Rolling mill bearings fault diagnosis based on improved multivariate variational mode decomposition and multivariate composite multiscale weighted permutation entropy. (31st May 2022)

• Record Type: Journal Article
• Authors: Zhao, Chen; Sun, Jianliang; Lin, Shuilin; Peng, Yan
• Is Part Of: Measurement. Volume 195(2022)
• Publication Date: 2022-05-31
• Subjects: Multivariate Composite Multiscale Weighted Permutation Entropy -- Improved Multivariate Variational Mode Decomposition -- Particle Swarm Optimization - Support Vector Machine -- Fault diagnosis -- Rolling mill multi-row bearing; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111190
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: MCMWPE is proposed to extract fault features from multivariate signals of rolling mill multi-row bearings. IMVMD is proposed by using WPE to optimize the parameters of MVMD and introducing an iterative acceleration factor. A novel fault diagnosis method is proposed based on IMVMD, MCMWPE, and PSO-SVM. Actual rolling mill data from a factory and experiment rolling mill data verify the diagnosis performance of the proposed method. Abstract: The multi-row bearings of rolling mills are subject to axial and radial loads. Due to the complex working conditions, it is difficult to achieve better results in fault diagnosis by analyzing signal directional vibration signals. In order to realize the fault diagnosis of bearings subjected to multiple directional loads, this paper introduces the idea of cooperative processing of multi-sensing signals and proposes a fault feature extraction method of improved multivariate variational mode decomposition (IMVMD) combined with multivariate composite multiscale weighted permutation entropy (MCMWPE). First, reconstruct the signal by variational modal decomposition (VMD). Upgrade VMD to multi-channel decomposition mode by considering the correlation of the multi-channel signal and optimizing its parameters. Secondly, propose the method to calculate multi-channel signal entropy. Represent bearing fault features by calculating the MCMWPE of multi-channel reconstructed signals. The method proposed in this paper is validated on experiment rolling mill datasets and actual rolling mill datasets from a factory. Entropy curves and PSO-SVM classification results show that IMVMD-MCMWPE can extract fault features better than other methods.