Vibration Fatigue, Rotordynamics, and FEA Procedure-Beginner

[Nklitzke]

I would like to predict if failure will occur due to fatigue induced from vibrations (I assume rotating unbalance will be the dominant excitation). I am a relatively new engineer, so I apologize for any ignorant questions, but I really appreciate any help and direction provided.

I have attached images illustrating a simplified version of this system. It will generally operate at 1 constant speed, though it can be operated at any speed. It will be supported and hung by brackets that are not shown (I do realize these may be important BCs that will need to be taken into account during the modal study). The spindle utilizes ball bearings.

I believe one of the first steps is to determine what frequency the damaging events may be occurring at (which order) and what the source of the vibration is.
I have a Campbell diagram (acceleration vs RPM) from an accelerometer measurement during a "run-up" and from this I also have:
Acceleration PSD
(integrated to) Velocity PSD
(double integrated to) Displacement PSD
High accelerations were recorded corresponding to the 1st and 3rd order of the rotational speed.
I have attached sketched recreations of the data as I do not have the actual diagrams. I believe they are all reported in RMS.
Can it be determined what orders contribute most to damage solely from the PSD and Campbell diagram? Which PSD is "best" to look at?
Since I am interested in fatigue, I am hypothesizing that the displacement PSD may be the best indicator of which order is the most damaging.

Possible sources of vibration that I have identified:
1st order: Unbalance
2nd order: Aero effects (2 blades)
3rd order: Unknown (but high accelerations were recorded)(maybe bearings?) Is there a recommended way to determine the source of vibration?

A FE model was created and correlated with empirical results from an impulse test (hammer excitation; multi-input single-output). The natural frequencies and mode shapes have strong correlation until the 3rd order range. I am concerned if the 3rd order is a major contributor to the total accumulated damage that I may not be able to accurately simulate this.
My current plan is to run a linear dynamic vibration (harmonic) analysis to determine the stress and ultimately, determine the fatigue life. Damping was an output from the impulse test, so I plan to use those values. I do not know what my loads are and what orders of excitation are significant. I can certainly calculate the maximum unbalance load, but it acts in the X-Z plane and the response is in the Y (vertical) plane. Is it the bending that would cause the vertical excitation and, if so, how would I model this? How could I estimate force values for the higher orders?
Again, any help on how to proceed is extremely appreciated.

 

[Nklitzke]

I already need to correct myself. I do not have the PSDs, I have the responses in the frequency domain.

 

[GregLocock]

I'm sorry, answering that lot would take hours

"Can it be determined what orders contribute most to damage solely from the PSD and Campbell diagram?" well obviulsy i your case first and third order would be most likely, but no, in general you need to do a proper fatigue analysis to be sure.

"Is there a recommended way to determine the source of vibration?" There is not one way, there are many ways, some of which might work in a particular case

" How could I estimate force values for the higher orders?" You don't even know what is causing 3rd order, how can you estimate the amplitude of something you know even less about?



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Nklitzke]

Thank you. I appreciate your time and fully understand that I asked too many questions to realistically expect any detailed response.

In regards to the higher order amplitudes, I believe your question is rhetorical, but I was considering applying unit loads in a dynamic simulation using my correlated FE model (with virtual sensors in the same positions as the accelerometers). I would then scale my unit loads so that the responses(vs frequency) match the measured responses, thus "backing out" the force magnitude by utilizing the reciprocity of the (largely) linear system. Is that poor logic?

 

[GregLocock]

Ah, I didn't realise you had vibration spectra for the higher orders. Yes, that approach might work, but deriving excitation forces from accelerations is cutting edge stuff in my field (whole vehicle loads), and if it were routinely possible would replace millions of dollars of equipment.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?