In many cases people only focus on the peaks in the vibration spectrum; especially the tops of the peaks. But what happens at the bottom of the spectrum is also very important. The "noise floor", and the shape of the base of the peak, provides useful information.
The Noise Floor
The bottom of the spectrum is the noise floor - but we will focus on the area of the spectrum between the peaks. If the machine is running smoothly, the area between the peaks will be flat and have a low amplitude. But if there is a source of "broadband" energy, it can generate vibration across the entire spectrum or just in specific areas. Rubs, impacts, looseness, cavitation, turbulence, and flow noise can all generate this "energy".
The "energy" can excite natural frequencies, so in specific areas we might see the noise floor rise up. Depending upon the amount of damping, and the amplitude of the energy, the spectrum might rise up quite significantly.
The noise itself can be quite focused (i.e. we will see it in a small band of frequency), so we might see the area around the pump vane rate rise up if there is cavitation. And we might see a large, broad peak below 1X if there is flow turbulence.
Broad Based Peaks
If there is a "pure" source of vibration, such as unbalance or the pump vane rate, then we expect to see a narrow peak. However if we look at the base of the peak and see that it is quite broad, we need to ask; "why is it so?"
There are actually a few reasons:
1. If that source of vibration is exciting a resonance, then the base will be broad - and the amplitude of the peak will be much higher than it would be if there were no resonance.
2. If there were sidebands, however your resolution was insufficient to be able to see the sidebands, then the peak may look broad. It is therefore important to use 3200+ lines to make sure you know what you are dealing with.
3. As mentioned above, if there is cavitation, the area around the pump vane rate peak may also be elevated.
There are two more very important points:
1. If you see an elevated noise floor, check the amplitude scale. If the highest peak in the spectrum has a very low amplitude, the noise floor will "appear" to be higher than what you are used to. Likewise, if the scale is set to a high amplitude, you may never notice a raised noise floor. That is just another reason why a logarithmic scale is very useful.
2. If there is a raised noise floor - look in the time waveform. The waveform will show you the impacts, rubs, bursts of energy from cavitation, and so on. Of course, you need to collect the time waveform correctly, but that is covered in another tip.
Here are a few examples:
Here is an interesting example. The first spectrum is 800 lines. Notices the broad bases of the peaks. Could it be resonance. The next spectrum is 3200 lines - you can see there is more to it. And the third spectrum is zoomed in to the base of the peaks.
This looks very noisy, however the amplitude is very low:
This case looks like there is no noise - but the peaks are very high in amplitude:
Here is the
...
