Transient structural dynamic calculations with ANSYS

[ARA1985]

Hi everybody,

I am launching time-domain simulations (ANSYS CLASSIC V12.1) on a mechanism by applying displacement time histories at its extremities(basin tests). I use a submodelling approach (use of a beam model, reapplying the loads (force/moment Wind + accélérations) issued from the beam model on submodels Meshed with shells). The results seems globally good (in accordance with the expectation) except at one node where the accélérations are very high. What it is strange is that at this node, the displacements are ok. When a calculate the accélérations by derivating twice the displacement, the results are ok. So how can you explain the fact that the acceleration are not well calculated?

It is probably due to the Newmark scheme but I am not sure (The scheme used is an unconditionnally stable scheme)?

ps: When I work in imposed accelerations, the results are perfect everywhere (in both displacement and acceleration). Working with acceleration is not convenient for me because of pre-processing, drift issues...

Thanks in advance for yout help.

 

[Drej]

Be careful with any derived data in a transient analysis, as these can contain numerical noise. You can ask for internally-calculated accelerations in the latest version of ansys but you'll need to first check that in v12 you can do this (I don't have access to the help files at the moment). If you can, then you need to ask for these before you solve (see the OUTRES command). If not then you may need to apply an appropriate filter to the accelerations by some third party software like Matlab or something, or if you're feeling really bold write your own filter code. Again, be careful about using filters as these can remove 'real' data and not just noise generated by the solver.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[ARA1985]

Thank you Drej for your answer.

I agree the fact that filtering the high frequency content can be dangerous (the response could be non linear).
From my point of view, the internal accelertions are calculated using Newmark (or HHT) method (numerical integration).

I can make the calculation using the latest version of ANSYS. I will try next week.

Regards