Sensitive method for detecting tooth faults in gearboxes based on wavelet denoising and empirical mode decomposition

Many signal processing methods have been developed to detect gear system faults. However, signal noise greatly influences the monitoring process. In addition, useful fatigue information can be misinterpreted by other useless oscillation components in vibration signal and noise. These conditions lead to unclear results that hinder researchers from effectively detecting faults. To overcome this problem, this study first adopts wavelet theory to remove noise and then utilizes the empirical mode decomposition characteristics of the Hilbert-Huang transform to analyze useful Intrinsic mode functions (IMFs) on the basis of signal modulation and correlation theory. Sifted IMFs are then reconstructed as new signals called D-E signals. Finally, Hilbert energy spectrum and kurtosis value are used to complete fault diagnosis. This study compares the proposed method with the Discrete wavelet transform (DWT) method to verify the superiority of the proposed method. Experiment results from using different degrees of gear crack demonstrate that the proposed method is more sensitive in gear fault detection than the DWT method.