Research On Feature Extraction Of Nonstationary Signals And Intelligent Diagnosis Method Based On Wavelet Theory

There are a large number of non-stationary signals in the fields of condition monitoring and fault diagnosis for mechanical equipment. Researching and developing effective engineering methods for processing non-stationary signals are necessary for promoting sustained development of fault diagnosis technology. Over the last few years, rapid expanding methods and theories for processing non-stationary signals, especially wavelet theory, provided powerful tools for condition monitoring and fault diagnosis for mechanical equipment. In this paper, application problems of wavelet theory are investigated, such as signal de-noising, fault feature extraction, modal parameters identification and intelligent fault diagnosis. The main research works are as follows:(1)The basic theories about wavelet transform are introduced. Then, the border distortion problem in the process of wavelet transform is discussed and merits of border expending methods for solving border distortion are compared. A new border expanding method based on Autoregressive Intergraded Moving Average model is employed. Compared with border expanding method based on AR model, ARIMA model changes unstable series into stable ones by difference and thus possesses better effects for non- stationary signals.(2)A multi-scale de-noising algorithm based on the convolution type of wavelet packet transformation is presented. This algorithm overcomes shortcomings of the classical wavelet packet transformation, in which the length of sequences obtained always decreases by decomposition scales. The new algorithm improves estimated method of white noise standard deviation at each scale and thus keeps the main edges of signal well. A new thresholding function is employed in this algorithm, which is simple in expression and as continuous as the Donoho's soft thresholding function, and overcomes the shortcoming that there is an invariable dispersion between the estimated wavelet coefficients and the decomposed wavelet coefficients of the soft-thresholding method. Simulation results indicate that the new de-noising method suppresses the Pseudo-Gibbs phenomena near the singularities of the signal effectively and achieves better MSR performance and SNR gains than de-noising method based on classical wavelet packet transformation...