Condition diagnosis of rotating machinery by the dimensionless symptom parameters of vibration signal has been widely used for the plant predictive maintenance procedures . This paper describes a new method to diagnose the plant machinery condition by the Information Divergence (ID) which is calculated from probability density functions of vibration signals. On the basis of the theoretical analysis, it is shown that the failure diagnosis sensitivity by the ID method is higher than by the one of Kullback-Leibler Information (KI). According to the information and the pattern recognition theory, a condition surveillance and a precise diagnosis method for rotating machinery by the ID method were also established. The availability of this methods has been verified by the experiments for gears and bearings condition diagnosis.
