Incipient Fault Feature Extraction of Rolling Bearing Based on Signal Reconstruction

Abstract

In the incipient fault vibration signals of rolling bearings, weak fault features are easily submerged in strong background noise and difficult to be extracted. The sparse decomposition method can perform well in the extraction of weak fault features, but the low signal-to-noise ratio (SNR) would cause excessive decomposition. To enhance the fault features and maintain the time–frequency structure of fault impulses, a novel incipient fault feature extraction of rolling bearing based on signal reconstruction is proposed. Firstly, the Teager energy operator (TEO) is used to obtain the envelope of the impulse components in the vibration signal, which is also sensitive to noise and would be seriously affected by strong noise. Secondly, a Savitzky–Golay (S–Golay) filter based on the particle swarm optimization (PSO) algorithm is adopted to suppress the noise in the TEO envelope and generate a smooth envelope signal. Finally, the fault signal is reconstructed by the multiplication of the filtered TEO envelope signal and the original signal. The reconstructed signal can maintain the structural characteristics of the original fault impact signal and can provide reliable feature enhancement signals for further sparse decomposition, multi-source vibration separation, and other operations. Simulation signals and experiments verify the effectiveness of this method in extracting early fault features under low SNRs.