eigen value analysis

[lutfulla_aeam]

Dear seismosoft,
I am working on seismostruct v7.0.2, there I noticed that during eigen value analysis the period is decreasing with the increase of subdivision of the same infrmFB element.I have seen in your verification example that for elfrm element the period is decreasing when subdivision is increased. But in case of infrmFB the period should remain unchanged.
Please tell me why the period is changing when we are subdividing the infrmFB element into a number of element.

[huffte]

It appears to me that an eigenvalue analysis, by its very nature, must use elastic stiffness properties and cannot account for inelastic behavior in elements. So, I would think that the element type would have virtually no impact upon eigenvalue analysis results. Obviously, this is not true for other types of analyses.

Best of luck, lutfulla_aeam.

[zenginse]

Greetings everyone.

I have the same problem in SeismoStruct v7.0.2
I need modal mass participation ratios so I've run a eigenvalue analysis but it consistently gives incorrect results. Beside, periods of the structure are not they should be according to the other software.

I'm not sure if i can share my model file here ( because i don't know it is allowed? ) but lutfulla_aeam's problem is a serious one.
When I change my infrmFB frames' subdivision number from 200 to 100, mass participations seems normal. ( It's normal because i am comparing results with another software) But Periods of the structure are not seems normal.

If you need more details or anything else about model i can give you the details.

[seismosoft]

Hi lutfulla_aeam,

By subdividing your member to smaller elements, you introduce more releases to your structure, hence the differences in the eigenvalue results (essentially you allow for a different deformed shape). This is the case in example 2.8 of the Verification Report, and probably in your model too. You will notice that by increasing the number of elements (larger discretisation) your eigensolutions will converge, as in example 2.8
Further, the infrm and elfrm elements are not very different in eigenvalue, the only difference being that the section stiffness is calculated by the integration of the stiffnesses of the monitoring points in the former case, and by the elastic section properties EA, EI, GJ in the latter. The element's formulation is very similar, after that.

Seismosoft Support

[seismosoft]

Hi zenginse,

How can you be so sure that the two models are the same? There is a number of variables that affect the response (lumped vs distributed plasticity, elements' formulation, geometry, material properties etc).

Packages like SeismoStruct have been extensively tested for many years and it is more likely that there are differences in the models, rather than bugs in the programs.

Seismosoft Support

[ruipinho]

zenginse,

I suspect you would benefit from reading carefully the Verification Report of SeismoStruct.

Regards,

Rui

[zenginse]

Hi everyone,

Thank you your answers firstly. seismosoft, I cannot be sure that two models exactly same but i am quite sure there must be some logical approximation. I don't allege that both of the package software use same formulation etc. But both software show me nearly exact mass participation ratio which is %85.

When i change my fibers in columns and beams from 312 and 318 to 328 and 326, my mass participations in Ux and Uy change very dramatically.

I don't know whether it is alright but i am sharing a link that shows number of fibers in elements according to the mass participations.

https://yadi.sk/i/ZBcg-ECxf4ucH

Thank you.

[z.gronti]

Dear Zengin,

I couldn't download your file. We are usually interested in the mass participation in Ux relatively to Uy. Have they changed in the same percentage? You may send the *.spf file to my email if you want me to take a closer look.

[zenginse]

Hi z.gronti,

Sorry for your trouble about download. I have changed the source of the file so that anyone that want to examine can access with ease.

I did not searched a ratio or anything else in Ux relatively to Uy. But they seemed to me very arbitrary values.

I subdivided my columns and beams to 250 fibers 'cos i thought this is a sufficient number to run Time History Analysis and other pushover analysis.

Any help would be appreciated.

Thank you for sparing some time.

[z.gronti]

Dear Zengin,

I have checked the file you sent me for different number of section fibres and i don't get the differences in mass participation percentages that you have mentioned. In fact the results i get are almost the same. Please, make sure that all the project settings are the same and the only feature that you change is the number of section fibres.
Also, i have to mention that, in case of using the Lanczos eigen-solver, it gets a little bit mixed in completely symmetric structures (100% symmetric), as your model. Hence, in such structures (100% symmetric) the best solution is to use the Jacobi with Ritz Transformation eigen-solver.