Using Structure-borne Sound Analysis to Predict Bearing Damages in High-speed Ventilators

Abstract: By measuring structure-borne sound from bearings, a conclusion about their condition can be made. The measurements are done with accelerometers which should be attached as close as possi-ble to the bearing. The signal from the accelerometer may have to be lowpass filtered before being A/D-converted and stored. These steps have to be done in real time. The analysis of the signal is done afterwards. First, the frequency spectrum of the measure-ments is produced. This has to be done in a certain way to maintain the amplitudes. A window function has to be applied to reduce the effect of spectral leakage. When the frequency spectrum is at hand, frequencies can be identified. Bearing damages cause special frequencies (rollover frequencies) which can be calculated if the geometry data of the bearing is known. The frequencies, their harmonics and sidebands of them and their harmonics can be observed in the spectrum. Other frequencies which are either depending on or independent of the rotation speed will also be found in the spectrum. The conclusion of the bearing condition can be made by observing the amplitudes of the bear-ing frequencies or by using a broadband diagnostic method, which measure the 'peakness' of the whole spectrum.Within this dissertation project, many trials were done to study the theory. In the beginning many failures disturbed the results, but the experiences from the trials made it possible to pre-vent failures later on. The analysing software was developed in Scilab which is an open sources version of Matlab.