Application of ISVD de-noising and correlation dimension in fault diagnosis of flue gas turbine

Key words:  flue gas turbine  fault diagnosis  correlation dimension  SVD de-noising  pseudo-phase portrait  low-pass filter

Abstract:

Considering the complicated and non-linear characteristics of the vibrational signal of flue gas turbine，the iterative singular value decomposition (ISVD) de-noising and correlation dimension were applied in fault diagnosis of flue gas turbine. To eliminate the noise, the low-pass filter and the ISVD de-noising were applied separately. Then the correlation dimensions of vibration signal of flue gas turbine under different fault conditions were estimated. The results show that the effect of the low-pass filter is not obvious while the ISVD de-noisijig can reduce noise effectively. Comparing with the pseudo-phase portrait reconstructed from signal containing noise, the pseudo-phase portrait reconstructed after ISVD de-noising is more regular. By ISVD de-noising, the scale region on the log-log plot of correlation integrals becomes wider. The correlation dimension is obviously different for different fault conditions after ISVD de-noising, so it can be used as the quantitative characteristic parameter for fault diagnosis. This offers a simple and effective method for fault diagnosis of flue gas turbine.