In normal modes analysis, different extraction methods produce very different resuts.

[fabian4GB8W]

As suggested by the title, I'm doing a normal modes analysis. The problem is that different extraction methods, i.e., Lanczos and subspace iteration, yield very different values for the frequency of the lowest mode.

For example, for exactly the same mechanical system, Lanczos teration yields 2.56 Hz, whereas for subspace iteration it gives 0.748 Hz. In both cases, for the second lowest mode, both methods yield roughly the same value of 71.35 Hz.

It's also strange that in the case of Lanczos iteration, when asking to report up to ten normal modes (rather than 3), the lowest one is claimed to have a frequency of 0.051 Hz instead of 2.56 Hz. There is no mode with 71.35 Hz either. Something similar happens when choosing subspace iteration.

The most important parameter I need to calculate is the frequency of the lowest resonant mode. Is there a way to determine which methods yields the more accurate result?

Any help will be appreciated. Thank you very much in advance.

Regards,

Fabián

[John_Holtz · Accepted solution]

Hi Fabian,

I think both methods (subspace and Lanczos) use an iterative method to "find the modes" by computing more modes than necessary and then removing the other modes. Therefore, specifying more modes (10 vs 3) will calculate even more modes (like 30 vs 9) which is how it finds a different set of results.

You need to look at the mode shape as well as the frequency to see if "mode 1" corresponds to "mode 1". It is possible that one solution version another misses some. 0.051 Hz sounds very low and I wonder if that is some rigid body mode. 

That's what I know at the moment.

John

[naibaf_omsare]

Hello John,

Thank you very much for your reply. Your answer is indeed insightful.

I came to realize the problem was that I needed to consider a prestress condition. Once the prestress state is calculated, the very low resonant frequency I mentioned (0.051 Hz) is gone from the set of solutions. Additionally, the normal modes calculated coincide with independent calculations.

I'm confused as when a prestress analysis is requires. However, I think it's better to ask in another thread.

Regards,

Fabián