MESMD and Its Application to Fault Diagnosis of Gear

Gear vibration signals collected from gearboxes typically exhibit weak fault characteristics. Consequently, the fault information present in a single-channel signal is frequently inadequate due to the transmission of vibrations and the presence of external noise. This article proposes a novel multivariate signal decomposition method, multivariate extreme-point symmetric mode decomposition (MESMD), to accurately decompose multichannel signals at the same time. A new multivariate soft-sifting stopping criterion (MSSC) and completely adaptive projection (CAP) are employed in MESMD. MSSC adaptively adjusts the number of sifting iterations for each component. CAP improves decomposition accuracy. Subsequently, MESMD is employed to analyze the simulated and experimental data of multivariate gear fault signals. Finally, MESMD is compared with multivariate empirical mode decomposition (MEMD), multivariate variational mode decomposition (MVMD), and CAP multivariate local characteristic-scale decomposition (CAPMLCD). The simulation and experimental results indicate that MESMD outperforms MEMD, MVMD, and CAPMLCD in decomposition accuracy, resistance to mode mixing ability, and computational efficiency.