The following screen shot from C-Trend shows an Asset Status table for two identical bench grinders. Both grinders have had their grinding wheels removed to reduce the out-of-balance vibration so that we can see the bearing noise more clearly. This was done because cheap bench grinders are generally not very well balanced, which results in a large amount of vibration at the running speed (3000 RPM or 50 Hz in this case). This out-of-balance vibration, even on a new machine, is so large it usually completely swamps any other vibrations, such as those emanating from the bearings.

There is still a relatively high degree of out-of-balance on both grinders however, which is why their ISO readings are showing as being above the warning level of 2.8 mm/s, as defined by the ISO standard for this size of machine.

Figure 1 – C-Trend asset status tables for two bench grinders

Looking at the bearing status however, shows distinct differences between the two machines. Grinder#1 is a brand new motor with nothing wrong with its bearings and consequently the Bearing Noise and DeMod readings for both measurement points are relatively low.

Grinder#2 however, has had one of its bearings (MP2) deliberately damaged, which has resulted in large values for both these indicators of bearing health (Bearing Noise and DeMod). These can be explained as follows.

If we look at the vibration waveform for the two grinders they appear as shown in Figure 2 (Grinder#1 with good bearings) and Figure 3 (Ginder#2 with one bad bearing).

Figure 2 – Vibration waveform from a “good” bearing, (MP1 on Grinder#1)

Figure 3 – Vibration waveform from a damaged bearing, (MP2 on Grinder#2)

The waveform in Figure 2 shows a clear sinusoidal signal with a period of 20 milliseconds corresponding to the running speed of 50Hz (3000 RPM). This is as expected, as the period of a waveform is given by 1/f, where f = 50 in this case. This signal is due to the residual amount of out-of-balance on the grinder rotor, resulting in a peak amplitude of approximately 0.25g. A small amount of bearing noise is also clearly visible as a high frequency modulation (spikes) superimposed on top of the out of balance run speed waveform.

By comparison the vibration waveform in Figure 3 shows a large number of high amplitude spikes that are typical of bearing damage. The spikes are caused by the damaged bearing parts knocking against each other. The amplitude of some of the spikes is as high as 5g, almost totally obscuring the underlying sinusoidal 50Hz out-of-balance signal.

The fact that the bearing noise spikes are of very short time duration means they give rise to high frequency components in the resulting vibration frequency spectrum. Consequently, examining the high frequency content of a vibration waveform is a very good indication of the degree of bearing wear, which is precisely what is done when the bearing noise figure in BDU is calculated. The waveform is first of all high pass filtered to remove any low speed vibration signals and then the magnitude of the signal is calculated to determine the bearing wear as explained below.

Bearing Noise

This is a measure of the high frequency noise in Bearing Damage Units (BDU), where 100 BDU corresponds to 1g RMS (average) high frequency vibration.

100 BDU or 1g (approx 9.8 metres/sec2) of high frequency vibration corresponds to a fairly high level of bearing noise and so can be considered indicative of a damaged bearing. In other words it may be helpful to think of the Bearing Noise figure in BDU as being very roughly equivalent to “percentage” of bearing wear. For example the bearing waveform shown in Figure 2 for a good bearing gave a Bearing Noise figure of only 6 BDU. However the Bearing Noise figure for the damaged bearing waveform in Figure 3 was 89 BDU, due to the high frequency noise spikes present in this waveform.

Envelope Demodulation (DeMod)

An even better measure of bearing wear can be achieved by looking at the magnitude of the Envelope Demodulated vibration waveform (DeMod).

The DeMod figure is derived by amplitude demodulating the vibration waveform. This employs the same method that is used to separate the sound (e.g. the speech or music) from the channel frequency in a radio broadcast. The technique works on the principle that the worn bearing parts strike against each other as they spin and cause localised “ringing” in the bearing. This is exactly like striking a bell where the harder the bell is struck the louder it rings. In this example the ringing of the bearing acts as a “carrier frequency” and the bearing “noise” caused by the striking of the worn bearing parts is the superimposed signal that we wish to separate.

Hence by demodulating the vibration waveform from the bearing it is possible to determine how hard the bearing parts are being struck and therefore how badly damaged the bearing is. By examining the DeMod spectrum it is also possible to determine the frequency at which they are striking and therefore identify the likely cause of the fault. For example, a peak at the ball pass outer frequency ((BPFO) would indicate a fault on the outer race of the bearing.

For example, the DeMod spectrum of the damaged bearing on Grinder#2, shown in Figure 4 below, clearly shows a series of peaks at harmonics of 152Hz, which corresponds to the BPFO for this particular type of bearing running at 50Hz. In fact this bearing had a deliberately damaged outer race.

Figure 4 – DeMod spectrum from a damaged bearing, (MP2 on Grinder#2)

A major advantage of the Envelope Demodulation technique however, is that it tends to localise the bearing noise and this makes it easier to identify the source of the bearing noise.

In the example of the two grinders given here, it is easy to see from Figure 5 below that the DeMod readings of the two different measurement points on the damaged grinder are significantly different. The good bearing (MP1) gave a DeMod reading of 3.5 while the damaged bearing (MP2) gave a reading of 24, a ratio between the two readings of approximately 7:1. By comparison the ratio between the BDU readings of the good and bad bearings on Grinder#2 is only approximately 5:1 (18 BDU compared to 89 BDU).

Figure 5 – C-Trend status table for Grinder#1 and Grinder#2

The reason for this localisation of the DeMod signal is due to what is sometimes called the “disco effect” where, just like the music at a discotheque, the higher frequencies do not travel as far as the low frequencies. This is why you only hear the low frequency sounds outside a disco. In the same way the “ringing” noise due to the bearing wear that is selectively detected by the Demod method, does not travel as far as the lower frequency bearing noises detected by the BDU reading.  This is why DeMod is a very good tool for localising bearing faults on any particular machine.