View Issue Details

IDProjectCategoryView StatusLast Update
0002108OpenFOAM[All Projects] Bugpublic2016-06-02 14:08
Reporterali.shAssigned Tohenry 
PrioritynormalSeveritymajorReproducibilityalways
Status resolvedResolutionfixed 
PlatformGNU/LinuxOSUbuntuOS Version14.04
Product Version 
Fixed in Version 
Summary0002108: error in SGS epsilon in dynamic K equation model Parallel simulation
DescriptionWhen running in parallel, with dynamic 1 equation eddy model, the epsilon value at the cells that are boundaries of processors is wrong.
This issue does not exist in OFv3.0+
Steps To ReproduceAttached you can find a simple modified solver to write epsilonSGS in two ways. using dynamically evaluated Ce and constant Ce=1.048
Interesting is that when looking at the epsilonSGS which is evaluated by the constant Ce=1.048, there is no problem.
Also it is not the matter of defining a volScalarField to write down epsilon since also using this code at the source level in dynamicKEq.C, correct function you can reproduce the issue:
if(this->runTime_.outputTime()) { volScalarField epsilon("epsilon", this->epsilon()); epsilon.write(); }
Additional InformationI could not attach a test case, you can simply use pitzDaily test case and run e.g. with 8 processors using dynamickEqn model.
Set the max range of epsilonSGS to sth low to see the processor boundaries.

in my pisoFoam I added two volScalarFields in createField.H and assigned value to them in main function.
TagsNo tags attached.

Activities

ali.sh

2016-06-02 08:54

reporter  

mypisoFoam.tar.gz (174,754 bytes)

henry

2016-06-02 12:20

manager   ~0006360

I have resolved the issue in OpenFOAM-dev:
    commit 2f1bc368e5ac532bce133dc112e2a153f7de0391

The problem relates to the inappropriate use of the boundary values of delta which have no meaning and I have temporarily worked-around the problem by correcting the BCs of the field before filtering.

My longe-term plan is to rewrite the sources of the turbulence models to use GeometricField::InternalFields and also make delta a volScalarField::InternalField which will avoid further problems such as this one.

If the dynamic LES models now work as expected in OpenFOAM-dev I can make the same change to OpenFOAM-3.0.x if required.

henry

2016-06-02 14:08

manager   ~0006361

Resolved in OpenFOAM-3.0.x by commit bd1209ae61b8857e3c8083734210dc8bbcc1c215
Resolved in OpenFOAM-dev by commit 2f1bc368e5ac532bce133dc112e2a153f7de0391

Issue History

Date Modified Username Field Change
2016-06-02 08:54 ali.sh New Issue
2016-06-02 08:54 ali.sh File Added: mypisoFoam.tar.gz
2016-06-02 12:20 henry Note Added: 0006360
2016-06-02 14:08 henry Note Added: 0006361
2016-06-02 14:08 henry Status new => resolved
2016-06-02 14:08 henry Fixed in Version => 3.0.x
2016-06-02 14:08 henry Resolution open => fixed
2016-06-02 14:08 henry Assigned To => henry