Wavelet threshold analysis combined with EMD method for mechanical equipment fault diagnosis

Concerning the problem of noise interference in the mechanical equipment fault signal acquisition, a novel mechanical equipment fault diagnosis method of wavelet shrinkage threshold based on Bayesian estimation combined with EMD is proposed. The fault signal denoising characteristics of different scales are considered in the proposed method. A new threshold which is suitable for the situation of noise distribution is selected. Noise reduction can be gotten by improving the threshold function. The signal components decomposed and denoised by EMD is extracted with cross-correlation and kurtosis criterion to highlight the high-frequency resonance components with which the blindness of IMF component selection can be avoided. The results of analysis applied to simulated signal and the measured signal show that the equipment fault detection performance can be improved.