Abstract:

The SSA(Singular Spectrum Analysis)was used to reconstruct the temperature signals of bearings in the phase space.According to the singular spectrum characteristic of the reconstructed attractor track matrix.the tempe rature signals were separated into two independent components:the trend and the noise. The trendfeature of bearing temperature was then obtained from the original signal.On this basis,a fault prediction sys-tem for the bearing of hydroelectric generating set was developed.This system includes three subsystems respe c-tively for data collection,trend prediction and fault diagnosis.Test results show that the proposed system can predict the increase trend of bearing temperature early,extract the fault characteristics contained by the original signal,give an advance warning to prevent the faults at the burning tile,and can be applied to the state analy-sis,state assessment and fault prediction of hydroelectric generating sets.

Key words: hydroelectric generating set, singular spectrum analysis, bearing temperature, temperature trend, fault prediction