161 / 2023-10-22 10:11:43

A Novel Bearing Fault Feature Extraction Method Based on Sparse Enhancement Dynamic Mode Decomposition

Dynamic mode decomposition, Rolling bearings, Fault feature extraction, Sparse optimization.

Special Sessions > SS13 - Signal Decomposition-Integrated Mechanical Fault Detection and Remaining Useful Life Prediction

In the field of fault diagnosis, the key issue is to extract fault characteristic information from noisy signals. To further resolve this problem, many mode decomposition algorithms have been proposed to separate fault features from signals. Dynamic mode decomposition (DMD) is a novel nonlinear mode decomposition method, which can effectively extract the dynamic characteristics of the signal by analyzing the specific DMD modes. However, mode selection and strong noise still limit the application of DMD in bearing fault diagnosis. In this paper, a novel sparse enhancement dynamic mode decomposition (SEDMD) is proposed to further improve the noise robustness of DMD. First, traditional DMD is used to process the acquired signal and obtain a series of DMD modes to be identified. Then, considering the prior knowledge that sparse distribution of fault features in one-dimensional signals, a sparse optimization model with GMC penalty is used to select fault related DMD modes. In this way, the identification of DMD modes and the suppression of noise mixed in the modes are completed simultaneously. Finally, the sparse enhancement DMD modes are reconstructed into one-dimensional time series. Subsequent experiments verified the superior noise reduction performance of the proposed method.