The vibration signal of the hobbing machine is susceptible to changes in the frequency domain distribution owing to the influence of the machine’s inherent vibration and random pulses, which affects the condition monitoring and wear prediction of the hobbing machine. Variational mode decomposition (VMD) can compensate for the mode mixing problem of ensemble empirical mode decomposition (EEMD) method owing to its inherent equivalent filtering property. However, the decomposition performance of VMD depends heavily on two hyperparameters that need to be set in advance, i.e., the number of bandwidth-limited intrinsic mode functions (BLIMFs) K and the penalty factor α. Thus, a hybrid signal denoising and feature enhancement method based on parameter adaptive variational mode decomposition (PAVMD) and autocorrelation analysis is proposed in this study. First, gradient-based optimizer (GBO) is introduced to optimally select the decomposition parameter of VMD, and then a series of BLIMFs are obtained via VMD. Further, an evaluation criterion called enhanced periodic modulation intensity (EPMI) based on autocorrelation analysis is built to quantify the noise-related degree of each BLIMF. Finally, the denoised signal is obtained based on the proposed reconstruction strategy. The comparison with other methods in both the simulation and the actual signal analysis reveals that the proposed method has a better performance in terms of eliminating the high-frequency noise, reserving higher effective frequencies, and the evaluation indicator.

No abstract