A Method for Identifying Weak Faults of Rolling Bearings Around Railway Based on PCA&SVD-LMD

The fault signal of high-speed train rolling bearing from the railway track is often overwhelmed by strong noise, so that it is difficult to extract the fault features. In order to improve the efficiency of the extraction of the fault feature, and realize the intelligent identification for the rolling bearing faults on the high-speed train, a fault diagnosis method based on principle component analysis (PCA)&singular value decomposition (SVD) and local mean decomposition (LMD) is applied to extract the fault feature signal. Firstly, the feature signal is acquired by PCA&SVD denoising the raw acoustic signal which always featured as the nonlinear and non-stationary signal. LMD, a new self-adaptive time-frequency analysis method, is applied to decompose the feature signal into a series of product functions (PFs), each of which is the product of an envelope signal and a purely frequency modulated signal. Finally the fault feature vectors that extracted from the PFs are used as the observed inputs of pre-trained DHMMs presenting the failure states to evaluate the probabilities of the possible fault. The experimental results show that the proposed method can accurately identify the different states of rolling bearing faults and have advantages in less time-consuming and human intervention relatively from the acoustic information under the complex conditions.