Bearing fault detection via wavelet packet transform and rough set theory

A novel method for bearing fault detection through the sound signal translated from the vibration signal is introduced. The wavelet packet transform is used to preprocess the signal, then the computed wavelet coefficients are divided into clusters accordingly, important frequency ranges have a larger number of clusters than less important frequency ranges. These clusters extract the time-frequency information to get the fault characteristic of the bearing. In order to identify the fault, the method based on rough set theory is adopted. The result from experience proves that fault types can be identified and diagnosed by the above method. Furthermore, it attains nearly the same accuracy, compared with artificial neural network.