Fault Prognosis of Steam Turbine Generator Set by Trend Analysis of Frequency

Abstract

A new fault prognosis method based on trend analysis of frequency component is proposed. The vibration signal of bearing or shaft of the steam turbo-generator set is transformed into frequency signal by Fast Fourier Transform. The different frequency components are classified according to sohre’s chart. The amplitudes of different frequency components are ranked by time series. The trend of each frequency component is analyzed in terms of polynomial fitting and significant level 0.05. A set of diagnostic relations induced mainly from sohre’s charts can be used to predict the fault based on the trend analysis of the frequency components. And the model is validated and discussed by a simulation example and a case.