Double-optimized symmetric geometric mode decomposition with dispersion entropy and its application in feature extraction

Abstract

Symmetric geometric mode decomposition (SGMD) offers notable advantages in preserving the basic features of time series and in noise robustness. However, SGMD faces issues related to inaccurate mode decomposition and parameter selection. To address these problems, this paper proposes a double-optimized symmetric geometric mode decomposition with dispersion entropy (DSGMDDE). This algorithm incorporates dispersion entropy(DisE) as an indicator for mode reconstruction, enhancing the accuracy of mode decomposition. Furthermore, a double optimization algorithm is introduced to optimize parameters, thereby improving the effectiveness of the algorithm. By combining DSGMDDE with DisE, a feature extraction method named DSGMDDE-DisE is proposed. Simulation results demonstrate that, compared to four other mode decomposition algorithms, DSGMDDE offers higher decomposition accuracy and better robustness. Furthermore, DSGMDDE-DisE shows superior feature extraction capability compared to the other four feature extraction methods. Real-world experiment results indicate that DSGMDDE-DisE can more accurately distinguish between eight types of ship radiated noises (SRNs) and five types of Southeast University bearings (SUBs) fault signals.