Mesh motion alternatives in OpenFOAM

CFD with OpenSource Software, Assignment 3

Andreu Oliver Gonzlez December 2009, Gteborg (Sweden)

Reviewed by: Aurlia Vallier Hkan Nilsson

Mesh motions alternatives

Andreu Oliver Gonzlez

INDEX
INDEX 1 INTRODUCTION 2 MESH MOTION APPROACHES AND THE DIFFERENT CLASSES 3 PROCEDURE FOR DEFINING A MESH WITH MOTION 3.1 Mesh motion method 3.1.1 Automatic mesh motion (dynamicFvMesh) 3.1.2 Topological changes in the mesh (topoChangerFvMesh) 3.2 Appropriate solver 3.3 Diffusivity model 4 dynamicInkJetFvMesh 4.1 Explanation of the dynamicInkJetFvMesh class 4.2 Example of use of the dynamicInkJetFvMesh class 4.3 Modification of the class dynamicInkJetFvMesh to dynamicMyClassFvMesh 4.4 Example of use of the dynamicMyClassFvMesh class 5 CONCLUSIONS 6 REFERENCES APPENDIX blockMeshDict system folder codes dynamicMeshDict dynamicMyClassFvMesh.C icoDyMFoamMesh.C 1 2 2 3 4 4 5 5 6 7 7 9 18 21 24 25 26 26 27 27 28 30

Mesh motions alternatives

Andreu Oliver Gonzlez

1 INTRODUCTION
The aim of this project is to study different alternatives of mesh motion in OpenFOAM. An overview of the different mesh motion classes is presented. The purpose of this is to give some information for the reader to be able to select an appropriate class for each specific situation. After this overview of the different classes available for mesh motion, a deep description is carried out for the dynamicInkJetFvMesh class and a modification of this class is done. The solver used for problems with moving meshes is the IcoDyMFoam solver (DyM: Dynamic Mesh). This is a transient solver for incompressible and laminar flow. In the case of turbulent flow the turbDyMFoam can be used. Those solvers are used in version 1.5.x of OpenFOAM. In the 1.6.x version, they have both been collected in the pimpleDyMFoam solver. [1]

2 MESH MOTION APPROACHES AND THE DIFFERENT CLASSES

There are two mesh manipulation approaches in OpenFOAM; the difference between them is the topology changing during the simulation or not. These two types are named dynamicFvMesh and topoChangerFvMesh of which the second includes topological changes. Each approach includes different classes and they are the ones that follow [6]: dynamicFvMesh: automatic mesh motion, for the case where the mesh topology does not change. There are five sub-classes for this class: 1) staticFvMesh, where the mesh has no motion. 2) dynamicMotionSolverFvMesh, which is used in cases where the motion of interval mesh points are solved for using boundary conditions and diffusivity models. It is the simplest type of mesh motion solver. 3) dynamicInkJetFvMesh, which is similar to the one before, but in this case the mesh points are explicitly specified instead of solved for. 4) dynamicRefineFvMesh, it is similar to the staticFvMesh class but in this case a refinement or unrefinement of the mesh in the three directions is carried out by adding or removing points. 5) solidBodyMotionFvMesh, which is used to describe a solid body motion of the mesh specified by a run-time selectable motion function.

Mesh motions alternatives

Andreu Oliver Gonzlez

topoChangerFvMesh: topological mesh changes, when the mesh topology changes during the simulation. There are four types of sub-classes for this class [6][8]: 1) linearValveFvMesh, which uses sliding meshes at the interface of two pieces of mesh in relative linear motion. The dictionary linearValveFvMeshCoeffs is used to select the motion solver type for mesh handling. 2) linearValveLayersFvMesh, which is similar to the class before but layer addition and removal is the extra feature instead of pure squeezing or stretching of the nodes and cells. The input variables needed are the same as the ones for the sub-class presented before and, moreover, the ones found in the extra subdictionary layer. 3) mixerFvMesh, which is used when a sliding interface is needed between one rotating part and a fixed one. A part from the dynamicMeshDict, the dictionary MRFZones is important because it is where the moving parts are determined. From this dictionary, different zones are generated and those are used by the slidingInterface class which gives the relative motion between the two sides of the sliding interface. 4) movingConeTopoFvMesh, which applies squeezing and stretching of the cells, but when cell layer thickness reaches a critical value a new cell layer is added or an old cell layer is removed. A part from the dynamicMeshDict file and the extra dictionary file meshModifiers inside constant folder, in the movingConeTopoFvMesh.C is required a sub-dictionary to specify the coefficients to define the moving and fixed boundaries and characteristics, besides the minimum and maximum cell layer thickness in each region.

3 PROCEDURE FOR DEFINING A MESH WITH MOTION

The structure of files in order to solve this kind of applications is the usual ones, where you find a folder with the initial values (0) and two folders, which are: constant, where it can be found some files and folders such as a dynamicMeshDict file, a transportProperties file and a polyMesh folder. system, where it is found a controlDict file, an fvSchemes file and an fvSolution file, but also other files are found depending on the approach followed for the mesh motion. First of all, the mesh has to be defined in the blockMeshDict file inside the constant folder. In order to have mesh motion in any direction, for some classes the boundary type should 3

Mesh motions alternatives

Andreu Oliver Gonzlez

be set to patch for the moving and changing cells in the direction where motion can be defined. Then, moving-mesh boundary conditions have to be specified to allow the movement in the desired direction. A part from that, a dynamicMeshDict file has to be added inside the constant folder, where the different definitions used and needed for the moving mesh are specified (mesh manipulation dictionaries, solvers, classes, diffusivities and coefficients required for the case).

3.1 Mesh motion method

As introduced before, in the dynamicMeshDict file some definitions have to be described, but, first of all, knowing the desired motion, the mesh motion method (dynamicFvMesh or topoChangerFvMesh) has to be chosen and also the used sub-class. If none of the sub-classes already provided by OpenFOAM fits with the desired one, a modification of one of the existing ones or the creation of a new one has to be carried out. 3.1.1 Automatic mesh motion (dynamicFvMesh) The mesh motion is given in different ways depending on the class. For the dynamicInkJetFvMesh, the mesh points are specified generally by an equation, so only inside the sub-class, the motion equation has to be added. For the dynamicRefineFvMesh, the points added or removed should be specified in the sub-class as well as the direction where the refinement or unrefinement is desired to take place. In the case of the dynamicMotionSolverFvMesh, a solver is needed to solve a mesh motion equation, where boundary motion acts as a boundary condition and determines the position of mesh points. The motion is characterized by the spacing between nodes, which changes by stretching and squeezing. This mesh motion equation can be simplified, and there are mainly four types [11]: Spring analogy, which is insufficiently robust. Linear plus torsional spring analogy, which is complex, expensive and non-linear. Laplace equation with constant and variable diffusivity. Linear pseudo-solid equation for small deformations.

The mesh spacing and quality is controlled by variable diffusivity (3.3 Diffusivity model). Changing the diffusivity implies redistribution of the boundary motion through the volume of the mesh. And the definition of valid motion from an initially valid mesh implies that no forces or cells are inverted during motion, which helps to preserve the mesh quality.

Mesh motions alternatives

Andreu Oliver Gonzlez

The

corresponding

library

for

this

mesh

manipulation

approach

is

the

libDynamicFvMesh.so. 3.1.2 Topological changes in the mesh (topoChangerFvMesh) The number of points, faces, cells and/or mesh connectivity changes during simulation. It is used for more demanding and complex mesh motion than the automatic approach where the original topology cannot be kept or the precision of the solution would be affected by keeping the original mesh settings during the simulations. For that, mesh modifiers are required to describe what kind of mesh manipulation action is carried out [7][8]: Attach or detach of boundary. Layer addition or removal. Sliding interface.

The class polyTopoChanger will look for the necessary data and extract it from the extra dictionary meshModifiers, otherwise, the data will be read from the dynamicMeshDict. The corresponding dynamic library is libtopoChangerFvMesh.so. Once the method for mesh motion and the sub-class is chosen, they have to be specified in the dynamicMeshDict file, as mentioned before.

3.2 Appropriate solver

It has to be pointed out that not always a solver is required because sometimes the motion is described by an equation inside the class definition and it is solved internally, as it was mentioned before. In the cases where a solver is needed, once the mesh is set, the moving points of the grid require models and corresponding mesh motion equations to be solved. The most used ones are [5]: displacementLaplacian, the equations of cell motion are solved based on the Laplacian of the diffusivity and the cell displacement (pointDisplacement extra file is required in the starting time folder). For this solver, the final displacement of the mesh components is needed as well as the mesh displacement of the internal field. velocityLaplacian, similar to the previous solver with the difference being the equation solved, which is the Laplacian of the diffusivity and the cell motion velocity (pointMotionU file has to be available to be read). A part from the input variables that are the same as those in the displacementLaplacian case, the user has to be 5

Mesh motions alternatives

Andreu Oliver Gonzlez

aware that the code deals with the boundary velocities instead of the final motions, so care have to be taken when determining the dimensions. It is used when each time an order of magnitude of the maximum displacement is known to be not too big. LaplaceFaceDecomposition, used when the order of magnitude of the maximum displacement is not known or known to be big. The mesh is rebuilt after a decomposition of all cells and faces and the Laplace smoothing equation is solved by the Finite Element Method. It increases the robustness but, on the other hand, it increases the computational cost compared to the velocityLaplacian solver. SBRStress, it is a displacement model, solving Laplacian of diffusivity and the cellDisplacement and it considers also the solid body rotation term in calculations (pointDisplacement file is also required in the constant directory).

3.3 Diffusivity model

As said for the solvers, for the classes where the motion is solved internally in the mesh class, the diffusivity model is not needed. The diffusivity model is used to determine how the points should be moved when solving the cell motion equation for each time step. There are two groups of diffusivity models [5]: 1) Quality based methods, where the diffusion field is a function of a cell quantity measure. There are four types and they are the following ones: a. uniform, where the mesh manipulation is done uniformly for all moving boundaries by stretching or squeezing with the same ratio all the cells in each region. b. directional, where the mesh stretching or squeezing is done proportionally to the direction of the motion. The main idea in this case is that the mesh manipulation is done by considering the slipping boundaries. Two scalar coefficients are required, one defining the mean cell non orthogonality and the other one to determine the mean cell skewness. c. motionDirectional, where the mesh manipulation is done by prioritizing the moving body and adjusting the cells in a way that is more appropriate for the moving body. The same coefficients than for the above method have to be specified. d. inverseDistance, where the user specifies one or more boundaries and the diffusivity of the field is based on the inverse of the distance from that boundary. 6

Mesh motions alternatives

Andreu Oliver Gonzlez

2) Distance based methods, used together with the quality based methods and in which the diffusion field will be a function of the inverse of cell centre distance l to the nearest selected boundary. There are three of them, which are: a. linear, the diffusivity field is based linearly on the inverse of the cell center distance to the nearest boundary. b. quadratic, as the one above except being a quadratic relation instead of a linear one. c. exponential, in this case the diffusivity of the field is based on the exponential of the inverse of cell-center distance to the selected boundaries.

4 dynamicInkJetFvMesh
In my master thesis I will model the vertebral column to make CFD simulations with the purpose of evaluating and studying the whiplash pain causes. For that reason, the dynamicInkJetFvMesh is here studied deeply. This class is the appropriate one because the motion of the model will be given from FEM simulations. Although the dynamicInkJetFvMesh class is the most suitable class, some changes will have to be done on it. In this project only a simple modification will be done in order to get a better understanding of the class and how to make motion modifications in that particular class.

4.1 Explanation of the dynamicInkJetFvMesh class

The most important parts of the code of dynamicInkJetFvMesh.C are provided below with some comments to understand how it works:
00036 defineTypeNameAndDebug(dynamicInkJetFvMesh, 0);

It calls the functions typeName and debug to specify the type class used, which is dynamicInkJetFvMesh in this case, and some information for debugging.
00037 IOobject) }; addToRunTimeSelectionTable(dynamicFvMesh, dynamicInkJetFvMesh,

It adds the dynamicInkJetFvMesh (which is thisType, dynamicInkJetFvMesh, inside the baseType, dynamicFvMesh) to the table where the classes used are defined. In the following part of the code, the sub-class is defined and it is said where to read this information as well as the coefficients needed for this sub-class (amplitude, frequency and refPlaneX):

Mesh motions alternatives

Andreu Oliver Gonzlez

00041 // * * * * * * * * * * * * * * * * Constructors * * * * * * * * * * * * * * // 00043 Foam::dynamicInkJetFvMesh::dynamicInkJetFvMesh(const IOobject& io) 00044 : 00045 dynamicFvMesh(io), 00046 dynamicMeshCoeffs_ 00047 ( 00048 IOdictionary 00049 ( 00050 IOobject 00051 ( 00052 "dynamicMeshDict", 00053 io.time().constant(),

From above, it can be seen that the dynamicMeshDict file is located in the folder constant.
00054 00055 00056 00057 00058 00059 *this, IOobject::MUST_READ, IOobject::NO_WRITE

),

) ).subDict(typeName + "Coeffs")

From above, a subdictionary called dynamicInkJetFvMeshCoeffs exists inside the dynamicFvMesh with the following scalar numbers:
00060 00061 00062 00063 00064 00065 00066 00067 00068 amplitude_(readScalar(dynamicMeshCoeffs_.lookup("amplitude"))), frequency_(readScalar(dynamicMeshCoeffs_.lookup("frequency"))), refPlaneX_(readScalar(dynamicMeshCoeffs_.lookup("refPlaneX"))), stationaryPoints_ ( IOobject ( "points", io.time().constant(),

Just above, it can be seen that the file Points is also located in the folder constant.
00069 00070 00071 00072 00073 00074 00075 { 00076 00077 00078 00079 00080 } meshSubDir, *this, IOobject::MUST_READ, IOobject::NO_WRITE

) ) Info<< << << <<

"Performing a dynamic mesh calculation: " << endl "amplitude: " << amplitude_ " frequency: " << frequency_ " refPlaneX: " << refPlaneX_ << endl;

In the next part of the code, the mesh points are updated by an equation, from which the coordinate x is calculated after using a scaling function:
00082 // * * * * * * * * * * * * * * * * Destructor * * * * * * * * * * * * * * * // 00084 Foam::dynamicInkJetFvMesh::~dynamicInkJetFvMesh() 00085 {}

Mesh motions alternatives

Andreu Oliver Gonzlez

00088 // * * * * * * * * * * * * * * * Member Functions * * * * * * * * * * * * * // 00090 bool Foam::dynamicInkJetFvMesh::update() 00091 { 00092 scalar scalingFunction = 00093 0.5*(::cos(2*mathematicalConstant::pi*frequency_*time().value()) 1.0);

Which its mathematical notation is scaling_function = 0.5(cos (2tf) 1):

00095 00096 00097 00098 Info<< "Mesh scaling. Time = " << time().value() << " scaling: " << scalingFunction << endl; pointField newPoints = stationaryPoints_;

Above, the newPoints matrix is given the values of the stationaryPoints one. And its values are replaced using the following equation.
00100 00101 00102 00103 00104 00105 00106 00107 00108 00109 00110 00111 00112 newPoints.replace ( vector::X, stationaryPoints_.component(vector::X)* ( 1.0 + pos ( - (stationaryPoints_.component(vector::X)) - refPlaneX_ )*amplitude_*scalingFunction ) );

With the function replace, the new points are recalculated following the motion equation described just above. With vector::X specification it is said that the motion is only changing the mesh in one direction, in this case in the X direction. Its mathematical notation is x = xold(1 + pos(-xold - refPlaneX)amplitudescaling_function)
00113 00114 fvMesh::movePoints(newPoints);

Mesh points are moved to the new points calculated. In order to get a better idea of how it works an example is developed to show it.

4.2 Example of use of the dynamicInkJetFvMesh class

The example is going to show the motion of a very simple mesh by using the icoDyMFoam solver, but only using the mesh manipulation of that solver. To start to make the example, create the example folder:
>> mkdir $WM_PROJECT_USER_DIR/myExample

This folder has to have the following structure with the following three folders: 9

Mesh motions alternatives

Andreu Oliver Gonzlez

0 p U dynamicMeshDict. polyMesh, where the blockMeshDict is located. transportProperties. controlDict. fvSchemes. fvSolution.

constant -

system -

All the codes for the files specified are available and this structure can be seen in course homepage [1] under my name in the link called files. First of all a simple mesh is defined in the blockMeshDict (code attached in the Appendix, like also the code for controlDict); the geometry chosen is a long and thin rectangle (0.006x0.075x0.001m) which is fixed from the bottom part, shown in Figure 1. The mesh is defined in the negative side of the x axis, which means that it goes from x = -0.006 until x = 0.

10

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 1: Mesh geometry.

Once the mesh is defined, the utility to generate the mesh is called:
>> blockMesh

Then, the file dynamicMeshDict has to be created inside the constant folder where the class for the for mesh manipulation (amplitude, is specified and and where plane of the subdictionary The dynamicInkJetFvMeshCoeffs has to be described with the values of the coefficients required
{ version format class object 2.0; ascii; dictionary; motionProperties;

the

class

frequency

reference).

dynamicMeshDict file is shown below:

} // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // dynamicFvMesh dynamicInkJetFvMesh; motionSolverLibs ("libfvMotionSolvers.so"); dynamicInkJetFvMeshCoeffs { amplitude 0.06; frequency 2; refPlaneX 0; } // ************************************************************************* //

11

Mesh motions alternatives

Andreu Oliver Gonzlez

The motion defined with this class makes the set of points to be compressed and expanded sinusoidally to impose a sinusoidal variation (Eq. 1 & 2 with the code):
scalingFunction = 0.5*(::cos(2*mathematicalConstant::pi*frequency_*time().value()) - 1.0);

scaling_function = 0.5(cos (2tf) 1)

Eq. 1

stationaryPoints_.component(vector::X)*(1.0 + pos((stationaryPoints_.component(vector::X)) refPlaneX_)*amplitude_*scalingFunction)

x = xold(1 + pos(-xold - refPlaneX)amplitudescaling_function) The coefficients, as seen in the code from dynamicMeshDict, are set to: Amplitude = 0.06 Frequency = 2

Eq. 2

Reference_Plane_X = 0

At this point, the mesh can be moved. As it is going to be used the icoDyMFoam solver to do that, the parts from the solver that are not used to manipulate the mesh are going to be deleted and therefore the solver is renamed as icoDyMFoamMesh. For that, the following steps should be followed:
>> cp r $FOAM_SOLVERS/incompressible/icoDyMFoam \ $WM_PROJECT_USER_DIR/icoDyMFoamMesh >> cd icoDyMFoamMesh >> wclean >> mv icoDyMFoam.C icoDyMFoamMesh.C

The parts from icoDyMFoamMesh.C that can be deleted are: Make the fluxes absolute. Make the fluxes relative to the mesh motion. Pimple loop.

The file code is available in the website provided before and also in the Appendix, where it can easily be seen which parts have to be deleted. After those modifications, the solver should be recompiled and to be able to do it the file Make/files should be as follows:
icoDyMFoamMesh.C EXE = $(FOAM_USER_APPBIN)/icoDyMFoamMesh

Now, the solver can be compiled:

>> wmake

12

Mesh motions alternatives

Andreu Oliver Gonzlez

Finally by calling the solver, the mesh manipulation can be seen in paraFoam:
>> icoDyMFoamMesh >> paraFoam

The example is used with different values for the three constants and the results achieved are shown in the figures below:

Figure 2: Initial position where motion is still not applied (t = 0s).

Figure 2 is showing the mesh in its initial position for all the cases that have been run, at t = 0s. And the arrow shown appears in all the figures to have the initial width of the mesh, which is 0.006m, and therefore have a reference to appreciate the mesh motion. In Figure 3, it is shown the mesh motion for the case using the constants defined before and it will be the reference to analyze how the three constants influence the mesh motion.

13

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 3: Mesh motion with 0.06m of amplitude, 2Hz of frequency and 0m of refPlaneX for t = 0.1, 0.25, 0.3 and 0.4s.

Figure 4: Mesh motion with 0.03m of amplitude, 2Hz of frequency and 0m of refPlaneX for t = 0.1, 0.2, 0.25 and 0.4s.

14

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 5: Mesh motion with 0.06m of amplitude, 4Hz of frequency and 0m of refPlaneX for t = 0.05, 0.125, 0.25 and 0.375s.

Figure 6: Mesh motion with 0.06m of amplitude, 2Hz of frequency and -0.003m of refPlaneX for t = 0.1, 0.25, 0.3 and 0.4s.

15

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 7: Mesh motion with 0.06m of amplitude, 2Hz of frequency and 0.003m of refPlaneX for t = 0.1, 0.15, 0.2, 0.25, 0.3 and 0.4s.

In Figure 7, it can be seen that the mesh connectivity changes; this can be observed by the third vertical line which end up between the 4th and 5th line at t = 0.25s. 16

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 8: Mesh motion with 0.06m of amplitude, 2Hz of frequency and 0.006m of refPlaneX for t = 0.1, 0.2, 0.25 and 0.4s.

Observing Figures 2-8 above and analyzing the equation of motion (Eq. 2), it can be seen that the three constants modify the motion in the following way: - amplitude: varies the length the mesh is deformed in the x direction. Looking at Figure 3 and 4, it can easily be seen how for the same time steps the position of the left side of the mesh has moved less; at t = 0.25s, when the maximum displacement takes place is the double in the Figure 3 because the amplitude value is the double. - frequency: modifies the number of periods for the same total time and therefore varies the speed of change. Comparing Figure 3 and 5, it can be seen that with the same time (0.5s) the sinusoidal motion in Figure 5 has been done twice and that is because the frequency is the double that configuration. In order to see it easily, equation (Eq. 1) has been analyzed; as the scaling_function is a cosinus function which is multiplied by 0.5 after 1 is substracted to it, therefore it goes from -1 to 0 which makes the motion to follow the sinus shape, as shown in the Figure 9:

17

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 9: Scaling function for 2Hz and 4Hz of frequency from 0 to 0.5s.

- refPlaneX: from Eq. 2, it can be noted that the sinusoidal motion is with respect to refPlaneX. But it affects in different ways depending on the intervals where it is located, as it can be seen in Figures 6-8. Comparing the different figures, it can be seen that Figure 6 is equal to Figure 3. Then, in Figure 7, only half of the mesh is being moved, which is because the refPlaneX is defined as 0.003 (half of the mesh width). And Figure 8 shows how mesh is moved having refPlaneX as 0.006, where only the mesh points of the left side are moving. Using the information above and some other values that were used, the refPlaneX intervals can be defined; all considering that the mesh defined is going in x direction from -0.006 until 0: - For refPlaneX [,0] , the values given by pos function are 1 for all the points of the mesh; therefore, the mesh motion will be the same for this interval of refPlaneX values. - For refPlaneX (0,0.006] , the values given by pos function are 1 only for the points that has an x position smaller than refPlaneX; therefore, only these points are moved, while the rest are kept in the initial position. - For refPlaneX (0.006,] , the values given by pos function are 0 for all the points of the mesh; so there is no motion.

4.3 Modification of the class dynamicInkJetFvMesh to dynamicMyClassFvMesh

In this part of the project it is going to be shown how to modify the class dynamicInkJetFvMesh with the purpose to define the desired motion. The new class it is called dynamicMyClassFvMesh where a polynomial motion is going to be carried out; shown in Figure 10.

18

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 10: Motion schema for the new class.

The equation to define this motion is the one that follows: x = aty2 + b Eq. 3

where x is the displacement in the x direction and y is the position from the bottom part of the geometry. It is dependent on time to see the movement step by step, where time will be going from 0 to 1s. As the bottom part is fixed: for y 0 x 0 b 0

Defining a displacement in the top part, for example 10cm, when t = 1s: for y 0.75 x 0.1 a 0.1778

Then the coordinate x is updated with the next function: x = xold + aty2 Eq. 4

A scaling function has been added (Eq. 5) to provide a more complex motion giving then a displacement in x direction but in both sides. Therefore the updating function for the x coordinate is shown below: scaling_function = cos (2tf) x = xold + aty2scaling_function Eq. 5 Eq. 6

In order to change the class, first the new class have to be created from the existing one: 19

Mesh motions alternatives

Andreu Oliver Gonzlez

>> cp r $FOAM_SRC/dynamicFvMesh/dynamicInkJetFvMesh \ $WM_PROJECT_USER_DIR/dynamicMyClassFvMesh >> cd $WM_PROJECT_USER_DIR/dynamicMyClassFvMesh >> sed s/dynamicInkJetFvMesh/dynamicMyClassFvMesh/g <dynamicInkJetFvMesh.C \ >dynamicMyClassFvMesh.C >> sed s/dynamicInkJetFvMesh/dynamicMyClassFvMesh/g <dynamicInkJetFvMesh.H \ >dynamicMyClassFvMesh.H >> rm r dynamicInkJetFvMesh.* >> cp r $FOAM_SRC/dynamicFvMesh/Make $WM_PROJECT_USER_DIR/dynamicMyClassFvMesh

At this point the new class has been created but only by changing the names from the original dynamicInkJetFvMesh class, therefore, it has to be compiled. To compile the dynamicMyClassFvMesh class, the files file and the options file have to be modified: files, rewritten as follows to only compile the dynamicMyClassFvMesh library:
dynamicMyClassFvMesh.C LIB=$(FOAM_USER_LIBBIN)/libdynamicMyClassFvMesh

options, the next line has been added to include the files from the original library:

-I$(LIB_SRC)/dynamicFvMesh/lnInclude

Now the compilation can be done:

>> cd $WM_PROJECT_USER_DIR/dynamicMyClassFvMesh >> wmake libso

When the compilation is done properly, then the modification can be done. This step defined just above, it is only done to ensure that the modification of name is done properly. To modify the motion equation, dynamicMyClassFvMesh.C has to be modified in the part where the equation is defined, red colour shows the modified code:
00060 00061 00062 00077 00078 00079 00092 00093 a_(readScalar(dynamicMeshCoeffs_.lookup("a"))), frequency_(readScalar(dynamicMeshCoeffs_.lookup("frequency"))), // refPlaneX_(readScalar(dynamicMeshCoeffs_.lookup("refPlaneX"))), << "a: " << a_ << " frequency: " << frequency_ << endl; // << " refPlaneX: " << refPlaneX_ << endl; scalar scalingFunction = (::cos(2*mathematicalConstant::pi*frequency_*time().value());

00100 newPoints.replace 00101 ( 00102 vector::X, 00103 stationaryPoints_.component(vector::X)+ 00104 a-*time().value()*(stationaryPoints_.component(vector::Y))* (stationaryPoints_.component(vector::Y))*scalingFunction 00105 );

The complete code of it is shown in the Appendix. The class has to be recompiled by typing:
>> wmake libso

20

Mesh motions alternatives

Andreu Oliver Gonzlez

4.4 Example of use of the dynamicMyClassFvMesh class

The example is the same that was done in part 4.2, but now it is going to show the motion of the mesh by using the new class. The example defined in part 4.2 can be copied:
>> cp r $WM_PROJECT_USER_DIR/myExample $WM_PROJECT_USER_DIR/myClassExample

But some changes have to be done in controlDict (where now the endTime is 1s, code attached in the Appendix) and dynamicMeshDict: reference to the new class library the needed coefficients (a and frequency) have to be redefined in the subdictionary dynamicMyClassFvMeshCoeffs inside the dynamicMeshDict. The values for the coefficients have been taken: a = 0.4
{ version format class object 2.0; ascii; dictionary; motionProperties;

frequency = 2

} // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // dynamicFvMeshLibs (libdynamicMyClassFvMesh.so); dynamicFvMesh dynamicMyClassFvMesh; motionSolverLibs ("libfvMotionSolvers.so"); dynamicMyClassFvMeshCoeffs { a 0.4; frequency 2; } // ************************************************************************* //

Now, the mesh can be moved and icoDyMFoamMesh is going to be used. By calling the solver, the mesh manipulation can be seen in the paraFoam:
>> icoDyMFoamMesh >> paraFoam

As for the example shown for the original class, different values for the two constants are defined and the results achieved are shown in the Figures 11-13:

21

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 11: Mesh motion with 0.4m of amplitude and 2Hz of frequency for t = 0.05, 0.1, 0.15, 0.2, 0.25, 0.4 and 0.5s.

Figure 12: Mesh motion with 0.8m of amplitude and 2Hz of frequency for t = 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4 and 0.5s.

22

Mesh motions alternatives

Andreu Oliver Gonzlez

Figure 13: Mesh motion with 0.4m of amplitude and 4Hz of frequency for t = 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45 and 0.5s.

It can be seen that the two constants modify the motion in the following way: - a: varies the total displacement in the x direction. Comparing Figures 11 and 12, it can be observed how the displacement of the top part of the mesh in the last time step (0.5s) for a equal to 0.8 is bigger than the one for a with a value of 0.4. - frequency: varies the speed of change; depending on the value it can be needed to change the write interval in the controlDict file to see the results of the change in paraFoam. Comparing Figures 11 and 13, where Figure 13 has a frequency which is the double of the one for the case solved in Figure 11, it can be observed that the mesh is making more fluctuations around the vertical axis for the same time. The frequency constant is used inside the scaling function (Eq. 5) and it is similar to the one used for the dynamicInkJetFvMesh class but in this case it is going from -1 to 1, therefore, plotting the x position without considering the dependency on the y position yields Figure 14:

23

Mesh motions alternatives

Andreu Oliver Gonzlez

0.4tscaling_function

0.4tscaling_function

t [s]

t [s]

Figure 14: 0.4tscaling_function for 2Hz and 4Hz of frequency and time from 0 to 0.5 s.

From Figure 14, it can be seen that the important part of the motion applied to the mesh is a sinusoidal movement and its amplitude is increasing with time. As the motion is dependent on y, the displacement for the x coordinates is the product shown in Figure 14 multiplied by the y coordinate squared therefore for the bottom points the movement is zero while for the top points the motion is the highest one.

5 CONCLUSIONS
It can be concluded that in order to have a mesh with motion there are two main ways to follow: The automatic mesh motion (dynamicFvMesh) with which the mesh topology does not change. The topological changes in the mesh (topoChangerFvMesh). Inside these two options of mesh motion manipulation, there are different classes with different motions specified. It has to be added that dynamicInkJetFvMesh defines a movement based on harmonic motion around a reference plane solved internally in the class, which means that an external solver is not needed to calculate the x coordinates after the motion. The modification of this class, dynamicMyClassFvMesh, defines another motion, a sinusoidal one along x direction depending on y position. Finally, as it has been seen with the dynamicInkJetFvMesh from the dynamicFvMesh, when the motion specified originally is not the demanded by the user, the class can be modified in order to define the desired motion.

24

Mesh motions alternatives

Andreu Oliver Gonzlez

6 REFERENCES
[1] Hkan Nilsson (2009-09). PhD course in CFD with OpenSource software, 2009. Slides from the homepage of the course, which is given at Chalmers TH (Gteborg, Sweden). Homepage: www.tfd.chalmers.se/~hani/kurser/OS_CFD_2009/ [2] OpenFOAM website The Open Source CFD Toolbox. Retrieved November 2009 from: www.opencfd.co.uk/openfoam/ [3] OpenFOAM Wiki. Retrieved November 2009 from: www.openfoamwiki.net [4] CFD Online Forums about OpenFOAM. Retrieved November 2009 from: www.cfdonline.com [5] Hrvoje Jasak and Henrik Rusche (2009-06). Dynamic Mesh Handling in OpenFOAM. Slides from the 4th OpenFOAM workshop (Montreal, Canada). [6] Pirooz Moradnia (2008). A tutorial on how to use Dynamic Mesh solver IcoDyMFoam. Report for the PhD course in OpenFOAM at Chalmers TH (Gteborg, Sverige). [7] Olivier Petit (2008). Different ways to treat rotating geometries. Report for the PhD course in OpenFOAM at Chalmers TH (Gteborg, Sverige). [8] Erik Bjerklund (2009). A modification of the movingConeTopoFvMesh library. Report for the PhD course in OpenFOAM at Chalmers TH (Gteborg, Sverige). [9] The OpenFOAM The Open Source CFD Toolbox. Retrieved November 2009 from: http://foam.sourceforge.net [10] The SfR Fresh Software Archive. Retrieved November 2009 from: www.sfr-fresh.com [11] Christophe Kassiotis (2008). Which strategy to move the mesh in the Computational Fluid Dynamic code OpenFOAM. Report for the PhD course in OpenFOAM at Chalmers TH (Gteborg, Sverige). 25

Mesh motions alternatives

Andreu Oliver Gonzlez

APPENDIX
The most important and used codes for this project are presented now [9][10]: blockMeshDict The code where the mesh is defined:
FoamFile { version 2.0; format ascii; class dictionary; object blockMeshDict; } // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // convertToMeters 0.1; vertices ( (-0.006 0 0) (0 0 0) (0 0.075 0) (-0.006 0.075 0) (-0.006 0 0.001) (0 0 0.001) (0 0.075 0.001) (-0.006 0.075 0.001) ); blocks ( hex (0 1 2 3 4 5 6 7) (4 50 1) simpleGrading (1 1 1) ); edges ( ); patches ( wall movingWall ( (3 7 6 2) (0 4 7 3) (1 2 6 5) ) wall fixedWalls ( (1 5 4 0) ) empty frontAndBack ( (0 3 2 1) (4 5 6 7) ) ); mergePatchPairs (

26

Mesh motions alternatives

);

Andreu Oliver Gonzlez

// ************************************************************************* //

system folder codes controlDict For myExample:

FoamFile { version 2.0; format ascii; class dictionary; location "system"; object controlDict; } // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // application icoFoam; startFrom startTime; startTime 0; stopAt endTime; endTime 0.5; deltaT 0.005; adjustTimeStep no; //added maxCo 0.5; writeControl timeStep; writeInterval 5; purgeWrite 0; writeFormat ascii; writePrecision 6; writeCompression uncompressed; timeFormat general; timePrecision 6; runTimeModifiable yes; // ************************************************************************* //

For myClassExample, endTime is 1. dynamicMeshDict The code where the mesh class and the library for the class is specified. For myExample:
FoamFile { version 2.0; format ascii; class dictionary; object motionProperties; } // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // dynamicFvMesh dynamicInkJetFvMesh; motionSolverLibs ("libfvMotionSolvers.so"); dynamicInkJetFvMeshCoeffs {

27

Mesh motions alternatives

amplitude frequency refPlaneX } 0.06; 2; 0;

Andreu Oliver Gonzlez

// ************************************************************************* //

And the code for myClassExample:

FoamFile { version 2.0; format ascii; class dictionary; object motionProperties; } // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // dynamicFvMeshLibs dynamicFvMesh motionSolverLibs ("libdynamicMyClassFvMesh.so"); dynamicMyClassFvMesh; ("libfvMotionSolvers.so");

dynamicMyClassFvMeshCoeffs { a 0.4; frequency 2; } // ************************************************************************* //

dynamicMyClassFvMesh.C
#include #include #include #include "dynamicMyClassFvMesh.H" "addToRunTimeSelectionTable.H" "volFields.H" "mathematicalConstants.H"

// * * * * * * * * * * * * * * Static Data Members * * * * * * * * * * * * * // namespace Foam { defineTypeNameAndDebug(dynamicMyClassFvMesh, 0); addToRunTimeSelectionTable(dynamicFvMesh, dynamicMyClassFvMesh, IOobject); } // * * * * * * * * * * * * * * * * Constructors * * * * * * * * * * * * * * //

Foam::dynamicMyClassFvMesh::dynamicMyClassFvMesh(const IOobject& io) : dynamicFvMesh(io), dynamicMeshCoeffs_ ( IOdictionary ( IOobject ( "dynamicMeshDict", io.time().constant(), *this,

28

Mesh motions alternatives

IOobject::MUST_READ, IOobject::NO_WRITE

Andreu Oliver Gonzlez

), a_(readScalar(dynamicMeshCoeffs_.lookup("a"))), frequency_(readScalar(dynamicMeshCoeffs_.lookup("frequency"))), stationaryPoints_ ( IOobject ( "points", io.time().constant(), meshSubDir, *this, IOobject::MUST_READ, IOobject::NO_WRITE ) ) Info<< "Performing a dynamic mesh calculation: " << endl << "a: " << a_ << " frequency: " << frequency_ << endl; * * * * * * * * * * * * * * * //

) ).subDict(typeName + "Coeffs")

} // * * * * * * * * * * * * * * * * Destructor

Foam::dynamicMyClassFvMesh::~dynamicMyClassFvMesh() {} // * * * * * * * * * * * * * * * Member Functions * * * * * * * * * * * * * //

bool Foam::dynamicMyClassFvMesh::update() { scalar scalingFunction = (::cos(2*mathematicalConstant::pi*frequency_*time().value()); Info<< "Mesh scaling. Time = " << time().value() << " scaling: " << scalingFunction << endl; pointField newPoints = stationaryPoints_; newPoints.replace ( vector::X, stationaryPoints_.component(vector::X)+ a_*time().values()*(stationaryPoints_.component(vector::Y))* (stationaryPoints_.component(vector::Y))*scalingFunction ); fvMesh::movePoints(newPoints); volVectorField& U = const_cast<volVectorField&>(lookupObject<volVectorField>("U")); U.correctBoundaryConditions(); } return true;

// ************************************************************************* //

29

Mesh motions alternatives

Andreu Oliver Gonzlez

icoDyMFoamMesh.C The code of the solver modified in order to solve only the mesh motion:
#include "fvCFD.H" #include "dynamicFvMesh.H" // * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // int main(int argc, char *argv[]) { # include "setRootCase.H" # # # # # # include include include include include include "createTime.H" "createDynamicFvMesh.H" "readPISOControls.H" "initContinuityErrs.H" "createFields.H" "readTimeControls.H"

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * // Info<< "\nStarting time loop\n" << endl; while (runTime.run()) { include "readControls.H" include "CourantNo.H" include "setDeltaT.H" runTime++; Info<< "Time = " << runTime.timeName() << nl << endl; mesh.update(); runTime.write(); Info<< "ExecutionTime = " << runTime.elapsedCpuTime() << " s" << " ClockTime = " << runTime.elapsedClockTime() << " s" << nl << endl; } Info<< "End\n" << endl; return(0); } // ************************************************************************* //

# # #

30