Cancel
Showing results for
Did you mean:

# condense stiffness matrix

Experimenter

Hello People,

I'm doing a CMS for a large model. I'm processing my results in Matlab for a target mode selection. My problem is, to solve the generalized eigenvalue problem in matlab my mass and stiffness matrices from NX should be the same size, and they are not.

In my SOL 103 real eigenvalues solution i obtain in a .pch file my original stiffness and mass matrices:

in the stiffness matrix all the nodes have 6 DOF

in the mass matrix all the nodes except those from RBE elements have 3 DOF, the former ones 6 DOF, so no rotational DOF for normal nodes.

as i understand i would require to make a static condensation of the matrix K, so Kred = Ktt - Ktr*(Krr)^-1*Krt. Where subscript t are translational DOF and r are rotational DOF.

As i do this in matlab and compare with the CMS (Craig-Bampton) from NX i obtain different results, specifically for the fixed interface modes. I'd like to know if there is a parameter or a easy way to make this static reduction. I have tried making a DOFset restraining the additional DOF of my stiffness matrix but this i would need to do it for all the nodes and it's no practical.

I would really much appreciate help,

best regards,

Jaime.

2 REPLIES

# Re: condense stiffness matrix

Experimenter
Hi Jaime,

I am currently facing the same problem. My model is a lumped mass-spring system. Performed CMS (Fixed-Boundary) reduction and my mass and stiffness matrices(from .pch file) are not of the same size.

Did you find any solution? If so, any help on this matter would be much appreciated.

Thanks,
Shilpa

# Re: condense stiffness matrix

Experimenter
hello Shilpa,

I still can't find a solution. The only thing I know is that the type of
elements you use define how many degrees of freedom you'll have, so you
won't face this problem if you use only CHEXA elements for instance. But if
you use shell elements they'll have 6 DOF but only 3 DOF with mass. I
haven't tried to disable the DOF, maybe that could work, but i'm also not
sure how to do it. Please tell me if you come with a solution

--

Jaime Andrés Bretón