Substructuring vs. CMS 1

[meshparts]

Hello,

I have done some comparison simulations using the Substructure and CMS (Component Mode Synthesis) techniques in ANSYS.

For statical analysis both methods are very accurate, compared with the results of the full model.

For modal analysis I'm getting totally wrong results when using the substructure technique. On the other hand CMS works just fine.

So the question is: Why dose the substructure technique works just for static simulations? Am I doing something wrong? I think so, but where? Both stiffness and mass matrices are written so why dose the statical analysis work but the modal not??

Thanks in advance!

Regards,
Alex

 

[cbrn]

Hi,
it happens because CMS includes normal modes' generalized coordinates, while the "classical" substructuring does not. This allows to correctly reconstruct the dynamic behaviour of the complete model, as if the complete dynamic matrix was present.

Regards

 

[meshparts]

Hi cbrn,

thank you for your answer!

So one can say, never use the standard substructure technique, when computing eigenmodes? Why still having the option of exporting both stiffness and mass matrix (seopt,,2)?

What kind of method is Ansys using for standard substructure: the Guyan reduction?

Regards
Alex

 

[cbrn]

Hi,
yes, the "standard" method should be a Guyan reduction or something very similar.
I suppose the option is not forced to CMS in any case because there may be some figures where the standard substructuring gives acceptable results with some savings in computational resources. I seem to remember that it's when only the low-freq eigenmodes are relevant. For example, if you look for the response to a 10-10000 Hz sine sweep and the only relevant eigenmodes are, say, in 5-20 Hz range. In this case, the error (which should increase with the eigenmode order) in the medium-high eigenfrequencies calculation would become irrelevant. But, for the exact meaning of "low" frequency, you'd better rely on Ansys Theory Manual rather than on me ! ;-)

Regards

 

[meshparts]

Thanks cbrn!

It is now clear. I also found the following in the Ansys Help:

3.14.1. Theoretical Basis of Matrix Reduction

The ANSYS program uses the Guyan Reduction procedure to calculate the reduced matrices. The key assumption in this procedure is that for the lower frequencies, inertia forces on the slave DOF (those DOF being reduced out) are negligible compared to elastic forces transmitted by the master DOF. [...] The net result is that the reduced stiffness matrix is exact, whereas the reduced mass and damping matrices are approximate.

Regards,
Alex