Preload in Modal Analysis - Restart 2

[adampar]

Hi everyone,
I use Femap v.11.1.2 with NX Nastran 9.0 and i face a rather difficult problem.
First of all i have a structure that i want to use in a frequency response analysis, but before that i run a modal analysis so that i can use these results for restart in frequency response.
Now, my structure is preloaded and i need this preload in order to have a first natural frequency of 50 Hz.I used two ways of extracting the preloaded modes:
I run a NL analysis with a temperature load to induce stress in certain preloading members and ask to calculate the natural frequencies using the stiffness matrix at the last step of the NL Analysis
I run a modal analysis having defined a static subcase with a *DEFORM card for the preloading members
The problem is that the first solution gives me first natural frequency of 52 Hz, whereas the second one gives 27 Hz!! What is the deal here?
Also if the results from the first analysis are correct, can i use them as restart files for frequency response analysis?
I wish i get an answer,
Thank you

 

[jbrackin]

I have a few questions as well as some suggestions. Why nonlinear for the preload run?
If you are modeling bolt preload, then thermal loading is not the only way, you might want to use the bolt preload feature to set the preload more easily.

You can calculate stiffened modes from any static loading using the STATSUB case control. This can be set up in Femap. If you save this solution, then you should be able to restart from it for your modal frequency response.

if you have not performed this analysis before, strongly sugggest a small pilot model to get comfortable with the setup before submitting a large run.

 

[adampar]

Hi and thanks for the reply,
I have a truss structure that i want to press down (a heavy weight applied on it) and i want to calculate these preloaded natural frequencies. Now, i have already run preloaded modal analysis using STATSUB with both a temperature load and DEFORM card on beams that i intentionally connected to my structure to take this preload, and in both cases the modes were the same as in the structure without preload... I have never used bolt preload, and i don't see how it could apply to my case.

I ran NL analysis because i know it's another way to preload the structure and calculate modes. In fact using this analysis output higher frequencies just as i wanted! But now i can't use them for subsequent frequency response...​

 

[jbrackin]

Why are you using temperature or deform loading? Those are methods are inducing a preload(like a bolted joint) that cannot be applied by normal mechanical loads. If you expect that your truss changes frequencies due to this heavy weight you refer to, then just apply the heavy weight in the linear case used by STATSUB. There is no need for any other loading to be applied.
Using linear static analysis and STATSUB only adds the differential stiffness from the loading to the stiffness used by modes. If you see no change in the modes, then this means that particular effect does not change the modes significantly.
If you run a nonlinear solution and calculate modes from the converged nonlinear solution, then you have now included differential stiffness as well large displacement effects. If you are seeing a change in the frequencies with nonlinear but not with linear statsub, then this would suggest to me, that large displacement is the most important contribution and would then need to be included.

It is possible to restart and use the converged nonlinear stiffness matrix. I would consider that a pretty advanced method and only for someone very experienced with Nastran or under the guidance of an experienced user. If you believe that is required, the place to start is a pilot model that is quick and easy to run and understand, then test all run setups and procedures before tackling the real model.

 

[adampar]

Ok thank you jbrackin, i see. Could you also give out some hints, how this procedure of restarting in modal analysis from the converged nonlinear stiffness matrix could be implemented? I 'll try first your suggestions, but in case i need it...