A new bearing weak fault diagnosis method based on improved singular spectrum decomposition and frequency-weighted energy slice bispectrum

Abstract

The embedding dimension parameter has a significant effect on the decomposition results of singular spectrum decomposition (SSD). However, in the conventional SSD method, the embedding dimension of each iteration is determined by the empirical formula, which may lead to the problem of mode mixing and over decomposition. Aiming at this issue, mode mixing index (MMI) and over decomposition index (ODI) are proposed in this paper, and the product of the two indices is used as the basis for selecting the optimal embedding dimension from preset interval for each iteration. Subsequently, to extract the weak fault symptoms from the susceptive mode components, a new spectrum analysis technique named frequency-weighted energy slice bispectrum (FWESB) is proposed. In this method, Gaussian white noise and uncoupled frequency component in the signal are suppressed, thereby highlighting the weak shock fault characteristics in the vibration signal. The analysis results of the simulation and the experiment prove the validity of the proposed method in alleviating mode mixing and extracting weak fault symptoms of rolling bearings. Besides, the average results of multiple tests in experimental and engineering applications show that for experimental signals, compared with SSD, empirical wavelet transform (EWT) and fast kurtogram (FK), the characteristic frequency intensity coefficient (CFIC) of the proposed method are increased by 210.36%, 46.30% and 187.86% respectively. For practical engineering signals, compared with SSD, EWT and FK, the CFIC of the proposed method are increased by 127.13%, 190.89% and 301.51%, respectively.