Random Vibration Analysis results

[bnengineer]

We are performing a Random Vibe FEA on a spacecraft component. My question is once we have stress values from the analysis (either 1 sigma or 3 sigma), what then do we do with these? Compare to yield? Superimpose with pressure and thermal stresses and then compare to yield? Are they suitable only for fatigue calculations, and again in combination with operating stresses from a separate analysis? A brief web search describes most folks in industry estimating a test series to apply to their component and does not describe the analytical avenues that can be explored.

 

[GBor]

Random Vibe on a "spacecraft component"? I have to believe that you would want an expert in the field to answer this one. Doesn't sound like something to ask the general populace no matter how knowledgable we may be in FEA. I think you should find someone a little closer to the situation, but if that isn't possible, here are a few questions:

What is the expectation of this component? How is it used? How critical is it that it not fail? Is this random vibration event something that happens once as it travels through the atmosphere at a certain speed? Or does it happen through the entire flight through the atmosphere and back? Are the other environmental factors that you mention occuring simulaneously with the vibration event (if so, I definitely think you need to consider them in the final post-processing)?

I'm sure an aerospace engineer would understand your situation, but for those of us with a vibrations background, could you provide a little more information?

Thanks!

Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.
Lower Alabama SolidWorks Users Group
Magnitude The Finite Element Analysis Magazine for the Engineering Community

 

[bnengineer]

Your suggestion to find an expert is appropriate and we are taking steps to do so, unfortunately our customer is not the proper source for said expert, making it a bit more difficult.

The random vibe PSD curve provided pertains to launch loads the component would experience, and for this particular component the operating environment would be happening at the same time (making it very critical to the application, hint, hint). So again, our plan is to combine the effect, but we were searching for any indication of what other folks do with random vibe results. We do have a fatigue life requirement as well as simple yield requirements. The random vibe stresses seem like an HCF-type phenomenon, but clearly if the component is reuseable and will experience launch after launch, that is more LCF.

Thanks for your input.

 

[GBor]

Sounds interesting. I definitely think you have to consider fatigue. I have no feel for how many cycles your particular part may experience during launch, so I can only suggest that you compare your analysis results to the appropriate Goodman curve to see how many cycles it would take to fail at the stress levels that you are seeing assuming a fully alternating stress.

I do think it is important for you to apply the pressure and thermal stresses. What software are you using for the analysis? Most of the better-supported FEA packages allow for taking the results from one analysis as the inputs to another.

If results from a random vibe analysis indicate excessive stress compared to yield, you have some significant design issues to overcome (fatigue limit is generally 60% of yield for metallics, 55% in some cases).

For the comparison to yield, I would suspect that you have some "worst case" loading. Either a peak g-loading during the acceleration to break through the atmosphere, or perhaps some impact load through the sound barriers...I have no idea what you experience in a given launch like this, but there should be something other than a vibrational input. I'm just thinking...waiting for someone with some experience in this area to "chime in".

 

[rb1957]

i reckon before the rubber hits the road (or in your case, before the spark hits the rocket fuel) that there will be a vibration test of the compeonent.

from the analysis point of view, i'd take the 3sigma vibration stresses and superimpose them with the othere relevant loads/stresses (thermal, pressure, inertia (maneourve?)) ... this'll give you the maximum stress experienced by the component. include the ultimate factors and complete the static stress analysis.

also fatigue is very relevant ... i think it's HCF (with a big GAG cycle) ... you'll apply the same stress combination as above (it's not only vibration, but vibration with the other applied stresses) and at the end of each flight the stress comes down to zero.

I'd think the agency that your customer is working for should have this defined ... maybe your customer has lots more information than they're letting on ... this should be covered by a product spec. from whoever owns the spacecraft, including requirements for test.

 

[btrueblood]

"Random Vibe FEA on a spacecraft component."

Probably 90% of a rocket is not operational during launch, so you are in a very small segment of a small segment of the aero community.

" My question is once we have stress values from the analysis (either 1 sigma or 3 sigma), what then do we do with these?"

Compare them to failure criteria

" Compare to yield?" yes, that is one criteria.

" Superimpose with pressure and thermal stresses and then compare to yield?"

Yes, if the combined stress is a realistic operating condition. HOWEVER, remember that a pressure load on a thin shell fuel tank, for instance, provides a great deal of stiffening to that shell structure. You may want to simulate the combined loads (launch vib + operational) to make sure you don't end up overpredicting (or underpredicting) the true combined stresses.

" Are they suitable only for fatigue calculations, and again in combination with operating stresses from a separate analysis?"

Realistically, you would want to create some combination of stresses from 1 to 3 sigma as a "normal" launch environment and examine the fatigue margins as a result of that. I.e. it is not very probable that the part will see a continuous series of 3-sigma events, what it will actually see is a time distribution of events, the probability for each level of loading occuring during any given amount of time is equal to the area under the bell curve...etc. etc. With luck, you can reduce the number of cases you must analyze by using judgement, and reduce the analysis to a subset of "worst case scenarios".