Noise reduction method for gear crack impact signals based on ITD-WPT

WU Donghao; LIU Lijuan; MA Liyong; KANG Kai; YAN Zexu

doi:10.16578/j.issn.1004.2539.2026.03.020

Abstract：[Objective]Aiming at the problems of poor denoising effect and insufficient retention of key features when traditional denoising methods process gear crack impact signals,a denoising method(ITD-WPT)integrating intrinsic time-scale decomposition(IDT)and wavelet packet transform(WPT)was proposed to achieve efficient denoising and accurate feature retention of gear crack impact signals.[Methods]Firstly,the noisy signal was decomposed into several proper rotation(PR)components by ITD,and high-frequency components containing noise were separated by identifying abnormal PR components to lay a foundation for precise denoising.Secondly,WPT was used for multi-scale decomposition of the noisy high-frequency components,and wavelet packet coefficients were extracted through refined threshold processing to achieve accurate noise suppression.Then,the wavelet packet denoised components were reconstructed,and signal reconstruction was performed using ITD combined with noise-free low-frequency components to obtain high-quality denoised signals.Finally,the method was applied to simulated cracked gear signals and test broken tooth signals,and comparative verification was conducted with classical methods such as moving average,empirical mode decomposition(EMD)and wavelet denoising.[Results]The results show that the denoising effect of this method is significantly superior to the comparative methods:when the noise variance is 50,the signal-to-noise ratio(SNR)reaches 9.490 dB and the correlation coefficient is 0.944;when the variance is 100,the SNR is 4.012 dB and the correlation coefficient is 0.838.Moreover,it could effectively retain the impact characteristics of signals,providing a reference for the denoising of gear crack impact signals.