Gearboxes are integral components in the transfer of mechanical power. Given the non-stationary nature of their vibration signals, this study proposes a novel fault diagnosis method for wind turbine gearboxes. This method integrates Dung Beetle Optimization (DBO)-optimized variational mode decomposition (VMD) with improved refined composite multiscale residual dispersion entropy (IRCMRDE). To enhance the accuracy and robustness of signal decomposition, the DBO algorithm adaptively optimizes the VMD parameters. It uses a multi-objective weighted fusion criterion incorporating envelope entropy, kurtosis, and spectral kurtosis, collectively enhancing sensitivity to early-stage fault features. Subsequently, key modes are selected based on the intrinsic mode function (IMF) energy criterion. The IRCMRDE algorithm then extracts multiscale features. This feature extraction method improves sensitivity to local dynamic changes by incorporating median filtering. Finally, a support vector machine (SVM) classifies the faults. Experimental validation demonstrates the superior performance of the proposed framework. It achieves a diagnostic accuracy of over 98.89% under various faults and challenging operating conditions, confirming its efficiency, accuracy, and robustness in practical gear fault diagnosis. The results also prove its stability against parameter variations and its efficacy in detecting weak faults.
Li et al. (Thu,) studied this question.