关键词: 发电机, 小波降噪, VMD, 信号处理, 麻雀搜索算法

Abstract: There are some constant errors between the original signals and the denoised signals filtered by the traditional wavelet threshold function. The variational mode decomposition (VMD) sets different penalty factors α and different numbers of modal decomposition layers K in signal processing, affecting the noise reduction enormously. Therefore, this paper introduces the improved wavelet threshold denoising method and sparrow search algorithm-variational mode decomposition (SSA-VMD) into generator vibration signal processing. First, SSA is employed to optimize the VMD decomposition parameters α and K with the minimum envelope entropy as the objective function to achieve the optimal noise reduction. Then, the noisy vibration signal is decomposed into K intrinsic mode functions (IMF) by VMD, and the improved wavelet threshold is employed to denoise again the components below the set IMF threshold. Finally, the denoised IMF is recombined to obtain the ultimate noise reduction signal. An analysis through Matlab shows the improved wavelet threshold function denoising method together with SSA-VMD effectively reduces the mean square error of the signals and markedly improves the denoising of generators’ vibration signals.