Nastran random response analysis on a free-free structure

[stefanix1972]

Dear all,
I'm investigating the response to a disturb of a spacecraft in flight.
The disturb can be thought as a random acceleration input applied to a node (as a PSD, for example).
My approach would be to use SOL111, DLOAD and RANDPS, but the point is that structure is a free-free one.

Normally, if I applied the load by an SPCD to a node, I would need to constraint the structure to that node at least in one DOF.
Then the node DOF would be "unlocked" by SPCD application in that specific DOF.

But in this way, the modal base would be impacted by a constraint that does not exist, whereas the structure is a free-free one.
Is there any other approach I could apply?
It seems to me that SPCD is no compatible with a free-free structure behavior.
Currently, the spacecraft is constrained in the CoG by a weak CBUSH (K~0.001), connected in one end to an SPC and in the other end to an RBE3 that is linked to all other model nodes.

Regards,
Stefanix1972.

 

[dmapguru]

A non-disjoint structure free to move in the 3 translational and 3 rotational directions of the Cartesian system defining those translations and rotations will have the same dynamic behaviour irrespective of where in space it is located. Moving a structure to a different location or orienting to some other angle (or both of course) implies a rigid body transformation from one location/orientation to another, but this relocation does not change the properties of the structure, so any disturbance applied to the structure oriented in the same relative direction to that structure will solicit the same response. This means we can separate the total dynamic response of a structure into a rigid body part and a part that is relative to some location, usually on the structure; this is called a particular solution, but it requires the definition of “relative to where?”.

When we define an SPCD in a dynamic response analysis, we are defining a “where”. The SPCD definition in dynamics gets the name enforced motion because the motion could be enforced displacement, velocity or acceleration, not just displacements like in statics. By definition, a PSD in random analysis is an acceleration quantity (the P in PSD means Power, and it comes from the testing world where power is a measure of acceleration). So the enforced motion is going to be enforced acceleration. The key word for the current discussion here is “enforced” – with an SPCD entry, you are saying “I already know the displacement/velocity/acceleration of the points on SPCD entries, so these points will have prescribed motion at every frequency”. At the enforced motion degrees of freedom, the structure will be obliged to follow your prescribed motion. Depending on which degrees of freedom have enforced motion defined, one or more of the rigid body movements will be removed from the problem, yielding a particular solution relative to the enforced motion degrees of freedom.

So when you say a constraint does not exist for the modal basis, oh yes it does. If the structure will have prescribed motion at the SPCD degrees of freedom, these must be restrained during the computation of the modal basis to obtain the correct modal behaviour relative to these DOF. With enforced motion, you are simply rendering the solution a particular one instead of one in an inertial frame of reference. You are defining another frame of reference when you use the RBE3 to connect all nodes in the model to a grounded spring located at the CG of the structure.

If you want to satisfy yourself that the modal basis computed when your enforced motion points are fixed by SPC/SPCD entries, create a small structure that will solve quickly, and define your enforced motion problem with SPCD entries. Run this problem is SOL 111 which will compute the modes you request then solve the frequency response problem on the modal degrees of freedom. Now re-run this problem in SOL 108 (direct frequency response). Just change the 111 to 108 and go. SOL 108 does not compute a modal basis, it solves the problem on the physical variables (which is why I suggested a small problem to try it out). If you requested enough modes in the SOL 111 run, you should get the same answers to a few decimal places in both SOL 111 and SOL 108. Tip: use PARAM,G damping in both the SOL 111 and 108 runs – modal damping is not available in SOL 108 because there are no modes.

DG

 

[stefanix1972]

Dear DG,
thanks for the reply, maybe my question was not well posed and I'm not sure to fully understand the answer.
The disturb in my structure is in a whatsoever node and can be of different nature. It can be for example a PSD of acceleration g²/Hz or a of a force (N²/Hz). Or it can be in the frequency domain or can be a transient force in time domain.
If it is of random nature, the approach is to calculate frequency response by a modal base and then use FRF to calculate random PSD acceleration of another point that is sensitive to acceleration.
My point of confusion is the modal base: if I use SPCD to prescribe acceleration PSD in a node, this must be restrained during the computation of the modal basis to obtain the correct modal behaviour relative to this DOF (I quoted your answer as I absolutely agree with you).
This means that the modal base is calculated with the model not free, therefore modal combination will be not the one of a free model.
Hence, my doubt arises and that's the reason of the workaround of a model which is constrained in the CoG by weak springs: it is not "numerically" free but in practice it is. The constraint in the CoG is just a dummy one to try to run the analysis by ACCE option in RLOAD1, it has nothing to do with the SPC related to ACCE application point.
To summarize my tests:
if I use ACCE option in RLOAD1, it doesn't work both for a fully free-free structure and a weakly free-free structure.
If I use LOAD option in RLOAD1, it works both for a fully free-free structure and a weakly free-free structure.

Regards,
Stefanix1972