Fault diagnosis method of rolling bearing based on Variational Modal Decomposition Multisynchrosqueezing Transform Combined with Long Short-term Memory Networks

• Published in: IEEE access p. 1
• Main Authors: Liang, Tao, Lv, Qingzhao, Tan, Jianxin, Jing, Yanwei, Sun, Hexu
• Format: Journal Article
• Language: English
• Published: IEEE 2024
• Subjects: Bearing fault diagnosis; Convolutional neural networks; Fault diagnosis; Feature extraction; Long short-term memory; Multisynchrosqueezing transform; Rolling bearings; Time-frequency analysis; Transforms; Variational modal decomposition; Vibrations
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2022.3177650

Aiming at the characteristics of weak vibration signal, strong interference, unevenness and nonlinearity in rolling bearing faults, this paper proposes a bearing fault intelligent diagnosis method based on variational modal decomposition (VMD) and Multisynchrosqueezing transform (MSST). First, the original vibration signal of the bearing is decomposed into multiple intrinsic mode functions (IMF) through the optimal parameter VMD, and then an effective IMF is selected for signal reconstruction according to the kurtosis value and mutual information. Secondly, MSST is applied to the reconstructed signal to obtain a time-frequency (TF) images with high energy concentration, and then extract the time-frequency and time-amplitude signals of the vibration signal in the TF images according to the ridge detection algorithm. Finally, these time series are used as input, using Long short-term memory (LSTM) network is trained to complete the intelligent classification and diagnosis of bearing failure. This method is compared with other fault diagnosis methods through test data sets. The results show that the method proposed in this paper is superior to other methods in recognition accuracy and recognition type, can accurately identify and classify bearing faults, and has strong generalization. Ability to better meet the needs of actual engineering.