Wavelet analysis applied to bearing vibration detection

Machinery condition monitoring methods contribute significantly to an industries' competitiveness under the current economic conditions. Reducing the chances of a major equipment failure not only reduces repair or replacement costs, but also may prevent environmental damage. The effectiveness of such a monitoring system depends on the signal processing approach used. Experience shows that, as yet, there is no single approach that is satisfactory to cover the detection of all damage events. The usual approach of using the Fourier transform to produce a frequency spectrum has the disadvantage that a whole range of features including those of short time duration, which disappear on averaging. A spectral line therefore represents the average of all the activity over the recorded time block, in that particular frequency window. Hence many features indicating early signs of failure could be lost. An alternative approach to such non-stationary vibrational data is the application of the Wavelet Transform.