Application And Research Of Time-frequency Analysis In Fault Diagnosis Of Bearing

The traditional signal analysis methods have great advantages when they deal with the stationary signals, but they are helpless at the presence of the nonstationary stochastic signals. Although a great many of methods in modern signal processing could process the nonstationary stochastic signals effectively, they also could not avoid the defects of separation between time and frequency domains. As a branch of the modern signal processing, TF(time-frequency) analysis is very important, it combines the time domain and the frequency domain. Signal is presented in terms of joint TFD, overcoming the shortcomings which are took place by processing signal only in time or frequency domain, and the characteristics which are difficultly obtained in time or frequency domain are clearly presented in TF plane. Apparently, TF analysis is an ideal signal processing method.The vibrating signal of the rolling bearing which is affected by various impacts when it is rolling is complicated and nonstationary. So it is suitable for analysis with TF methods. In this thesis, the bearing is studied, and the vibration signal is analysed under the three modes of normal, single point fault in inner ring and ball. TF characteristics of the fault signal are extracted by the TF analysis methods, so the fault diagnosis is realized.In this thesis, we thoroughly study the theory of TF, and emphasize research of the conflict between cross-term interference and TF resolution of the quadratic TFD simultaneously. Therefore, the main problem to be discussed is how to restrain the cross-term interference and keep high TF resolution simultaneously. The main work is as follows:(1) Several typical TFDs and their properties are studied, and the disadvantages which resulted from the inherent nature of their kenel functions are pointed out, then the developed algorithm is further studied.(2) Adaptive chirplet transform and it's fast algorithm is studied in order to further solve the aforementioned problems, and then explore the properties and its applications.(3) Some typical TFDs are applied to fault diagnosis of rolling bearing and the diagnostic results are analyzed thoroughly. From comparison we can obtain that the adaptive chirplet transform can restrain cross-term interference and keep high TF resolution simultaneously, and the TF characteristic is better extracted in fault diagnosis. Therefore, it is an much more ideal method for fault diagnosis of rolling bearing.