0
Temperature = 298
Displacement 1 = 0.0
Displacement 2 = 0.0
Displacement 3 = 0.0
End

Boundary Condition 2
Target Boundaries = 2
Potential = 7
Temperature = 298
Displacement 1 = 0.0
Displacement 2 = 0.0
Displacement 3 = 0.0
End

Results
The problem converges after 27 steady state iterations on the tolerance limits defined above. The calcu-
lation takes about 180 cpu seconds of which 40 cpus is spent in solving the stress analysis equation. The
calculations were performed on a Compaq Alpha Server with a 1 GHz central processor.
Result for temperature distribution and the displacement are shown in Figs 17.2 and 17.3. The temper-
ature rises unrealistically high in this example because all heat transfer mechanisms out of the structure are

c CSC – IT Center for Science

17. Thermal actuator driven with electrostatic currents 85

Figure 17.2: Temperature distribution.

Figure 17.3: The displacement of the actuator.

neglected. Presumambly at least the heat radiation is of major importance in this case. For displacement, the
results show a movement of about 3.3 micrometers for the actuator tip.

c CSC – IT Center for Science

Tutorial 18

Axisymmetric coating process

Solvers: FlowSolve, FreeSurfaceReduced

Tools: ElmerGrid, editor
Dimensions: 2D, Steady-state

Case definition
The optical fibers are quite fragile and must therefore be coated with a layer of polymer before they are
stored. This means that the coating process must be done with the same speed as the drawing of optical
fibers. When the diameter of the fiber is only 125 µm this sets high demands for the coating geometry since
it must provide even coating at high draw speeds. In Elmer a tailored free surface boundary condition allows
an efficient solution of this particular problem.

Solution procedure
The mesh is done with ElmerGrid in the directory coat by the command
ElmerGrid 1 2 coat.grd
Therefore the header reads
Header
Mesh DB "." "coat"
End
The geometry is axisymmetric and the problem is solved in steady state. Typically around 10 iterations is
needed to solve the problem but to be on the safe side 30 is set as the maximum.
Simulation
Coordinate System = Axi Symmetric
Simulation Type = Steady State
Steady State Max Iterations = 30
Output Intervals = 1
Output File = "coat.result"
Post File = "coat.ep"
End
In this case there is only one body which comprises of the polymer floating between the coating cup and the
optical fiber.
Body 1
Equation = 1
Material = 1
End

86
18. Axisymmetric coating process 87

The presented solution used four different solvers. The Navier-Stokes solver is required to solve the flow
field for the polymer.
Solver 1
Equation = Navier-Stokes
Stabilize = True
Internal Move Boundary = Logical False
Nonlinear System Max Iterations = 5
Nonlinear System Convergence Tolerance = 1.0e-7
Nonlinear System Newton After Iterations = 2
Nonlinear System Newton After Tolerance = 1.0e-2
Nonlinear System Relaxation Factor = 0.7
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU1
Linear System Max Iterations = 100
Linear System Convergence Tolerance = 1.0e-10
Steady State Convergence Tolerance = 1.0e-7
End
A tailored free surface solver is used to find the position of the free surface with a given flow field. The
variable being solved is the displacement of the free surface. Relaxation is used to avoid over-shooting
during the itaration. This solver does not solve any matrix equations. Instead it solves the radius from the
mass conservation constraint for each node on the free surface separately. There is a possibility to do the
mapping also within the solver using a 1D scheme but this is disabled by setting the Perform Mapping
to be False.
Solver 2
Equation = "Free Surface Reduced"
Procedure = "FreeSurfaceReduced" "FreeSurfaceReduced"
Variable = Dx
Variable DOFs = 1
Nonlinear System Relaxation Factor = 0.7
Nonlinear System Convergence Tolerance = 1.0e-3
Steady State Convergence Tolerance = 1.0e-3
Perform Mapping = Logical False
End
The mesh update solver is required to map the computational mesh so that it corresponds to the altered
geometry. Here the displacements of the free surface have already been computed and this solver solves the
displacements inside the domain. Note that solvers 1, 2 and 3 are coupled and therefore the system must be
solved iteratively
Solver 3
Equation = Mesh Update
Linear System Solver = Iterative
Linear System Iterative Method = BiCGSTAB
Linear System Preconditioning = ILU
Linear System Convergence Tolerance = 1.0e-12
Linear System Max Iterations = 200
Linear System Symmetric = True
Steady State Convergence Tolerance = 1.0e-4
End

In the end, an additional solver is used to compute the forces acting on the fiber. This does not affect the
results.

c CSC – IT Center for Science

18. Axisymmetric coating process 88

Solver 4
Equation = Fluidic Force
Procedure = "FluidicForce" "ForceCompute"
Calculate Viscous Force = Logical True
End

Addiationally there are two solvers for saving the results in a form that is more useful than plain pictures.
The SaveScalars saves the scalar values, such as the diameter and force values, and the SaveLine
saves the free surface.
Solver 5
Equation = SaveScalars
Procedure = "SaveData" "SaveScalars"
Filename = "scalars.dat"
End

Solver 6
Equation = SaveLine
Procedure = "SaveData" "SaveLine"
Filename = "kurvi.dat"
End
The equation includes only the solvers that need a permutation vector pointing out the active nodes. There-
fore the save utilities do not need to belong to the set of active solvers.

Equation 1
Active Solvers(4) = 1 2 3 4
End
The material parameters are those of the polymer. Additionally elasticity parameters are needed because the
solver that updates the mesh is actually a linear elasticity solver.
Material 1
Density = 1.0
Viscosity = 1.0
Poisson Ratio = 0.3
Youngs Modulus = 1.0
End
Five different boundary conditions are needed. The origin is a symmetry axis and thefore the radial
velocity is set to zero. The axial velovity is the draw velocity.
Boundary Condition 1
Name = "Symmetry"
Target Boundaries = 1
Velocity 2 = -10.0 ! The draw velocity
Velocity 1 = 0.0
Compute Fluidic Force = Logical True
Mesh Update 1 = 0.0
End
The free surface has a condition stating that the reduced order free surface solver should be solved for that.
Additionally the free surface is a boundary condition for the mesh update, and a line to be saved.
Boundary Condition 2
Name = "Free"
Target Boundaries = 2
Mesh Update 1 = Equals Dx

c CSC – IT Center for Science

18. Axisymmetric coating process 89

Mesh Update 2 = 0.0

Free Surface Reduced = Logical True
Save Line = Logical True
End
At the outlet the radial velocity should vanish and the axial coordinate should be fixed.

Boundary Condition 3
Name = "Outlet"
Target Boundaries = 3
Velocity 1 = 0.0
Mesh Update 2 = 0.0
End
At the inlet it is assumed that there is no radial velocity and that the pressure acting on the surface is zero.
Boundary Condition 4
Name = "Inlet"
Target Boundaries = 4
Velocity 1 = 0.0
Pressure = 0.0
Mesh Update 2 = 0.0
End

Finally, no-slip conditions are set for the boundaries with the walls of the coater.
Boundary Condition 5
Name = "No-slip"
Target Boundaries = 5
Velocity 1 = 0.0
Velocity 2 = 0.0
Mesh Update 1 = 0.0
Mesh Update 2 = 0.0
End

Results
In the given case solution is obtained after 13 iterations. The solution gives the final radius, the forces, and
the profile of the free surface. To visualize the true free surface you may do the following. Read in the only
the last timestep and in ElmerPost give the following commands:

math nodes0 = nodes

math nodes = nodes0 + Mesh.Update
Note that this does not work if there is more than one set of variable values.

c CSC – IT Center for Science

18. Axisymmetric coating process 90

Figure 18.1: The velocity and pressure fields in a simple coating geometry. The solution utilizes the reduced
dimensional free surface solver.

c CSC – IT Center for Science

Tutorial 19

Blood ejection from a ventricle into

aorta

Directory: ArteryFlow
Solvers: FlowSolve,ElasticSolve, OutletCompute
Tools: Editor, Fortran 90 compiler, ElmerGrid
Dimensions: 2D, Transient

Case description
This tutorial is about simulating blood ejection in to the elastic human aorta. The idea is to mimic left ven-
tricle contration and resulting pulse propagation in an elastic conduit. In the simulation about 0.8 desiliters
of blood is ejected to a 50 cm long elastic aorta during a time period of 400 ms. In order to get the outlet
of the model behave physiologically more realistic way, a one dimensional model is coupled with the higher
order model.

Solution procedure
First we generate the mesh of 366 eight-node quadrilaterals elements with the command
ElmerGrid 1 2 contra
Next we generate one dimensional mesh to the outlet of the 2D model. The program AddOneDim is posed to
be run in the mesh directory contra. The length, the number of the elements, and the coordinate direction
of the 1D section will be asked.
In the simulation block the timestep is set equal to 1 ms and total simulation time equal to 600 ms. The
geometry consists of five bodies of which the first three are for the fluid volume. Body number 1 os the
contracting volume. Body 2 is a short rigid channel between the body 1 and the elastic artery. Artificial
compressibility method is used for the fluid volume (body 3) which is in contact with the elastic wall (body
4). One dimesional model is the body 5. Material settings for those are following:

! Bodies 1 and 2 (blood)

Material 1
Density = 1000
Viscosity = 3.5e-3
Youngs Modulus = 1
Poisson Ratio = 0.3
End

! Body 3 (blood)

91
19. Blood ejection from a ventricle into aorta 92

Material 2
Density = 1000
Viscosity = 3.5e-3
Youngs Modulus = 1
Poisson Ratio = 0.3
Compressibility Model = Artificial Compressible
Artificial Compressibility = 3.3E-5
End

! Body 4 (elastic wall)

Material 3
Density = 1010
Youngs Modulus = 3.0e5
Poisson Ratio = 0.45
End

! One dimesional model

Material 4
Density = 1010.0
Artery Wall Youngs Modulus = Real 3.0e5
Artery Radius = Real 0.0135
Artery Wall Thickness = Real 0.002
Artery Poisson Ratio = Real 0.45
End
Notice that the radius of the one dimesional model (Artery Radius) is to the midplane of the wall
(inner radius + half of the wall thickness). The overall FSI iteration scheme is started by one dimesional
solver (OutletCompute, see the solver manual), after that Navier-Stokes, elasticity and mesh update
solvers are run. Steady state convergence tolerance is set equalt to 1.0E-4 for each of the solvers. The
nonlinearities of each of the solvers are computed within the FSI scheme loop, that is, the flag Nonlinear
System Max Iterations is set equal to 1. Artificial compressibility coefficient is computed by the
equation c = (1 − ν 2 )[D/(E h)], where ν is the Poisson ratio of the artery wall, D, E and h are the inner
diameter, Young’s modulus and the thichness of the artery, respectively.
The only driving force of the system, the wall motion of the contracting fluid domain is given by the
fortran function Motion, see the figure 19.1. The boundary condition setting is
! Moving boundary
Boundary Condition 1
Target Boundaries = 1
Velocity 1 = 0
Velocity 2 = Equals Mesh Velocity 2
Mesh Update 1 = Real 0
Mesh Update 2 = Variable Time
Real Procedure "./Motion" "Motion"
End
At the outlet, the pressure boundary condition is given by the function OutletPres and the corre-
sponding radial displacement of the end wall of the outlet is given by the function OutletdX
! Outlet pressure of the 2D model
Boundary Condition 2
Target Boundaries = 2
Flux Integrate = Logical True
Flow Force BC = True
Pressure 2 = Variable Time
Real Procedure "./ArteryOutlet" "OutletPres"

c CSC – IT Center for Science

19. Blood ejection from a ventricle into aorta 93

Mesh Update 2 = Real 0

End

! Radial displacement of the end wall at the outlet of 2D model

Boundary Condition 9
Target Boundaries = 9
Displacement 1 = Variable Time
Real Procedure "ArteryOutlet" "OutletdX"
Displacement 2 = 0
End

FSI interface boundary is described as following

! FSI interface boundary
Boundary Condition 11
Target Boundaries = 11
Velocity 1 = Equals Mesh Velocity 1
Velocity 2 = Equals Mesh Velocity 2
Mesh Update 1 = Equals Displacement 1
Mesh Update 2 = Equals Displacement 2
Force BC = Logical True
End
Finally, the coupling of the 1D model with the 2D is done at the inlet boundary as

Boundary Condition 16
Target Boundaries = 16
Fluid Coupling With Boundary = Integer 2
Structure Coupling With Boundary = Integer 9
End

Results
The contraction is curve seen in the figure 19.1 and the velocity fields at different time levels are presented
in the figure 19.2. Postprocessing instructions are given in the file PostProcessingInstr.txt.

c CSC – IT Center for Science

19. Blood ejection from a ventricle into aorta 94

Figure 19.1: Contraction curve generated by the function Motion.

c CSC – IT Center for Science

19. Blood ejection from a ventricle into aorta 95

Figure 19.2: The geometry of the model and velocity fields at 5 time steps, 100, 200, 300, 400 and 500 ms.
The displacements of the wall are magnified by factor of 10.

c CSC – IT Center for Science

Part III

Utility Problems

96
Tutorial 20

Operator splitting in the evolutionary

heat equation

Directory: TemperatureOperatorSplitting
Solvers: HeatSolve, TransportEquationSolver, RateOfChangeSolver
Tools: Editor
Dimensions: 2D, Transient simulation

Introduction
The drawback of the stabilized finite element formulations available in Elmer to solve the convection-
diffusion equation and Navier-Stokes equations is that these methods are computationally expensive, in
particular when the residual-free-bubbles formulation is used. In evolutionary problems the reduction of
computational cost may be attempted by applying operator splitting techniques in which the original equa-
tion at each time step is splitted up into subproblems that are easier to solve numerically.
The aim of this chapter is to provide an illustrative example on using operator splitting capability in the
solution of the time-dependent heat equation. For the theoretical background of the operator splitting scheme
applied here the reader is referred to Elmer Models Manual and references therein.

Case description
The problem considered here is the solution of the time-dependent heat equation in the homogeneous L-
shaped plate the geometry of which is shown in Figure 27.1. The density of the material is taken to be unity,
while the heat conductivity k of the material is taken to be a parameter with values ranging between 0.05
and 0.01. The plate is heated by a constant internal heat source magnitude of which equals to unity. The
convection velocity field is assumed to be constant with the two Cartesian components of the velocity vector
equaling to unity. The boundaries Γi , i = 1, 4, 5, 6, are kept at constant temperature 0, while the boundaries
Γ2 and Γ3 are assumed to be insulated, i.e. the heat flux on this part of the boundary vanishes.
The problem is to solve the temperature distribution in the plate. The time interval for the simulation is
taken to be [0,2] and the temperature of the plate is 0 at the time t = 0.

Solution procedure
Using operator splitting the solution of the heat equation may be replaced at each time step by the solu-
tion of three subproblems consisting of two time-dependent Poisson equations and the convective transport
problem. The effects of diffusion and convection are decoupled by this splitting so that the diffusion and
convection phenomena are taken into account by the steps involving the solution of the Poisson equation and
the convective transport problem, respectively.

97
20. Operator splitting in the evolutionary heat equation 98

The time-dependent Poisson equations can be solved using the basic heat equation solver in Elmer. To
avoid the use of stabilized finite element formulations in the solution of the convective transport problem
this subproblem is solved by discretizating an equivalent wave-like equation formulation (second order in
time equation).
When the operator splitting method is applied, specific care is needed in prescribing boundary and initial
conditions for the simulation. While the boundary conditions (and initial conditions) for the steps involving
the solution of the Poisson equation may be defined in the usual manner, the boundary conditions for the con-
vective transport problem may be prescribed only on the part of inflow boundary on which the temperature
is prescribed. In the example the inflow boundary is the union of the boundary segments Γi , i = 1, 2, 5, 6, so
the boundary conditions for the convective transport equation should be given on Γi , i = 1, 5, 6. In addition
to prescribing the boundary conditions, the rate of change of the field subject to the convection operator
(here temperature) is needed as an initial condition at the beginning of pure convection step. This field can
be solved using a specific solver (RateOfChangeSolver) prior to the solution of the convective transport
problem. The boundary conditions for this solver should be prescribed on the same part of the boundary on
which the boundary conditions for the convective transport problem are prescribed.
It should be noted that in connection with the operator splitting method the user should specify an even
number of time steps. Although the details of implementation need not be known in order to use the operator
splitting ability, it is noted that the the running of two successive time steps actually constitutes a single
operator splitting scheme step as described in Elmer Models Manual. There are thus three equations (referred
to as Heat Equation, Rate Of Change Equation and Transport Equation) which may be
solved at one time step. At odd time steps all these equations are solved meaning that both the diffusion and
convection steps are taken, whereas at even time steps the solution of the convective transport problem is
omitted so that the diffusion step is performed only. Physically meaningful results satisfying all the essential
boundary conditions may thus be written to the results file only after even time steps.
The mesh for this example was created using Femlab and tools for converting meshes from Femlab to
Elmer format. The mesh in Elmer format is given in the tutorial files. This unstructured mesh consists of
8458 elements each of which having three nodes.
The Header section of the solver input file is used to declare the directory containing the Elmer mesh
files:

Header
Mesh DB "." "femlab-mesh"
End

The Simulation section of the solver input file is used to declare the coordinate system and simulation
type as well as simulation parameters related to the time discretization scheme:

Simulation
Coordinate System = String "Cartesian 2D"
Simulation Type = String "Transient"
Timestepping Method = String "Crank-Nicolson"
Timestep Intervals(1) = Integer 200
Timestep Sizes(1) = Real 0.01
Output Intervals(1) = Integer 1
Steady State Max Iterations = Integer 1
Post File = File "os-example.ep"
End

Here the keyword Timestepping Method is used to define the time discretization scheme that is
used in the solution of the time-dependent Poisson equations. Note that one should not use multi-step
methods in connection with the operator splitting method. The time discretization scheme that is used in the
solution of the convective transport problem is fixed and need not be specified. It should be noted also that
the number of time steps should be even. Since the equations solved at one time step are not coupled and
can thus be solved in sequential manner, Steady State Max Iterations may be taken to be 1.
The Body section of the solver input file is used to declare integer identifiers for body forces, equations,
initial conditions and materials:

c CSC – IT Center for Science

20. Operator splitting in the evolutionary heat equation 99

Body 1
Body Force = Integer 1
Equation = Integer 1
Initial Condition = Integer 1
Material = Integer 1
End

The Equation section of the solver input file in turn has the following declaration

Equation 1
Active Solvers(3) = Integer 1 2 3
End

indicating that three equations are to be solved using solvers with the integer identifiers 1, 2 and 3. Accord-
ingly, three Solver sections are needed. The first Solver section is used for the Poisson equation and has the
following declarations:

Solver 1
Equation = String "Heat Equation"
Variable = String "Temperature"
Variable Dofs = Integer 1
Linear System Solver = String "Iterative"
Linear System Iterative Method = String "BiCGStab"
Linear System Max Iterations = Integer 350
Linear System Convergence Tolerance = Real 1.0e-08
Linear System Abort Not Converged = Logical True
Linear System Preconditioning = String "ILU0"
Steady State Convergence Tolerance = Real 1.0e-05
Stabilize = Logical False
Bubbles = Logical False
End

Note that there is no need to use stabilization, so the values of the keywords Stabilize and Bubbles
may be set to be False to reduce the computational cost.
The remaining two Solver sections are related to the convective transport problem. The first one of these
sections is used to declare the solver parameters related to the problem (Rate Of Change Equation)
the solution of which gives the approximation to the rate of change of the temperature at the beginning of
pure convection step. The contents of this solver section is

Solver 2
Equation = String "Rate Of Change Equation"
Procedure = File "RateOfChange" "RateOfChangeSolver"
Variable = String "Udot0"
Variable Dofs = Integer 1
Advection = String "Constant"
Advection Variable = String "Temperature"
Linear System Solver = String "Iterative"
Linear System Iterative Method = String "BiCGStab"
Linear System Max Iterations = Integer 350
Linear System Convergence Tolerance = Real 1.0e-08
Linear System Abort Not Converged = Logical True
Linear System Preconditioning = String "ILU0"
Steady State Convergence Tolerance = Real 1.0e-05
End

c CSC – IT Center for Science

20. Operator splitting in the evolutionary heat equation 100

Here the keyword Advection Variable is used to declare the quantity which is subject to the
convection operator, while the keyword Advection is used to define the type of the velocity vector. The
name Udot0 is used for the solution of this problem.
Finally, the Solver section for the wave-like equation, which is equivalent to the convective transport
equation, has the following declarations:

Solver 3
Equation = String "Transport Equation"
Procedure = File "TransportEquation" "TransportEquationSolver"
Time Derivative Order = Integer 2
Variable = String "U"
Variable Dofs = Integer 1
Advection = String "Constant"
Advection Variable = String "Temperature"
Rate Of Change Equation Variable = String "Udot0"
Linear System Solver = String "Iterative"
Linear System Iterative Method = String "BiCGStab"
Linear System Max Iterations = Integer 350
Linear System Convergence Tolerance = Real 1.0e-8
Linear System Abort Not Converged = Logical True
Linear System Preconditioning = String "ILU0"
Steady State Convergence Tolerance = Real 1.0e-05
End

The name U is used for the solution of the convective transport problem. The value of the keyword Time
Derivative Order must be 2. The use of the keywords Advection Variable and Advection
is similar to that explained in connection with the second solver section.
The Material section is used to declare the material properties and, as the velocity vector in the convection
operator is of constant type, also the components of the velocity vector. In the case k = 0.01 the contents of
this section is

Material 1
Density = Real 1
Heat Capacity = Real 1
Heat Conductivity = Real 0.01
Advection Velocity 1 = Real 1
Advection Velocity 2 = Real 1
End

Body Force section is used to declare the body force in the Poisson equations.

Body Force 1
Heat Source = Real 1
End

Finally, the initial conditions and boundary conditions are specified. The temperature at t = 0 is defined
by giving the declaration

Initial Condition 1
Temperature = Real 0
End

Two Boundary Condition sections are needed. The first one is used to prescribe the boundary conditions
on the part of inflow boundary where the temperature is given:

c CSC – IT Center for Science

20. Operator splitting in the evolutionary heat equation 101

Table 20.1: The maximum temperature at t = 2.0. For comparison the maximum temperature according to
the steady state solution is also recorded.

Method k
0.05 0.025 0.01
OS 1.0235 1.0424 1.0677
S 1.0269 1.0436 1.0418
S(steady state) 1.0279 1.0437 1.0418
RFB 1.0271 1.0458 1.0853
RFB(steady state) 1.0286 1.0462 1.1050

Boundary Condition 1
Target Boundaries(3) = Integer 1 2 4
Temperature = Real 0
Udot0 1 = Real 0
U 1 = Real 0
End

The rate of change of the temperature is zero on this part of the boundary as the temperature is kept
fixed. Thus the zero boundary conditions for Udot0 are defined. The boundary value of the solution to the
convective transport problem should equal to the temperature. Therefore zero boundary conditions for U are
also defined.
The second Boundary Condition section is used to define the Dirichlet boundary conditions on the out-
flow boundary:

Boundary Condition 2
Target Boundaries(1) = Integer 3
Temperature = Real 0
End

Note that the zero heat flux condition need not be specified explicitly. Similarly, the treatment of the
outflow boundary conditions for the wave-like equation are handled automatically by the code and need not
be specified.

Results
From a numerical point of view the solution of the problem becomes increasingly difficult as the heat con-
ductivity k tends to zero. In order to examine possible dependence on the heat conductivity parameter the
problem was solved in three cases with k taking values 0.05, 0.025 and 0.01. For comparison the same case
was solved using three alternative methods. Here the operator splitting scheme is referred to as OS, S is the
stabilized finite element method and RFB is the method based on the residual-free-bubbles formulation. In
the case of methods S and RFB the simulation was performed using 100 time steps which corresponds to the
number of 200 time steps used in the case of operator splitting scheme.
The maximum temperature at t = 2.0 is recorded in Table 20.1. The CPU time used in the simulation to
obtain solution for a certain value of the heat conductivity is shown in Table 20.2.

Discussion
The benefit of the wave-like equation formulation of the convective transport problem is that this formulation
can be discretizated without using stabilized finite element formulations. Thus all the subproblems arising
from operator splitting can be solved using standard FE techniques. On the other hand, the expense of this
approach is that one is lead to handle the second order in time equation.

c CSC – IT Center for Science

20. Operator splitting in the evolutionary heat equation 102

Table 20.2: CPU time used by the different methods for simulation (k = 0.01).

Method CPU
OS 211.13
S 82.16
RFB 340.70

In the current implementation of the method the wave-like equation is discretizated in time using the
Crank-Nicolson scheme which is expected to perform well if the solution to the convective transport problem
is smooth. Spurious oscillations may however occur in the case of a rough solution changing rapidly in short
length scales. The user of the method should be aware that the deterioration of accuracy may thus occur if
the solution is not smooth and = k/||~u||∞ −→ 0, meaning that convection dominates.
In the cases considered convection dominates, the parameter ranging between 3.5 · 10−2 and 7.1 · 10−3 ,
and the solution has also sharp boundary layer near the upper edge. No spurious oscillations are however
detected in the solution. Nevertheless, the results recorded in Table 20.1 indicate that the differences between
the methods become larger as −→ 0.
In view of the results shown in Table 20.2 the accuracy of the operator splitting scheme in predicting the
maximum temperature is comparable to that of the residual-free-bubbles method while the computational
cost measured in CPU time is reduced by approximately 40 % when the operator splitting scheme is used.
It should be noted that the operator splitting scheme has the favourable feature that small spurious os-
cillations present in the solution after pure convection step may naturally be damped by the step involving
diffusion phenomena. Note also that each of the subproblems arising from the operator splitting may be
solved very efficiently using multigrid techniques, whereas robust multigrid solvers amenable to solving
linear systems arising from direct discretization of the convection-diffusion equation are still to be found.
This makes the operator splitting scheme attractive in the solution of problems having a very large number
of unknowns.

c CSC – IT Center for Science

Tutorial 21

Temperature distribution with BEM

Directory: PoissonBEM
Solvers: PoissonBEMSolver
Tools: ElmerGrid, editor
Dimensions: 2D

Case definition
This tutorial uses boundary element method (BEM) to solve Poisson equation. Even though Elmer is pri-
marily a finite element software the are limited support also for BEM computation. One should however
note that Elmer does not include any multilevel strategies essential for a good performance. For more details
about BEM check the Elmer Models Manual. The simulation setting is described in Figure 21.1. A heater
with constant heat flux is placed inside a box and the walls of the box are in fixed temperature. We are in-
terested in the temperature distribution in the medium around the heater (Ω) and on the surface of the heater
(Γ1 ). We also want to know the heat flux through the walls of the box (Γ2 ).

Ω, medium

heater

Γ1 , − ∂T
∂n = 1

Γ2 , T = 0

Figure 21.1: Simulation setting

Solution Procedure
First we create a mesh with ElmerGrid. The mesh is defined in heater.grd and it is created with com-
mand
ElmerGrid 1 2 heater
The solver input file PoissonBEM.sif starts with the definition of the mesh directory.

103
21. Temperature distribution with BEM 104

Header
Mesh DB "." "heater"
End
The simulation uses 2D cartesian geometry, searches a steady state and since there is no coupled solvers only
one iteration is needed. Numerical results are written to file BEM_Temperature.result and ElmerPost
file is BEM_Temperature.ep.
Simulation
Coordinate System = Cartesian 2D
Coordinate Mapping(3) = 1 2 3

Simulation Type = Steady

Steady State Max Iterations = 1

Output Intervals = 1
Post File = "BEM_Temperature.ep"
Output File = "BEM_Temperature.result"
End
There is just one body, the medium around the heater, and it uses equation 1.
Body 1
Name = "medium"
Equation = 1
End
In equation block we say that we use the solver named PoissonBEM.
Equation 1
PoissonBEM = Logical True
End
In solver block the Equation keyword must match the one in equation block. We also need to define the
procedure, name the variable (Temperature) and tell the degrees of freedom of the variable. Keyword
Optimize Bandwidth must be set to false with BEM solver. Since we were interested in the flux, we
must now export it to the results. The lines beginning Exported must be exactly as below. Keywords
beginning Linear System can be used except that the preconditioning cannot be ILU.
Solver 1
Equation = PoissonBEM
Procedure = "PoissonBEM" "PoissonBEMSolver"
Variable = Temperature
Variable DOFs = 1

Optimize Bandwidth = False

Exported Variable 1 = String Flux

Exported Variable 1 DOFs = 1

Linear System Solver = Iterative

Linear System Iterative Method = BiCGStab
Linear System Preconditioning = Jacobi
Linear System Max Iterations = 100
Linear System Convergence Tolerance = 1.0e-8

Steady State Convergence Tolerance = 1.0e-6

End

c CSC – IT Center for Science

21. Temperature distribution with BEM 105

Finally we give the boundary conditions for the heater surface and for the walls of the box. The keyword
Body Id tells the reference body of this boundary. Here it is 1. The keyword Normal Target Body
tells the direction of the outer normal. Value -1 means the side where there are no volume elements. We
didn’t mesh the inside of the heater and so we can use value -1 in both cases. The heat flux from heater to
medium is 1 and the walls of the box are set to zero temperature. The keyword Temperature matches the
name of the variable in solver block.
Boundary Condition 1
Name = "heater_surface"
Target Boundaries = 1

Body Id = 1
Normal Target Body = Integer -1
Flux = Real 1
End

Boundary Condition 2
Name = "box_walls"
Target Boundaries = 2

Body Id = 1
Normal Target Body = Integer -1
Temperature = 0
End

Results
Problem is solved with command Solver. The results are then viewed with ElmerPost. In Figure 21.2
is the temperature distribution.

Figure 21.2: The temperature distribution.

c CSC – IT Center for Science

Tutorial 22

Adding user defined equation solver

Directory: Temperature1D
Solvers: PoissonSolver
Tools: Editor, Fortran 90 compiler, ElmerGrid
Dimensions: 1D, Steady-state

Problem description
This tutorial is about creating the code for a simple poisson equation solver. The solver is applied to 1d case
with internal source term and fixed boundaries.
Mathematically the problem we solve is

−∆Φ = f in Ω
(22.1)
Φ = 0 on Γ

Allthough this example is in 1d the same solver code also applies to 2D and 3D problems.

Solution procedure
Own codes solving some specific equation may be added dynamically to Elmer software. Here we create a
very simple equation solver code. The final code may be found in the tutorial directory as well as the files
for running the example. The solution may be attempted as follows:

• Copy all the files from tutorial directory to current directory

• Setup Elmer

• Give the following commands:

elmerf90 -o Poisson Poisson.f90

ElmerGrid 1 2 1dheat
ElmerSolver
ElmerPost

The solver code

The example Fortran code may be found in the tutorial files under the name Poisson.f90. The example run
is defined in 1dheat.sif. Only a rough guidline is given here of both of the files, refer to the files themselves
for more details.
All the equation solvers in Elmer have the following common interface

106
22. Adding user defined equation solver 107

SUBROUTINE PoissonSolver( Model, Solver, dt, TransientSimulation )

USE SolverUtils

TYPE(Model) :: Model
TYPE(Solver_t), POINTER :: Solver
REAL(KIND=dp) :: dt
LOGICAL :: TransientSimulation

...
END SUBROUTINE PoissonSolver
The argument Model contains pointers to the whole definition of the Elmer run. The argument Solver
contains parameters specific to our equation solver. The argument dt and TransientSimulation are the current
timestep size, and a flag if this run is steady or transient. These don’t concern us this time.
When starting the ElmerSolver looks the solver input (.sif) file for a Solver section with keyword "Pro-
cedure". This should contain reference to the compiled code

Procedure = "Poisson" "PoissonSolver"

where the first string in the right hand side is the file name of the compiled code, and second argument is the
name of the subroutine to look for in the given file.
In the Solver section one also gives the name of the field variable (here Poisson) and the DOFs/node
(here 1).
The basic duty of the equation solver is to solve one or more field variables inside the time progressing-
or steady state iteration-loop of ElmerSolver. Here we use FEM to discretize the Poisson equation and finally
solve the equation by calling ElmerSolver utility SolveSystem.
The solution progresses the following way:

• Get the space for variables and temporaries from ElmerSolver and compiler. The matrix structure and
space for solution and RHS vector have already been allocated for you before you enter the equation
solver.
The matrix is of type Matrix_t and may be obtained from the arguments as

TYPE(Matrix_t), POINTER :: StiffMatrix

StiffMatrix => Solver % Matrix

Usually one doesn’t need to know the internal storage scheme or the fields of the Matrix type, but one
just passes this pointer further to ElmerSolver utility routines.
Similarly, the force vector may be accessed as follows:

REAL(KIND=dp), POINTER :: ForceVector(:)

ForceVector => StiffMatrix % RHS

The solution vector is obtainable similarily

TYPE(Variable_t), POINTER :: Solution

Solution => Solver % Variable

The Variable_t structure contains the following fields

– DOFs: the number of degrees of freedom for one node. This value is for information only and
should’nt be modified.
– Perm: an integer array that is nonzero for nodes that belong to computational volume for this
equation. The entry P erm(i) holds the index of the global matrix row (for 1 DOF) for nodal
point i. This array should’nt be modified by the equation solver.

c CSC – IT Center for Science

22. Adding user defined equation solver 108

– Values: Space for the solution vector values. Note that the values are ordered the same way as
the matrix rows, .i.e. the value of Potential at node n is stored at
val = Solution % Values( Solution % Perm(n) )
• Initialize the global system to zero. Calling the utility routing

CALL InitializeToZero( StiffMatrix, ForceVector )

is usually enough.
• Go trough the elements for which this equation is to be solved, get the elemental matrices and vectors
and add them to the global system:

DO i=1,Solver % NumberOfActiveElements
CurrentElement => Solver % Mesh % Elements( Solver % ActiveElements(i) )
...
CALL LocalMatrix( ... )
CALL UpdateGlobalEquations( ... )
END DO
CALL FinishAssembly( ... )

Here the LocalMatrix is your own subroutine computing elemental matrices and vectors. In the exam-
ple code LocalMatrix uses three routines from ElmerSolver utilities. The function

dim = CoordinateSystemDimension()

returns the dimension of the current coordinate system, i.e. the return value is 1, 2 or 3 depending
on the input file setting of keyword "Coordinate System". The function GaussPoints returns structure
containing the integration point local coordinates and weights

TYPE(GaussIntegrationPoints_t) :: IntegStuff
IntegStuff = GaussPoints( Element )

The fields of the type GaussIntegrationPoints_t are

INTEGER :: n
REAL(KIND=dp) :: u(:), v(:), w(:), s(:)

the integer value n is the number of points selected. The arrays u,v and w are the local coordinates of
the points, and the array s contains the weights of the points. One may call the GaussPoints-routine
with second argument,

IntegStuff = GaussPoints( Element, n )

if the default number of integration points for given element is not suitable.
Inside the integration loop the function ElementInfo is called:

TYPE(Element_t), POINTER :: Element

TYPE(Nodes_t) :: Nodes
REAL(KIND=dp) :: U,V,W,detJ, Basis(n), dBasisdx(n,3), ddBasisddx(n,3,3)

stat = ElementInfo( Element, Nodes, U, V, W, detJ, &

Basis, dBasisdx, ddBasisddx, .FALSE. )

c CSC – IT Center for Science

22. Adding user defined equation solver 109

0.14

0.12

0.1

0.08
Potential

0.06

0.04

0.02

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Figure 22.1: Solution of the Poisson Equation.

This routine returns determinant of the element jacobian (detJ), basis function values (Basis(n)), basis
function global derivative values (dBasisdx(n,3)), basis function second derivative values ( ddBasis-
ddx(n,3,3) ). The second derivatives are only computed if the next logical flag is set to true. All the
values are computed at the point U,V,W inside element defined by structures Element and Nodes.
Refer to the code for more details.
• Set boundary conditions. Here only dirichlet boundary conditions are used. These may be set by using
the utility routine SetDirichletBoundaries.
• Solve the system by calling utility routine SolveSystem.

Results
In the elmerpost file there is a variable called Potential which contains the solution of this simple example.
See figure 22.1

c CSC – IT Center for Science

Tutorial 23

Volume flow boundary condition

Directory: FlowLinearRestriction
Solvers: FlowSolve, SolveWithLinearRestriction
Tools: Editor, Fortran 90 compiler, ElmerGrid
Dimensions: 2D, Transient

Case definition
This tutorial gives an example how to use SolveWithLinearRestriction. It also describes how to execute own
functions before the original system is solved. In order to understant the case reader should be familiar with
compressed row storage matrixes and elmer basics. This tutorial gives only the guidelines and reader is
adviced to read the files in order to get more through understanding.
We simulate the flow of incompressible fluid in a pipe. The pipe has a length of 5 and a width of 1.
On the left end we want to describe a certain timedependent volume flow. In other words, we don’t want
to describe the velocity field here but we want the velocity field be such that it transports certain amount of
volume in timeinterval. We could integrate the correct volume flow, but let’s now approximate it to make
the more important aspects more visible. Our approximation here is that the volume flow is proportional to
average velocity on the edge i.e.
N
1 X volume
ui = (23.1)
N i=1 time
Here ui are the nodal velocities parallel to the pipe on the left edge and N is the number of nodes on the left
edge. We want to set a nicely scaled sinusoidal volume flow on the edge, which leads to
N
X
ui = 10N sin(2Πt) (23.2)
i=1

This equation we can (easily) force with lagrange multiplier.

Solution procedure
First we make a uniform mesh of 800 four-node quadrilaterals with command
ElmerGrid 1 2 mflow
Next we construct the solver input file. Header is simply
Header
Mesh DB "." "mflow"
End
The simulation block is also very simple. Here we need to define the timestepping method and timescale.

110
23. Volume flow boundary condition 111

Simulation
Coordinate System = Cartesian 2D

Simulation Type = Transient

Steady State Max Iterations = 1

Timestepping Method = BDF

BDF Order = 1

Timestep Sizes = 0.02

Timestep Intervals = 100

Output Intervals = 1

Output File = "mflow.result"

Post File = "mflow.ep"
End

The body, material and equation blocks are as usual. The material parameters, of course, have affect on the
solution and interested reader is encouraged to modify these values and recalculate the solution.
Body 1
Material = 1
Equation = 1
End

Material 1
Density = 3.0
Viscosity = 0.1
End

Equation 1
Navier-Stokes = TRUE
Active Solvers(1) = 1
End

The solver block has the usual Navier-Stokes keywords and two keywords for volume flow boundary. The
Before Linsolve keyword defines binaryfile and function that is called before the system is solved.
This function we must write and compile and we will come to it shortly. The following keyword, Export
Lagrange Multiplier, states that we are not interested in the value of the Lagrenge multiplier and it
is therefore not saved.

Solver 1
Equation = Navier-Stokes
Stabilize = True

Before Linsolve = "./AddMassFlow" "AddMassFlow"

Export Lagrange Multiplier = Logical FALSE

Linear System Solver = Iterative

Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU1
Linear System Max Iterations = 500
Linear System Scaling = False
Linear System Convergence Tolerance = 1.0e-8

c CSC – IT Center for Science

23. Volume flow boundary condition 112

Nonlinear System Max Iterations = 15

Nonlinear System Convergence Tolerance = 1.0e-8
Nonlinear System Newton After Tolerance = 1.0e-4
Nonlinear System Newton After Iterations = 8
Nonlinear System Relaxation Factor = 1.0

Steady State Convergence Tolerance = 1.0e-7

End
In boundary conditions we state that both parallel and perpendiculer velocities are zero on the pipe sides and
on both edges the perpendicular velocity is zero. Here we also define the number tags for the boundaries.
The tag 2 is assigned to boundary that has number 4 in grd-file, which is the left edge of the pipe. To this tag
number 2 we shall refer in our AddMassFlow-function.
Boundary Condition 1
Target Boundaries(2) = 1 3
Velocity 1 = 0.0
Velocity 2 = 0.0
End

Boundary Condition 2
Target Boundaries = 4
Velocity 2 = 0.0
End

Boundary Condition 3
Target Boundaries = 2
Velocity 2 = 0.0
End

AddMassFlow function
Here we shall only give some rough guidelines of the function, for more information check the code. This
function creates the constraint matrix and RHS that forces the equation mentioned above. Then it calls
SolveWithLinearRestriction to solve the system. The coinstraint matrix is actually only a row-vector and the
RHS is only one value.
• The function parameters are defined in Elmer so you shouldn’t change them.
• First we set a pointer to EMatrix-field of the given system matrix. If the pointed matrix is not yet
allocated, calculate the number of nodes on the edge we want to define the volume flow. This gives us
the number of non-zeros in our constraint matrix and we can allocate the matrix.
• Set the rows, cols and diag -fields of the matrix. This sets the non-zeros on their right places in the
constraint matrix.
• Set all values of the constraint matrix to unity.
• Calculate the RHS-value. The current time was checked in the beginning of the function, so this is
possible.
• Call SolveWithLinearRestriction
• Return 1 which tells the ElmerSolver that the system is already solved.
The function is the compiled with command
elmerf90 -o AddMassFlow AddMassFlow.f90
Here it is assumed that the source file name is AddMassFlow.f90.

c CSC – IT Center for Science

23. Volume flow boundary condition 113

Results
Just say ElmerSolver and you should get the solution in few minutes. The velocity perpendicular to the
pipe is practically zero and the velocity parallel to the pipe is an example of Womersley velocity profile 1 .
An interesting feature of this velocity profile is that on some timesteps the fluid flows to both directions, see
figure 23.1.

Figure 23.1: Solution of the velocity field. Note the flow to both directions.

1 J.Physiol (1955) 127, 553-563

c CSC – IT Center for Science

Tutorial 24

Streamlines

Directory: FlowStreamlines
Solvers: StreamSolver, FlowSolve
Tools: ElmerGrid, editor
Dimensions: 2D

Case definition
The case definition is the same as in the incompressible flow passing a step. The mathematical definition of
the stream function ψ is
∂ψ ∂ψ
u = , v = − . (24.1)
∂y ∂x
where u, v are the velocity components in x, y geometry. For more info check Elmer Models Manual.

Solution Procedure
First we create a mesh with ElmerGrid. The mesh is defined in step.grd and it is created with command

ElmerGrid 1 2 step
You may need to compile the StreamSolver yourself. If the Elmer environment is succesfully setup the
compilation command should look like the following lines,
elmerf90 -o StreamSolver StreamSolver.f90

The solver input file streamlines.sif starts with the definition of the mesh directory.
Header
Mesh DB "." "step"
End

The simulation uses 2D cartesian geometry and searches a Steady State. There is no coupled solvers so only
one iteration is needed. Numerical results are written to file streamlines.result and ElmerPost file
is streamlines.ep.
Simulation
Coordinate System = Cartesian 2D
Coordinate Mapping(3) = 1 2 3

Simulation Type = Steady

Steady State Max Iterations = 1

114
24. Streamlines 115

Output Intervals = 1
Post File = "streamlines.ep"
Output File = "streamlines.result"
End
There is just one body and it uses equation 1 and is of material 1.

Body 1
Equation = 1
Material = 1
End

The equation block states that we use Solvers 1 and 2 to solve the problem and that we use Navier-Stokes
equations.
Equation 1
Active Solvers(2) = 1 2
Navier-Stokes = True
End
In material block we define the density and the viscosity of the fluid.
Material 1
Density = 1
Viscosity = 0.01
End
Solver 1 is for the Navier-Stokes equations. Here we give the linear system solver 1 and convergence crite-
rions for linear, nonlinear and steady state solution of the Navier-Stokes equations.
Solver 1
Equation = "Navier-Stokes"
Stabilize = True

Linear System Solver = Iterative

Linear System Iterative Method = BiCGStab
Linear System Max Iterations = 500
Linear System Convergence Tolerance = 1.0e-8
Linear System Preconditioning = ILU1

Nonlinear System Convergence Tolerance = 1.0e-6

Nonlinear System Max Iterations = 15
Nonlinear System Newton After Iterations = 8
Nonlinear System Newton After Tolerance = 1.0e-4
Nonlinear System Relaxation Factor = 1.0

Steady State Convergence Tolerance = 1.0e-6

End

Then the solver for streamlines.

• Name of the equation. This may be what ever you like.
• Name of the binary file and the subroutine. If you compiled the StreamSolver yourself, then you may
need to change this to Procedure = "./StreamSolver" "StreamSolver".

• Name of the variable. This may be what ever you like.

1 Biconjugate gradient method with incomplete LU preconditioning

c CSC – IT Center for Science

24. Streamlines 116

• Stream function is scalar, so the degree of freedom is 1.

Next set of keywords is for the StreamSolver. More info on keywords is in the Elmer Models Manual.
• Name of the flow field variable. The name of the FlowSolves variable is FlowSolution.

• Global number of the offset node. 1 is always a safe choise.

• Shift the smallest value to zero.
• Scale the maximum value to 1.
• Use the normal stream function i.e. don’t use Stokes stream function.

Then we define the linear system solver and convergence criterions.

Solver 2
Equation = "StreamSolver"
Procedure = "StreamSolver" "StreamSolver"
Variable = "StreamFunction"
Variable DOFs = 1

Stream Function Velocity Variable = String "Flow Solution"

Stream Function First Node = Integer 1
Stream Function Shifting = Logical TRUE
Stream Function Scaling = Logical TRUE
Stokes Stream Function = Logical FALSE

Linear System Solver = Iterative

Linear System Iterative Method = BiCGStab
Linear System Max Iterations = 500
Linear System Convergence Tolerance = 1.0e-8
Linear System Preconditioning = ILU1

Steady State Convergence Tolerance = 1.0e-6

End
Finally we give the boundary conditions. The condition 1 is for the lower and upper side of the step
(Γ1 ,Γ2 ,Γ3 ,Γ5 in case definition). Here both velocities are zero. The condition 2 is for the output edge
(Γ4 ). Here vertical velocity is zero. The condition 3 is for the input edge (Γ6 ). Here horzontal velocity is 1
and vertical velocity is zero.
Boundary Condition 1
Target Boundaries = 1
Velocity 1 = 0
Velocity 2 = 0
End

Boundary Condition 2
Target Boundaries = 2
Velocity 2 = 0
End

Boundary Condition 3
Target Boundaries = 3
Velocity 1 = 1
Velocity 2 = 0
End

c CSC – IT Center for Science

24. Streamlines 117

Results
Problem is solved with command Solver. The results are then viewed with ElmerPost. In figure 24.1
are some contour lines of the stream function. These are also flows streamlines. The contour values are
manually selected to get a nice picture. Note the swirl after the step.

Figure 24.1: The streamlines of the flow.

c CSC – IT Center for Science

Tutorial 25

Electroosmotic flow and advected

species

Directory: Microfluidic
Solvers: StatElecSolve, FlowSolve , AdvectionDiffusion, Electrokinetics
Tools: ElmerGrid, editor
Dimensions: 2D

Case definition
This tutorial is an example of setting up a simulation for (microfluidic) electroosmotic flow advecting a
passive scalar quantity. Diffusion of the species is also included. The geometry of the system is a simple 2D
microchannel with T crossing. The flow is induced by the applied electric field and the electric double layer
at the channel walls. The analyte (species) is inserted into the system from the left hand side inlet.
More details on the electrokinetic capabilities of Elmer are found on the Models Manual, chapter “Elec-
trokinetics”.

Solution Procedure
The computatonal mesh is done with ElmerGrid in directory Tcross with the command
ElmerGrid 1 2 Tcross -scale 1e-5 1e-5 1e-5
The scale option above is used to obtain appropriate dimensions from a geometry template defined in nondi-
mensional units.
The command file may be written with a text editor. The file includes the following information.
The mesh directory is given in the header of the command file
Header
Mesh DB "." "Tcross"
End

The simulation block of the command file defines, eg., the case to be time dependent (transient) with 10−5
second timesteps and altogether 120 time intervals. Results from every second timestep are saved into the
files diffusion1.*.
Simulation
Coordinate System = Cartesian 2D

Simulation Type = "Transient"

Steady State Max Iterations = 20

118
25. Electroosmotic flow and advected species 119

Timestep Intervals = 120

Timestep Sizes = 1e-5
Output Intervals = 2

Timestepping Method = BDF

BDF Order = 2

Binary Output = Logical True

Output File = "diffusion1.res"

Post File = "diffusion1.ep"

Output Version Numbers = Logical True

Max Output Level = 32
End
The electrostatics and electrokinetics solvers require the value of the permittivity of vacuum. This is actually
not even needed here since the case deals with conducting media. Thus the value has been fixed as 1.0 to
avoid warnings on missing constant definitions.
Constants
! Permittivity Of Vacuum = 8.8542e-12 ! C^2/Nm^2
Permittivity Of Vacuum = 1.0 ! manipulation for conducting material
End
The case includes only one body. The corresponding equation definitions are found in section Equation
1 and material parameters from section Material 1.
Body 1
Equation = 1
Material = 1
End
All three solvers are active in this equation set. Further definitions include, first, that the convection of the
species is switched on (besides diffusion). Then that for Navier-Stokes equations the convective term may
be left out resulting in laminar Stokes flow, and finally that the elctric field is computed by the electrostatics
rather than given by the user.
Equation 1
Active Solvers(3) = 1 2 3
Convection = Computed
NS Convect = False
Electric Field = String "computed"
End
Following are the solver definitions. Solver 1 is the electrostatics solver. The equation is linear and thus no
nonlinear iterations are needed. The equation is solved using a fast direct method UMFPack.
Solver 1
Equation = "Stat Elec Solver"
Procedure = "StatElecSolve" "StatElecSolver"
Variable = String "Potential"
Variable DOFs = 1

Calculate Electric Field = True

Calculate Electric Flux = False
Linear System Convergence Tolerance = 1.0E-10

c CSC – IT Center for Science

25. Electroosmotic flow and advected species 120

Linear System Solver = Direct

Linear System Direct Method = UMFPack
Linear System Preconditioning = ILU1
Linear System Residual Output = 1

Nonlinear System Max Iterations = 1

Nonlinear System Convergence Tolerance = 1.0e-10
Steady State Convergence Tolerance = 1.0E-10
End

The next solver is for the Navier-Stokes equations. Here nonlinear iterations are required.
Solver 2
Equation = "Navier-Stokes"

Linear System Convergence Tolerance = 1.0D-08

Linear System Solver = Iterative
Linear System Iterative Method = "BiCGStab"
Linear System Max Iterations = 500
Linear System Abort Not Converged = False ! True
Linear System Preconditioning = ILU1
Linear System Residual Output = 10

Nonlinear System Convergence Tolerance = 1.0e-6

Nonlinear System Max Iterations = 30
Nonlinear System Newton After Iterations = 10
Nonlinear System Newton After Tolerance = Real 1.0D-8
Nonlinear System Relaxation Factor = 1.0
Steady State Convergence Tolerance = 1.0D-03

Stabilize = True
End
The advection-diffusion equation does not affect neither the electrostatic field nor the flow, thus it may be
solved only after a converged solution for the previous two equations is available. This is achieved with
the Exec Solver definition below. The advected quantity is given the name Analyte. The advection-
diffusion solver uses bubble stabilization method to avoid numerical problems associated with convection
type equations.
Solver 3
Exec Solver = After Timestep
Equation = "Analyte transfer"
Procedure = "AdvectionDiffusion" "AdvectionDiffusionSolver"
Variable = String "Analyte"
Variable DOFs = 1

Bubbles = True

Linear System Convergence Tolerance = 1.0E-06

Linear System Solver = "Iterative"
Linear System Iterative Method = "BiCGStab"
Linear System Max Iterations = 500
Linear System Preconditioning = ILU2
Linear System Residual Output = 1

c CSC – IT Center for Science

25. Electroosmotic flow and advected species 121

Nonlinear System Max Iterations = 1

Nonlinear System Convergence Tolerance = 1.0e-5
Steady State Convergence Tolerance = 1.0D-06
End
The material parameters are given below.

Material 1
Density = 1e3

Viscosity = 1e-03

Relative Permittivity = 1.0 ! this is actually electric conductivity

Analyte Diffusivity = Real 1e-10

End
Finally the boundary conditions are defined. The first BC is given for the channel walls. Here, tangential
velocity (velocity components 1 and 2) is computed by the Helmholtz-Smoluchowski slip velocity condition,
which means that the velocity is computed using the computed electric field and the electroosmotic mobility
as inputs.
Boundary Condition 1
Name = "channel-walls"
Target Boundaries(2) = 4 5

EO Mobility = Real 5e-08

Velocity 1 = Variable Pressure

Real Procedure "Electrokinetics" "helmholtz_smoluchowski1"
Velocity 2 = Variable Pressure
Real Procedure "Electrokinetics" "helmholtz_smoluchowski2"
End
The next BC is the inlet condition. We give a potential of 100 Volts and define that there is no flow in
y-direction. The analyte concentration at the inlet is defined as a function of time using table format. The
concentration is 1.0 up until time instant 3 · 10−5 , is zero after 4 · 10−5 s and decreases linearly between these
two time instants.
Boundary Condition 2
Name = "el_A"
Target Boundaries = 1

Potential = 100.0

Velocity 2 = 0.0

Analyte = Variable Time

Real
0.0 1.0
3.0e-5 1.0
4.0e-5 0.0
0.5 0.0
End

End

c CSC – IT Center for Science

25. Electroosmotic flow and advected species 122

The final two boundary conditions are for the outlets. Different potentials for these are defined as well as a
condition for velocity component.
Boundary Condition 3
Name = "el_B"
Target Boundaries = 2

Potential = 30.0

Velocity 1 = 0.0
End

Boundary Condition 4
Name = "el_C"
Target Boundaries = 3

Potential = 0.0

Velocity 1 = 0.0
End
After writing the command file is finished, the problem can be solved by entering the command ElmerSolver
on the command line. The results can be examined with ElmerPost.

Results
Solving the problems takes less than a minute cpu time on a PC. The maximum and minimum concentration
over the whole simulation are 1.0235 and -0.075748. The solution of this problem should be between 0
and 1. This shows that some numerical discretization errors are present in the simulation. The errors would
diminish when using smaller timesteps and also with denser mesh. Simulation results at the time instant of
0.00025 seconds are shown in Fig. 25.1.

Figure 25.1: Analyte concentration on time instant 0.00025 seconds.

c CSC – IT Center for Science

25. Electroosmotic flow and advected species 123

The maximum value of the magnitude of the velocity in the results is 0.105 m/s.
The electric field is written into the output only componentwise. The magnitude of the field may be
visualised after giving the following commands on the ElmerPost command line (assuming one has read in
all 61 timesteps):
math E(0,time(0):time(61)-1) = Electric.field.1
math E(1,time(0):time(61)-1) = Electric.field.2
math E(2,time(0):time(61)-1) = 0
math E_abs = sqrt(vdot(E,E))
Now visualising the variable E_abs reveals that the electric field magnitude is between 2453 and 2.18 · 106 .

Notes
When checking the simulation results the user will notice that the electric potential does not change in time
and that the flow reaches steady-state within a few timesteps. This is quite clear also from the problem
setup: the electric field is due to invarying potential with constant material parameters. Also the flow in
microchannels is usually laminar.
Thus the most efficient way to simulate the current case would be to compute first the steady-state
solution for the electrokinetic flow and use the steady flow to advect the analyte. This would be done by
running two separate simulations; resolve first the steady flow, and to use this flow solution as restart for the
advection-diffusion equation.

c CSC – IT Center for Science

Tutorial 26

Active and passive elements

Directory: PassiveElements
Solvers: HeatSolve
Tools: ElmerGrid, editor
Dimensions: 2D

Case definition
This tutorial shows an example of using passive elements in Elmer. This feature allows the activation and
deactivation of parts of the geometry during the simulation. This tutorial uses the heat eqution solver to
demonstrate this capability. Use with other solvers is analogous.
The geometry of the problem consists of two parts. The lower edge of the lower part is held at constant
temperature of 1 degrees. The upper body is heated with a constant heating power. Between time steps 5
and 6 the two bodies are connected by two heat conductors, and the heat is conducted from the higher body
to the lower one. The goal of the simulation is to model the temperature distribution in the system over time.
The problem is a pure heat transfer problem that may be solved with HeatSolve.

Solution Procedure
The computatonal mesh is done with ElmerGrid in directory tmesh with the command
ElmerGrid 1 2 tmesh
The command file may be written with a text editor. The file includes the following information.
The mesh directory is given in the header of the command file
Header
Mesh DB "." "tmesh"
End
The simulation block of the command file defines, eg., the case to be time dependent (transient) with 1
second timesteps and altogether 15 time intervals.
Simulation
Max Output Level = 32
Coordinate System = Cartesian 2D
Simulation Type = Transient
Timestepping Method = BDF
BDF Order = 2
Timestep Intervals = 15
Timestep Sizes = 1
Output Intervals = 1

124
26. Active and passive elements 125

Steady State Max Iterations = 1

Output Version Numbers = Logical True
Output File = heat.res
Post File = heat.ep
End
The heat equation solver asks for the Stefan-Boltzmann constant that gives the relationship between temper-
ature and radiation power, although radiation is not needed here. Let us define it anyway to avoid warnings
of missing parameters.
Constants
Stefan Boltzmann = 5.67e-8
End
There are three bodies with the same equation but different material properties. Body 3 is heated by a
constant body force. Body 2 forms the connecting parts of the system. An initial condition as well as a body
force is defined for this body. The body force contains the initial deactivation, and later activation, of the
connecting part. Note that this part is included in the geometry all the time, but the command file is used to
define when they are included into the simulation.
Body 1
Equation = 1
Material = 1
End

Body 2
Equation = 1
Material = 2
Body Force = 2
Initial Condition = 1
End

Body 3
Equation = 1
Material = 1
Body Force = 1
End
The only solver is the heat solver (Solver 1)
Equation 1
Active Solvers = 1
End
The initial condition for the initially passive elements is taken to be 1 degrees; the same temperature than
the colder part of the system has as a boundary condition.
Initial Condition 1
Temperature = 1.0
End
The heating power is defined to be 10 W/kg
Body Force 1
Heat Source = 10
End
Now the passive condition for the connecting part is defined. When the parameter Temperature Passive
has a value larger than zero, the current element is excluded from the solution, otherwise it is included as a
normal element. The parameter may depend on variables, coordinates or time. Here it is defined to depend
on time using a tabular format.

c CSC – IT Center for Science

26. Active and passive elements 126

Body Force 2
Temperature Passive = Variable Time
Real
0.0 1.0
5.0 1.0
5.2 -1.0
8.0 -1.0
End

End
The material properties of the system are artificial. The following three properties are needed for each
material.
Material 1
Heat Capacity = 1
Heat Conductivity = 1
Density = 1
End

Material 2
Heat Capacity = 10
Heat Conductivity = 1
Density = 1
End
The heat equation is solved with an iterative method. The system is linear, thus multiple nonlinear iterations
are not needed.
Solver 1
Equation = heat equation
Linear System Solver = Iterative
Linear System Iterative Method = BiCGStab
Linear System Preconditioning = ILU0
Linear System Max Iterations = 300
Linear System Convergence Tolerance = 1.0e-6
Linear System Abort Not Converged = Logical False
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-5
Steady State Convergence Tolerance = 1.0e-5
End
The boundary conditions are simple. The lower boundary of the lower body is held at 1 degrees and the
upper boundary of the upper body at 10 degrees.
Boundary Condition 1
Target Boundaries = 1

Temperature = 1
End

Boundary Condition 2
Target Boundaries = 4

Temperature = 10
End
After writing the command file is finished, the problem can be solved by entering the command ElmerSolver
on the command line. The results can be examined with ElmerPost.

c CSC – IT Center for Science

26. Active and passive elements 127

Figure 26.1: Temperature distribution of the system at the final time instant (with spectral_32 color map).

Results
With the given computational mesh the problem is solved in a few seconds. The maximum and minimum
temperatures in the system over the whole simulation are 15.466 degrees and 0.6525 degrees respectively.
The maximum and minimum temperature at the final time instant are 14.207 degrees and 1.000 degrees,
respectively. The results at the time instant of 15 seconds are shown in Fig. 26.1.

Notes
For equations with more than one components (such as displacement for Stress Analysis solver in 2D or
3D) the passive elements feature apply to all the components. The feature is activated by defining, eg.,
Displacement Passive in the Body Force section. Note that for Navier-Stokes equations one should
use Flow Solution Passive, and that this affects the Pressure as well as the Velocity components.
However, when using multiple solvers, one can define some of them passive and some of them active at
the same time.

c CSC – IT Center for Science

Part IV

Obsolite Problems

128
129

These tutorials have become obsolite due to the transition from ElmerFront to ElmerGUI. If you still
want to continue using ElmerFront these tutorials may be perfectly valid.

c CSC – IT Center for Science

Tutorial 27

Temperature distribution

Directory: TemperatureAngle
Solvers: HeatSolve
Tools: ElmerFront
Dimensions: 2D, Steady-state

Problem description
An L-shaped structure (see figure 27.1) is heated by an internal heat source, which magnitude is 1 W/m3 .
The density of the structure is 1 kg/m3 and the heat conductivity is 1 W/mK. All the boundaries Γi are kept
on constant temperature of 0 K. The problem is to solve the temperature distribution in the structure.
Mathematically the problem to be solved is

−κ∆T = ρf inΩ
(27.1)
T = 0 onΓ

where κ is the heat conductivity, T is the temperature and f is the heat source. It is assumed that density and
heat conductivity are constants.

Solution procedure
• Start Elmer in the desired directory with the command

ElmerFront

• Open the file that contains the geometry of the structure from the File menu. Select also the working
directory for the model.

File -> Open cad-file

File = TempDist.egf
Model name = TempDist
Model directory = temp_tutorial

• Select the equations to be solved from the Problem menu. In this case there is only one equation, the
heat equation.

Problem -> Equations

Heat equation

• Apply the body forces from the Model menu. Give the value for the body force (heat source).

130
27. Temperature distribution 131

Γ4

Γ5 Ω

Γ3
Γ6

Γ1

Γ2

Figure 27.1: L-shaped structure

Model -> Body forces

Heat source = 1

• Define the material properties from the Model menu. Give the values for the density and the heat
conductivity.

Model -> Materials

Density = 1
Heat conductivity = 1

• Define boundary conditions from the Model menu. Give the value of the temperature at all boundaries
Γi (i = 1, . . . , 6).

Model -> Boundary conditions

Temperature = 0

• Define mesh for the structure from the Mesh menu. First give name for the mesh and then define the
element size. Press finally the “Generate Mesh” button.

Mesh -> Define mesh

Mesh name = Mesh1
Model Mesh H [m] = 0.1
Generate mesh

• Now the problem may be solved.

Run -> Solver

• After the solution is done, view the results by selecting the Postprocessor from the Run menu.

Run -> Postprocessor

• To save the created model, select Save model file from the File menu.

File -> Save model file

c CSC – IT Center for Science

27. Temperature distribution 132

Table 27.1: Results with different element sizes

Model Mesh H [m] Elements max(T ) [K] cpu time [s]
0.2 187 0.1452 0.25
0.15 330 0.1462 0.28
0.1 735 0.1478 0.38
0.05 2793 0.1487 0.85
0.025 11100 0.1492 2.88
0.01 69654 0.1493 19.37

Results
As a result the maximum temperature in the structure is given. For a comparison the same problem was
solved six times with different element sizes. The maximum temperature obtained by using different meshes
is recorded in Table 27.1. From the results one can see that the result converges. With a denser mesh the
result is more accurate, but solving the problem takes more calculation time. For reference, the central
processor (cpu) time used in each case is also shown in the table.

c CSC – IT Center for Science

Tutorial 28

Loaded elastic beam

Directory: ElasticBeam

28.1 Solution with linear model

Solvers: StressSolve
Tools: ElmerFront
Dimensions: 2D, Steady-state

Case definition
A homogenous, elastic beam (Ω) is rigidly supported on one end (boundary Γ4 ). On boundary Γ3 the beam
is subjected to a load q(x), which grows linearly from zero to q0 (see figure 28.1). Material properties of the
beam are the Poisson ratio 0.3 and Young’s modulus 200 · 109 N/m2 . Problem is to solve the displacement
of the beam.
y
q(x)

Γ3
Ω Γ2
Γ4 x

Γ1

Figure 28.1: Beam and loading.

Problem is solved according to linear elasticity theory. Mathematically the problem to be solved is


 −divσ = 0 in Ω
σ = λtr[ε(u)]I + 2µε(u) in Ω



u = 0 on Γ4 (28.1)
σn = 0 on Γ ∪ Γ

1 2



σn = −q on Γ3


where λ and µ are the Lamé constants (which can be expressed in terms of the Poisson ratio and Young’s
modulus), ε is the linearized strain tensor, u is the displacement vector, q is the given surface traction and n
is the outward unit normal to the boundary.

133
28. Loaded elastic beam 134

Solution procedure
• Start ElmerFront.
• Open the file that contains the geometry of the beam from the File menu. Select also the working
directory for the model.

File -> Open cad-file

File = Beam.egf
Model name = Beam
Model directory = beam_tutorial

• Select the equations to be solved from the Problem menu. In this case stress analysis is selected. It
solves the problem according to linear elastic theory.

Problem -> Equations

Stress analysis

• Define the material properties from the Model menu. Give the values for Young’s modulus and the
Poisson ratio. Add the defined material properties to the material property sets so they become attached
to Ω.

Model -> Materials

Young’s modulus = 200e9
Poisson ratio = 0.3

• Define the Dirichlet boundary condition and the load from the Model menu. Give the value zero for
displacements at the boundary Γ4 and press Add. The linearly varying load is defined in the same
panel as follows. Select the boundary 3, click the cursor on to the Force-y line, check the box Table,
and press finally the Edit button. A window opens in which the tabular bc entry is defined. Select first
Coordinate 1 as the variable on which the Force-y depends. Write on the line below entry 0 0. Click
Add. Write 1 -1.0e7 on the line and click again Add. Now, the space below contains two lines
written in two columns. The first colums holds the values for the Coordinate 1 and the second column
for the Force-y. The value of the force is interpolated according to these definitions. Now click OK on
the Table entry panel. In Boundary Conditions panel, click Add and then OK.

Model -> Boundary conditions

Boundary 4
Displacement-X = 0
Displacement-Y = 0
Add
Boundary 3
Force-Y
Table
Edit
Variable = Coordinate 1
0 0
Add
1 -1e7
Add
OK
Add
OK

• Define mesh from the Mesh menu. First give name for the mesh. Then select “Mesh structure” and
define element type and the number of elements. Attach the defined mesh structure to the Body 1 and
click OK. Create the mesh by pressing “Generate mesh” button.

c CSC – IT Center for Science

28. Loaded elastic beam 135

Mesh -> Define mesh

Mesh name = Mesh1
Mesh structure
Element type = Quad
Nof elements (1st and 3rd edge) = 40
Nof elements (2nd and 4th edge) = 4
Add
OK
Generate mesh

• Now to solve the problem defined with the constant load select from the Run menu item Solver.

Run -> Solver

• Results may be viewed with the ElmerPost program

Run -> Postprocessor

or click the Results button on the main window.

Results
As a result the absolute value of maximum displacement is given. The displacements calculated with differ-
ent load values q0 are tabulated in table 28.1. Note that the absolute value of the displacement varies linearly
with respect to the load since the model is linear.

Figure 28.2: The displacement of an elastic beam with different loads using a linear model

If you look at the results you can see that the displacement values become relatively large. The linear
theory is valid only to small displacements. From Fig 28.2 you can also notice that the beam does not
maintain its original form. This means that the linear elasticity theory can not take into consideration all the
necessary phenomenona that are related to the problem, anymore. To be able to solve the problem we must
use general elasticity theory. This is done in the following subsection.

c CSC – IT Center for Science

28. Loaded elastic beam 136

Table 28.1: Displacements with different load values

q0 [N/m2 ] max |u| [m]

-1.0e7 0.04862
-1.0e8 0.4862
-1.0e9 4.862

28.2 Solution with nonlinear model

Solvers: ElasticSolve
Tools: editor
Dimensions: 2D, Steady-state

Case definition
In the following the beam problem is solved with general elasticity theory. That is done by using the nonlin-
ear elasticity solver of Elmer. In the case of homogenous elastic material the problem can be written into a
following mathematical form


 −div[(I + ∇u)Σ] = 0 in Ω
Σ = λ(trE)I + 2µE in Ω




E = 21 (∇uT + ∇u + ∇uT ∇u) in Ω


 u = 0 on Γ4
(I + ∇u)Σn = 0 on Γ1 ∪ Γ2




(I + ∇u)Σn = −q on Γ3


where u is the displacement vector, q is the given surface load, Σ is the second Piola-Kirchhoff stress tensor,
λ and µ are the Lamé constants and E is the Green-St Venant strain tensor.

Solution procedure
The problem is solved here without the graphical user interface. Open the solver input file of the linear case,
Beam.sif, and edit the following changes. Define nonlinear elasticity as the only equation
Equation 1
Name = "Equation1"
Nonlinear Elasticity = Logical True
End
Change the name correspondingly in the Solver block and add information about the procedure needed.
Leave all other keywords on the solver block unchanged.
Solver 1
Equation = "Nonlinear Elasticity"
Procedure = "ElasticSolve" "ElasticSolver"
...
End
Finally, the force load may be changed in boundary condition 2 as
Force 2 = Variable Coordinate 1
0 0
1 -1.0000e+09
End
The problem may now be solved from the command line by typing ElmerSolver.

c CSC – IT Center for Science

28. Loaded elastic beam 137

Results

Table 28.2: Maximum displacements with different load values calculated according to general elasticity
theory and linear theory.

q0 [N/m2 ] max |u| [m] max |u| (linear) [m]

-1.0e7 0.04890 0.04862
-1.0e8 0.4532 0.4862
-1.0e9 1.297 4.861

From table 28.2 you can see the difference between the results calculated according to nonlinear and
linear theory. According to the linear theory the displacement increases linearly as the load increases. This
can be seen clearly from the results. The last loading level (-1.0e9 N/m2 ) is fairly large and the beam would
probably break under that load. So the value of displacement might be unrealistic in that case.

c CSC – IT Center for Science

Tutorial 29

Incompressible flow passing a step

29.1 Solution with linear triangles

Directory: FlowStepIncompressible
Solvers: FlowSolve
Tools: ElmerFront
Dimensions: 2D, Steady-state

Case definition
A fluid, flowing past a step (see figure 29.1), has the density 1 kg/m and viscosity 0.01 kg/ms. The velocity
of the fluid on the incoming boundary Γ6 in the x-direction is 1 m/s and in the y-direction 0 m/s (see
figure 29.1). On the outcoming boundary Γ4 the velocity is 0 m/s in the y-direction and the pressure is 0 Pa.
The problem is to solve the velocity field and the pressure change in Ω.

Γ5
Γ6 u6,x
Ω Γ4
y Γ1 Γ2

x
Γ3

Figure 29.1: Step.

Mathematically the problem to be solved is

−∇ · (2µε) + ρ~u · ∇~u + ∇p = 0 in Ω
(29.1)
∇ · ~u = 0 in Ω
with the boundary conditions

 ~ux = 1 on Γ6
~ux = 0 on Γi , i = 1, 2, 3, 5 (29.2)
~uy = 0 on Γi , i = 1, . . . , 6,


where µ is the viscosity, ε is the strain tensor, ρ is the density, ~u is the velocity and p is the pressure. It is
assumed that the density and viscosity are constants.

Solution procedure
• Start ElmerFront.

138
29. Incompressible flow passing a step 139

• Open the file that contains the geometry of the step from the File menu. Select also the working
directory for the model.

File -> Open cad-file

File = StepFlow.egf
Model name = StepFlow
Model directory = step_tutorial

• Select the equations to be solved from the Problem menu. In this case we solve the Navier-Stokes
equations.

Problem -> Equations

Navier-Stokes

• Define the material properties from the Model menu. Give the values for the density and the viscosity.

Model -> Materials

Density = 1
Viscosity = 0.01

• Define boundary conditions from the Model menu. Give the values of the velocities at each boundary
Γi . Add the different boundary conditions to boundary condition sets and attach each constraint to a
boundary or boundaries Γi that the constraint concernes (see figure 29.1).

Model -> Boundary conditions

on Γ_6: Velocity-X = 1 and Velocity-Y = 0
on Γ_i, i = 1, 2, 3, 5: Velocity-X = 0 and Velocity-Y = 0
on Γ_4: Velocity-Y = 0

• Define mesh from the Mesh menu. First give name for the mesh and then define the element size.
Create the mesh by pressing “Generate mesh” button.

Mesh -> Define mesh

Mesh name = Mesh1
Model Mesh H [m] = 0.2
Generate mesh

• Now to solve the problem select from the Run menu item Solver. This starts the solver.

Run -> Solver

• After the solution is done, view the results by selecting from the Run menu item Postprocessor.

Run -> Postprocessor

• To save the created model, select from the File menu item Save model file.

File -> Save model file

• To exit Elmer select from the File menu item Exit.

File -> Exit

Results
As a result the maximum pressure difference and maximum velocity is given (see table 29.1). One special
result of interest is the point, on the x-axis, at which the direction of the flow changes. In this case its position
is about 8.3 m.

c CSC – IT Center for Science

29. Incompressible flow passing a step 140

Table 29.1: Pressure difference and velocity

Elements max(∆p) [Pa] max |~u| [m/s]

1426 1.039 1.489

29.2 Solution with 2nd order rectangles

Solvers: FlowSolve
Tools: ElmerGrid, editor
Dimensions: 2D, Steady-state

Case definition
In the following the flow past a step -problem is solved with eight-noded quadrilateral elements. The mesh
is done with ElmerGrid which is a simple mesh generator that can be downloaded via the Elmer internet
pages. Here a grd file is introduced. It contains the geometry data and parameters needed in defining the
mesh of the structure. ElmerGrid transforms this file into Elmer mesh files (mesh.boundary, mesh.nodes,
mesh.header and mesh.elements).

Solution procedure
The problem might be solved using ElmerFront by reading an external mesh into the program but here
instructions for command line usage of Elmer are given.
• First generate mesh with ElmerGrid with the following command.

ElmerGrid 1 2 Step.grd

• Make the necessary changes to the .sif file. Changes are made to header section, boundary conditions
and boundaries. The sif file is conveniently edited using a text editor. The sections should be edited
into the following form

Header
CHECK KEYWORDS Warn
Mesh DB "." "Step"
End

Boundary Condition 1
Name = "Constraint1"
Target Boundaries(1) = 1

Velocity 1 = 1
Velocity 2 = 0
End

Boundary Condition 2
Name = "Constraint2"
Target Boundaries(1) = 3

Velocity 1 = 0
Velocity 2 = 0
End

c CSC – IT Center for Science

29. Incompressible flow passing a step 141

Boundary Condition 3
Name = "Constraint3"
Target Boundaries(1) = 2

Velocity 2 = 0
End

• To solve the problem run the solver by typing Solver.

Results
In Table 29.2 are the results of the problem solved with eight-noded quadrilateral (408) and three-noded
triangular (303) elements.

Element type Elements max(∆p) [Pa] max |~u| [m/s]

408 531 1.182 1.404
303 1416 1.039 1.489

Table 29.2: Pressure difference and maximum velocity with 2nd order rectangles and first order triangles.

When the problem is solved with eight-noded quadrilateral elements the point at which the flowing
direction changes is about 8.5 m.

c CSC – IT Center for Science

Tutorial 30

Transient flow and heat equations - the

Rayleigh-Benard instability

Directory: RayleighBenard
Solvers: HeatSolve, FlowSolve
Tools: ElmerGrid,Editor
Dimensions: 2D, Transient

Case definition
This tutorial is about simulating the developing of the Rayleigh-Benard instability in a rectangular domain
(Figure 30.1) of dimensions 0.01 m height and 0.06 m length. The simulation is performed with water and
the needed material parameters of water are presented in Table 30.1. The temperature difference between
the upper and lower boundary is set to 0.5 so that lower one has the temperature of 283.5 K and the upper
one has the temperature of 283 K.
The density of water is inversely proportional to its temperature. Thus, heated water starts to flow upwards,
and colder downwards due to gravity. In this case we assume that the Boussinesq approximation is valid for
thermal incompressible fluid flow. In other words, the density of the term ρf~ in the incompressible
Navier-Stokes equation can be redefined by the Boussinesq approximation

ρ = ρ0 (1 − β(T − T0 ))

where β is the heat expansion coefficient and the subscript 0 refers to a reference state.

Figure 30.1: Domain.

142
30. Transient flow and heat equations - the Rayleigh-Benard instability 143

Table 30.1: Material parameters.

parameter value
density 1000 kg/m3
viscosity 1040e-6 Ns/m2
heat capacity 4190 J/(kg·K)
heat conductivity 0.6 W/(m·K)
heat expansion coefficient 1.8e-4 K−1
reference temperature 283 K

Solution procedure
The mesh has been constructed in advance and it consists of 2646 elements. The mesh files locate under the
directory Mesh.

Header
Mesh DB "." "Mesh"
Include Path ""
Results Directory ""
End

The simulation is carried out in 2-dimensional cartesian coordinates. Newmark time-stepping method is
selected with 200 steps and with step size of two seconds. In addition, we have to set the value
Transient for the keyword Simulation Type.

Simulation
Coordinate System = "Cartesian 2D"
Coordinate Mapping(3) = 1 2 3
Timestepping Method = "Newmark"
Newmark Beta = 1
Timestep Intervals(1) = 200
Timestep Sizes(1) = 2
Output Intervals(1) = 1
Simulation Type = "Transient"
Steady State Max Iterations = 10
Solver Input File = "RayleighBenard.sif"
Output File = "RayleighBenard.result"
Post File = "RayleighBenard.ep"
End

Only one body is needed during the simulation.

Body 1
Equation = 1
Material = 1
Body force = 1
Initial Condition = 1
End

We assume that the Boussinesq approximation is valid.

Body Force 1
Boussinesq = True
End

c CSC – IT Center for Science

30. Transient flow and heat equations - the Rayleigh-Benard instability 144

In next section the needed initial conditions are defined.

Initial Condition 1
Temperature = 283
Velocity 1 = 1.0e-9
Velocity 2 = 0
End

The equations of the model are coupled and that is why the convection is computed.

Equation 1
Navier-Stokes = True
Heat Equation = True
Convection = "Computed"
End

We must also define the material parameters of water.

Material 1
Density = 1000
Viscosity = 1040e-6
Heat Capacity = 4190
Heat Conductivity = 0.6
Heat Expansion Coefficient = 1.8e-4
Reference Temperature = 283
End

The heat equation is solved first using direct solver.

Solver 1
Equation = "Heat Equation"
Stabilize = Logical True
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-6
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-3
Nonlinear System Relaxation Factor = 1
Linear System Solver = "Direct"
Linear System Iterative Method = "BiCGStab"
Linear System Convergence Tolerance = 1.0e-12
Linear System Max Iterations = 100
Steady State Convergence Tolerance = 1.0e-6
End

The second solver solves the incompressible Navier-Stokes equation. We have used iterative solver this
time. Also all the convergence criterions are defined in the following section.

Solver 2
Equation = "Navier-Stokes"
Stabilize = Logical True
Nonlinear System Max Iterations = 1
Nonlinear System Convergence Tolerance = 1.0e-6
Nonlinear System Newton After Iterations = 3
Nonlinear System Newton After Tolerance = 1.0e-3
Nonlinear System Relaxation Factor = 1
Linear System Solver = "Iterative"

c CSC – IT Center for Science

30. Transient flow and heat equations - the Rayleigh-Benard instability 145

Linear System Iterative Method = "BiCGStab"

Linear System Convergence Tolerance = 1.0e-9
Linear System Max Iterations = 100
Linear System Preconditioning = ILU1
Steady State Convergence Tolerance = 1.0e-05
End

Finally we have to specify the boundary conditions. The temperature difference between the top and
bottom boundary (Boundary Conditions 1 and 3) is set to be 0.5 degrees. The velocities in x- and
y-direction have to be zero in all the boundaries.

Boundary Condition 1
Target Boundaries = 1
Velocity 1 = 0
Velocity 2 = 0
Temperature = 283
End

Boundary Condition 2
Target Boundaries = 2
Velocity 1 = 0
Velocity 2 = 0
End

Boundary Condition 3
Target Boundaries = 3
Velocity 1 = 0
Velocity 2 = 0
Temperature = 283.5
End

Boundary Condition 4
Target Boundaries = 4
Velocity 1 = 0
Velocity 2 = 0
End

Results
Due to the number of the time-steps the simulation may easily take more than ten minutes. After the solver
has finished, the results can be postprocessed with ElmerPost. When opening the result-file it is important
to press the button All which selects all the calculated time steps. A video of the results can be viewed by
selecting the option Timestep Control and pressing the button Loop under the Edit menu.
In Figures 30.2 and 30.3 the obtained temperature distribution and the velocity vectors are presented. We
have also measured the maximum velocity of the flow (Table 30.2). However, more accurate presentation of
the simulation results is presented in Elmer’s example page in the internet address
http://www.csc.fi/elmer/examples/rayleighbenard/index.html.

Table 30.2: Maximum velocity.

maximum velocity 3.98e-4 m/s

c CSC – IT Center for Science

30. Transient flow and heat equations - the Rayleigh-Benard instability 146

Figure 30.2: Temperature distribution at 260 s.

Figure 30.3: Velocity vectors at 260 s.

c CSC – IT Center for Science

Index

acoustic impedance, 57 StatMagSolve, 70

acoustic waves, 68 Stokes equation, 57
AdvectionDiffusion, 117 StreamSolver, 113
Ansys, 80 StressSolve, 15, 43, 80, 132

BEM, 102 tetgen, 10

Boussinesq, 29, 141
Womersley velocity profile, 112
capacitance, 34, 64
compressed row storage, 109

EigenSolve, 43
ElasticSolve, 74, 90, 135
ElectricForce, 64
Electrokinetics, 117
ElmerFront, 68, 129, 132, 137
ElmerGrid, 39, 43, 48, 52, 57, 64, 70, 74, 80, 85, 90,
102, 105, 109, 113, 117, 123, 139, 141
ElmerGUI, 5, 10, 15, 19, 24, 29, 34

FlowSolve, 19, 24, 29, 52, 57, 74, 85, 90, 109, 113,
137, 139, 141
FlowSolve , 117
Fortran 90 compiler, 74, 90, 105, 109
FreeSurfaceReduced, 85

HeatSolve, 5, 10, 29, 39, 52, 80, 123, 129, 141

HelmholtzSolve, 68

KESolver, 24

MEMS, 60, 65
MeshSolve, 74

OpenCascade, 10
OutletCompute, 90

passive elements, 123

PoissonBEMSolver, 102
PoissonSolver, 105

SaveData, 59, 66
SmitcSolver, 48
SolveWithLinearRestriction, 109
StatCurrentSolve, 80
StatElecSolve, 64, 117
StatElecSolver, 34

147

Elmer Models Manual
No ratings yet
Elmer Models Manual
347 pages
BMRS Maths Reas Stats
No ratings yet
BMRS Maths Reas Stats
574 pages
A history of cepstrum analysis and its application to mechanical problems Robert B. Randall download pdf
100% (5)
A history of cepstrum analysis and its application to mechanical problems Robert B. Randall download pdf
43 pages
Elmer Models Manual
No ratings yet
Elmer Models Manual
363 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
186 pages
Elmer Grid Manual
No ratings yet
Elmer Grid Manual
45 pages
Guided Examples Using Elmerfem
No ratings yet
Guided Examples Using Elmerfem
49 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
201 pages
Get Started Elmer
No ratings yet
Get Started Elmer
101 pages
Elmergrid Manual: Peter R Aback CSC - It Center For Science 28 Aug 2020
No ratings yet
Elmergrid Manual: Peter R Aback CSC - It Center For Science 28 Aug 2020
45 pages
GGPS Syllabus 11
No ratings yet
GGPS Syllabus 11
42 pages
Basic Programming With Elmer: 1.1 Calling Convention
No ratings yet
Basic Programming With Elmer: 1.1 Calling Convention
14 pages
Elmerfem-fsic2024
No ratings yet
Elmerfem-fsic2024
15 pages
Solutions Manual Introduction To Finite Elements in Engineering 4th Edition Tirupathi R Chandrupatla Ashok D Belegundu PDF
50% (4)
Solutions Manual Introduction To Finite Elements in Engineering 4th Edition Tirupathi R Chandrupatla Ashok D Belegundu PDF
10 pages
Free Surface Modeling Methods - CFD-101 - FLOW-3D
No ratings yet
Free Surface Modeling Methods - CFD-101 - FLOW-3D
7 pages
Lightweight Aggregates for Civil Engineering - SINTEF Report
No ratings yet
Lightweight Aggregates for Civil Engineering - SINTEF Report
20 pages
Elmer Gui Manual
No ratings yet
Elmer Gui Manual
41 pages
Elmer Overview
No ratings yet
Elmer Overview
10 pages
Elmer, A Computational Tool For Pdes - Application To Vibroacoustics
No ratings yet
Elmer, A Computational Tool For Pdes - Application To Vibroacoustics
7 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
81 pages
ELMER-Guide To FEM Simulations v03 PDF
No ratings yet
ELMER-Guide To FEM Simulations v03 PDF
53 pages
ISO-11479-2-2011
No ratings yet
ISO-11479-2-2011
9 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
65 pages
9 - ING SMP-MTS, PAKET-1
No ratings yet
9 - ING SMP-MTS, PAKET-1
25 pages
Parallel Computing With Elmer
No ratings yet
Parallel Computing With Elmer
54 pages
Elmer Gui Manual
No ratings yet
Elmer Gui Manual
49 pages
Elmer Finite Element Software For Multiphysical Optimization Problems
No ratings yet
Elmer Finite Element Software For Multiphysical Optimization Problems
28 pages
16 - Shear, Torsion, and Multiaxial Testing
No ratings yet
16 - Shear, Torsion, and Multiaxial Testing
10 pages
Real Time Systems - Midterm Exam - Feb 2025
No ratings yet
Real Time Systems - Midterm Exam - Feb 2025
4 pages
ME-ST-EL-001
No ratings yet
ME-ST-EL-001
13 pages
finishing and polishing
No ratings yet
finishing and polishing
19 pages
Elmer Overview
No ratings yet
Elmer Overview
10 pages
Elmer Fem From Pcs To Supercomputers: A 3-Day Course What Is CSC?
No ratings yet
Elmer Fem From Pcs To Supercomputers: A 3-Day Course What Is CSC?
8 pages
Elmer Solver
No ratings yet
Elmer Solver
100 pages
Chemical Sensors: 4B6 Solid State Devices and Chemical/Biological Sensors
No ratings yet
Chemical Sensors: 4B6 Solid State Devices and Chemical/Biological Sensors
45 pages
Lab - Introduction To Finite Element Methods and COMSOL Multiphysics
No ratings yet
Lab - Introduction To Finite Element Methods and COMSOL Multiphysics
5 pages
FF Basic Tutorial
No ratings yet
FF Basic Tutorial
70 pages
Elmer Gui Manual
No ratings yet
Elmer Gui Manual
49 pages
Module Exercise 1 - KTG and Thermodynamics 1677647255263
No ratings yet
Module Exercise 1 - KTG and Thermodynamics 1677647255263
23 pages
2021-SRL-Wu-Statistical Correlation of Seismicity and Geodetic Strain Rate in The Chinese Mainland
No ratings yet
2021-SRL-Wu-Statistical Correlation of Seismicity and Geodetic Strain Rate in The Chinese Mainland
9 pages
Numme Lab: Finite Elements
No ratings yet
Numme Lab: Finite Elements
5 pages
Beyond The Veil: Can The 4th Dimension Explain Spirituality (Original)
No ratings yet
Beyond The Veil: Can The 4th Dimension Explain Spirituality (Original)
5 pages
scheufler_abstract_openfoam_2019
No ratings yet
scheufler_abstract_openfoam_2019
2 pages
Diinsox86 Help Eng
No ratings yet
Diinsox86 Help Eng
98 pages
02 ElmerWebinar Multiphysics
No ratings yet
02 ElmerWebinar Multiphysics
51 pages
Note 1 Introduction To The Research Methodology
No ratings yet
Note 1 Introduction To The Research Methodology
45 pages
Concrete Structures and Construction 5000. Concrete Structures and Construction 5200. Prestressed Concrete 5200. Prestressed Concrete
No ratings yet
Concrete Structures and Construction 5000. Concrete Structures and Construction 5200. Prestressed Concrete 5200. Prestressed Concrete
26 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
152 pages
FPDE User Guide
No ratings yet
FPDE User Guide
77 pages
Finite Elements Using Flex Pde
No ratings yet
Finite Elements Using Flex Pde
12 pages
Elmer Tutorials For SVALI
No ratings yet
Elmer Tutorials For SVALI
152 pages
Electric Fusion Welded UNS N08020, UNS N08024, and UNS N08026 Nickel Alloy Pipe
No ratings yet
Electric Fusion Welded UNS N08020, UNS N08024, and UNS N08026 Nickel Alloy Pipe
5 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
148 pages
Numerical Solution of PDE: Comsol Multiphysics: Example 1, The Poisson Equation On An Ellipse
No ratings yet
Numerical Solution of PDE: Comsol Multiphysics: Example 1, The Poisson Equation On An Ellipse
4 pages
1.units and Measurements
No ratings yet
1.units and Measurements
27 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
103 pages
ELMER - FEM Sif Files and Simulations
No ratings yet
ELMER - FEM Sif Files and Simulations
44 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
148 pages
Datesheet - 2023-24 - 2023-24 - III To IX and XI - PT3 - UT3
No ratings yet
Datesheet - 2023-24 - 2023-24 - III To IX and XI - PT3 - UT3
1 page
Ate 15090371
No ratings yet
Ate 15090371
9 pages
Pragmatism and Logical Positivism
No ratings yet
Pragmatism and Logical Positivism
3 pages
FreeFem - PDE Solver
No ratings yet
FreeFem - PDE Solver
32 pages
Physicsl Lab Manual 12 Supplement
No ratings yet
Physicsl Lab Manual 12 Supplement
6 pages
STAT613
No ratings yet
STAT613
295 pages
Elmer Tutorials
No ratings yet
Elmer Tutorials
96 pages
Elmer: Beoynd Elmergui - About Pre-And Postprocessing, Derived Data and Manually Working With The Case
No ratings yet
Elmer: Beoynd Elmergui - About Pre-And Postprocessing, Derived Data and Manually Working With The Case
30 pages
For Macs, PCS, Linux: Freefem+
No ratings yet
For Macs, PCS, Linux: Freefem+
45 pages
Elmer GUI Tutorials: CSC - IT Center For Science December 19, 2018
No ratings yet
Elmer GUI Tutorials: CSC - IT Center For Science December 19, 2018
96 pages
PHY4 June 2004
No ratings yet
PHY4 June 2004
2 pages
Chapter 6 Thermal Energy Webnotes
No ratings yet
Chapter 6 Thermal Energy Webnotes
2 pages
Introduction To: Elmer FEM Software
No ratings yet
Introduction To: Elmer FEM Software
29 pages
Elmer Programmers Tutorial
No ratings yet
Elmer Programmers Tutorial
14 pages
Lab Report
100% (1)
Lab Report
3 pages
Transient Flow and Heat Equations - The Rayleigh-Benard Instability
No ratings yet
Transient Flow and Heat Equations - The Rayleigh-Benard Instability
5 pages
Speedblue S01: Stainless Steel Aisi 316 Cyclon Filters
No ratings yet
Speedblue S01: Stainless Steel Aisi 316 Cyclon Filters
2 pages
Elmer
No ratings yet
Elmer
4 pages
Menum Lab: Finite Elements
No ratings yet
Menum Lab: Finite Elements
4 pages
Bks MaiSL 1103 ws00 Xxaann
No ratings yet
Bks MaiSL 1103 ws00 Xxaann
2 pages
ENM316 - Finite Element Analysis
No ratings yet
ENM316 - Finite Element Analysis
49 pages
Annual Plan Grade 2
No ratings yet
Annual Plan Grade 2
1 page
Elmer Tut 1
No ratings yet
Elmer Tut 1
8 pages
HVAC Electrical for Idiots
From Everand
HVAC Electrical for Idiots
Brien Hollis
No ratings yet
Introduction to Optical Components
From Everand
Introduction to Optical Components
Simone Malacrida
No ratings yet
Thermal Properties Measurement of Materials
From Everand
Thermal Properties Measurement of Materials
Yves Jannot
No ratings yet
Electronics And Circuit Design Made Easy
From Everand
Electronics And Circuit Design Made Easy
Kerwin Mathew
3/5 (9)
Influence of System Parameters Using Fuse Protection of Regenerative DC Drives
From Everand
Influence of System Parameters Using Fuse Protection of Regenerative DC Drives
Dr. Hedaya Alasooly
No ratings yet
Analog Dialogue, Volume 48, Number 1: Analog Dialogue, #13
From Everand
Analog Dialogue, Volume 48, Number 1: Analog Dialogue, #13
Analog Dialogue
4/5 (1)
Feynman Lectures Simplified 2C: Electromagnetism: in Relativity & in Dense Matter
From Everand
Feynman Lectures Simplified 2C: Electromagnetism: in Relativity & in Dense Matter
Robert Piccioni
No ratings yet
Reference Guide To Useful Electronic Circuits And Circuit Design Techniques - Part 1
From Everand
Reference Guide To Useful Electronic Circuits And Circuit Design Techniques - Part 1
Kerwin Mathew
2.5/5 (3)
Analog Dialogue, Volume 46, Number 2: Analog Dialogue, #6
From Everand
Analog Dialogue, Volume 46, Number 2: Analog Dialogue, #6
Analog Dialogue
No ratings yet