Modal analysis -- specify frequency range?

dculp

For modal analysis, can the frequency range be specified? For example, only extract frequencies between 18 kHz and 22 kHz? I want to do this for two reasons --

1. To reduce analysis times for large models by limiting the number of extracted modes.
2. Currently I must specify N modes. If this is the first time that I am analyzing the model then I may not have a good feel for N. If the initial N is too small then I must increase N and rerun the analysis. If the initial N is too large then extra analysis time is wasted extracting unwanted higher-order modes.

If this can't be done within the UI, can the liml file be edited to do this? If so, exactly how?

Thanks,
Don C.

Victor

I don't think this is really possible with the eigenvalue solver that Mecway and CCX use. Most of the solving time is spent finding the frequencies, and all at once, so by the time it can choose them, it's too late. CCX has options to specify a frequency range but it finds only as many as you specify then discards those that are outside the range so you wouldn't know if it missed some modes at the top.

dculp

Victor --

I always know the approximate frequency of the mode that I'm interested in because I'm designing to operate at a particular resonant frequency. For example, the nominal operating frequency of my design might be 20 kHz. Then I would only be interested in modes between 18 kHz and 22 kHz. For certain models there may be hundreds of unwanted modes below 18 kHz.

I understand what you're saying about runtimes. However, there are other advantages to limiting the frequency range. For example, the CCXtoParaview utility outputs a separate Paraview vtu or vtk file for every mode that it finds in the CCX frd file. If there are a lot of unwanted modes in the frd file then this leads to a lot of disk clutter. Also, I have written a utility to extract all mode-frequencies from either the liml or frd files. If the number of modes in these files could be reduced then the search time would also be reduced.

Can you explain how to insert the frequency-limiting info into the CCX file?

(I hope that you may sometime add a liml to Paraview converter.)

Victor

Here's an example. The changes are to replace

*FREQUENCY
8

with

*FREQUENCY
8, 1e8, 3e8

to limit it to modes between 1e8 and 3e8 Hz

dculp

Victor --

I ran a Calculix model (77806 nodes) with *FREQUENCY set to 50. The analysis included stress calculations.

When a frequency range wasn't specified, the total analysis time was 14:49 minutes of which about 7 minutes were for stress calculations (one for every extracted mode). The resulting Calculix frd file size was 641 Mb.

When the frequency range was set so that only the last 11 modes were extracted, the total analysis time was 9:31 minutes (36% reduction) of which only about 2 minutes were for stress calculations (limited to the 11 extracted modes). The resulting Calculix frd file size was 150 Mb.

Thus, for moderate sized models where stresses are calculated, limiting the frequency range can substantially reduce the analysis time and greatly reduce the Calculix frd file size. I haven't investigated for analyses where stresses aren't calculated but I assume that limiting the frequency range wouldn't have a substantial advantage for the analysis time or frd file size.

Victor

Thanks for testing that @dculp. I hadn't considered the cost of the stress. Looks like it may be more useful.