it's not a limitation but a limit of validity of linear Dynamics.
Linear dinamics implies that the analysis should stay linear. That's why the code can use linear superimposition method.
Contact is non linear, that's why you can't use it in random vibration (which is one of the possible linear dynamic analyses.
You can overcome that limitation by doing non linear dynamic analysis I think.
I only fear that you can perform only a time history analysis there and not a random vibration analysis where you input PSD as A function of frequency.
I would need to check this before telling you if it can be done or not.
Hi David: Alessandro is correct. A changing boundary condition makes it a nonlinear problem, because the software has to stop what it's doing and reformulate the stiffness matrix when any boundary conditions change (such as parts separating or coming back into contact). Hence it's not a limitation of the software since it does not fall under the perview of Linear Dynamics (in any FEA software, i.e., ANSYS won't do that either). He is also correct that you can run a nonlinear dynamic but it's only available in the time domain. If you do have a PSD curve and have the overall GRMS specification, you can synthesize a time-history input that meets the specification, so you can read that into the Nonlinear dynamic. I know that Tom Irvine writes a code to do that. You would need to contact Tom at http://www.vibrationdata.com/. He is really good at that stuff. Hope that helps. - Tony B.
Thanks both of you for your help - I think I understand the limitation correctly now about how a dynamic analysis needs linearity for superpostioning to be valid. And we do know the GRMS level so I will definitely look into what Tom Irvine has to say and offer on this issue. I've been meaning to sign up for his online course material for some time and now I have a great reason to do so.