A rotating machinery fault feature extraction approach based on an adaptive wavelet denoising method and synthetic detection index

Feature extraction from vibration signals plays a vital role in rotating machinery fault diagnosis. The noise contained in the signals will interfere with the fault feature extraction result. Wavelet denoising (WD) is a commonly used method to reduce the noise, but its parameters are generally selected based on subjective experience. With this problem in mind, an adaptive wavelet denoising (AWD) method is proposed in this paper. Using permutation entropy to evaluate the signal noise level and taking its minimum value as the fitness function, the whale optimization algorithm is applied to optimize the WD parameters. Based on the AWD method and a synthetic detection index, a new feature extraction approach is proposed. Results from simulation experiments and engineering applications prove that the signal denoising performance of the AWD method and the fault feature extraction approach are satisfactory.