The Application Of Blind Source Separation Algorithm In Rotor Fault Diagnosis

In the study of rotating machinery equipment condition monitoring and fault diagnosis, The fault feature extraction and pattern recognition are related to the reliability and accuracy of fault diagnosis, and they are core problem of rotating machinery fault diagnosis. As an important part of rotating machinery, the merits of the rotor performance affect the whole running state of rotating machinery directly. This paper adopted the Fast ICA method based on negative entropy and the blind source separation method based on maximum signal to noise ratio. The research work has been carried out through two methods of blind source separation(BSS) upon the problem of the fault feature extraction of rotor system.The four fault mechanisms of typical rotor are analyzed based on the research of the relationship of source signals and the acceleration sensor array, The blind source separation model is established according to the statistical features of vibration signal. In the view of the aliasing phenomenon of characteristics of stationary signals, Fast ICA algorithm based on negative entropy was put forward. The objective function is established by using the principle of maximum negative entropy and the criteria based on non gaussian maximum. The experiment proved that: The signal separation waveform and amplitude were consistent with the source signal’s, the main frequency are recovery accuratly, implementing the third-order convergence and the convergence speed is enhanced, the relevant analysis result is 0.9767. In order to improve the effect of time-varying non-stationary signal blind source separation, reduce the information redundancy of separated signals. The method of maximum signal-to-noise ratio is proposed based on blind source separation. The cost function of this method is maximum signal-to-noise ratio function and crosstalk error curve was astringed rapidly and stably,Combining with the rotor vibration signals separation experiments based on order analysis,order spectrum separation of the mixed signal were implemented,the order spectra of rotor vibration signal and noise signal spectra were obtained. The rotor fault characteristics were extracted.Based on the forefathers’ research, the rotor fault detection system is established by combineing with BK data acquisition device,the running state of the rotor in the practical work was simulated, it meet the requirements of signal blind source separation test and analysis.