Mass Matrix

[LJacq]

Hi,

I'm calculating the eigenmodes of a problem and I'd neet to also determine the mass matrix. How can i retrieve it?

Thx in advance

[tzwinger]

Hi,
Do you want to have the system matrix written out in ASCII? If this is your goal, a simple

Code: Select all

 Dump system Matrix = filename

in the solver section should do it. This was explained in the 6.2 Release Notes viewtopic.php?f=11&t=1443&p=4290&hilit= ... dump#p4290

Best Regards,

Thomas

[LJacq]

Thx for your response. I tried the manipulation but get this message:

Model Input: Unlisted keyword: [dump system matrix] in section: [solver 1].

I though downloaded the last version

[tzwinger]

Could you then try to cast it, i.e., write

Code: Select all

Dump Mass Matrix = String "TheFilename.suffix"

instead. Does that help?

Best Regards,

Thomas

[tzwinger]

Correction to previously sent message:

Try

Code: Select all

Dump system Matrix = String "TheFilename.suffix"

instead.

BR,

Thomas

[LJacq]

OK, I now have a file with the response. But in this response, there's a lot of negatives numbers ...
I think the system matrix is not what i really need, it must be a combination of dumped and mass matrix, to create a "linearized" system (Ax=b).
I probably have to extract myself the mass elements. I try to find in
http://elmerfem.svn.sourceforge.net/vie ... k/fem/src/
where is definite this matrix i have a low level of programmation and it's quite hard for me ... I was waiting "simple files" like "linear elasticity solver" but it's a little more harder to find the right way in all the .src ...
I really don't want to take you much time and please scuse me for all my problems but could you just give me a direction, a start point to find where is made this matrix?

[tzwinger]

Hello,
perhaps you can find some answer in chapter 6 of the ElmerSolverManual http://www.nic.funet.fi/pub/sci/physics ... Manual.pdf: Solving eigenvalue problems

Best Regards,

Thomas

[raback]

Hi

You could perhaps use a dummy solver just to get the mass matrix. For example, apply the DivergenceSolver which is one of the simplest solvers and use the same "Dump System Matrix" keyword and you should get the mass matrix (without density accounted for).

-Peter

[stoykov]

Hello,

Including the lines:
Dump System Matrix = String "matrix"
Dump System RHS = String "rhs"
into the .sif file in the solver section, creates a files containing the values of a matrix and a vector, let's say [A]{x} = {b} , I obtain the vector {b} equal to zeros.

But the eigenvalue problems is w^2[M]{u} = [K]{u}

Can someone tell me the relation between both equations? Do we have [A] = -w^2[M] + [K], for a fixed w?

Regards,
Stan

[raback]

Hi

I think that for eigenvalue problems the full matrix that you mark with A is never formed. Arpack needs both M and K. I would think that the system matrix corresponds then to K. For transient problems, for example, the global M is never formed but it's contributions are added to K depending on the time integration scheme.

So for your problem I think the code should be tweaked so that also M is saved, or you should use some other dummy solver where the system matrix is simply the mass matrix.

-Peter