Impulsive Gear Fault Diagnosis Using Adaptive Morlet Wavelet Filter Based on Alpha-Stable Distribution and Kurtogram

The vibration signals extracted from gearbox are usually masked in heavy background noise. Analyzing the feature components of these signals is challenging and crucial to incipient fault diagnosis. The Kurtogram has been verified as a very powerful and practical tool in the mechanical fault diagnosis due to its advantages in detecting and characterizing transients in signals. However, the accuracy of the kurtogram is limited because it is based on short-time Fourier transform (STFT) or FIR filter in extracting transient characteristics from a noisy signal. Therefore, it is necessary to develop a more accurate filter to overcome its shortcomings and further improve its fault detection accuracy. Wavelet transform has been widely applied in the past as a powerful tool to analyze the non-stationary signals, and an alpha-stable distribution model is often used to describe the statistical features of non-Gaussian signals. In this paper, a Morlet wavelet filter is optimized based on the kurtogram and the alpha parameter of the alpha-stable distribution model. Through the analysis of simulation signals at different fault degrees and operating conditions, it can be concluded that alpha-stable distribution has better performance than kurtosis in measuring the non-Gaussian characteristics of an impulsive gear fault signal. The results obtained from the simulation and practical experiments confirm the superiority of the proposed method for incipient gear fault diagnosis.