Intelligent Fault Identification for Rolling Element Bearings in Impulsive Noise Environments Based on Cyclic Correntropy Spectra and LSSVM

Abstract

Rolling element bearings are important components in various types of industrial equipment. It is necessary to develop advanced fault diagnosis techniques to prevent unexpected accidents caused by bearing failures. However, impulsive background noise in industrial fields also presents a similar fault-excited characteristic, which brings interference to the fault diagnosis of rolling element bearings. Focusing on this issue, this paper proposes a new feature extraction method based on the cyclic correntropy spectrum (CCES) for intelligent fault identification.. In this study, the cyclic correntropy (CCE) function is introduced to suppress the impulsive noise. A simplified frequency spectrum named CCES is obtained for the feature extraction. Then, narrowband kurtosis vectors are extracted from the CCES. Finally, these extracted features are used to train the least squares support vector machine (LSSVM) for the fault pattern identification. Analyses of two bearing datasets, including train axle bearing data that are contaminated by impulsive noise are used as case studies for the validation of the proposed method. To illustrate the advancement of the new method, performance comparisons with two recently developed methods are conducted. The experimental results verify that the proposed method not only outperforms these two methods but also exhibits a stable self-adaptation ability.

Review of research on signal decomposition and fault diagnosis of rolling bearing based on vibration signal

“…The classical machine learning algorithms: BP, DT, SVM and LSSVM, the intelligent optimization algorithms improved Y NL and NS [189] Y Small-sample, NL [190] Y NL and NS, impulsive noise [191] Y NL and NS [192] -Big data, quantum Intelligent optimization algorithms-SVM [194] Improve SVM parameters Slow optimization speed, many adjustment parameters -Mixed noise [195] Y Complex imbalanced data [196] Y NL and NS [197] -Multi-channel signals [199] Y EFS [200] Y NL and NS [201] Y SVM and the deep learning algorithms: CNN and LSTM are compared from the key features, application difficulties and application occasions (table 3). By combing and comparing the common methods of fault feature identification, the classical machine learning algorithms and deep learning algorithms can achieve better application results in different occasions.…”

“…Li et al [189] proposed a fault diagnosis method based on deep structure and sparse LSSVM, which combined LSSVM with sparse theory and effectively solved the insufficient sparsity problem of LSSVM. Zhao et al [190] proposed a fault diagnosis method based on cyclic correntropy spectrum and LSSVM, which had stable adaptive ability. Jiao et al [191] compared LSSVM with radial basis function neural network and probabilistic neural network, indicating that LSSVM had better generalization performance in bearing fault classification.…”

The Prediction of the Remaining Useful Life of Rotating Machinery Based on an Adaptive Maximum Second-Order Cyclostationarity Blind Deconvolution and a Convolutional LSTM Autoencoder

“…Rotating machinery environments are severe and involve many reasons for failures. Sensors are used to collect the data from the rotating machinery; signal processing methods are deployed to remove the noise; and deep learning methods are applied as prognostics to assess the rotating machinery for prognostics and health management (PHM) [ 1 , 2 , 3 ].…”

Rolling Bearing Fault Diagnosis Based on WOA-VMD-MPE and MPSO-LSSVM

“…Gao et al [ 30 ] fused singular entropy, energy entropy, and permutation entropy to obtain complementary features, combined with the PSO algorithm to optimize LSSVM, and successfully completed the diagnosis of bearing faults. Zhao et al [ 31 ] extracted narrowband kurtosis vectors from the cyclic correntropy spectrum (CCES) as feature vectors of LSSVM for the early detection and classification of locomotive axle bearing faults. Zhu et al [ 32 ] used VMD to decompose the bearing vibration signal, used the fuzzy entropy of each IMF as the feature vector, optimized the LSSVM model by the gray wolf optimizer (GWO) algorithm, and finally completed the identification of the rolling bearing faults.…”