Fault diagnosis of rolling bearing based on improved CEEMDAN and distance evaluation technique

Abstract

In order to accurately identify the fault conditions of rolling bearing, this paper presents a fault diagnosis method based on improved complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and distance evaluation technique. In this method, to effectively extract potential fault-related information, vibration signals of rolling bearing in different fault conditions are decomposed into a set of intrinsic mode functions (IMFs) through improved CEEMDAN. The first eight IMFs containing most fault information are selected for extracting fault features. The original feature set is obtained including energy values, singular values and envelope sample entropy values. Then distance evaluation technique is implemented for selecting sensitive feature set and discarding irrelevant or redundant features. Subsequently, the sensitive feature set is fed into support vector machine (SVM) for automatically identifying rolling bearing fault conditions. The simulation results demonstrate that improved CEEMDAN is able to solve the problem of mode mixing and achieve a numerically negligible reconstruction error. Meanwhile experimental consequences indicate that the proposed method can acquire higher identification accuracy, as well as reduce the classifier computational burden.

A bearing fault diagnosis approach based on an improved neural network combined with transfer learning

“…In contrast to EEMD, the CEEMDAN adds the white noise to the residuals after each order of Intrinsic Mode Function (IMF), averages the IMF at that point and iterates over it. The CEEMDAN has the advantages of faster computational speed, better modal decomposition result and completeness compared with EMD or EEMD [25,26]. CEEMDAN-WPTD combines CEEMDAN for signal decomposition and WPTD for noise reduction.…”

A novel two-stream multi-head self-attention convolutional neural network for bearing fault diagnosis

“…Wang et al 30 proposed a multi-scale convolutional attention network (MSCAN) for bearing remaining lifetime prediction by adding the attention module to each layer of CNN. Ding et al 31 proposed a multi-scale attentional convolutional neural network model based on the squeeze-and-excitation network attention module. Wang et al 32 proposed a fault diagnosis method by introducing an attention mechanism in each residual module of the residual network.…”

Dynamic analysis of localized defects in rolling bearing systems

“…For instance, Jiang et al [5] studied rolling bearing fault identification using multilayer deep learning convolutional neural network. Ding et al [6] presented fault diagnosis of rolling bearing based on improved CEEMDAN and distance evaluation technique. In recent years, A lot of research works have been conducted to dynamic behaviors for rolling bearings.…”