Permutation entropy-based improved uniform phase empirical mode decomposition for mechanical fault diagnosis

Gearbox Fault Diagnosis Based on Improved Variational Mode Extraction

“…Many indicators have been proposed for impulse detection and fault feature extraction from rotating machinery vibration signals. Examples of widely used indicators are kurtosis [1]; the weighted kurtosis index (also known as KCI) [24,29,32,38]; envelope kurtosis (EK) [39]; the Gini index (GI) [40]; and entropy-class indices, such as information entropy (IE, also known as Shannon entropy) [26,34], sample entropy (SE) [35,41], permutation entropy (PE) [15], power spectral entropy (PSE) [42], envelope entropy (EE) [30], and envelope spectrum entropy (ESE) [43]. To verify the superiority of the proposed SDE index, the values of these 10 indices are calculated for the desired modes obtained by VME with different penalty parameters, α n (n = 1, 2, .…”

“…Essentially, the process of fault feature extraction comprises the elimination of noise and interference components in vibration signals. An effective approach to solve this problem is signal decomposition, the variants of which include wavelet decomposition (WT) [12][13][14], empirical mode decomposition (EMD) [11,15,16], local mean decomposition (LMD) [9,10,17,18], and empirical wavelet transform [19][20][21][22]. However, WT is not a selfadaptive signal analysis method because it is restricted by the selection of the wavelet basis function and number of decomposition levels [17].…”

“…However, WT is not a selfadaptive signal analysis method because it is restricted by the selection of the wavelet basis function and number of decomposition levels [17]. Although EMD can self-adaptively decompose a multimodulated signal into a series of intrinsic mode functions (IMFs), it lacks a theoretical basis and has some inherent defects, such as sensitivity to noise, end effects, and mode mixing [4,5,15,16]. Like EMD, LMD adaptively decomposes a multicomponent signal into several single-component AM-FM signals but encounters several technical problems, such as end effects and mode mixing [17,18].…”

Composite Multivariate Multi-Scale Permutation Entropy and Laplacian Score Based Fault Diagnosis of Rolling Bearing

“…(1) The CMMPE values of vibration signals are computed to extract the hidden fault features. (2) The LS algorithm is used to sort the raw features according to their importance from high to low for sensitive feature selection. (3) The BA-SVM based multi-classifier is built and applied to realize automatic fault recognition.…”

“…Rolling bearings has an indispensable role in many large rotating machines; once it works with a local failure, the normal operation of mechanical equipment will be disturbed, and serious economic losses and safety incidents will be caused if the fault cannot be detected in time [ 1 , 2 ]. Hence, it is of significance to implement health monitoring and early fault diagnosis of rolling bearing for a safe production [ 3 , 4 ].…”