Rolling element bearing fault diagnosis using recursive wavelet and SOM neural network

This paper is focused on fault diagnosis of rolling element bearing due to localized defects i.e. rolling element and outer raceway on the bearing component, which is essential to the design of high performance rotor bearing system. A new fault diagnosis method based on recursive wavelet (RW) and SOM neural network, RW-SOM neural network is proposed. First, wavelet threshold de-noising is utilized to preprocess the raw vibration signal obtained by QPZZ-Ċ system, which can reduce the influence from the noise and to benefit to extract the characteristic signal. Then, a new method of feature extract based on recursive wavelet is proposed in order to solve the problems of bad real-time and the long window, which are born in traditional wavelet decomposition. Finally, bearing faults are classified using SOM neural network. The simulation results show that recursive wavelet combined with SOM neural network for fault diagnosis is effective and is superior to traditional wavelet decomposition.