Denoising the hob vibration signal using improved complete ensemble empirical mode decomposition with adaptive noise and noise quantization strategies

A novel denoising method combining complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and noise quantization strategies is proposed to solve the problem of the noise of the hob vibration signals disturbing the condition monitoring and feature extraction. The vibration signal is decomposed into several intrinsic mode functions (IMFs) and a residual based on CEEMDAN first. Considering that statistical indicators such as correlation coefficient and kurtosis are not effective in the presence of non-Gaussian noises and modulation because they primarily focus on the signal statistical distribution while ignoring the characteristics of the mechanism, a novel index based on the autocorrelation function analysis called periodic modulation for noise assessment (PMNA) is proposed to quantify the noise of IMFs. Further, IMFs are rearranged in the decreasing order of PMNA. A novel threshold joint with IMFs noise assessment (TJINA) varying with the combination of PMNA and the rearranged IMF retrieval is designed, which has advantages in the local smoothness and small fluctuation. On that basis, IMFs are divided into noise domain and signal domain, IMFs in the noise domain are denoised with TJINA and soft threshold function strategies. The proposed method is applied to the simulated signals with different input signal to noise ratios (SNRin) and two measured gear hobbing vibration signals. The comparison with some state-of-the-art approaches and the ablation experiment reveals that the proposed method performs better in enhancing the effective components and eliminating noise.

Automated summary

This paper proposes a novel denoising method combining complete ensemble empirical mode decomposition with adaptive noise to solve the problem of noise in gear hobbing vibration signals. The method effectively extracts the signal and eliminates noise through noise assessment and processing of intrinsic mode functions (IMFs).