Random Response Analysis on Large Models

Steve Bell_21840

Hi,

 

We are developing some large FE models, in excess of 4 million elements. Which must

be subjected to a series of random vibration analyses. Have there been any developments

or tips for keeping the huge temporary files created by Optistruct to a minimum. It appears

we are restricted to only creating results for sets, which means re-running the analyses

several times on the same model.

Note:

1. We do ensure no results are created for the frequency response sub-step.
2. We have to use a 1Hz increment for the random vibration analysis.
3. The frequency range is also large typically 5 to 1000Hz.
4. We are exceeding 4tb of discuss space.

 

Kind regards,

 

Steve

Adriano_Koga

it seems indeed a large problem to solve.

 

Is it possible to break into smaller frequency intervals? instead of running it from 5 to 1000Hz directly.

You can use EIGSAVE and EIGRETRIEVE to avoid running the normal modes everytime.

 

Also, instead of using FREQ1 with 1Hz interval, what about reducing FREQ1 interval, but using FREQ4 to get a better frequency interval close to your ressonant modes? (if not so many)