defining contact in a PSD analysis

[alexboldea]

i've been doing a few PSD analyses using Abaqus and i was interested in modelling contacts between some parts. so far i've been using tie constraints which means i have no relative displacement between the node set pairs in that particular interaction. however, this causes a lot of stiffness.
i've talked to the guys from Abaqus support and i've also done some research and apparently the reason why i can't use contact in a PSD simulation is because that would change the stiffness matrix during the simulation and make the problem nonlinear. by "contact" i mean to allow two bodies to move away from each other but not intersect if they're moving towards one another, but rather get some tensions (like it would happen in reality).

i am interested if there is any workaround to this matter, i saw someone on the solid works forum saying something about creating a time-dependent analysis but it was unclear to me as i've never used solid works (isn't it a designer software?)
i'm not set on using only Abaqus, we have access to quite a bit of software here but the problem is - is there any software that could do this PSD simulation keeping in mind the contact issue? or is there a special workaround, as the issue seems to be related rather to the mathematics of it all then to a software limitation.

any thoughts on this would be greatly appreciated, thank you!

alex

 

[GregLocock]

I agree, using a time domain approach is the most likely to succeed.

So roughly speaking the procedure would be:

create excitation time vs force signal that meets your input psd (see other threads on this).

Use this to excite your model.

This will give you plots of stress and strain and accelerations vs time

then analyse those back into the frequency domain using pwelch or whatever.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[spongebob007]

If you want to include contact in a dynamic analysis you must perform a transient analysis.

 

[caemagic]

Hey alex

I am not familiar with abaqus , hence I am not sure what sort of features it has , anyways if I were you I would have two approachs to this problem

1- synthesis a time history for your given PSD and make a transient analysis , if abaqus cannot do this conversion , then download the MATLAB code from the following website:

http://vibrationdata.wordpress.com/...-history-to-satisfy-a-power-spectral-density/

if you do not have MATLAB , then try to download the opensource software scilab , it's almost identical to matlab and you be able to the use the code to be downloaded from the previous link
2- the second approach is to try to model linear contact , I am not sure whether this would applicable to your case or not. for more info abt linear contact , do refer to the attachement

cheers
cae magic

 

[alexboldea]

thanks everyone for the answers! just because i haven't said anything doesn't mean i've been ignoring the issue. the conclusion is that, indeed, i would need to perform a transient analysis and this is simply not an option at the moment due to the complexity of the assembly / limited hardware.

since other projects have come i'm putting this personal issue to the side for now until i'll get some time to try a transient on a simple assembly.