Post Processing and Advanced Running Options

POLITECNICO DI MILANO CHALMERS

OpenFOAM postprocessing and advancedrunning options

Gianluca Montenegro

Department of EnergyPolitecnico di Milano

Gianluca Montenegro/ OpenFOAM postprocessing and advanced running options

POLITECNICO DI MILANO CHALMERSThe post processing tool: paraFoam

The main post-processing tool provided with OpenFOAM is the reader module to runwith ParaView, an open-source, visualization application

The module is compiled into 2 libraries, PVFoamReader and vtkFoam, using version2.4.4 of ParaView supplied with the OpenFOAM release

ParaView uses the Visualization Toolkit (VTK) as its data processing and renderingengine and can therefore read any data in VTK format

OpenFOAM includes the foamToVTK utility to convert data from its native format toVTK format, which means that any VTK-based graphics tools can be used topost-process OpenFOAM cases. This provides an alternative means for usingParaView with OpenFOAM

paraFoam is strictly a script that launches ParaView using the reader module suppliedwith OpenFOAM

It is executed like any of the OpenFOAM utilities with the root directory path and thecase directory name as arguments:

paraFoam <root> <case>


POLITECNICO DI MILANO CHALMERSBefore starting

Before starting with the user should copy two tutorials in his own run folder:

cp -r $FOAM_TUTORIALS/simpleFoam/pitzDaily/ $FOAM_RUN/cp -r $FOAM_TUTORIALS/dieselFoam/aachenBomb $FOAM_RUN/

SimpleFoam is an incompressible steady-state flow solver, while dieselFoam isdedicated to combustion process with spray injection

The two cases should also be executed

runblockMesh . pitzDailysimpleFoam . pitzDailyblockMesh . aachenBombdieselFoam . aachenBomb

The user should care about reducing the endTime and the writeInterval to haveimmediately some calculations to post process for the pitzDaily case

For the aachenBomb case the user should switch off the chemistry to reduce thecomputational time


POLITECNICO DI MILANO CHALMERSViewing the mesh

The mesh can be viewed only in paraFoam since there is no pre-processing tool toview the mesh

We shall start with the pitzDaily case of the simpleFoam tutorial:

paraFoam $FOAM RUN/ pitzDaily

Click the Accept button which will bring up an image of the case geometry in the imagedisplay window.

In the parameters panel it is possible to choose what you can visualize:

the time step

the region

the vol field

the point field if exists

To visualize the mesh make sure that all the regions are toggled on


POLITECNICO DI MILANO CHALMERSThe post processing tool: paraFoam

When paraView is launched the case is controlled from the left panel, which containsthe following:

Selection Window lists the modules opened in ParaView, where the selectedmodules are highlighted in yellow and the graphics for the given module can beenabled/disabled by clicking the eye button alongside

Parameters panel contains the input selections for the case, such as times,regions and fields

Display panel controls the visual representation of the selected module, e.g.colors

Information panel gives case statistics such as mesh geometry and size

ParaView operates a tree-based structure in which data can be filtered from the top-level case module to create sets of sub-modules



Select property and wireframe of surface to visualize the mesh. Play wit theactor color menu to select the color you prefer



To improve the quality of the mesh Visualization it is possible to overlap the surface ofyour mesh

Select the extract parts from the filter menu and then select surfacevisualization of the mesh. Play with the color option to create the desired contrast


POLITECNICO DI MILANO CHALMERSViewing fields: display panel

The Display panel contains the settings for visualizing the data/fields for a given casemodule. The following points are particularly important:

the data range is not automatically updated to the max/min limits of a field, so theuser should take care to select Edit Color Map and Reset range at appropriateintervals, in particular after loading the initial case module

the style of legend, e.g. font, for the color bar is controlled by in the Scalar Barpanel of the Edit Color Map window

the underlying mesh can be represented by selecting Wireframe in theRepresentation menu

the geometry, e.g. a mesh (if Wireframe is selected), can be visualized as a singlecolor by selecting Property from the Color by menu and specifying the Actor color

the image can be made translucent by editing the value in Opacity (1 = solid, 0 =invisible)

the activated field can be translated, scaled, rotated with respect the other ones

In the color windows it is possible to select the field that the user wants to plot


POLITECNICO DI MILANO CHALMERSViewing fields: vectorFields (U)

Selecting the velocity field from the color window in the display panel, two options areavailable:

point volPointInterpolate , the velocity field is defined on the grid vertex andinterpolated from the cell centers

cell U the velocity field is displayed in the cell center

It is possible to visualize only the magnitude of velocity, without any information aboutthe direction of the velocity vector:

magU . pitzDaily

Time = 50Reading UCalculating magU

mag(U): max: 10.619 min: 0.12244

Then select the magU filed in the color window

If you run the magU utility after having launched paraFoam you will need to change thetime step or to update the GUI to visualize the new field

Alternatively the user may use the calculator filter. Keep in mind that it works only forthe selected time step



To visualize the velocity filed as a field of vectors you must act on filters once again:

The user must first create a filter to select the quantities in the cell centers



Starting from the cell center filtered field, the user must select the Glyphs filter andthen accept

The vector field will be displayed in the window on the right


POLITECNICO DI MILANO CHALMERSViewing fields: glyphs (U)

The filter reads an Input and offers a range of Glyphs for which the Arrow0 provides aclear vector plot images

In the Orient/Scale window, the most common options for Scale Mode are: VectorMagnitude, where the glyph length is proportional to the vector magnitude, whereasselecting Data Scaling Off each glyph is the same length

An additional Scale Factor parameter controls the base length of the glyphs

It is possible to select the maximum number of Glyphs to be displayed. Putting anumber lower than the cell number can speed up the visualization

Other option rather than arrows are available: cones, cubes, lines, spheres andarrow2d

It is possible to select the glyphs filter directly from the ParaView toolbar:

Glyphs button


POLITECNICO DI MILANO CHALMERSViewing fields: contour plots

A contour plot is created by selecting Contour from the Filter menu at the top menu bar

Iso−surface button

The filter acts on a given module so that, if the module is the 3D case module itself, thecontours will be a set of 2D surfaces that represent a constant value, i.e. isosurfaces

The Parameters panel for contours contains the list of constant values, that the usercan edit, most conveniently by the Generate range of values window. The Input andScalars menus present the module and fields, respectively, that are read by the filter


POLITECNICO DI MILANO CHALMERSViewing fields: streamlines

Streamlines are created by first creating an append attributes filter

Streamlines are created selecting the tracer lines using the Stream Tracer filter

The tracer Seed window specifies a distribution of tracer points over a Line or PointCloud

The distance the tracer travels and the length of steps the tracer takes are specified inthe text boxes below


POLITECNICO DI MILANO CHALMERSViewing fields: probes

To plot a graph in ParaView, the users must first extract the internal mesh using theExtract Parts filter

The user can then plot a graph by selecting Probe from the Filter menu

In the Probe Object window, the user should select Line and position the line verticallyup the center of the domain with a resolution of 50

The fields that are plotted are selected in the Point scalars window of the Probe panel



The user can select the probe filter directly by clicking onto the probes button on thetoolbar

Sample button

It is possible to write the selected sampled fields onto a csv file. The output will be acomma separated file as follows:

volPointInterpolate(p),0.200097,5.12835,8.24178,10.3533,...volPointInterpolate(magU),8.63369,8.13691,7.70859,7.37819,....volPointInterpolate(epsilon),48.1914,52.6348,59.0361,67.6491,7,...volPointInterpolate(k),1.08187,1.13229,1.19894,1.28333,1.3918,...


POLITECNICO DI MILANO CHALMERSViewing fields: Data analysis

To plot field values in a certain point (cell, vertex, line) as a function of time (plottedtime steps) the user can select the filter Data analysis form the filter menu or byclicking onto the Data Analysis button on the toolbar:

Data analysis



Pick the cell you want to select by positioning the pointer onto it and pressing the p key(pick)

The user can pick the cell by means of different query methods: cell, point(interpolated), line, cell ID, point ID

Plotted fields can be activated or deactivated

The time period can be selected by acting on the slide bar in thetemporal parameters window

Once again the plotted fields can be written onto a csv file by clicking thesave as CSV button


POLITECNICO DI MILANO CHALMERSCut planes

To create a contour plot across a plane rather than producing isosurfaces, the usermust first use the Cut filter to create the cutting plane, on which the contours can beplotted

The Cut filter allows the user to specify acutting Plane or Sphere in the Cut Func-tion menu by a center and normal/radiusrespectively

The user can then run the Contour fil-ter on the cut plane to generate contourlines

Multiple cut planes can be generatedusing the Add value or the Generaterange of values sub panel


POLITECNICO DI MILANO CHALMERSClip filter

The Clip filter works similarly to the Cut one, but keeps the mesh and fieldinformation on one side of the cutting plane

The Clip filter allows the user to specifya cutting Plane, Sphere, Box or on thebasis of a value for a scalar field

The clip selection can be inverted byactivating the Inside out option

The visibility option allows the user toactivate or deactivate the visibility of thecutting plane, sphere or box


POLITECNICO DI MILANO CHALMERSThreshold

The Threshold filter allows to visualize only the cells having the values of theselected field within a specified range

The range can be specified by means of moving the Upper threshold and lowerthreshold bars


POLITECNICO DI MILANO CHALMERSVisualizing Lagrangian

To visualize the Lagrangian particle tracking the case must be converted into paraviewformat

The user should run the application foamToVTK on the case containing the Lagrangianparticle tracking (in our case the aachenBomb tutorial for the dieselFoam application)

foamToVTK . aachenBomb

This command will generate an additional folder named VTK inside the case directory

The user must start paraview typing the following command:

paraview

The case can be opened clicking on the file menu, open, and selecting the VTKdirectory. By clicking on the *.vtk file the case will be opened

The user may load all the time steps by clicking on the timesteps button and selectingthe Add all choice on the new window

The use must then open the files contained in the lagrangian directory inside the VTKfolder


POLITECNICO DI MILANO CHALMERSVisualizing Lagrangian

To display the particles the user must filter the lagrangian field with the Glyphs filter

Then the scalar option must be selected in the scale mode window along with thespheres as glyphs


POLITECNICO DI MILANO CHALMERSThe 3D view button

There are settings controlling the 3D view Properties, selected from the View menu, oralternatively by clicking the small 3D View toggle button at the top left of the imagedisplay window

The General panel includes the following items which are often worth setting at startup

the background color, where white is often a preferred choice for printed material

Use parallel projection which is the usual choice for CFD, especially for 2D cases

the level of detail (LOD) which controls the rendering of the image while it is beingmanipulated, e.g. translated, resized, rotated; lowering the levels set by thesliders, allows cases with large numbers of cells to be re-rendered quickly duringmanipulation

The Annotate panel includes options for including annotations in the image

The Display Orientation Axes feature is particularly useful.

The Camera panel includes buttons of Standard Views, settings for Camera Controlsand the options for detailed Camera Orientation

The user can store up to 6 camera positions by clicking on a given button using theright mouse button, for subsequent retrieval using the left mouse button


POLITECNICO DI MILANO CHALMERSViewing sets with paraview

The user may create sets (pintSet, faceSet and cellSet) to perform certain additionaloperation over the fields, i.e. moving certain points or applying a source term only in acertain area

A set, i.e. a cellSet, can be created in OpenFOAM using the command cellSet

$FOAM_APP/utilities/mesh/manipulation/cellSet

This application creates a list of cells on the base of data read from the cellSetDictfile

The user shall try to create a new cellSet in the pitzDaily tutorial case

runcp -r $FOAM_TUTORIAL/simpleFoam/pitzDailycd pitzDaily/systemcp $FOAM_UTILITIES/mesh/manipulation/cellSet/cellSetDict .



The user must modify the cellSetDict file in the following way:

// Name of set to operate onname myCellSet;

// One of clear/new/invert/add/delete|subset/listaction new;

topoSetSources(

// Cells with cell centre within boxboxToCell{

box (-0.05 -0.25 -0.001) (0.1 0.25 0.001);}

);

To create the cell set the user must run the cellSet application

cellSet . pitzDaily



Once the cellSet has been created run the foamToVTK command on the pitzDailycase using the cellSet option:

foamToVTK . pitzDaily -cellSet myCellSet

To know all the available option of the foamToVTK application type foamToVTK andpress enter. It will be displayed the usage of that application

Usage: foamToVTK <root> <case> [-surfaceFields] [-ascii][-faceSet faceSet name] [-mesh mesh name] [-nearCellValue][-pointSet pointSet name] [-excludePatches patches to exclude][-allPatches] [-cellSet cellSet name] [-parallel] [-fields fields][-constant] [-noPointValues] [-latestTime] [-noInternal] [-time time]

To visualize the cellSet the user must open the VTK folder (it works also form a alreadyopened paraFoam windows) in the pitzDaily directory and select the file with the nameof the cellSet: myCellSet_*.*


POLITECNICO DI MILANO CHALMERSData manipulation

It possible to rely on data manipulation utilities which OpanFOAM provides inside itsutilities folder

OpenFOAM utilities can also be copied and customized to create user defined utilities

An example could be the ptot utilities which calculates the total pressure starting fromthe velocity and pressure field

$FOAM_UTILITIES/postProcessing/miscellaneous/ptot

It is also possible to export data to be visualized with other post-processing tools

>cd $FOAM_UTILITIES/postProcessing/dataConversion>lsfoamDataToFluent foamToFieldview9 foamToVTKfoamToEnsight foamToGMV smapToFoam


POLITECNICO DI MILANO CHALMERSRunning in parallel

The method of parallel computing used by OpenFOAM is known as domaindecomposition, in which the geometry and associated fields are broken into pieces andallocated to separate processors for solution

The first step is to decompose the domain using the decomposePar utility

The user must edit the dictionary associated with decomposePar nameddecomposeParDict which is located in the system directory of the case

The first entry is numberOfSubdomains which specifies the number of sub-domainsinto which the case will be decomposed, usually corresponding to the number ofprocessors available for the case

The parallel running uses the public domain openMPI implementation of the standardmessage passing interface (MPI)

The process of parallel computation involves: decomposition of mesh and fields;running the application in parallel; and, post-processing the decomposed case


POLITECNICO DI MILANO CHALMERSRunning in parallel

We shall use the pitzDaily case for the simpleFoam application tutorial

tutcd simpleFoamcd pitzDaily

We can copy the pitzDaily case to the pitzDailyParallel

cp -r pitzDaily pitzDailyParallel

And copy into the system directory the directory the decomposeParDict file available inthe pitzdailtExptInlet case

cp pitzDailyExptInlet/system/decomposeParDict pitzDailyParallel/system/


POLITECNICO DI MILANO CHALMERSRunning in parallel: domain decomposition

The mesh and fields are decomposed using the decomposePar utility

The underlying aim is to break up the domain with minimal effort but in such a way toguarantee a fairly economic solution

The geometry and fields are broken up according to a set of parameters specified in adictionary named decomposeParDict that must be located in the system directoryof the case of interest

In the decomposeParDict file the user must set the number of domains which thecase should be decomposed into: usually it corresponds to the number of coresavailable for the calculation

numberOfSubdomains 2;



The user has a choice of four methods of decomposition, specified by the methodkeyword

For each method there are a set of coefficients specified in a sub-dictionary ofdecompositionDict, named <method>Coeffs, used to instruct the decompositionprocess

method simple/hierarchical/metis/manual;

simple : simple geometric decomposition in which the domain is split into piecesby direction, e.g. 2 pieces in the x direction, 1 in y etc.

hierarchical : Hierarchical geometric decomposition which is the same as simpleexcept the user specifies the order in which the directional split is done, e.g. first inthe y-direction, then the x-direction etc

metis : METIS decomposition which requires no geometric input from the user andattempts to minimize the number of processor boundaries. The user can specify aweighting for the decomposition between processors which can be useful onmachines with differing performance between processors

manual : Manual decomposition, where the user directly specifies the allocation ofeach cell to a particular processor



The coefficients sub-dictionary for the simple method are set as follows:

simpleCoeffs{

n (2 1 1);delta 0.001;

}

Where n is the number of sub-domains in the x, y, z directions (nx ny nz) and delta isthe cell skew factor, typically 10−3

The coefficients sub-dictionary for the hierarchical method are set as follows:

hierarchicalCoeffs{

n (2 2 1);delta 0.001;order xyz;

}

Where n is the number of sub-domains in the x, y, z directions, delta is the cell skewfactor and order is the order of decomposition xyz/xzy/yxz...



The coefficients sub-dictionary for the metis method are set as follows:

metisCoeffs{

processorWeights(

1...1

);}

It is the list of weighting factors for allocation of cells to processors: <wt1> is theweighting factor for processor 1, etc.

The coefficients sub-dictionary for the manual method contains the reference to a file

manualCoeffs{

dataFile "<fileName>";}

It is the name of file containing data of allocation of cells to processors



Data files may need to be distributed if only local disks are used in order to improveperformance

The root path to the case directory may differ between machines. The paths must thenbe specified in the decomposeParDict dictionary using distributed and rootskeywords

The distributed entry should read

distributed yes;

and the roots entry is a list of root paths, <root0>, <root1>, . . . , for each node

roots<nRoots>("<root0>""<root1>"...);

where <nRoots> is the number of roots

Each of the processorN directories should be placed in the case directory at each ofthe root paths specified in the decomposeParDict dictionary



The decomposePar utility is executed in the normal manner by typing

decomposePar <root> <case>

The output will look like:

Calculating distribution of cellsSelecting decompositionMethod metis...Processor 0

Number of cells = 6199Number of faces shared with processor 1 = 71Number of processor patches = 1Number of processor faces = 71Number of boundary faces = 12667

Processor 1Number of cells = 6026Number of faces shared with processor 0 = 71Number of processor patches = 1Number of processor faces = 71Number of boundary faces = 12343



On completion, a set of subdirectories will have been created, one for each processor,in the case directory

The directories are named processorN where N = 0, 1, ... represents a processornumber and contains a time directory, containing the decomposed field descriptions,and a constant/polyMesh directory containing the decomposed mesh description.

pitzDailyParallel|------> 0|------> constant|------> processor0|------> processor1|------> system

openMPI can be run on a local multiprocessor machine very simply but when runningon machines across a network, a file must be created that contains the host names ofthe machines. The file can be given any name and located at any path


POLITECNICO DI MILANO CHALMERSRunning in parallel: execution

An application is run in parallel using the mpirun command:

mpirun --hostfile <machines> -np <nProcs> <foamExec><root> <case> <otherArgs> -parallel > log

where: <nProcs> is the number of processors; <foamExec> is the executable, e.g.simpleFoam; and, the output is redirected to a file named log.

The <machines> file contains the names of the machines listed, one machine per line

The names must correspond to a fully resolved hostname in the /etc/hosts file of themachine on which the openMPI is run. The <machines> file would contain:

hostname1hostname2 cpu=2hostname3

In our case it becomes:

mpirun -np 2 simpleFoam . pitzDailyParallel -parallel > log &


POLITECNICO DI MILANO CHALMERSRunning in parallel: post-processing

When post-processing cases that have been run in parallel the user has two options:

reconstruction of the mesh and field data to recreate the complete domain andfields, which can be post-processed as normal;

post-processing each segment of decomposed domain individually


POLITECNICO DI MILANO CHALMERSRunning in parallel: post-processing

After a case has been run in parallel, it can be reconstructed for post-processing bymerging the sets of time directories from each processorN directory into a single set oftime directories

The reconstructPar utility performs such a reconstruction by executing the command:

reconstructPar <root> <case>

In our case it will become:

reconstructPar . pitzDailyParallelTime = 0Reconstructing FV fields

Reconstructing volScalarFieldsp

.....Reconstructing volVectorFields

UReconstructing volTensorFields

RNo point fieldsNo lagrangian fields


POLITECNICO DI MILANO CHALMERSThe sample utility

OpenFOAM provides the sample utility to sample field data for plotting on graphs

The sampling locations are specified for a case through a sampleDict dictionarylocated in the case system directory

The data can be written in a range of formats including well-known graphing packagessuch as Grace/xmgr, gnuplot and jPlot

The sampling can be executed by running the utility application sample according tothe application syntax

sample . pitzDaily

The sampleDict dictionary can be generated with the standard set of keyword entriesusing FoamX, or by copying a detailed sampleDict from a the sample utility folder:

$FOAM_UTILITIES/postProcessing/miscellaneous/sample


POLITECNICO DI MILANO CHALMERSThe sampleDict file

The sampleDict contains entries to be set according to the user needs

interpolationScheme cellPoint;

The choice for the interpolationScheme can be one of the following:

cell : use cell-centre value only, constant over cells, no interpolation at all

cellPoint : use cell-centre and vertex values

cellPointFace : use cell-centre, vertex and face values

Vertex values are determined from neighboring cell-centre values whereas face valuesdetermined using the current face interpolation scheme for the field (linear, gamma,etc.)

writeFormat raw;

The choice for the write format depends on the application used to plot the series. Theuser can choose one of the following:

xmgr

jplot

gnuplot

raw



The sampleDict file contains a sampleSet subdictionary which defines where thefields should be sampled

sampleSets(

uniform{

name data;axis x;start (-0.0206 0 0);end (0.29 0 0);nPoints 1000;

})

In this subdictionary the user can specify the type of sampling method, the name ofsamples and how to write point coordinate plus other parameters depending on themethod chosen



The user may choose one of the following sampling methods:uniform : evenly distributed points on lineface : one point per face intersectionmidPoint : one point per cell, in between two face intersectionsmidPointAndFace : combination of face and midPointcurve : specified points, not necessary on line, uses trackingcloud : specified points, uses findCell

name lineX1;

The name entry is typically a filename. After running the sample application the userwill find a new folder in the case directory named sample

Each data file is given a name containing the field name, the sample set name, and anextension relating to the output format, including .xy for raw data, .agr for Grace/xmgrand .dat for jPlot

axis distance;

Specifies how to write point coordinates, options are:x/y/z: x/y/z coordinate onlyxyz: three columnsdistance: distance from start of sampling line (if uses line) or distance from firstspecified sampling point



There are also some type specific options:

start and end coordinate for uniform, face, midPoint and midPointAndFace

nPoints number of sampling points for uniform

list of coordinates for curve and cloud

The fields list contains the fields that the user wishes to sample

fields(

pmag(U)U.component(1)R.component(0)

);

The sample utility can parse the following restricted set of functions to enable the userto manipulate vector and tensor fields

U.component(n) writes the nth component of the vector/tensor

mag(U) writes the magnitude of the vector/tensor


POLITECNICO DI MILANO CHALMERSProbes

To plot values of fields as functions of the time the user can rely on the newfunctionObject class

This is a modular plugin that evaluates at every time step the specified field at thespecified location

Function objects are created adding them to a function subdict in the controlDict file

The user may find an example of how to modify the controlDict file in$FOAM_TUTORIAL/oodles/pitzDaily/ case

This function cannot parse the set of functions as the sample utility does, however inthe case of the Ux component extraction the user can easily do that for the U probedfield


POLITECNICO DI MILANO CHALMERSProbes

functions(

probes1{

type probes; // Type of functionObject// Where to load it from (if not already in solver)functionObjectLibs ("libsampling.so");probeLocations // Locations to be probed. runTime modifiable!(

(0.1778 0.0253 0.0));// Fields to be probed. runTime modifiable!fields(

p);

});

During the job run it will be created, inside the case directory, a folder named probescontaining the probed values


POLITECNICO DI MILANO CHALMERSfoamLog

There are limitations to monitoring a job by reading the log file, in particular it is difficultto extract trends over a long period of time

The foamLog script is therefore provided to extract data of residuals, iterations, Courantnumber etc. from a log file and present it in a set of files that can be plotted graphically

The user must run the application in the following manner:

simpleFoam . pitzDaily > log

Where the log file can be named arbitrarilyThe script is executed by

foamLog <root> <case> <logFile>

The files are stored in a subdirectory of the case directory named logs


POLITECNICO DI MILANO CHALMERSfoamLog

Once the log file has been created the foamLog must be executed to generate therequired statistics

foamLog <root> <case> <logFile>

The files are stored in a subdirectory of the case directory named logs

Each file has the name <var> <subIter> where <var> is the name of the variablespecified in the log file and <subIter> is the iteration number within the time step

Those variables that are solved for, the initial residual takes the variable name <var>and final residual takes ¡var¿FinalRes. By default, the files are presented intwo-column format of time and the extracted values

>cd pitzDaily>ls

contCumulative_0 epsilonIters_0 kIters_0 Time_0 UyFinalRes_0contGlobal_0 executionTime_0 p_0 Ux_0 UyIters_0contLocal_0 foamLog.awk pFinalRes_0 UxFinalRes_0epsilon_0 k_0 pIters_0 UxIters_0epsilonFinalRes_0 kFinalRes_0 Separator_0 Uy_0


Post Processing and Advanced Running Options

Documents