eigenVectors ?
 

Hi Foamers,
what is the exact function in openFoam to get the eigenVectors of a volsymmTensorField A ? If it will return a tensor , then how are the vectors aligned, they will be as rows or as columns in the tensor?

thanks


T.D.

Hello,

The function is really easy:
eigenVectors(A) should work
And it will return in your case a volTensorField T

For each cell i, the first eigen vector is T.internalField()[i].x(), the second T.internalField()[i].y() and the third T.internalField()[i].z()

But to answer directly they are store by line
so the first one is composed of xx, xy and xz.

In the code they use the constructor line 67 there:
http://foam.sourceforge.net/docs/cpp..._8H_source.php


Best regards,

Fred

Hi,
i am trying to get eigenVectors of the strain rate tensor E, so i tried

volTensorField Gr=fvc::grad(U);
volSymmTensorField E = symm(Gr);

volTensorField T=eigenVectors(E);

and it didn't work !!!! i get this error:

i also tried volSymmTensorField T=eigenVectors(E); but it also didn't work !!
Any ideas ????

Hello T.D.,

my fault, eigenVectors is implemented only with volTensorField as input parameter not with volSymmTensorField.

So you will have to convert your E in a volTensorField. Hopefully a trick like tensor(1, 0, 0, 0, 1, 0, 0, 0, 1)*E will do the trick...

Good luck,

Frederic

Hi
it didn't work as you said, i think it seems that eigenVectors() function does not accept a tensor with dimensions, any idea how to take only the values of the dimensioned volSymmTensorField "E" and put them in another dimensionless volTensorField (for example "E2" ) ???

thanks a lot

yours,

T.D.

Hi it should work for a dimensionedTensor, since its implemented in the dimensionedTensor class. Probably it does not work on the boundaries... Try to work only on the internal field.

dimensionedTensorField myTensorFieldOfTheInternal = myVolTensorFieldWhatIHave.internal();
dimensionedTensorField myDesiredEigenvectors = eigenVectors(myTensorFieldOfTheInternal);

Then you give it back as

volTensorField myEigenVectors
(
IOobject
(
"myEigenVectors",
myVolTensorFieldWhatIHave.time().timeName(),
myVolTensorFieldWhatIHave.mesh(),
IOobject::NO_READ
),
myVolTensorFieldWhatIHave.mesh(),
dimensionedTensor::zero
);

myEigenVectors.internal() = myDesiredEigenvectors;


If you need the values on the boundary, maybe you have to loop over those and do the same.
Hope this helps

Kathrin

Hi,
thanks a lot for your anwser, but i tried this inside the time loop:

volTensorField Gr=fvc::grad(U);
volSymmTensorField E = symm(Gr);

Srate.component(0)=E.component(0);
Srate.component(1)=E.component(1);
Srate.component(2)=E.component(2);
Srate.component(3)=E.component(3);
Srate.component(4)=E.component(4);
Srate.component(5)=E.component(5);
Srate.component(6)=E.component(6);
Srate.component(7)=E.component(7);
Srate.component(8)=E.component(8);

volTensorField Ei=eigenVectors(Srate);

and it compiles well. What i need to know now, is it OK like this for filling all Srate components ? however E is symmetric tensor but Srate is not;

RK: ----inside .H file, Srate defined as:
volTensorField Srate
(
IOobject
(
"Srate",
runTime.timeName(),
mesh,
IOobject::NO_READ,
IOobject::AUTO_WRITE
),
mesh
);
---------------------------------------------------------------------


yours,
T.D.

I tried the above recommendations to obtain the eigenVectors but I get a NO_READ error. Also, can anyone on this forum tell me how to obtain the eigenvectors as three different vectors instead of a tensor so that I can easily plot the individual vectors as arrows.

Dear Foamers,

thanks for the helpful discussion. Nevertheless, I have two remaining questions:

1.) How can I calculate the eigenvector for known eigenvalues?
(Reason: My problem has one real and two complex conjugate eigenvalues. I get an error calculating the complex eigenvectors, hence I would like to calculate only the eigenvector for the real eigenvalue.)

Therefor I would like to use the following function (defined in 'tensor.C' at line 207 (for OF 3.0.1).

Is this a reasonable approach? (Does not work for simulation.)

2.) Assigning the return value like this (for a simple test case)

leads to the error:
error: reference to ‘vector’ is ambiguous

Why that?! The return type should be 'vector'?!

I would be glad for help,
Peter

Stop - it works :)

For Foamers with similar problems:
-> 1.)
I should have defined the return value as a volVectorField (and not as a volTensorField like for the function 'eigenVectors').
(Another remaining problem is how to declare a 'volVectorField' like 'EV'. I use a nasty solution here...)

-> 2.)
Still not clear, but it works, so what!

Peter

Hi guys,

About the eigenvectors and eigenvalue, Alberto has implemented a solver to handle this. check it out here: https://bitbucket.org/albertop/eigensolver

Good luck!