The vibration at the surface of the bearing housing has a great story to tell. It can tell you about the smallest spall on the surface of a bearing, it can tell you about damaged rotor bars, it can tell you about worn teeth in the critical gearbox, and so much more - but you'll never know unless you take the measurement correctly.
There are two main opportunities to make mistakes: the mounting of the accelerometer, and the settings used by the analyzer.
Mounting the accelerometer
There are a few simple rules:
1. Mount the accelerometer at a location with a good transmission path to the bearing. Fins, covers, and brackets are not good locations.
2. Mount the accelerometer in the same place every time. Repeatability is essential.
3. Prepare a clean flat surface on the bearing or use a mounting pad. If you are interested in high frequency vibration (and you should be!) you should use flat magnets, quick-connect mounts, or stud mounts.
Analyzer settings
There are a few simple rules:
1. Make sure the Fmax is high enough to cover 3.5 times the highest forcing frequency of interest (e.g. gearmesh), and at least 10 times the highest bearing frequency expected. This ensures you don't miss any harmonics. If you don't know the highest frequency generated by the bearings, you should be safe by using an Fmax of 70 times the turning speed of the shaft.
2. Make sure the resolution is sufficient to identify all the individual frequencies and sidebands. 3200 lines is a good place to start.
3. Make sure you take plenty of averages, especially if you use overlap averaging (- and you should). The shaft of the machine should turn at least 50 times during your measurement (and if you were to test it again immediately the data should look the same). For example: 1800 RPM machine, Fmax: 3000 Hz, Overlap: 67%, Averages: 5 - there are 75 shaft rotations. That's good.
4. Consider acquiring two measurements so that you can quickly and easily identify unbalance, misalignment, looseness, resonance and other faults AND gearbox, motor, bearing, and cavitation faults without having to toggle units or zooming the frequency scale:
a. Fmax: 70 times the turning speed and units of acceleration
b. Fmax: 10 times the turning speed and units of velocity
Remember, just because the spectrum you view in the analysis software looks like a normal spectrum does not mean the measurement was taken correctly. That's the problem with vibration analysis; it is hard to tell when you are making mistakes (unless the machine fails unexpectedly).
There is a great deal more we could say on this important topic, but hopefully these comments will give you something to think about.
Tip provided by Jason Tranter, Managing Director, Mobius Institute
