Rotating Machinery Vector Dual-spectrum Analysis And Fault Diagnosis Applied Research

The vector spectrum analysis based on data fusion is one of the full information analysis method. Based on the full study of the dynamic character of the rotor, it fuses the double channel signal effectively, and reflects the vibration factually. While, the HOS (Higher order spectral) technique is a powerful tool to non-linear, non-Gaussian signal processing. It can suppress the additive colored Gaussian noise of unknown power spectrum effectively, and retain the signal phase information while it reflect the signal energy. As one style of the HOS, bispectrum, it posses all the properties of HOS, but has the lowest order and the minimum compute load. But, there is still a problem on bispectrum analysis, that is , this method is based on single channel signal, it means that it can only reflect the vibration future of the rotor in the sensor orientation. In this paper, the data fusion idea was introduced to the bispectrum analysis, and some innovation was worked out. The main content is as follows:First, it presented the HOS theory briefly, and studied the bispectrum analysis method, and its algorithm, characteristics, physical significance, and then put it into the fault diagnosis of rolling bearing. There different situation ware studied: 1. different fault styles; 2. different serious of the same fault under the same loading; 3. different loadings. The result indicates that: the bispectrum analysis method is feasible and effective.Second, to the question of the single channel bispectrum analysis, that is the uncertainty of different result from different sensor orientation, this paper extended the bispectrum analysis to vector signal, put forward the vector-bispectrum analysis method, and produced its definition, the algorithm and the nature, and put it to the feature extraction of rotating machinery. The simulation and experimentation indicate that, the vector-bispectrum analysis can fuses the information of the double channel effectively, and reflect the vibration roundly and factually. The method is feasible and effective.Third, the cesptrum and polycesptrum is introduced because of its advantage in the analysis of complex signal and modulated signal, and is extended to the analysis of vector signal. Then a new method, vector biscesptrum analysis, was put forward on the base of vector bispectrum, and with its definition, algorithm and properties. The analysis results of engineering signal proved that it can depress noise and identify periodic component of the signal effectively. The more important function is that this method can fuse the double channel information effectively, and provide reliable evidence for the fault diagnosis.And at last, a new analysis method based on vector bispectrum-RBFN was put forward after a brief introduction of RBF network. The research result of three typical faults in the rotating machinery indicates that the method is useful.