A Novel Fault Diagnosis Method of Rolling Bearings Based on AFEWT-KDEMI

Abstract

According to the dynamic characteristics of the rolling bearing vibration signal and the distribution characteristics of its noise, a fault identification method based on the adaptive filtering empirical wavelet transform (AFEWT) and kernel density estimation mutual information (KDEMI) classifier is proposed. First, we use AFEWT to extract the feature of the rolling bearing vibration signal. The hypothesis test of the Gaussian distribution is carried out for the sub-modes that are obtained by the twice decomposition of EWT, and Gaussian noise is filtered out according to the test results. In this way, we can overcome the noise interference and avoid the mode selection problem when we extract the feature of the signal. Then we combine the advantages of kernel density estimation (KDE) and mutual information (MI) and put forward a KDEMI classifier. The mutual information of the probability density combining the unknown signal feature vector and the probability density of the known type signal is calculated. The type of the unknown signal is determined via the value of the mutual information, so as to achieve the purpose of fault identification of the rolling bearing. In order to verify the effectiveness of AFEWT in feature extraction, we extract signal features using three methods, AFEWT, EWT, and EMD, and then use the same classifier to identify fault signals. Experimental results show that the fault signal has the highest recognition rate by using AFEWT for feature extraction. At the same time, in order to verify the performance of the AFEWT-KDEMI method, we compare two classical fault signal identification methods, SVM and BP neural network, with the AFEWT-KDEMI method. Through experimental analysis, we found that the AFEWT-KDEMI method is more stable and effective.

Bearing Fault Diagnosis Approach under Data Quality Issues

“…The level of induced noise depends on the application and the used measuring device. In the vibration measurements, the additive noise followed a Gaussian distribution [43,44]. To study the effect of the noise on the performance of the Detrended Fluctuation Analysis (DFA) and the proposed method, we considered an additive zeromean white Gaussian noise with a variable standard deviation (σ) to simulate the different levels of the noise.…”

An Adaptive Spectrum Segmentation Method to Optimize Empirical Wavelet Transform for Rolling Bearings Fault Diagnosis

“…The root mean square, kurtosis, and skewness of empirical mode are combined into feature vectors, and then kernel density estimation and mutual information are used to classify fault features. Furthermore, Ge et al [32] makes the second EWT for empirical mode and extracts the part of sub-mode which does not satisfy the 95% confidence interval Gaussian distribution hypothesis test. Similarly, correlation coefficients are also used to calculate the correlation among empirical modes to extract the most valuable component [33].…”

Fault Feature Extraction and Enhancement of Rolling Element Bearings Based on Maximum Correlated Kurtosis Deconvolution and Improved Empirical Wavelet Transform

“…Combining advantages of both WT and EMD, the EWT can construct the wavelet basis in an adaptive way and decompose a signal according to the contained information. Therefore, EWT has attracted much attention and been used in a variety of applications, including medicine, biology [21,22], and machinery [23][24][25][26][27][28][29][30][31][32][33]. The traditional spectrum segmentation is achieved by detecting local maxima or minima of the spectrum.…”