In any running machine, there are a large number of exciting forces . Frequency of some of these forcesmay be close enough to the natural frequency of the machine, thereby causing respnance. My question.
1. Is it possible to catch natural frequency through simplke spectrum analysis without bump test?
2.I have heard, logarithmic scale is helpful in finding such cases. What ratio between its peak amplitude and its broad base frequency amplitude around that peak can tell , yes it is resonance. Also, how many averages should be taken to reduce the effect of random m noise and the window to be used.

Depending on your instrumentation, you may be able to do a coastdown, and look for the frequencies that do not decrease in frequency as the machine coasts down. Especially look for those that might tend to get larger or smaller, but not decrease in frequency.

http://vibrotek.com/notes/bumptest/index.htm

If the issue is identification of natural frequencies without shutting down the motor, you can achieve this by doing a Bump Test while the machine is running & using negative averaging.

The data collector will cancel out the common components & you effectively get the same spectra as a Bump Test at standstill. CSI & Pruftechnik offer this feature, Commtest does not. Not sure about SKF.

Regards,

Aditya

I find negative averaging for bump test difficult to analyse. Since the running machine may already excites the natural frequency the energy introduce in the system with an impact hammer is often negligible. This makes it hard to analyse the result. Cost down and bump test are the perfect tools. I believe that the only other mean of verifying a resonance problem on a running machine is to install a dynamic absorber. You have to know the natural frequency of the system. Then you design a dynamic absorber, I like to tune it on a heavy milling stand prior installation, if the vibration goes down at the suspected frequency then its resonance if not, well wrong frequency or vibration from an other source.
Best regard, Marcel

quote:

Originally posted by Aditya:
Aditya

Aditya, I actually mean different from what is being interpreted. I know of linear minus averaging with respect to bump test for running machines. I mean to say that for any running machine, there are a large no. of exciting forces with broadband frequencies. Any of these forces must be exciting resonance in the running machine. Can we diagnose this resonance and thus natural frequency only by analysis of its spectrun without bumping it and capturing its spectrum during normal run
any comments.
regards.
Irshad.

It seems like Duncan provided a good paper to illustrate how to inpsect the noise floor to get a good idea where the resonances lie. It works best with high resolution and log scale as was stated.

I'm sure it's not a perfect method, but I think it's a great start and relatively easy - sometimes retrievable from data that you have already collected.