http://findarticles.com/p/articles/mi_m1215/is_n3_v198/ai_19239924

These guys mention a drive-by acoustic survey detecting spalling (classic fatigue), water etch (caused by lube contamination), roller defects, and "spun cone" (taper roller brg inner race?). In my opinion the complete list of possible bearing defects is a long one, and the symptomatic vibration chart kind of complicated.

See the first and second tables in:

http://vibrotek.com/article.php?article=articles/new94vi/index.htm

This includes complication details with probabilities! It's not for drive-by!

Dan's article is Circa 1997 and Duncan's article has data Circa 1993, so what's new? The rail bearings detection must have to deal with Doppler frequency shift due to train motion.

Walt

More rail bearing data that illustrates the difference between accelerometer and microphone data:

http://vibrotek.com/article.php?article=articles/factors/index.htm

The following is relatively new(1996):

http://vibrotek.com/article.php?article=articles/dlcvi96/anew.htm

There's also some articles on vibrotek.com on diagnosing fluid film bearings using envelope detection but that's not exactly a new tecnique; I have most of a copy of an article on that topic written jointly by Exxon and Boeing employee that dates bck to 1973.

Many things that are promoted as new merely have expired patents.

Duncan,

Do you have an explanation on why the microphone spectra showed higher relative energy above 7000 Hz compared to the accelerometer? Could it be non-bearing machine noise, non-machine background, bearing windage, bearing seal friction, or what?

Walt

VENDOR WARNING

If you measure bearing signals using accelerometers, there's a considerable amount of attenuation above the friction mode resonances in the 1 kHz to 4 kHz (roughly) range. That's why higher frequency systems like shock pulse have difficulty measuring signals from the opposite, usually inner, race. In our diagnostic software, DREAM, we include a calculation for the path loss through the bearing to compensate for the loss. For small bearings, the loss at medium frequencies is usually fairly small but for larger bearings like the rail car bearing in the example, it can be considerable.
On the other hand, for microphone based sound measurements, attenuation is less and the inner race is actually a better radiator than the outer race. Sound based measurements have the difficulty of greater interference from other sources. Signal attenuation, if handled correctly, can be a useful property, localizing the signal to the bearing under test.

Everything has tradeoffs.