Different behaviour according to discretization

[Danae]

I run an eigenvalue analysis to my model with columns consisting of 1 element. The results of the 3 first periods are: output1 T=0.954, output 2 T=0.945, output 3 T=0.858. When I run absolutely the same model but with 2,4 or more elements at the columns, the results are the upcoming: output 1 T=0.858, output 2 T=0.954, output 3 T = 0.945. Is there something wrong? Does the output 1 means that its period is the first one, or I choose as the first period, myself (and first modes), the maximum of the periods of all the outputs, consequently, the period of output 2? As well as this, the results in the pushover analysis are different according to the discretization of the columns. How can I explain this? I hope I don't write in a very complicated way...
Thank you very much!

[Danae]

I have another similar problem. I run an eigenvalue analysis and I get the results with a certain range of the periods, output1 T=0.954, output 2 T=0.945, output 3 T=0.858 (as I mentioned before). Then I run an adaptive pushover analysis and I get in the first output with Load 0, a different range of periods: output 1 T=0.858, output 2 T=0.954, output 3 T = 0.945 . I have the impression that the two results should be the same, but they are not. Is there something wrong again?

[seismosoft]

The results are the same, simply the order in which the different modes are output has changed. Check modal masses (or modal participation factors) and you see what we mean.

Seismosoft Support

[Danae]

Yes, I know that the results are the same but in a different order and that was my problem. I was afraid that there was something wrong with my model. So, you mean that it is not necessary to take the first in order mode as the fundamental, I can choose myself the maximum of all the outputs and set it as the fundamental one? In that case the outputs are only in the order that the results are calculated and not in the right final one. Please confirm to me that I have correctly understood.
And can you please explain to me the different results of pushover analysis depending on the discretisation of the columns?

[seismosoft]

Yes, according to the theory of eigenvalue analysis, the fundamental period of vibration is that featuring the highest modal participation mass. Please do refer to any of the Structural Dynamics books quoted in the Help System for further details on these topics.
As for what concerns issues of mesh discretization, we advise you to read both the relevant sections of the Help System as well as the technical literature on the matter, including that discussing localizations issues.

Seismosoft Support

[Danae]

Thank you very much for the immediate help! I am aware of the theory of eigenvalue analysis and that is why I was confused, having the impression that the program should automatically show first the fundamental mode. As far as the discretization issues are concerned, it is true that I don't know a lot of things, I can have a look in the literature mentioned. My main problem was if that difference in the behaviour was logical and now you confirmed it. Thank you!

[Danae]

Hello again!
I have an unexpected problem which has to do again with the discretization. I run static pushover analyses for absolutely the same building but with different number of elements at the columns, and I get the forthcoming results: For columns with 1 element, the displacement collapse occurs at 0.078m, for columns with 2 at 0.189m, for colums with 3 at 0.252m, for colums with 4 at 0.238m and for colums with 6 at 0.308m and takes very long time to finish. According to the finite elements theory, the greatest number of elements, the closest to reality it is, is that right? But in my case, which number of elements is better and more realistic to use for the rest of the analyses?
Thank you very much! I would be very grateful for your help, as I'm really confused with it.

[seismosoft]

Hi,
Only in elastic analysis does the principle you quote (the more refined the mesh is, the more accurate the results are) hold. In inelastic analysis, localization phenomena prevent such rule from being fully accurate.
This topic (localization) is extremely complex, and we can only suggest you to search the literature for further advice.
In any case, we can certainly state that the results obtained with 1 or 2 (displacement-based) elements are not accurate enough, as explicitly indicated in the Help System, hence they are not even worth considering.
Seismosoft Support

[Danae]

Thank you so much for the help, it is really usefull for my dissertation!