Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Per node excitation of random vibration
Posted 10.05.2020, 12:58 GMT-4 Acoustics & Vibrations Version 5.5 4 Replies
Please login with a confirmed email address before reporting spam
Dear Comsol community,
It is of interest to excite a plate at each node with an imported matrix that is interpolated. Using the bracket random vibration model as reference -ref, for the procedure I define the interpolation as input in the RO model. The model uses shell elements and by so I apply a face load to the surface. As described in ref I run a rvibr study before defining the PSD characteristics. Though the calculation runs without a fault there is no rom defined in global parameters ROM.rvib (image) so as to define random vibration characteristics. The interpolation - a force over space type, is working by the way. Any thoughts?
Thank you for your time.
Attachments:
Please login with a confirmed email address before reporting spam
Hi Theodore,
It is not clear to me wheter the pressure field is distributed, but is fully correlated, or whether there is also a non-trivial cross-correlation with a spatial diestribution.
-------------------Henrik Sönnerlind
COMSOL
Please login with a confirmed email address before reporting spam
Hej Henrik,
The purpose is to apply a delta correlated excitation over the boundary, i.e. white noise, which is incoherent and by such has no cross correlation. The model is solved fast and the process is as explained inconsistent to the psd part. My conception was to apply the interpolation, as force per area related to nodes, as input for the RO model and use the RO input for the boundary load.
Tack, Theodore
Please login with a confirmed email address before reporting spam
Hi Theodore,
This is non-trivial, for two reasons.
- For each independent (uncorrelated) input, you need one ROM input.
- Since any boundary condition in COMSOL Multiphysics is applied on a geometric entity, you cannot directly apply loads on nodes.
What can you do then?
As there will have to be a lot of ROM inputs, I do not think this is feasible without automation using a model method or similar.
You can then add a large number of points, each of them equipped with a point load, represented by its own ROM input.
Alternatively, you can subdivide the boundary into a large number of smaller boundaries, each of them with its own boundary load.
In the Random Vibration node, you will have to add as many PSD entries as there are ROM inputs.
How feasible this in terms of computational efficiency, I cannot really tell. That, of course, depend on the number of geometrical objects with corresponding ROM inputs that you use.
-------------------Henrik Sönnerlind
COMSOL
Please login with a confirmed email address before reporting spam
Hej Henrik,
I appreciate the input and yes, more or less I have these processes in mind from Comsol articles and the forum. What you are describing with the multiple ROMs is what I expected from the PSD part as a subroutine. I understand it needs build up from my part.
The process I follow is one described by Mace and Shorter under the overall scope of a mid frequency approach . The main method is characterised of low processing demands. Considering your comment at the end, a question is raised about the process created, since the nodal points are above 1500. It looks computationally demanding.
Thank you, Theodore