A Fault Diagnosis Method Of Rotating Machine Based On The HHT And Box Dimension

The vibration signal analysing is an effective way for the machine condition monitoring and fault diagnosis, and the important fault information are often reflected by the transient signals and time-varying signals. Because the traditional signal processing method based on the FFT has been unable to meet certain need, the new signal processing methods such as wavelet transform, fractal etc. are proposed. The Hilbert-Huang Transform put forward by Huang etc. in 1998 is used to deal with the stationary and non-stationary signals. Two important steps of the Hilbert-Huang Transform and its relectant problerms were derived and analysed, and the application of non-stationary signal processing and rotating machinery fault diagnosis are studied specialy.It is proposed in this thesis that the decompostion number of the EMD is controlled by the box dimension of the residual signal, and the causes of the end problems of the HHT dealt with by the AR model are analysed. In the IMF criteria, the track of the energy difference has a good physical meaning, and it is also used to discuss the shortcoming of other methods including IMF and data extension. In another aspect, the impact of the random noise signal and sampling points per period of signal on HHT are analysed. Comparison on diagnostic methods and resultes between the HHT and FFT is proposed. The simulation on the common fault signals of the rotating machinery is studied and applied to some fault data of speeding-up rotor which has the non-stationary characteristics and local rubbing rotor fault data, to obtain some useful conclusions.In addtion, the failure characteristics mechanism of rotating machinery, the rotor rubbing fault diagnosis mechanism and diagnosis methods are discussed.