Micro Gas Turbine Bearing Monitoring 1

You have the advantage of having a baseline measurement. Any deviation from the baseline indicates a failure. Rolling element bearings don't really wear, they fail by fatigue. Pits develop on the elements and races which show up immediately as a change in vibration profiles.

The vibration frequency can indicate exactly where the failure occured.

Outside race rotation is another failure mode for rolling element bearings but I don't know how that identifies in a vibration profile.

Thanks. My research indicates it typically follows the curve attached. My guess is that vibration levels increasing by 50% from the "as new" or "normal" condition indicates some form of failure is occurring. I thought 50% would be a good starting point as failure has an exponential function.

Why limit yourself to magnitude only? Software makes it possible to monitor vibration at many more levels using the same instruments. The Fourier transform is a key tool because you can monitor magnitudes at various frequencies. If a new frequency shows up on your spectrum analysis that will give an early warning of a developing failure.

section 9.2 here-


Note the different measurement units ( velocity, acceleration), acceptable values, and monitoring stuff up to 10khz.
Shoulders on bearing seats not being sufficiently square or bearing seat errors of cylindricity and size can generate measurable vibration when brand new.
Hence the various GM's narrow band, filtered, and defined wide band parameters.

"Overall" vibration measurements need to have defined frequency limits.
Some ISO standards measure velocity (RMS!) and stopped at 60,000 cpm / 1 kHz. I believe that was an attempt to create "something" of generic usefulness in commercial applications, and recognizing that
velocity measurements at higher frequency will mathematically tend to downplay and miss the tiny amplitudes that are still important. Acceleration measurements are generally necessary to get a whiff of those problems.

Thanks for your input guys, great read. Yes I'm using FFT and checking amplitudes at frequency. My software also produces RMS for velocity and acceleration so its good to see how this plays out.