Fault diagnosis of flywheel bearing based on parameter optimization variational mode decomposition energy entropy and deep learning. (15th January 2022)

• Record Type: Journal Article
• Authors: He, Deqiang; Liu, Chenyu; Jin, Zhenzhen; Ma, Rui; Chen, Yanjun; Shan, Sheng
• Is Part Of: Energy. Volume 239:Part B(2022)
• Publication Date: 2022-01-15
• Subjects: Energy recovery -- Flywheel energy storage system -- Fault diagnosis -- Inverted residual neural network; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.122108
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Flywheel energy storage system is widely used in train braking energy recovery, and has achieved excellent energy-saving effect. As a key component of the flywheel energy storage system, the health of the bearing is greatly significant to realize the effective recovery of train braking energy. The vibration signal of the bearing presents complex nonlinear and non-stationary characteristics, which makes it difficult to diagnose the fault of the bearing. To solve this problem, a fault diagnosis method for bearing of flywheel energy storage system based on parameter optimization Variational Mode Decomposition (VMD) energy entropy is proposed. Firstly, the improved Sparrow Search Algorithm is used to optimize VMD parameters with the dispersion entropy as the fitness value. Then, the original signal is decomposed into a series of intrinsic mode components by using the optimized VMD algorithm, and the energy entropy of each component is calculated to construct the feature vector. Finally, an Inverted Residual Convolutional Neural Network (IRCNN) is used as feature vector input model for fault diagnosis. The experimental results show that the proposed method can effectively extract the bearing fault characteristics and realize accurate fault diagnosis, and the recognition rate reaches 97.5%, which is better than the comparison method. Highlights: The CSSA-VMD method can effectively carry out bearing vibration signal analysis. The energy entropy is introduced into the feature extraction. The input model of IRCNN as feature vector is superior to the comparison models. Experiments verify the effectiveness and superiority of the proposed method.