Function of subcases in random respons analysis

[ProductivCADUser]

Hello togheter

What is the purpose of subcases in random response analysis?
Do the results add up? Or does that make no sense?

In my case, I have three PSD graph, one for each axis.
I have created one subcase per direction; I defined the load direction using the direction of gravity-load.

I obtain individual results by leaving only one subcase active. When I calculate subcases in combination, I get different results. I haven’t found any information on this in the manual.

[John_Holtz · Accepted solution]

Hi @ProductivCADUser 

In general, a random response analysis does a Frequency Response analysis and outputs those results as "STEP n, FREQ=f". The STEP results are scaled based on the entered PSD and outputs the result as "PSD n, FREQ=f". The PSD results are integrated to give the RMS results.

When there are multiple subcases, the solver calculates the Frequency Response for each subcase and outputs separate "STEP n, FREQ=f". All subcases results are scaled by the PSD and combined, so there is only one "PSD n, FREQ=f". The PSD results are integrated to give the RMS results.

Here's the problem and why you should not use multiple subcases in Random Response analysis. (I used version 2026 for this test.)

1. Inventor may show that there are multiple "STEP n, FREQ=f" results, but it only shows the results from the last subcase. Thus, you cannot check the Frequency Response calculation for other subcases.
2. The solver should be calculating the "PSD n, FREQ=f" results for each subcase and combining them for one "PSD n, FREQ=f" result, but the solver may not use the correct PSD for each subcase. (This item has been reported to the developers, but I do not have any feedback about what circumstances cause this problem. Some models work properly; some do not.)

John

[ProductivCADUser]

Thanks John.
In this case, we ensure that only one subcase is active at one time for the calculation.

This seems to me to be a serious error in the programme. At the very least, NASTRAN should warn the user if this hasn’t been programmed correctly.

I only realised this when I compared the results.

[Dr_Nastran]

The Implicit Nastran User's Manual found in the Editor Help has an example on page 216 for when 2 subcases would be required.  This is for cross-spectral density.  Otherwise use a single subcase for Random unless you need cross-spectral density.