Fault identification of rolling bearing with variable speed based on generalized broadband mode decomposition and distance evaluation technique

Abstract

In the fault diagnosis of rolling bearing, fault signals are often interfered by other noise signals, so that the fault characteristics are not obvious. In particular, vibration signals of rolling bearing with variable speed show non-stationary features, and the existing analysis methods are difficult to accurately extract their fault characteristics. The main idea of the previously proposed broadband mode decomposition (BMD) method is to search in the association dictionary containing both broadband and narrowband signals, so it can accurately extract the characteristics of stationary broadband signals. However, when applied to variable speed signals disturbed by strong noises, BMD algorithm is likely to produce modal confusion and has some errors in decomposition. Therefore, in this paper, a generalized broadband mode decomposition (GBMD) method is proposed to remove the rotating speed, and more accurate decomposition results are obtained. In order to realize the fault diagnosis of rolling bearing with variable speed, firstly, the original signals are decomposed by GBMD. Secondly, the eigenvalues of the first three components of the signals are calculated and the eigenvalue matrix is constructed. And distance evaluation technique (DET) method is utilized to screen out marked features. Finally, the marked features are input into backward propagation neural network (BP) for training and testing to identify the fault types of rolling bearing. Compared with ensemble empirical mode decomposition (EEMD), variational mode decomposition (VMD), and BMD, GBMD has good effect in anti-noise performance and fault feature extraction. Therefore, the combination of BP can achieve higher identification accuracy.