Multi-objective Informative Frequency Band Selection Based on Negentropy-induced Grey Wolf Optimizer for Fault Diagnosis of Rolling Element Bearings

Abstract

Informative frequency band (IFB) selection is a challenging task in envelope analysis for the localized fault detection of rolling element bearings. In previous studies, it was often conducted with a single indicator, such as kurtosis, etc., to guide the automatic selection. However, in some cases, it is difficult for that to fully depict and balance the fault characters from impulsiveness and cyclostationarity of the repetitive transients. To solve this problem, a novel negentropy-induced multi-objective optimized wavelet filter is proposed in this paper. The wavelet parameters are determined by a grey wolf optimizer with two independent objective functions i.e., maximizing the negentropy of squared envelope and squared envelope spectrum to capture impulsiveness and cyclostationarity, respectively. Subsequently, the average negentropy is utilized in identifying the IFB from the obtained Pareto set, which are non-dominated by other solutions to balance the impulsive and cyclostationary features and eliminate the background noise. Two cases of real vibration signals with slight bearing faults are applied in order to evaluate the performance of the proposed methodology, and the results demonstrate its effectiveness over some fast and optimal filtering methods. In addition, its stability in tracking the IFB is also tested by a case of condition monitoring data sets.

Early weak fault diagnosis of rolling element bearing based on resonance sparse decomposition and multi-objective information frequency band selection method

“…Decomposition residual is presented in Figure 4(c). The parameters of MOGWO are set same as reference (Gu et al, 2020), and the calculated values of SE negentropy and SES negentropy are presented in Figure 5(a) after iteration. Then, the average value of the SE and SES as shown in Figure 5(a) is calculated, and the maximum average value of 1.5 is selected as the best optimal one.…”

“…In Wang et al (2013), the vibration signal of fault bearing is filtered by the wavelet packet transform first; then, the kurtosis of the power spectral of the envelope of the filtered signal was used as index to improve performance of SK. However, fewer harmonics of fault characteristic frequency (FCF) in the ES might be induced by using the above-stated frequency-domain indexes (Gu et al, 2020). Another aspect of signal processing of REB mainly focuses on measuring cyclostationarity (Borghesani et al, 2014) rather than the impulsiveness in OIFB selection (Antoni and Borghesani, 2019; Moshrefzadeh and Fasana, 2018; Smith et al, 2019).…”

Weak fault feature extraction of rolling element bearing based on variational mode extraction and multi-objective information fusion band-pass filter

“…where 𝜃 represent the cyclic frequency. The multi-objective grey wolf optimizer (MOGWO) algorithm [24][25] is used to optimize the calculation process of Eqs. ( 24) and ( 25).…”

“…5. The parameters of MOGWO are set same as reference [25], and the calculated values of SE Nenentropy and SES Nenentropy are presented in Fig. 5(a) after iteration.…”

Simultaneously Low Rank and Group Sparse Decomposition for Rolling Bearing Fault Diagnosis

“…To precisely keep abreast of the dynamic health condition for REBs, it is important to develop techniques to detect the incipient fault, in which vibration feature extraction is assumed to be an effective approach [ 4 , 5 , 6 ]. When the local fault happens in the REBs, there will be periodic impulses generated by striking the defect surface with rollers in the vibration signal [ 7 , 8 ]. However, the periodic impulses are expected for weak to be detected in the early fault stage, leading to a great challenge for bearing fault diagnosis [ 9 , 10 ].…”