Method of rotor fault information enhancement based on robust independent component analysis

Key words:  robust independent component analysis  dual-tree complex wavelet transform  rotor system  information enhancement  fault diagnosis

Abstract:

Considering the frequency aliasing and weakness of rotor vibration signal, a novel method of fault information enhancement based on robust independent component analysis (RICA) was presented. Firstly, the signals were denoised using dual-tree complex wavelet transform (DTCWT) to reduce the noise so as to improve the performance of RICA algorithm. Then the denoised signals were separated and information was enhanced by means of RICA. Compared with the other two classical blind source separation algorithms, the effectiveness of the proposed method was validated with the simulative signal. The results show that the global optimal value is given by optimizing the step factor using the method. The optimal step size parameters are gotten and prewhitening is avoided using algebraic methods with lower computational cost. The method shows high convergence speed and good source separation property, especially applicable for small-data records. The method can effectively separate fault signals and extract quantitative fault characteristics.