measuring simple shaft vibration at a bearing

January 29, 2013 4:26 PM

i am trying to measure the vibration of a bearing, which is at one end of a simple fan shaft. the fan is large but rotates at reasonably slow speed. fan diameter approx 3m, shaft speed approx 145 rpm, 2 dodge s2000 bearings. i am getting FFT data but when i look in the low frequency range the data is not meaningful. i am using sample rate 4096 Hz, block time 0.125 sec. 20 sec sample time and i think the acceleromter is set up correctly. where might i be falling over?
posted by edtut to Science & Nature (11 answers total)

I think you need to clarify some of your terms:

* What do you mean by "not meaningful"?
* What do you mean by "vibration of a bearing"? Are you thinking of internal vibration of its elements, vibration induced on it by the action of the fan, vibration from an irregular shaft, something else?
* How big an acceleration do you expect? What part of the bearing do you have instrumented?

My first thought is that even at a shaft speed of 145rpm (or about 2.25Hz) I wouldn't necessarily expect to see vibration of the bearing in the same frequency domain -- there's a decent chance that you'd get many disturbances in one shaft revolution, from either irregularities in the shaft or bearing elements, and there's no reason to think they'd be regularly spaced or (simply) periodic, so any vibration that would really be "seen" by Fourier is going to be composed of natural frequencies of the steel elements themselves, which is most likely way above anything Mr. Nyquist would predict you see at your current sampling rate. Without a huge amount of data I would guess that any low-frequency components are going to look a lot like noise, since it could easily be that (for example) you've got one bearing element knocking at 3Hz, another at 3.2Hz, another at 2.9Hz, another at 2.95Hz, etc, etc. all the way across your low-frequency band.
posted by range at 5:02 PM on January 29, 2013

First, I would verify the setup. Take off the accelerometer and use a sine wave from a signal source. see if you see the peaks where you expect. Use an attentuator if you need to to drop the signal level down to a comparable level.

I presume you are using a data acquisition system that has some analog outputs? If not, you can always use a soundcard and an open source spectrum program. (I'll recommend one if you don't have one.)

Second, once you trust your FFT settings, get some time domain data from the accelerometer. You should see something on a scope.

There are only a few possibilities, and those are that your FFT settings are wrong, your time domain data isn't what you think it is, or it is missing. Smack the bearing with a little hammer and see if some data shows up in the lower FFT bins. Should be an impulse spectrum.

Your 20 seconds worth of time data seems extreme. You need some real time displays, and they would be shorter block intervals... Why such a long FFT? Are you zero padding or what?

Verify setup first, then post more info/results and/or memail.
posted by FauxScot at 5:12 PM on January 29, 2013

Response by poster: good answers. "range" your on the right track, for the 2.25 Hz rotations we are seeing peaks at 2.8 Hz, 5.6, 8.4 etc. doesnt seem to fit. i think i am seeing harmonic peaks, but they dont quite match the shaft "speed".
scot i will try and hammer test, that should give me a clue. and i will cut down the 20sec, maybe the block size is hurting my definition? i think i can work out a sine wave generator, i am using an LMS system and should be able to work out something.
thanks all for the advice
posted by edtut at 6:20 PM on January 29, 2013

If you want to verify your setup you can put your accelerometer on a speaker cone and feed that with a LF sine wave. Sweep the sine wave through the range you are curious about.

You'd expect the blade pass rate to be reflected in your data. I bet you have a five-bladed fan. The fundamental is 14 Hz. 14/5 = 2.8.
posted by jet_silver at 8:13 PM on January 29, 2013

Response by poster: the impellor is in 2 halves, like 2 cones stuck together, and has 9 blades on each cone.
so i guess at 145 rpm the blade passing frequency should be around 20.25 Hz
posted by edtut at 8:54 PM on January 29, 2013

Response by poster: am i just being silly or is there a quick way to drop a screen shot in here?
posted by edtut at 9:19 PM on January 29, 2013

Another thing to look at is the mounting of the accelerometer and its sensitivity axis/axes vs. the direction of the vibration you're trying to look at. I would normally try to measure stuff like this with a set of strain gages, not an accelerometer, partly because of those issues.

Finally, and this may just be an artifact of me trying to do a free body diagram in my head at 1am, doesn't most of the vibration from the blades turn into a longitudinal wave running down the axis of the drive shaft, and therefore not really couple well into the bearing? If you were getting big torques on the shaft or other forces perpendicular to the axis I could imaging transmitting to the bearing that way, but axial forces should barely register in the bearing.
posted by range at 10:13 PM on January 29, 2013

Response by poster: i agree with your comments on longitudinal load relative to blade passing freq, but believe you should still see a response consistent with the rotating speed.
posted by edtut at 10:26 PM on January 29, 2013

am i just being silly or is there a quick way to drop a screen shot in here?

nope. stick it on dropbox or flickr or imgur and send a link.

good suggestion on the speaker. however, a function generator has the advantages of waveform shape and implied spectral components, amplitude and freq adjustments. validate your FFT settings and verify your software is set up.

then, speaker test to verify the accelerometer. you need to wiggle it in its sensitive axis and SEE some output.

at that point, you can test and assess your target with confidence.
posted by FauxScot at 10:45 PM on January 29, 2013

crickets......


what happened, OP? Did you get things working?
posted by FauxScot at 4:49 AM on January 31, 2013

Response by poster: hi sorry my delay.
there wasnt a chance to go further on the setup,
so i didnt run more testing but may later in the year.
i think i need to play around a lot more with mounting location.
posted by edtut at 8:36 PM on April 3, 2013