Feature extraction of combined failures of rolling bearings based on adaptive variance symplectic geometry model decomposition

Abstract

Vibration signals of rolling bearings with faults are characterized by strong nonlinearity and non-stationarity, making it difficult to extract fault information. In the engineering practice, the fault of bearing is often represented as compound. Compared with single fault, it is more difficult to extract feature information of combined faults. Symplectic geometric mode decomposition represents better decomposition performance and can provide better protection to geometry structure of phase space. However, the performance of feature extraction is affected by ineffective symplectic geometrical components when processing signals with strong noise and weak failure feature. Meanwhile, there is a lack of effective standard for component option. To solve these problems, an adaptive variance symplectic geometric mode decomposition method is proposed. To decrease the interference of noise and strengthen fault features in original signal, the variance sequence of original signal is constructed. To prevent the influence of improper embedding dimension on decomposition, the embedding dimension of track matrix of variance sequence is adaptively determined with the maximum of margin factor as criterion. To solve the problem of symplectic geometric mode decomposition being lack of effective standard for component option, optimal fault component is determined by the maximum of activity parameter. Faults identification of bearings is accomplished with the power spectrum of the optimal fault component. To ascertain the efficacy and superiority of the proposed method, it was compared with symplectic geometric mode decomposition method. Results indicate that the adaptive variance symplectic geometric mode decomposition method proposed can effectively suppress the noise, reduce the influence of invalid symplectic geometric component and accomplish the adaptive option of effective components, effectively extract features of combined faults and enables precise judgment of combined faults of bearings. Furthermore, compared with the contrast method, the feature frequencies are distributed in a lower frequency band, which is beneficial for the real-time monitoring of faults in engineering applications.